| blob | d25e3e30a68aa910ec9828c74ee53486f9bc86d4 |
1 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4 step to make sure stack always aligned.
6 2018-04-20 Carl Love <cel@us.ibm.com>
8 PR target/83402
9 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
10 size check for arg0.
12 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
13 Tom de Vries <tom@codesourcery.com>
15 PR target/85445
16 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
17 Emit insns for calls too.
18 (nvptx_find_par): Always look for worker-level predecessor insn.
19 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
20 calls.
21 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
22 (nvptx_process_pars): Propagate frames for calls.
24 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
26 PR target/85469
27 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
28 Removed.
29 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
30 (ix86_handle_option): Don't handle OPT_mibt.
31 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
32 __SHSTK__.
33 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
34 has_ibt and ibt.
35 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
36 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
37 (ix86_target_macros): Define __CET__ with flag_cf_protection
38 for -fcf-protection.
39 * config/i386/i386.c (isa2_opts): Remove -mibt.
40 * config/i386/i386.h (TARGET_IBT): Removed.
41 (TARGET_IBT_P): Likewise.
42 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
43 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
44 * config/i386/i386.opt (mcet): Update help message.
45 (mshstk): Likewise.
46 (mibt): Removed.
47 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
48 -mcet as an alias for -mshstk.
50 2018-04-20 Richard Biener <rguenther@suse.de>
52 PR middle-end/85475
53 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
54 complexity by forcing a single use of the multiply operand.
56 2018-04-20 Martin Jambor <mjambor@suse.cz>
58 ipa/85449
59 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
60 recursion dependency to only apply to non-clones.
62 2018-04-20 Martin Jambor <mjambor@suse.cz>
64 ipa/85447
65 * ipa-cp.c (create_specialized_node): Check that clones of
66 self-recursive edges exist during IPA-CP.
68 2018-04-19 Toon Moene <toon@moene.org>
70 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
71 by -O3.
73 2018-04-19 Jakub Jelinek <jakub@redhat.com>
75 PR tree-optimization/85467
76 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
77 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
78 VECTOR_CST element to type.
80 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
82 PR target/85397
83 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
84 * config/i386/i386.md (builtin_setjmp_setup): Removed.
85 (builtin_longjmp): Likewise.
86 (save_stack_nonlocal): New pattern.
87 (restore_stack_nonlocal): Likewise.
89 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
91 PR target/85404
92 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
93 Replace ASM_OUTPUT_LABEL with fprintf.
95 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
97 PR target/85417
98 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
99 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
100 * config/i386/i386-c.c (ix86_target_macros_internal): Also
101 define __IBT__ and __SHSTK__ for -fcf-protection.
102 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
103 TARGET_IBT.
104 (ix86_trampoline_init): Likewise.
105 (x86_output_mi_thunk): Likewise.
106 (ix86_notrack_prefixed_insn_p): Likewise.
107 (ix86_option_override_internal): Don't disallow -fcf-protection.
108 * config/i386/i386.md (rdssp<mode>): Also enable for
109 -fcf-protection.
110 (incssp<mode>): Likewise.
111 (nop_endbr): Likewise.
112 * config/i386/i386.opt (mcet): Change help message to built-in
113 functions only.
114 (mibt): Likewise.
115 (mshstk): Likewise.
116 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
117 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
118 enable CET built-in functions.
120 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
122 * common/config/i386/i386-common.c
123 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
124 OPTION_MASK_ISA_MOVDIRI_UNSET,
125 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
126 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
127 * config.gcc (movdirintrin.h): New header.
128 * config/i386/cpuid.h (bit_MOVDIRI,
129 bit_MOVDIR64B): New bits.
130 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
131 and -mmvodir64b.
132 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
133 (VOID, PVOID, PCVOID)): New function types.
134 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
135 __builtin_ia32_directstoreu_u64,
136 __builtin_ia32_movdir64b): New builtins.
137 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
138 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
139 and -mmovdiri.
140 (ix86_valid_target_attribute_inner_p): Ditto.
141 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
142 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
143 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
144 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
145 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
146 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
147 (movdiri<mode>, movdir64b_<mode>): New.
148 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
149 * config/i386/immintrin.h: Include movdirintrin.h.
150 * config/i386/movdirintrin.h: New file.
151 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
153 2018-04-19 Richard Biener <rguenther@suse.de>
155 PR middle-end/85455
156 * cfg.c (clear_bb_flags): When loop state says we have
157 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
159 2018-04-19 Richard Biener <rguenther@suse.de>
161 PR tree-optimization/84737
162 * tree-vect-data-refs.c (vect_copy_ref_info): New function
163 copying restrict info.
164 (vect_setup_realignment): Use it.
165 * tree-vectorizer.h (vect_copy_ref_info): Declare.
166 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
167 the first DR to all generated stores.
168 (vectorizable_load): Likewise for loads.
170 2018-04-19 Jakub Jelinek <jakub@redhat.com>
172 PR tree-optimization/85446
173 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
174 the integral and pointer types to have the same precision.
176 * doc/install.texi: Document --disable-cet being the default and
177 --enable-cet=auto.
179 2018-04-18 Martin Liska <mliska@suse.cz>
181 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
182 style.
184 2018-04-18 Martin Liska <mliska@suse.cz>
186 Revert
187 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
189 PR ipa/83983
190 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
191 arguments if they are comparable.
193 2018-04-18 Martin Liska <mliska@suse.cz>
195 Revert
196 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
198 PR lto/84805
199 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
200 incomplete types.
202 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
204 PR target/85388
205 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
206 ENDBR after calling __morestack.
208 2018-04-18 David Malcolm <dmalcolm@redhat.com>
210 PR jit/85384
211 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
212 by using gcc_base_ver to generate a gcc_driver_version, and use
213 it when generating GCC_DRIVER_NAME.
214 * configure: Regenerate.
216 2018-04-18 Jakub Jelinek <jakub@redhat.com>
218 PR target/81084
219 * config.gcc: Obsolete powerpc*-*-*spe*.
221 2018-04-17 Jakub Jelinek <jakub@redhat.com>
223 PR debug/84637
224 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
225 (stabstr_D): Change type of unum from unsigned int to
226 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
227 type.
229 2018-04-17 Jim Wilson <jimw@sifive.com>
231 PR 84856
232 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
233 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
234 Set arg_pointer_offset after using pretend_args_size.
236 2018-04-17 Jakub Jelinek <jakub@redhat.com>
238 PR rtl-optimization/85431
239 * dse.c (record_store): Ignore zero width stores.
241 PR sanitizer/85230
242 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
243 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
244 __builtin_stack_restore rather than after it.
245 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
246 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
247 argument instead of virtual_dynamic_stack_rtx.
249 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
251 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
252 New prototype.
253 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
254 Add note to error message to explain internal mapping of overloaded
255 built-in function name to non-overloaded built-in function name.
256 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
257 function.
259 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
261 PR target/85424
262 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
263 where the inputs overlap with the output.
265 2018-04-17 Jakub Jelinek <jakub@redhat.com>
267 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
268 (=v, v) alternative and explicit "memory" attribute.
269 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
270 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
271 attributes.
272 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
273 "sselog1" type instead of "sselog".
274 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
275 "sselog". Remove explicit "memory" attribute.
276 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
277 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
278 attributes.
279 (vec_extract_hi_v32hi): Merge all alternatives into one, use
280 "sselog1" type instead of "sselog". Remove explicit "memory"
281 attribute.
282 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
283 use "sselog1" type instead of "sselog". Remove explicit "memory"
284 attribute.
285 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
286 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
287 attributes.
288 (vec_extract_hi_v64qi): Merge all alternatives into one, use
289 "sselog1" type instead of "sselog". Remove explicit "memory"
290 attribute.
291 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
292 use "sselog1" type instead of "sselog". Remove explicit "memory"
293 attribute.
295 PR target/85430
296 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
298 PR middle-end/85414
299 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
300 on a SUBREG.
302 2018-04-17 Martin Jambor <mjambor@suse.cz>
304 PR ipa/85421
305 * ipa-cp.c (create_specialized_node): Call
306 expand_all_artificial_thunks if necessary.
308 2018-04-17 Martin Liska <mliska@suse.cz>
310 PR lto/85405
311 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
312 in message, remote space in between '_G' and '('.
314 2018-04-17 Jakub Jelinek <jakub@redhat.com>
316 PR target/85281
317 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
318 avx512f_vmcmp<mode>3<round_saeonly_name>,
319 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
320 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
321 avx512f_rndscale<mode><round_saeonly_name>,
322 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
323 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
324 Use %<iptr>2 instead of %2 for -masm=intel.
325 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
326 avx512f_vcvttss2usi<round_saeonly_name>,
327 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
328 -masm=intel.
329 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
330 avx512f_vcvttsd2usi<round_saeonly_name>,
331 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
332 Use %q1 instead of %1 for -masm=intel.
333 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
334 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
335 of %3 for -masm=intel.
336 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
337 -masm=intel.
338 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
339 -masm=intel.
340 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
341 -masm=intel.
342 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
343 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
344 %g1.
345 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
346 -masm=intel.
347 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
348 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
349 %g1 and one with %0 and %1.
350 (avx512er_vmrcp28<mode><round_saeonly_name>,
351 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
352 %1 for -masm=intel.
353 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
354 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
355 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
356 of %0 and %{%4%} for -masm=intel.
357 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
358 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
359 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
360 order of %0 and %{%5%}%{z%} for -masm=intel.
362 2018-04-17 Jan Hubicka <jh@suse.cz>
364 PR lto/85405
365 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
367 2018-04-17 Martin Liska <mliska@suse.cz>
369 PR ipa/85329
370 * multiple_target.c (create_dispatcher_calls): Set apostrophes
371 for target_clone error message. Make default implementation
372 clone to be a local declaration.
373 (separate_attrs): Add new argument and check for an empty
374 string.
375 (expand_target_clones): Handle it.
376 (ipa_target_clone): Make redirection just for target_clones
377 functions.
379 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
380 Tom de Vries <tom@codesourcery.com>
382 PR middle-end/84955
383 * omp-expand.c (expand_oacc_for): Add dummy false branch for
384 tiled basic blocks without omp continue statements.
386 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
388 PR target/83660
389 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
390 vec_extract expression as having side effects to make sure it gets
391 a cleanup point.
393 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
395 PR target/85403
396 * config/i386/i386.c (get_builtin_code_for_version): Check
397 error_mark_node.
399 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
401 PR target/84331
402 * gcc/config.gcc: Support "skylake".
403 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
404 PROCESSOR_SKYLAKE.
405 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
406 (processor_target_table): Add "skylake".
407 (ix86_option_override_internal): Add "skylake".
408 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
409 PROCESSOR_CANNONLAKE.
410 (get_builtin_code_for_version): Fix priority for
411 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
412 PROCESSOR_SKYLAKE-AVX512.
413 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
414 (processor_type): Add PROCESSOR_SKYLAKE.
416 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
417 Jason Merrill <jason@redhat.com>
419 PR c++/85112
420 * convert.c (convert_to_integer_1): Use direct recursion for
421 enumeral types and types with a precision less than the number
422 of bits in their mode.
424 2018-04-16 Julia Koval <julia.koval@intel.com>
426 PR target/84413
427 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
428 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
430 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
432 PR target/85293
433 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
434 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
435 and -mno-direct-move.
437 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
439 PR target/83402
440 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
441 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
442 Ensure negative shifts result in {0}.
444 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
446 PR rtl-optimization/79916
447 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
448 regs (if any) to define how to gnerate SD moves when LRA is in
449 progress.
451 2018-04-13 Jakub Jelinek <jakub@redhat.com>
453 PR rtl-optimization/85393
454 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
455 * except.c (expand_dw2_landing_pad_for_region): Make static.
456 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
457 a label and unconditional jump to old_bb, rather than
458 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
459 basic block.
461 PR rtl-optimization/85376
462 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
463 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
464 instead of a specific value.
466 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
467 Bin Cheng <bin.cheng@arm.com>
469 PR tree-optimization/82965
470 PR tree-optimization/83991
471 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
472 by_profile_only parameter.
473 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
474 information if the loop was predicted to iterate too many times.
475 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
477 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
479 PR lto/71991
480 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
481 always inline.
483 2018-04-13 Martin Liska <mliska@suse.cz>
484 Jakub Jelinek <jakub@redhat.com>
486 PR middle-end/81657
487 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
488 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
489 * builtins.c (expand_builtin_memory_copy_args): Use
490 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
491 handle dest_addr == pc_rtx.
493 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
495 PR target/85291
496 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
497 asked to not generate direct moves.
498 (fix_trunc<mode>si2_stfiwx): Similar.
499 (fix_trunc<mode>si2_internal): Similar.
501 2018-04-12 Jakub Jelinek <jakub@redhat.com>
503 PR debug/83157
504 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
505 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
506 lookup if dest in some wider mode is known to be const0_rtx and
507 if so, record permanent equivalence for it to be ZERO_EXTEND of
508 the narrower mode destination.
510 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
512 * lto-streamer-out.c (output_function): Revert 259346.
513 * omp-expand.c (expand_oacc_for): Likewise.
515 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
517 PR rtl-optimization/85354
518 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
519 * sel-sched.c (sel_global_init): ... here.
521 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
523 PR target/85238
524 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
525 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
526 mode for PE-COFF targets.
527 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
528 (i386_pe_asm_lto_end): Likewise.
529 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
530 (TARGET_ASM_LTO_END): Likewise.
531 * config/i386/winnt.c (saved_debug_info_level): New static variable.
532 (i386_pe_asm_lto_start): New function.
533 (i386_pe_asm_lto_end): Likewise.
535 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
536 Richard Biener <rguenther@suse.de>
538 PR middle-end/84955
539 * lto-streamer-out.c (output_function): Fix CFG loop state before
540 streaming out.
541 * omp-expand.c (expand_oacc_for): Handle calls to internal
542 functions like regular functions.
544 2018-04-12 Richard Biener <rguenther@suse.de>
546 PR lto/85371
547 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
548 for the early LTO debug to properly generate references to it
549 during DIE emission. Do not re-use that for the skeleton for
550 split-dwarf.
551 (dwarf2out_early_finish): Likewise.
553 2018-04-12 Jakub Jelinek <jakub@redhat.com>
555 PR target/85328
556 * config/i386/sse.md
557 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
558 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
559 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
560 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
561 and output is a reg, avoid creating invalid lowpart subreg, but
562 instead split into a 512-bit move. Don't split if not AVX512VL,
563 input is xmm16+ reg and output is a mem.
564 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
565 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
566 xmm16+ reg and output is a mem.
568 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
570 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
571 also for flag_dwarf2_cfi_asm.
573 2018-04-12 Jakub Jelinek <jakub@redhat.com>
575 PR rtl-optimization/85342
576 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
577 a bool scalar var inside of the loop instead. Don't try to update
578 recog_data.operand after failed apply_change_group.
580 2018-04-12 Tom de Vries <tom@codesourcery.com>
582 PR target/85296
583 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
584 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
585 array with flexible array member as array without given dimension.
586 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
587 argument for undefined param to true.
589 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
591 PR target/85321
592 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
593 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
594 from PowerPC section.
595 * config/rs6000/sysv4.opt (mcall-): Improve help text.
596 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
597 help text that is too long.
598 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
599 help text that is too long.
600 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
601 help text that is too long.
603 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
605 * config/alpha/alpha.md (stack_probe_internal): Rename
606 from "probe_stack". Update all callers.
608 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
610 PR rtl-optimization/84566
611 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
612 sched_macro_fuse_insns.
614 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
616 PR target/84301
617 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
618 (compute_block_dependences): ... from here.
620 2018-04-11 Jakub Jelinek <jakub@redhat.com>
622 PR tree-optimization/85331
623 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
624 from int to HOST_WIDE_INT.
626 2018-04-11 Martin Jambor <mjambor@suse.cz>
628 PR ipa/84149
629 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
630 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
631 not the same as the source val.
632 (cgraph_edge_brings_value_p): New parameter.
633 (gather_edges_for_value): Pass destination value to
634 cgraph_edge_brings_value_p.
635 (perhaps_add_new_callers): Likewise.
636 (get_info_about_necessary_edges): Likewise and exclude values brought
637 only by self-recursive edges.
638 (create_specialized_node): Redirect only clones of self-calling edges.
639 (+self_recursive_pass_through_p): New function.
640 (find_more_scalar_values_for_callers_subset): Use it.
641 (find_aggregate_values_for_callers_subset): Likewise.
642 (known_aggs_to_agg_replacement_list): Removed.
643 (decide_whether_version_node): Re-calculate known constants for all
644 remaining context clones.
646 2018-04-11 Richard Biener <rguenther@suse.de>
648 PR lto/85339
649 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
650 from early DWARF output.
651 (dwarf2out_early_finish): Output line info unconditionally into
652 early DWARF and add reference to it.
654 2018-04-11 Jakub Jelinek <jakub@redhat.com>
656 PR target/85281
657 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
658 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
659 other than V2DFmode using iptr mode attribute.
660 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
662 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
664 PR rtl-optimization/84659
665 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
667 2018-04-11 Jakub Jelinek <jakub@redhat.com>
669 PR debug/85302
670 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
671 SIZEP is NULL.
672 (output_loc_list): Pass address of a dummy size variable even in the
673 locview handling loop.
674 (index_location_lists): Add comment on why skip_loc_list_entry can't
675 call size_of_locs.
677 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
679 PR target/85261
680 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
681 into register.
683 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
685 PR target/85321
686 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
687 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
688 and -mstring-compare-inline-limit.
690 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
692 PR target/85287
693 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
694 for stack clash protection in a register whenever we need it to be in
695 a register.
697 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
699 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
700 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
702 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
704 PR target/85321
705 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
706 the help text.
707 (mlong-double-): Ditto.
708 * config/rs6000/sysv4.opt (msdata=): Ditto.
709 (mtls-size=): Ditto.
711 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
713 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
714 erroneous entries for
715 "vector int vec_ldl (int, long int *)", and
716 "vector unsigned int vec_ldl (int, unsigned long int *)".
717 Add comments and entries for
718 "vector bool char vec_ldl (int, bool char *)",
719 "vector bool short vec_ldl (int, bool short *)",
720 "vector bool int vec_ldl (int, bool int *)",
721 "vector bool long long vec_ldl (int, bool long long *)",
722 "vector pixel vec_ldl (int, pixel *)",
723 "vector long long vec_ldl (int, long long *)",
724 "vector unsigned long long vec_ldl (int, unsigned long long *)".
725 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
726 type tree bool_long_long_type_node and correct definition of
727 bool_V2DI_type_node to make reference to this new type tree.
728 (rs6000_mangle_type): Replace erroneous reference to
729 bool_long_type_node with bool_long_long_type_node.
730 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
731 comments to emphasize sign distinctions for char and int types and
732 replace RS6000_BTI_bool_long constant with
733 RS6000_BTI_bool_long_long constant. Also add comment to restrict
734 use of RS6000_BTI_pixel.
735 (bool_long_type_node): Remove this macro definition.
736 (bool_long_long_type_node): New macro definition
738 2018-04-10 Jakub Jelinek <jakub@redhat.com>
740 PR rtl-optimization/85300
741 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
742 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
743 simplify_unary_operation fails.
745 2018-04-10 Martin Liska <mliska@suse.cz>
747 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
748 cgraph_edge and ipa_ref.
750 2018-04-10 Jakub Jelinek <jakub@redhat.com>
752 PR target/85177
753 PR target/85255
754 * config/i386/sse.md
755 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
756 computation of the VEC_MERGE selector from mask.
757 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
758 Fix decoding of the VEC_MERGE selector into mask.
760 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
762 PR tree-optimization/85286
763 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
765 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
767 * final.c (final_1): Set insn_last_address as well as
768 insn_current_address.
770 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
772 PR target/85173
773 * explow.c (emit_stack_probe): Call validize_mem on memory location
774 before passing it to gen_probe_stack. Create address operand and
775 legitimize it for the probe_stack_address case.
777 2018-04-09 Jan Hubicka <jh@suse.cz>
779 PR lto/85078
780 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
781 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
782 * tree.c (free_lang_data_in_type): Fix handling of binfos;
783 walk basetypes.
784 (free_lang_data): Rebuild type inheritance graph.
786 2018-04-09 Martin Sebor <msebor@redhat.com>
788 * invoke.texi (-finline-small-functions): Mention other optimization
789 options.
790 (-findirect-inlining, -fpartial-inlining): Same.
791 (-finline-functions-called-once): Same.
792 (-freorder-blocks-and-partition): Same.
794 2018-04-09 Jan Hubicka <jh@suse.cz>
796 PR rtl/84058
797 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
798 jumps; choose last target that matches the criteria (i.e.
799 no partition changes for non-crossing jumps).
800 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
801 support for redirecting crossing jumps to non-crossing.
803 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
805 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
806 also for naked functions.
808 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
810 * config/arc/arc.md (add_shift): New pattern.
811 (add_shift2): Likewise.
812 (sub_shift): Likewise.
813 (sub_shift_cmp0_noout): Likewise.
814 (compare_si_ashiftsi): Likewise.
815 (xbfu_cmp0_noout): New combine pattern.
816 (xbfu_cmp0"): Likewise.
817 (movsi_set_cc_insn): Place the predicable variant first.
818 (commutative_binary_cmp0_noout): Remove clobber.
819 (commutative_binary_cmp0): New pattern.
820 (noncommutative_binary_cmp0): Likewise.
821 (noncommutative_binary_cmp0_noout): Likewise.
822 (noncommutative_binary_comparison_result_used): Removed.
823 (rsub_cmp0): New pattern.
824 (rsub_cmp0_noout): Likewise.
825 (extzvsi): Changed, keep only meaningful variants.
826 (SQH, SEZ): New iterators.
827 (SQH_postfix): New mode attribute.
828 (SEZ_prefix): New code attribute.
829 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
830 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
831 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
832 of numerical value.
833 (noncommutative_operator): Check the availability of barrel
834 shifter option.
836 2018-04-09 Richard Biener <rguenther@suse.de>
838 PR tree-optimization/85284
839 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
840 Only use the niter constraining form of simple_iv when the exit
841 is always executed.
843 2018-04-09 Tom de Vries <tom@codesourcery.com>
845 PR target/84041
846 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
847 (define_expand "*memory_barrier"): New define_expand.
848 (define_insn "memory_barrier"): New insn.
850 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
852 PR rtl-optimization/80463
853 PR rtl-optimization/83972
854 PR rtl-optimization/83480
856 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
857 correct producer for the insn.
858 (tidy_control_flow): Fixup seqnos in case of debug insns.
860 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
862 PR rtl-optimization/83913
864 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
865 different sched-times when merging exprs.
867 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
869 PR rtl-optimization/83962
871 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
872 tidy_fallthru_edge and tidy_control_flow.
874 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
876 PR rtl-optimization/83530
878 * sel-sched.c (force_next_insn): New global variable.
879 (remove_insn_for_debug): When force_next_insn is true, also leave only
880 next insn in the ready list.
881 (sel_sched_region): When the region wasn't scheduled, make another pass
882 over it with force_next_insn set to 1.
884 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
886 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
887 into tm_file.
888 * config/nds32/constants.md (unspec_volatile_element): Add enum values
889 for interrupt control.
890 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
891 functions for interrupt control.
892 * config/nds32/nds32-intrinsic.md: Likewise.
893 * config/nds32/nds32_intrinsic.h: Likewise.
894 * config/nds32/nds32.h (nds32_builtins): Likewise.
896 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
898 * config/nds32/nds32.c (nds32_init_machine_status,
899 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
900 strict_aligned_p field.
901 (nds32_expand_to_rtl_hook): New function.
902 (TARGET_EXPAND_TO_RTL_HOOK): Define.
903 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
905 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
906 Chung-Ju Wu <jasonwucj@gmail.com>
908 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
909 * config/nds32/nds32-n7.md: New file.
910 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
911 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
912 pipeline.
913 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
914 * config/nds32/nds32.md (pipeline_model): Add n7.
915 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
916 * config/nds32/pipelines.md: Include n7 settings.
918 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
919 Chung-Ju Wu <jasonwucj@gmail.com>
921 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
922 * config/nds32/nds32-e8.md: New file.
923 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
924 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
925 pipeline.
926 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
927 * config/nds32/nds32.md (pipeline_model): Add e8.
928 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
929 * config/nds32/pipelines.md: Include e8 settings.
931 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
932 Chung-Ju Wu <jasonwucj@gmail.com>
934 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
935 * config/nds32/nds32-n8.md: New file.
936 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
937 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
938 pipeline.
939 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
940 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
941 * config/nds32/nds32.md (pipeline_model): Add n8.
942 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
943 * config/nds32/pipelines.md: Include n8 settings.
945 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
946 Chung-Ju Wu <jasonwucj@gmail.com>
948 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
949 * config/nds32/nds32-n9-2r1w.md: New file.
950 * config/nds32/nds32-n9-3r2w.md: New file.
951 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
952 nds32_register_ports): New or modify for cpu n9.
953 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
954 pipeline.
955 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
956 * config/nds32/nds32-utils.c: New file.
957 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
958 TARGET_MUL_SLOW): Define.
959 * config/nds32/nds32.md (pipeline_model): New attribute.
960 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
961 New options that support cpu n9.
962 * config/nds32/pipelines.md: Include n9 settings.
963 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
965 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
967 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
968 information if necessary.
969 (output_cond_branch_compare_zero): Likewise.
970 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
971 (nds32_target_alignment): Refine for alignment.
972 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
973 (FUNCTION_BOUNDARY): Modify.
974 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
975 align case.
976 * config/nds32/nds32.opt (malways-align, malign-functions): New.
978 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
980 * config/nds32/constants.md (unspec_volatile_element): Add values for
981 TLB operation and data prefetch.
982 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
983 functions for TLB operation and data prefetch.
984 * config/nds32/nds32-intrinsic.md: Likewise.
985 * config/nds32/nds32_intrinsic.h: Likewise.
986 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
987 (nds32_print_operand): Likewise.
988 * config/nds32/nds32.h (nds32_builtins): Likewise.
990 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
991 Andrew Pinski <pinsika@gcc.gnu.org>
993 PR middle-end/82976
994 * match.pd: Use constant_boolean_node of correct type instead of
995 boolean_true_node or boolean_false_node for simplifying
996 pointer comparisons to zero.
998 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1000 PR tree-optimization/80021
1001 * tree.c (verify_type_variant): Make error call in verify_variant_match
1002 translatable and remove final full stop.
1004 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1006 * config/nds32/constants.md (unspec_volatile_element): Add
1007 UNSPEC_VOLATILE_EH_RETURN.
1008 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1009 nds32_output_stack_pop): Support dwarf exception handling process.
1010 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1011 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1012 exception handling process.
1013 (nds32_compute_stack_frame): Likewise.
1014 (nds32_return_addr_rtx): Likewise.
1015 (nds32_initial_elimination_offset): Likewise.
1016 (nds32_expand_prologue): Likewise.
1017 (nds32_expand_epilogue): Likewise.
1018 (nds32_dynamic_chain_address): New function.
1019 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1020 exception handling.
1021 (DYNAMIC_CHAIN_ADDRESS): Define.
1022 (EH_RETURN_DATA_REGNO): Define.
1023 (EH_RETURN_STACKADJ_RTX): Define.
1024 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1025 patterns for dwarf exception handling.
1027 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1029 * config/nds32/nds32.h: Clean up obsolete macros.
1031 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1033 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1034 Add enum values for particular instructions.
1035 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1036 particular intrinsic functions.
1037 * config/nds32/nds32-intrinsic.md: Likewise.
1038 * config/nds32/nds32_intrinsic.h: Likewise.
1039 * config/nds32/nds32.h (nds32_builtins): Likewise.
1040 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1041 (btst, ave): New patterns for particular instructions.
1043 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1045 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1046 Add enum values for atomic load/store and memory sync.
1047 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1048 and memory sync.
1049 * config/nds32/nds32-intrinsic.md: Likewise.
1050 * config/nds32/nds32_intrinsic.h: Likewise.
1051 * config/nds32/nds32.h (nds32_builtins): Likewise.
1053 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1055 PR tree-optimization/85257
1056 * fold-const.c (native_encode_vector): If not all elts could fit
1057 and off is -1, return 0 rather than offset.
1058 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1059 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1060 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1061 adjust buffer in native_interpret_expr call.
1063 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1065 * config/nds32/constants.md (unspec_volatile_element): Add cache
1066 control enum values.
1067 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1068 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1069 * config/nds32/nds32.c (nds32_cctl_names): New.
1070 (nds32_print_operand): Handle cache control register names.
1071 * config/nds32/nds32.h (nds32_builtins): New enum values.
1072 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1073 macros.
1074 * config/nds32/nds32.md (type): Add mmu.
1075 * config/nds32/pipelines.md (simple_insn): Add mmu.
1077 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1079 * config/nds32/nds32.md (type): Remove call.
1080 * config/nds32/pipelines.md (simple_insn): Likewise.
1082 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1084 * config/nds32/constants.md (unspec_volatile_element): Add
1085 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1086 UNSPEC_VOLATILE_FMFCFG.
1087 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1088 description for fmfcfg and fmfcsr.
1089 (bdesc_1arg): Add fmtcsr.
1090 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1091 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1092 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1093 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1094 unspec_fmfcfg): New patterns.
1095 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1096 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1097 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1098 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1099 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1100 __nds32__fmfcfg): Define.
1102 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1104 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1105 intrinsic register names.
1106 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1107 intrinsic register enum values and macros.
1109 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1111 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1112 for load/store addressing form.
1113 (nds32_print_operand_address): Likewise.
1115 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1117 PR target/85196
1118 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1119 based on LABEL_REF. Remove useless assertion.
1120 (pic_address_needs_scratch): Fix formatting.
1121 (sparc_legitimize_pic_address): Minor tweaks.
1122 (sparc_delegitimize_address): Adjust assertion accordingly.
1123 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1124 into symbolic_operand.
1125 (movsi_high_pic_label_ref): Likewise.
1126 (movsi_lo_sum_pic_label_ref): Likewise.
1127 (movdi_pic_label_ref): Likewise.
1128 (movdi_high_pic_label_ref): Likewise.
1129 (movdi_lo_sum_pic_label_ref): Likewise.
1131 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1133 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1134 custom LIB_SPEC setup.
1136 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1137 Kito Cheng <kito.cheng@gmail.com>
1139 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1140 * config/riscv/freebsd.h: New.
1142 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1144 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1145 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1146 file.
1148 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1149 Kito Cheng <kito.cheng@gmail.com>
1151 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1152 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1153 * config/nds32/nds32-protos.h (nds32_output_call,
1154 nds32_output_return): Declare.
1155 * config/nds32/nds32.md: Refine all the call and return patterns.
1157 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1159 PR debug/85252
1160 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1161 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1163 PR rtl-optimization/84872
1164 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1165 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1166 EDGE_CROSSING edge.
1168 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1170 * expr.c (copy_blkmode_to_reg): Revert 254862.
1171 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1173 2018-04-06 Richard Biener <rguenther@suse.de>
1175 PR middle-end/85244
1176 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1177 after seeing a component reference with an adjacent field. Treat
1178 refs to arrays at struct end of external decls similar to
1179 refs to unconstrained commons.
1181 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1183 PR sanitizer/85213
1184 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1185 look through SAVE_EXPRs with non-side-effects argument. Adjust
1186 recursive calls.
1187 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1188 save_p here.
1190 2018-04-06 Richard Biener <rguenther@suse.de>
1192 PR middle-end/85180
1193 * alias.c (find_base_term): New wrapper around find_base_term
1194 unwinding CSELIB_VAL_PTR changes.
1195 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1196 recursion.
1198 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1200 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1201 instructions.
1202 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1203 constant definitions.
1204 ("nop"): lr 0,0 -> nopr r0
1205 ("nop_lr0", "nop_lr1"): New insn definitions.
1207 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1209 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1210 NDS32_V3PUSH_AVAILABLE_P macro.
1212 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1213 Chung-Ju Wu <jasonwucj@gmail.com>
1215 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1216 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1217 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1218 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1219 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1220 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1221 * config/nds32/constraints.md: New constraints and checking for hard
1222 float configuration.
1223 * config/nds32/iterators.md: New mode iterator and attribute for hard
1224 float configuration.
1225 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1226 patterns.
1227 * config/nds32/nds32-fpu.md: New file.
1228 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1229 deal with hard float code generation.
1230 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1231 ARCH_V3S.
1232 (abi_type, float_reg_number): New enum type.
1233 * config/nds32/nds32-predicates.c: New predicates for hard float.
1234 * config/nds32/nds32-protos.h: Declare functions for hard float.
1235 * config/nds32/nds32.c: Implementation for hard float configuration.
1236 * config/nds32/nds32.h: Definitions for hard float configuration.
1237 * config/nds32/nds32.md: Include hard float machine description and
1238 modify patterns for hard float configuration.
1239 * config/nds32/nds32.opt: New options for hard float configuration.
1240 * config/nds32/predicates.md: New predicates for hard float
1241 configuration.
1243 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1245 * common/config/nds32/nds32-common.c
1246 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1248 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1250 PR middle-end/85195
1251 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1252 CONSTRUCTOR_ELT (ctor, ...)->value.
1254 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1256 PR target/85193
1257 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1259 2018-04-05 Tom de Vries <tom@codesourcery.com>
1261 PR target/85204
1262 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1263 cond jump.
1265 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1266 Kito Cheng <kito.cheng@gmail.com>
1268 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1269 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1270 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1271 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1273 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1274 Kito Cheng <kito.cheng@gmail.com>
1276 * config/nds32/constraints.md (Ufe): New memory constraint.
1277 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1278 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1279 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1280 operands.
1281 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1282 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1284 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1286 * config/nds32/nds32.md: Use optimize_size in the condition for
1287 alu-shift instructions.
1289 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1291 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1293 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1295 * config/nds32/nds32.md (negsi2): Refine pattern.
1297 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1298 Chung-Ju Wu <jasonwucj@gmail.com>
1300 * config/nds32/iterators.md (shift_rotate): New code iterator.
1301 (shift): New code attribute.
1302 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1303 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1304 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1305 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1306 bit-wise operations.
1307 (andsi3, *andsi3): Ditto.
1308 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1309 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1310 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1311 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1312 nds32_ior_operand, nds32_xor_operand): New predicates.
1314 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1316 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1317 (addsi3, subsi3): ... this.
1319 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1321 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1323 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1325 * config/nds32/nds32.md: Adjust indention.
1327 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1329 * config/nds32/nds32.md (feature): New attribute.
1331 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1333 * config/nds32/nds32.md (subtype): New attribute.
1335 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1337 PR target/85203
1338 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1339 expansion to perform a bitwise AND of the argument followed by a
1340 boolean negation of the result.
1342 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1344 PR rtl-optimization/84878
1345 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1346 the basic block. Assert the use reference is not artificial and that
1347 it has an associated insn.
1349 2018-04-04 Michael Matz <matz@suse.de>
1351 * builtins.c (compute_objsize): Pass correct operand
1352 to array_at_struct_end_p.
1354 2018-04-04 Richard Biener <rguenther@suse.de>
1356 PR lto/85176
1357 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1358 from contexts for DINFO_LEVEL_TERSE and below.
1360 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1362 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1363 resiter_operand condition.
1364 * config/nds32/nds32.md (*move<mode>): Ditto.
1366 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1367 Monk Chiang <sh.chiang04@gmail.com>
1369 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1371 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1373 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1375 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1376 Kito Cheng <kito.cheng@gmail.com>
1378 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1379 nds32_cond_code_str, output_cond_branch,
1380 output_cond_branch_compare_zero, nds32_expand_cbranch,
1381 nds32_expand_cstore, nds32_expand_movcc,
1382 nds32_output_cbranchsi4_equality_zero,
1383 nds32_output_cbranchsi4_equality_reg,
1384 nds32_output_cbranchsi4_equality_reg_or_const_int,
1385 nds32_output_cbranchsi4_greater_less_zero: New functions.
1386 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1387 nds32_expand_cstore, nds32_expand_movcc,
1388 nds32_output_cbranchsi4_equality_zero,
1389 nds32_output_cbranchsi4_equality_reg,
1390 nds32_output_cbranchsi4_equality_reg_or_const_int,
1391 nds32_output_cbranchsi4_greater_less_zero): Declare.
1392 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1393 nds32_rimm11s_operand): New predicates.
1394 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1395 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1396 patterns.
1398 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1400 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1401 * config/nds32/nds32.md: Ditto.
1402 * config/nds32/pipelines.md: Ditto.
1404 2018-04-04 Richard Biener <rguenther@suse.de>
1406 PR tree-optimization/85168
1407 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1408 propagating abnormals.
1410 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1412 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1414 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1415 Kito Cheng <kito.cheng@gmail.com>
1417 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1418 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1419 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1420 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1421 * config/nds32/nds32.md (sibcall_internal): New.
1422 (sibcall_register): Remove.
1423 (sibcall_immediate): Remove.
1424 (sibcall_value_internal): New.
1425 (sibcall_value_register): Remove.
1426 (sibcall_value_immediate): Remove.
1427 * config/nds32/predicates.md (nds32_general_register_operand): New.
1428 (nds32_call_address_operand): New.
1430 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1432 PR rtl-optimization/85167
1433 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1434 bb_defs if *split_p, instead preinitialize it to NULL.
1436 PR tree-optimization/85156
1437 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1438 evaluating the argument multiple times.
1440 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1442 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1443 than vector.
1444 (_mm_cvtpd_ps): Likewise.
1445 (_mm_cvttpd_epi32): Likewise.
1446 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1447 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1448 vector, pixel, and bool following altivec.h include.
1450 2018-04-03 Martin Sebor <msebor@redhat.com>
1452 * doc/extend.texi (Common Function Attributes): Clarify.
1453 (const attribute): Likewise.
1454 (pure attribute): Likewise.
1456 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1458 PR target/85169
1459 * config/i386/i386.c (ix86_expand_vector_set): Use
1460 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1462 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1464 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1465 instructions when changing rounding bits to preserve precision bits
1466 in the x87 control word.
1468 2018-04-03 Martin Liska <mliska@suse.cz>
1470 PR tree-optimization/82491
1471 * rtl.h (strip_offset_and_add): Replace += suboffset with
1472 poly_uint64 () + suboffset.
1474 2018-03-29 Martin Liska <mliska@suse.cz>
1475 Martin Jambor <mjambor@suse.cz>
1477 PR ipa/84947
1478 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1479 param_type is not an integral or pointer type.
1481 2018-04-03 Richard Biener <rguenther@suse.de>
1483 * sese.h (recompute_all_dominators): Remove.
1485 2018-04-02 Martin Sebor <msebor@redhat.com>
1487 * doc/invoke.texi (-Wrestrict): Fix typos.
1489 2018-04-02 Jim Wilson <jimw@sifive.com>
1491 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1492 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1493 (<optab>di3, <optab>si3_extend): Likewise.
1494 (<optab>si3_mask, <optab>si3_mask_1): New.
1495 (<optab>di3_mask, <optab>di3_mask_1): New.
1496 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1497 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1498 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1500 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1502 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1503 example.
1505 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1507 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1508 (nds32_canonicalize_comparison): New function.
1510 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1511 Kito Cheng <kito.cheng@gmail.com>
1512 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1514 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1515 * config/nds32/constants.md (unspec_volatile_element): Add
1516 UNSPEC_VOLATILE_RELAX_GROUP.
1517 * config/nds32/nds32-relax-opt.c: New file.
1518 * config/nds32/nds32-predicates.c
1519 (nds32_symbol_load_store_p): New function.
1520 * config/nds32/nds32-protos.h
1521 (nds32_symbol_load_store_p): Declare function.
1522 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1523 * config/nds32/nds32.c
1524 (nds32_register_pass): New function to register pass.
1525 (nds32_register_passes): New function to register passes.
1526 * config/nds32/nds32.md (relax_group): New pattern.
1527 * config/nds32/nds32.opt (mrelax-hint): New option.
1528 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1530 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1532 * config/nds32/t-nds32: Modify files dependency.
1534 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1536 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1537 (PROFILE_HOOK): Define its implementation.
1539 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1541 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1542 type and 32-bit size.
1544 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1546 PR middle-end/85090
1547 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1548 (V_128_256): New mode iterator.
1549 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1550 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1551 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1552 of V.
1553 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1554 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1556 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1558 PR target/83315
1559 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1560 NaN inputs correctly.
1562 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1564 PR target/80546
1565 * config/rs6000/vsx.md (??r): New mode attribute.
1566 (*vsx_mov<mode>_64bit): Use it.
1567 (*vsx_mov<mode>_32bit): Likewise.
1569 2018-03-30 Martin Sebor <msebor@redhat.com>
1571 PR tree-optimization/84818
1572 * builtins.c (check_access): Use warning_n.
1574 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1576 PR target/83822
1577 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1578 condition.
1579 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1580 condition.
1582 2018-03-30 Julia Koval <julia.koval@intel.com>
1584 PR target/84413
1585 * x86-tune.def (movx, partial_reg_dependency): Enable for
1586 m_SKYLAKE_AVX512.
1588 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1590 PR inline-asm/84985
1591 * lra-constraints.c (process_alt_operands): Move setting
1592 this_alternative_matches below.
1594 2018-03-29 Martin Liska <mliska@suse.cz>
1596 PR lto/84995.
1597 * doc/invoke.texi: Document how LTO works with debug info.
1598 Describe auto-load support of binutils. Mention 'x86-64'
1599 as valid option value of -march option.
1601 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1603 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1605 PR c/85094
1606 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1607 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1608 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1609 checking.
1611 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1613 PR target/84912
1614 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1615 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1616 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1617 for RS6000_BTM_POWERPC64.
1618 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1619 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1620 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1621 definition.
1622 (DIVDE): Use it.
1623 (DIVDEU): Likewise.
1625 2018-03-28 Carl Love <cel@us.ibm.com>
1627 Revert
1628 2017-09-27 Carl Love <cel@us.ibm.com>
1630 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1631 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1632 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1633 fctiw instruction.
1635 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1637 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1638 instead of __vector bool.
1639 (_mm_max_pu8): Likewise.
1640 (_mm_min_pi16): Likewise.
1642 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1644 PR target/84912
1645 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1646 (DIVWEUO): Likewise.
1647 (DIVDEO): Likewise.
1648 (DIVDEUO): Likewise.
1649 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1650 DIVWEUO and DIVDEUO.
1651 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1652 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1653 (div_extend): Likewise.
1654 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1655 builtin function.
1656 (__builtin_divweuo): Likewise.
1657 (__builtin_divdeo): Likewise.
1658 (__builtin_divdeuo): Likewise.
1660 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1662 PR target/85095
1663 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1664 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1666 PR tree-optimization/82004
1667 * gimple-match-head.c (optimize_pow_to_exp): New function.
1668 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1669 Don't fold to exp if optimize_pow_to_exp is false.
1671 2018-03-28 Martin Liska <mliska@suse.cz>
1673 PR other/84819
1674 * calls.c (initialize_argument_information): Fix trailing space.
1675 * common.opt: Fix typo and provide better explanation for
1676 -fsanitize-coverage option.
1677 * config/i386/i386.opt: Fix typo.
1679 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1680 Martin Liska <mliska@suse.cz>
1682 PR sanitizer/85081
1683 * gimplify.c (asan_poison_variable): Don't do the check for
1684 gimplify_omp_ctxp here.
1685 (gimplify_decl_expr): Do it here.
1686 (gimplify_target_expr): Likewise.
1688 2018-03-28 Martin Liska <mliska@suse.cz>
1690 PR target/84988
1691 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1692 chkp_type_bounds_count if MPX is not enabled.
1694 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1696 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1698 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1700 PR target/84914
1701 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1702 function to create the function decl for complex long double
1703 multiply and divide for -mabi=ieeelongdouble.
1704 (init_float128_ieee): Call it.
1706 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1708 PR target/85044
1709 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1710 -fcf-protection=branch -mibt.
1711 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1713 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1715 PR target/81863
1716 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1718 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1720 PR target/85056
1721 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1722 extern array declarations.
1724 2018-03-27 Richard Biener <rguenther@suse.de>
1726 PR middle-end/84067
1727 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1728 explicit single_use checks.
1730 2018-03-27 Richard Biener <rguenther@suse.de>
1732 PR tree-optimization/85082
1733 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1734 Valueize the VUSE.
1736 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1738 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1739 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1740 Turn on fasynchronous-unwind-tables and funwind-tables.
1742 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1744 PR target/85073
1745 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1746 (*bmi_blsr_<mode>_ccz): Ditto.
1748 2018-03-26 Tom de Vries <tom@codesourcery.com>
1750 PR tree-optimization/85063
1751 * omp-general.c (offloading_function_p): New function. Factor out
1752 of ...
1753 * omp-offload.c (pass_omp_target_link::gate): ... here.
1754 * omp-general.h (offloading_function_p): Declare.
1755 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1756 with attribute omp declare target for offloading functions.
1758 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1760 PR tree-optimization/84005
1761 * tree-data-ref.h (get_base_for_alignment): Declare.
1762 * tree-data-ref.c (get_base_for_alignment_1): New function.
1763 (get_base_for_alignment): Likewise.
1764 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1765 get_base_for_alignment to find a suitable base object, instead
1766 of always using drb->base_address.
1768 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1770 PR inline-asm/85022
1771 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1772 known size by default.
1774 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1776 PR inline-asm/85030
1777 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1778 and non BLKmode operands.
1780 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1782 PR target/85026
1783 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1784 Clean up attributes.
1786 2018-03-23 Richard Biener <rguenther@suse.de>
1788 PR debug/85020
1789 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1790 we are going to emit early debug for LTO.
1792 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1794 PR inline-asm/85034
1795 * function.c (match_asm_constraints_1): Don't optimize if input
1796 doesn't satisfy general_operand predicate for output's mode.
1798 PR inline-asm/85022
1799 * alias.c (write_dependence_p): Don't require for x_canonicalized
1800 non-VOIDmode if x has VOIDmode.
1802 PR sanitizer/85029
1803 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1804 just don't try to optimize it rather than assert it never happens.
1806 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1808 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1809 macro expansions for definition of ST_INTERNAL_<mode> and
1810 LD_INTERNAL_<mode> builtins.
1811 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1812 Remove prototype.
1813 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1814 function.
1815 (altivec_expand_st_builtin): Likewise.
1816 (altivec_expand_builtin): Remove calls to deleted functions.
1817 (rs6000_address_for_altivec): Delete this function.
1818 * config/rs6000/vector.md: Remove expands for
1819 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1821 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1823 PR target/84826
1824 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1825 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1826 re-computing once computed.
1827 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1828 (arm_init_machine_status): Initialize
1829 machine->static_chain_stack_bytes.
1831 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1833 PR target/84760
1834 * doc/extend.texi: Add four new prototypes for vec_ld.
1835 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1836 definitions for more logical presentation.
1837 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1838 entries for V1TI variants of __builtin_altivec_ld builtin.
1839 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1840 handling of V1TI variant of LVX icode pattern.
1841 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1842 (rs6000_gimple_fold_builtin): Likewise.
1843 (altivec_init_builtins): Add code to define
1844 __builtin_altivec_lvx_v1ti function.
1846 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1848 PR inline-asm/84941
1849 * function.c (match_asm_constraints_1): Don't do the optimization
1850 if input isn't a REG, SUBREG, MEM or constant.
1852 2018-03-22 Tom de Vries <tom@codesourcery.com>
1854 PR tree-optimization/84956
1855 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1856 bb_has_abnormal_pred.
1858 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1860 PR sanitizer/85018
1861 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1862 DECL_INITIAL (decl) to decl at the end.
1863 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1864 adjust the comment.
1866 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1868 * doc/extend.texi (__builtin_tgmath): Document when complex
1869 integer types are treated as _Complex _Float64.
1871 2018-03-21 Tom de Vries <tom@codesourcery.com>
1873 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1875 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1877 PR tree-optimization/84960
1878 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1879 if it is ENTRY block, move them into single succ of ENTRY in that case.
1881 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1883 PR tree-optimization/84811
1884 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1885 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1886 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1887 (known_subrange_p): Update accordingly. Cast each value involved
1888 in the size comparison, rather than casting the result of the
1889 subtraction.
1891 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1893 PR tree-optimization/84982
1894 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1895 by flipping the least significant bit rather than all bits from
1896 bitpos to bitpos + bitsize - 1.
1898 2018-03-21 Nathan Sidwell <nathan@acm.org>
1900 * doc/extend.texi (Deprecated Features): Remove mention of
1901 long-deleted deprecations.
1903 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1905 PR jit/84288
1906 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1907 * configure: Regenerate.
1909 2018-03-21 Tom de Vries <tom@codesourcery.com>
1911 PR tree-optimization/83126
1912 * tree-parloops.c (num_phis): New function.
1913 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1915 2018-03-21 Nathan Sidwell <nathan@acm.org>
1917 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1918 mention anon-struct/union members and trailing attributes.
1920 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1922 PR tree-optimization/84969
1923 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1924 builtin memset partitions if they set different rhs values.
1926 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1928 PR rtl-optimization/84989
1929 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1930 VEC_DUPLICATE with scalar result mode.
1932 2018-03-21 Martin Liska <mliska@suse.cz>
1934 PR ipa/84963
1935 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1936 not intended return statement.
1938 2018-03-21 Martin Liska <mliska@suse.cz>
1940 PR target/84988
1941 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1942 (chkp_find_bound_slots_1): Limit number of iterations.
1944 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1946 PR target/84838
1947 * Minor grammar fixes for x86 options.
1949 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1951 PR debug/84875
1952 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1953 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1955 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1957 PR target/83789
1958 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1959 (altivec_lvx_<mode>_1op): Likewise.
1960 (altivec_stvx_<mode>_2op): Likewise.
1961 (altivec_stvx_<mode>_1op): Likewise.
1962 (altivec_lvx_<VM2:mode>): New define_expand.
1963 (altivec_stvx_<VM2:mode>): Likewise.
1964 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1965 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1966 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1967 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1968 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1969 (rs6000_gen_lvx): Likewise.
1970 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1971 (altivec_expand_stv_builtin): Likewise.
1972 (altivec_expand_builtin): Likewise.
1973 * config/rs6000/vector.md: Likewise.
1975 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1977 PR target/82518
1978 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1979 BYTES_BIG_ENDIAN.
1981 2018-03-20 Richard Biener <rguenther@suse.de>
1983 PR target/84986
1984 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1985 sign-conversions as zero, fall back to standard scalar_stmt
1986 cost for the rest.
1988 2018-03-20 Martin Liska <mliska@suse.cz>
1990 PR ipa/84825
1991 * predict.c (rebuild_frequencies): Handle case when we have
1992 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1994 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1996 PR target/84990
1997 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1998 flag_section_anchors.
1999 * varasm.c (use_blocks_for_decl_p): Remove hack for
2000 dw2_force_const_mem.
2002 PR target/84845
2003 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2004 to ...
2005 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2006 be created, use lowpart_subreg of operands[0] rather than operands[0]
2007 itself.
2008 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2009 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2010 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2011 and n constraint instead of aarch64_shift_imm_di and Usd.
2012 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2013 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2015 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2017 PR target/82989
2018 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2019 to favor GPR over NEON registers.
2020 (<shift>di3_neon): Likewise.
2022 2018-03-20 Tom de Vries <tom@codesourcery.com>
2024 PR target/84952
2025 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2026 (nvptx_process_pars): Emit bar.sync asap and alap.
2028 2018-03-20 Tom de Vries <tom@codesourcery.com>
2030 PR target/84954
2031 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2032 seen_label if seen_label is already set.
2034 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2036 PR target/84945
2037 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2038 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2039 Use 1U instead of 1. Formatting fixes.
2041 PR c/84953
2042 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2043 instead of TREE_TYPE (s1) for the return value.
2045 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2047 PR tree-optimization/84946
2048 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2049 bitsize + bitsize in poly_uint64 rather than poly_int64.
2051 PR sanitizer/78651
2052 * dwarf2asm.c: Include fold-const.c.
2053 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2054 of decl rather than decl itself.
2056 PR rtl-optimization/84643
2057 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2059 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2061 PR sanitizer/78651
2062 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2063 calling assemble_variable.
2065 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2067 PR target/81647
2068 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2069 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2071 2018-03-19 Jim Wilson <jimw@sifive.com>
2073 PR bootstrap/84856
2074 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2075 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2076 (riscv_first_stack_step): Likewise.
2077 (riscv_option_override): Use STACK_BOUNDARY instead of
2078 MIN_STACK_BOUNDARY.
2079 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2080 MIN_STACK_BOUNDARY.
2081 (BIGGEST_ALIGNMENT): Set to 128.
2082 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2083 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2084 STACK_BOUNDARY.
2086 2018-03-19 Richard Biener <rguenther@suse.de>
2088 PR tree-optimization/84933
2089 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2090 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2092 2018-03-19 Richard Biener <rguenther@suse.de>
2094 PR tree-optimization/84859
2095 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2096 (cond_if_else_store_replacement): Perform sinking operation on
2097 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2098 Generalize what a BB with a single eligible store is.
2100 2018-03-19 Richard Biener <rguenther@suse.de>
2102 PR tree-optimization/84929
2103 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2104 chrec_is_positive against non-chrec arg.
2106 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2108 PR target/84711
2109 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2111 2018-03-18 Martin Liska <mliska@suse.cz>
2113 PR rtl-optimization/84635
2114 * regrename.c (build_def_use): Use matches_mode only when
2115 matches >= 0.
2117 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2119 PR tree-optimization/84913
2120 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2121 vectorize chains of COND_EXPRs.
2123 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2125 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2127 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2129 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2131 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2133 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2135 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2136 Kito Cheng <kito.cheng@gmail.com>
2138 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2139 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2140 (nds32_adjust_reg_alloc_order): New function.
2141 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2143 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2145 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2146 nds32_print_operand, nds32_print_operand_address): Use
2147 HOST_WIDE_INT_PRINT_DEC instead.
2149 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2151 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2153 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2155 PR target/84902
2156 * config/i386/i386.c (initial_ix86_tune_features,
2157 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2158 unsigned long long.
2159 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2160 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2161 rather than 1u << ix86_tune. Formatting fix.
2162 (ix86_option_override_internal): Change ix86_arch_mask from
2163 unsigned int to unsigned HOST_WIDE_INT, initialize to
2164 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2165 (ix86_function_specific_restore): Likewise.
2167 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2169 PR target/84899
2170 * postreload.c (reload_combine_recognize_pattern): Perform
2171 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2172 truncate_int_for_mode the result for the destination's mode.
2174 PR c/84909
2175 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2176 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2177 Likewise.
2178 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2179 Likewise.
2181 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2183 PR target/84876
2184 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2185 regno_allocno_class_array and sorted_pseudos.
2186 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2187 insns where regno is used.
2189 2018-03-16 Martin Liska <mliska@suse.cz>
2191 PR ipa/84833
2192 * multiple_target.c (create_dispatcher_calls): Redirect
2193 reference in the symbol table.
2195 2018-03-16 Martin Liska <mliska@suse.cz>
2197 PR ipa/84722
2198 * multiple_target.c (create_dispatcher_calls): Redirect also
2199 an alias.
2201 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2203 PR c++/79937
2204 PR c++/82410
2205 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2206 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2207 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2209 2018-03-16 Julia Koval <julia.koval@intel.com>
2211 * doc/invoke.texi (Skylake Server): Add CLWB.
2212 Cannonlake): Remove CLWB.
2214 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2216 PR tree-optimization/84841
2217 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2218 1 << 3.
2219 (FLOAT_ONE_CONST_TYPE): Define.
2220 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2221 (sort_by_operand_rank): Put entries with higher constant_type last
2222 rather than first to match comments.
2224 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2226 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2227 split predicate.
2229 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2231 PR c++/79085
2232 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2233 check and use address of target always.
2235 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2237 PR target/84574
2238 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2239 (indirect_thunk_bnd_needed): Likewise.
2240 (indirect_thunks_used): Likewise.
2241 (indirect_thunks_bnd_used): Likewise.
2242 (indirect_return_needed): New.
2243 (indirect_return_bnd_needed): Likewise.
2244 (output_indirect_thunk_function): Add a bool argument for
2245 function return.
2246 (output_indirect_thunk_function): Don't generate alias for
2247 function return thunk.
2248 (ix86_code_end): Call output_indirect_thunk_function to generate
2249 function return thunks.
2250 (ix86_output_function_return): Set indirect_return_bnd_needed
2251 and indirect_return_needed instead of indirect_thunk_bnd_needed
2252 and indirect_thunk_needed.
2254 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2256 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2257 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2258 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2260 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2261 Paul Hua <paul.hua.gm@gmail.com>
2263 PR c/84852
2264 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2266 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2268 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2269 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2270 resp. SFmode cases.
2272 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2274 PR target/84711
2275 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2276 instead of GET_MODE_SIZE when comparing Units.
2278 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2280 PR target/68256
2281 * varasm.c (hash_section): Return an unchangeble hash value
2282 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2283 Return !aarch64_can_use_per_function_literal_pools_p ().
2285 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2287 PR target/84860
2288 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2289 rather than address of cmode as last argument to prepare_cmp_insn.
2291 2018-03-15 Julia Koval <julia.koval@intel.com>
2293 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2294 F_AVX512VNNI, F_AVX512BITALG): New.
2296 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2298 PR target/83451
2299 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2300 insn for floating-point loads and stores.
2302 2018-03-14 Carl Love <cel@us.ibm.com>
2304 * config/rs6000/rs6000-c.c: Add macro definitions for
2305 ALTIVEC_BUILTIN_VEC_PERMXOR.
2306 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2307 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2308 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2309 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2310 UNSPEC_VPERMXOR.
2311 * config/doc/extend.texi: Add prototypes for vec_permxor.
2313 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2315 PR c/84852
2316 * diagnostic-show-locus.c (class layout_point): Convert m_line
2317 from int to linenum_type.
2318 (line_span::comparator): Use linenum "compare" function when
2319 comparing line numbers.
2320 (test_line_span): New function.
2321 (layout_range::contains_point): Convert param "row" from int to
2322 linenum_type.
2323 (layout_range::intersects_line_p): Likewise.
2324 (layout::will_show_line_p): Likewise.
2325 (layout::print_source_line): Likewise.
2326 (layout::should_print_annotation_line_p): Likewise.
2327 (layout::print_annotation_line): Likewise.
2328 (layout::print_leading_fixits): Likewise.
2329 (layout::annotation_line_showed_range_p): Likewise.
2330 (struct line_corrections): Likewise for field m_row.
2331 (line_corrections::line_corrections): Likewise for param "row".
2332 (layout::print_trailing_fixits): Likewise.
2333 (layout::get_state_at_point): Likewise.
2334 (layout::get_x_bound_for_row): Likewise.
2335 (layout::print_line): Likewise.
2336 (diagnostic_show_locus): Likewise for locals "last_line" and
2337 "row".
2338 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2339 * input.c (selftest::test_linenum_comparisons): New function.
2340 (selftest::input_c_tests): Call it.
2341 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2342 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2343 * selftest.h (ASSERT_GT): New macro.
2344 (ASSERT_GT_AT): New macro.
2345 (ASSERT_LT): New macro.
2346 (ASSERT_LT_AT): New macro.
2348 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2350 PR rtl-optimization/84780
2351 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2353 2018-03-14 Martin Liska <mliska@suse.cz>
2355 * tree.c (record_node_allocation_statistics): Use
2356 get_stats_node_kind.
2357 (get_stats_node_kind): New function extracted from
2358 record_node_allocation_statistics.
2359 (free_node): Use get_stats_node_kind.
2361 2018-03-14 Richard Biener <rguenther@suse.de>
2363 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2364 that the value-set of ANTIC_IN doesn't grow.
2366 Revert
2367 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2368 member.
2369 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2370 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2372 2018-03-14 Julia Koval <julia.koval@intel.com>
2374 * config.gcc (icelake-client, icelake-server): New.
2375 (icelake): Remove.
2376 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2377 (initial_ix86_arch_features): Ditto.
2378 (PTA_SKYLAKE): Add SGX.
2379 (PTA_ICELAKE): Remove.
2380 (PTA_ICELAKE_CLIENT): New.
2381 (PTA_ICELAKE_SERVER): New.
2382 (ix86_option_override_internal): Split up icelake on icelake client and
2383 icelake server.
2384 (get_builtin_code_for_version): Ditto.
2385 (fold_builtin_cpu): Ditto.
2386 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2387 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2388 * config/i386/i386.h (processor_type): Ditto.
2389 * doc/invoke.texi: Ditto.
2391 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2393 PR sanitizer/83392
2394 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2395 INTEGER_CST offset, add it together with bitpos / 8 and
2396 sign extend based on POINTER_SIZE.
2398 PR target/84844
2399 Revert
2400 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2402 PR target/78090
2403 * config/i386/constraints.md (Yc): New register constraint.
2404 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2405 Use Yc constraint for alternative 2 of operand 0. Remove
2406 preferred_for_speed attribute.
2408 2018-03-14 Richard Biener <rguenther@suse.de>
2410 PR tree-optimization/84830
2411 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2412 with the old one to avoid oscillations.
2414 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2416 PR target/83712
2417 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2418 pseudos.
2419 (assign_by_spills): Return a flag of reload assignment failure.
2420 Do not process the reload assignment failures. Do not spill other
2421 reload pseudos if they has the same reg class. Update n if
2422 necessary.
2423 (lra_assign): Add a return arg. Set up from the result of
2424 assign_by_spills call.
2425 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2426 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2427 usage_insns if it is not NULL.
2428 (spill_hard_reg_in_range): New function.
2429 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2430 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2431 function prototypes.
2432 (lra_assign): Change prototype.
2433 * lra.c (lra): Add code to deal with fails by splitting hard reg
2434 live ranges.
2436 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2438 * config/riscv/riscv.opt (mrelax): New option.
2439 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2440 "norelax" when riscv_mrelax is disabled.
2441 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2443 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2445 PR target/84743
2446 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2447 reassociation for int modes.
2449 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2451 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2452 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2453 for big-endian.
2454 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2455 * config/aarch64/aarch64-sve.md
2456 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2457 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2458 (*extend<mode><Vwide>2): Rename to...
2459 (aarch64_sve_extend<mode><Vwide>2): ...this.
2460 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2461 renaming the old pattern to...
2462 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2463 unsigned packs.
2464 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2465 define_expand, renaming the old pattern to...
2466 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2467 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2468 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2469 account when deciding which SVE instruction the optab should use.
2470 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2472 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2474 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2475 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2476 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2477 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2478 (tlsdesc_small_<mode>): Turn a define_expand and use
2479 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2480 (tlsdesc_small_advsimd_<mode>): ...this.
2481 (tlsdesc_small_sve_<mode>): New pattern.
2483 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2485 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2486 (UNSPEC_UMUL_HIGHPART): New constants.
2487 (MUL_HIGHPART): New int iteraor.
2488 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2489 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2490 define_expand.
2491 (*<su>mul<mode>3_highpart): New define_insn.
2493 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2495 PR lto/84805
2496 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2497 incomplete types.
2499 2018-03-13 Martin Liska <mliska@suse.cz>
2501 PR ipa/84658.
2502 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2503 vector.
2504 (sem_item_optimizer::~sem_item_optimizer): Release it.
2505 (sem_item_optimizer::merge_classes): Register variable aliases.
2506 (sem_item_optimizer::fixup_pt_set): New function.
2507 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2508 * ipa-icf.h: Declare new variables and functions.
2510 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2512 PR middle-end/84834
2513 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2514 integer_pow2p@2 and test integer_pow2p in condition.
2515 (A < 0 ? C : 0): Similarly for @1.
2517 PR middle-end/84831
2518 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2519 characters starting at p contain '\0' character, don't look beyond
2520 that.
2522 PR target/84827
2523 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2524 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2526 PR target/84828
2527 * reg-stack.c (change_stack): Change update_end var from int to
2528 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2529 also call set_block_for_insn on the newly added insns and rescan.
2531 PR target/84786
2532 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2533 on the last operand.
2535 PR c++/84704
2536 * tree.c (stabilize_reference_1): Return save_expr (e) for
2537 STATEMENT_LIST even if it doesn't have side-effects.
2539 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2541 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2543 2018-03-12 Renlin Li <renlin.li@arm.com>
2545 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2546 aarch64_output_scalar_simd_mov_immediate.
2548 2018-03-12 Martin Sebor <msebor@redhat.com>
2550 PR tree-optimization/83456
2551 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2552 for perfectly overlapping calls to memcpy.
2553 (gimple_fold_builtin_memory_chk): Same.
2554 (gimple_fold_builtin_strcpy): Handle no-warning.
2555 (gimple_fold_builtin_stxcpy_chk): Same.
2556 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2558 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2560 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2561 parameter. Use it for SFmode.
2562 (rs6000_function_arg_advance_1): Adjust.
2563 (rs6000_function_arg): Adjust.
2564 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2566 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2568 PR rtl-optimization/84169
2569 PR rtl-optimization/84780
2570 * combine.c (can_combine_p): Check for a 2-insn combination whether
2571 the destination register is used between the two insns, too.
2573 2018-03-12 Richard Biener <rguenther@suse.de>
2575 PR tree-optimization/84803
2576 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2577 for refs DR analysis didn't process.
2579 2018-03-12 Richard Biener <rguenther@suse.de>
2581 PR tree-optimization/84777
2582 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2583 force-vectorize loops ignore whether we are optimizing for size.
2585 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2587 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2588 (TARGET_MD_ASM_ADJUST): Define.
2590 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2591 Kito Cheng <kito.cheng@gmail.com>
2592 Chung-Ju Wu <jasonwucj@gmail.com>
2594 * config/nds32/nds32.c (nds32_compute_stack_frame,
2595 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2596 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2597 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2598 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2599 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2600 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2601 * config/nds32/nds32.md (prologue, epilogue): Use macro
2602 NDS32_V3PUSH_AVAILABLE_P to do checking.
2604 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2606 PR debug/58150
2607 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2608 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2609 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2610 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2611 addition of most attributes on !orig_type_die or the attribute not
2612 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2614 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2615 Chung-Ju Wu <jasonwucj@gmail.com>
2617 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2618 __NDS32_VH__ macro.
2619 * config/nds32/nds32.opt (mvh): New option.
2621 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2622 Chung-Ju Wu <jasonwucj@gmail.com>
2624 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2625 function.
2626 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2627 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2628 definition.
2630 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2631 Chung-Ju Wu <jasonwucj@gmail.com>
2633 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2634 function.
2635 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2636 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2638 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2639 Kito Cheng <kito.cheng@gmail.com>
2640 Chung-Ju Wu <jasonwucj@gmail.com>
2642 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2643 UNSPEC_FFMISM and UNSPEC_FLMISM.
2644 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2645 for ffb, ffmism and flmism.
2646 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2647 (unspec_ffmism): Ditto.
2648 (unspec_flmism): Ditto.
2649 (nds32_expand_builtin_impl): Check if string extension is available.
2650 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2651 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2653 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2655 Reverting patch:
2656 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2658 PR target/83712
2659 * lra-assigns.c (assign_by_spills): Return a flag of reload
2660 assignment failure. Do not process the reload assignment
2661 failures. Do not spill other reload pseudos if they has the same
2662 reg class.
2663 (lra_assign): Add a return arg. Set up from the result of
2664 assign_by_spills call.
2665 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2666 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2667 usage_insns if it is not NULL.
2668 (spill_hard_reg_in_range): New function.
2669 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2670 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2671 function prototypes.
2672 (lra_assign): Change prototype.
2673 * lra.c (lra): Add code to deal with fails by splitting hard reg
2674 live ranges.
2676 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2678 PR target/84807
2679 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2681 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2683 PR debug/84620
2684 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2685 (dw_val_node): Add val_symbolic_view.
2686 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2687 (symview_upper_bound): New.
2688 (new_line_info_table): Initialize symviews_since_reset.
2689 (dwarf2out_source_line): Count symviews_since_reset and set
2690 symview_upper_bound.
2691 (dw_val_equal_p): Handle symview.
2692 (add_AT_symview): New.
2693 (print_dw_val): Handle symview.
2694 (attr_checksum, attr_checksum_ordered): Likewise.
2695 (same_dw_val_p, size_of_die): Likewise.
2696 (value_format, output_die): Likewise.
2697 (add_high_low_attributes): Use add_AT_symview for entry_view.
2698 (dwarf2out_finish): Reset symview_upper_bound, clear
2699 zero_view_p.
2701 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2703 PR target/83969
2704 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2705 Add strict argument and use it.
2706 (rs6000_split_multireg_move): Update for new strict argument.
2707 (mem_operand_gpr): Disallow all non-offsettable addresses.
2708 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2710 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2712 PR target/84772
2713 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2714 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2715 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2717 PR c++/84767
2718 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2719 decl, use remap_type if we want to use the type.
2721 2018-03-09 Martin Sebor <msebor@redhat.com>
2723 PR tree-optimization/84526
2724 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2725 Remove dead code.
2726 (builtin_access::generic_overlap): Be prepared to handle non-array
2727 base objects.
2729 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2731 PR rtl-optimization/84682
2732 * lra-constraints.c (process_address_1): Check is_address flag
2733 for address constraints.
2734 (process_alt_operands): Likewise.
2735 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2736 preprocess_constraints.
2737 * recog.h (preprocess_constraints): Add oploc parameter.
2738 Adjust callers.
2739 * recog.c (preprocess_constraints): Test address_operand for
2740 CT_ADDRESS constraints.
2742 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2744 PR target/83712
2745 * lra-assigns.c (assign_by_spills): Return a flag of reload
2746 assignment failure. Do not process the reload assignment
2747 failures. Do not spill other reload pseudos if they has the same
2748 reg class.
2749 (lra_assign): Add a return arg. Set up from the result of
2750 assign_by_spills call.
2751 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2752 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2753 usage_insns if it is not NULL.
2754 (spill_hard_reg_in_range): New function.
2755 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2756 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2757 function prototypes.
2758 (lra_assign): Change prototype.
2759 * lra.c (lra): Add code to deal with fails by splitting hard reg
2760 live ranges.
2762 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2764 PR target/83193
2765 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2766 Accept complain bool parameter. Only emit errors if it is true.
2767 (arm_parse_cpu_option_name): Likewise.
2768 (arm_target_thumb_only): Adjust callers of the above.
2769 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2770 prototype to take a default true bool parameter.
2771 (arm_parse_arch_option_name): Likewise.
2773 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2774 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2776 PR jit/64089
2777 PR jit/84288
2778 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2779 * configure: Regenerate.
2780 * configure.ac ("linker --version-script option"): New.
2781 ("linker soname option"): New.
2783 2018-03-09 Richard Biener <rguenther@suse.de>
2785 PR tree-optimization/84775
2786 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2787 immediate uses of predicate stmts and mark them modified.
2789 Revert
2790 PR tree-optimization/84178
2791 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2792 to caller.
2793 (version_loop_for_if_conversion): Delay update_ssa call.
2794 (tree_if_conversion): Delay update_ssa until after predicate
2795 insertion.
2797 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2799 PR target/84763
2800 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2801 when the function accesses prior frames.
2803 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2805 PR debug/84456
2806 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2807 gen_llsym, otherwise call maybe_gen_llsym.
2809 PR inline-asm/84742
2810 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2811 has ',' character inside of it.
2813 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2815 PR target/84748
2816 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2817 as clobbering CC_REGNUM.
2819 2018-03-08 Richard Biener <rguenther@suse.de>
2821 PR middle-end/84552
2822 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2823 (follow_copies_to_constant): Do not follow SSA names registered
2824 for update.
2826 2018-03-08 Richard Biener <rguenther@suse.de>
2828 PR tree-optimization/84178
2829 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2830 to caller.
2831 (version_loop_for_if_conversion): Delay update_ssa call.
2832 (tree_if_conversion): Delay update_ssa until after predicate
2833 insertion.
2835 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2837 PR tree-optimization/84178
2838 * tree-if-conv.c (release_bb_predicate): Remove the
2839 the assertion that the stmts have NULL use_ops.
2840 Discard the statements, asserting that they haven't
2841 yet been added to a BB.
2843 2018-03-08 Richard Biener <rguenther@suse.de>
2845 PR tree-optimization/84746
2846 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2847 (phi_translate): Pass in destination ANTIC_OUT set.
2848 (phi_translate_1): Likewise. For a simplified result lookup
2849 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2850 (phi_translate_set): Adjust.
2851 (do_pre_regular_insertion): Likewise.
2852 (do_pre_partial_partial_insertion): Likewise.
2854 2018-03-08 Martin Liska <mliska@suse.cz>
2856 PR gcov-profile/84735
2857 * doc/gcov.texi: Document usage of profile files.
2858 * gcov-io.h: Document changes in the format.
2860 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2862 PR debug/84404
2863 PR debug/84408
2864 * dwarf2out.c (struct dw_line_info_table): Update comments for
2865 view == -1.
2866 (FORCE_RESET_NEXT_VIEW): New.
2867 (FORCE_RESETTING_VIEW_P): New.
2868 (RESETTING_VIEW_P): Check for -1 too.
2869 (ZERO_VIEW_P): Likewise.
2870 (new_line_info_table): Force-reset next view.
2871 (dwarf2out_begin_function): Likewise.
2872 (dwarf2out_source_line): Simplify zero_view_p initialization.
2873 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2874 view directly. Omit view when omitting .loc at line 0.
2876 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2878 PR tree-optimization/84740
2879 * tree-switch-conversion.c (process_switch): Call build_constructors
2880 only if info.phi_count is non-zero.
2882 PR tree-optimization/84739
2883 * tree-tailcall.c (find_tail_calls): Check call arguments against
2884 DECL_ARGUMENTS (current_function_decl) rather than
2885 DECL_ARGUMENTS (func) when checking for tail recursion.
2887 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2889 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2890 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2891 Volker Reichelt's entry and add entries for people that perform
2892 GCC fuzzy testing and report numerous bugs.
2894 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2896 PR target/82411
2897 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2898 readonly data in sdata, if that is disabled.
2899 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2900 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2901 -mreadonly-in-sdata option.
2903 2018-03-07 Martin Sebor <msebor@redhat.com>
2905 PR tree-optimization/84468
2906 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2907 basic block when looking for nul assignment.
2909 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2911 PR target/84277
2912 * except.h (output_function_exception_table): Adjust prototype.
2913 * except.c (output_function_exception_table): Remove FNNAME parameter
2914 and add SECTION parameter. Ouput one part of the table at a time.
2915 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2916 the first part of the exception table and emit unwind directives.
2917 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2918 (i386_pe_seh_cold_init): Likewise.
2919 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2920 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2921 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2922 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2923 * config/i386/winnt.c: Include except.h.
2924 (struct seh_frame_state): Add reg_offset, after_prologue and
2925 in_cold_section fields.
2926 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2927 (i386_pe_seh_cold_init): New function.
2928 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2929 to seh->in_cold_section.
2930 (seh_emit_push): Record the offset of the push.
2931 (seh_emit_save): Record the offet of the save.
2932 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2933 Test seh->after_prologue to disregard the epilogue.
2934 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2935 (i386_pe_end_cold_function): New function.
2937 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2939 PR fortran/84565
2940 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2941 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2943 PR c++/84704
2944 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2945 on tmp_var.
2946 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2947 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2949 PR middle-end/84723
2950 * multiple_target.c: Include tree-inline.h and intl.h.
2951 (expand_target_clones): Diagnose and fail if node->definition and
2952 !tree_versionable_function_p (node->decl).
2954 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2956 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2957 sprint_ul.
2958 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2959 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2960 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2962 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2964 PR target/84710
2965 * combine.c (try_combine): Use reg_or_subregno instead of handling
2966 just paradoxical SUBREGs and REGs.
2968 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2970 * config/arc/arc.c (arc_finalize_pic): Remove function.
2971 (arc_must_save_register): We use single base PIC register, remove
2972 checks to save/restore the PIC register.
2973 (arc_expand_prologue): Likewise.
2974 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2975 Remove.
2976 (arc_verify_short): Likewise.
2977 (arc_attr_type): Likewise.
2978 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2979 (walk_stores): Likewise.
2980 (arc_address_cost): Make it static.
2981 (arc_verify_short): Likewise.
2982 (branch_dest): Likewise.
2983 (arc_attr_type): Likewise.
2984 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2985 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2986 (arc_final_prescan_insn): Remove inserting the nops due to
2987 hardware hazards. It is done in reorg step.
2988 (insn_length_variant_t): Remove.
2989 (insn_length_parameters_t): Likewise.
2990 (arc_insn_length_parameters): Likewise.
2991 (arc_get_insn_variants): Likewise.
2992 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2994 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2996 PR inline-asm/84683
2997 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2998 assertion failure.
3000 PR tree-optimization/84687
3001 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3002 on new_node->decl.
3003 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3005 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3007 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3008 Rename to ppc_speculation_barrier.
3009 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3010 __builtin_ppc_speculation_barrier.
3012 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3014 PR target/84700
3015 * combine.c (combine_simplify_rtx): Don't try to simplify if
3016 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3017 are equal to x.
3019 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3021 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3022 to 32 bytes when compiling for POWER9.
3024 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3026 PR target/84564
3027 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3028 regparm >= 3 with no arg reg available also for calls with
3029 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3031 PR target/84524
3032 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3033 orig,vex.
3034 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3036 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3038 PR target/84264
3039 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3041 2018-03-05 Richard Biener <rguenther@suse.de>
3043 PR tree-optimization/84486
3044 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3045 When inserting a __builtin_assume_aligned call set the LHS
3046 SSA name alignment info accordingly.
3048 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3050 PR tree-optimization/84114
3051 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3052 Avoid reassociation of FLOAT_MODE addition.
3054 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3056 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3057 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3058 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3059 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3060 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3061 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3062 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3063 and -mwbnoinvd.
3064 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3065 __builtin_ia32_wbinvd): New builtins.
3066 (SPECIAL_ARGS2): New.
3067 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3068 (SPECIAL_ARGS2): New.
3069 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3070 (ix86_valid_target_attribute_inner_p): Ditto.
3071 (ix86_init_mmx_sse_builtins): Add special_args2.
3072 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3073 TARGET_WBNOINVD_P): New.
3074 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3075 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3076 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3077 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3078 * config/i386/pconfigintrin.h: New file.
3079 * config/i386/wbnoinvdintrin.h: Ditto.
3080 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3081 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3083 2018-03-05 Richard Biener <rguenther@suse.de>
3085 PR tree-optimization/84670
3086 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3087 member.
3088 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3089 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3090 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3091 doesn't grow if all successors (recursively) were visited at least
3092 once.
3094 2018-03-05 Richard Biener <rguenther@suse.de>
3096 PR tree-optimization/84650
3097 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3098 if executed in the loop pipeline.
3100 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3102 * doc/configfiles.texi (Configuration Files): Move info about
3103 conditionalizing $target-protos.h to...
3104 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3105 differs from $target-protos.h.
3107 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3108 Chung-Ju Wu <jasonwucj@gmail.com>
3110 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3111 * config/nds32/nds32-multiple.md (setmemsi): Define.
3112 * config/nds32/nds32-memory-manipulation.c
3113 (nds32_gen_dup_4_byte_to_word_value): New.
3114 (emit_setmem_word_loop): New.
3115 (emit_setmem_byte_loop): New.
3116 (nds32_expand_setmem_loop): New.
3117 (nds32_expand_setmem_loop_v3m): New.
3118 (nds32_expand_setmem_unroll): New.
3119 (nds32_expand_setmem): New.
3121 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3122 Chung-Ju Wu <jasonwucj@gmail.com>
3124 * config/nds32/nds32-memory-manipulation.c
3125 (nds32_emit_load_store): New.
3126 (nds32_emit_post_inc_load_store): New.
3127 (nds32_emit_mem_move): New.
3128 (nds32_emit_mem_move_block): New.
3129 (nds32_expand_movmemsi_loop_unknown_size): New.
3130 (nds32_expand_movmemsi_loop_known_size): New.
3131 (nds32_expand_movmemsi_loop): New.
3132 (nds32_expand_movmemsi_unroll): New.
3133 (nds32_expand_movmemqi): Rename ...
3134 (nds32_expand_movmemsi): ... to this.
3135 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3136 (movmemsi): ... to this.
3137 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3138 (nds32_expand_movmemsi): ... to this.
3140 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3141 Monk Chiang <sh.chiang04@gmail.com>
3142 Chung-Ju Wu <jasonwucj@gmail.com>
3144 * config/nds32/nds32-protos.h
3145 (nds32_expand_load_multiple): New arguments.
3146 (nds32_expand_store_multiple): Ditto.
3147 (nds32_valid_multiple_load_store): Rename ...
3148 (nds32_valid_multiple_load_store_p): ... to this.
3149 * config/nds32/nds32-memory-manipulation.c
3150 (nds32_expand_load_multiple): Refine implementation.
3151 (nds32_expand_store_multiple): Ditto.
3152 * config/nds32/nds32-multiple.md
3153 (load_multiple): Update nds32_expand_load_multiple interface.
3154 (store_multiple): Update nds32_expand_store_multiple interface.
3155 * config/nds32/nds32-predicates.c
3156 (nds32_valid_multiple_load_store): Rename ...
3157 (nds32_valid_multiple_load_store_p): ... to this and refine
3158 implementation.
3159 * config/nds32/predicates.md
3160 (nds32_load_multiple_and_update_address_operation): New predicate.
3161 (nds32_store_multiple_and_update_address_operation): New predicate.
3163 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3164 Chung-Ju Wu <jasonwucj@gmail.com>
3166 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3167 (combo): New attribute.
3168 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3170 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3172 * config/nds32/nds32.opt: Change -mcmodel= default value.
3174 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3175 Monk Chiang <sh.chiang04@gmail.com>
3176 Chung-Ju Wu <jasonwucj@gmail.com>
3178 * config/nds32/constants.md (unspec_element): New enum.
3179 * config/nds32/constraints.md (Umw): New constraint.
3180 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3181 * config/nds32/nds32-intrinsic.md: Likewise.
3182 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3183 (nds32_valid_smw_lwm_base_p): New.
3184 (nds32_output_smw_single_word): New.
3185 (nds32_output_lmw_single_word): New.
3186 (nds32_expand_unaligned_load): New.
3187 (nds32_expand_unaligned_store): New.
3188 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3189 (nds32_output_smw_single_word): Declare.
3190 (nds32_output_lmw_single_word): Declare.
3191 (nds32_expand_unaligned_load): Declare.
3192 (nds32_expand_unaligned_store): Declare.
3193 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3194 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3195 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3196 NDS32_BUILTIN_UASTORE_DW.
3197 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3198 predicate.
3200 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3201 Kito Cheng <kito.cheng@gmail.com>
3202 Chung-Ju Wu <jasonwucj@gmail.com>
3204 * config/nds32/nds32-intrinsic.c
3205 (nds32_expand_builtin_null_ftype_reg): Delete.
3206 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3207 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3208 (nds32_read_argument): New.
3209 (nds32_legitimize_target): Ditto.
3210 (nds32_legitimize_argument): Ditto.
3211 (nds32_check_constant_argument): Ditto.
3212 (nds32_expand_unop_builtin): Ditto.
3213 (nds32_expand_unopimm_builtin): Ditto.
3214 (nds32_expand_binop_builtin): Ditto.
3215 (nds32_builtin_decl_impl): Ditto.
3216 (builtin_description): Ditto.
3217 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3218 (nds32_init_builtins_impl): Ditto.
3219 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3220 (nds32_builtin_decl): New.
3221 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3222 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3224 2018-03-02 Jeff Law <law@redhat.com>
3226 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3227 (redundant_insn, fill_simple_delay_slots): Likewise.
3228 (fill_slots_from_thread): Likewise.
3229 * resource.c (mark_referenced_resources): Likewise.
3230 (mark_set_resources, find_dead_or_set_registers): Likewise.
3232 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3234 * substring-locations.h (format_warning_va): Formatting fix for
3235 ATTRIBUTE_GCC_DIAG.
3236 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3237 argument.
3238 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3239 * substring-locations.c: Include intl.h.
3240 (format_warning_va): Turned into small wrapper around
3241 format_warning_n_va, renamed to ...
3242 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3243 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3244 use ngettext.
3245 (format_warning_at_substring_n): New function.
3246 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3247 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3248 format_warning_at_substring with just a shorter name instead of
3249 const function pointer.
3250 (fmtwarn_n): New function.
3251 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3252 appropriate, get rid of all the fmtstr temporaries, move conditionals
3253 with G_() wrapped string literals directly into fmtwarn arguments,
3254 cast dir.len to (int), formatting fixes.
3256 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3258 * doc/invoke.texi: Remove "Cilk Plus" references.
3260 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3261 Richard Biener <rguenther@suse.de>
3263 PR ipa/84628
3264 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3265 for error or warning attributes if CALL_FROM_THUNK_P is set.
3266 Formatting fixes.
3268 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3270 PR target/56540
3271 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3272 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3274 PR target/56540
3275 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3276 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3278 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3279 instead of -1U in last predictors element's probability member.
3281 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3283 PR ipa/83983
3284 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3285 arguments if they are comparable.
3287 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3289 PR tree-optimization/84634
3290 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3291 masks and masked_loop_p with a single loop_masks, making sure it's
3292 null for bb vectorization.
3294 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3296 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3297 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3298 loop->force_vectorize to check whether there is no alias.
3300 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3302 PR target/84614
3303 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3304 prototypes.
3305 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3306 comments.
3307 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3308 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3309 instead of a loop around prev_real_insn.
3310 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3311 prev_real_insn.
3313 PR inline-asm/84625
3314 * config/i386/i386.c (ix86_print_operand): Use conditional
3315 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3316 zero vector.
3318 2018-03-02 Richard Biener <rguenther@suse.de>
3320 PR tree-optimization/84427
3321 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3322 (bitmap_set_subtract_values): Rewrite to handle multiple
3323 exprs per value.
3324 (clean): Likewise.
3325 (prune_clobbered_mems): Likewise.
3326 (phi_translate): Take edge instead of pred/phiblock.
3327 (phi_translate_1): Likewise.
3328 (phi_translate_set): Likewise. Insert all translated
3329 exprs for a value into the set, keeping possibly multiple
3330 expressions per value.
3331 (compute_antic_aux): Adjust for phi_translate changes.
3332 When intersecting union the expressions and prune those
3333 not in the final value set, keeping possibly multiple
3334 expressions per value. Do not use value-insertion
3335 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3336 all expressions. Add verification that the value-sets
3337 only shrink during iteration.
3338 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3339 (do_pre_regular_insertion): Likewise.
3340 (do_pre_partial_partial_insertion): Likewise.
3342 2018-03-02 Richard Biener <rguenther@suse.de>
3344 PR target/82005
3345 * config/darwin.c (saved_debug_info_level): New static global.
3346 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3347 (darwin_asm_lto_end): Restore debug info generation settings.
3349 2018-03-01 Martin Liska <mliska@suse.cz>
3351 PR sanitizer/82484
3352 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3353 volatile arguments.
3355 2018-03-01 Richard Biener <rguenther@suse.de>
3357 PR debug/84645
3358 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3359 type annotation with LTO when debug was disabled at compile-time.
3361 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3363 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3364 XINT with INTVAL.
3365 (mips_final_postscan_insn): Likewise.
3367 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3369 PR rtl-optimization/84528
3370 * alias.c (init_alias_target): Add commentary.
3371 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3372 a unique base value if the frame pointer is not eliminated
3373 to the stack pointer.
3375 2018-03-01 Tom de Vries <tom@codesourcery.com>
3377 PR rtl-optimization/83327
3378 * lra-int.h (hard_regs_spilled_into): Declare.
3379 * lra.c (hard_regs_spilled_into): Define.
3380 (init_reg_info): Init hard_regs_spilled_into.
3381 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3382 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3383 (process_bb_lives): Handle hard_regs_spilled_into.
3384 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3385 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3387 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3389 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3390 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3391 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3392 * config/rs6000/aix72.h: New file.
3394 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3396 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3397 instead of warning_at with conditional singular and plural messages
3398 where possible.
3400 PR target/52991
3401 * stor-layout.c (update_alignment_for_field): For
3402 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3403 && !DECL_PACKED (field), do the alignment update, just use
3404 only desired_align instead of MAX (type_align, desired_align)
3405 as the alignment.
3406 (place_field): Don't do known_align < desired_align handling
3407 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3408 is non-NULL, instead do it after rli->prev_field handling and
3409 only if not within a bitfield word. For DECL_PACKED (field)
3410 use type_align of BITS_PER_UNIT.
3412 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3414 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3415 superfluous parentheses and trailing spaces.
3417 2018-02-28 Richard Biener <rguenther@suse.de>
3419 PR tree-optimization/84584
3420 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3421 SCoPs with fake exit edge.
3423 2018-02-28 Martin Liska <mliska@suse.cz>
3425 PR testsuite/84597
3426 * timevar.c (timer::print): Fix format to properly print 100%
3427 values.
3429 2018-02-28 Richard Biener <rguenther@suse.de>
3431 PR middle-end/84607
3432 * genmatch.c (capture_info::walk_match): Do not mark
3433 captured expressions without operands as expr_p given
3434 they act more like predicates and should be subject to
3435 "lost tail" side-effect preserving.
3437 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3439 PR rtl-optimization/81611
3440 * auto-inc-dec.c (attempt_change): Move dead note from
3441 mem_insn if it's the next use of regno
3442 (find_address): Take address use of reg holding
3443 non-incremented value. Add parm to limit search to the named
3444 reg only.
3445 (merge_in_block): Attempt to use a mem insn that is the next
3446 use of the original regno.
3448 2018-02-27 Martin Sebor <msebor@redhat.com>
3450 PR c++/83871
3451 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3452 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3454 2018-02-27 Martin Sebor <msebor@redhat.com>
3456 PR translation/84207
3457 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3458 n argument to unsigned HOST_WIDE_INT.
3459 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3460 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3461 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3462 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3464 2018-02-27 Richard Biener <rguenther@suse.de>
3466 PR tree-optimization/84512
3467 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3468 Do not use the estimate returned from record_stmt_cost for
3469 the scalar iteration cost but sum properly using add_stmt_cost.
3471 2018-02-27 Richard Biener <rguenther@suse.de>
3473 PR tree-optimization/84466
3474 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3475 Adjust last change to less strictly validate use operands.
3477 2018-02-27 Martin Liska <mliska@suse.cz>
3479 PR gcov-profile/84548
3480 * gcov.c (process_file): Allow partial overlap and consider it
3481 also as group functions.
3482 (output_lines): Properly calculate range of lines for a group.
3484 2018-02-27 Martin Liska <mliska@suse.cz>
3486 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3487 'ggc' suffixes. Change first column width.
3488 (timer::print): Fix formatting of the column.
3490 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3492 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3493 preserve inline entry blocks for the sake of debug inline
3494 entry point markers alone.
3495 (remove_unused_locals): Suggest in comments a better place to
3496 force the preservation of inline entry blocks that are
3497 otherwise unused, but do not preserve them.
3499 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3501 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3503 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3505 PR target/84039
3506 * config/i386/constraints.md (Bs): Replace
3507 ix86_indirect_branch_register with
3508 TARGET_INDIRECT_BRANCH_REGISTER.
3509 (Bw): Likewise.
3510 * config/i386/i386.md (indirect_jump): Likewise.
3511 (tablejump): Likewise.
3512 (*sibcall_memory): Likewise.
3513 (*sibcall_value_memory): Likewise.
3514 Peepholes of indirect call and jump via memory: Likewise.
3515 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3516 (*sibcall_value_GOT_32): Likewise.
3517 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3518 (GOT_memory_operand): Likewise.
3519 (call_insn_operand): Likewise.
3520 (sibcall_insn_operand): Likewise.
3521 (GOT32_symbol_operand): Likewise.
3522 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3524 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3526 PR rtl-optimization/83496
3527 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3528 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3529 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3530 redundant insn, if any.
3531 (relax_delay_slots): Likewise.
3532 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3534 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3536 PR tree-optimization/83965
3537 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3538 that grouped statements are part of a reduction chain. Return
3539 true if the statement is not marked as a reduction itself but
3540 is part of a group.
3541 (vect_recog_dot_prod_pattern): Don't check whether the statement
3542 is part of a group here.
3543 (vect_recog_sad_pattern): Likewise.
3544 (vect_recog_widen_sum_pattern): Likewise.
3546 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3548 PR debug/84545
3549 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3551 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3553 PR target/84530
3554 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3555 the bool argument.
3556 (ix86_output_indirect_function_return): New prototype.
3557 (ix86_split_simple_return_pop_internal): Likewise.
3558 * config/i386/i386.c (indirect_return_via_cx): New.
3559 (indirect_return_via_cx_bnd): Likewise.
3560 (indirect_thunk_name): Handle return va CX_REG.
3561 (output_indirect_thunk_function): Create alias for
3562 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3563 (ix86_output_indirect_jmp): Remove the bool argument.
3564 (ix86_output_indirect_function_return): New function.
3565 (ix86_split_simple_return_pop_internal): Likewise.
3566 * config/i386/i386.md (*indirect_jump): Don't pass false
3567 to ix86_output_indirect_jmp.
3568 (*tablejump_1): Likewise.
3569 (simple_return_pop_internal): Change it to define_insn_and_split.
3570 Call ix86_split_simple_return_pop_internal to split it for
3571 -mfunction-return=.
3572 (simple_return_indirect_internal): Call
3573 ix86_output_indirect_function_return instead of
3574 ix86_output_indirect_jmp.
3576 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3578 PR bootstrap/84405
3579 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3580 memset and value initialization afterwards.
3582 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3584 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3586 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3588 PR target/84521
3589 * common/config/aarch64/aarch64-common.c
3590 (aarch_option_optimization_table[]): Switch
3591 off fomit-frame-pointer
3593 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3594 Chung-Ju Wu <jasonwucj@gmail.com>
3596 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3597 volatile memory.
3598 (store_multiple): Ditto.
3600 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3602 * config.gcc: Add --with-cpu support for nds32 target.
3603 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3604 * config/nds32/nds32.opt: Add -mcpu= option.
3606 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3608 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3609 isel=yes): Warn for these deprecated options.
3611 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3613 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3614 ISA_2_5_MASKS_EMBEDDED.
3616 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3618 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3619 p->max as pointers rather than using iterative_hash_expr.
3621 2018-02-23 Carl Love <cel@us.ibm.com>
3623 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3624 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3625 BU_P8V_OVERLOAD_2.
3626 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3627 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3628 P8V_BUILTIN_VEC_VUNSIGNED2.
3630 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3632 PR target/81572
3633 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3634 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3635 LRA_UNKNOWN_ALT.
3636 * lra-constraints.c (curr_insn_transform): Set up
3637 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3638 LRA_UNKNOWN_ALT.
3639 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3640 * lra-eliminations.c (spill_pseudos): Ditto.
3641 (process_insn_for_elimination): Ditto.
3642 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3643 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3644 LRA_NON_CLOBBERED_ALT.
3646 2018-02-22 Martin Sebor <msebor@redhat.com>
3648 PR tree-optimization/84480
3649 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3650 to maybe_diag_stxncpy_trunc. Call it.
3651 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3652 from gimple_fold_builtin_strcpy. Print inlining stack.
3653 (handle_builtin_stxncpy): Print inlining stack.
3654 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3656 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3658 PR target/84176
3659 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3660 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3661 and -fcheck-pointer-bounds are used together.
3662 (indirect_thunk_prefix): New enum.
3663 (indirect_thunk_need_prefix): New function.
3664 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3665 "_nt" instead of "_bnd" for NOTRACK prefix.
3666 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3667 (output_indirect_thunk_function): Likewise.
3668 (): Likewise.
3669 (ix86_code_end): Update output_indirect_thunk_function calls.
3670 (ix86_output_indirect_branch_via_reg): Replace
3671 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3672 (ix86_output_indirect_branch_via_push): Likewise.
3673 (ix86_output_function_return): Likewise.
3674 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3675 incompatible with -fcf-protection=branch and
3676 -fcheck-pointer-bounds.
3678 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3680 PR target/83335
3681 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3682 Change gcc_assert call to output_operand_lossage.
3684 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3686 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3688 2018-02-22 DJ Delorie <dj@redhat.com>
3689 Sebastian Perta <sebastian.perta@renesas.com>
3690 Oleg Endo <olegendo@gcc.gnu.org>
3692 * config/rx/rx.c (rx_rtx_costs): New function.
3693 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3695 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3697 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3699 2018-02-22 Martin Liska <mliska@suse.cz>
3701 PR driver/83193
3702 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3703 Add "native" as a possible value.
3705 2018-02-22 Martin Liska <mliska@suse.cz>
3707 PR driver/83193
3708 * config/i386/i386.c (ix86_option_override_internal):
3709 Add "native" as a possible value for -march and -mtune.
3711 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3713 PR target/84502
3714 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3715 to all type variants.
3717 PR tree-optimization/84503
3718 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3719 width as info->bitpos + info->bitsize - start.
3720 (merged_store_group::merge_overlapping): Simplify width computation.
3721 (check_no_overlap): New function.
3722 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3723 start + width and last_order of the group, fail if check_no_overlap
3724 fails.
3725 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3726 to group if check_no_overlap fails.
3728 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3730 * config/rs6000/altivec.md: Delete contraint arguments to
3731 define_expand, define_split, and define_peephole2, and in
3732 define_insn_and_split if always unused.
3733 * config/rs6000/darwin.md: Ditto.
3734 * config/rs6000/dfp.md: Ditto.
3735 * config/rs6000/rs6000.md: Ditto.
3736 * config/rs6000/sync.md: Ditto.
3737 * config/rs6000/vector.md: Ditto.
3738 * config/rs6000/vsx.md: Ditto.
3740 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3742 * config/rs6000/altivec.md: Write output control strings as braced
3743 blocks instead of double-quoted strings.
3744 * config/rs6000/darwin.md: Ditto.
3745 * config/rs6000/rs6000.md: Ditto.
3746 * config/rs6000/vector.md: Ditto.
3747 * config/rs6000/vsx.md: Ditto.
3749 2018-02-21 Jason Merrill <jason@redhat.com>
3751 PR c++/84314 - ICE with templates and fastcall attribute.
3752 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3754 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3756 * ipa-cp.c (determine_versionability): Fix comment typos.
3758 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3760 PR c/84229
3761 * ipa-cp.c (determine_versionability): Do not version functions caling
3762 va_arg_pack.
3764 2018-02-21 Martin Liska <mliska@suse.cz>
3766 PR driver/83193
3767 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3768 Add "native" as a possible value.
3769 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3770 the macro when native cpu detection is available.
3772 2018-02-21 Martin Liska <mliska@suse.cz>
3774 PR driver/83193
3775 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3776 Add "native" as a possible value.
3777 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3778 when native cpu detection is available.
3780 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3781 Martin Sebor <msebor@redhat.com>
3783 PR tree-optimization/84478
3784 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3785 false.
3786 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3787 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3788 support which is conservatively correct, for 2 only stay conservative
3789 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3790 argument to the 2 argument get_range_strlen, adjust 6 arg
3791 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3792 false.
3793 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3794 (gimple_fold_builtin_strlen): Pass true as last argument to
3795 get_range_strlen.
3797 2018-02-20 Martin Sebor <msebor@redhat.com>
3799 PR middle-end/84095
3800 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3801 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3802 (builtin_memref::builtin_memref): Factor out parts into
3803 set_base_and_offset and call it.
3805 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3807 PR middle-end/84406
3808 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3809 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3810 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3811 search at the associated MODE_INT.
3813 2018-02-20 Jeff Law <law@redhat.com>
3815 PR middle-end/82123
3816 PR tree-optimization/81592
3817 PR middle-end/79257
3818 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3819 for range data rather than using global data.
3820 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3821 range data rather than using global data.
3822 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3823 pass it to children as needed.
3824 (struct directive::fmtresult): Similarly.
3825 (struct directive::set_width): Similarly.
3826 (struct directive::set_precision): Similarly.
3827 (format_integer, format_directive, parse_directive): Similarly.
3828 (format_none): Accept unnamed vr_values parameter.
3829 (format_percent, format_floating, format_character): Similarly.
3830 (format_string, format_plain): Similarly.
3831 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3832 the EVRP range analyzer for range data rather than using global data.
3833 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3834 gimple-ssa-evrp-analyze.h
3835 (class sprintf_dom_walker): Add after_dom_children member function.
3836 Add evrp_range_analyzer member.
3837 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3838 range analyzer as needed.
3839 (sprintf_dom_walker::after_dom_children): New member function.
3840 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3841 if not optimizing.
3842 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3843 (evrp_range_analyzer::pop_to_marker): Likewise.
3845 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3847 PR tree-optimization/84419
3848 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3849 with the required type if its current type is compatible but
3850 different.
3852 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3854 PR middle-end/82004
3855 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3856 after vectorization.
3858 2018-02-20 Martin Liska <mliska@suse.cz>
3860 PR driver/83193
3861 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3862 possible values if we don't have a hint.
3864 2018-02-20 Martin Liska <mliska@suse.cz>
3866 PR c/84310
3867 PR target/79747
3868 * final.c (shorten_branches): Build align_tab array with one
3869 more element.
3870 * opts.c (finish_options): Add alignment option limit check.
3871 (MAX_CODE_ALIGN): Likewise.
3872 (MAX_CODE_ALIGN_VALUE): Likewise.
3873 * doc/invoke.texi: Document maximum allowed option value for
3874 all -falign-* options.
3876 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3878 PR target/84146
3879 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3880 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3881 * var-tracking.c (emit_note_insn_var_location): Remove all references
3882 to NOTE_INSN_CALL_ARG_LOCATION.
3883 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3884 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3885 Use copy_rtx_if_shared.
3886 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3887 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3888 (dwarf2out_var_location): Remove handling of
3889 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3890 on call_insn.
3891 * final.c (final_scan_insn): Remove all references to
3892 NOTE_INSN_CALL_ARG_LOCATION.
3893 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3894 before dumping final insns.
3895 * except.c (emit_note_eh_region_end): Remove all references to
3896 NOTE_INSN_CALL_ARG_LOCATION.
3897 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3898 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3899 * config/arc/arc.c (hwloop_optimize): Likewise.
3900 * config/arm/arm.c (create_fix_barrier): Likewise.
3901 * config/s390/s390.c (s390_chunkify_start): Likewise.
3902 * config/sh/sh.c (find_barrier): Likewise.
3903 * config/i386/i386.c (rest_of_insert_endbranch,
3904 ix86_seh_fixup_eh_fallthru): Likewise.
3905 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3906 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3907 * config/frv/frv.c (frv_function_prologue): Likewise.
3908 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3909 reg note.
3910 (note_outside_basic_block_p): Remove all references to
3911 NOTE_INSN_CALL_ARG_LOCATION.
3912 * gengtype.c (adjust_field_rtx_def): Likewise.
3913 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3914 Likewise.
3915 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3916 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3918 PR c++/84444
3919 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3920 is ADDR_EXPR.
3922 PR tree-optimization/84452
3923 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3924 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3925 is NULL.
3927 2018-02-19 Martin Liska <mliska@suse.cz>
3929 PR sanitizer/82183
3930 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3932 2018-02-19 Martin Liska <mliska@suse.cz>
3933 Richard Sandiford <richard.sandiford@linaro.org>
3935 PR tree-optimization/82491
3936 * gimple-fold.c (get_base_constructor): Make earlier bail out
3937 to prevent ubsan.
3939 2018-02-19 Carl Love <cel@us.ibm.com>
3941 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3942 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3943 BU_P8V_OVERLOAD_1.
3944 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3945 P8V_BUILTIN_VEC_NEG.
3947 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3949 * config/rl78/rl78.md (movdf): New define expand.
3951 2018-02-19 Martin Liska <mliska@suse.cz>
3953 PR other/80589
3954 * doc/invoke.texi: Fix typo.
3955 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3957 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3959 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3960 handle rs6000_single_float and rs6000_double_float specially for
3961 e500 family CPUs.
3963 2018-02-16 Jeff Law <law@redhat.com>
3965 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3966 (pop_regs): Use it.
3968 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3970 PR ipa/84425
3971 * ipa-inline.c (inline_small_functions): Fix a typo.
3973 2018-02-16 Nathan Sidwell <nathan@acm.org>
3975 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3977 2018-02-16 Carl Love <cel@us.ibm.com>
3979 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3980 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3981 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3982 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3983 expansion to P8V_BUILTIN_VEC_FLOAT2.
3985 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3987 PR rtl-optimization/70023
3988 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3989 src_regno into account.
3991 2018-02-16 Carl Love <cel@us.ibm.com>
3993 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3994 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3995 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3996 * config/rs6000/rs6000.c: Remove case statements for
3997 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3998 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3999 and P9V_BUILTIN_VEC_VINSERT4B.
4000 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4001 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4002 * config/rs6000/vsx.md:
4003 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4004 vec_insert4b.
4006 2018-02-16 Carl Love <cel@us.ibm.com>
4008 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4009 vec_insert4b.
4010 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4011 definitions.
4012 * config/rs6000/rs6000-c.c: Add the definitions for
4013 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4014 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4015 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4016 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4017 definition for insert4b and define insn *insert3b_internal.
4018 * doc/extend.texi: Add documentation for vec_extract4b.
4020 2018-02-16 Nathan Sidwell <nathan@acm.org>
4022 * doc/extend.texi (Backwards Compatibility): Mention friend
4023 injection. Note for-scope is deprecated.
4024 * doc/invoke.texi (-ffriend-injection): Deprecate.
4026 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4028 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4029 that moved to I2, also allow destinations that are a paradoxical
4030 subreg (instead of a normal reg).
4032 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4034 PR target/83831
4035 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4036 to QImode.
4038 2018-02-16 Richard Biener <rguenther@suse.de>
4040 PR tree-optimization/84037
4041 PR tree-optimization/84016
4042 PR target/82862
4043 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4044 Adjust vec_construct for the fact we need additional higher latency
4045 128bit inserts for AVX256 and AVX512 vector builds.
4046 (ix86_add_stmt_cost): Scale vector construction cost for
4047 elementwise loads.
4049 2018-02-16 Richard Biener <rguenther@suse.de>
4051 PR tree-optimization/84417
4052 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4053 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4054 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4056 2018-02-16 Martin Liska <mliska@suse.cz>
4058 PR sanitizer/84307
4059 * internal-fn.def (ASAN_CHECK): Set proper flags.
4060 (ASAN_MARK): Likewise.
4062 2018-02-16 Julia Koval <julia.koval@intel.com>
4064 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4065 from PTA_CANNONLAKE.
4067 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4069 PR target/84272
4070 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4071 Use ++iter rather than iter++ for std::list iterators.
4072 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4073 defer deleting them until all nodes in the forest are processed. Do
4074 free even leaf nodes. Change to_process into auto_vec.
4076 PR bootstrap/84405
4077 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4078 * vec.h (vec_default_construct): Use memset instead of placement new
4079 if BROKEN_VALUE_INITIALIZATION is defined.
4080 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4081 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4082 is defined.
4084 PR rtl-optimization/83723
4085 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4086 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4087 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4088 recursive calls.
4089 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4090 callers.
4091 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4093 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4095 PR rtl-optimization/81443
4096 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4097 from inner REGs to paradoxical SUBREGs.
4099 2018-02-16 Richard Biener <rguenther@suse.de>
4101 PR tree-optimization/84399
4102 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4103 For operands we can analyze at their definition make sure we can
4104 analyze them at each use as well.
4106 2018-02-16 Richard Biener <rguenther@suse.de>
4108 PR tree-optimization/84190
4109 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4110 volatile accesses if the decl isn't volatile.
4112 2018-02-15 Jason Merrill <jason@redhat.com>
4114 PR c++/84314 - ICE with templates and fastcall attribute.
4115 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4116 TYPE_CANONICAL on an existing type.
4118 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4120 PR tree-optimization/84383
4121 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4122 dstoff nor call operand_equal_p if dstbase is NULL.
4124 PR tree-optimization/84334
4125 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4126 also a CONSTANT_CLASS_P, punt.
4128 2018-02-14 Jim Wilson <jimw@sifive.com>
4130 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4131 first SMALL_OPERAND check. New local min_second_step. Move assert
4132 to where locals are set. Add TARGET_RVC support.
4133 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4135 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4137 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4139 2018-02-14 Martin Sebor <msebor@redhat.com>
4141 PR tree-optimization/83698
4142 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4143 arrays constrain the offset range to their bounds.
4144 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4145 (builtin_access::overlap): Avoid setting the size of overlap if it's
4146 already been set.
4147 (maybe_diag_overlap): Also consider arrays when deciding what values
4148 of offsets to include in diagnostics.
4150 2018-02-14 Martin Sebor <msebor@redhat.com>
4152 PR c/84108
4153 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4154 that correspond to the kind of a declaration.
4156 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4158 PR target/83984
4159 * config/pa/pa.md: Load address of PIC label using the linkage table
4160 if the label is nonlocal.
4162 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4164 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4165 warning message if user requests -maltivec=be.
4166 * doc/invoke.texi: Document deprecation of -maltivec=be.
4168 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4170 PR target/84220
4171 * config/rs6000/rs6000-c.c: Update definitions for
4172 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4173 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4175 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4177 PR target/84239
4178 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4179 add _get_ssp intrinsics. Remove argument from
4180 __builtin_ia32_rdssp[d|q].
4181 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4182 * config/i386/i386-builtin.def: Remove argument from
4183 __builtin_ia32_rdssp[d|q].
4184 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4185 ix86_expand_special_args_builtin for _rdssp[d|q].
4186 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4187 Clear register before usage.
4188 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4189 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4191 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4193 PR tree-optimization/84357
4194 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4195 operand 1 of an ARRAY_REF too.
4197 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4199 PR target/83831
4200 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4201 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4202 declarations.
4203 (set_of_reg): New struct.
4204 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4205 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4206 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4207 functions.
4208 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4209 Split into bitclr, bitset, bitinvert patterns if appropriate.
4210 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4211 use rx_fuse_in_memory_bitop.
4212 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4213 to named insn, correct maximum insn length.
4215 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4217 PR target/79242
4218 * machmode.def: Define a complex mode for PARTIAL_INT.
4219 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4220 MODE_PARTIAL_INT.
4221 * doc/rtl.texi: Document CSPImode.
4222 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4223 handling.
4224 (msp430_hard_regno_nregs_with_padding): Likewise.
4226 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4228 PR target/84279
4229 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4231 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4233 PR rtl-optimization/84169
4234 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4235 we generated a parallel as new i3 and we split that to new i2 and i3
4236 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4237 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4238 those to i2, not i1. Partially rewrite this scan code.
4240 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4242 PR c/82210
4243 * stor-layout.c (place_field): For variable length fields, adjust
4244 offset_align afterwards not just based on the field's alignment,
4245 but also on the size.
4247 PR middle-end/84309
4248 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4249 of exps and logs in the use_exp2 case.
4251 2018-02-13 Jeff Law <law@redhat.com>
4253 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4254 entry for "vector".
4256 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4257 ARGS as unused.
4259 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4261 PR debug/84342
4262 PR debug/84319
4263 * common.opt (gas-loc-support, gas-locview-support): New.
4264 (ginline-points, ginternal-reset-location-views): New.
4265 * doc/invoke.texi: Document them. Use @itemx where intended.
4266 (gvariable-location-views): Adjust.
4267 * target.def (reset_location_view): New.
4268 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4269 (TARGET_RESET_LOCATION_VIEW): New.
4270 * doc/tm.texi: Rebuilt.
4271 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4272 (dwarf2out_default_as_locview_support): New.
4273 (output_asm_line_debug_info): Use option variables.
4274 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4275 (output_loc_list): Likewise.
4276 (add_high_low_attributes): Check option variables.
4277 Don't output entry view attribute in strict mode.
4278 (gen_inlined_subroutine_die): Check option variables.
4279 (dwarf2out_inline_entry): Likewise.
4280 (init_sections_and_labels): Likewise.
4281 (dwarf2out_early_finish): Likewise.
4282 (maybe_reset_location_view): New, from...
4283 (dwarf2out_var_location): ... here. Call it.
4284 * debug.h (dwarf2out_default_as_loc_support): Declare.
4285 (dwarf2out_default_as_locview_support): Declare.
4286 * hooks.c (hook_int_rtx_insn_0): New.
4287 * hooks.h (hook_int_rtx_insn_0): Declare.
4288 * toplev.c (process_options): Take -gas-loc-support and
4289 -gas-locview-support from dwarf2out. Enable
4290 -gvariable-location-views by default only with locview
4291 assembler support. Enable -ginternal-reset-location-views by
4292 default only if the target defines the corresponding hook.
4293 Enable -ginline-points by default if location views are
4294 enabled; force it disabled if statement frontiers are
4295 disabled.
4296 * tree-inline.c (expand_call_inline): Check option variables.
4297 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4299 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4301 PR tree-optimization/84321
4302 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4303 handling. Also check whether the anti-range contains any values
4304 that satisfy the mask; switch to a VR_RANGE if not.
4306 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4308 PR sanitizer/84340
4309 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4311 2018-02-13 Martin Jambor <mjambor@suse.cz>
4313 PR c++/83990
4314 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4315 of call statements, also set location of a load to a temporary.
4317 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4319 * config/rl78/rl78.c (add_vector_labels): New function.
4320 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4321 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4322 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4323 which checks that no arguments are passed.
4324 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4325 * doc/extend.texi: Documentation for the new attribute.
4327 2018-02-13 Andreas Schwab <schwab@suse.de>
4329 * config/riscv/linux.h (CPP_SPEC): Define.
4331 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4333 PR target/84335
4334 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4335 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4336 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4337 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4338 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4339 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4340 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4341 temporarily for AES and PCLMUL builtins.
4343 PR tree-optimization/84339
4344 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4345 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4346 Formatting fixes.
4348 PR middle-end/84309
4349 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4350 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4351 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4352 inline function.
4353 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4354 inline function.
4355 * omp-simd-clone.h: New file.
4356 * omp-simd-clone.c: Include omp-simd-clone.h.
4357 (expand_simd_clones): No longer static.
4358 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4359 cgraph.h and omp-simd-clone.h.
4360 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4361 (vect_recog_widen_shift_pattern): Formatting fix.
4362 (vect_pattern_recog_1): Don't check optab for calls.
4364 PR target/84336
4365 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4366 operands[2] into a REG before using gen_lowpart on it.
4368 2018-02-12 Jeff Law <law@redhat.com>
4370 PR target/83760
4371 * config/sh/sh.c (find_barrier): Consider a sibling call
4372 a barrier as well.
4374 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4375 successfully back substituting a reg.
4377 2018-02-12 Richard Biener <rguenther@suse.de>
4379 PR tree-optimization/84037
4380 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4381 parameter, move visited init to caller.
4382 (vect_slp_analyze_operations): Separate cost from validity
4383 check, initialize visited once for all instances.
4384 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4385 for all instances.
4386 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4387 out an assert.
4388 (vect_model_promotion_demotion_cost): Likewise.
4389 (vectorizable_bswap): Guard cost modeling with !slp_node
4390 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4391 SLP stmts.
4392 (vectorizable_call): Likewise.
4393 (vectorizable_conversion): Likewise.
4394 (vectorizable_assignment): Likewise.
4395 (vectorizable_shift): Likewise.
4396 (vectorizable_operation): Likewise.
4397 (vectorizable_store): Likewise.
4398 (vectorizable_load): Likewise.
4399 (vectorizable_condition): Likewise.
4400 (vectorizable_comparison): Likewise.
4402 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4404 PR sanitizer/84307
4405 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4406 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4407 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4409 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4411 PR middle-end/83665
4412 * params.def (inline-min-speedup): Increase from 8 to 15.
4413 (max-inline-insns-auto): Decrease from 40 to 30.
4414 * ipa-split.c (consider_split): Add some buffer for function to
4415 be considered inlining candidate.
4416 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4417 default values.
4419 2018-02-12 Richard Biener <rguenther@suse.de>
4421 PR tree-optimization/84037
4422 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4423 matched stmts if we cannot swap the non-matched ones.
4425 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4427 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4428 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4429 _mm_maskz_scalef_round_ss): New intrinsics.
4430 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4431 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4432 __builtin_ia32_scalefss_round): Remove.
4433 (__builtin_ia32_scalefsd_mask_round,
4434 __builtin_ia32_scalefss_mask_round): New intrinsics.
4435 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4436 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4437 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4438 "<round_constraint>")): Changed to ...
4439 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4440 "<round_scalar_constraint>")): ... this.
4441 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4442 %0, %1, %2<round_op3>}"): Changed to ...
4443 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4444 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4445 %2<round_scalar_mask_op3>}"): ... this.
4446 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4448 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4450 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4451 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4452 (_mm_maskz_sqrt_round_ss): New intrinsics.
4453 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4454 (__builtin_ia32_sqrtsd_mask_round)
4455 (__builtin_ia32_sqrtss_mask_round): New builtins.
4456 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4457 (__builtin_ia32_sqrtss_round): Remove.
4458 (__builtin_ia32_sqrtsd_mask_round)
4459 (__builtin_ia32_sqrtss_mask_round): New builtins.
4460 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4461 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4462 ((match_operand:VF_128 1 "vector_operand"
4463 "xBm,<round_constraint>")): Changed to ...
4464 ((match_operand:VF_128 1 "vector_operand"
4465 "xBm,<round_scalar_constraint>")): ... this.
4466 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4467 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4468 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4469 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4470 %<iptr>1<round_scalar_mask_op3>}): ... this.
4471 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4472 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4474 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4476 PR target/84266
4477 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4478 Cast vec_cmpeq result to correct type.
4479 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4480 Cast vec_cmpgt result to correct type.
4482 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4484 * final.c (final_scan_insn_1): Renamed from...
4485 (final_scan_insn): ... this. New wrapper, to recover
4486 seen from the outermost call in recursive ones.
4487 * config/sparc/sparc.c (output_return): Drop seen from call.
4488 (output_sibcall): Likewise.
4489 * config/visium/visium.c (output_branch): Likewise.
4491 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4493 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4494 function label.
4496 2018-02-10 Alan Modra <amodra@gmail.com>
4498 PR target/84300
4499 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4500 Specify LR as an input.
4502 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4504 PR sanitizer/83987
4505 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4506 remove_member_access_dummy_vars): New functions.
4507 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4508 lower_omp_1, execute_lower_omp): Use them.
4510 PR rtl-optimization/84308
4511 * shrink-wrap.c (spread_components): Release todo vector.
4513 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4515 PR rtl-optimization/57193
4516 * ira-color.c (struct allocno_color_data): Add member
4517 conflict_allocno_hard_prefs.
4518 (update_conflict_allocno_hard_prefs): New.
4519 (bucket_allocno_compare_func): Add a preference based on
4520 conflict_allocno_hard_prefs.
4521 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4522 (color_allocnos): Remove a dead code. Initiate
4523 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4525 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4527 PR target/84226
4528 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4529 constraint from =wa to wa. Avoid a subreg on the output operand,
4530 instead use a pseudo and subreg it in a move.
4531 (p9_xxbrd_<mode>): Changed to ...
4532 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4533 (p9_xxbrd_v2df): New expander.
4534 (p9_xxbrw_<mode>): Changed to ...
4535 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4536 (p9_xxbrw_v4sf): New expander.
4538 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4540 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4542 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4544 PR target/83926
4545 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4546 multiply in 32-bit mode.
4547 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4548 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4549 mode.
4551 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4553 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4554 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4555 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4556 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4558 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4560 PR lto/84213
4561 * dwarf2out.c (is_trivial_indirect_ref): New function.
4562 (dwarf2out_late_global_decl): Do not generate a location
4563 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4564 and that are not defined in the current unit.
4566 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4568 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4569 instead of a libcall for UNORDERED.
4571 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4573 PR target/82641
4574 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4575 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4577 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4579 PR target/PR84295
4580 * config/s390/s390.c (s390_set_current_function): Invoke
4581 s390_indirect_branch_settings also if fndecl didn't change.
4583 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4585 * config/rs6000/rs6000.md (blockage): Set length to zero.
4587 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4589 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4591 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4593 PR sanitizer/84285
4594 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4595 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4596 -static-lib*san.
4598 PR debug/84252
4599 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4600 PARALLEL incoming that failed vt_get_decl_and_offset check.
4602 PR middle-end/84237
4603 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4604 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4605 TREE_READONLY bit.
4606 (get_variable_section): For decls in named .bss* sections pass true as
4607 second argument to bss_initializer_p.
4609 2018-02-09 Marek Polacek <polacek@redhat.com>
4610 Jakub Jelinek <jakub@redhat.com>
4612 PR c++/83659
4613 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4614 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4615 Sync some changes from cxx_fold_indirect_ref.
4617 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4619 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4620 markers.
4621 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4622 (BLOCK_INLINE_ENTRY_LABEL): New.
4623 (dwarf2out_var_location): Disregard inline entry markers.
4624 (inline_entry_data): New struct.
4625 (inline_entry_data_hasher): New hashtable type.
4626 (inline_entry_data_hasher::hash): New.
4627 (inline_entry_data_hasher::equal): New.
4628 (inline_entry_data_table): New variable.
4629 (add_high_low_attributes): Add DW_AT_entry_pc and
4630 DW_AT_GNU_entry_view attributes if a pending entry is found
4631 in inline_entry_data_table. Add old entry_pc attribute only
4632 if debug nonbinding markers are disabled.
4633 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4634 markers are enabled.
4635 (block_within_block_p, dwarf2out_inline_entry): New.
4636 (dwarf2out_finish): Check that no entries remained in
4637 inline_entry_data_table.
4638 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4639 (final_scan_insn, notice_source_line): Likewise.
4640 (rest_of_clean_state): Skip inline entry markers.
4641 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4642 markers.
4643 * gimple.c (gimple_build_debug_inline_entry): New.
4644 * gimple.h (enum gimple_debug_subcode): Add
4645 GIMPLE_DEBUG_INLINE_ENTRY.
4646 (gimple_build_debug_inline_entry): Declare.
4647 (gimple_debug_inline_entry_p): New.
4648 (gimple_debug_nonbind_marker_p): Adjust.
4649 * insn-notes.def (INLINE_ENTRY): New.
4650 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4651 inline entry marker notes.
4652 (print_insn): Likewise.
4653 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4654 (INSN_DEBUG_MARKER_KIND): Likewise.
4655 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4656 * tree-inline.c (expand_call_inline): Build and insert
4657 debug_inline_entry stmt.
4658 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4659 inline entry blocks early, if nonbind markers are enabled.
4660 (dump_scope_block): Dump fragment info.
4661 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4662 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4663 (gimple_build_debug_inline_entry): New.
4664 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4665 Enable/disable inline entry points too.
4666 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4667 (DEBUG_INSN): Describe inline entry markers.
4669 * common.opt (gvariable-location-views): New.
4670 (gvariable-location-views=incompat5): New.
4671 * config.in: Rebuilt.
4672 * configure: Rebuilt.
4673 * configure.ac: Test assembler for view support.
4674 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4675 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4676 * dwarf2out.c (var_loc_view): New typedef.
4677 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4678 (dwarf2out_locviews_in_attribute): New.
4679 (dwarf2out_locviews_in_loclist): New.
4680 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4681 (enum dw_line_info_opcode): Add LI_adv_address.
4682 (struct dw_line_info_table): Add view.
4683 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4684 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4685 (zero_view_p): New variable.
4686 (ZERO_VIEW_P): New macro.
4687 (output_asm_line_debug_info): New.
4688 (struct var_loc_node): Add view.
4689 (add_AT_view_list, AT_loc_list): New.
4690 (add_var_loc_to_decl): Add view param. Test it against last.
4691 (new_loc_list): Add view params. Record them.
4692 (AT_loc_list_ptr): Handle loc and view lists.
4693 (view_list_to_loc_list_val_node): New.
4694 (print_dw_val): Handle dw_val_class_view_list.
4695 (size_of_die): Likewise.
4696 (value_format): Likewise.
4697 (loc_list_has_views): New.
4698 (gen_llsym): Set vl_symbol too.
4699 (maybe_gen_llsym, skip_loc_list_entry): New.
4700 (dwarf2out_maybe_output_loclist_view_pair): New.
4701 (output_loc_list): Output view list or entries too.
4702 (output_view_list_offset): New.
4703 (output_die): Handle dw_val_class_view_list.
4704 (output_dwarf_version): New.
4705 (output_compilation_unit_header): Use it.
4706 (output_skeleton_debug_sections): Likewise.
4707 (output_rnglists, output_line_info): Likewise.
4708 (output_pubnames, output_aranges): Update version comments.
4709 (output_one_line_info_table): Output view numbers in asm comments.
4710 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4711 Call maybe_gen_llsym.
4712 (loc_list_from_tree_1): Adjust.
4713 (add_AT_location_description): Create view list attribute if
4714 needed, check it's absent otherwise.
4715 (convert_cfa_to_fb_loc_list): Adjust.
4716 (maybe_emit_file): Call output_asm_line_debug_info for test.
4717 (dwarf2out_var_location): Reset views as needed. Precompute
4718 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4719 attribute. Set view.
4720 (new_line_info_table): Reset next view.
4721 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4722 (dwarf2out_source_line): Likewise. Output view resets and labels to
4723 the assembler, or select appropriate line info opcodes.
4724 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4725 (optimize_string_length): Catch it. Adjust.
4726 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4727 dw_val_class_view_list, and remove it if no longer needed.
4728 (hash_loc_list): Hash view numbers.
4729 (loc_list_hasher::equal): Compare them.
4730 (optimize_location_lists): Check whether a view list symbol is
4731 needed, and whether the locview attribute is present, and
4732 whether they match. Remove the locview attribute if no longer
4733 needed.
4734 (index_location_lists): Call skip_loc_list_entry for test.
4735 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4736 Use output_dwarf_version.
4737 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4738 (struct dw_val_node): Add val_view_list.
4739 * final.c (SEEN_NEXT_VIEW): New.
4740 (set_next_view_needed): New.
4741 (clear_next_view_needed): New.
4742 (maybe_output_next_view): New.
4743 (final_start_function): Rename to...
4744 (final_start_function_1): ... this. Take pointer to FIRST,
4745 add SEEN parameter. Emit param bindings in the initial view.
4746 (final_start_function): Reintroduce SEEN-less interface.
4747 (final): Rename to...
4748 (final_1): ... this. Take SEEN parameter. Output final pending
4749 next view at the end.
4750 (final): Reintroduce seen-less interface.
4751 (final_scan_insn): Output pending next view before switching
4752 sections or ending a block. Mark the next view as needed when
4753 outputting variable locations. Notify debug backend of section
4754 changes, and of location view changes.
4755 (rest_of_handle_final): Adjust.
4756 * toplev.c (process_options): Autodetect value for debug variable
4757 location views option. Warn on incompat5 without -gdwarf-5.
4758 * doc/invoke.texi (gvariable-location-views): New.
4759 (gvariable-location-views=incompat5): New.
4760 (gno-variable-location-views): New.
4762 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4764 PR tree-optimization/84136
4765 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4766 that the result of find_edge is non-NULL.
4768 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4770 PR target/83008
4771 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4772 storing integer register in SImode. Fix cost of 256 and 512
4773 byte aligned SSE register store.
4775 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4777 * config/i386/i386.c (ix86_multiplication_cost): Fix
4778 multiplication cost for TARGET_AVX512DQ.
4780 2018-02-08 Marek Polacek <polacek@redhat.com>
4782 PR tree-optimization/84238
4783 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4784 get_range_strlen.
4786 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4788 PR tree-optimization/84265
4789 * tree-vect-stmts.c (vectorizable_store): Don't treat
4790 VMAT_CONTIGUOUS accesses as grouped.
4791 (vectorizable_load): Likewise.
4793 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4795 PR tree-optimization/81635
4796 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4797 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4798 (test_round_for_mask): New functions.
4799 (wide_int_cc_tests): Call test_round_for_mask.
4800 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4801 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4802 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4803 range returned by get_range_info.
4805 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4807 PR ipa/81360
4808 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4809 * symtab.c: Include builtins.h
4810 (symtab_node::output_to_lto_symbol_table_p): Move here
4811 from lto-streamer-out.c:output_symbol_p.
4812 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4813 (output_symbol_p): Move all logic to symtab.c
4814 (produce_symtab): Update.
4816 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4818 * config/s390/s390-opts.h (enum indirect_branch): Define.
4819 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4820 (s390_indirect_branch_via_thunk)
4821 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4822 (enum s390_indirect_branch_type): Define.
4823 * config/s390/s390.c (struct s390_frame_layout, struct
4824 machine_function): Remove.
4825 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4826 (indirect_branch_table_label_no, indirect_branch_table_name):
4827 Define variables.
4828 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4829 (enum s390_indirect_branch_option): Define.
4830 (s390_return_addr_from_memory): New function.
4831 (s390_handle_string_attribute): New function.
4832 (s390_attribute_table): Add new attribute handler.
4833 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4834 (s390_indirect_branch_via_thunk): New function.
4835 (s390_indirect_branch_via_inline_thunk): New function.
4836 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4837 sibling call optimization for non z10 compiles.
4838 (s390_emit_call): Force indirect branch target to be a single
4839 register. Add r1 clobber for non-z10 compiles.
4840 (s390_emit_epilogue): Emit return jump via return_use expander.
4841 (s390_reorg): Handle JUMP_INSNs as execute targets.
4842 (s390_option_override_internal): Perform validity checks for the
4843 new command line options.
4844 (s390_indirect_branch_attrvalue): New function.
4845 (s390_indirect_branch_settings): New function.
4846 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4847 (s390_output_indirect_thunk_function): New function.
4848 (s390_code_end): Implement target hook.
4849 (s390_case_values_threshold): Implement target hook.
4850 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4851 macros.
4852 * config/s390/s390.h (struct s390_frame_layout)
4853 (struct machine_function): Move here from s390.c.
4854 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4855 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4856 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4857 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4858 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4859 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4860 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4861 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4862 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4863 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4864 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4865 (mnemonic attribute): Add values which aren't recognized
4866 automatically.
4867 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4868 pattern for branch conversion. Fix mnemonic attribute.
4869 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4870 indirect branch via thunk if requested.
4871 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4872 ("*indirect_jump"): Disable for branch conversion using out of
4873 line thunks.
4874 ("indirect_jump_via_thunk<mode>_z10")
4875 ("indirect_jump_via_thunk<mode>")
4876 ("indirect_jump_via_inlinethunk<mode>_z10")
4877 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4878 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4879 ("casesi_jump_via_inlinethunk<mode>_z10")
4880 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4881 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4882 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4883 ("*indirect2_jump"): Disable for branch conversion.
4884 ("casesi_jump"): Turn into expander and expand patterns for branch
4885 conversion.
4886 ("return_use"): New expander.
4887 ("*return"): Emit return via thunk and rename it to ...
4888 ("*return<mode>"): ... this one.
4889 * config/s390/s390.opt: Add new options and and enum for the
4890 option values.
4892 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4894 * lra-constraints.c (match_reload): Unconditionally use
4895 gen_lowpart_SUBREG, rather than selecting between that
4896 and equivalent gen_rtx_SUBREG code.
4898 2018-02-08 Richard Biener <rguenther@suse.de>
4900 PR tree-optimization/84233
4901 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4902 changed flag instead of boguously re-using phi_inserted.
4904 2018-02-08 Martin Jambor <mjambor@suse.cz>
4906 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4907 static local variables.
4909 2018-02-08 Richard Biener <rguenther@suse.de>
4911 PR tree-optimization/84278
4912 * tree-vect-stmts.c (vectorizable_store): When looking for
4913 smaller vector types to perform grouped strided loads/stores
4914 make sure the mode is supported by the target.
4915 (vectorizable_load): Likewise.
4917 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4919 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4920 Increase LDP/STP opportunities by adding adjacent callee-saves.
4922 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4924 PR rtl-optimization/84068
4925 PR rtl-optimization/83459
4926 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4928 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4930 PR tree-optimization/84224
4931 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4932 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4933 non-zero arguments.
4935 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4937 PR target/84113
4938 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4939 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4940 count, remove one USE.
4942 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4944 * doc/install.texi (Configuration): Document the
4945 --with-long-double-format={ibm,ieee} PowerPC configuration
4946 options.
4948 PR target/84154
4949 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4950 Convert from define_expand to be define_insn_and_split. Rework
4951 float/double/_Float128 conversions to QI/HI/SImode to work with
4952 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4953 conversions to QI/HImode types did a store and then a load to
4954 truncate the value. For conversions to VSX registers, don't split
4955 the insn, instead emit the code directly. Use the code iterator
4956 any_fix to combine signed and unsigned conversions.
4957 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4958 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4959 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4960 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4961 (fix_<mode>di2_hw): Likewise.
4962 (fixuns_<mode>di2_hw): Likewise.
4963 (fix_<mode>si2_hw): Likewise.
4964 (fixuns_<mode>si2_hw): Likewise.
4965 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4966 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4967 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4968 fix<uns>_trunc<SFDF:mode>si2_p8.
4969 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4970 used.
4971 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4972 (fix<uns>_<mode>_mem): Likewise.
4973 (fctiw<u>z_<mode>_mem): Likewise.
4974 (fix<uns>_<mode>_mem): Likewise.
4975 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4976 the register allocator from doing a direct move to the GPRs to do
4977 a store, and instead use the ISA 3.0 store byte/half-word from
4978 vector register instruction. For IEEE 128-bit floating point,
4979 also optimize stores of 32-bit ints.
4980 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4982 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4984 * genextract.c (push_pathstr_operand): New function to support
4985 [a-zA-Z].
4986 (walk_rtx): Call push_pathstr_operand.
4987 (print_path): Support [a-zA-Z].
4989 2018-02-07 Richard Biener <rguenther@suse.de>
4991 PR tree-optimization/84037
4992 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4993 (cse_and_gimplify_to_preheader): Declare.
4994 (vect_get_place_in_interleaving_chain): Likewise.
4995 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4996 ivexpr_map.
4997 (_loop_vec_info::~_loop_vec_info): Delete it.
4998 (cse_and_gimplify_to_preheader): New function.
4999 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5000 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5001 (vectorizable_load): Likewise. For grouped stores always base
5002 the IV on the first element.
5003 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5004 condition before gimplifying.
5006 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5008 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5009 *DIV_EXPR and *MOD_EXPR.
5011 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5013 PR target/84248
5014 * config/i386/i386.c (ix86_option_override_internal): Mask out
5015 the CF_SET bit when checking -fcf-protection.
5017 2018-02-07 Tom de Vries <tom@codesourcery.com>
5019 PR libgomp/84217
5020 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5021 enough.
5023 2018-02-07 Richard Biener <rguenther@suse.de>
5025 PR tree-optimization/84204
5026 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5027 this place.
5029 PR tree-optimization/84205
5030 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5031 special-case isl_ast_op_zdiv_r.
5033 PR tree-optimization/84223
5034 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5035 Only add conditions from within the region.
5036 (gather_bbs::after_dom_children): Adjust.
5038 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5040 PR target/84209
5041 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5042 * config/avr/avr.md: Only post-reload split REG-REG moves if
5043 either register is GENERAL_REG_P.
5045 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5047 PR tree-optimization/84235
5048 * tree-ssa-scopedtables.c
5049 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5050 if the subtraction is performed in floating point type where NaNs are
5051 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5052 build 1. Formatting fix.
5054 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5056 PR target/84146
5057 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5058 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5059 and skip it regardless of bb boundaries. Use CALL_P macro,
5060 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5061 unnecessarily, formatting fix.
5063 2018-02-06 Michael Collison <michael.collison@arm.com>
5065 * config/arm/thumb2.md:
5066 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5067 (*thumb_mov_notscc): Ditto.
5069 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5071 PR target/84154
5072 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5073 unsigned_fix, not "s".
5075 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5077 * configure.ac (gcc_fn_eh_frame_ro): New function.
5078 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5079 correct .eh_frame permissions.
5080 * configure: Regenerate.
5082 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5084 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5085 irrelevant options.
5087 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5089 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5090 Display warning message for -mno-speculate-indirect-jumps.
5092 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5094 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5095 Undocumented.
5096 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5098 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5100 PR tree-optimization/84225
5101 * tree-eh.c (find_trapping_overflow): Only call
5102 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5104 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5106 PR target/84145
5107 * config/i386/i386.c: Reimplement the check of possible options
5108 -mibt/-mshstk conbination. Change error messages.
5109 * doc/invoke.texi: Fix a typo: remove extra '='.
5111 2018-02-06 Marek Polacek <polacek@redhat.com>
5113 PR tree-optimization/84228
5114 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5116 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5118 PR target/82641
5119 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5120 emitted arch directives.
5121 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5122 __ARM_FEATURE_COPROC before changing architectures.
5124 2018-02-06 Richard Biener <rguenther@suse.de>
5126 * config/i386/i386.c (print_reg): Fix typo.
5127 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5129 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5131 * configure: Regenerate.
5133 2018-02-05 Martin Sebor <msebor@redhat.com>
5135 PR tree-optimization/83369
5136 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5137 inlining context.
5139 2018-02-05 Martin Liska <mliska@suse.cz>
5141 * doc/invoke.texi: Cherry-pick upstream r323995.
5143 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5145 * ira.c (ira_init_register_move_cost): Adjust comment.
5147 2018-02-05 Martin Liska <mliska@suse.cz>
5149 PR gcov-profile/84137
5150 * doc/gcov.texi: Fix typo in documentation.
5152 2018-02-05 Martin Liska <mliska@suse.cz>
5154 PR gcov-profile/83879
5155 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5156 using dlopen functionality.
5158 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5160 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5161 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5162 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5163 _mm_maskz_range_round_ss): New intrinsics.
5164 (__builtin_ia32_rangesd128_round)
5165 (__builtin_ia32_rangess128_round): Remove.
5166 (__builtin_ia32_rangesd128_mask_round,
5167 __builtin_ia32_rangess128_mask_round): New builtins.
5168 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5169 __builtin_ia32_rangess128_round): Remove.
5170 (__builtin_ia32_rangesd128_mask_round,
5171 __builtin_ia32_rangess128_mask_round): New builtins.
5172 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5173 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5174 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5175 "<round_saeonly_constraint>")): Changed to ...
5176 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5177 "<round_saeonly_scalar_constraint>")): ... this.
5178 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5179 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5180 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5181 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5182 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5184 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5186 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5187 options.
5188 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5189 Remove all values except native, 8540 and 8548.
5191 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5193 * config/i386/i386.c (ix86_output_function_return): Pass
5194 INVALID_REGNUM, instead of -1, as invalid register number to
5195 indirect_thunk_name and output_indirect_thunk.
5197 2018-02-02 Julia Koval <julia.koval@intel.com>
5199 * config.gcc: Add -march=icelake.
5200 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5201 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5202 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5203 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5204 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5205 (processor_target_table): Add icelake.
5206 (ix86_option_override_internal): Handle new PTAs.
5207 (get_builtin_code_for_version): Handle icelake.
5208 (M_INTEL_COREI7_ICELAKE): New.
5209 (fold_builtin_cpu): Handle icelake.
5210 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5211 * doc/invoke.texi: Add -march=icelake.
5213 2018-02-02 Julia Koval <julia.koval@intel.com>
5215 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5216 to wide_int_bitmask.
5217 * wide-int-bitmask.h: New.
5219 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5221 PR target/84066
5222 * config/i386/i386.md: Replace Pmode with word_mode in
5223 builtin_setjmp_setup and builtin_longjmp to support x32.
5225 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5227 PR target/56010
5228 PR target/83743
5229 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5230 #include "opts.h".
5231 (rs6000_supported_cpu_names): New static variable.
5232 (linux_cpu_translation_table): Likewise.
5233 (elf_platform) <cpu>: Define new static variable and use it.
5234 Translate kernel AT_PLATFORM name to canonical name if needed.
5235 Error if platform name is unknown.
5237 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5239 PR target/84089
5240 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5242 2018-02-01 Jeff Law <law@redhat.com>
5244 PR target/84128
5245 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5246 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5247 the scratch if RELEASE_VIA_POP is false.
5248 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5249 If we have to save a temporary register, decrement SIZE appropriately.
5250 Pass new arguments to release_scratch_register_on_entry.
5251 (ix86_adjust_stack_and_probe): Likewise.
5252 (ix86_emit_probe_stack_range): Pass new arguments to
5253 release_scratch_register_on_entry.
5255 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5257 PR rtl-optimization/84157
5258 * combine.c (change_zero_ext): Use REG_P predicate in
5259 front of HARD_REGISTER_P predicate.
5261 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5263 * config/avr/avr.c (avr_option_override): Move disabling of
5264 -fdelete-null-pointer-checks to...
5265 * common/config/avr/avr-common.c (avr_option_optimization_table):
5266 ...here.
5268 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5270 PR tree-optimization/81635
5271 * tree-data-ref.c (split_constant_offset_1): For types that
5272 wrap on overflow, try to use range info to prove that wrapping
5273 cannot occur.
5275 2018-02-01 Renlin Li <renlin.li@arm.com>
5277 PR target/83370
5278 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5279 TAILCALL_ADDR_REGS.
5280 (aarch64_register_move_cost): Likewise.
5281 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5282 TAILCALL_ADDR_REGS.
5283 (REG_CLASS_NAMES): Likewise.
5284 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5285 TAILCALL_ADDR_REGS. Remove IP registers.
5286 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5288 2018-02-01 Richard Biener <rguenther@suse.de>
5290 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5291 for specifying RPO order and allow NULL for that.
5292 * domwalk.c (dom_walker::dom_walker): Likewise.
5293 (dom_walker::walk): Handle NULL RPO order.
5294 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5295 in RPO order.
5296 (rewrite_update_dom_walker): Likewise.
5297 (mark_def_dom_walker): Likewise.
5299 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5301 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5302 (aarch64_maybe_expand_sve_subreg_move): Declare.
5303 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5304 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5305 predicate.
5306 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5307 that are semantically a reverse operation.
5308 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5309 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5310 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5311 functions.
5312 (aarch64_can_change_mode_class): For big-endian, forbid changes
5313 between two SVE modes if they have different element sizes.
5315 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5317 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5318 the TImode handling for big-endian targets.
5320 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5322 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5323 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5324 not just bytes.
5325 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5326 Remove BSWAP handing for big-endian targets and use the form of
5327 LD1RQ appropariate for the mode.
5329 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5331 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5332 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5333 duplicated element.
5335 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5337 PR tearget/83845
5338 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5339 check for operands that need to go through aarch64_sve_reload_be.
5341 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5343 PR tree-optimization/81661
5344 PR tree-optimization/84117
5345 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5346 * tree-eh.c: Include gimplify.h.
5347 (find_trapping_overflow, replace_trapping_overflow,
5348 rewrite_to_non_trapping_overflow): New functions.
5349 * tree-vect-loop.c: Include tree-eh.h.
5350 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5351 * tree-data-ref.c: Include tree-eh.h.
5352 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5354 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5356 PR rtl-optimization/84123
5357 * combine.c (change_zero_ext): Check if hard register satisfies
5358 can_change_dest_mode before calling gen_lowpart_SUBREG.
5360 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5362 PR target/82444
5363 * ira.c (ira_init_register_move_cost): Remove assert.
5365 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5367 PR rtl-optimization/84071
5368 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5369 * doc/tm.texi: Regenerate.
5371 2018-01-31 Richard Biener <rguenther@suse.de>
5373 PR tree-optimization/84132
5374 * tree-data-ref.c (analyze_miv_subscript): Properly
5375 check whether evolution_function_is_affine_multivariate_p
5376 before calling gcd_of_steps_may_divide_p.
5378 2018-01-31 Julia Koval <julia.koval@intel.com>
5380 PR target/83618
5381 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5382 * config/i386/i386.md (rdpid_rex64) New.
5383 (rdpid): Make 32bit only.
5385 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5387 PR lto/84105
5388 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5389 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5391 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5393 Revert
5394 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5396 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5398 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5400 PR rtl-optimization/84071
5401 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5402 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5404 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5406 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5407 (arc_attribute_table): Add 'aux' attribute.
5408 (arc_in_small_data_p): Consider aux like variables.
5409 (arc_is_aux_reg_p): New function.
5410 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5411 (arc_get_aux_arg): New function.
5412 (prepare_move_operands): Handle aux-register access.
5413 (arc_handle_aux_attribute): New function.
5414 * doc/extend.texi (ARC Variable attributes): Add subsection.
5416 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5418 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5419 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5420 (arc_attribute_table): Add 'uncached' attribute.
5421 (arc_print_operand): Print '.di' flag for uncached memory
5422 accesses.
5423 (arc_in_small_data_p): Do not consider for small data the uncached
5424 types.
5425 (arc_is_uncached_mem_p): New function.
5426 * config/arc/predicates.md (compact_store_memory_operand): Check
5427 for uncached memory accesses.
5428 (nonvol_nonimm_operand): Likewise.
5429 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5431 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5433 PR c/84100
5434 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5435 falign-loops=): Add Optimization flag.
5437 2018-01-30 Jeff Law <law@redhat.com>
5439 PR target/84064
5440 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5441 INT_REGISTERS_SAVED. Check it prior to calling
5442 get_scratch_register_on_entry.
5443 (ix86_adjust_stack_and_probe): Similarly.
5444 (ix86_emit_probe_stack_range): Similarly.
5445 (ix86_expand_prologue): Corresponding changes.
5447 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5449 PR target/40411
5450 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5451 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5453 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5455 PR target/84112
5456 * lra-constraints.c (curr_insn_transform): Process AND in the
5457 address.
5459 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5461 PR rtl-optimization/83986
5462 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5463 dependence against last_pending_memory_flush in addition to
5464 pending_jump_insns.
5466 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5468 PR tree-optimization/81611
5469 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5470 copies.
5472 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5474 PR target/83758
5475 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5476 a reg rtx.
5478 2018-01-30 Richard Biener <rguenther@suse.de>
5479 Jakub Jelinek <jakub@redhat.com>
5481 PR tree-optimization/84111
5482 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5483 inner loops added during recursion, as they don't have up-to-date
5484 SSA form.
5486 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5488 PR ipa/81360
5489 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5490 (can_inline_edge_by_limits_p): ... here.
5491 (can_early_inline_edge_p, check_callers,
5492 update_caller_keys, update_callee_keys, recursive_inlining,
5493 add_new_edges_to_heap, speculation_useful_p,
5494 inline_small_functions,
5495 inline_small_functions, flatten_function,
5496 inline_to_all_callers_1): Update.
5498 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5500 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5501 zeros correctly.
5503 2018-01-30 Richard Biener <rguenther@suse.de>
5505 PR tree-optimization/83008
5506 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5507 invariant and constant vector uses in stmts when they need
5508 more than one stmt.
5510 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5512 PR bootstrap/84017
5513 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5514 * configure: Regenerate.
5516 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5518 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5519 pattern.
5520 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5521 Use gen_rtx_REG rather than gen_lowpart.
5523 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5525 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5526 rather than 0 when creating partial subregs.
5528 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5530 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5531 of usage.
5533 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5535 PR target/81550
5536 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5537 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5538 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5539 flags. This restores the settings used before the 2017-07-24.
5540 Turning off pre increment/decrement/modify allows IVOPTS to
5541 optimize DF/SF loops where the index is an int.
5543 2018-01-29 Richard Biener <rguenther@suse.de>
5544 Kelvin Nilsen <kelvin@gcc.gnu.org>
5546 PR bootstrap/80867
5547 * tree-vect-stmts.c (vectorizable_call): Don't call
5548 targetm.vectorize_builtin_md_vectorized_function if callee is
5549 NULL.
5551 2018-01-22 Carl Love <cel@us.ibm.com>
5553 * doc/extend.tex: Fix typo in second arg in
5554 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5556 2018-01-29 Richard Biener <rguenther@suse.de>
5558 PR tree-optimization/84086
5559 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5560 (flush_ssaname_freelist): When SSA names were released reset
5561 the SCEV hash table.
5563 2018-01-29 Richard Biener <rguenther@suse.de>
5565 PR tree-optimization/84057
5566 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5567 removed paths when removing edges.
5569 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5571 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5572 -mfunction-return=@var{choice}.
5574 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5576 PR diagnostic/84034
5577 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5578 Handle CR like TAB.
5579 (layout::print_source_line): Likewise.
5580 (test_get_line_width_without_trailing_whitespace): Add test cases.
5582 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5584 PR middle-end/84040
5585 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5586 debug insns.
5588 2018-01-26 Jim Wilson <jimw@sifive.com>
5590 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5592 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5593 specified.
5595 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5597 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5598 and CMP + SUB-immediate -> SUBS.
5600 2018-01-26 Martin Sebor <msebor@redhat.com>
5602 PR tree-optimization/83896
5603 * tree-ssa-strlen.c (get_string_len): Rename...
5604 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5605 Avoid assuming length is constant.
5606 (handle_char_store): Use HOST_WIDE_INT for string length.
5608 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5610 PR target/81763
5611 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5612 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5614 2018-01-26 Richard Biener <rguenther@suse.de>
5616 PR rtl-optimization/84003
5617 * dse.c (record_store): Only record redundant stores when
5618 the earlier store aliases at least all accesses the later one does.
5620 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5622 PR rtl-optimization/83985
5623 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5624 REG_CFA_RESTORE insns.
5625 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5626 REG_CFA_RESTORE insns here.
5628 PR c/83989
5629 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5630 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5632 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5634 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5635 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5636 (arc_init): Likewise.
5637 (arc_override_options): Likewise.
5638 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5639 value.
5640 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5641 support.
5642 * config/arc/arc.h (TARGET_DBNZ): Define.
5643 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5644 properly set the tune attribute.
5645 (dbnz): Use TARGET_DBNZ guard.
5646 * config/arc/arc.opt (mtune): Add core3 option.
5648 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5650 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5651 recognize new pic like addresses.
5652 (arc_delegitimize_address): Clean up.
5654 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5656 * config/arc/arc-arches.def: Option mrf16 valid for all
5657 architectures.
5658 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5659 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5660 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5661 * config/arc/arc-tables.opt: Regenerate.
5662 * config/arc/arc.c (arc_conditional_register_usage): Handle
5663 reduced register file case.
5664 (arc_file_start): Set must have build attributes.
5665 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5666 mrf16 option value.
5667 * config/arc/arc.opt (mrf16): Add new option.
5668 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5669 * config/arc/genmultilib.awk: Handle new mrf16 option.
5670 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5671 * config/arc/t-multilib: Regenerate.
5672 * doc/invoke.texi (ARC Options): Document mrf16 option.
5674 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5676 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5677 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5678 (arc_attribute_table): Add 'secure_call' attribute.
5679 (arc_print_operand): Print secure call operand.
5680 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5681 secure.
5682 (arc_is_secure_call_p): New function. * config/arc/arc.md
5683 (call_i): Add support for sjli instruction.
5684 (call_value_i): Likewise.
5685 * config/arc/constraints.md (Csc): New constraint.
5687 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5688 John Eric Martin <John.Martin@emmicro-us.com>
5690 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5691 * config/arc/arc.c (_arc_jli_section): New struct.
5692 (arc_jli_section): New type.
5693 (rc_jli_sections): New static variable.
5694 (arc_handle_jli_attribute): New function.
5695 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5696 (arc_file_end): New function.
5697 (TARGET_ASM_FILE_END): Define.
5698 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5699 (arc_add_jli_section): New function.
5700 (jli_call_scan): Likewise.
5701 (arc_reorg): Call jli_call_scan.
5702 (arc_output_addsi): Remove 'S' from printing asm operand.
5703 (arc_is_jli_call_p): New function.
5704 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5705 operand.
5706 (movhi_insn): Likewise.
5707 (movsi_insn): Likewise.
5708 (movsi_set_cc_insn): Likewise.
5709 (loadqi_update): Likewise.
5710 (load_zeroextendqisi_update): Likewise.
5711 (load_signextendqisi_update): Likewise.
5712 (loadhi_update): Likewise.
5713 (load_zeroextendhisi_update): Likewise.
5714 (load_signextendhisi_update): Likewise.
5715 (loadsi_update): Likewise.
5716 (loadsf_update): Likewise.
5717 (movsicc_insn): Likewise.
5718 (bset_insn): Likewise.
5719 (bxor_insn): Likewise.
5720 (bclr_insn): Likewise.
5721 (bmsk_insn): Likewise.
5722 (bicsi3_insn): Likewise.
5723 (cmpsi_cc_c_insn): Likewise.
5724 (movsi_ne): Likewise.
5725 (movsi_cond_exec): Likewise.
5726 (clrsbsi2): Likewise.
5727 (norm_f): Likewise.
5728 (normw): Likewise.
5729 (swap): Likewise.
5730 (divaw): Likewise.
5731 (flag): Likewise.
5732 (sr): Likewise.
5733 (kflag): Likewise.
5734 (ffs): Likewise.
5735 (ffs_f): Likewise.
5736 (fls): Likewise.
5737 (call_i): Remove 'S' asm letter, add jli instruction.
5738 (call_value_i): Likewise.
5739 * config/arc/arc.op (mjli-always): New option.
5740 * config/arc/constraints.md (Cji): New constraint.
5741 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5742 operand.
5743 (subsf3_fpx): Likewise.
5744 (mulsf3_fpx): Likewise.
5745 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5746 asm operand.
5747 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5748 function attrbutes.
5749 * doc/invoke.texi (ARC): Document mjli-always option.
5751 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5753 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5754 avoid addition with 0 and use incw and decw where possible.
5756 2018-01-26 Richard Biener <rguenther@suse.de>
5758 PR tree-optimization/81082
5759 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5760 association if it requires casting to unsigned.
5761 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5762 from fold_plusminus_mult_expr to catch important cases late when
5763 range info is available.
5765 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5767 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5768 * configure.ac (hidden_linkonce): New test.
5769 * configure: Regenerate.
5770 * config.in: Regenerate.
5772 2018-01-26 Julia Koval <julia.koval@intel.com>
5774 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5775 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5776 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5777 _mm_mask_bitshuffle_epi64_mask): Fix type.
5778 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5779 USI_FTYPE_V4DI_V4DI_USI): Remove.
5780 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5781 __builtin_ia32_vpshufbitqmb256_mask,
5782 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5783 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5784 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5786 2018-01-26 Alan Modra <amodra@gmail.com>
5788 PR target/84033
5789 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5790 UNSPEC_VBPERMQ. Sort other unspecs.
5792 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5794 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5796 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5798 PR middle-end/83055
5799 * predict.c (drop_profile): Do not push/pop cfun; update also
5800 node->count.
5801 (handle_missing_profiles): Fix logic looking for zero profiles.
5803 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5805 PR middle-end/83977
5806 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5807 on functions with #pragma omp declare simd or functions with simd
5808 attribute.
5809 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5810 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5811 Remove trailing \n from warning_at calls.
5813 2018-01-25 Tom de Vries <tom@codesourcery.com>
5815 PR target/84028
5816 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5817 for neutered workers.
5819 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5821 PR target/68467
5822 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5823 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5825 2018-01-24 Jeff Law <law@redhat.com>
5827 PR target/83994
5828 * i386.c (get_probe_interval): Move to earlier point.
5829 (ix86_compute_frame_layout): If -fstack-clash-protection and
5830 the frame is larger than the probe interval, then use pushes
5831 to save registers rather than reg->mem moves.
5832 (ix86_expand_prologue): Remove conditional for int_registers_saved
5833 assertion.
5835 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5837 PR target/84014
5838 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5839 min/max for never referenced object.
5841 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5843 PR middle-end/83977
5844 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5845 here.
5846 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5847 attributes from DECL_ATTRIBUTES (decl) without affecting
5848 DECL_ATTRIBUTES (current_function_decl).
5849 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5850 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5852 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5854 PR tree-optimization/83979
5855 * fold-const.c (fold_comparison): Use constant_boolean_node
5856 instead of boolean_{true,false}_node.
5858 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5860 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5861 with zero counts.
5863 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5865 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5866 Simplify the clause that sets the length attribute.
5867 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5868 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5869 clause that sets the length attribute.
5870 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5872 2018-01-24 Tom de Vries <tom@codesourcery.com>
5874 PR target/83589
5875 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5876 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5877 Add strict parameter.
5878 (prevent_branch_around_nothing): Insert dummy insn between branch to
5879 label and label with no ptx insn inbetween.
5880 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5882 2018-01-24 Tom de Vries <tom@codesourcery.com>
5884 PR target/81352
5885 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5886 for neutered threads in warp.
5887 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5889 2018-01-24 Richard Biener <rguenther@suse.de>
5891 PR tree-optimization/83176
5892 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5893 operands.
5895 2018-01-24 Richard Biener <rguenther@suse.de>
5897 PR tree-optimization/82819
5898 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5899 code generating pluses that are no-ops in the target precision.
5901 2018-01-24 Richard Biener <rguenther@suse.de>
5903 PR middle-end/84000
5904 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5906 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5908 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5909 to merge probabilities.
5910 * predict.c (probably_never_executed): Also mark as cold functions
5911 with global 0 profile and guessed local profile.
5912 * profile-count.c (profile_probability::combine_with_count): New
5913 member function.
5914 * profile-count.h (profile_probability::operator*,
5915 profile_probability::operator*=, profile_probability::operator/,
5916 profile_probability::operator/=): Reduce precision to adjusted
5917 and set value to guessed on contradictory divisions.
5918 (profile_probability::combine_with_freq): Remove.
5919 (profile_probability::combine_wiht_count): Declare.
5920 (profile_count::force_nonzero):: Set to adjusted.
5921 (profile_count::probability_in):: Set quality to adjusted.
5922 * tree-ssa-tail-merge.c (replace_block_by): Use
5923 combine_with_count.
5925 2018-01-23 Andrew Waterman <andrew@sifive.com>
5926 Jim Wilson <jimw@sifive.com>
5928 * config/riscv/riscv.c (riscv_stack_boundary): New.
5929 (riscv_option_override): Set riscv_stack_boundary. Handle
5930 riscv_preferred_stack_boundary_arg.
5931 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5932 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5933 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5934 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5935 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5936 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5938 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5940 PR target/83905
5941 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5942 of struct ix86_frame.
5943 (ix86_expand_epilogue): Likewise. Add a local variable for
5944 the reg_save_offset field in struct ix86_frame.
5946 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5948 PR tree-optimization/82604
5949 * tree-loop-distribution.c (enum partition_kind): New enum item
5950 PKIND_PARTIAL_MEMSET.
5951 (partition_builtin_p): Support above new enum item.
5952 (generate_code_for_partition): Ditto.
5953 (compute_access_range): Differentiate cases that equality can be
5954 proven at all loops, the innermost loops or no loops.
5955 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5956 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5957 (finalize_partitions, distribute_loop): Don't fuse partition of
5958 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5959 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5960 parloop is enabled.
5962 2018-01-23 Martin Liska <mliska@suse.cz>
5964 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5965 order to ignore the predictor.
5966 (PRED_POLYMORPHIC_CALL): Likewise.
5967 (PRED_RECURSIVE_CALL): Likewise.
5969 2018-01-23 Martin Liska <mliska@suse.cz>
5971 * tree-profile.c (tree_profiling): Print function header to
5972 aware reader which function we are working on.
5973 * value-prof.c (gimple_find_values_to_profile): Do not print
5974 not interesting value histograms.
5976 2018-01-23 Martin Liska <mliska@suse.cz>
5978 * profile-count.h (enum profile_quality): Add
5979 profile_uninitialized as the first value. Do not number values
5980 as they are zero based.
5981 (profile_count::verify): Update sanity check.
5982 (profile_probability::verify): Likewise.
5984 2018-01-23 Nathan Sidwell <nathan@acm.org>
5986 * doc/invoke.texi (ffor-scope): Deprecate.
5988 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5990 PR tree-optimization/83510
5991 * domwalk.c (set_all_edges_as_executable): New function.
5992 (dom_walker::dom_walker): Convert bool param
5993 "skip_unreachable_blocks" to enum reachability. Move setup of
5994 edge flags to set_all_edges_as_executable and only do it when
5995 reachability is REACHABLE_BLOCKS.
5996 * domwalk.h (enum dom_walker::reachability): New enum.
5997 (dom_walker::dom_walker): Convert bool param
5998 "skip_unreachable_blocks" to enum reachability.
5999 (set_all_edges_as_executable): New decl.
6000 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6001 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6002 "reachability".
6003 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6004 but converting true to REACHABLE_BLOCKS.
6005 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6006 * tree-vrp.c
6007 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6008 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6009 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6010 REACHABLE_BLOCKS.
6011 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6012 if check_all_array_refs will be called.
6014 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6016 * tree.c (selftest::test_location_wrappers): Add more test
6017 coverage.
6019 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6021 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6022 (selftest::test_bit_in_range): Likewise.
6024 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6026 PR testsuite/83888
6027 * doc/sourcebuild.texi (vect_float): Say that the selector
6028 only describes the situation when -funsafe-math-optimizations is on.
6029 (vect_float_strict): Document.
6031 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6033 PR tree-optimization/83965
6034 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6035 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6036 instead of checking only for a reduction.
6037 (vect_recog_widen_sum_pattern): Likewise.
6039 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6041 * predict.c (probably_never_executed): Only use precise profile info.
6042 (compute_function_frequency): Skip after inlining hack since we now
6043 have quality checking.
6045 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6047 * profile-count.h (profile_probability::very_unlikely,
6048 profile_probability::unlikely, profile_probability::even): Set
6049 precision to guessed.
6051 2018-01-23 Richard Biener <rguenther@suse.de>
6053 PR tree-optimization/83963
6054 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6055 Properly terminate dominator walk when crossing the exit edge not
6056 when visiting its source block.
6058 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6060 PR c++/83918
6061 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6062 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6064 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6066 PR tree-optimization/83957
6067 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6068 semicolon after for body surrounded by braces.
6070 PR tree-optimization/83081
6071 * profile-count.h (profile_probability::split): New method.
6072 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6073 Use profile_probability::split.
6074 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6075 when splitting a single conditional jump into 2.
6077 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6079 PR tree-optimization/69452
6080 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6081 decl.
6083 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6085 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6086 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6087 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6089 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6091 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6092 * config/rl78/rl78.md (movdi): New define_expand.
6093 * config/rl78/rl78.c (rl78_split_movdi): New function.
6095 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6097 PR target/83862
6098 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6099 no longer used.
6100 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6101 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6102 128-bit to produce an UNSPEC move to get the double word with the
6103 signbit and then a shift directly to do signbit.
6104 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6105 implementation with a new version that just does either a direct
6106 move or a regular move. Move memory interface to separate insns.
6107 Move insns so they are next to the expander.
6108 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6109 with signbit move. Split big and little endian case.
6110 (signbit<mode>2_dm_mem_le): Likewise.
6111 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6112 (signbit<mode>2_dm2): Likewise.
6114 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6116 * config/rl78/rl78.md (anddi3): New define_expand.
6118 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6120 * config/rl78/rl78.md (umindi3): New define_expand.
6122 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6124 * config/rl78/rl78.md (smindi3): New define_expand.
6126 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6128 * config/rl78/rl78.md (smaxdi3): New define_expand.
6130 2018-01-22 Carl Love <cel@us.ibm.com>
6132 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6133 LVX_V1TI): Add macro expansion.
6134 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6135 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6136 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6137 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6138 Change check to determine if the instruction is a byte reversing
6139 entry. Fix typo in comment.
6140 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6141 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6142 Add def_builtin calls for new builtins.
6143 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6144 Add define_insn expansion.
6146 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6148 * config/rl78/rl78.md (umaxdi3): New define_expand.
6150 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6152 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6153 for non-QImode registers.
6155 2018-01-22 Richard Biener <rguenther@suse.de>
6157 PR tree-optimization/83963
6158 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6159 including the loop exit block.
6160 (scop_detection::merge_sese): Likewise.
6161 (scop_detection::add_scop): Do it here instead.
6163 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6165 * doc/sourcebuild.texi (arm_softfloat): Document.
6167 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6169 PR gcc/77734
6170 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6171 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6172 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6174 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6175 David Edelsohn <dje.gcc@gmail.com>
6177 PR target/83946
6178 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6179 Change "crset eq" to "crset 2".
6180 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6181 (*call_indirect_aix<mode>_nospec): Likewise.
6182 (*call_value_indirect_aix<mode>_nospec): Likewise.
6183 (*call_indirect_elfv2<mode>_nospec): Likewise.
6184 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6185 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6186 change assembly output from . to $.
6187 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6188 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6189 (*tablejump<mode>_internal1_nospec): Likewise.
6191 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6193 PR target/80870
6194 * config/sh/sh_optimize_sett_clrt.cc:
6195 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6197 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6199 PR tree-optimization/83940
6200 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6201 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6202 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6203 use it to pass back the definition type.
6204 (vect_check_store_rhs): Likewise rhs_dt_out.
6205 (vect_build_gather_load_calls): Add a mask_dt argument and use
6206 it instead of a call to vect_is_simple_use.
6207 (vectorizable_store): Update calls to vect_check_load_store_mask
6208 and vect_check_store_rhs. Use the dt returned by the latter instead
6209 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6210 instead of calls to vect_is_simple_use. Pass the scalar rather
6211 than the vector operand to vect_is_simple_use when handling
6212 second and subsequent copies of an rhs value.
6213 (vectorizable_load): Update calls to vect_check_load_store_mask
6214 and vect_build_gather_load_calls. Use the cached mask_dt and
6215 gs_info.offset_dt instead of calls to vect_is_simple_use.
6217 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6219 PR middle-end/83945
6220 * tree-emutls.c: Include gimplify.h.
6221 (lower_emutls_2): New function.
6222 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6223 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6224 it before further processing.
6226 PR target/83930
6227 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6228 UINTVAL (trueop1) instead of INTVAL (op1).
6230 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6232 PR debug/81570
6233 PR debug/83728
6234 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6235 INCOMING_FRAME_SP_OFFSET if not defined.
6236 (scan_trace): Add ENTRY argument. If true and
6237 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6238 emit a note to adjust the CFA offset.
6239 (create_cfi_notes): Adjust scan_trace callers.
6240 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6241 INCOMING_FRAME_SP_OFFSET in the CIE.
6242 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6243 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6244 Likewise.
6245 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6246 * doc/tm.texi: Regenerated.
6248 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6250 PR rtl-optimization/83147
6251 * lra-constraints.c (remove_inheritance_pseudos): Use
6252 lra_substitute_pseudo_within_insn.
6254 2018-01-19 Tom de Vries <tom@codesourcery.com>
6255 Cesar Philippidis <cesar@codesourcery.com>
6257 PR target/83920
6258 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6260 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6262 PR target/83790
6263 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6264 spaces for function labels.
6266 2018-01-19 Martin Liska <mliska@suse.cz>
6268 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6269 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6270 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6271 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6272 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6273 (PRED_CONST_RETURN): Change from 69 to 65.
6274 (PRED_NULL_RETURN): Change from 91 to 71.
6275 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6276 (PRED_LOOP_GUARD): Change from 66 to 73.
6278 2018-01-19 Martin Liska <mliska@suse.cz>
6280 * predict.c (predict_insn_def): Add new assert.
6281 (struct branch_predictor): Change type to signed integer.
6282 (test_prediction_value_range): Amend test to cover
6283 PROB_UNINITIALIZED.
6284 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6285 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6286 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6287 (PRED_LOOP_IV_COMPARE): Likewise.
6288 * predict.h (PROB_UNINITIALIZED): Define new constant.
6290 2018-01-19 Martin Liska <mliska@suse.cz>
6292 * predict.c (dump_prediction): Add new format for
6293 analyze_brprob.py script which is enabled with -details
6294 suboption.
6295 * profile-count.h (precise_p): New function.
6297 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6299 PR tree-optimization/83922
6300 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6301 there are no statements that need masking.
6302 (vect_active_double_reduction_p): New function.
6303 (vect_analyze_loop_operations): Use it when handling phis that
6304 are not in the loop header.
6306 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6308 PR tree-optimization/83914
6309 * tree-vect-loop.c (vectorizable_induction): Don't convert
6310 init_expr or apply the peeling adjustment for inductions
6311 that are nested within the vectorized loop.
6313 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6315 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6316 instead of NEG.
6318 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6320 PR sanitizer/81715
6321 PR testsuite/83882
6322 * function.h (gimplify_parameters): Add gimple_seq * argument.
6323 * function.c: Include gimple.h and options.h.
6324 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6325 for the added local temporaries if needed.
6326 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6327 if there are any parameter cleanups, wrap whole body into a
6328 try/finally with the cleanups.
6330 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6332 PR target/82964
6333 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6334 Use GET_MODE_CLASS for scalar floating point.
6336 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6338 PR ipa/82256
6339 patch by PaX Team
6340 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6341 Fix call of call_cgraph_insertion_hooks.
6343 2018-01-18 Martin Sebor <msebor@redhat.com>
6345 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6347 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6349 PR ipa/83619
6350 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6351 frequencies.
6353 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6355 PR other/70268
6356 * common.opt: (-ffile-prefix-map): New option.
6357 * opts.c (common_handle_option): Defer it.
6358 * opts-global.c (handle_common_deferred_options): Handle it.
6359 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6360 * file-prefix-map.h: New file.
6361 (remap_debug_filename, add_debug_prefix_map): ...here.
6362 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6363 * final.c (debug_prefix_map, add_debug_prefix_map
6364 remap_debug_filename): Move to...
6365 * file-prefix-map.c: New file.
6366 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6367 generalize, get rid of alloca(), use strrchr() instead of strchr().
6368 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6369 Implement in terms of add_prefix_map().
6370 (remap_macro_filename, remap_debug_filename): Implement in term of
6371 remap_filename().
6372 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6373 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6374 * dbxout.c: Include file-prefix-map.h.
6375 * varasm.c: Likewise.
6376 * vmsdbgout.c: Likewise.
6377 * xcoffout.c: Likewise.
6378 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6379 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6380 * doc/invoke.texi (-ffile-prefix-map): Document.
6381 (-fdebug-prefix-map): Update description.
6383 2018-01-18 Martin Liska <mliska@suse.cz>
6385 * config/i386/i386.c (indirect_thunk_name): Document that also
6386 lfence is emitted.
6387 (output_indirect_thunk): Document why both instructions
6388 (pause and lfence) are generated.
6390 2018-01-18 Richard Biener <rguenther@suse.de>
6392 PR tree-optimization/83887
6393 * graphite-scop-detection.c
6394 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6395 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6396 (scop_detection::merge_sese): Re-implement with a flood-fill
6397 algorithm that properly finds a SESE region if it exists.
6399 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6401 PR c/61240
6402 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6403 pointer_diff optimizations use view_convert instead of convert.
6405 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6407 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6408 Generate different code for -mno-speculate-indirect-jumps.
6409 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6410 (*call_indirect_aix<mode>): Disable for
6411 -mno-speculate-indirect-jumps.
6412 (*call_indirect_aix<mode>_nospec): New define_insn.
6413 (*call_value_indirect_aix<mode>): Disable for
6414 -mno-speculate-indirect-jumps.
6415 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6416 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6417 -mno-speculate-indirect-jumps.
6418 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6420 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6422 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6423 long double type, set the flags for noting the default long double
6424 type, even if we don't pass or return a long double type.
6426 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6428 PR ipa/83051
6429 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6430 failure.
6432 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6434 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6435 support for merge[hl].
6436 (fold_mergehl_helper): New helper function.
6437 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6438 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6439 (altivec_vmrglw_direct): Add xxmrglw insn.
6441 2018-01-17 Andrew Waterman <andrew@sifive.com>
6443 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6444 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6446 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6448 PR lto/83121
6449 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6450 call the lto_location_cache before reading the
6451 DECL_SOURCE_LOCATION of the types.
6453 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6454 Richard Sandiford <richard.sandiford@linaro.org>
6456 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6457 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6458 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6459 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6460 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6461 Add declaration.
6462 * config/aarch64/constraints.md (aarch64_movti_operand):
6463 Limit immediates.
6464 * config/aarch64/predicates.md (Uti): Add new constraint.
6466 2018-01-17 Carl Love <cel@us.ibm.com>
6468 * config/rs6000/vsx.md (define_expand xl_len_r,
6469 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6470 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6471 lxvll.
6472 (define_expand, define_insn): Move the shift left from the
6473 define_insn to the define_expand for lxvl and stxvl instructions.
6474 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6475 and XL_LEN_R definitions to PURE.
6477 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6479 * config/i386/i386.c (indirect_thunk_name): Declare regno
6480 as unsigned int. Compare regno with INVALID_REGNUM.
6481 (output_indirect_thunk): Ditto.
6482 (output_indirect_thunk_function): Ditto.
6483 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6484 in the call to output_indirect_thunk_function.
6486 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6488 PR middle-end/83884
6489 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6490 rather than the size of inner_type to determine the stack slot size
6491 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6493 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6495 PR target/83546
6496 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6497 to PTA_SILVERMONT.
6499 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6501 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6502 endian Linux systems to optionally enable multilibs for selecting
6503 the long double type if the user configured an explicit type.
6504 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6505 have no long double multilibs if not defined.
6506 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6507 warn if the user used -mabi={ieee,ibm}longdouble and we built
6508 multilibs for long double.
6509 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6510 appropriate multilib option.
6511 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6512 multilib options.
6513 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6514 for building long double multilibs.
6515 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6517 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6519 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6520 copies.
6522 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6523 64 bits.
6524 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6525 128 bits.
6527 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6528 variables.
6530 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6531 return value.
6533 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6535 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6536 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6538 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6540 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6541 different rtl trees depending on TARGET_64BIT.
6542 (rs6000_gen_lvx): Likewise.
6544 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6546 * config/visium/visium.md (nop): Tweak comment.
6547 (hazard_nop): Likewise.
6549 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6551 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6552 -mspeculate-indirect-jumps.
6553 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6554 for -mno-speculate-indirect-jumps.
6555 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6556 (*call_value_indirect_elfv2<mode>): Disable for
6557 -mno-speculate-indirect-jumps.
6558 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6559 (indirect_jump): Emit different RTL for
6560 -mno-speculate-indirect-jumps.
6561 (*indirect_jump<mode>): Disable for
6562 -mno-speculate-indirect-jumps.
6563 (*indirect_jump<mode>_nospec): New define_insn.
6564 (tablejump): Emit different RTL for
6565 -mno-speculate-indirect-jumps.
6566 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6567 (tablejumpsi_nospec): New define_expand.
6568 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6569 (tablejumpdi_nospec): New define_expand.
6570 (*tablejump<mode>_internal1): Disable for
6571 -mno-speculate-indirect-jumps.
6572 (*tablejump<mode>_internal1_nospec): New define_insn.
6573 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6574 option.
6576 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6578 * caller-save.c (insert_save): Drop unnecessary parameter. All
6579 callers updated.
6581 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6582 Richard Biener <rguenth@suse.de>
6584 PR libgomp/83590
6585 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6586 return early, inline manually is_gimple_sizepos. Make sure if we
6587 call gimplify_expr we don't end up with a gimple constant.
6588 * tree.c (variably_modified_type_p): Don't return true for
6589 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6590 * gimplify.h (is_gimple_sizepos): Remove.
6592 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6594 PR tree-optimization/83857
6595 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6596 vectorizable_live_operation for pure SLP statements.
6597 (vectorizable_live_operation): Handle PHIs.
6599 2018-01-16 Richard Biener <rguenther@suse.de>
6601 PR tree-optimization/83867
6602 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6603 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6605 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6607 PR c/83844
6608 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6609 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6610 If off is not INTEGER_CST, issue a may not be aligned warning
6611 rather than isn't aligned. Use isn%'t rather than isn't.
6612 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6613 into MULT_EXPR.
6614 <case MULT_EXPR>: Improve the case when bottom and one of the
6615 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6616 operand, in that case check if the other operand is multiple of
6617 bottom divided by the INTEGER_CST operand.
6619 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6621 PR target/83858
6622 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6623 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6624 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6625 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6626 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6627 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6628 (pa_function_arg_size): New function.
6630 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6632 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6633 in a separate statement.
6635 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6637 PR tree-optimization/83847
6638 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6639 group gathers and scatters.
6641 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6643 PR rtl-optimization/86620
6644 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6646 PR rtl-optimization/83213
6647 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6648 to last if both are JUMP_INSNs.
6650 PR tree-optimization/83843
6651 * gimple-ssa-store-merging.c
6652 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6653 store_immediate_info for bswap/nop orig_stores.
6655 2018-01-15 Andrew Waterman <andrew@sifive.com>
6657 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6658 !TARGET_MUL.
6659 <UDIV>: Increase cost if !TARGET_DIV.
6661 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6663 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6664 (define_attr "cr_logical_3op"): New.
6665 (cceq_ior_compare): Adjust.
6666 (cceq_ior_compare_complement): Adjust.
6667 (*cceq_rev_compare): Adjust.
6668 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6669 (is_cracked_insn): Adjust.
6670 (insn_must_be_first_in_group): Adjust.
6671 * config/rs6000/40x.md: Adjust.
6672 * config/rs6000/440.md: Adjust.
6673 * config/rs6000/476.md: Adjust.
6674 * config/rs6000/601.md: Adjust.
6675 * config/rs6000/603.md: Adjust.
6676 * config/rs6000/6xx.md: Adjust.
6677 * config/rs6000/7450.md: Adjust.
6678 * config/rs6000/7xx.md: Adjust.
6679 * config/rs6000/8540.md: Adjust.
6680 * config/rs6000/cell.md: Adjust.
6681 * config/rs6000/e300c2c3.md: Adjust.
6682 * config/rs6000/e500mc.md: Adjust.
6683 * config/rs6000/e500mc64.md: Adjust.
6684 * config/rs6000/e5500.md: Adjust.
6685 * config/rs6000/e6500.md: Adjust.
6686 * config/rs6000/mpc.md: Adjust.
6687 * config/rs6000/power4.md: Adjust.
6688 * config/rs6000/power5.md: Adjust.
6689 * config/rs6000/power6.md: Adjust.
6690 * config/rs6000/power7.md: Adjust.
6691 * config/rs6000/power8.md: Adjust.
6692 * config/rs6000/power9.md: Adjust.
6693 * config/rs6000/rs64.md: Adjust.
6694 * config/rs6000/titan.md: Adjust.
6696 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6698 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6699 ix86_indirect_branch_register logic.
6701 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6703 * config/i386/constraints.md (Bs): Update
6704 ix86_indirect_branch_register check. Don't check
6705 ix86_indirect_branch_register with GOT_memory_operand.
6706 (Bw): Likewise.
6707 * config/i386/predicates.md (GOT_memory_operand): Don't check
6708 ix86_indirect_branch_register here.
6709 (GOT32_symbol_operand): Likewise.
6711 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6713 * config/i386/predicates.md (constant_call_address_operand):
6714 Rewrite ix86_indirect_branch_register logic.
6715 (sibcall_insn_operand): Likewise.
6717 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6719 * config/i386/constraints.md (Bs): Replace
6720 ix86_indirect_branch_thunk_register with
6721 ix86_indirect_branch_register.
6722 (Bw): Likewise.
6723 * config/i386/i386.md (indirect_jump): Likewise.
6724 (tablejump): Likewise.
6725 (*sibcall_memory): Likewise.
6726 (*sibcall_value_memory): Likewise.
6727 Peepholes of indirect call and jump via memory: Likewise.
6728 * config/i386/i386.opt: Likewise.
6729 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6730 (GOT_memory_operand): Likewise.
6731 (call_insn_operand): Likewise.
6732 (sibcall_insn_operand): Likewise.
6733 (GOT32_symbol_operand): Likewise.
6735 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6737 PR middle-end/83837
6738 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6739 type rather than type addr's type points to.
6740 (expand_omp_atomic_mutex): Likewise.
6741 (expand_omp_atomic): Likewise.
6743 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6745 PR target/83839
6746 * config/i386/i386.c (output_indirect_thunk_function): Use
6747 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6748 for __x86_return_thunk.
6750 2018-01-15 Richard Biener <rguenther@suse.de>
6752 PR middle-end/83850
6753 * expmed.c (extract_bit_field_1): Fix typo.
6755 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6757 PR target/83687
6758 * config/arm/iterators.md (VF): New mode iterator.
6759 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6760 Remove integer-related logic from pattern.
6761 (neon_vabd<mode>_3): Likewise.
6763 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6765 PR middle-end/82694
6766 * common.opt (fstrict-overflow): No longer an alias.
6767 (fwrapv-pointer): New option.
6768 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6769 also for pointer types based on flag_wrapv_pointer.
6770 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6771 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6772 opts->x_flag_wrapv got set.
6773 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6774 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6775 POINTER_TYPE_OVERFLOW_UNDEFINED.
6776 * match.pd: Likewise in address comparison pattern.
6777 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6779 2018-01-15 Richard Biener <rguenther@suse.de>
6781 PR lto/83804
6782 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6783 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6784 Reset type names to their identifier if their TYPE_DECL doesn't
6785 have linkage (and thus is used for ODR and devirt).
6786 (save_debug_info_for_decl): Remove.
6787 (save_debug_info_for_type): Likewise.
6788 (add_tree_to_fld_list): Adjust.
6789 * tree-pretty-print.c (dump_generic_node): Make dumping of
6790 type names more robust.
6792 2018-01-15 Richard Biener <rguenther@suse.de>
6794 * BASE-VER: Bump to 8.0.1.
6796 2018-01-14 Martin Sebor <msebor@redhat.com>
6798 PR other/83508
6799 * builtins.c (check_access): Avoid warning when the no-warning bit
6800 is set.
6802 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6804 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6805 * ira-color (allocno_hard_regs_compare): Likewise.
6807 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6809 PR target/83013
6810 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6811 Use .pushsection/.popsection.
6813 2018-01-14 Martin Sebor <msebor@redhat.com>
6815 PR c++/81327
6816 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6818 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6820 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6821 entry from extra_headers.
6822 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6823 extra_headers, make the list bitwise identical to the i?86-*-* one.
6825 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6827 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6828 -mcmodel=large with -mindirect-branch=thunk,
6829 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6830 -mfunction-return=thunk-extern.
6831 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6832 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6833 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6835 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6837 * config/i386/i386.c (print_reg): Print the name of the full
6838 integer register without '%'.
6839 (ix86_print_operand): Handle 'V'.
6840 * doc/extend.texi: Document 'V' modifier.
6842 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6844 * config/i386/constraints.md (Bs): Disallow memory operand for
6845 -mindirect-branch-register.
6846 (Bw): Likewise.
6847 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6848 (GOT_memory_operand): Likewise.
6849 (call_insn_operand): Likewise.
6850 (sibcall_insn_operand): Likewise.
6851 (GOT32_symbol_operand): Likewise.
6852 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6853 for -mindirect-branch-register.
6854 (tablejump): Likewise.
6855 (*sibcall_memory): Likewise.
6856 (*sibcall_value_memory): Likewise.
6857 Disallow peepholes of indirect call and jump via memory for
6858 -mindirect-branch-register.
6859 (*call_pop): Replace m with Bw.
6860 (*call_value_pop): Likewise.
6861 (*sibcall_pop_memory): Replace m with Bs.
6862 * config/i386/i386.opt (mindirect-branch-register): New option.
6863 * doc/invoke.texi: Document -mindirect-branch-register option.
6865 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6867 * config/i386/i386-protos.h (ix86_output_function_return): New.
6868 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6869 set function_return_type.
6870 (indirect_thunk_name): Add ret_p to indicate thunk for function
6871 return.
6872 (output_indirect_thunk_function): Pass false to
6873 indirect_thunk_name.
6874 (ix86_output_indirect_branch_via_reg): Likewise.
6875 (ix86_output_indirect_branch_via_push): Likewise.
6876 (output_indirect_thunk_function): Create alias for function
6877 return thunk if regno < 0.
6878 (ix86_output_function_return): New function.
6879 (ix86_handle_fndecl_attribute): Handle function_return.
6880 (ix86_attribute_table): Add function_return.
6881 * config/i386/i386.h (machine_function): Add
6882 function_return_type.
6883 * config/i386/i386.md (simple_return_internal): Use
6884 ix86_output_function_return.
6885 (simple_return_internal_long): Likewise.
6886 * config/i386/i386.opt (mfunction-return=): New option.
6887 (indirect_branch): Mention -mfunction-return=.
6888 * doc/extend.texi: Document function_return function attribute.
6889 * doc/invoke.texi: Document -mfunction-return= option.
6891 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6893 * config/i386/i386-opts.h (indirect_branch): New.
6894 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6895 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6896 with local indirect jump when converting indirect call and jump.
6897 (ix86_set_indirect_branch_type): New.
6898 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6899 (indirectlabelno): New.
6900 (indirect_thunk_needed): Likewise.
6901 (indirect_thunk_bnd_needed): Likewise.
6902 (indirect_thunks_used): Likewise.
6903 (indirect_thunks_bnd_used): Likewise.
6904 (INDIRECT_LABEL): Likewise.
6905 (indirect_thunk_name): Likewise.
6906 (output_indirect_thunk): Likewise.
6907 (output_indirect_thunk_function): Likewise.
6908 (ix86_output_indirect_branch_via_reg): Likewise.
6909 (ix86_output_indirect_branch_via_push): Likewise.
6910 (ix86_output_indirect_branch): Likewise.
6911 (ix86_output_indirect_jmp): Likewise.
6912 (ix86_code_end): Call output_indirect_thunk_function if needed.
6913 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6914 needed.
6915 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6916 (ix86_attribute_table): Add indirect_branch.
6917 * config/i386/i386.h (machine_function): Add indirect_branch_type
6918 and has_local_indirect_jump.
6919 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6920 to true.
6921 (tablejump): Likewise.
6922 (*indirect_jump): Use ix86_output_indirect_jmp.
6923 (*tablejump_1): Likewise.
6924 (simple_return_indirect_internal): Likewise.
6925 * config/i386/i386.opt (mindirect-branch=): New option.
6926 (indirect_branch): New.
6927 (keep): Likewise.
6928 (thunk): Likewise.
6929 (thunk-inline): Likewise.
6930 (thunk-extern): Likewise.
6931 * doc/extend.texi: Document indirect_branch function attribute.
6932 * doc/invoke.texi: Document -mindirect-branch= option.
6934 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6936 PR ipa/83051
6937 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6939 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6941 * ipa-inline.c (want_inline_small_function_p): Return false if
6942 inlining has already failed with CIF_FINAL_ERROR.
6943 (update_caller_keys): Call want_inline_small_function_p before
6944 can_inline_edge_p.
6945 (update_callee_keys): Likewise.
6947 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6949 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6950 New function.
6951 (rs6000_quadword_masked_address_p): Likewise.
6952 (quad_aligned_load_p): Likewise.
6953 (quad_aligned_store_p): Likewise.
6954 (const_load_sequence_p): Add comment to describe the outer-most loop.
6955 (mimic_memory_attributes_and_flags): New function.
6956 (rs6000_gen_stvx): Likewise.
6957 (replace_swapped_aligned_store): Likewise.
6958 (rs6000_gen_lvx): Likewise.
6959 (replace_swapped_aligned_load): Likewise.
6960 (replace_swapped_load_constant): Capitalize argument name in
6961 comment describing this function.
6962 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6963 and stores that access quad-word aligned addresses and replace
6964 with stvx or lvx instructions when appropriate.
6965 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6966 New function prototype.
6967 (rs6000_quadword_masked_address_p): Likewise.
6968 (rs6000_gen_lvx): Likewise.
6969 (rs6000_gen_stvx): Likewise.
6970 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6971 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6972 when memory address is aligned.
6973 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6974 this split to select lvx instruction when memory address is aligned.
6975 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6976 instruction when memory address is aligned.
6977 (*vsx_le_perm_load_v16qi): Likewise.
6978 (four unnamed splitters): Modify to select the stvx instruction
6979 when memory is aligned.
6981 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6983 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6984 which appears in the queue multiple times.
6986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6987 Alan Hayward <alan.hayward@arm.com>
6988 David Sherwood <david.sherwood@arm.com>
6990 * tree-vectorizer.h (vec_lower_bound): New structure.
6991 (_loop_vec_info): Add check_nonzero and lower_bounds.
6992 (LOOP_VINFO_CHECK_NONZERO): New macro.
6993 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6994 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6995 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6996 fields. Make seg_len the distance travelled, not including the
6997 access size.
6998 (dr_direction_indicator): Declare.
6999 (dr_zero_step_indicator): Likewise.
7000 (dr_known_forward_stride_p): Likewise.
7001 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7002 tree-ssanames.h.
7003 (runtime_alias_check_p): Allow runtime alias checks with
7004 variable strides.
7005 (operator ==): Compare access_size and align.
7006 (prune_runtime_alias_test_list): Rework for new distinction between
7007 the access_size and seg_len.
7008 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7009 segment lengths.
7010 (get_segment_min_max): New function.
7011 (create_intersect_range_checks): Use it.
7012 (dr_step_indicator): New function.
7013 (dr_direction_indicator): Likewise.
7014 (dr_zero_step_indicator): Likewise.
7015 (dr_known_forward_stride_p): Likewise.
7016 * tree-loop-distribution.c (data_ref_segment_size): Return
7017 DR_STEP * (niters - 1).
7018 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7019 constructor.
7020 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7021 (vect_preserves_scalar_order_p): New function, split out from...
7022 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7023 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7024 (vect_vfa_access_size): New function.
7025 (vect_vfa_align): Likewise.
7026 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7027 (dump_lower_bound): New function.
7028 (vect_check_lower_bound): Likewise.
7029 (vect_small_gap_p): Likewise.
7030 (vectorizable_with_step_bound_p): Likewise.
7031 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7032 depencies if the vectorization factor is 1. Convert the checks
7033 for nonzero steps into checks on the bounds of DR_STEP. Try using
7034 a bunds check for variable steps if the minimum required step is
7035 relatively small. Update calls to the dr_with_seg_len
7036 constructor and to vect_compile_time_alias.
7037 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7038 function.
7039 (vect_loop_versioning): Call it.
7040 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7041 when retrying.
7042 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7044 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7045 Alan Hayward <alan.hayward@arm.com>
7046 David Sherwood <david.sherwood@arm.com>
7048 * doc/sourcebuild.texi (vect_scatter_store): Document.
7049 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7050 optabs.
7051 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7052 Document.
7053 * genopinit.c (main): Add supports_vec_scatter_store and
7054 supports_vec_scatter_store_cached to target_optabs.
7055 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7056 IFN_MASK_SCATTER_STORE.
7057 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7058 functions.
7059 * internal-fn.h (internal_store_fn_p): Declare.
7060 (internal_fn_stored_value_index): Likewise.
7061 * internal-fn.c (scatter_store_direct): New macro.
7062 (expand_scatter_store_optab_fn): New function.
7063 (direct_scatter_store_optab_supported_p): New macro.
7064 (internal_store_fn_p): New function.
7065 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7066 IFN_MASK_SCATTER_STORE.
7067 (internal_fn_mask_index): Likewise.
7068 (internal_fn_stored_value_index): New function.
7069 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7070 for scatter stores.
7071 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7072 * optabs-query.c (supports_vec_scatter_store_p): New function.
7073 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7074 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7075 true for scatter stores.
7076 (vect_gather_scatter_fn_p): Handle scatter stores too.
7077 (vect_check_gather_scatter): Consider using scatter stores if
7078 supports_vec_scatter_store_p.
7079 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7080 scatter stores too.
7081 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7082 internal_fn_stored_value_index.
7083 (check_load_store_masking): Handle scatter stores too.
7084 (vect_get_store_rhs): Make public.
7085 (vectorizable_call): Use internal_store_fn_p.
7086 (vectorizable_store): Handle scatter store internal functions.
7087 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7088 when deciding whether the end of the group has been reached.
7089 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7090 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7091 (mask_scatter_store<mode>): New insns.
7093 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7094 Alan Hayward <alan.hayward@arm.com>
7095 David Sherwood <david.sherwood@arm.com>
7097 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7098 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7099 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7100 function.
7101 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7102 Use vect_truncate_gather_scatter_offset if we can't treat the
7103 operation as a normal gather load or scatter store.
7104 (get_group_load_store_type): Take the gather_scatter_info
7105 as argument. Try using a gather load or scatter store for
7106 single-element groups.
7107 (get_load_store_type): Update calls to get_group_load_store_type
7108 and vect_use_strided_gather_scatters_p.
7110 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7111 Alan Hayward <alan.hayward@arm.com>
7112 David Sherwood <david.sherwood@arm.com>
7114 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7115 optional tree argument.
7116 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7117 null target hooks.
7118 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7119 but continue to use the current value as a fallback.
7120 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7121 to compare the updates.
7122 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7123 (get_load_store_type): Use it when handling a strided access.
7124 (vect_get_strided_load_store_ops): New function.
7125 (vect_get_data_ptr_increment): Likewise.
7126 (vectorizable_load): Handle strided gather loads. Always pass
7127 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7129 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7130 Alan Hayward <alan.hayward@arm.com>
7131 David Sherwood <david.sherwood@arm.com>
7133 * doc/md.texi (gather_load@var{m}): Document.
7134 (mask_gather_load@var{m}): Likewise.
7135 * genopinit.c (main): Add supports_vec_gather_load and
7136 supports_vec_gather_load_cached to target_optabs.
7137 * optabs-tree.c (init_tree_optimization_optabs): Use
7138 ggc_cleared_alloc to allocate target_optabs.
7139 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7140 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7141 functions.
7142 * internal-fn.h (internal_load_fn_p): Declare.
7143 (internal_gather_scatter_fn_p): Likewise.
7144 (internal_fn_mask_index): Likewise.
7145 (internal_gather_scatter_fn_supported_p): Likewise.
7146 * internal-fn.c (gather_load_direct): New macro.
7147 (expand_gather_load_optab_fn): New function.
7148 (direct_gather_load_optab_supported_p): New macro.
7149 (direct_internal_fn_optab): New function.
7150 (internal_load_fn_p): Likewise.
7151 (internal_gather_scatter_fn_p): Likewise.
7152 (internal_fn_mask_index): Likewise.
7153 (internal_gather_scatter_fn_supported_p): Likewise.
7154 * optabs-query.c (supports_at_least_one_mode_p): New function.
7155 (supports_vec_gather_load_p): Likewise.
7156 * optabs-query.h (supports_vec_gather_load_p): Declare.
7157 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7158 and memory_type field.
7159 (NUM_PATTERNS): Bump to 15.
7160 * tree-vect-data-refs.c: Include internal-fn.h.
7161 (vect_gather_scatter_fn_p): New function.
7162 (vect_describe_gather_scatter_call): Likewise.
7163 (vect_check_gather_scatter): Try using internal functions for
7164 gather loads. Recognize existing calls to a gather load function.
7165 (vect_analyze_data_refs): Consider using gather loads if
7166 supports_vec_gather_load_p.
7167 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7168 (vect_get_gather_scatter_offset_type): Likewise.
7169 (vect_convert_mask_for_vectype): Likewise.
7170 (vect_add_conversion_to_patterm): Likewise.
7171 (vect_try_gather_scatter_pattern): Likewise.
7172 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7173 (vect_vect_recog_func_ptrs): Add it.
7174 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7175 internal_fn_mask_index and internal_gather_scatter_fn_p.
7176 (check_load_store_masking): Take the gather_scatter_info as an
7177 argument and handle gather loads.
7178 (vect_get_gather_scatter_ops): New function.
7179 (vectorizable_call): Check internal_load_fn_p.
7180 (vectorizable_load): Likewise. Handle gather load internal
7181 functions.
7182 (vectorizable_store): Update call to check_load_store_masking.
7183 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7184 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7185 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7186 (aarch64_gather_scale_operand_d): New predicates.
7187 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7188 (mask_gather_load<mode>): New insns.
7190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7191 Alan Hayward <alan.hayward@arm.com>
7192 David Sherwood <david.sherwood@arm.com>
7194 * optabs.def (fold_left_plus_optab): New optab.
7195 * doc/md.texi (fold_left_plus_@var{m}): Document.
7196 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7197 * internal-fn.c (fold_left_direct): Define.
7198 (expand_fold_left_optab_fn): Likewise.
7199 (direct_fold_left_optab_supported_p): Likewise.
7200 * fold-const-call.c (fold_const_fold_left): New function.
7201 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7202 * tree-parloops.c (valid_reduction_p): New function.
7203 (gather_scalar_reductions): Use it.
7204 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7205 (vect_finish_replace_stmt): Declare.
7206 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7207 (needs_fold_left_reduction_p): New function, split out from...
7208 (vect_is_simple_reduction): ...here. Accept reductions that
7209 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7210 (vect_force_simple_reduction): Also store the reduction type in
7211 the assignment's STMT_VINFO_REDUC_TYPE.
7212 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7213 (merge_with_identity): New function.
7214 (vect_expand_fold_left): Likewise.
7215 (vectorize_fold_left_reduction): Likewise.
7216 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7217 scalar phi in place for it. Check for target support and reject
7218 cases that would reassociate the operation. Defer the transform
7219 phase to vectorize_fold_left_reduction.
7220 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7221 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7222 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7224 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7226 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7227 call to ifc_temp_var.
7229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7230 Alan Hayward <alan.hayward@arm.com>
7231 David Sherwood <david.sherwood@arm.com>
7233 * target.def (legitimize_address_displacement): Take the original
7234 offset as a poly_int.
7235 * targhooks.h (default_legitimize_address_displacement): Update
7236 accordingly.
7237 * targhooks.c (default_legitimize_address_displacement): Likewise.
7238 * doc/tm.texi: Regenerate.
7239 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7240 as an argument, moving assert of ad->disp == ad->disp_term to...
7241 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7242 Try calling targetm.legitimize_address_displacement before expanding
7243 the address rather than afterwards, and adjust for the new interface.
7244 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7245 Match the new hook interface. Handle SVE addresses.
7246 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7247 new hook interface.
7249 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7251 * Makefile.in (OBJS): Add early-remat.o.
7252 * target.def (select_early_remat_modes): New hook.
7253 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7254 * doc/tm.texi: Regenerate.
7255 * targhooks.h (default_select_early_remat_modes): Declare.
7256 * targhooks.c (default_select_early_remat_modes): New function.
7257 * timevar.def (TV_EARLY_REMAT): New timevar.
7258 * passes.def (pass_early_remat): New pass.
7259 * tree-pass.h (make_pass_early_remat): Declare.
7260 * early-remat.c: New file.
7261 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7262 function.
7263 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7266 Alan Hayward <alan.hayward@arm.com>
7267 David Sherwood <david.sherwood@arm.com>
7269 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7270 vfm1 with a bound_epilog parameter.
7271 (vect_do_peeling): Update calls accordingly, and move the prologue
7272 call earlier in the function. Treat the base bound_epilog as 0 for
7273 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7274 this base when peeling for gaps.
7275 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7276 with fully-masked loops.
7277 (vect_estimate_min_profitable_iters): Handle the single peeled
7278 iteration in that case.
7280 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7281 Alan Hayward <alan.hayward@arm.com>
7282 David Sherwood <david.sherwood@arm.com>
7284 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7285 single-element interleaving even if the size is not a power of 2.
7286 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7287 accesses for single-element interleaving if the group size is
7288 not a power of 2.
7290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7291 Alan Hayward <alan.hayward@arm.com>
7292 David Sherwood <david.sherwood@arm.com>
7294 * doc/md.texi (fold_extract_last_@var{m}): Document.
7295 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7296 * optabs.def (fold_extract_last_optab): New optab.
7297 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7298 * internal-fn.c (fold_extract_direct): New macro.
7299 (expand_fold_extract_optab_fn): Likewise.
7300 (direct_fold_extract_optab_supported_p): Likewise.
7301 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7302 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7303 EXTRACT_LAST_REDUCTION.
7304 (get_initial_def_for_reduction): Do not create an initial vector
7305 for EXTRACT_LAST_REDUCTION reductions.
7306 (vectorizable_reduction): Leave the scalar phi in place for
7307 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7308 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7309 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7310 transform phase to vectorizable_condition.
7311 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7312 split out from...
7313 (vect_finish_stmt_generation): ...here.
7314 (vect_finish_replace_stmt): New function.
7315 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7316 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7317 pattern.
7318 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7320 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7321 Alan Hayward <alan.hayward@arm.com>
7322 David Sherwood <david.sherwood@arm.com>
7324 * doc/md.texi (extract_last_@var{m}): Document.
7325 * optabs.def (extract_last_optab): New optab.
7326 * internal-fn.def (EXTRACT_LAST): New internal function.
7327 * internal-fn.c (cond_unary_direct): New macro.
7328 (expand_cond_unary_optab_fn): Likewise.
7329 (direct_cond_unary_optab_supported_p): Likewise.
7330 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7331 loops using EXTRACT_LAST.
7332 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7333 (extract_last_<mode>): ...this optab.
7334 (vec_extract<mode><Vel>): Update accordingly.
7336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7337 Alan Hayward <alan.hayward@arm.com>
7338 David Sherwood <david.sherwood@arm.com>
7340 * target.def (empty_mask_is_expensive): New hook.
7341 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7342 * doc/tm.texi: Regenerate.
7343 * targhooks.h (default_empty_mask_is_expensive): Declare.
7344 * targhooks.c (default_empty_mask_is_expensive): New function.
7345 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7346 if the target says that empty masks are expensive.
7347 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7348 New function.
7349 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7352 Alan Hayward <alan.hayward@arm.com>
7353 David Sherwood <david.sherwood@arm.com>
7355 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7356 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7357 (vect_use_loop_mask_for_alignment_p): New function.
7358 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7359 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7360 niters_skip argument. Make sure that the first niters_skip elements
7361 of the first iteration are inactive.
7362 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7363 Update call to vect_set_loop_masks_directly.
7364 (get_misalign_in_elems): New function, split out from...
7365 (vect_gen_prolog_loop_niters): ...here.
7366 (vect_update_init_of_dr): Take a code argument that specifies whether
7367 the adjustment should be added or subtracted.
7368 (vect_update_init_of_drs): Likewise.
7369 (vect_prepare_for_masked_peels): New function.
7370 (vect_do_peeling): Skip prologue peeling if we're using a mask
7371 instead. Update call to vect_update_inits_of_drs.
7372 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7373 mask_skip_niters.
7374 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7375 alignment. Do not include the number of peeled iterations in
7376 the minimum threshold in that case.
7377 (vectorizable_induction): Adjust the start value down by
7378 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7379 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7380 Take the number of skipped iterations into account when calculating
7381 the loop bounds.
7382 * tree-vect-stmts.c (vect_gen_while_not): New function.
7384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7385 Alan Hayward <alan.hayward@arm.com>
7386 David Sherwood <david.sherwood@arm.com>
7388 * doc/sourcebuild.texi (vect_fully_masked): Document.
7389 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7390 default value to 0.
7391 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7392 split out from...
7393 (vect_analyze_loop_2): ...here. Don't check the vectorization
7394 factor against the number of loop iterations if the loop is
7395 fully-masked.
7397 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7398 Alan Hayward <alan.hayward@arm.com>
7399 David Sherwood <david.sherwood@arm.com>
7401 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7402 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7403 (dump_groups): Update accordingly.
7404 (iv_use::mem_type): New member variable.
7405 (address_p): New function.
7406 (record_use): Add a mem_type argument and initialize the new
7407 mem_type field.
7408 (record_group_use): Add a mem_type argument. Use address_p.
7409 Remove obsolete null checks of base_object. Update call to record_use.
7410 (find_interesting_uses_op): Update call to record_group_use.
7411 (find_interesting_uses_cond): Likewise.
7412 (find_interesting_uses_address): Likewise.
7413 (get_mem_type_for_internal_fn): New function.
7414 (find_address_like_use): Likewise.
7415 (find_interesting_uses_stmt): Try find_address_like_use before
7416 calling find_interesting_uses_op.
7417 (addr_offset_valid_p): Use the iv mem_type field as the type
7418 of the addressed memory.
7419 (add_autoinc_candidates): Likewise.
7420 (get_address_cost): Likewise.
7421 (split_small_address_groups_p): Use address_p.
7422 (split_address_groups): Likewise.
7423 (add_iv_candidate_for_use): Likewise.
7424 (autoinc_possible_for_pair): Likewise.
7425 (rewrite_groups): Likewise.
7426 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7427 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7428 (get_alias_ptr_type_for_ptr_address): New function.
7429 (rewrite_use_address): Rewrite address uses in calls that were
7430 identified by find_address_like_use.
7432 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7433 Alan Hayward <alan.hayward@arm.com>
7434 David Sherwood <david.sherwood@arm.com>
7436 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7437 TARGET_MEM_REFs.
7438 * gimple-expr.h (is_gimple_addressable: Likewise.
7439 * gimple-expr.c (is_gimple_address): Likewise.
7440 * internal-fn.c (expand_call_mem_ref): New function.
7441 (expand_mask_load_optab_fn): Use it.
7442 (expand_mask_store_optab_fn): Likewise.
7444 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7445 Alan Hayward <alan.hayward@arm.com>
7446 David Sherwood <david.sherwood@arm.com>
7448 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7449 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7450 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7451 (cond_umax@var{mode}): Document.
7452 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7453 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7454 (cond_umin_optab, cond_umax_optab): New optabs.
7455 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7456 (COND_IOR, COND_XOR): New internal functions.
7457 * internal-fn.h (get_conditional_internal_fn): Declare.
7458 * internal-fn.c (cond_binary_direct): New macro.
7459 (expand_cond_binary_optab_fn): Likewise.
7460 (direct_cond_binary_optab_supported_p): Likewise.
7461 (get_conditional_internal_fn): New function.
7462 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7463 Cope with reduction statements that are vectorized as calls rather
7464 than assignments.
7465 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7466 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7467 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7468 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7469 (UNSPEC_COND_EOR): New unspecs.
7470 (optab): Add mappings for them.
7471 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7472 (sve_int_op, sve_fp_op): New int attributes.
7474 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7475 Alan Hayward <alan.hayward@arm.com>
7476 David Sherwood <david.sherwood@arm.com>
7478 * optabs.def (while_ult_optab): New optab.
7479 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7480 * internal-fn.def (WHILE_ULT): New internal function.
7481 * internal-fn.h (direct_internal_fn_supported_p): New override
7482 that takes two types as argument.
7483 * internal-fn.c (while_direct): New macro.
7484 (expand_while_optab_fn): New function.
7485 (convert_optab_supported_p): Likewise.
7486 (direct_while_optab_supported_p): New macro.
7487 * wide-int.h (wi::udiv_ceil): New function.
7488 * tree-vectorizer.h (rgroup_masks): New structure.
7489 (vec_loop_masks): New typedef.
7490 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7491 and fully_masked_p.
7492 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7493 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7494 (vect_max_vf): New function.
7495 (slpeel_make_loop_iterate_ntimes): Delete.
7496 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7497 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7498 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7499 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7500 internal-fn.h, stor-layout.h and optabs-query.h.
7501 (vect_set_loop_mask): New function.
7502 (add_preheader_seq): Likewise.
7503 (add_header_seq): Likewise.
7504 (interleave_supported_p): Likewise.
7505 (vect_maybe_permute_loop_masks): Likewise.
7506 (vect_set_loop_masks_directly): Likewise.
7507 (vect_set_loop_condition_masked): Likewise.
7508 (vect_set_loop_condition_unmasked): New function, split out from
7509 slpeel_make_loop_iterate_ntimes.
7510 (slpeel_make_loop_iterate_ntimes): Rename to..
7511 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7512 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7513 (vect_do_peeling): Update call accordingly.
7514 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7515 loops.
7516 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7517 mask_compare_type, can_fully_mask_p and fully_masked_p.
7518 (release_vec_loop_masks): New function.
7519 (_loop_vec_info): Use it to free the loop masks.
7520 (can_produce_all_loop_masks_p): New function.
7521 (vect_get_max_nscalars_per_iter): Likewise.
7522 (vect_verify_full_masking): Likewise.
7523 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7524 retries, and free the mask rgroups before retrying. Check loop-wide
7525 reasons for disallowing fully-masked loops. Make the final decision
7526 about whether use a fully-masked loop or not.
7527 (vect_estimate_min_profitable_iters): Do not assume that peeling
7528 for the number of iterations will be needed for fully-masked loops.
7529 (vectorizable_reduction): Disable fully-masked loops.
7530 (vectorizable_live_operation): Likewise.
7531 (vect_halve_mask_nunits): New function.
7532 (vect_double_mask_nunits): Likewise.
7533 (vect_record_loop_mask): Likewise.
7534 (vect_get_loop_mask): Likewise.
7535 (vect_transform_loop): Handle the case in which the final loop
7536 iteration might handle a partial vector. Call vect_set_loop_condition
7537 instead of slpeel_make_loop_iterate_ntimes.
7538 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7539 (check_load_store_masking): New function.
7540 (prepare_load_store_mask): Likewise.
7541 (vectorizable_store): Handle fully-masked loops.
7542 (vectorizable_load): Likewise.
7543 (supportable_widening_operation): Use vect_halve_mask_nunits for
7544 booleans.
7545 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7546 (vect_gen_while): New function.
7547 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7548 (aarch64_uqdec<mode>): New insn.
7550 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7551 Alan Hayward <alan.hayward@arm.com>
7552 David Sherwood <david.sherwood@arm.com>
7554 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7555 (reduc_xor_scal_optab): New optabs.
7556 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7557 (reduc_xor_scal_@var{m}): Document.
7558 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7559 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7560 internal functions.
7561 * fold-const-call.c (fold_const_call): Handle them.
7562 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7563 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7564 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7565 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7566 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7567 (UNSPEC_XORV): New unspecs.
7568 (optab): Add entries for them.
7569 (BITWISEV): New int iterator.
7570 (bit_reduc_op): New int attributes.
7572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7573 Alan Hayward <alan.hayward@arm.com>
7574 David Sherwood <david.sherwood@arm.com>
7576 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7577 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7578 * optabs.def (vec_shl_insert_optab): New optab.
7579 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7580 (duplicate_and_interleave): Likewise.
7581 * tree-vect-loop.c: Include internal-fn.h.
7582 (neutral_op_for_slp_reduction): New function, split out from
7583 get_initial_defs_for_reduction.
7584 (get_initial_def_for_reduction): Handle option 2 for variable-length
7585 vectors by loading the neutral value into a vector and then shifting
7586 the initial value into element 0.
7587 (get_initial_defs_for_reduction): Replace the code argument with
7588 the neutral value calculated by neutral_op_for_slp_reduction.
7589 Use gimple_build_vector for constant-length vectors.
7590 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7591 but the first group_size elements have a neutral value.
7592 Use duplicate_and_interleave otherwise.
7593 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7594 Update call to get_initial_defs_for_reduction. Handle SLP
7595 reductions for variable-length vectors by creating one vector
7596 result for each scalar result, with the elements associated
7597 with other scalar results stubbed out with the neutral value.
7598 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7599 Require IFN_VEC_SHL_INSERT for double reductions on
7600 variable-length vectors, or SLP reductions that have
7601 a neutral value. Require can_duplicate_and_interleave_p
7602 support for variable-length unchained SLP reductions if there
7603 is no neutral value, such as for MIN/MAX reductions. Also require
7604 the number of vector elements to be a multiple of the number of
7605 SLP statements when doing variable-length unchained SLP reductions.
7606 Update call to vect_create_epilog_for_reduction.
7607 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7608 and remove initial values.
7609 (duplicate_and_interleave): Make public.
7610 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7611 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7614 Alan Hayward <alan.hayward@arm.com>
7615 David Sherwood <david.sherwood@arm.com>
7617 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7618 (can_duplicate_and_interleave_p): New function.
7619 (vect_get_and_check_slp_defs): Take the vector of statements
7620 rather than just the current one. Remove excess parentheses.
7621 Restriction rejectinon of vect_constant_def and vect_external_def
7622 for variable-length vectors to boolean types, or types for which
7623 can_duplicate_and_interleave_p is false.
7624 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7625 (duplicate_and_interleave): New function.
7626 (vect_get_constant_vectors): Use gimple_build_vector for
7627 constant-length vectors and suitable variable-length constant
7628 vectors. Use duplicate_and_interleave for other variable-length
7629 vectors. Don't defer the update when inserting new statements.
7631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7632 Alan Hayward <alan.hayward@arm.com>
7633 David Sherwood <david.sherwood@arm.com>
7635 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7636 min_profitable_iters doesn't go negative.
7638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7639 Alan Hayward <alan.hayward@arm.com>
7640 David Sherwood <david.sherwood@arm.com>
7642 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7643 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7644 * optabs.def (vec_mask_load_lanes_optab): New optab.
7645 (vec_mask_store_lanes_optab): Likewise.
7646 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7647 (MASK_STORE_LANES): Likewise.
7648 * internal-fn.c (mask_load_lanes_direct): New macro.
7649 (mask_store_lanes_direct): Likewise.
7650 (expand_mask_load_optab_fn): Handle masked operations.
7651 (expand_mask_load_lanes_optab_fn): New macro.
7652 (expand_mask_store_optab_fn): Handle masked operations.
7653 (expand_mask_store_lanes_optab_fn): New macro.
7654 (direct_mask_load_lanes_optab_supported_p): Likewise.
7655 (direct_mask_store_lanes_optab_supported_p): Likewise.
7656 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7657 parameter.
7658 (vect_load_lanes_supported): Likewise.
7659 * tree-vect-data-refs.c (strip_conversion): New function.
7660 (can_group_stmts_p): Likewise.
7661 (vect_analyze_data_ref_accesses): Use it instead of checking
7662 for a pair of assignments.
7663 (vect_store_lanes_supported): Take a masked_p parameter.
7664 (vect_load_lanes_supported): Likewise.
7665 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7666 vect_store_lanes_supported and vect_load_lanes_supported.
7667 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7668 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7669 parameter. Don't allow gaps for masked accesses.
7670 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7671 and vect_load_lanes_supported.
7672 (get_load_store_type): Take a masked_p parameter and update
7673 call to get_group_load_store_type.
7674 (vectorizable_store): Update call to get_load_store_type.
7675 Handle IFN_MASK_STORE_LANES.
7676 (vectorizable_load): Update call to get_load_store_type.
7677 Handle IFN_MASK_LOAD_LANES.
7679 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7680 Alan Hayward <alan.hayward@arm.com>
7681 David Sherwood <david.sherwood@arm.com>
7683 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7684 modes for SVE.
7685 * config/aarch64/aarch64-protos.h
7686 (aarch64_sve_struct_memory_operand_p): Declare.
7687 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7688 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7689 (VPRED, vpred): Handle SVE structure modes.
7690 * config/aarch64/constraints.md (Utx): New constraint.
7691 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7692 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7693 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7694 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7695 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7696 structure modes. Split into pieces after RA.
7697 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7698 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7699 New patterns.
7700 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7701 SVE structure modes.
7702 (aarch64_classify_address): Likewise.
7703 (sizetochar): Move earlier in file.
7704 (aarch64_print_operand): Handle SVE register lists.
7705 (aarch64_array_mode): New function.
7706 (aarch64_sve_struct_memory_operand_p): Likewise.
7707 (TARGET_ARRAY_MODE): Redefine.
7709 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7710 Alan Hayward <alan.hayward@arm.com>
7711 David Sherwood <david.sherwood@arm.com>
7713 * target.def (array_mode): New target hook.
7714 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7715 * doc/tm.texi: Regenerate.
7716 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7717 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7718 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7719 targetm.array_mode.
7720 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7721 type sizes.
7723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7724 Alan Hayward <alan.hayward@arm.com>
7725 David Sherwood <david.sherwood@arm.com>
7727 * fold-const.c (fold_binary_loc): Check the argument types
7728 rather than the result type when testing for a vector operation.
7730 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7732 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7733 * doc/tm.texi: Regenerate.
7735 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7736 Alan Hayward <alan.hayward@arm.com>
7737 David Sherwood <david.sherwood@arm.com>
7739 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7740 (sve): Document new AArch64 extension.
7741 * doc/md.texi (w): Extend the description of the AArch64
7742 constraint to include SVE vectors.
7743 (Upl, Upa): Document new AArch64 predicate constraints.
7744 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7745 enum.
7746 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7747 (msve-vector-bits=): New option.
7748 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7749 SVE when these are disabled.
7750 (sve): New extension.
7751 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7752 modes. Adjust their number of units based on aarch64_sve_vg.
7753 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7754 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7755 aarch64_addr_query_type.
7756 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7757 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7758 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7759 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7760 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7761 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7762 (aarch64_simd_imm_zero_p): Delete.
7763 (aarch64_check_zero_based_sve_index_immediate): Declare.
7764 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7765 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7766 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7767 (aarch64_sve_float_mul_immediate_p): Likewise.
7768 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7769 rather than an rtx.
7770 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7771 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7772 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7773 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7774 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7775 (aarch64_regmode_natural_size): Likewise.
7776 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7777 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7778 left one place.
7779 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7780 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7781 for VG and the SVE predicate registers.
7782 (V_ALIASES): Add a "z"-prefixed alias.
7783 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7784 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7785 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7786 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7787 (REG_CLASS_NAMES): Add entries for them.
7788 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7789 and the predicate registers.
7790 (aarch64_sve_vg): Declare.
7791 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7792 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7793 (REGMODE_NATURAL_SIZE): Define.
7794 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7795 SVE macros.
7796 * config/aarch64/aarch64.c: Include cfgrtl.h.
7797 (simd_immediate_info): Add a constructor for series vectors,
7798 and an associated step field.
7799 (aarch64_sve_vg): New variable.
7800 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7801 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7802 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7803 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7804 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7805 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7806 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7807 (aarch64_get_mask_mode): New functions.
7808 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7809 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7810 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7811 predicate modes and predicate registers. Explicitly restrict
7812 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7813 to store a vector mode if it is recognized by
7814 aarch64_classify_vector_mode.
7815 (aarch64_regmode_natural_size): New function.
7816 (aarch64_hard_regno_caller_save_mode): Return the original mode
7817 for predicates.
7818 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7819 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7820 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7821 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7822 functions.
7823 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7824 does not overlap dest if the function is frame-related. Handle
7825 SVE constants.
7826 (aarch64_split_add_offset): New function.
7827 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7828 them aarch64_add_offset.
7829 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7830 and update call to aarch64_sub_sp.
7831 (aarch64_add_cfa_expression): New function.
7832 (aarch64_expand_prologue): Pass extra temporary registers to the
7833 functions above. Handle the case in which we need to emit new
7834 DW_CFA_expressions for registers that were originally saved
7835 relative to the stack pointer, but now have to be expressed
7836 relative to the frame pointer.
7837 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7838 functions above.
7839 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7840 IP0 and IP1 values for SVE frames.
7841 (aarch64_expand_vec_series): New function.
7842 (aarch64_expand_sve_widened_duplicate): Likewise.
7843 (aarch64_expand_sve_const_vector): Likewise.
7844 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7845 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7846 into the register, rather than emitting a SET directly.
7847 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7848 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7849 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7850 (offset_9bit_signed_scaled_p): New functions.
7851 (aarch64_replicate_bitmask_imm): New function.
7852 (aarch64_bitmask_imm): Use it.
7853 (aarch64_cannot_force_const_mem): Reject expressions involving
7854 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7855 (aarch64_classify_index): Handle SVE indices, by requiring
7856 a plain register index with a scale that matches the element size.
7857 (aarch64_classify_address): Handle SVE addresses. Assert that
7858 the mode of the address is VOIDmode or an integer mode.
7859 Update call to aarch64_classify_symbol.
7860 (aarch64_classify_symbolic_expression): Update call to
7861 aarch64_classify_symbol.
7862 (aarch64_const_vec_all_in_range_p): New function.
7863 (aarch64_print_vector_float_operand): Likewise.
7864 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7865 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7866 and the FP immediates 1.0 and 0.5.
7867 (aarch64_print_address_internal): Handle SVE addresses.
7868 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7869 (aarch64_regno_regclass): Handle predicate registers.
7870 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7871 data modes.
7872 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7873 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7874 (aarch64_convert_sve_vector_bits): New function.
7875 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7876 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7877 rather than an rtx.
7878 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7879 Handle SVE vector and predicate modes. Accept VL-based constants
7880 that need only one temporary register, and VL offsets that require
7881 no temporary registers.
7882 (aarch64_conditional_register_usage): Mark the predicate registers
7883 as fixed if SVE isn't available.
7884 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7885 Return true for SVE vector and predicate modes.
7886 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7887 rather than an unsigned int. Handle SVE modes.
7888 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7889 SVE modes.
7890 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7891 if SVE is enabled.
7892 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7893 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7894 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7895 (aarch64_sve_float_mul_immediate_p): New functions.
7896 (aarch64_sve_valid_immediate): New function.
7897 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7898 Explicitly reject structure modes. Check for INDEX constants.
7899 Handle PTRUE and PFALSE constants.
7900 (aarch64_check_zero_based_sve_index_immediate): New function.
7901 (aarch64_simd_imm_zero_p): Delete.
7902 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7903 vector modes. Accept constants in the range of CNT[BHWD].
7904 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7905 ask for an Advanced SIMD mode.
7906 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7907 (aarch64_simd_vector_alignment): Handle SVE predicates.
7908 (aarch64_vectorize_preferred_vector_alignment): New function.
7909 (aarch64_simd_vector_alignment_reachable): Use it instead of
7910 the vector size.
7911 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7912 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7913 functions.
7914 (MAX_VECT_LEN): Delete.
7915 (expand_vec_perm_d): Add a vec_flags field.
7916 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7917 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7918 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7919 for SVE modes.
7920 (aarch64_evpc_rev): Rename to...
7921 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7922 (aarch64_evpc_rev_global): New function.
7923 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7924 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7925 MAX_VECT_LEN.
7926 (aarch64_evpc_sve_tbl): New function.
7927 (aarch64_expand_vec_perm_const_1): Update after rename of
7928 aarch64_evpc_rev. Handle SVE permutes too, trying
7929 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7930 than aarch64_evpc_tbl.
7931 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7932 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7933 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7934 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7935 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7936 (aarch64_expand_sve_vcond): New functions.
7937 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7938 of aarch64_vector_mode_p.
7939 (aarch64_dwarf_poly_indeterminate_value): New function.
7940 (aarch64_compute_pressure_classes): Likewise.
7941 (aarch64_can_change_mode_class): Likewise.
7942 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7943 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7944 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7945 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7946 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7947 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7948 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7949 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7950 constraints.
7951 (Dn, Dl, Dr): Accept const as well as const_vector.
7952 (Dz): Likewise. Compare against CONST0_RTX.
7953 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7954 of "vector" where appropriate.
7955 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7956 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7957 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7958 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7959 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7960 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7961 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7962 (v_int_equiv): Extend to SVE modes.
7963 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7964 mode attributes.
7965 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7966 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7967 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7968 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7969 (SVE_COND_FP_CMP): New int iterators.
7970 (perm_hilo): Handle the new unpack unspecs.
7971 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7972 attributes.
7973 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7974 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7975 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7976 (aarch64_equality_operator, aarch64_constant_vector_operand)
7977 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7978 (aarch64_sve_nonimmediate_operand): Likewise.
7979 (aarch64_sve_general_operand): Likewise.
7980 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7981 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7982 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7983 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7984 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7985 (aarch64_sve_float_arith_immediate): Likewise.
7986 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7987 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7988 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7989 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7990 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7991 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7992 (aarch64_sve_float_arith_operand): Likewise.
7993 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7994 (aarch64_sve_float_mul_operand): Likewise.
7995 (aarch64_sve_vec_perm_operand): Likewise.
7996 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7997 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7998 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7999 as well as const_vector.
8000 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8001 in file. Use CONST0_RTX and CONSTM1_RTX.
8002 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8003 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8004 Use aarch64_simd_imm_zero.
8005 * config/aarch64/aarch64-sve.md: New file.
8006 * config/aarch64/aarch64.md: Include it.
8007 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8008 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8009 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8010 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8011 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8012 (sve): New attribute.
8013 (enabled): Disable instructions with the sve attribute unless
8014 TARGET_SVE.
8015 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8016 aarch64_expand_mov_immediate.
8017 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8018 CNT[BHSD] immediates.
8019 (movti): Split CONST_POLY_INT moves into two halves.
8020 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8021 Split additions that need a temporary here if the destination
8022 is the stack pointer.
8023 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8024 (*add<mode>3_poly_1): New instruction.
8025 (set_clobber_cc): New expander.
8027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8029 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8030 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8031 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8032 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8033 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8034 Change innermode from fixed_mode_size to machine_mode.
8035 (simplify_subreg): Update call accordingly. Handle a constant-sized
8036 subreg of a variable-length CONST_VECTOR.
8038 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8039 Alan Hayward <alan.hayward@arm.com>
8040 David Sherwood <david.sherwood@arm.com>
8042 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8043 (add_offset_to_base): New function, split out from...
8044 (create_mem_ref): ...here. When handling a scale other than 1,
8045 check first whether the address is valid without the offset.
8046 Add it into the base if so, leaving the index and scale as-is.
8048 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8050 PR c++/83778
8051 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8052 fold_for_warn before checking if arg2 is INTEGER_CST.
8054 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8056 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8057 (store_multiple_operation): Delete.
8058 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8059 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8060 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8061 guarded by TARGET_STRING.
8062 (rs6000_output_load_multiple): Delete.
8063 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8064 OPTION_MASK_STRING / TARGET_STRING handling.
8065 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8066 (const rs6000_opt_masks) <"string">: Change mask to 0.
8067 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8068 (MASK_STRING): Delete.
8069 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8070 parts. Simplify.
8071 (load_multiple): Delete.
8072 (*ldmsi8): Delete.
8073 (*ldmsi7): Delete.
8074 (*ldmsi6): Delete.
8075 (*ldmsi5): Delete.
8076 (*ldmsi4): Delete.
8077 (*ldmsi3): Delete.
8078 (store_multiple): Delete.
8079 (*stmsi8): Delete.
8080 (*stmsi7): Delete.
8081 (*stmsi6): Delete.
8082 (*stmsi5): Delete.
8083 (*stmsi4): Delete.
8084 (*stmsi3): Delete.
8085 (movmemsi_8reg): Delete.
8086 (corresponding unnamed define_insn): Delete.
8087 (movmemsi_6reg): Delete.
8088 (corresponding unnamed define_insn): Delete.
8089 (movmemsi_4reg): Delete.
8090 (corresponding unnamed define_insn): Delete.
8091 (movmemsi_2reg): Delete.
8092 (corresponding unnamed define_insn): Delete.
8093 (movmemsi_1reg): Delete.
8094 (corresponding unnamed define_insn): Delete.
8095 * config/rs6000/rs6000.opt (mno-string): New.
8096 (mstring): Replace by deprecation warning stub.
8097 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8099 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8101 * regrename.c (regrename_do_replace): If replacing the same
8102 reg multiple times, try to reuse last created gen_raw_REG.
8104 PR debug/81155
8105 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8106 main to workaround a bug in GDB.
8108 2018-01-12 Tom de Vries <tom@codesourcery.com>
8110 PR target/83737
8111 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8113 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8115 PR rtl-optimization/80481
8116 * ira-color.c (get_cap_member): New function.
8117 (allocnos_conflict_by_live_ranges_p): Use it.
8118 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8119 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8121 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8123 PR target/83628
8124 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8125 (*saddl_se_1): Ditto.
8126 (*ssubsi_1): Ditto.
8127 (*ssubl_se_1): Ditto.
8129 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8131 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8132 rather than wi::to_widest for DR_INITs.
8133 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8134 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8135 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8136 INTEGER_CSTs.
8137 (vect_analyze_group_access_1): Note that here.
8139 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8141 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8142 polynomial type sizes.
8144 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8146 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8147 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8148 (gimple_add_tmp_var): Likewise.
8150 2018-01-12 Martin Liska <mliska@suse.cz>
8152 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8153 (gimple_alloc_sizes): Likewise.
8154 (dump_gimple_statistics): Use PRIu64 in printf format.
8155 * gimple.h: Change uint64_t to int.
8157 2018-01-12 Martin Liska <mliska@suse.cz>
8159 * tree-core.h: Use uint64_t instead of int.
8160 * tree.c (tree_node_counts): Likewise.
8161 (tree_node_sizes): Likewise.
8162 (dump_tree_statistics): Use PRIu64 in printf format.
8164 2018-01-12 Martin Liska <mliska@suse.cz>
8166 * Makefile.in: As qsort_chk is implemented in vec.c, add
8167 vec.o to linkage of gencfn-macros.
8168 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8169 passing the info to record_node_allocation_statistics.
8170 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8171 and pass the info.
8172 * ggc-common.c (struct ggc_usage): Add operator== and use
8173 it in operator< and compare function.
8174 * mem-stats.h (struct mem_usage): Likewise.
8175 * vec.c (struct vec_usage): Remove operator< and compare
8176 function. Can be simply inherited.
8178 2018-01-12 Martin Jambor <mjambor@suse.cz>
8180 PR target/81616
8181 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8182 * tree-ssa-math-opts.c: Include domwalk.h.
8183 (convert_mult_to_fma_1): New function.
8184 (fma_transformation_info): New type.
8185 (fma_deferring_state): Likewise.
8186 (cancel_fma_deferring): New function.
8187 (result_of_phi): Likewise.
8188 (last_fma_candidate_feeds_initial_phi): Likewise.
8189 (convert_mult_to_fma): Added deferring logic, split actual
8190 transformation to convert_mult_to_fma_1.
8191 (math_opts_dom_walker): New type.
8192 (math_opts_dom_walker::after_dom_children): New method, body moved
8193 here from pass_optimize_widening_mul::execute, added deferring logic
8194 bits.
8195 (pass_optimize_widening_mul::execute): Moved most of code to
8196 math_opts_dom_walker::after_dom_children.
8197 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8198 * config/i386/i386.c (ix86_option_override_internal): Added
8199 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8201 2018-01-12 Richard Biener <rguenther@suse.de>
8203 PR debug/83157
8204 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8205 inline instance vars.
8207 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8209 PR target/81819
8210 * config/rx/rx.c (rx_is_restricted_memory_address):
8211 Handle SUBREG case.
8213 2018-01-12 Richard Biener <rguenther@suse.de>
8215 PR tree-optimization/80846
8216 * target.def (split_reduction): New target hook.
8217 * targhooks.c (default_split_reduction): New function.
8218 * targhooks.h (default_split_reduction): Declare.
8219 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8220 target requests first reduce vectors by combining low and high
8221 parts.
8222 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8223 (get_vectype_for_scalar_type_and_size): Export.
8224 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8225 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8226 * doc/tm.texi: Regenerate.
8227 * config/i386/i386.c (ix86_split_reduction): Implement
8228 TARGET_VECTORIZE_SPLIT_REDUCTION.
8230 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8232 PR target/83368
8233 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8234 in PIC mode except for TARGET_VXWORKS_RTP.
8235 * config/sparc/sparc.c: Include cfgrtl.h.
8236 (TARGET_INIT_PIC_REG): Define.
8237 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8238 (sparc_pic_register_p): New predicate.
8239 (sparc_legitimate_address_p): Use it.
8240 (sparc_legitimize_pic_address): Likewise.
8241 (sparc_delegitimize_address): Likewise.
8242 (sparc_mode_dependent_address_p): Likewise.
8243 (gen_load_pcrel_sym): Remove 4th parameter.
8244 (load_got_register): Adjust call to above. Remove obsolete stuff.
8245 (sparc_expand_prologue): Do not call load_got_register here.
8246 (sparc_flat_expand_prologue): Likewise.
8247 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8248 (sparc_use_pseudo_pic_reg): New function.
8249 (sparc_init_pic_reg): Likewise.
8250 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8251 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8253 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8255 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8256 Add item for branch_cost.
8258 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8260 PR rtl-optimization/83565
8261 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8262 not extend the result to a larger mode for rotate operations.
8263 (num_sign_bit_copies1): Likewise.
8265 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8267 PR target/40411
8268 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8269 -symbolic.
8270 Use values-Xc.o for -pedantic.
8271 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8273 2018-01-12 Martin Liska <mliska@suse.cz>
8275 PR ipa/83054
8276 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8277 New function.
8278 (possible_polymorphic_call_targets): Use it.
8279 (ipa_devirt): Likewise.
8281 2018-01-12 Martin Liska <mliska@suse.cz>
8283 * profile-count.h (enum profile_quality): Use 0 as invalid
8284 enum value of profile_quality.
8286 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8288 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8289 -mext-string options.
8291 2018-01-12 Richard Biener <rguenther@suse.de>
8293 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8294 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8295 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8296 Likewise.
8297 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8299 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8301 * configure.ac (--with-long-double-format): Add support for the
8302 configuration option to change the default long double format on
8303 PowerPC systems.
8304 * config.gcc (powerpc*-linux*-*): Likewise.
8305 * configure: Regenerate.
8306 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8307 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8308 used without modification.
8310 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8312 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8313 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8314 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8315 MISC_BUILTIN_SPEC_BARRIER.
8316 (rs6000_init_builtins): Likewise.
8317 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8318 enum value.
8319 (speculation_barrier): New define_insn.
8320 * doc/extend.texi: Document __builtin_speculation_barrier.
8322 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8324 PR target/83203
8325 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8326 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8327 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8328 iterators.
8329 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8330 integral modes instead of "ss" and "sd".
8331 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8332 vectors with 32-bit and 64-bit elements.
8333 (vecdupssescalarmodesuffix): New mode attribute.
8334 (vec_dup<mode>): Use it.
8336 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8338 PR target/83330
8339 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8340 frame if argument is passed on stack.
8342 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8344 PR target/82682
8345 * ree.c (combine_reaching_defs): Optimize also
8346 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8347 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8349 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8351 PR middle-end/83189
8352 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8354 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8356 PR middle-end/83718
8357 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8358 after they are computed.
8360 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8362 PR tree-optimization/83695
8363 * gimple-loop-linterchange.cc
8364 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8365 reset cached scev information after interchange.
8366 (pass_linterchange::execute): Remove call to scev_reset_htab.
8368 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8370 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8371 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8372 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8373 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8374 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8375 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8376 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8377 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8378 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8379 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8380 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8381 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8382 (V_lane_reg): Likewise.
8383 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8384 New define_expand.
8385 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8386 (vfmal_lane_low<mode>_intrinsic,
8387 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8388 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8389 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8390 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8391 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8392 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8394 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8396 * config/arm/arm-cpus.in (fp16fml): New feature.
8397 (ALL_SIMD): Add fp16fml.
8398 (armv8.2-a): Add fp16fml as an option.
8399 (armv8.3-a): Likewise.
8400 (armv8.4-a): Add fp16fml as part of fp16.
8401 * config/arm/arm.h (TARGET_FP16FML): Define.
8402 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8403 when appropriate.
8404 * config/arm/arm-modes.def (V2HF): Define.
8405 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8406 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8407 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8408 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8409 vfmsl_low, vfmsl_high): New set of builtins.
8410 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8411 (vfml_op): New code attribute.
8412 (VFMLHALVES): New int iterator.
8413 (VFML, VFMLSEL): New mode attributes.
8414 (V_reg): Define mapping for V2HF.
8415 (V_hi, V_lo): New mode attributes.
8416 (VF_constraint): Likewise.
8417 (vfml_half, vfml_half_selector): New int attributes.
8418 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8419 define_expand.
8420 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8421 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8422 New define_insn.
8423 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8424 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8425 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8426 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8427 documentation.
8428 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8429 Document new effective target and option set.
8431 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8433 * config/arm/arm-cpus.in (armv8_4): New feature.
8434 (ARMv8_4a): New fgroup.
8435 (armv8.4-a): New arch.
8436 * config/arm/arm-tables.opt: Regenerate.
8437 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8438 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8439 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8440 Add matching rules for -march=armv8.4-a and extensions.
8441 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8443 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8445 PR target/81821
8446 * config/rx/rx.md (BW): New mode attribute.
8447 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8449 2018-01-11 Richard Biener <rguenther@suse.de>
8451 PR tree-optimization/83435
8452 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8453 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8454 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8456 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8457 Alan Hayward <alan.hayward@arm.com>
8458 David Sherwood <david.sherwood@arm.com>
8460 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8461 field.
8462 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8463 (aarch64_print_address_internal): Use it to check for a zero offset.
8465 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8466 Alan Hayward <alan.hayward@arm.com>
8467 David Sherwood <david.sherwood@arm.com>
8469 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8470 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8471 Return a poly_int64 rather than a HOST_WIDE_INT.
8472 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8473 rather than a HOST_WIDE_INT.
8474 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8475 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8476 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8477 final_offset from HOST_WIDE_INT to poly_int64.
8478 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8479 to_constant when getting the number of units in an Advanced SIMD
8480 mode.
8481 (aarch64_builtin_vectorized_function): Check for a constant number
8482 of units.
8483 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8484 GET_MODE_SIZE.
8485 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8486 attribute instead of GET_MODE_NUNITS.
8487 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8488 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8489 GET_MODE_SIZE for fixed-size registers.
8490 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8491 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8492 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8493 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8494 (aarch64_print_operand, aarch64_print_address_internal)
8495 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8496 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8497 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8498 Handle polynomial GET_MODE_SIZE.
8499 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8500 wider than SImode without modification.
8501 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8502 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8503 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8504 passing and returning SVE modes.
8505 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8506 rather than GEN_INT.
8507 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8508 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8509 (aarch64_allocate_and_probe_stack_space): Likewise.
8510 (aarch64_layout_frame): Cope with polynomial offsets.
8511 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8512 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8513 polynomial offsets.
8514 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8515 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8516 poly_int64 rather than a HOST_WIDE_INT.
8517 (aarch64_get_separate_components, aarch64_process_components)
8518 (aarch64_expand_prologue, aarch64_expand_epilogue)
8519 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8520 (aarch64_anchor_offset): New function, split out from...
8521 (aarch64_legitimize_address): ...here.
8522 (aarch64_builtin_vectorization_cost): Handle polynomial
8523 TYPE_VECTOR_SUBPARTS.
8524 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8525 GET_MODE_NUNITS.
8526 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8527 number of elements from the PARALLEL rather than the mode.
8528 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8529 rather than GET_MODE_BITSIZE.
8530 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8531 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8532 (aarch64_expand_vec_perm_const_1): Handle polynomial
8533 d->perm.length () and d->perm elements.
8534 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8535 Apply to_constant to d->perm elements.
8536 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8537 polynomial CONST_VECTOR_NUNITS.
8538 (aarch64_move_pointer): Take amount as a poly_int64 rather
8539 than an int.
8540 (aarch64_progress_pointer): Avoid temporary variable.
8541 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8542 the mode attribute instead of GET_MODE.
8544 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8545 Alan Hayward <alan.hayward@arm.com>
8546 David Sherwood <david.sherwood@arm.com>
8548 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8549 x exists before using it.
8550 (aarch64_add_constant_internal): Rename to...
8551 (aarch64_add_offset_1): ...this. Replace regnum with separate
8552 src and dest rtxes. Handle the case in which they're different,
8553 including when the offset is zero. Replace scratchreg with an rtx.
8554 Use 2 additions if there is no spare register into which we can
8555 move a 16-bit constant.
8556 (aarch64_add_constant): Delete.
8557 (aarch64_add_offset): Replace reg with separate src and dest
8558 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8559 Use aarch64_add_offset_1.
8560 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8561 an rtx rather than an int. Take the delta as a poly_int64
8562 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8563 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8564 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8565 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8566 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8567 and aarch64_add_sp.
8568 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8569 aarch64_add_constant.
8571 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8573 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8574 Use scalar_float_mode.
8576 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8578 * config/aarch64/aarch64-simd.md
8579 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8580 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8581 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8582 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8583 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8584 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8585 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8586 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8587 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8588 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8590 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8592 PR target/83514
8593 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8594 targ_options->x_arm_arch_string is non NULL.
8596 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8598 * config/aarch64/aarch64.h
8599 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8601 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8603 PR target/82096
8604 * expmed.c (emit_store_flag_force): Swap if const op0
8605 and change VOIDmode to mode of op0.
8607 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8609 PR rtl-optimization/83761
8610 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8611 than bytes to mode_for_size.
8613 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8615 PR middle-end/83189
8616 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8617 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8618 profile.
8620 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8622 PR middle-end/83575
8623 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8624 when in layout mode.
8625 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8626 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8627 partition fixup.
8629 2018-01-10 Michael Collison <michael.collison@arm.com>
8631 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8632 * config/aarch64/aarch64-option-extension.def: Add
8633 AARCH64_OPT_EXTENSION of 'fp16fml'.
8634 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8635 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8636 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8637 * config/aarch64/constraints.md (Ui7): New constraint.
8638 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8639 (VFMLA_SEL_W): Ditto.
8640 (f16quad): Ditto.
8641 (f16mac1): Ditto.
8642 (VFMLA16_LOW): New int iterator.
8643 (VFMLA16_HIGH): Ditto.
8644 (UNSPEC_FMLAL): New unspec.
8645 (UNSPEC_FMLSL): Ditto.
8646 (UNSPEC_FMLAL2): Ditto.
8647 (UNSPEC_FMLSL2): Ditto.
8648 (f16mac): New code attribute.
8649 * config/aarch64/aarch64-simd-builtins.def
8650 (aarch64_fmlal_lowv2sf): Ditto.
8651 (aarch64_fmlsl_lowv2sf): Ditto.
8652 (aarch64_fmlalq_lowv4sf): Ditto.
8653 (aarch64_fmlslq_lowv4sf): Ditto.
8654 (aarch64_fmlal_highv2sf): Ditto.
8655 (aarch64_fmlsl_highv2sf): Ditto.
8656 (aarch64_fmlalq_highv4sf): Ditto.
8657 (aarch64_fmlslq_highv4sf): Ditto.
8658 (aarch64_fmlal_lane_lowv2sf): Ditto.
8659 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8660 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8661 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8662 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8663 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8664 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8665 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8666 (aarch64_fmlal_lane_highv2sf): Ditto.
8667 (aarch64_fmlsl_lane_highv2sf): Ditto.
8668 (aarch64_fmlal_laneq_highv2sf): Ditto.
8669 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8670 (aarch64_fmlalq_lane_highv4sf): Ditto.
8671 (aarch64_fmlsl_lane_highv4sf): Ditto.
8672 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8673 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8674 * config/aarch64/aarch64-simd.md:
8675 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8676 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8677 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8678 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8679 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8680 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8681 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8682 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8683 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8684 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8685 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8686 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8687 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8688 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8689 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8690 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8691 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8692 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8693 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8694 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8695 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8696 (vfmlsl_low_u32): Ditto.
8697 (vfmlalq_low_u32): Ditto.
8698 (vfmlslq_low_u32): Ditto.
8699 (vfmlal_high_u32): Ditto.
8700 (vfmlsl_high_u32): Ditto.
8701 (vfmlalq_high_u32): Ditto.
8702 (vfmlslq_high_u32): Ditto.
8703 (vfmlal_lane_low_u32): Ditto.
8704 (vfmlsl_lane_low_u32): Ditto.
8705 (vfmlal_laneq_low_u32): Ditto.
8706 (vfmlsl_laneq_low_u32): Ditto.
8707 (vfmlalq_lane_low_u32): Ditto.
8708 (vfmlslq_lane_low_u32): Ditto.
8709 (vfmlalq_laneq_low_u32): Ditto.
8710 (vfmlslq_laneq_low_u32): Ditto.
8711 (vfmlal_lane_high_u32): Ditto.
8712 (vfmlsl_lane_high_u32): Ditto.
8713 (vfmlal_laneq_high_u32): Ditto.
8714 (vfmlsl_laneq_high_u32): Ditto.
8715 (vfmlalq_lane_high_u32): Ditto.
8716 (vfmlslq_lane_high_u32): Ditto.
8717 (vfmlalq_laneq_high_u32): Ditto.
8718 (vfmlslq_laneq_high_u32): Ditto.
8719 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8720 (AARCH64_FL_FOR_ARCH8_4): New.
8721 (AARCH64_ISA_F16FML): New ISA flag.
8722 (TARGET_F16FML): New feature flag for fp16fml.
8723 (doc/invoke.texi): Document new fp16fml option.
8725 2018-01-10 Michael Collison <michael.collison@arm.com>
8727 * config/aarch64/aarch64-builtins.c:
8728 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8729 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8730 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8731 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8732 (AARCH64_ISA_SHA3): New ISA flag.
8733 (TARGET_SHA3): New feature flag for sha3.
8734 * config/aarch64/iterators.md (sha512_op): New int attribute.
8735 (CRYPTO_SHA512): New int iterator.
8736 (UNSPEC_SHA512H): New unspec.
8737 (UNSPEC_SHA512H2): Ditto.
8738 (UNSPEC_SHA512SU0): Ditto.
8739 (UNSPEC_SHA512SU1): Ditto.
8740 * config/aarch64/aarch64-simd-builtins.def
8741 (aarch64_crypto_sha512hqv2di): New builtin.
8742 (aarch64_crypto_sha512h2qv2di): Ditto.
8743 (aarch64_crypto_sha512su0qv2di): Ditto.
8744 (aarch64_crypto_sha512su1qv2di): Ditto.
8745 (aarch64_eor3qv8hi): Ditto.
8746 (aarch64_rax1qv2di): Ditto.
8747 (aarch64_xarqv2di): Ditto.
8748 (aarch64_bcaxqv8hi): Ditto.
8749 * config/aarch64/aarch64-simd.md:
8750 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8751 (aarch64_crypto_sha512su0qv2di): Ditto.
8752 (aarch64_crypto_sha512su1qv2di): Ditto.
8753 (aarch64_eor3qv8hi): Ditto.
8754 (aarch64_rax1qv2di): Ditto.
8755 (aarch64_xarqv2di): Ditto.
8756 (aarch64_bcaxqv8hi): Ditto.
8757 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8758 (vsha512h2q_u64): Ditto.
8759 (vsha512su0q_u64): Ditto.
8760 (vsha512su1q_u64): Ditto.
8761 (veor3q_u16): Ditto.
8762 (vrax1q_u64): Ditto.
8763 (vxarq_u64): Ditto.
8764 (vbcaxq_u16): Ditto.
8765 * config/arm/types.md (crypto_sha512): New type attribute.
8766 (crypto_sha3): Ditto.
8767 (doc/invoke.texi): Document new sha3 option.
8769 2018-01-10 Michael Collison <michael.collison@arm.com>
8771 * config/aarch64/aarch64-builtins.c:
8772 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8773 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8774 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8775 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8776 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8777 (AARCH64_ISA_SM4): New ISA flag.
8778 (TARGET_SM4): New feature flag for sm4.
8779 * config/aarch64/aarch64-simd-builtins.def
8780 (aarch64_sm3ss1qv4si): Ditto.
8781 (aarch64_sm3tt1aq4si): Ditto.
8782 (aarch64_sm3tt1bq4si): Ditto.
8783 (aarch64_sm3tt2aq4si): Ditto.
8784 (aarch64_sm3tt2bq4si): Ditto.
8785 (aarch64_sm3partw1qv4si): Ditto.
8786 (aarch64_sm3partw2qv4si): Ditto.
8787 (aarch64_sm4eqv4si): Ditto.
8788 (aarch64_sm4ekeyqv4si): Ditto.
8789 * config/aarch64/aarch64-simd.md:
8790 (aarch64_sm3ss1qv4si): Ditto.
8791 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8792 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8793 (aarch64_sm4eqv4si): Ditto.
8794 (aarch64_sm4ekeyqv4si): Ditto.
8795 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8796 (sm3part_op): Ditto.
8797 (CRYPTO_SM3TT): Ditto.
8798 (CRYPTO_SM3PART): Ditto.
8799 (UNSPEC_SM3SS1): New unspec.
8800 (UNSPEC_SM3TT1A): Ditto.
8801 (UNSPEC_SM3TT1B): Ditto.
8802 (UNSPEC_SM3TT2A): Ditto.
8803 (UNSPEC_SM3TT2B): Ditto.
8804 (UNSPEC_SM3PARTW1): Ditto.
8805 (UNSPEC_SM3PARTW2): Ditto.
8806 (UNSPEC_SM4E): Ditto.
8807 (UNSPEC_SM4EKEY): Ditto.
8808 * config/aarch64/constraints.md (Ui2): New constraint.
8809 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8810 * config/arm/types.md (crypto_sm3): New type attribute.
8811 (crypto_sm4): Ditto.
8812 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8813 (vsm3tt1aq_u32): Ditto.
8814 (vsm3tt1bq_u32): Ditto.
8815 (vsm3tt2aq_u32): Ditto.
8816 (vsm3tt2bq_u32): Ditto.
8817 (vsm3partw1q_u32): Ditto.
8818 (vsm3partw2q_u32): Ditto.
8819 (vsm4eq_u32): Ditto.
8820 (vsm4ekeyq_u32): Ditto.
8821 (doc/invoke.texi): Document new sm4 option.
8823 2018-01-10 Michael Collison <michael.collison@arm.com>
8825 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8826 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8827 (AARCH64_FL_FOR_ARCH8_4): New.
8828 (AARCH64_FL_V8_4): New flag.
8829 (doc/invoke.texi): Document new armv8.4-a option.
8831 2018-01-10 Michael Collison <michael.collison@arm.com>
8833 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8834 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8835 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8836 * config/aarch64/aarch64-option-extension.def: Add
8837 AARCH64_OPT_EXTENSION of 'sha2'.
8838 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8839 (crypto): Disable sha2 and aes if crypto disabled.
8840 (crypto): Enable aes and sha2 if enabled.
8841 (simd): Disable sha2 and aes if simd disabled.
8842 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8843 New flags.
8844 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8845 (TARGET_SHA2): New feature flag for sha2.
8846 (TARGET_AES): New feature flag for aes.
8847 * config/aarch64/aarch64-simd.md:
8848 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8849 conditional on TARGET_AES.
8850 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8851 (aarch64_crypto_sha1hsi): Make pattern conditional
8852 on TARGET_SHA2.
8853 (aarch64_crypto_sha1hv4si): Ditto.
8854 (aarch64_be_crypto_sha1hv4si): Ditto.
8855 (aarch64_crypto_sha1su1v4si): Ditto.
8856 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8857 (aarch64_crypto_sha1su0v4si): Ditto.
8858 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8859 (aarch64_crypto_sha256su0v4si): Ditto.
8860 (aarch64_crypto_sha256su1v4si): Ditto.
8861 (doc/invoke.texi): Document new aes and sha2 options.
8863 2018-01-10 Martin Sebor <msebor@redhat.com>
8865 PR tree-optimization/83781
8866 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8867 as string arrays.
8869 2018-01-11 Martin Sebor <msebor@gmail.com>
8870 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8872 PR tree-optimization/83501
8873 PR tree-optimization/81703
8875 * tree-ssa-strlen.c (get_string_cst): Rename...
8876 (get_string_len): ...to this. Handle global constants.
8877 (handle_char_store): Adjust.
8879 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8880 Jim Wilson <jimw@sifive.com>
8882 * config/riscv/riscv-protos.h (riscv_output_return): New.
8883 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8884 (riscv_attribute_table, riscv_output_return),
8885 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8886 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8887 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8888 function.
8889 (riscv_expand_prologue): Add early return for naked function.
8890 (riscv_expand_epilogue): Likewise.
8891 (riscv_function_ok_for_sibcall): Return false for naked function.
8892 (riscv_set_current_function): New.
8893 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8894 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8895 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8896 * doc/extend.texi (RISC-V Function Attributes): New.
8898 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8900 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8901 check for 128-bit long double before checking TCmode.
8902 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8903 128-bit long doubles before checking TFmode or TCmode.
8904 (FLOAT128_IBM_P): Likewise.
8906 2018-01-10 Martin Sebor <msebor@redhat.com>
8908 PR tree-optimization/83671
8909 * builtins.c (c_strlen): Unconditionally return zero for the empty
8910 string.
8911 Use -Warray-bounds for warnings.
8912 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8913 for non-constant array indices with COMPONENT_REF, arrays of
8914 arrays, and pointers to arrays.
8915 (gimple_fold_builtin_strlen): Determine and set length range for
8916 non-constant character arrays.
8918 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8920 PR middle-end/81897
8921 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8922 empty blocks.
8924 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8926 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8928 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8930 PR target/83399
8931 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8932 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8933 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8934 indexed_or_indirect_operand predicate.
8935 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8936 (*vsx_le_perm_load_v8hi): Likewise.
8937 (*vsx_le_perm_load_v16qi): Likewise.
8938 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8939 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8940 (*vsx_le_perm_store_v8hi): Likewise.
8941 (*vsx_le_perm_store_v16qi): Likewise.
8942 (eight unnamed splitters): Likewise.
8944 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8946 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8947 * config/rs6000/emmintrin.h: Likewise.
8948 * config/rs6000/mmintrin.h: Likewise.
8949 * config/rs6000/xmmintrin.h: Likewise.
8951 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8953 PR c++/43486
8954 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8955 "public_flag".
8956 * tree.c (tree_nop_conversion): Return true for location wrapper
8957 nodes.
8958 (maybe_wrap_with_location): New function.
8959 (selftest::check_strip_nops): New function.
8960 (selftest::test_location_wrappers): New function.
8961 (selftest::tree_c_tests): Call it.
8962 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8963 (maybe_wrap_with_location): New decl.
8964 (EXPR_LOCATION_WRAPPER_P): New macro.
8965 (location_wrapper_p): New inline function.
8966 (tree_strip_any_location_wrapper): New inline function.
8968 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8970 PR target/83735
8971 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8972 stack_realign_offset for the largest alignment of stack slot
8973 actually used.
8974 (ix86_find_max_used_stack_alignment): New function.
8975 (ix86_finalize_stack_frame_flags): Use it. Set
8976 max_used_stack_alignment if we don't realign stack.
8977 * config/i386/i386.h (machine_function): Add
8978 max_used_stack_alignment.
8980 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8982 * config/arm/arm.opt (-mbranch-cost): New option.
8983 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8984 account.
8986 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8988 PR target/83629
8989 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8990 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8992 2018-01-10 Richard Biener <rguenther@suse.de>
8994 PR debug/83765
8995 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8996 early out so it also covers the case where we have a non-NULL
8997 origin.
8999 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9001 PR tree-optimization/83753
9002 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9003 for non-strided grouped accesses if the number of elements is 1.
9005 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9007 PR target/81616
9008 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9009 * i386.h (TARGET_USE_GATHER): Define.
9010 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9012 2018-01-10 Martin Liska <mliska@suse.cz>
9014 PR bootstrap/82831
9015 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9016 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9017 partitioning.
9018 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9019 CLEANUP_NO_PARTITIONING is not set.
9021 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9023 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9024 for vectors, as a partial revert of r254296.
9025 * rtl.h (const_vec_p): Delete.
9026 (const_vec_duplicate_p): Don't test for vector CONSTs.
9027 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9028 * expmed.c (make_tree): Likewise.
9030 Revert:
9031 * common.md (E, F): Use CONSTANT_P instead of checking for
9032 CONST_VECTOR.
9033 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9034 checking for CONST_VECTOR.
9036 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9038 PR middle-end/83575
9039 * predict.c (force_edge_cold): Handle in more sane way edges
9040 with no prediction.
9042 2018-01-09 Carl Love <cel@us.ibm.com>
9044 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9045 V4SI, V4SF types.
9046 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9047 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9048 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9049 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9050 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9051 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9052 * config/rs6000/rs6000-protos.h: Add extern defition for
9053 rs6000_generate_float2_double_code.
9054 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9055 function.
9056 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9057 (float2_v2df): Add define_expand.
9059 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9061 PR target/83628
9062 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9063 op_mode in the force_to_mode call.
9065 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9067 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9068 instead of checking each element individually.
9069 (aarch64_evpc_uzp): Likewise.
9070 (aarch64_evpc_zip): Likewise.
9071 (aarch64_evpc_ext): Likewise.
9072 (aarch64_evpc_rev): Likewise.
9073 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9074 instead of checking each element individually. Return true without
9075 generating rtl if
9076 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9077 whether all selected elements come from the same input, instead of
9078 checking each element individually. Remove calls to gen_rtx_REG,
9079 start_sequence and end_sequence and instead assert that no rtl is
9080 generated.
9082 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9084 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9085 order of HIGH and CONST checks.
9087 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9089 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9090 if the destination isn't an SSA_NAME.
9092 2018-01-09 Richard Biener <rguenther@suse.de>
9094 PR tree-optimization/83668
9095 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9096 move prologue...
9097 (canonicalize_loop_form): ... here, renamed from ...
9098 (canonicalize_loop_closed_ssa_form): ... this and amended to
9099 swap successor edges for loop exit blocks to make us use
9100 the RPO order we need for initial schedule generation.
9102 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9104 PR tree-optimization/64811
9105 * match.pd: When optimizing comparisons with Inf, avoid
9106 introducing or losing exceptions from comparisons with NaN.
9108 2018-01-09 Martin Liska <mliska@suse.cz>
9110 PR sanitizer/82517
9111 * asan.c (shadow_mem_size): Add gcc_assert.
9113 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9115 Don't save registers in main().
9117 PR target/83738
9118 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9119 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9120 * config/avr/avr.c (avr_set_current_function): Don't error if
9121 naked, OS_task or OS_main are specified at the same time.
9122 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9123 OS_main.
9124 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9125 attribute.
9126 * common/config/avr/avr-common.c (avr_option_optimization_table):
9127 Switch on -mmain-is-OS_task for optimizing compilations.
9129 2018-01-09 Richard Biener <rguenther@suse.de>
9131 PR tree-optimization/83572
9132 * graphite.c: Include cfganal.h.
9133 (graphite_transform_loops): Connect infinite loops to exit
9134 and remove fake edges at the end.
9136 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9138 * ipa-inline.c (edge_badness): Revert accidental checkin.
9140 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9142 PR ipa/80763
9143 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9144 symbols; not inline clones.
9146 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9148 PR target/83507
9149 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9150 hard registers. Formatting fixes.
9152 PR preprocessor/83722
9153 * gcc.c (try_generate_repro): Pass
9154 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9155 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9156 do_report_bug.
9158 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9159 Kito Cheng <kito.cheng@gmail.com>
9161 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9162 (riscv_leaf_function_p): Delete.
9163 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9165 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9167 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9168 function.
9169 (do_ifelse): New function.
9170 (do_isel): New function.
9171 (do_sub3): New function.
9172 (do_add3): New function.
9173 (do_load_mask_compare): New function.
9174 (do_overlap_load_compare): New function.
9175 (expand_compare_loop): New function.
9176 (expand_block_compare): Call expand_compare_loop() when appropriate.
9177 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9178 option description.
9179 (-mblock-compare-inline-loop-limit): New option.
9181 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9183 PR target/83677
9184 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9185 Reverse order of second and third operands in first alternative.
9186 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9187 of first and second elements in UNSPEC_VPERMR vector.
9188 (altivec_expand_vec_perm_le): Likewise.
9190 2018-01-08 Jeff Law <law@redhat.com>
9192 PR rtl-optimizatin/81308
9193 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9194 (process_switch): If group_case_labels makes a change, then set
9195 cfg_altered.
9196 (pass_convert_switch::execute): If a switch is converted, then
9197 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9199 PR rtl-optimization/81308
9200 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9201 splitting insns.
9203 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9205 PR target/83663 - Revert r255946
9206 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9207 generation for cases where splatting a value is not useful.
9208 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9209 across a vec_duplicate and a paradoxical subreg forming a vector
9210 mode to a vec_concat.
9212 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9214 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9215 -march=armv8.3-a variants.
9216 * config/arm/t-multilib: Likewise.
9217 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9219 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9221 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9222 to generate rtl.
9223 (cceq_ior_compare_complement): Give it a name so I can use it, and
9224 change boolean_or_operator predicate to boolean_operator so it can
9225 be used to generate a crand.
9226 (eqne): New code iterator.
9227 (bd/bd_neg): New code_attrs.
9228 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9229 a single define_insn.
9230 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9231 decrement (bdnzt/bdnzf/bdzt/bdzf).
9232 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9233 with the new names of the branch decrement patterns, and added the
9234 names of the branch decrement conditional patterns.
9236 2018-01-08 Richard Biener <rguenther@suse.de>
9238 PR tree-optimization/83563
9239 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9240 cache.
9242 2018-01-08 Richard Biener <rguenther@suse.de>
9244 PR middle-end/83713
9245 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9247 2018-01-08 Richard Biener <rguenther@suse.de>
9249 PR tree-optimization/83685
9250 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9251 references to abnormals.
9253 2018-01-08 Richard Biener <rguenther@suse.de>
9255 PR lto/83719
9256 * dwarf2out.c (output_indirect_strings): Handle empty
9257 skeleton_debug_str_hash.
9258 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9260 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9262 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9263 (emit_store_direct): Likewise.
9264 (arc_trampoline_adjust_address): Likewise.
9265 (arc_asm_trampoline_template): New function.
9266 (arc_initialize_trampoline): Use asm_trampoline_template.
9267 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9268 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9269 * config/arc/arc.md (flush_icache): Delete pattern.
9271 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9273 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9274 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9275 munaligned-access.
9277 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9279 PR target/83681
9280 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9281 by not USED_FOR_TARGET.
9282 (make_pass_resolve_sw_modes): Likewise.
9284 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9286 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9287 USED_FOR_TARGET.
9289 2018-01-08 Richard Biener <rguenther@suse.de>
9291 PR middle-end/83580
9292 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9294 2018-01-08 Richard Biener <rguenther@suse.de>
9296 PR middle-end/83517
9297 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9299 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9301 PR middle-end/81897
9302 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9303 basic blocks with a small number of successors.
9304 (convert_control_dep_chain_into_preds): Improve handling of
9305 forwarder blocks.
9306 (dump_predicates): Split apart into...
9307 (dump_pred_chain): ...here...
9308 (dump_pred_info): ...and here.
9309 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9310 (can_chain_union_be_invalidated_p): Improve check for invalidation
9311 of paths.
9312 (uninit_uses_cannot_happen): Avoid unnecessary if
9313 convert_control_dep_chain_into_preds yielded nothing.
9315 2018-01-06 Martin Sebor <msebor@redhat.com>
9317 PR tree-optimization/83640
9318 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9319 subtracting negative offset from size.
9320 (builtin_access::overlap): Adjust offset bounds of the access to fall
9321 within the size of the object if possible.
9323 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9325 PR rtl-optimization/83699
9326 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9327 extract_bit_field_as_subreg to cases in which the extracted
9328 value is also a vector.
9330 * lra-constraints.c (process_alt_operands): Test for the equivalence
9331 substitutions when detecting a possible reload cycle.
9333 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9335 PR debug/83480
9336 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9337 by default if flag_selective_schedling{,2}. Formatting fixes.
9339 PR rtl-optimization/83682
9340 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9341 if it has non-VECTOR_MODE element mode.
9342 (vec_duplicate_p): Likewise.
9344 PR middle-end/83694
9345 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9346 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9348 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9350 PR target/83604
9351 * config/i386/i386-builtin.def
9352 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9353 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9354 Require also OPTION_MASK_ISA_AVX512F in addition to
9355 OPTION_MASK_ISA_GFNI.
9356 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9357 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9358 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9359 to OPTION_MASK_ISA_GFNI.
9360 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9361 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9362 OPTION_MASK_ISA_AVX512BW.
9363 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9364 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9365 addition to OPTION_MASK_ISA_GFNI.
9366 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9367 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9368 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9369 to OPTION_MASK_ISA_GFNI.
9370 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9371 a requirement for all ISAs rather than any of them with a few
9372 exceptions.
9373 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9374 processing.
9375 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9376 bitmasks to be enabled with 3 exceptions, instead of requiring any
9377 enabled ISA with lots of exceptions.
9378 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9379 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9380 Change avx512bw in isa attribute to avx512f.
9381 * config/i386/sgxintrin.h: Add license boilerplate.
9382 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9383 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9384 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9385 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9386 defined.
9387 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9388 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9389 temporarily sse2 rather than sse if not enabled already.
9391 PR target/83604
9392 * config/i386/sse.md (VI248_VLBW): Rename to ...
9393 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9394 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9395 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9396 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9397 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9398 mode iterator instead of VI248_VLBW.
9400 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9402 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9403 (record_modified): Skip clobbers; add debug output.
9404 (param_change_prob): Use sreal frequencies.
9406 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9408 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9409 punt for user-aligned variables.
9411 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9413 * tree-chrec.c (chrec_contains_symbols): Return true for
9414 POLY_INT_CST.
9416 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9418 PR target/82439
9419 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9420 of (x|y) == x for BICS pattern.
9422 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9424 PR tree-optimization/83605
9425 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9426 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9427 can throw.
9429 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9431 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9432 * config/epiphany/rtems.h: New file.
9434 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9435 Uros Bizjak <ubizjak@gmail.com>
9437 PR target/83554
9438 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9439 QIreg_operand instead of register_operand predicate.
9440 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9441 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9442 comments instead of -fmitigate[-_]rop.
9444 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9446 PR bootstrap/81926
9447 * cgraphunit.c (symbol_table::compile): Switch to text_section
9448 before calling assembly_start debug hook.
9449 * run-rtl-passes.c (run_rtl_passes): Likewise.
9450 Include output.h.
9452 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9454 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9455 range_int_cst_p rather than !symbolic_range_p before calling
9456 extract_range_from_multiplicative_op_1.
9458 2018-01-04 Jeff Law <law@redhat.com>
9460 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9461 redundant test in assertion.
9463 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9465 * doc/rtl.texi: Document machine_mode wrapper classes.
9467 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9469 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9470 using tree_to_uhwi.
9472 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9474 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9475 the VEC_PERM_EXPR fold to fail.
9477 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9479 PR debug/83585
9480 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9481 to switched_sections.
9483 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9485 PR target/83680
9486 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9487 test for d.testing.
9489 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9491 PR target/83387
9492 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9493 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9495 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9497 PR debug/83666
9498 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9499 is BLKmode and bitpos not zero or mode change is needed.
9501 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9503 PR target/83675
9504 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9505 TARGET_VIS2.
9507 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9509 PR target/83628
9510 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9511 instead of MULT rtx. Update all corresponding splitters.
9512 (*saddl_se): Ditto.
9513 (*ssub<modesuffix>): Ditto.
9514 (*ssubl_se): Ditto.
9515 (*cmp_sadd_di): Update split patterns.
9516 (*cmp_sadd_si): Ditto.
9517 (*cmp_sadd_sidi): Ditto.
9518 (*cmp_ssub_di): Ditto.
9519 (*cmp_ssub_si): Ditto.
9520 (*cmp_ssub_sidi): Ditto.
9521 * config/alpha/predicates.md (const23_operand): New predicate.
9522 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9523 Look for ASHIFT, not MULT inner operand.
9524 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9526 2018-01-04 Martin Liska <mliska@suse.cz>
9528 PR gcov-profile/83669
9529 * gcov.c (output_intermediate_file): Add version to intermediate
9530 gcov file.
9531 * doc/gcov.texi: Document new field 'version' in intermediate
9532 file format. Fix location of '-k' option of gcov command.
9534 2018-01-04 Martin Liska <mliska@suse.cz>
9536 PR ipa/82352
9537 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9539 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9541 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9543 2018-01-03 Martin Sebor <msebor@redhat.com>
9545 PR tree-optimization/83655
9546 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9547 checking calls with invalid arguments.
9549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9551 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9552 (vectorizable_mask_load_store): Delete.
9553 (vectorizable_call): Return false for masked loads and stores.
9554 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9555 instead of gimple_assign_rhs1.
9556 (vectorizable_load): Handle IFN_MASK_LOAD.
9557 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9561 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9562 split out from..,
9563 (vectorizable_mask_load_store): ...here.
9564 (vectorizable_load): ...and here.
9566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9568 * tree-vect-stmts.c (vect_build_all_ones_mask)
9569 (vect_build_zero_merge_argument): New functions, split out from...
9570 (vectorizable_load): ...here.
9572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9574 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9575 split out from...
9576 (vectorizable_mask_load_store): ...here.
9577 (vectorizable_store): ...and here.
9579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9581 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9582 split out from...
9583 (vectorizable_mask_load_store): ...here.
9585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9587 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9588 (vect_model_store_cost): Take a vec_load_store_type instead of a
9589 vect_def_type.
9590 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9591 (vect_model_store_cost): Take a vec_load_store_type instead of a
9592 vect_def_type.
9593 (vectorizable_mask_load_store): Update accordingly.
9594 (vectorizable_store): Likewise.
9595 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9599 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9600 IFN_MASK_LOAD calls here rather than...
9601 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9604 Alan Hayward <alan.hayward@arm.com>
9605 David Sherwood <david.sherwood@arm.com>
9607 * expmed.c (extract_bit_field_1): For vector extracts,
9608 fall back to extract_bit_field_as_subreg if vec_extract
9609 isn't available.
9611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9612 Alan Hayward <alan.hayward@arm.com>
9613 David Sherwood <david.sherwood@arm.com>
9615 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9616 they are variable or constant sized.
9617 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9618 slots for constant-sized data.
9620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9621 Alan Hayward <alan.hayward@arm.com>
9622 David Sherwood <david.sherwood@arm.com>
9624 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9625 handling COND_EXPRs with boolean comparisons, try to find a better
9626 basis for the mask type than the boolean itself.
9628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9630 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9631 is calculated and how it can be overridden.
9632 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9633 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9634 if defined.
9635 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9636 if nonzero.
9638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9639 Alan Hayward <alan.hayward@arm.com>
9640 David Sherwood <david.sherwood@arm.com>
9642 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9643 Remove the mode argument.
9644 (aarch64_simd_valid_immediate): Remove the mode and inverse
9645 arguments.
9646 * config/aarch64/iterators.md (bitsize): New iterator.
9647 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9648 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9649 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9650 aarch64_simd_valid_immediate.
9651 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9652 (aarch64_reg_or_bic_imm): Likewise.
9653 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9654 with an insn_type enum and msl with a modifier_type enum.
9655 Replace element_width with a scalar_mode. Change the shift
9656 to unsigned int. Add constructors for scalar_float_mode and
9657 scalar_int_mode elements.
9658 (aarch64_vect_float_const_representable_p): Delete.
9659 (aarch64_can_const_movi_rtx_p)
9660 (aarch64_simd_scalar_immediate_valid_for_move)
9661 (aarch64_simd_make_constant): Update call to
9662 aarch64_simd_valid_immediate.
9663 (aarch64_advsimd_valid_immediate_hs): New function.
9664 (aarch64_advsimd_valid_immediate): Likewise.
9665 (aarch64_simd_valid_immediate): Remove mode and inverse
9666 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9667 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9668 and aarch64_float_const_representable_p on the result.
9669 (aarch64_output_simd_mov_immediate): Remove mode argument.
9670 Update call to aarch64_simd_valid_immediate and use of
9671 simd_immediate_info.
9672 (aarch64_output_scalar_simd_mov_immediate): Update call
9673 accordingly.
9675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9676 Alan Hayward <alan.hayward@arm.com>
9677 David Sherwood <david.sherwood@arm.com>
9679 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9680 (mode_nunits): Likewise CONST_MODE_NUNITS.
9681 * machmode.def (ADJUST_NUNITS): Document.
9682 * genmodes.c (mode_data::need_nunits_adj): New field.
9683 (blank_mode): Update accordingly.
9684 (adj_nunits): New variable.
9685 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9686 parameter.
9687 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9688 listed in adj_nunits.
9689 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9690 listed in adj_nunits. Don't emit case statements for such modes.
9691 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9692 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9693 nothing if adj_nunits is nonnull.
9694 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9695 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9696 (emit_mode_fbit): Update use of print_maybe_const_decl.
9697 (emit_move_size): Likewise. Treat the array as non-const
9698 if adj_nunits.
9699 (emit_mode_adjustments): Handle adj_nunits.
9701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9703 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9704 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9705 (VECTOR_MODES): Use it.
9706 (make_vector_modes): Take the prefix as an argument.
9708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9709 Alan Hayward <alan.hayward@arm.com>
9710 David Sherwood <david.sherwood@arm.com>
9712 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9713 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9714 for MODE_VECTOR_BOOL.
9715 * machmode.def (VECTOR_BOOL_MODE): Document.
9716 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9717 (make_vector_bool_mode): New function.
9718 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9719 MODE_VECTOR_BOOL.
9720 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9721 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9722 Likewise.
9723 * stor-layout.c (int_mode_for_mode): Likewise.
9724 * tree.c (build_vector_type_for_mode): Likewise.
9725 * varasm.c (output_constant_pool_2): Likewise.
9726 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9727 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9728 for MODE_VECTOR_BOOL.
9729 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9730 of mode class checks.
9731 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9732 instead of a list of mode class checks.
9733 (expand_vector_scalar_condition): Likewise.
9734 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9737 Alan Hayward <alan.hayward@arm.com>
9738 David Sherwood <david.sherwood@arm.com>
9740 * machmode.h (mode_size): Change from unsigned short to
9741 poly_uint16_pod.
9742 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9743 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9744 or if measurement_type is not polynomial.
9745 (fixed_size_mode::includes_p): Check for constant-sized modes.
9746 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9747 return a poly_uint16 rather than an unsigned short.
9748 (emit_mode_size): Change the type of mode_size from unsigned short
9749 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9750 (emit_mode_adjustments): Cope with polynomial vector sizes.
9751 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9752 for GET_MODE_SIZE.
9753 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9754 for GET_MODE_SIZE.
9755 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9756 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9757 * caller-save.c (setup_save_areas): Likewise.
9758 (replace_reg_with_saved_mem): Likewise.
9759 * calls.c (emit_library_call_value_1): Likewise.
9760 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9761 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9762 (gen_lowpart_for_combine): Likewise.
9763 * convert.c (convert_to_integer_1): Likewise.
9764 * cse.c (equiv_constant, cse_insn): Likewise.
9765 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9766 (cselib_subst_to_values): Likewise.
9767 * dce.c (word_dce_process_block): Likewise.
9768 * df-problems.c (df_word_lr_mark_ref): Likewise.
9769 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9770 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9771 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9772 (rtl_for_decl_location): Likewise.
9773 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9774 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9775 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9776 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9777 (expand_expr_real_1): Likewise.
9778 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9779 (pad_below): Likewise.
9780 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9781 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9782 * ira.c (get_subreg_tracking_sizes): Likewise.
9783 * ira-build.c (ira_create_allocno_objects): Likewise.
9784 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9785 (ira_sort_regnos_for_alter_reg): Likewise.
9786 * ira-costs.c (record_operand_costs): Likewise.
9787 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9788 (resolve_simple_move): Likewise.
9789 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9790 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9791 (lra_constraints): Likewise.
9792 (CONST_POOL_OK_P): Reject variable-sized modes.
9793 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9794 (add_pseudo_to_slot, lra_spill): Likewise.
9795 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9796 * optabs-query.c (get_best_extraction_insn): Likewise.
9797 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9798 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9799 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9800 * recog.c (offsettable_address_addr_space_p): Likewise.
9801 * regcprop.c (maybe_mode_change): Likewise.
9802 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9803 * regrename.c (build_def_use): Likewise.
9804 * regstat.c (dump_reg_info): Likewise.
9805 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9806 (find_reloads, find_reloads_subreg_address): Likewise.
9807 * reload1.c (eliminate_regs_1): Likewise.
9808 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9809 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9810 (simplify_binary_operation_1, simplify_subreg): Likewise.
9811 * targhooks.c (default_function_arg_padding): Likewise.
9812 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9813 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9814 (verify_gimple_assign_ternary): Likewise.
9815 * tree-inline.c (estimate_move_cost): Likewise.
9816 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9817 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9818 (get_address_cost_ainc): Likewise.
9819 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9820 (vect_supportable_dr_alignment): Likewise.
9821 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9822 (vectorizable_reduction): Likewise.
9823 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9824 (vectorizable_operation, vectorizable_load): Likewise.
9825 * tree.c (build_same_sized_truth_vector_type): Likewise.
9826 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9827 * var-tracking.c (emit_note_insn_var_location): Likewise.
9828 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9829 (ADDR_VEC_ALIGN): Likewise.
9831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9832 Alan Hayward <alan.hayward@arm.com>
9833 David Sherwood <david.sherwood@arm.com>
9835 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9836 unsigned short.
9837 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9838 or if measurement_type is polynomial.
9839 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9840 * combine.c (make_extraction): Likewise.
9841 * dse.c (find_shift_sequence): Likewise.
9842 * dwarf2out.c (mem_loc_descriptor): Likewise.
9843 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9844 (extract_bit_field, extract_low_bits): Likewise.
9845 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9846 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9847 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9848 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9849 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9850 * reload.c (find_reloads): Likewise.
9851 * reload1.c (alter_reg): Likewise.
9852 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9853 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9854 * tree-if-conv.c (predicate_mem_writes): Likewise.
9855 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9856 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9857 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9858 * valtrack.c (dead_debug_insert_temp): Likewise.
9859 * varasm.c (mergeable_constant_section): Likewise.
9860 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9863 Alan Hayward <alan.hayward@arm.com>
9864 David Sherwood <david.sherwood@arm.com>
9866 * expr.c (expand_assignment): Cope with polynomial mode sizes
9867 when assigning to a CONCAT.
9869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9870 Alan Hayward <alan.hayward@arm.com>
9871 David Sherwood <david.sherwood@arm.com>
9873 * machmode.h (mode_precision): Change from unsigned short to
9874 poly_uint16_pod.
9875 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9876 short.
9877 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9878 or if measurement_type is not polynomial.
9879 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9880 in which the mode is already known to be a scalar_int_mode.
9881 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9882 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9883 initializer.
9884 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9885 for GET_MODE_PRECISION.
9886 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9887 for GET_MODE_PRECISION.
9888 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9889 as polynomial.
9890 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9891 (expand_field_assignment, make_extraction): Likewise.
9892 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9893 (get_last_value): Likewise.
9894 * convert.c (convert_to_integer_1): Likewise.
9895 * cse.c (cse_insn): Likewise.
9896 * expr.c (expand_expr_real_1): Likewise.
9897 * lra-constraints.c (simplify_operand_subreg): Likewise.
9898 * optabs-query.c (can_atomic_load_p): Likewise.
9899 * optabs.c (expand_atomic_load): Likewise.
9900 (expand_atomic_store): Likewise.
9901 * ree.c (combine_reaching_defs): Likewise.
9902 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9903 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9904 * tree.h (type_has_mode_precision_p): Likewise.
9905 * ubsan.c (instrument_si_overflow): Likewise.
9907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9908 Alan Hayward <alan.hayward@arm.com>
9909 David Sherwood <david.sherwood@arm.com>
9911 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9912 polynomial numbers of units.
9913 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9914 (valid_vector_subparts_p): New function.
9915 (build_vector_type): Remove temporary shim and take the number
9916 of units as a poly_uint64 rather than an int.
9917 (build_opaque_vector_type): Take the number of units as a
9918 poly_uint64 rather than an int.
9919 * tree.c (build_vector_from_ctor): Handle polynomial
9920 TYPE_VECTOR_SUBPARTS.
9921 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9922 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9923 (build_vector_from_val): If the number of units is variable,
9924 use build_vec_duplicate_cst for constant operands and
9925 VEC_DUPLICATE_EXPR otherwise.
9926 (make_vector_type): Remove temporary is_constant ().
9927 (build_vector_type, build_opaque_vector_type): Take the number of
9928 units as a poly_uint64 rather than an int.
9929 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9930 VECTOR_CST_NELTS.
9931 * cfgexpand.c (expand_debug_expr): Likewise.
9932 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9933 (store_constructor, expand_expr_real_1): Likewise.
9934 (const_scalar_mask_from_tree): Likewise.
9935 * fold-const-call.c (fold_const_reduction): Likewise.
9936 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9937 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9938 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9939 (fold_relational_const): Likewise.
9940 (native_interpret_vector): Likewise. Change the size from an
9941 int to an unsigned int.
9942 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9943 TYPE_VECTOR_SUBPARTS.
9944 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9945 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9946 duplicating a non-constant operand into a variable-length vector.
9947 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9948 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9949 * ipa-icf.c (sem_variable::equals): Likewise.
9950 * match.pd: Likewise.
9951 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9952 * print-tree.c (print_node): Likewise.
9953 * stor-layout.c (layout_type): Likewise.
9954 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9955 * tree-cfg.c (verify_gimple_comparison): Likewise.
9956 (verify_gimple_assign_binary): Likewise.
9957 (verify_gimple_assign_ternary): Likewise.
9958 (verify_gimple_assign_single): Likewise.
9959 * tree-pretty-print.c (dump_generic_node): Likewise.
9960 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9961 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9962 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9963 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9964 (vect_shift_permute_load_chain): Likewise.
9965 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9966 (expand_vector_condition, optimize_vector_constructor): Likewise.
9967 (lower_vec_perm, get_compute_type): Likewise.
9968 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9969 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9970 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9971 (vect_recog_mask_conversion_pattern): Likewise.
9972 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9973 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9974 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9975 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9976 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9977 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9978 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9979 (supportable_widening_operation): Likewise.
9980 (supportable_narrowing_operation): Likewise.
9981 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9982 Likewise.
9983 * varasm.c (output_constant): Likewise.
9985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9986 Alan Hayward <alan.hayward@arm.com>
9987 David Sherwood <david.sherwood@arm.com>
9989 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9990 so that both the length == 3 and length != 3 cases set up their
9991 own permute vectors. Add comments explaining why we know the
9992 number of elements is constant.
9993 (vect_permute_load_chain): Likewise.
9995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9996 Alan Hayward <alan.hayward@arm.com>
9997 David Sherwood <david.sherwood@arm.com>
9999 * machmode.h (mode_nunits): Change from unsigned char to
10000 poly_uint16_pod.
10001 (ONLY_FIXED_SIZE_MODES): New macro.
10002 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10003 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10004 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10005 New typedefs.
10006 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10007 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10008 or if measurement_type is not polynomial.
10009 * genmodes.c (ZERO_COEFFS): New macro.
10010 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10011 poly_uint16.
10012 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10013 Use ZERO_COEFFS when emitting initializers.
10014 * data-streamer.h (bp_pack_poly_value): New function.
10015 (bp_unpack_poly_value): Likewise.
10016 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10017 for GET_MODE_NUNITS.
10018 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10019 for GET_MODE_NUNITS.
10020 * tree.c (make_vector_type): Remove temporary shim and make
10021 the real function take the number of units as a poly_uint64
10022 rather than an int.
10023 (build_vector_type_for_mode): Handle polynomial nunits.
10024 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10025 * emit-rtl.c (const_vec_series_p_1): Likewise.
10026 (gen_rtx_CONST_VECTOR): Likewise.
10027 * fold-const.c (test_vec_duplicate_folding): Likewise.
10028 * genrecog.c (validate_pattern): Likewise.
10029 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10030 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10031 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10032 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10033 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10034 * rtlanal.c (subreg_get_info): Likewise.
10035 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10036 (vect_grouped_load_supported): Likewise.
10037 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10038 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10039 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10040 (simplify_const_unary_operation, simplify_binary_operation_1)
10041 (simplify_const_binary_operation, simplify_ternary_operation)
10042 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10043 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10044 instead of CONST_VECTOR_NUNITS.
10045 * varasm.c (output_constant_pool_2): Likewise.
10046 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10047 explicit-encoded elements in the XVEC for variable-length vectors.
10049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10051 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10054 Alan Hayward <alan.hayward@arm.com>
10055 David Sherwood <david.sherwood@arm.com>
10057 * coretypes.h (fixed_size_mode): Declare.
10058 (fixed_size_mode_pod): New typedef.
10059 * builtins.h (target_builtins::x_apply_args_mode)
10060 (target_builtins::x_apply_result_mode): Change type to
10061 fixed_size_mode_pod.
10062 * builtins.c (apply_args_size, apply_result_size, result_vector)
10063 (expand_builtin_apply_args_1, expand_builtin_apply)
10064 (expand_builtin_return): Update accordingly.
10066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10068 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10069 * cselib.c (cselib_hash_rtx): Likewise.
10070 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10071 CONST_VECTOR encoding.
10073 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10074 Jeff Law <law@redhat.com>
10076 PR target/83641
10077 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10078 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10079 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10080 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10082 PR target/83641
10083 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10084 explicitly probe *sp in a noreturn function if there were any callee
10085 register saves or frame pointer is needed.
10087 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10089 PR debug/83621
10090 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10091 BLKmode for ternary, binary or unary expressions.
10093 PR debug/83645
10094 * var-tracking.c (delete_vta_debug_insn): New inline function.
10095 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10096 insns from get_insns () to NULL instead of each bb separately.
10097 Use delete_vta_debug_insn. No longer static.
10098 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10099 delete_vta_debug_insns callers.
10100 * rtl.h (delete_vta_debug_insns): Declare.
10101 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10102 instead of variable_tracking_main.
10104 2018-01-03 Martin Sebor <msebor@redhat.com>
10106 PR tree-optimization/83603
10107 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10108 arguments past the endof the argument list in functions declared
10109 without a prototype.
10110 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10111 Avoid checking when arguments are null.
10113 2018-01-03 Martin Sebor <msebor@redhat.com>
10115 PR c/83559
10116 * doc/extend.texi (attribute const): Fix a typo.
10117 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10118 issuing -Wsuggest-attribute for void functions.
10120 2018-01-03 Martin Sebor <msebor@redhat.com>
10122 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10123 offset_int::from instead of wide_int::to_shwi.
10124 (maybe_diag_overlap): Remove assertion.
10125 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10126 * gimple-ssa-sprintf.c (format_directive): Same.
10127 (parse_directive): Same.
10128 (sprintf_dom_walker::compute_format_length): Same.
10129 (try_substitute_return_value): Same.
10131 2018-01-03 Jeff Law <law@redhat.com>
10133 PR middle-end/83654
10134 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10135 non-constant residual for zero at runtime and avoid probing in
10136 that case. Reorganize code for trailing problem to mirror handling
10137 of the residual.
10139 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10141 PR tree-optimization/83501
10142 * tree-ssa-strlen.c (get_string_cst): New.
10143 (handle_char_store): Call get_string_cst.
10145 2018-01-03 Martin Liska <mliska@suse.cz>
10147 PR tree-optimization/83593
10148 * tree-ssa-strlen.c: Include tree-cfg.h.
10149 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10150 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10151 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10152 to false.
10153 (strlen_dom_walker::before_dom_children): Call
10154 gimple_purge_dead_eh_edges. Dump tranformation with details
10155 dump flags.
10156 (strlen_dom_walker::before_dom_children): Update call by adding
10157 new argument cleanup_eh.
10158 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10160 2018-01-03 Martin Liska <mliska@suse.cz>
10162 PR ipa/83549
10163 * cif-code.def (VARIADIC_THUNK): New enum value.
10164 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10165 thunks.
10167 2018-01-03 Jan Beulich <jbeulich@suse.com>
10169 * sse.md (mov<mode>_internal): Tighten condition for when to use
10170 vmovdqu<ssescalarsize> for TI and OI modes.
10172 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10174 Update copyright years.
10176 2018-01-03 Martin Liska <mliska@suse.cz>
10178 PR ipa/83594
10179 * ipa-visibility.c (function_and_variable_visibility): Skip
10180 functions with noipa attribure.
10182 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10184 * gcc.c (process_command): Update copyright notice dates.
10185 * gcov-dump.c (print_version): Ditto.
10186 * gcov.c (print_version): Ditto.
10187 * gcov-tool.c (print_version): Ditto.
10188 * gengtype.c (create_file): Ditto.
10189 * doc/cpp.texi: Bump @copying's copyright year.
10190 * doc/cppinternals.texi: Ditto.
10191 * doc/gcc.texi: Ditto.
10192 * doc/gccint.texi: Ditto.
10193 * doc/gcov.texi: Ditto.
10194 * doc/install.texi: Ditto.
10195 * doc/invoke.texi: Ditto.
10197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10199 * vector-builder.h (vector_builder::m_full_nelts): Change from
10200 unsigned int to poly_uint64.
10201 (vector_builder::full_nelts): Update prototype accordingly.
10202 (vector_builder::new_vector): Likewise.
10203 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10204 (vector_builder::operator ==): Likewise.
10205 (vector_builder::finalize): Likewise.
10206 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10207 Take the number of elements as a poly_uint64 rather than an
10208 unsigned int.
10209 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10210 from unsigned int to poly_uint64.
10211 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10212 (vec_perm_indices::new_vector): Likewise.
10213 (vec_perm_indices::length): Likewise.
10214 (vec_perm_indices::nelts_per_input): Likewise.
10215 (vec_perm_indices::input_nelts): Likewise.
10216 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10217 number of elements per input as a poly_uint64 rather than an
10218 unsigned int. Use the original encoding for variable-length
10219 vectors, rather than clamping each individual element.
10220 For the second and subsequent elements in each pattern,
10221 clamp the step and base before clamping their sum.
10222 (vec_perm_indices::series_p): Handle polynomial element counts.
10223 (vec_perm_indices::all_in_range_p): Likewise.
10224 (vec_perm_indices_to_tree): Likewise.
10225 (vec_perm_indices_to_rtx): Likewise.
10226 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10227 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10228 (tree_vector_builder::new_binary_operation): Handle polynomial
10229 element counts. Return false if we need to know the number
10230 of elements at compile time.
10231 * fold-const.c (fold_vec_perm): Punt if the number of elements
10232 isn't known at compile time.
10234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10236 * vec-perm-indices.h (vec_perm_builder): Change element type
10237 from HOST_WIDE_INT to poly_int64.
10238 (vec_perm_indices::element_type): Update accordingly.
10239 (vec_perm_indices::clamp): Handle polynomial element_types.
10240 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10241 (vec_perm_indices::all_in_range_p): Likewise.
10242 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10243 than shwi trees.
10244 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10245 polynomial vec_perm_indices element types.
10246 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10247 * fold-const.c (fold_vec_perm): Likewise.
10248 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10249 * tree-vect-generic.c (lower_vec_perm): Likewise.
10250 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10251 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10252 element type to HOST_WIDE_INT.
10254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10255 Alan Hayward <alan.hayward@arm.com>
10256 David Sherwood <david.sherwood@arm.com>
10258 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10259 rather than an int. Use plus_constant.
10260 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10261 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10264 Alan Hayward <alan.hayward@arm.com>
10265 David Sherwood <david.sherwood@arm.com>
10267 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10268 a HOST_WIDE_INT to a poly_int64.
10270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10271 Alan Hayward <alan.hayward@arm.com>
10272 David Sherwood <david.sherwood@arm.com>
10274 * calls.c (load_register_parameters): Cope with polynomial
10275 mode sizes. Require a constant size for BLKmode parameters
10276 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10277 forces a parameter to be padded at the lsb end in order to
10278 fill a complete number of words, require the parameter size
10279 to be ordered wrt UNITS_PER_WORD.
10281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10282 Alan Hayward <alan.hayward@arm.com>
10283 David Sherwood <david.sherwood@arm.com>
10285 * reload1.c (spill_stack_slot_width): Change element type
10286 from unsigned int to poly_uint64_pod.
10287 (alter_reg): Treat mode sizes as polynomial.
10289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10290 Alan Hayward <alan.hayward@arm.com>
10291 David Sherwood <david.sherwood@arm.com>
10293 * reload.c (complex_word_subreg_p): New function.
10294 (reload_inner_reg_of_subreg, push_reload): Use it.
10296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10297 Alan Hayward <alan.hayward@arm.com>
10298 David Sherwood <david.sherwood@arm.com>
10300 * lra-constraints.c (process_alt_operands): Reject matched
10301 operands whose sizes aren't ordered.
10302 (match_reload): Refer to this check here.
10304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10305 Alan Hayward <alan.hayward@arm.com>
10306 David Sherwood <david.sherwood@arm.com>
10308 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10309 that the mode size is in the set {1, 2, 4, 8, 16}.
10311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10312 Alan Hayward <alan.hayward@arm.com>
10313 David Sherwood <david.sherwood@arm.com>
10315 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10316 Use plus_constant instead of gen_rtx_PLUS.
10318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10319 Alan Hayward <alan.hayward@arm.com>
10320 David Sherwood <david.sherwood@arm.com>
10322 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10323 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10324 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10325 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10326 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10327 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10328 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10329 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10330 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10331 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10332 a poly_int64.
10333 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10334 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10335 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10336 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10337 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10338 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10339 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10340 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10341 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10342 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10343 function.
10344 * expr.c (emit_move_resolve_push): Treat the input and result
10345 of PUSH_ROUNDING as a poly_int64.
10346 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10347 (emit_push_insn): Likewise.
10348 * lra-eliminations.c (mark_not_eliminable): Likewise.
10349 * recog.c (push_operand): Likewise.
10350 * reload1.c (elimination_effects): Likewise.
10351 * rtlanal.c (nonzero_bits1): Likewise.
10352 * calls.c (store_one_arg): Likewise. Require the padding to be
10353 known at compile time.
10355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10356 Alan Hayward <alan.hayward@arm.com>
10357 David Sherwood <david.sherwood@arm.com>
10359 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10360 Use plus_constant instead of gen_rtx_PLUS.
10362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10363 Alan Hayward <alan.hayward@arm.com>
10364 David Sherwood <david.sherwood@arm.com>
10366 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10367 rather than an int.
10369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10370 Alan Hayward <alan.hayward@arm.com>
10371 David Sherwood <david.sherwood@arm.com>
10373 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10374 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10375 via stack temporaries. Treat the mode size as polynomial too.
10377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10378 Alan Hayward <alan.hayward@arm.com>
10379 David Sherwood <david.sherwood@arm.com>
10381 * expr.c (expand_expr_real_2): When handling conversions involving
10382 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10383 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10384 as a poly_uint64 too.
10386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10387 Alan Hayward <alan.hayward@arm.com>
10388 David Sherwood <david.sherwood@arm.com>
10390 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10393 Alan Hayward <alan.hayward@arm.com>
10394 David Sherwood <david.sherwood@arm.com>
10396 * combine.c (can_change_dest_mode): Handle polynomial
10397 REGMODE_NATURAL_SIZE.
10398 * expmed.c (store_bit_field_1): Likewise.
10399 * expr.c (store_constructor): Likewise.
10400 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10401 and polynomial REGMODE_NATURAL_SIZE.
10402 (gen_lowpart_common): Likewise.
10403 * reginfo.c (record_subregs_of_mode): Likewise.
10404 * rtlanal.c (read_modify_subreg_p): Likewise.
10406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10407 Alan Hayward <alan.hayward@arm.com>
10408 David Sherwood <david.sherwood@arm.com>
10410 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10411 numbers of elements.
10413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10414 Alan Hayward <alan.hayward@arm.com>
10415 David Sherwood <david.sherwood@arm.com>
10417 * match.pd: Cope with polynomial numbers of vector elements.
10419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10420 Alan Hayward <alan.hayward@arm.com>
10421 David Sherwood <david.sherwood@arm.com>
10423 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10424 in a POINTER_PLUS_EXPR.
10426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10427 Alan Hayward <alan.hayward@arm.com>
10428 David Sherwood <david.sherwood@arm.com>
10430 * omp-simd-clone.c (simd_clone_subparts): New function.
10431 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10432 (ipa_simd_modify_function_body): Likewise.
10434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10435 Alan Hayward <alan.hayward@arm.com>
10436 David Sherwood <david.sherwood@arm.com>
10438 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10439 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10440 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10441 (expand_vector_condition, vector_element): Likewise.
10442 (subparts_gt): New function.
10443 (get_compute_type): Use subparts_gt.
10444 (count_type_subparts): Delete.
10445 (expand_vector_operations_1): Use subparts_gt instead of
10446 count_type_subparts.
10448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10449 Alan Hayward <alan.hayward@arm.com>
10450 David Sherwood <david.sherwood@arm.com>
10452 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10453 (vect_compile_time_alias): ...this new function. Do the calculation
10454 on poly_ints rather than trees.
10455 (vect_prune_runtime_alias_test_list): Update call accordingly.
10457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10458 Alan Hayward <alan.hayward@arm.com>
10459 David Sherwood <david.sherwood@arm.com>
10461 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10462 numbers of units.
10463 (vect_schedule_slp_instance): Likewise.
10465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10466 Alan Hayward <alan.hayward@arm.com>
10467 David Sherwood <david.sherwood@arm.com>
10469 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10470 constant and extern definitions for variable-length vectors.
10471 (vect_get_constant_vectors): Note that the number of units
10472 is known to be constant.
10474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10475 Alan Hayward <alan.hayward@arm.com>
10476 David Sherwood <david.sherwood@arm.com>
10478 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10479 of units as polynomial. Choose between WIDE and NARROW based
10480 on multiple_p.
10482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10483 Alan Hayward <alan.hayward@arm.com>
10484 David Sherwood <david.sherwood@arm.com>
10486 * tree-vect-stmts.c (simd_clone_subparts): New function.
10487 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10490 Alan Hayward <alan.hayward@arm.com>
10491 David Sherwood <david.sherwood@arm.com>
10493 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10494 vectors as polynomial. Use build_index_vector for
10495 IFN_GOMP_SIMD_LANE.
10497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10498 Alan Hayward <alan.hayward@arm.com>
10499 David Sherwood <david.sherwood@arm.com>
10501 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10502 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10503 for variable-length vectors.
10504 (vectorizable_mask_load_store): Treat the number of units as
10505 polynomial, asserting that it is constant if the condition has
10506 already been enforced.
10507 (vectorizable_store, vectorizable_load): Likewise.
10509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10510 Alan Hayward <alan.hayward@arm.com>
10511 David Sherwood <david.sherwood@arm.com>
10513 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10514 of units as polynomial. Punt if we can't tell at compile time
10515 which vector contains the final result.
10517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10518 Alan Hayward <alan.hayward@arm.com>
10519 David Sherwood <david.sherwood@arm.com>
10521 * tree-vect-loop.c (vectorizable_induction): Treat the number
10522 of units as polynomial. Punt on SLP inductions. Use an integer
10523 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10524 cast of such a series for variable-length floating-point
10525 reductions.
10527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10528 Alan Hayward <alan.hayward@arm.com>
10529 David Sherwood <david.sherwood@arm.com>
10531 * tree.h (build_index_vector): Declare.
10532 * tree.c (build_index_vector): New function.
10533 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10534 of units as polynomial, forcibly converting it to a constant if
10535 vectorizable_reduction has already enforced the condition.
10536 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10537 to create a {1,2,3,...} vector.
10538 (vectorizable_reduction): Treat the number of units as polynomial.
10539 Choose vectype_in based on the largest scalar element size rather
10540 than the smallest number of units. Enforce the restrictions
10541 relied on above.
10543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10544 Alan Hayward <alan.hayward@arm.com>
10545 David Sherwood <david.sherwood@arm.com>
10547 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10548 number of units as polynomial.
10550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10551 Alan Hayward <alan.hayward@arm.com>
10552 David Sherwood <david.sherwood@arm.com>
10554 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10555 * target.def (autovectorize_vector_sizes): Return the vector sizes
10556 by pointer, using vector_sizes rather than a bitmask.
10557 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10558 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10559 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10560 Likewise.
10561 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10562 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10563 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10564 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10565 * omp-general.c (omp_max_vf): Likewise.
10566 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10567 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10568 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10569 * tree-vect-slp.c (vect_slp_bb): Likewise.
10570 * doc/tm.texi: Regenerate.
10571 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10572 to a poly_uint64.
10573 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10574 the vector size as a poly_uint64 rather than an unsigned int.
10575 (current_vector_size): Change from an unsigned int to a poly_uint64.
10576 (get_vectype_for_scalar_type): Update accordingly.
10577 * tree.h (build_truth_vector_type): Take the size and number of
10578 units as a poly_uint64 rather than an unsigned int.
10579 (build_vector_type): Add a temporary overload that takes
10580 the number of units as a poly_uint64 rather than an unsigned int.
10581 * tree.c (make_vector_type): Likewise.
10582 (build_truth_vector_type): Take the number of units as a poly_uint64
10583 rather than an unsigned int.
10585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10586 Alan Hayward <alan.hayward@arm.com>
10587 David Sherwood <david.sherwood@arm.com>
10589 * target.def (get_mask_mode): Take the number of units and length
10590 as poly_uint64s rather than unsigned ints.
10591 * targhooks.h (default_get_mask_mode): Update accordingly.
10592 * targhooks.c (default_get_mask_mode): Likewise.
10593 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10594 * doc/tm.texi: Regenerate.
10596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10597 Alan Hayward <alan.hayward@arm.com>
10598 David Sherwood <david.sherwood@arm.com>
10600 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10601 * omp-general.c (omp_max_vf): Likewise.
10602 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10603 (expand_omp_simd): Handle polynomial safelen.
10604 * omp-low.c (omplow_simd_context): Add a default constructor.
10605 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10606 (lower_rec_simd_input_clauses): Update accordingly.
10607 (lower_rec_input_clauses): Likewise.
10609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10610 Alan Hayward <alan.hayward@arm.com>
10611 David Sherwood <david.sherwood@arm.com>
10613 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10614 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10615 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10616 (vect_analyze_slp_cost): Likewise.
10617 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10618 (vect_model_load_cost): Likewise.
10620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10621 Alan Hayward <alan.hayward@arm.com>
10622 David Sherwood <david.sherwood@arm.com>
10624 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10625 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10626 from an unsigned int * to a poly_uint64_pod *.
10627 (calculate_unrolling_factor): New function.
10628 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10631 Alan Hayward <alan.hayward@arm.com>
10632 David Sherwood <david.sherwood@arm.com>
10634 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10635 from an unsigned int to a poly_uint64.
10636 (_loop_vec_info::slp_unrolling_factor): Likewise.
10637 (_loop_vec_info::vectorization_factor): Change from an int
10638 to a poly_uint64.
10639 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10640 (vect_get_num_vectors): New function.
10641 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10642 (vect_get_num_copies): Use vect_get_num_vectors.
10643 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10644 to an unsigned int *.
10645 (vect_analyze_data_refs): Change min_vf from an int * to a
10646 poly_uint64 *.
10647 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10648 than an unsigned HOST_WIDE_INT.
10649 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10650 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10651 to an unsigned int *.
10652 (vect_analyze_data_ref_dependences): Likewise.
10653 (vect_compute_data_ref_alignment): Handle polynomial vf.
10654 (vect_enhance_data_refs_alignment): Likewise.
10655 (vect_prune_runtime_alias_test_list): Likewise.
10656 (vect_shift_permute_load_chain): Likewise.
10657 (vect_supportable_dr_alignment): Likewise.
10658 (dependence_distance_ge_vf): Take the vectorization factor as a
10659 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10660 (vect_analyze_data_refs): Change min_vf from an int * to a
10661 poly_uint64 *.
10662 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10663 vfm1 as a poly_uint64 rather than an int. Make the same change
10664 for the returned bound_scalar.
10665 (vect_gen_vector_loop_niters): Handle polynomial vf.
10666 (vect_do_peeling): Likewise. Update call to
10667 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10668 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10669 be constant.
10670 * tree-vect-loop.c (vect_determine_vectorization_factor)
10671 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10672 (vect_get_known_peeling_cost): Likewise.
10673 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10674 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10675 (vect_transform_loop): Likewise. Use the lowest possible VF when
10676 updating the upper bounds of the loop.
10677 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10678 rather than an int.
10679 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10680 polynomial unroll factors.
10681 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10682 (vect_make_slp_decision): Likewise.
10683 (vect_supported_load_permutation_p): Likewise, and polynomial
10684 vf too.
10685 (vect_analyze_slp_cost): Handle polynomial vf.
10686 (vect_slp_analyze_node_operations): Likewise.
10687 (vect_slp_analyze_bb_1): Likewise.
10688 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10689 than an unsigned HOST_WIDE_INT.
10690 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10691 (vectorizable_load): Handle polynomial vf.
10692 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10693 a poly_uint64.
10694 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10697 Alan Hayward <alan.hayward@arm.com>
10698 David Sherwood <david.sherwood@arm.com>
10700 * match.pd: Handle bit operations involving three constants
10701 and try to fold one pair.
10703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10705 * tree-vect-loop-manip.c: Include gimple-fold.h.
10706 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10707 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10708 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10709 Add a path that uses a step of VF instead of 1, but disable it
10710 for now.
10711 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10712 and niters_no_overflow parameters. Update calls to
10713 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10714 Create a new SSA name if the latter choses to use a ste other
10715 than zero, and return it via niters_vector_mult_vf_var.
10716 * tree-vect-loop.c (vect_transform_loop): Update calls to
10717 vect_do_peeling, vect_gen_vector_loop_niters and
10718 slpeel_make_loop_iterate_ntimes.
10719 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10720 (vect_gen_vector_loop_niters): Update declarations after above changes.
10722 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10724 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10725 128-bit round to integer instructions.
10726 (ceil<mode>2): Likewise.
10727 (btrunc<mode>2): Likewise.
10728 (round<mode>2): Likewise.
10730 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10732 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10733 unaligned VSX load/store on P8/P9.
10734 (expand_block_clear): Allow the use of unaligned VSX
10735 load/store on P8/P9.
10737 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10739 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10740 New function.
10741 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10742 swap associated with both a load and a store.
10744 2018-01-02 Andrew Waterman <andrew@sifive.com>
10746 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10747 * config/riscv/riscv.md (clear_cache): Use it.
10749 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10751 * web.c: Remove out-of-date comment.
10753 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10755 * expr.c (fixup_args_size_notes): Check that any existing
10756 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10757 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10758 (emit_single_push_insn): ...here.
10760 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10762 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10763 (const_vector_encoded_nelts): New function.
10764 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10765 (const_vector_int_elt, const_vector_elt): Declare.
10766 * emit-rtl.c (const_vector_int_elt_1): New function.
10767 (const_vector_elt): Likewise.
10768 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10769 of CONST_VECTOR_ELT.
10771 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10773 * expr.c: Include rtx-vector-builder.h.
10774 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10775 directly on the tree encoding.
10776 (const_vector_from_tree): Likewise.
10777 * optabs.c: Include rtx-vector-builder.h.
10778 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10779 sequence of "u" values.
10780 * vec-perm-indices.c: Include rtx-vector-builder.h.
10781 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10782 directly on the vec_perm_indices encoding.
10784 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10786 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10787 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10788 * rtx-vector-builder.h: New file.
10789 * rtx-vector-builder.c: Likewise.
10790 * rtl.h (rtx_def::u2): Add a const_vector field.
10791 (CONST_VECTOR_NPATTERNS): New macro.
10792 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10793 (CONST_VECTOR_DUPLICATE_P): Likewise.
10794 (CONST_VECTOR_STEPPED_P): Likewise.
10795 (CONST_VECTOR_ENCODED_ELT): Likewise.
10796 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10797 (unwrap_const_vec_duplicate): Likewise.
10798 (const_vec_series_p): Check for a non-duplicated vector encoding.
10799 Say that the function only returns true for integer vectors.
10800 * emit-rtl.c: Include rtx-vector-builder.h.
10801 (gen_const_vec_duplicate_1): Delete.
10802 (gen_const_vector): Call gen_const_vec_duplicate instead of
10803 gen_const_vec_duplicate_1.
10804 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10805 (gen_const_vec_duplicate): Use rtx_vector_builder.
10806 (gen_const_vec_series): Likewise.
10807 (gen_rtx_CONST_VECTOR): Likewise.
10808 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10809 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10810 Build a new vector rather than modifying a CONST_VECTOR in-place.
10811 (handle_special_swappables): Update call accordingly.
10812 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10813 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10814 Build a new vector rather than modifying a CONST_VECTOR in-place.
10815 (handle_special_swappables): Update call accordingly.
10817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10819 * simplify-rtx.c (simplify_const_binary_operation): Use
10820 CONST_VECTOR_ELT instead of XVECEXP.
10822 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10824 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10825 the selector elements to be different from the data elements
10826 if the selector is a VECTOR_CST.
10827 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10828 ssizetype for the selector.
10830 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10832 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10833 before testing each element individually.
10834 * tree-vect-generic.c (lower_vec_perm): Likewise.
10836 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10838 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10839 * selftest-run-tests.c (selftest::run_tests): Call it.
10840 * vector-builder.h (vector_builder::operator ==): New function.
10841 (vector_builder::operator !=): Likewise.
10842 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10843 (vec_perm_indices::all_from_input_p): New function.
10844 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10845 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10846 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10847 instead of reading the VECTOR_CST directly. Detect whether both
10848 vector inputs are the same before constructing the vec_perm_indices,
10849 and update the number of inputs argument accordingly. Use the
10850 utility functions added above. Only construct sel2 if we need to.
10852 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10854 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10855 the broadcast of the low byte.
10856 (expand_mult_highpart): Use an explicit encoding for the permutes.
10857 * optabs-query.c (can_mult_highpart_p): Likewise.
10858 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10859 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10860 (vectorizable_bswap): Likewise.
10861 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10862 explicit encoding for the power-of-2 permutes.
10863 (vect_permute_store_chain): Likewise.
10864 (vect_grouped_load_supported): Likewise.
10865 (vect_permute_load_chain): Likewise.
10867 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10869 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10870 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10871 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10872 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10873 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10874 (vect_gen_perm_mask_any): Likewise.
10876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10878 * int-vector-builder.h: New file.
10879 * vec-perm-indices.h: Include int-vector-builder.h.
10880 (vec_perm_indices): Redefine as an int_vector_builder.
10881 (auto_vec_perm_indices): Delete.
10882 (vec_perm_builder): Redefine as a stand-alone class.
10883 (vec_perm_indices::vec_perm_indices): New function.
10884 (vec_perm_indices::clamp): Likewise.
10885 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10886 (vec_perm_indices::new_vector): New function.
10887 (vec_perm_indices::new_expanded_vector): Update for new
10888 vec_perm_indices class.
10889 (vec_perm_indices::rotate_inputs): New function.
10890 (vec_perm_indices::all_in_range_p): Operate directly on the
10891 encoded form, without computing elided elements.
10892 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10893 encoding. Update for new vec_perm_indices class.
10894 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10895 the given vec_perm_builder.
10896 (expand_vec_perm_var): Update vec_perm_builder constructor.
10897 (expand_mult_highpart): Use vec_perm_builder instead of
10898 auto_vec_perm_indices.
10899 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10900 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10901 or double series encoding as appropriate.
10902 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10903 vec_perm_indices instead of auto_vec_perm_indices.
10904 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10905 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10906 (vect_permute_store_chain): Likewise.
10907 (vect_grouped_load_supported): Likewise.
10908 (vect_permute_load_chain): Likewise.
10909 (vect_shift_permute_load_chain): Likewise.
10910 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10911 (vect_transform_slp_perm_load): Likewise.
10912 (vect_schedule_slp_instance): Likewise.
10913 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10914 (vectorizable_mask_load_store): Likewise.
10915 (vectorizable_bswap): Likewise.
10916 (vectorizable_store): Likewise.
10917 (vectorizable_load): Likewise.
10918 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10919 vec_perm_indices instead of auto_vec_perm_indices. Use
10920 tree_to_vec_perm_builder to read the vector from a tree.
10921 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10922 vec_perm_builder instead of a vec_perm_indices.
10923 (have_whole_vector_shift): Use vec_perm_builder and
10924 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10925 truncation to calc_vec_perm_mask_for_shift.
10926 (vect_create_epilog_for_reduction): Likewise.
10927 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10928 from auto_vec_perm_indices to vec_perm_indices.
10929 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10930 instead of changing individual elements.
10931 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10932 the vector in d.perm.
10933 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10934 from auto_vec_perm_indices to vec_perm_indices.
10935 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10936 instead of changing individual elements.
10937 (arm_vectorize_vec_perm_const): Use new_vector to install
10938 the vector in d.perm.
10939 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10940 Update vec_perm_builder constructor.
10941 (rs6000_expand_interleave): Likewise.
10942 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10943 (rs6000_expand_interleave): Likewise.
10945 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10947 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10948 to qimode could truncate the indices.
10949 * optabs.c (expand_vec_perm_var): Likewise.
10951 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10953 * Makefile.in (OBJS): Add vec-perm-indices.o.
10954 * vec-perm-indices.h: New file.
10955 * vec-perm-indices.c: Likewise.
10956 * target.h (vec_perm_indices): Replace with a forward class
10957 declaration.
10958 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10959 * optabs.h: Include vec-perm-indices.h.
10960 (expand_vec_perm): Delete.
10961 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10962 (expand_vec_perm_const): Declare.
10963 * target.def (vec_perm_const_ok): Replace with...
10964 (vec_perm_const): ...this new hook.
10965 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10966 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10967 * doc/tm.texi: Regenerate.
10968 * optabs.def (vec_perm_const): Delete.
10969 * doc/md.texi (vec_perm_const): Likewise.
10970 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10971 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10972 expand_vec_perm for constant permutation vectors. Assert that
10973 the mode of variable permutation vectors is the integer equivalent
10974 of the mode that is being permuted.
10975 * optabs-query.h (selector_fits_mode_p): Declare.
10976 * optabs-query.c: Include vec-perm-indices.h.
10977 (selector_fits_mode_p): New function.
10978 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10979 is defined, instead of checking whether the vec_perm_const_optab
10980 exists. Use targetm.vectorize.vec_perm_const instead of
10981 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10982 fit in the vector mode before using a variable permute.
10983 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10984 vec_perm_indices instead of an rtx.
10985 (expand_vec_perm): Replace with...
10986 (expand_vec_perm_const): ...this new function. Take the selector
10987 as a vec_perm_indices rather than an rtx. Also take the mode of
10988 the selector. Update call to shift_amt_for_vec_perm_mask.
10989 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10990 Use vec_perm_indices::new_expanded_vector to expand the original
10991 selector into bytes. Check whether the indices fit in the vector
10992 mode before using a variable permute.
10993 (expand_vec_perm_var): Make global.
10994 (expand_mult_highpart): Use expand_vec_perm_const.
10995 * fold-const.c: Includes vec-perm-indices.h.
10996 * tree-ssa-forwprop.c: Likewise.
10997 * tree-vect-data-refs.c: Likewise.
10998 * tree-vect-generic.c: Likewise.
10999 * tree-vect-loop.c: Likewise.
11000 * tree-vect-slp.c: Likewise.
11001 * tree-vect-stmts.c: Likewise.
11002 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11003 Delete.
11004 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11005 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11006 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11007 (aarch64_vectorize_vec_perm_const): ...this new function.
11008 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11009 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11010 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11011 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11012 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11013 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11014 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11015 into...
11016 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11017 check for NEON modes.
11018 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11019 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11020 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11021 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11022 into...
11023 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11024 the old VEC_PERM_CONST conditions.
11025 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11026 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11027 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11028 (ia64_vectorize_vec_perm_const_ok): Merge into...
11029 (ia64_vectorize_vec_perm_const): ...this new function.
11030 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11031 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11032 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11033 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11034 * config/mips/mips.c (mips_expand_vec_perm_const)
11035 (mips_vectorize_vec_perm_const_ok): Merge into...
11036 (mips_vectorize_vec_perm_const): ...this new function.
11037 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11038 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11039 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11040 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11041 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11042 (rs6000_expand_vec_perm_const): Delete.
11043 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11044 Delete.
11045 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11046 (altivec_expand_vec_perm_const_le): Take each operand individually.
11047 Operate on constant selectors rather than rtxes.
11048 (altivec_expand_vec_perm_const): Likewise. Update call to
11049 altivec_expand_vec_perm_const_le.
11050 (rs6000_expand_vec_perm_const): Delete.
11051 (rs6000_vectorize_vec_perm_const_ok): Delete.
11052 (rs6000_vectorize_vec_perm_const): New function.
11053 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11054 an element count and rtx array.
11055 (rs6000_expand_extract_even): Update call accordingly.
11056 (rs6000_expand_interleave): Likewise.
11057 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11058 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11059 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11060 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11061 (rs6000_expand_vec_perm_const): Delete.
11062 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11063 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11064 (altivec_expand_vec_perm_const_le): Take each operand individually.
11065 Operate on constant selectors rather than rtxes.
11066 (altivec_expand_vec_perm_const): Likewise. Update call to
11067 altivec_expand_vec_perm_const_le.
11068 (rs6000_expand_vec_perm_const): Delete.
11069 (rs6000_vectorize_vec_perm_const_ok): Delete.
11070 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11071 reference to the SPE evmerge intructions.
11072 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11073 an element count and rtx array.
11074 (rs6000_expand_extract_even): Update call accordingly.
11075 (rs6000_expand_interleave): Likewise.
11076 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11077 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11078 new function.
11079 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11081 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11083 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11084 vector mode and that that mode matches the mode of the data
11085 being permuted.
11086 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11087 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11088 directly using expand_vec_perm_1 when forcing selectors into
11089 registers.
11090 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11092 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11094 * optabs-query.h (can_vec_perm_p): Delete.
11095 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11096 * optabs-query.c (can_vec_perm_p): Split into...
11097 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11098 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11099 particular selector is valid.
11100 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11101 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11102 (vect_grouped_load_supported): Likewise.
11103 (vect_shift_permute_load_chain): Likewise.
11104 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11105 (vect_transform_slp_perm_load): Likewise.
11106 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11107 (vectorizable_bswap): Likewise.
11108 (vect_gen_perm_mask_checked): Likewise.
11109 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11110 implementations of variable permutation vectors into account
11111 when deciding which selector to use.
11112 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11113 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11114 with a false third argument.
11115 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11116 to test whether the constant selector is valid and can_vec_perm_var_p
11117 to test whether a variable selector is valid.
11119 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11121 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11122 * optabs-query.c (can_vec_perm_p): Likewise.
11123 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11124 instead of vec_perm_indices.
11125 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11126 (vect_gen_perm_mask_checked): Likewise,
11127 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11128 (vect_gen_perm_mask_checked): Likewise,
11130 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11132 * optabs-query.h (qimode_for_vec_perm): Declare.
11133 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11134 (qimode_for_vec_perm): ...this new function.
11135 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11137 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11139 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11140 does not have a conditional at the top.
11142 2018-01-02 Richard Biener <rguenther@suse.de>
11144 * ipa-inline.c (big_speedup_p): Fix expression.
11146 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11148 PR target/81616
11149 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11150 for generic 4->6.
11152 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11154 PR target/81616
11155 Generic tuning.
11156 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11157 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11158 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11159 cond_taken_branch_cost 3->4.
11161 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11163 PR tree-optimization/83581
11164 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11165 TODO_cleanup_cfg if any changes have been made.
11167 PR middle-end/83608
11168 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11169 convert_modes if target mode has the right side, but different mode
11170 class.
11172 PR middle-end/83609
11173 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11174 last argument when extracting from CONCAT. If either from_real or
11175 from_imag is NULL, use expansion through memory. If result is not
11176 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11177 the parts directly to inner mode, if even that fails, use expansion
11178 through memory.
11180 PR middle-end/83623
11181 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11182 check for bswap in mode rather than HImode and use that in expand_unop
11183 too.
11184 \f
11185 Copyright (C) 2018 Free Software Foundation, Inc.
11187 Copying and distribution of this file, with or without modification,
11188 are permitted in any medium without royalty provided the copyright
11189 notice and this notice are preserved.