1 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
7 * config/nds32/nds32-protos.h (nds32_data_alignment,
8 nds32_local_alignment): Declare.
9 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
10 nds32_local_alignment): New functions.
11 (TARGET_CONSTANT_ALIGNMENT): Define.
12 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
14 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16 * config/nds32/nds32.c
17 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
18 (TARGET_MODES_TIEABLE_P): Likewise.
20 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
22 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
25 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
26 Chung-Ju Wu <jasonwucj@gmail.com>
28 * config/nds32/constants.md (unspec_volatile_element): Add enum values
30 * config/nds32/nds32-intrinsic.c: Implementation of expanding
32 * config/nds32/nds32-intrinsic.md: Likewise.
33 * config/nds32/nds32_intrinsic.h: Likewise.
34 * config/nds32/nds32.h (nds32_builtins): Likewise.
35 * config/nds32/nds32.opt (munaligned-access): New option.
36 * config/nds32/nds32.c (nds32_asm_file_start): Display
37 flag_unaligned_access status.
39 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
41 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
42 -mno-relax is present.
43 * config/riscv/linux.h (LINK_SPEC): Ditto.
45 2018-04-20 Martin Sebor <msebor@redhat.com>
48 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
50 (gimple_fold_builtin_stxcpy_chk): Same.
51 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
53 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
56 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
57 __powikf2 when long double is IEEE 128-bit.
59 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
61 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
62 step to make sure stack always aligned.
64 2018-04-20 Carl Love <cel@us.ibm.com>
67 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
70 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
71 Tom de Vries <tom@codesourcery.com>
74 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
75 Emit insns for calls too.
76 (nvptx_find_par): Always look for worker-level predecessor insn.
77 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
79 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
80 (nvptx_process_pars): Propagate frames for calls.
82 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
85 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
87 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
88 (ix86_handle_option): Don't handle OPT_mibt.
89 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
91 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
93 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
94 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
95 (ix86_target_macros): Define __CET__ with flag_cf_protection
97 * config/i386/i386.c (isa2_opts): Remove -mibt.
98 * config/i386/i386.h (TARGET_IBT): Removed.
99 (TARGET_IBT_P): Likewise.
100 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
101 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
102 * config/i386/i386.opt (mcet): Update help message.
105 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
106 -mcet as an alias for -mshstk.
108 2018-04-20 Richard Biener <rguenther@suse.de>
111 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
112 complexity by forcing a single use of the multiply operand.
114 2018-04-20 Martin Jambor <mjambor@suse.cz>
117 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
118 recursion dependency to only apply to non-clones.
120 2018-04-20 Martin Jambor <mjambor@suse.cz>
123 * ipa-cp.c (create_specialized_node): Check that clones of
124 self-recursive edges exist during IPA-CP.
126 2018-04-19 Toon Moene <toon@moene.org>
128 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
131 2018-04-19 Jakub Jelinek <jakub@redhat.com>
133 PR tree-optimization/85467
134 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
135 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
136 VECTOR_CST element to type.
138 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
141 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
142 * config/i386/i386.md (builtin_setjmp_setup): Removed.
143 (builtin_longjmp): Likewise.
144 (save_stack_nonlocal): New pattern.
145 (restore_stack_nonlocal): Likewise.
147 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
150 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
151 Replace ASM_OUTPUT_LABEL with fprintf.
153 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
156 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
157 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
158 * config/i386/i386-c.c (ix86_target_macros_internal): Also
159 define __IBT__ and __SHSTK__ for -fcf-protection.
160 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
162 (ix86_trampoline_init): Likewise.
163 (x86_output_mi_thunk): Likewise.
164 (ix86_notrack_prefixed_insn_p): Likewise.
165 (ix86_option_override_internal): Don't disallow -fcf-protection.
166 * config/i386/i386.md (rdssp<mode>): Also enable for
168 (incssp<mode>): Likewise.
169 (nop_endbr): Likewise.
170 * config/i386/i386.opt (mcet): Change help message to built-in
174 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
175 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
176 enable CET built-in functions.
178 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
180 * common/config/i386/i386-common.c
181 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
182 OPTION_MASK_ISA_MOVDIRI_UNSET,
183 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
184 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
185 * config.gcc (movdirintrin.h): New header.
186 * config/i386/cpuid.h (bit_MOVDIRI,
187 bit_MOVDIR64B): New bits.
188 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
190 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
191 (VOID, PVOID, PCVOID)): New function types.
192 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
193 __builtin_ia32_directstoreu_u64,
194 __builtin_ia32_movdir64b): New builtins.
195 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
196 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
198 (ix86_valid_target_attribute_inner_p): Ditto.
199 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
200 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
201 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
202 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
203 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
204 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
205 (movdiri<mode>, movdir64b_<mode>): New.
206 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
207 * config/i386/immintrin.h: Include movdirintrin.h.
208 * config/i386/movdirintrin.h: New file.
209 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
211 2018-04-19 Richard Biener <rguenther@suse.de>
214 * cfg.c (clear_bb_flags): When loop state says we have
215 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
217 2018-04-19 Richard Biener <rguenther@suse.de>
219 PR tree-optimization/84737
220 * tree-vect-data-refs.c (vect_copy_ref_info): New function
221 copying restrict info.
222 (vect_setup_realignment): Use it.
223 * tree-vectorizer.h (vect_copy_ref_info): Declare.
224 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
225 the first DR to all generated stores.
226 (vectorizable_load): Likewise for loads.
228 2018-04-19 Jakub Jelinek <jakub@redhat.com>
230 PR tree-optimization/85446
231 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
232 the integral and pointer types to have the same precision.
234 * doc/install.texi: Document --disable-cet being the default and
237 2018-04-18 Martin Liska <mliska@suse.cz>
239 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
242 2018-04-18 Martin Liska <mliska@suse.cz>
245 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
248 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
249 arguments if they are comparable.
251 2018-04-18 Martin Liska <mliska@suse.cz>
254 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
257 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
260 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
263 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
264 ENDBR after calling __morestack.
266 2018-04-18 David Malcolm <dmalcolm@redhat.com>
269 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
270 by using gcc_base_ver to generate a gcc_driver_version, and use
271 it when generating GCC_DRIVER_NAME.
272 * configure: Regenerate.
274 2018-04-18 Jakub Jelinek <jakub@redhat.com>
277 * config.gcc: Obsolete powerpc*-*-*spe*.
279 2018-04-17 Jakub Jelinek <jakub@redhat.com>
282 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
283 (stabstr_D): Change type of unum from unsigned int to
284 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
287 2018-04-17 Jim Wilson <jimw@sifive.com>
290 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
291 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
292 Set arg_pointer_offset after using pretend_args_size.
294 2018-04-17 Jakub Jelinek <jakub@redhat.com>
296 PR rtl-optimization/85431
297 * dse.c (record_store): Ignore zero width stores.
300 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
301 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
302 __builtin_stack_restore rather than after it.
303 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
304 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
305 argument instead of virtual_dynamic_stack_rtx.
307 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
309 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
311 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
312 Add note to error message to explain internal mapping of overloaded
313 built-in function name to non-overloaded built-in function name.
314 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
317 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
320 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
321 where the inputs overlap with the output.
323 2018-04-17 Jakub Jelinek <jakub@redhat.com>
325 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
326 (=v, v) alternative and explicit "memory" attribute.
327 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
328 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
330 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
331 "sselog1" type instead of "sselog".
332 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
333 "sselog". Remove explicit "memory" attribute.
334 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
335 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
337 (vec_extract_hi_v32hi): Merge all alternatives into one, use
338 "sselog1" type instead of "sselog". Remove explicit "memory"
340 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
341 use "sselog1" type instead of "sselog". Remove explicit "memory"
343 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
344 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
346 (vec_extract_hi_v64qi): Merge all alternatives into one, use
347 "sselog1" type instead of "sselog". Remove explicit "memory"
349 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
350 use "sselog1" type instead of "sselog". Remove explicit "memory"
354 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
357 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
360 2018-04-17 Martin Jambor <mjambor@suse.cz>
363 * ipa-cp.c (create_specialized_node): Call
364 expand_all_artificial_thunks if necessary.
366 2018-04-17 Martin Liska <mliska@suse.cz>
369 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
370 in message, remote space in between '_G' and '('.
372 2018-04-17 Jakub Jelinek <jakub@redhat.com>
375 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
376 avx512f_vmcmp<mode>3<round_saeonly_name>,
377 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
378 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
379 avx512f_rndscale<mode><round_saeonly_name>,
380 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
381 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
382 Use %<iptr>2 instead of %2 for -masm=intel.
383 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
384 avx512f_vcvttss2usi<round_saeonly_name>,
385 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
387 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
388 avx512f_vcvttsd2usi<round_saeonly_name>,
389 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
390 Use %q1 instead of %1 for -masm=intel.
391 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
392 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
393 of %3 for -masm=intel.
394 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
396 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
398 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
400 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
401 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
403 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
405 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
406 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
407 %g1 and one with %0 and %1.
408 (avx512er_vmrcp28<mode><round_saeonly_name>,
409 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
411 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
412 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
413 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
414 of %0 and %{%4%} for -masm=intel.
415 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
416 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
417 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
418 order of %0 and %{%5%}%{z%} for -masm=intel.
420 2018-04-17 Jan Hubicka <jh@suse.cz>
423 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
425 2018-04-17 Martin Liska <mliska@suse.cz>
428 * multiple_target.c (create_dispatcher_calls): Set apostrophes
429 for target_clone error message. Make default implementation
430 clone to be a local declaration.
431 (separate_attrs): Add new argument and check for an empty
433 (expand_target_clones): Handle it.
434 (ipa_target_clone): Make redirection just for target_clones
437 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
438 Tom de Vries <tom@codesourcery.com>
441 * omp-expand.c (expand_oacc_for): Add dummy false branch for
442 tiled basic blocks without omp continue statements.
444 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
447 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
448 vec_extract expression as having side effects to make sure it gets
451 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
454 * config/i386/i386.c (get_builtin_code_for_version): Check
457 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
460 * gcc/config.gcc: Support "skylake".
461 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
463 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
464 (processor_target_table): Add "skylake".
465 (ix86_option_override_internal): Add "skylake".
466 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
467 PROCESSOR_CANNONLAKE.
468 (get_builtin_code_for_version): Fix priority for
469 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
470 PROCESSOR_SKYLAKE-AVX512.
471 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
472 (processor_type): Add PROCESSOR_SKYLAKE.
474 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
475 Jason Merrill <jason@redhat.com>
478 * convert.c (convert_to_integer_1): Use direct recursion for
479 enumeral types and types with a precision less than the number
480 of bits in their mode.
482 2018-04-16 Julia Koval <julia.koval@intel.com>
485 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
486 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
488 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
491 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
492 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
493 and -mno-direct-move.
495 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
498 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
499 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
500 Ensure negative shifts result in {0}.
502 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
504 PR rtl-optimization/79916
505 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
506 regs (if any) to define how to gnerate SD moves when LRA is in
509 2018-04-13 Jakub Jelinek <jakub@redhat.com>
511 PR rtl-optimization/85393
512 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
513 * except.c (expand_dw2_landing_pad_for_region): Make static.
514 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
515 a label and unconditional jump to old_bb, rather than
516 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
519 PR rtl-optimization/85376
520 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
521 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
522 instead of a specific value.
524 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
525 Bin Cheng <bin.cheng@arm.com>
527 PR tree-optimization/82965
528 PR tree-optimization/83991
529 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
530 by_profile_only parameter.
531 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
532 information if the loop was predicted to iterate too many times.
533 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
535 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
538 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
541 2018-04-13 Martin Liska <mliska@suse.cz>
542 Jakub Jelinek <jakub@redhat.com>
545 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
546 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
547 * builtins.c (expand_builtin_memory_copy_args): Use
548 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
549 handle dest_addr == pc_rtx.
551 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
554 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
555 asked to not generate direct moves.
556 (fix_trunc<mode>si2_stfiwx): Similar.
557 (fix_trunc<mode>si2_internal): Similar.
559 2018-04-12 Jakub Jelinek <jakub@redhat.com>
562 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
563 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
564 lookup if dest in some wider mode is known to be const0_rtx and
565 if so, record permanent equivalence for it to be ZERO_EXTEND of
566 the narrower mode destination.
568 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
570 * lto-streamer-out.c (output_function): Revert 259346.
571 * omp-expand.c (expand_oacc_for): Likewise.
573 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
575 PR rtl-optimization/85354
576 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
577 * sel-sched.c (sel_global_init): ... here.
579 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
582 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
583 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
584 mode for PE-COFF targets.
585 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
586 (i386_pe_asm_lto_end): Likewise.
587 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
588 (TARGET_ASM_LTO_END): Likewise.
589 * config/i386/winnt.c (saved_debug_info_level): New static variable.
590 (i386_pe_asm_lto_start): New function.
591 (i386_pe_asm_lto_end): Likewise.
593 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
594 Richard Biener <rguenther@suse.de>
597 * lto-streamer-out.c (output_function): Fix CFG loop state before
599 * omp-expand.c (expand_oacc_for): Handle calls to internal
600 functions like regular functions.
602 2018-04-12 Richard Biener <rguenther@suse.de>
605 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
606 for the early LTO debug to properly generate references to it
607 during DIE emission. Do not re-use that for the skeleton for
609 (dwarf2out_early_finish): Likewise.
611 2018-04-12 Jakub Jelinek <jakub@redhat.com>
615 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
616 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
617 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
618 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
619 and output is a reg, avoid creating invalid lowpart subreg, but
620 instead split into a 512-bit move. Don't split if not AVX512VL,
621 input is xmm16+ reg and output is a mem.
622 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
623 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
624 xmm16+ reg and output is a mem.
626 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
628 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
629 also for flag_dwarf2_cfi_asm.
631 2018-04-12 Jakub Jelinek <jakub@redhat.com>
633 PR rtl-optimization/85342
634 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
635 a bool scalar var inside of the loop instead. Don't try to update
636 recog_data.operand after failed apply_change_group.
638 2018-04-12 Tom de Vries <tom@codesourcery.com>
641 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
642 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
643 array with flexible array member as array without given dimension.
644 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
645 argument for undefined param to true.
647 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
650 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
651 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
652 from PowerPC section.
653 * config/rs6000/sysv4.opt (mcall-): Improve help text.
654 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
655 help text that is too long.
656 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
657 help text that is too long.
658 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
659 help text that is too long.
661 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
663 * config/alpha/alpha.md (stack_probe_internal): Rename
664 from "probe_stack". Update all callers.
666 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
668 PR rtl-optimization/84566
669 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
670 sched_macro_fuse_insns.
672 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
675 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
676 (compute_block_dependences): ... from here.
678 2018-04-11 Jakub Jelinek <jakub@redhat.com>
680 PR tree-optimization/85331
681 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
682 from int to HOST_WIDE_INT.
684 2018-04-11 Martin Jambor <mjambor@suse.cz>
687 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
688 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
689 not the same as the source val.
690 (cgraph_edge_brings_value_p): New parameter.
691 (gather_edges_for_value): Pass destination value to
692 cgraph_edge_brings_value_p.
693 (perhaps_add_new_callers): Likewise.
694 (get_info_about_necessary_edges): Likewise and exclude values brought
695 only by self-recursive edges.
696 (create_specialized_node): Redirect only clones of self-calling edges.
697 (+self_recursive_pass_through_p): New function.
698 (find_more_scalar_values_for_callers_subset): Use it.
699 (find_aggregate_values_for_callers_subset): Likewise.
700 (known_aggs_to_agg_replacement_list): Removed.
701 (decide_whether_version_node): Re-calculate known constants for all
702 remaining context clones.
704 2018-04-11 Richard Biener <rguenther@suse.de>
707 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
708 from early DWARF output.
709 (dwarf2out_early_finish): Output line info unconditionally into
710 early DWARF and add reference to it.
712 2018-04-11 Jakub Jelinek <jakub@redhat.com>
715 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
716 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
717 other than V2DFmode using iptr mode attribute.
718 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
720 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
722 PR rtl-optimization/84659
723 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
725 2018-04-11 Jakub Jelinek <jakub@redhat.com>
728 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
730 (output_loc_list): Pass address of a dummy size variable even in the
731 locview handling loop.
732 (index_location_lists): Add comment on why skip_loc_list_entry can't
735 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
738 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
741 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
744 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
745 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
746 and -mstring-compare-inline-limit.
748 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
751 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
752 for stack clash protection in a register whenever we need it to be in
755 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
757 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
758 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
760 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
763 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
765 (mlong-double-): Ditto.
766 * config/rs6000/sysv4.opt (msdata=): Ditto.
769 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
771 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
772 erroneous entries for
773 "vector int vec_ldl (int, long int *)", and
774 "vector unsigned int vec_ldl (int, unsigned long int *)".
775 Add comments and entries for
776 "vector bool char vec_ldl (int, bool char *)",
777 "vector bool short vec_ldl (int, bool short *)",
778 "vector bool int vec_ldl (int, bool int *)",
779 "vector bool long long vec_ldl (int, bool long long *)",
780 "vector pixel vec_ldl (int, pixel *)",
781 "vector long long vec_ldl (int, long long *)",
782 "vector unsigned long long vec_ldl (int, unsigned long long *)".
783 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
784 type tree bool_long_long_type_node and correct definition of
785 bool_V2DI_type_node to make reference to this new type tree.
786 (rs6000_mangle_type): Replace erroneous reference to
787 bool_long_type_node with bool_long_long_type_node.
788 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
789 comments to emphasize sign distinctions for char and int types and
790 replace RS6000_BTI_bool_long constant with
791 RS6000_BTI_bool_long_long constant. Also add comment to restrict
792 use of RS6000_BTI_pixel.
793 (bool_long_type_node): Remove this macro definition.
794 (bool_long_long_type_node): New macro definition
796 2018-04-10 Jakub Jelinek <jakub@redhat.com>
798 PR rtl-optimization/85300
799 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
800 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
801 simplify_unary_operation fails.
803 2018-04-10 Martin Liska <mliska@suse.cz>
805 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
806 cgraph_edge and ipa_ref.
808 2018-04-10 Jakub Jelinek <jakub@redhat.com>
813 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
814 computation of the VEC_MERGE selector from mask.
815 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
816 Fix decoding of the VEC_MERGE selector into mask.
818 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
820 PR tree-optimization/85286
821 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
823 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
825 * final.c (final_1): Set insn_last_address as well as
826 insn_current_address.
828 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
831 * explow.c (emit_stack_probe): Call validize_mem on memory location
832 before passing it to gen_probe_stack. Create address operand and
833 legitimize it for the probe_stack_address case.
835 2018-04-09 Jan Hubicka <jh@suse.cz>
838 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
839 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
840 * tree.c (free_lang_data_in_type): Fix handling of binfos;
842 (free_lang_data): Rebuild type inheritance graph.
844 2018-04-09 Martin Sebor <msebor@redhat.com>
846 * invoke.texi (-finline-small-functions): Mention other optimization
848 (-findirect-inlining, -fpartial-inlining): Same.
849 (-finline-functions-called-once): Same.
850 (-freorder-blocks-and-partition): Same.
852 2018-04-09 Jan Hubicka <jh@suse.cz>
855 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
856 jumps; choose last target that matches the criteria (i.e.
857 no partition changes for non-crossing jumps).
858 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
859 support for redirecting crossing jumps to non-crossing.
861 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
863 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
864 also for naked functions.
866 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
868 * config/arc/arc.md (add_shift): New pattern.
869 (add_shift2): Likewise.
870 (sub_shift): Likewise.
871 (sub_shift_cmp0_noout): Likewise.
872 (compare_si_ashiftsi): Likewise.
873 (xbfu_cmp0_noout): New combine pattern.
874 (xbfu_cmp0"): Likewise.
875 (movsi_set_cc_insn): Place the predicable variant first.
876 (commutative_binary_cmp0_noout): Remove clobber.
877 (commutative_binary_cmp0): New pattern.
878 (noncommutative_binary_cmp0): Likewise.
879 (noncommutative_binary_cmp0_noout): Likewise.
880 (noncommutative_binary_comparison_result_used): Removed.
881 (rsub_cmp0): New pattern.
882 (rsub_cmp0_noout): Likewise.
883 (extzvsi): Changed, keep only meaningful variants.
884 (SQH, SEZ): New iterators.
885 (SQH_postfix): New mode attribute.
886 (SEZ_prefix): New code attribute.
887 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
888 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
889 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
891 (noncommutative_operator): Check the availability of barrel
894 2018-04-09 Richard Biener <rguenther@suse.de>
896 PR tree-optimization/85284
897 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
898 Only use the niter constraining form of simple_iv when the exit
901 2018-04-09 Tom de Vries <tom@codesourcery.com>
904 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
905 (define_expand "*memory_barrier"): New define_expand.
906 (define_insn "memory_barrier"): New insn.
908 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
910 PR rtl-optimization/80463
911 PR rtl-optimization/83972
912 PR rtl-optimization/83480
914 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
915 correct producer for the insn.
916 (tidy_control_flow): Fixup seqnos in case of debug insns.
918 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
920 PR rtl-optimization/83913
922 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
923 different sched-times when merging exprs.
925 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
927 PR rtl-optimization/83962
929 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
930 tidy_fallthru_edge and tidy_control_flow.
932 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
934 PR rtl-optimization/83530
936 * sel-sched.c (force_next_insn): New global variable.
937 (remove_insn_for_debug): When force_next_insn is true, also leave only
938 next insn in the ready list.
939 (sel_sched_region): When the region wasn't scheduled, make another pass
940 over it with force_next_insn set to 1.
942 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
944 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
946 * config/nds32/constants.md (unspec_volatile_element): Add enum values
947 for interrupt control.
948 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
949 functions for interrupt control.
950 * config/nds32/nds32-intrinsic.md: Likewise.
951 * config/nds32/nds32_intrinsic.h: Likewise.
952 * config/nds32/nds32.h (nds32_builtins): Likewise.
954 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
956 * config/nds32/nds32.c (nds32_init_machine_status,
957 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
958 strict_aligned_p field.
959 (nds32_expand_to_rtl_hook): New function.
960 (TARGET_EXPAND_TO_RTL_HOOK): Define.
961 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
963 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
964 Chung-Ju Wu <jasonwucj@gmail.com>
966 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
967 * config/nds32/nds32-n7.md: New file.
968 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
969 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
971 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
972 * config/nds32/nds32.md (pipeline_model): Add n7.
973 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
974 * config/nds32/pipelines.md: Include n7 settings.
976 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
977 Chung-Ju Wu <jasonwucj@gmail.com>
979 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
980 * config/nds32/nds32-e8.md: New file.
981 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
982 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
984 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
985 * config/nds32/nds32.md (pipeline_model): Add e8.
986 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
987 * config/nds32/pipelines.md: Include e8 settings.
989 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
990 Chung-Ju Wu <jasonwucj@gmail.com>
992 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
993 * config/nds32/nds32-n8.md: New file.
994 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
995 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
997 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
998 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
999 * config/nds32/nds32.md (pipeline_model): Add n8.
1000 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1001 * config/nds32/pipelines.md: Include n8 settings.
1003 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1004 Chung-Ju Wu <jasonwucj@gmail.com>
1006 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1007 * config/nds32/nds32-n9-2r1w.md: New file.
1008 * config/nds32/nds32-n9-3r2w.md: New file.
1009 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1010 nds32_register_ports): New or modify for cpu n9.
1011 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1013 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1014 * config/nds32/nds32-utils.c: New file.
1015 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1016 TARGET_MUL_SLOW): Define.
1017 * config/nds32/nds32.md (pipeline_model): New attribute.
1018 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1019 New options that support cpu n9.
1020 * config/nds32/pipelines.md: Include n9 settings.
1021 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1023 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1025 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1026 information if necessary.
1027 (output_cond_branch_compare_zero): Likewise.
1028 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1029 (nds32_target_alignment): Refine for alignment.
1030 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1031 (FUNCTION_BOUNDARY): Modify.
1032 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1034 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1036 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1038 * config/nds32/constants.md (unspec_volatile_element): Add values for
1039 TLB operation and data prefetch.
1040 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1041 functions for TLB operation and data prefetch.
1042 * config/nds32/nds32-intrinsic.md: Likewise.
1043 * config/nds32/nds32_intrinsic.h: Likewise.
1044 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1045 (nds32_print_operand): Likewise.
1046 * config/nds32/nds32.h (nds32_builtins): Likewise.
1048 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1049 Andrew Pinski <pinsika@gcc.gnu.org>
1052 * match.pd: Use constant_boolean_node of correct type instead of
1053 boolean_true_node or boolean_false_node for simplifying
1054 pointer comparisons to zero.
1056 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1058 PR tree-optimization/80021
1059 * tree.c (verify_type_variant): Make error call in verify_variant_match
1060 translatable and remove final full stop.
1062 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1064 * config/nds32/constants.md (unspec_volatile_element): Add
1065 UNSPEC_VOLATILE_EH_RETURN.
1066 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1067 nds32_output_stack_pop): Support dwarf exception handling process.
1068 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1069 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1070 exception handling process.
1071 (nds32_compute_stack_frame): Likewise.
1072 (nds32_return_addr_rtx): Likewise.
1073 (nds32_initial_elimination_offset): Likewise.
1074 (nds32_expand_prologue): Likewise.
1075 (nds32_expand_epilogue): Likewise.
1076 (nds32_dynamic_chain_address): New function.
1077 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1079 (DYNAMIC_CHAIN_ADDRESS): Define.
1080 (EH_RETURN_DATA_REGNO): Define.
1081 (EH_RETURN_STACKADJ_RTX): Define.
1082 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1083 patterns for dwarf exception handling.
1085 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1087 * config/nds32/nds32.h: Clean up obsolete macros.
1089 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1091 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1092 Add enum values for particular instructions.
1093 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1094 particular intrinsic functions.
1095 * config/nds32/nds32-intrinsic.md: Likewise.
1096 * config/nds32/nds32_intrinsic.h: Likewise.
1097 * config/nds32/nds32.h (nds32_builtins): Likewise.
1098 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1099 (btst, ave): New patterns for particular instructions.
1101 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1103 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1104 Add enum values for atomic load/store and memory sync.
1105 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1107 * config/nds32/nds32-intrinsic.md: Likewise.
1108 * config/nds32/nds32_intrinsic.h: Likewise.
1109 * config/nds32/nds32.h (nds32_builtins): Likewise.
1111 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1113 PR tree-optimization/85257
1114 * fold-const.c (native_encode_vector): If not all elts could fit
1115 and off is -1, return 0 rather than offset.
1116 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1117 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1118 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1119 adjust buffer in native_interpret_expr call.
1121 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1123 * config/nds32/constants.md (unspec_volatile_element): Add cache
1124 control enum values.
1125 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1126 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1127 * config/nds32/nds32.c (nds32_cctl_names): New.
1128 (nds32_print_operand): Handle cache control register names.
1129 * config/nds32/nds32.h (nds32_builtins): New enum values.
1130 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1132 * config/nds32/nds32.md (type): Add mmu.
1133 * config/nds32/pipelines.md (simple_insn): Add mmu.
1135 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1137 * config/nds32/nds32.md (type): Remove call.
1138 * config/nds32/pipelines.md (simple_insn): Likewise.
1140 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1142 * config/nds32/constants.md (unspec_volatile_element): Add
1143 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1144 UNSPEC_VOLATILE_FMFCFG.
1145 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1146 description for fmfcfg and fmfcsr.
1147 (bdesc_1arg): Add fmtcsr.
1148 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1149 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1150 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1151 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1152 unspec_fmfcfg): New patterns.
1153 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1154 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1155 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1156 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1157 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1158 __nds32__fmfcfg): Define.
1160 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1162 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1163 intrinsic register names.
1164 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1165 intrinsic register enum values and macros.
1167 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1169 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1170 for load/store addressing form.
1171 (nds32_print_operand_address): Likewise.
1173 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1176 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1177 based on LABEL_REF. Remove useless assertion.
1178 (pic_address_needs_scratch): Fix formatting.
1179 (sparc_legitimize_pic_address): Minor tweaks.
1180 (sparc_delegitimize_address): Adjust assertion accordingly.
1181 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1182 into symbolic_operand.
1183 (movsi_high_pic_label_ref): Likewise.
1184 (movsi_lo_sum_pic_label_ref): Likewise.
1185 (movdi_pic_label_ref): Likewise.
1186 (movdi_high_pic_label_ref): Likewise.
1187 (movdi_lo_sum_pic_label_ref): Likewise.
1189 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1191 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1192 custom LIB_SPEC setup.
1194 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1195 Kito Cheng <kito.cheng@gmail.com>
1197 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1198 * config/riscv/freebsd.h: New.
1200 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1202 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1203 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1206 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1207 Kito Cheng <kito.cheng@gmail.com>
1209 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1210 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1211 * config/nds32/nds32-protos.h (nds32_output_call,
1212 nds32_output_return): Declare.
1213 * config/nds32/nds32.md: Refine all the call and return patterns.
1215 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1218 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1219 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1221 PR rtl-optimization/84872
1222 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1223 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1226 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1228 * expr.c (copy_blkmode_to_reg): Revert 254862.
1229 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1231 2018-04-06 Richard Biener <rguenther@suse.de>
1234 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1235 after seeing a component reference with an adjacent field. Treat
1236 refs to arrays at struct end of external decls similar to
1237 refs to unconstrained commons.
1239 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1242 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1243 look through SAVE_EXPRs with non-side-effects argument. Adjust
1245 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1248 2018-04-06 Richard Biener <rguenther@suse.de>
1251 * alias.c (find_base_term): New wrapper around find_base_term
1252 unwinding CSELIB_VAL_PTR changes.
1253 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1256 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1258 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1260 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1261 constant definitions.
1262 ("nop"): lr 0,0 -> nopr r0
1263 ("nop_lr0", "nop_lr1"): New insn definitions.
1265 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1267 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1268 NDS32_V3PUSH_AVAILABLE_P macro.
1270 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1271 Chung-Ju Wu <jasonwucj@gmail.com>
1273 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1274 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1275 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1276 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1277 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1278 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1279 * config/nds32/constraints.md: New constraints and checking for hard
1280 float configuration.
1281 * config/nds32/iterators.md: New mode iterator and attribute for hard
1282 float configuration.
1283 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1285 * config/nds32/nds32-fpu.md: New file.
1286 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1287 deal with hard float code generation.
1288 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1290 (abi_type, float_reg_number): New enum type.
1291 * config/nds32/nds32-predicates.c: New predicates for hard float.
1292 * config/nds32/nds32-protos.h: Declare functions for hard float.
1293 * config/nds32/nds32.c: Implementation for hard float configuration.
1294 * config/nds32/nds32.h: Definitions for hard float configuration.
1295 * config/nds32/nds32.md: Include hard float machine description and
1296 modify patterns for hard float configuration.
1297 * config/nds32/nds32.opt: New options for hard float configuration.
1298 * config/nds32/predicates.md: New predicates for hard float
1301 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1303 * common/config/nds32/nds32-common.c
1304 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1306 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1309 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1310 CONSTRUCTOR_ELT (ctor, ...)->value.
1312 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1315 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1317 2018-04-05 Tom de Vries <tom@codesourcery.com>
1320 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1323 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1324 Kito Cheng <kito.cheng@gmail.com>
1326 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1327 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1328 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1329 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1331 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1332 Kito Cheng <kito.cheng@gmail.com>
1334 * config/nds32/constraints.md (Ufe): New memory constraint.
1335 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1336 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1337 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1339 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1340 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1342 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1344 * config/nds32/nds32.md: Use optimize_size in the condition for
1345 alu-shift instructions.
1347 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1349 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1351 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1353 * config/nds32/nds32.md (negsi2): Refine pattern.
1355 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1356 Chung-Ju Wu <jasonwucj@gmail.com>
1358 * config/nds32/iterators.md (shift_rotate): New code iterator.
1359 (shift): New code attribute.
1360 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1361 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1362 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1363 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1364 bit-wise operations.
1365 (andsi3, *andsi3): Ditto.
1366 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1367 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1368 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1369 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1370 nds32_ior_operand, nds32_xor_operand): New predicates.
1372 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1374 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1375 (addsi3, subsi3): ... this.
1377 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1379 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1381 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1383 * config/nds32/nds32.md: Adjust indention.
1385 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1387 * config/nds32/nds32.md (feature): New attribute.
1389 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1391 * config/nds32/nds32.md (subtype): New attribute.
1393 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1396 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1397 expansion to perform a bitwise AND of the argument followed by a
1398 boolean negation of the result.
1400 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1402 PR rtl-optimization/84878
1403 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1404 the basic block. Assert the use reference is not artificial and that
1405 it has an associated insn.
1407 2018-04-04 Michael Matz <matz@suse.de>
1409 * builtins.c (compute_objsize): Pass correct operand
1410 to array_at_struct_end_p.
1412 2018-04-04 Richard Biener <rguenther@suse.de>
1415 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1416 from contexts for DINFO_LEVEL_TERSE and below.
1418 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1420 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1421 resiter_operand condition.
1422 * config/nds32/nds32.md (*move<mode>): Ditto.
1424 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1425 Monk Chiang <sh.chiang04@gmail.com>
1427 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1429 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1431 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1433 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1434 Kito Cheng <kito.cheng@gmail.com>
1436 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1437 nds32_cond_code_str, output_cond_branch,
1438 output_cond_branch_compare_zero, nds32_expand_cbranch,
1439 nds32_expand_cstore, nds32_expand_movcc,
1440 nds32_output_cbranchsi4_equality_zero,
1441 nds32_output_cbranchsi4_equality_reg,
1442 nds32_output_cbranchsi4_equality_reg_or_const_int,
1443 nds32_output_cbranchsi4_greater_less_zero: New functions.
1444 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1445 nds32_expand_cstore, nds32_expand_movcc,
1446 nds32_output_cbranchsi4_equality_zero,
1447 nds32_output_cbranchsi4_equality_reg,
1448 nds32_output_cbranchsi4_equality_reg_or_const_int,
1449 nds32_output_cbranchsi4_greater_less_zero): Declare.
1450 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1451 nds32_rimm11s_operand): New predicates.
1452 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1453 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1456 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1458 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1459 * config/nds32/nds32.md: Ditto.
1460 * config/nds32/pipelines.md: Ditto.
1462 2018-04-04 Richard Biener <rguenther@suse.de>
1464 PR tree-optimization/85168
1465 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1466 propagating abnormals.
1468 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1470 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1472 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1473 Kito Cheng <kito.cheng@gmail.com>
1475 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1476 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1477 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1478 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1479 * config/nds32/nds32.md (sibcall_internal): New.
1480 (sibcall_register): Remove.
1481 (sibcall_immediate): Remove.
1482 (sibcall_value_internal): New.
1483 (sibcall_value_register): Remove.
1484 (sibcall_value_immediate): Remove.
1485 * config/nds32/predicates.md (nds32_general_register_operand): New.
1486 (nds32_call_address_operand): New.
1488 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1490 PR rtl-optimization/85167
1491 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1492 bb_defs if *split_p, instead preinitialize it to NULL.
1494 PR tree-optimization/85156
1495 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1496 evaluating the argument multiple times.
1498 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1500 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1502 (_mm_cvtpd_ps): Likewise.
1503 (_mm_cvttpd_epi32): Likewise.
1504 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1505 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1506 vector, pixel, and bool following altivec.h include.
1508 2018-04-03 Martin Sebor <msebor@redhat.com>
1510 * doc/extend.texi (Common Function Attributes): Clarify.
1511 (const attribute): Likewise.
1512 (pure attribute): Likewise.
1514 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1517 * config/i386/i386.c (ix86_expand_vector_set): Use
1518 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1520 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1522 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1523 instructions when changing rounding bits to preserve precision bits
1524 in the x87 control word.
1526 2018-04-03 Martin Liska <mliska@suse.cz>
1528 PR tree-optimization/82491
1529 * rtl.h (strip_offset_and_add): Replace += suboffset with
1530 poly_uint64 () + suboffset.
1532 2018-03-29 Martin Liska <mliska@suse.cz>
1533 Martin Jambor <mjambor@suse.cz>
1536 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1537 param_type is not an integral or pointer type.
1539 2018-04-03 Richard Biener <rguenther@suse.de>
1541 * sese.h (recompute_all_dominators): Remove.
1543 2018-04-02 Martin Sebor <msebor@redhat.com>
1545 * doc/invoke.texi (-Wrestrict): Fix typos.
1547 2018-04-02 Jim Wilson <jimw@sifive.com>
1549 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1550 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1551 (<optab>di3, <optab>si3_extend): Likewise.
1552 (<optab>si3_mask, <optab>si3_mask_1): New.
1553 (<optab>di3_mask, <optab>di3_mask_1): New.
1554 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1555 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1556 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1558 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1560 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1563 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1565 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1566 (nds32_canonicalize_comparison): New function.
1568 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1569 Kito Cheng <kito.cheng@gmail.com>
1570 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1572 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1573 * config/nds32/constants.md (unspec_volatile_element): Add
1574 UNSPEC_VOLATILE_RELAX_GROUP.
1575 * config/nds32/nds32-relax-opt.c: New file.
1576 * config/nds32/nds32-predicates.c
1577 (nds32_symbol_load_store_p): New function.
1578 * config/nds32/nds32-protos.h
1579 (nds32_symbol_load_store_p): Declare function.
1580 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1581 * config/nds32/nds32.c
1582 (nds32_register_pass): New function to register pass.
1583 (nds32_register_passes): New function to register passes.
1584 * config/nds32/nds32.md (relax_group): New pattern.
1585 * config/nds32/nds32.opt (mrelax-hint): New option.
1586 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1588 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1590 * config/nds32/t-nds32: Modify files dependency.
1592 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1594 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1595 (PROFILE_HOOK): Define its implementation.
1597 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1599 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1600 type and 32-bit size.
1602 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1605 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1606 (V_128_256): New mode iterator.
1607 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1608 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1609 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1611 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1612 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1614 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1617 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1618 NaN inputs correctly.
1620 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1623 * config/rs6000/vsx.md (??r): New mode attribute.
1624 (*vsx_mov<mode>_64bit): Use it.
1625 (*vsx_mov<mode>_32bit): Likewise.
1627 2018-03-30 Martin Sebor <msebor@redhat.com>
1629 PR tree-optimization/84818
1630 * builtins.c (check_access): Use warning_n.
1632 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1635 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1637 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1640 2018-03-30 Julia Koval <julia.koval@intel.com>
1643 * x86-tune.def (movx, partial_reg_dependency): Enable for
1646 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1649 * lra-constraints.c (process_alt_operands): Move setting
1650 this_alternative_matches below.
1652 2018-03-29 Martin Liska <mliska@suse.cz>
1655 * doc/invoke.texi: Document how LTO works with debug info.
1656 Describe auto-load support of binutils. Mention 'x86-64'
1657 as valid option value of -march option.
1659 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1661 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1664 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1665 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1666 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1669 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1672 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1673 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1674 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1675 for RS6000_BTM_POWERPC64.
1676 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1677 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1678 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1683 2018-03-28 Carl Love <cel@us.ibm.com>
1686 2017-09-27 Carl Love <cel@us.ibm.com>
1688 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1689 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1690 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1693 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1695 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1696 instead of __vector bool.
1697 (_mm_max_pu8): Likewise.
1698 (_mm_min_pi16): Likewise.
1700 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1703 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1704 (DIVWEUO): Likewise.
1706 (DIVDEUO): Likewise.
1707 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1708 DIVWEUO and DIVDEUO.
1709 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1710 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1711 (div_extend): Likewise.
1712 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1714 (__builtin_divweuo): Likewise.
1715 (__builtin_divdeo): Likewise.
1716 (__builtin_divdeuo): Likewise.
1718 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1721 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1722 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1724 PR tree-optimization/82004
1725 * gimple-match-head.c (optimize_pow_to_exp): New function.
1726 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1727 Don't fold to exp if optimize_pow_to_exp is false.
1729 2018-03-28 Martin Liska <mliska@suse.cz>
1732 * calls.c (initialize_argument_information): Fix trailing space.
1733 * common.opt: Fix typo and provide better explanation for
1734 -fsanitize-coverage option.
1735 * config/i386/i386.opt: Fix typo.
1737 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1738 Martin Liska <mliska@suse.cz>
1741 * gimplify.c (asan_poison_variable): Don't do the check for
1742 gimplify_omp_ctxp here.
1743 (gimplify_decl_expr): Do it here.
1744 (gimplify_target_expr): Likewise.
1746 2018-03-28 Martin Liska <mliska@suse.cz>
1749 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1750 chkp_type_bounds_count if MPX is not enabled.
1752 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1754 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1756 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1759 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1760 function to create the function decl for complex long double
1761 multiply and divide for -mabi=ieeelongdouble.
1762 (init_float128_ieee): Call it.
1764 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1767 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1768 -fcf-protection=branch -mibt.
1769 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1771 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1774 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1776 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1779 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1780 extern array declarations.
1782 2018-03-27 Richard Biener <rguenther@suse.de>
1785 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1786 explicit single_use checks.
1788 2018-03-27 Richard Biener <rguenther@suse.de>
1790 PR tree-optimization/85082
1791 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1794 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1796 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1797 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1798 Turn on fasynchronous-unwind-tables and funwind-tables.
1800 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1803 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1804 (*bmi_blsr_<mode>_ccz): Ditto.
1806 2018-03-26 Tom de Vries <tom@codesourcery.com>
1808 PR tree-optimization/85063
1809 * omp-general.c (offloading_function_p): New function. Factor out
1811 * omp-offload.c (pass_omp_target_link::gate): ... here.
1812 * omp-general.h (offloading_function_p): Declare.
1813 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1814 with attribute omp declare target for offloading functions.
1816 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1818 PR tree-optimization/84005
1819 * tree-data-ref.h (get_base_for_alignment): Declare.
1820 * tree-data-ref.c (get_base_for_alignment_1): New function.
1821 (get_base_for_alignment): Likewise.
1822 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1823 get_base_for_alignment to find a suitable base object, instead
1824 of always using drb->base_address.
1826 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1829 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1830 known size by default.
1832 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1835 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1836 and non BLKmode operands.
1838 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1841 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1842 Clean up attributes.
1844 2018-03-23 Richard Biener <rguenther@suse.de>
1847 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1848 we are going to emit early debug for LTO.
1850 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1853 * function.c (match_asm_constraints_1): Don't optimize if input
1854 doesn't satisfy general_operand predicate for output's mode.
1857 * alias.c (write_dependence_p): Don't require for x_canonicalized
1858 non-VOIDmode if x has VOIDmode.
1861 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1862 just don't try to optimize it rather than assert it never happens.
1864 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1866 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1867 macro expansions for definition of ST_INTERNAL_<mode> and
1868 LD_INTERNAL_<mode> builtins.
1869 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1871 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1873 (altivec_expand_st_builtin): Likewise.
1874 (altivec_expand_builtin): Remove calls to deleted functions.
1875 (rs6000_address_for_altivec): Delete this function.
1876 * config/rs6000/vector.md: Remove expands for
1877 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1879 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1882 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1883 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1884 re-computing once computed.
1885 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1886 (arm_init_machine_status): Initialize
1887 machine->static_chain_stack_bytes.
1889 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1892 * doc/extend.texi: Add four new prototypes for vec_ld.
1893 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1894 definitions for more logical presentation.
1895 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1896 entries for V1TI variants of __builtin_altivec_ld builtin.
1897 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1898 handling of V1TI variant of LVX icode pattern.
1899 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1900 (rs6000_gimple_fold_builtin): Likewise.
1901 (altivec_init_builtins): Add code to define
1902 __builtin_altivec_lvx_v1ti function.
1904 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1907 * function.c (match_asm_constraints_1): Don't do the optimization
1908 if input isn't a REG, SUBREG, MEM or constant.
1910 2018-03-22 Tom de Vries <tom@codesourcery.com>
1912 PR tree-optimization/84956
1913 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1914 bb_has_abnormal_pred.
1916 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1919 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1920 DECL_INITIAL (decl) to decl at the end.
1921 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1924 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1926 * doc/extend.texi (__builtin_tgmath): Document when complex
1927 integer types are treated as _Complex _Float64.
1929 2018-03-21 Tom de Vries <tom@codesourcery.com>
1931 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1933 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1935 PR tree-optimization/84960
1936 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1937 if it is ENTRY block, move them into single succ of ENTRY in that case.
1939 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1941 PR tree-optimization/84811
1942 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1943 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1944 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1945 (known_subrange_p): Update accordingly. Cast each value involved
1946 in the size comparison, rather than casting the result of the
1949 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1951 PR tree-optimization/84982
1952 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1953 by flipping the least significant bit rather than all bits from
1954 bitpos to bitpos + bitsize - 1.
1956 2018-03-21 Nathan Sidwell <nathan@acm.org>
1958 * doc/extend.texi (Deprecated Features): Remove mention of
1959 long-deleted deprecations.
1961 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1964 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1965 * configure: Regenerate.
1967 2018-03-21 Tom de Vries <tom@codesourcery.com>
1969 PR tree-optimization/83126
1970 * tree-parloops.c (num_phis): New function.
1971 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1973 2018-03-21 Nathan Sidwell <nathan@acm.org>
1975 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1976 mention anon-struct/union members and trailing attributes.
1978 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1980 PR tree-optimization/84969
1981 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1982 builtin memset partitions if they set different rhs values.
1984 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1986 PR rtl-optimization/84989
1987 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1988 VEC_DUPLICATE with scalar result mode.
1990 2018-03-21 Martin Liska <mliska@suse.cz>
1993 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1994 not intended return statement.
1996 2018-03-21 Martin Liska <mliska@suse.cz>
1999 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2000 (chkp_find_bound_slots_1): Limit number of iterations.
2002 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2005 * Minor grammar fixes for x86 options.
2007 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2010 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2011 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2013 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2016 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2017 (altivec_lvx_<mode>_1op): Likewise.
2018 (altivec_stvx_<mode>_2op): Likewise.
2019 (altivec_stvx_<mode>_1op): Likewise.
2020 (altivec_lvx_<VM2:mode>): New define_expand.
2021 (altivec_stvx_<VM2:mode>): Likewise.
2022 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2023 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2024 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2025 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2026 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2027 (rs6000_gen_lvx): Likewise.
2028 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2029 (altivec_expand_stv_builtin): Likewise.
2030 (altivec_expand_builtin): Likewise.
2031 * config/rs6000/vector.md: Likewise.
2033 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2036 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2039 2018-03-20 Richard Biener <rguenther@suse.de>
2042 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2043 sign-conversions as zero, fall back to standard scalar_stmt
2046 2018-03-20 Martin Liska <mliska@suse.cz>
2049 * predict.c (rebuild_frequencies): Handle case when we have
2050 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2052 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2055 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2056 flag_section_anchors.
2057 * varasm.c (use_blocks_for_decl_p): Remove hack for
2058 dw2_force_const_mem.
2061 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2063 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2064 be created, use lowpart_subreg of operands[0] rather than operands[0]
2066 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2067 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2068 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2069 and n constraint instead of aarch64_shift_imm_di and Usd.
2070 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2071 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2073 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2076 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2077 to favor GPR over NEON registers.
2078 (<shift>di3_neon): Likewise.
2080 2018-03-20 Tom de Vries <tom@codesourcery.com>
2083 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2084 (nvptx_process_pars): Emit bar.sync asap and alap.
2086 2018-03-20 Tom de Vries <tom@codesourcery.com>
2089 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2090 seen_label if seen_label is already set.
2092 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2095 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2096 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2097 Use 1U instead of 1. Formatting fixes.
2100 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2101 instead of TREE_TYPE (s1) for the return value.
2103 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2105 PR tree-optimization/84946
2106 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2107 bitsize + bitsize in poly_uint64 rather than poly_int64.
2110 * dwarf2asm.c: Include fold-const.c.
2111 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2112 of decl rather than decl itself.
2114 PR rtl-optimization/84643
2115 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2117 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2120 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2121 calling assemble_variable.
2123 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2126 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2127 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2129 2018-03-19 Jim Wilson <jimw@sifive.com>
2132 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2133 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2134 (riscv_first_stack_step): Likewise.
2135 (riscv_option_override): Use STACK_BOUNDARY instead of
2137 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2139 (BIGGEST_ALIGNMENT): Set to 128.
2140 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2141 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2144 2018-03-19 Richard Biener <rguenther@suse.de>
2146 PR tree-optimization/84933
2147 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2148 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2150 2018-03-19 Richard Biener <rguenther@suse.de>
2152 PR tree-optimization/84859
2153 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2154 (cond_if_else_store_replacement): Perform sinking operation on
2155 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2156 Generalize what a BB with a single eligible store is.
2158 2018-03-19 Richard Biener <rguenther@suse.de>
2160 PR tree-optimization/84929
2161 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2162 chrec_is_positive against non-chrec arg.
2164 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2167 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2169 2018-03-18 Martin Liska <mliska@suse.cz>
2171 PR rtl-optimization/84635
2172 * regrename.c (build_def_use): Use matches_mode only when
2175 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2177 PR tree-optimization/84913
2178 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2179 vectorize chains of COND_EXPRs.
2181 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2183 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2185 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2187 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2189 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2191 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2193 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2194 Kito Cheng <kito.cheng@gmail.com>
2196 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2197 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2198 (nds32_adjust_reg_alloc_order): New function.
2199 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2201 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2203 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2204 nds32_print_operand, nds32_print_operand_address): Use
2205 HOST_WIDE_INT_PRINT_DEC instead.
2207 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2209 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2211 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2214 * config/i386/i386.c (initial_ix86_tune_features,
2215 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2217 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2218 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2219 rather than 1u << ix86_tune. Formatting fix.
2220 (ix86_option_override_internal): Change ix86_arch_mask from
2221 unsigned int to unsigned HOST_WIDE_INT, initialize to
2222 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2223 (ix86_function_specific_restore): Likewise.
2225 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2228 * postreload.c (reload_combine_recognize_pattern): Perform
2229 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2230 truncate_int_for_mode the result for the destination's mode.
2233 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2234 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2236 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2239 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2242 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2243 regno_allocno_class_array and sorted_pseudos.
2244 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2245 insns where regno is used.
2247 2018-03-16 Martin Liska <mliska@suse.cz>
2250 * multiple_target.c (create_dispatcher_calls): Redirect
2251 reference in the symbol table.
2253 2018-03-16 Martin Liska <mliska@suse.cz>
2256 * multiple_target.c (create_dispatcher_calls): Redirect also
2259 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2263 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2264 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2265 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2267 2018-03-16 Julia Koval <julia.koval@intel.com>
2269 * doc/invoke.texi (Skylake Server): Add CLWB.
2270 Cannonlake): Remove CLWB.
2272 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2274 PR tree-optimization/84841
2275 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2277 (FLOAT_ONE_CONST_TYPE): Define.
2278 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2279 (sort_by_operand_rank): Put entries with higher constant_type last
2280 rather than first to match comments.
2282 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2284 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2287 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2290 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2291 check and use address of target always.
2293 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2296 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2297 (indirect_thunk_bnd_needed): Likewise.
2298 (indirect_thunks_used): Likewise.
2299 (indirect_thunks_bnd_used): Likewise.
2300 (indirect_return_needed): New.
2301 (indirect_return_bnd_needed): Likewise.
2302 (output_indirect_thunk_function): Add a bool argument for
2304 (output_indirect_thunk_function): Don't generate alias for
2305 function return thunk.
2306 (ix86_code_end): Call output_indirect_thunk_function to generate
2307 function return thunks.
2308 (ix86_output_function_return): Set indirect_return_bnd_needed
2309 and indirect_return_needed instead of indirect_thunk_bnd_needed
2310 and indirect_thunk_needed.
2312 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2314 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2315 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2316 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2318 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2319 Paul Hua <paul.hua.gm@gmail.com>
2322 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2324 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2326 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2327 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2330 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2333 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2334 instead of GET_MODE_SIZE when comparing Units.
2336 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2339 * varasm.c (hash_section): Return an unchangeble hash value
2340 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2341 Return !aarch64_can_use_per_function_literal_pools_p ().
2343 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2346 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2347 rather than address of cmode as last argument to prepare_cmp_insn.
2349 2018-03-15 Julia Koval <julia.koval@intel.com>
2351 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2352 F_AVX512VNNI, F_AVX512BITALG): New.
2354 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2357 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2358 insn for floating-point loads and stores.
2360 2018-03-14 Carl Love <cel@us.ibm.com>
2362 * config/rs6000/rs6000-c.c: Add macro definitions for
2363 ALTIVEC_BUILTIN_VEC_PERMXOR.
2364 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2365 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2366 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2367 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2369 * config/doc/extend.texi: Add prototypes for vec_permxor.
2371 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2374 * diagnostic-show-locus.c (class layout_point): Convert m_line
2375 from int to linenum_type.
2376 (line_span::comparator): Use linenum "compare" function when
2377 comparing line numbers.
2378 (test_line_span): New function.
2379 (layout_range::contains_point): Convert param "row" from int to
2381 (layout_range::intersects_line_p): Likewise.
2382 (layout::will_show_line_p): Likewise.
2383 (layout::print_source_line): Likewise.
2384 (layout::should_print_annotation_line_p): Likewise.
2385 (layout::print_annotation_line): Likewise.
2386 (layout::print_leading_fixits): Likewise.
2387 (layout::annotation_line_showed_range_p): Likewise.
2388 (struct line_corrections): Likewise for field m_row.
2389 (line_corrections::line_corrections): Likewise for param "row".
2390 (layout::print_trailing_fixits): Likewise.
2391 (layout::get_state_at_point): Likewise.
2392 (layout::get_x_bound_for_row): Likewise.
2393 (layout::print_line): Likewise.
2394 (diagnostic_show_locus): Likewise for locals "last_line" and
2396 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2397 * input.c (selftest::test_linenum_comparisons): New function.
2398 (selftest::input_c_tests): Call it.
2399 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2400 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2401 * selftest.h (ASSERT_GT): New macro.
2402 (ASSERT_GT_AT): New macro.
2403 (ASSERT_LT): New macro.
2404 (ASSERT_LT_AT): New macro.
2406 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2408 PR rtl-optimization/84780
2409 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2411 2018-03-14 Martin Liska <mliska@suse.cz>
2413 * tree.c (record_node_allocation_statistics): Use
2414 get_stats_node_kind.
2415 (get_stats_node_kind): New function extracted from
2416 record_node_allocation_statistics.
2417 (free_node): Use get_stats_node_kind.
2419 2018-03-14 Richard Biener <rguenther@suse.de>
2421 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2422 that the value-set of ANTIC_IN doesn't grow.
2425 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2427 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2428 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2430 2018-03-14 Julia Koval <julia.koval@intel.com>
2432 * config.gcc (icelake-client, icelake-server): New.
2434 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2435 (initial_ix86_arch_features): Ditto.
2436 (PTA_SKYLAKE): Add SGX.
2437 (PTA_ICELAKE): Remove.
2438 (PTA_ICELAKE_CLIENT): New.
2439 (PTA_ICELAKE_SERVER): New.
2440 (ix86_option_override_internal): Split up icelake on icelake client and
2442 (get_builtin_code_for_version): Ditto.
2443 (fold_builtin_cpu): Ditto.
2444 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2445 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2446 * config/i386/i386.h (processor_type): Ditto.
2447 * doc/invoke.texi: Ditto.
2449 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2452 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2453 INTEGER_CST offset, add it together with bitpos / 8 and
2454 sign extend based on POINTER_SIZE.
2458 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2461 * config/i386/constraints.md (Yc): New register constraint.
2462 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2463 Use Yc constraint for alternative 2 of operand 0. Remove
2464 preferred_for_speed attribute.
2466 2018-03-14 Richard Biener <rguenther@suse.de>
2468 PR tree-optimization/84830
2469 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2470 with the old one to avoid oscillations.
2472 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2475 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2477 (assign_by_spills): Return a flag of reload assignment failure.
2478 Do not process the reload assignment failures. Do not spill other
2479 reload pseudos if they has the same reg class. Update n if
2481 (lra_assign): Add a return arg. Set up from the result of
2482 assign_by_spills call.
2483 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2484 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2485 usage_insns if it is not NULL.
2486 (spill_hard_reg_in_range): New function.
2487 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2488 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2489 function prototypes.
2490 (lra_assign): Change prototype.
2491 * lra.c (lra): Add code to deal with fails by splitting hard reg
2494 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2496 * config/riscv/riscv.opt (mrelax): New option.
2497 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2498 "norelax" when riscv_mrelax is disabled.
2499 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2501 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2504 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2505 reassociation for int modes.
2507 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2509 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2510 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2512 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2513 * config/aarch64/aarch64-sve.md
2514 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2515 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2516 (*extend<mode><Vwide>2): Rename to...
2517 (aarch64_sve_extend<mode><Vwide>2): ...this.
2518 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2519 renaming the old pattern to...
2520 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2522 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2523 define_expand, renaming the old pattern to...
2524 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2525 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2526 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2527 account when deciding which SVE instruction the optab should use.
2528 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2530 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2532 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2533 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2534 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2535 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2536 (tlsdesc_small_<mode>): Turn a define_expand and use
2537 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2538 (tlsdesc_small_advsimd_<mode>): ...this.
2539 (tlsdesc_small_sve_<mode>): New pattern.
2541 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2543 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2544 (UNSPEC_UMUL_HIGHPART): New constants.
2545 (MUL_HIGHPART): New int iteraor.
2546 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2547 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2549 (*<su>mul<mode>3_highpart): New define_insn.
2551 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2554 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2557 2018-03-13 Martin Liska <mliska@suse.cz>
2560 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2562 (sem_item_optimizer::~sem_item_optimizer): Release it.
2563 (sem_item_optimizer::merge_classes): Register variable aliases.
2564 (sem_item_optimizer::fixup_pt_set): New function.
2565 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2566 * ipa-icf.h: Declare new variables and functions.
2568 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2571 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2572 integer_pow2p@2 and test integer_pow2p in condition.
2573 (A < 0 ? C : 0): Similarly for @1.
2576 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2577 characters starting at p contain '\0' character, don't look beyond
2581 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2582 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2585 * reg-stack.c (change_stack): Change update_end var from int to
2586 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2587 also call set_block_for_insn on the newly added insns and rescan.
2590 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2591 on the last operand.
2594 * tree.c (stabilize_reference_1): Return save_expr (e) for
2595 STATEMENT_LIST even if it doesn't have side-effects.
2597 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2599 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2601 2018-03-12 Renlin Li <renlin.li@arm.com>
2603 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2604 aarch64_output_scalar_simd_mov_immediate.
2606 2018-03-12 Martin Sebor <msebor@redhat.com>
2608 PR tree-optimization/83456
2609 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2610 for perfectly overlapping calls to memcpy.
2611 (gimple_fold_builtin_memory_chk): Same.
2612 (gimple_fold_builtin_strcpy): Handle no-warning.
2613 (gimple_fold_builtin_stxcpy_chk): Same.
2614 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2616 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2618 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2619 parameter. Use it for SFmode.
2620 (rs6000_function_arg_advance_1): Adjust.
2621 (rs6000_function_arg): Adjust.
2622 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2624 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2626 PR rtl-optimization/84169
2627 PR rtl-optimization/84780
2628 * combine.c (can_combine_p): Check for a 2-insn combination whether
2629 the destination register is used between the two insns, too.
2631 2018-03-12 Richard Biener <rguenther@suse.de>
2633 PR tree-optimization/84803
2634 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2635 for refs DR analysis didn't process.
2637 2018-03-12 Richard Biener <rguenther@suse.de>
2639 PR tree-optimization/84777
2640 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2641 force-vectorize loops ignore whether we are optimizing for size.
2643 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2645 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2646 (TARGET_MD_ASM_ADJUST): Define.
2648 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2649 Kito Cheng <kito.cheng@gmail.com>
2650 Chung-Ju Wu <jasonwucj@gmail.com>
2652 * config/nds32/nds32.c (nds32_compute_stack_frame,
2653 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2654 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2655 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2656 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2657 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2658 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2659 * config/nds32/nds32.md (prologue, epilogue): Use macro
2660 NDS32_V3PUSH_AVAILABLE_P to do checking.
2662 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2665 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2666 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2667 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2668 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2669 addition of most attributes on !orig_type_die or the attribute not
2670 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2672 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2673 Chung-Ju Wu <jasonwucj@gmail.com>
2675 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2677 * config/nds32/nds32.opt (mvh): New option.
2679 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2680 Chung-Ju Wu <jasonwucj@gmail.com>
2682 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2684 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2685 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2688 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2689 Chung-Ju Wu <jasonwucj@gmail.com>
2691 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2693 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2694 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2696 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2697 Kito Cheng <kito.cheng@gmail.com>
2698 Chung-Ju Wu <jasonwucj@gmail.com>
2700 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2701 UNSPEC_FFMISM and UNSPEC_FLMISM.
2702 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2703 for ffb, ffmism and flmism.
2704 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2705 (unspec_ffmism): Ditto.
2706 (unspec_flmism): Ditto.
2707 (nds32_expand_builtin_impl): Check if string extension is available.
2708 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2709 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2711 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2714 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2717 * lra-assigns.c (assign_by_spills): Return a flag of reload
2718 assignment failure. Do not process the reload assignment
2719 failures. Do not spill other reload pseudos if they has the same
2721 (lra_assign): Add a return arg. Set up from the result of
2722 assign_by_spills call.
2723 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2724 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2725 usage_insns if it is not NULL.
2726 (spill_hard_reg_in_range): New function.
2727 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2728 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2729 function prototypes.
2730 (lra_assign): Change prototype.
2731 * lra.c (lra): Add code to deal with fails by splitting hard reg
2734 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2737 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2739 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2742 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2743 (dw_val_node): Add val_symbolic_view.
2744 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2745 (symview_upper_bound): New.
2746 (new_line_info_table): Initialize symviews_since_reset.
2747 (dwarf2out_source_line): Count symviews_since_reset and set
2748 symview_upper_bound.
2749 (dw_val_equal_p): Handle symview.
2750 (add_AT_symview): New.
2751 (print_dw_val): Handle symview.
2752 (attr_checksum, attr_checksum_ordered): Likewise.
2753 (same_dw_val_p, size_of_die): Likewise.
2754 (value_format, output_die): Likewise.
2755 (add_high_low_attributes): Use add_AT_symview for entry_view.
2756 (dwarf2out_finish): Reset symview_upper_bound, clear
2759 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2762 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2763 Add strict argument and use it.
2764 (rs6000_split_multireg_move): Update for new strict argument.
2765 (mem_operand_gpr): Disallow all non-offsettable addresses.
2766 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2768 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2771 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2772 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2773 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2776 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2777 decl, use remap_type if we want to use the type.
2779 2018-03-09 Martin Sebor <msebor@redhat.com>
2781 PR tree-optimization/84526
2782 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2784 (builtin_access::generic_overlap): Be prepared to handle non-array
2787 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2789 PR rtl-optimization/84682
2790 * lra-constraints.c (process_address_1): Check is_address flag
2791 for address constraints.
2792 (process_alt_operands): Likewise.
2793 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2794 preprocess_constraints.
2795 * recog.h (preprocess_constraints): Add oploc parameter.
2797 * recog.c (preprocess_constraints): Test address_operand for
2798 CT_ADDRESS constraints.
2800 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2803 * lra-assigns.c (assign_by_spills): Return a flag of reload
2804 assignment failure. Do not process the reload assignment
2805 failures. Do not spill other reload pseudos if they has the same
2807 (lra_assign): Add a return arg. Set up from the result of
2808 assign_by_spills call.
2809 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2810 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2811 usage_insns if it is not NULL.
2812 (spill_hard_reg_in_range): New function.
2813 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2814 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2815 function prototypes.
2816 (lra_assign): Change prototype.
2817 * lra.c (lra): Add code to deal with fails by splitting hard reg
2820 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2823 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2824 Accept complain bool parameter. Only emit errors if it is true.
2825 (arm_parse_cpu_option_name): Likewise.
2826 (arm_target_thumb_only): Adjust callers of the above.
2827 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2828 prototype to take a default true bool parameter.
2829 (arm_parse_arch_option_name): Likewise.
2831 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2832 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2836 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2837 * configure: Regenerate.
2838 * configure.ac ("linker --version-script option"): New.
2839 ("linker soname option"): New.
2841 2018-03-09 Richard Biener <rguenther@suse.de>
2843 PR tree-optimization/84775
2844 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2845 immediate uses of predicate stmts and mark them modified.
2848 PR tree-optimization/84178
2849 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2851 (version_loop_for_if_conversion): Delay update_ssa call.
2852 (tree_if_conversion): Delay update_ssa until after predicate
2855 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2858 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2859 when the function accesses prior frames.
2861 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2864 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2865 gen_llsym, otherwise call maybe_gen_llsym.
2868 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2869 has ',' character inside of it.
2871 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2874 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2875 as clobbering CC_REGNUM.
2877 2018-03-08 Richard Biener <rguenther@suse.de>
2880 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2881 (follow_copies_to_constant): Do not follow SSA names registered
2884 2018-03-08 Richard Biener <rguenther@suse.de>
2886 PR tree-optimization/84178
2887 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2889 (version_loop_for_if_conversion): Delay update_ssa call.
2890 (tree_if_conversion): Delay update_ssa until after predicate
2893 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2895 PR tree-optimization/84178
2896 * tree-if-conv.c (release_bb_predicate): Remove the
2897 the assertion that the stmts have NULL use_ops.
2898 Discard the statements, asserting that they haven't
2899 yet been added to a BB.
2901 2018-03-08 Richard Biener <rguenther@suse.de>
2903 PR tree-optimization/84746
2904 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2905 (phi_translate): Pass in destination ANTIC_OUT set.
2906 (phi_translate_1): Likewise. For a simplified result lookup
2907 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2908 (phi_translate_set): Adjust.
2909 (do_pre_regular_insertion): Likewise.
2910 (do_pre_partial_partial_insertion): Likewise.
2912 2018-03-08 Martin Liska <mliska@suse.cz>
2914 PR gcov-profile/84735
2915 * doc/gcov.texi: Document usage of profile files.
2916 * gcov-io.h: Document changes in the format.
2918 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2922 * dwarf2out.c (struct dw_line_info_table): Update comments for
2924 (FORCE_RESET_NEXT_VIEW): New.
2925 (FORCE_RESETTING_VIEW_P): New.
2926 (RESETTING_VIEW_P): Check for -1 too.
2927 (ZERO_VIEW_P): Likewise.
2928 (new_line_info_table): Force-reset next view.
2929 (dwarf2out_begin_function): Likewise.
2930 (dwarf2out_source_line): Simplify zero_view_p initialization.
2931 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2932 view directly. Omit view when omitting .loc at line 0.
2934 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2936 PR tree-optimization/84740
2937 * tree-switch-conversion.c (process_switch): Call build_constructors
2938 only if info.phi_count is non-zero.
2940 PR tree-optimization/84739
2941 * tree-tailcall.c (find_tail_calls): Check call arguments against
2942 DECL_ARGUMENTS (current_function_decl) rather than
2943 DECL_ARGUMENTS (func) when checking for tail recursion.
2945 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2947 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2948 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2949 Volker Reichelt's entry and add entries for people that perform
2950 GCC fuzzy testing and report numerous bugs.
2952 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2955 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2956 readonly data in sdata, if that is disabled.
2957 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2958 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2959 -mreadonly-in-sdata option.
2961 2018-03-07 Martin Sebor <msebor@redhat.com>
2963 PR tree-optimization/84468
2964 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2965 basic block when looking for nul assignment.
2967 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2970 * except.h (output_function_exception_table): Adjust prototype.
2971 * except.c (output_function_exception_table): Remove FNNAME parameter
2972 and add SECTION parameter. Ouput one part of the table at a time.
2973 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2974 the first part of the exception table and emit unwind directives.
2975 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2976 (i386_pe_seh_cold_init): Likewise.
2977 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2978 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2979 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2980 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2981 * config/i386/winnt.c: Include except.h.
2982 (struct seh_frame_state): Add reg_offset, after_prologue and
2983 in_cold_section fields.
2984 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2985 (i386_pe_seh_cold_init): New function.
2986 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2987 to seh->in_cold_section.
2988 (seh_emit_push): Record the offset of the push.
2989 (seh_emit_save): Record the offet of the save.
2990 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2991 Test seh->after_prologue to disregard the epilogue.
2992 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2993 (i386_pe_end_cold_function): New function.
2995 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2998 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2999 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3002 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3004 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3005 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3008 * multiple_target.c: Include tree-inline.h and intl.h.
3009 (expand_target_clones): Diagnose and fail if node->definition and
3010 !tree_versionable_function_p (node->decl).
3012 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3014 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3016 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3017 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3018 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3020 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3023 * combine.c (try_combine): Use reg_or_subregno instead of handling
3024 just paradoxical SUBREGs and REGs.
3026 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3028 * config/arc/arc.c (arc_finalize_pic): Remove function.
3029 (arc_must_save_register): We use single base PIC register, remove
3030 checks to save/restore the PIC register.
3031 (arc_expand_prologue): Likewise.
3032 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3034 (arc_verify_short): Likewise.
3035 (arc_attr_type): Likewise.
3036 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3037 (walk_stores): Likewise.
3038 (arc_address_cost): Make it static.
3039 (arc_verify_short): Likewise.
3040 (branch_dest): Likewise.
3041 (arc_attr_type): Likewise.
3042 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3043 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3044 (arc_final_prescan_insn): Remove inserting the nops due to
3045 hardware hazards. It is done in reorg step.
3046 (insn_length_variant_t): Remove.
3047 (insn_length_parameters_t): Likewise.
3048 (arc_insn_length_parameters): Likewise.
3049 (arc_get_insn_variants): Likewise.
3050 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3052 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3055 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3058 PR tree-optimization/84687
3059 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3061 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3063 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3065 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3066 Rename to ppc_speculation_barrier.
3067 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3068 __builtin_ppc_speculation_barrier.
3070 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3073 * combine.c (combine_simplify_rtx): Don't try to simplify if
3074 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3077 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3079 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3080 to 32 bytes when compiling for POWER9.
3082 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3085 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3086 regparm >= 3 with no arg reg available also for calls with
3087 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3090 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3092 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3094 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3097 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3099 2018-03-05 Richard Biener <rguenther@suse.de>
3101 PR tree-optimization/84486
3102 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3103 When inserting a __builtin_assume_aligned call set the LHS
3104 SSA name alignment info accordingly.
3106 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3108 PR tree-optimization/84114
3109 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3110 Avoid reassociation of FLOAT_MODE addition.
3112 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3114 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3115 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3116 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3117 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3118 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3119 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3120 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3122 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3123 __builtin_ia32_wbinvd): New builtins.
3124 (SPECIAL_ARGS2): New.
3125 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3126 (SPECIAL_ARGS2): New.
3127 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3128 (ix86_valid_target_attribute_inner_p): Ditto.
3129 (ix86_init_mmx_sse_builtins): Add special_args2.
3130 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3131 TARGET_WBNOINVD_P): New.
3132 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3133 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3134 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3135 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3136 * config/i386/pconfigintrin.h: New file.
3137 * config/i386/wbnoinvdintrin.h: Ditto.
3138 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3139 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3141 2018-03-05 Richard Biener <rguenther@suse.de>
3143 PR tree-optimization/84670
3144 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3146 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3147 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3148 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3149 doesn't grow if all successors (recursively) were visited at least
3152 2018-03-05 Richard Biener <rguenther@suse.de>
3154 PR tree-optimization/84650
3155 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3156 if executed in the loop pipeline.
3158 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3160 * doc/configfiles.texi (Configuration Files): Move info about
3161 conditionalizing $target-protos.h to...
3162 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3163 differs from $target-protos.h.
3165 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3166 Chung-Ju Wu <jasonwucj@gmail.com>
3168 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3169 * config/nds32/nds32-multiple.md (setmemsi): Define.
3170 * config/nds32/nds32-memory-manipulation.c
3171 (nds32_gen_dup_4_byte_to_word_value): New.
3172 (emit_setmem_word_loop): New.
3173 (emit_setmem_byte_loop): New.
3174 (nds32_expand_setmem_loop): New.
3175 (nds32_expand_setmem_loop_v3m): New.
3176 (nds32_expand_setmem_unroll): New.
3177 (nds32_expand_setmem): New.
3179 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3180 Chung-Ju Wu <jasonwucj@gmail.com>
3182 * config/nds32/nds32-memory-manipulation.c
3183 (nds32_emit_load_store): New.
3184 (nds32_emit_post_inc_load_store): New.
3185 (nds32_emit_mem_move): New.
3186 (nds32_emit_mem_move_block): New.
3187 (nds32_expand_movmemsi_loop_unknown_size): New.
3188 (nds32_expand_movmemsi_loop_known_size): New.
3189 (nds32_expand_movmemsi_loop): New.
3190 (nds32_expand_movmemsi_unroll): New.
3191 (nds32_expand_movmemqi): Rename ...
3192 (nds32_expand_movmemsi): ... to this.
3193 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3194 (movmemsi): ... to this.
3195 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3196 (nds32_expand_movmemsi): ... to this.
3198 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3199 Monk Chiang <sh.chiang04@gmail.com>
3200 Chung-Ju Wu <jasonwucj@gmail.com>
3202 * config/nds32/nds32-protos.h
3203 (nds32_expand_load_multiple): New arguments.
3204 (nds32_expand_store_multiple): Ditto.
3205 (nds32_valid_multiple_load_store): Rename ...
3206 (nds32_valid_multiple_load_store_p): ... to this.
3207 * config/nds32/nds32-memory-manipulation.c
3208 (nds32_expand_load_multiple): Refine implementation.
3209 (nds32_expand_store_multiple): Ditto.
3210 * config/nds32/nds32-multiple.md
3211 (load_multiple): Update nds32_expand_load_multiple interface.
3212 (store_multiple): Update nds32_expand_store_multiple interface.
3213 * config/nds32/nds32-predicates.c
3214 (nds32_valid_multiple_load_store): Rename ...
3215 (nds32_valid_multiple_load_store_p): ... to this and refine
3217 * config/nds32/predicates.md
3218 (nds32_load_multiple_and_update_address_operation): New predicate.
3219 (nds32_store_multiple_and_update_address_operation): New predicate.
3221 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3222 Chung-Ju Wu <jasonwucj@gmail.com>
3224 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3225 (combo): New attribute.
3226 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3228 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3230 * config/nds32/nds32.opt: Change -mcmodel= default value.
3232 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3233 Monk Chiang <sh.chiang04@gmail.com>
3234 Chung-Ju Wu <jasonwucj@gmail.com>
3236 * config/nds32/constants.md (unspec_element): New enum.
3237 * config/nds32/constraints.md (Umw): New constraint.
3238 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3239 * config/nds32/nds32-intrinsic.md: Likewise.
3240 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3241 (nds32_valid_smw_lwm_base_p): New.
3242 (nds32_output_smw_single_word): New.
3243 (nds32_output_lmw_single_word): New.
3244 (nds32_expand_unaligned_load): New.
3245 (nds32_expand_unaligned_store): New.
3246 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3247 (nds32_output_smw_single_word): Declare.
3248 (nds32_output_lmw_single_word): Declare.
3249 (nds32_expand_unaligned_load): Declare.
3250 (nds32_expand_unaligned_store): Declare.
3251 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3252 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3253 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3254 NDS32_BUILTIN_UASTORE_DW.
3255 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3258 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3259 Kito Cheng <kito.cheng@gmail.com>
3260 Chung-Ju Wu <jasonwucj@gmail.com>
3262 * config/nds32/nds32-intrinsic.c
3263 (nds32_expand_builtin_null_ftype_reg): Delete.
3264 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3265 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3266 (nds32_read_argument): New.
3267 (nds32_legitimize_target): Ditto.
3268 (nds32_legitimize_argument): Ditto.
3269 (nds32_check_constant_argument): Ditto.
3270 (nds32_expand_unop_builtin): Ditto.
3271 (nds32_expand_unopimm_builtin): Ditto.
3272 (nds32_expand_binop_builtin): Ditto.
3273 (nds32_builtin_decl_impl): Ditto.
3274 (builtin_description): Ditto.
3275 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3276 (nds32_init_builtins_impl): Ditto.
3277 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3278 (nds32_builtin_decl): New.
3279 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3280 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3282 2018-03-02 Jeff Law <law@redhat.com>
3284 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3285 (redundant_insn, fill_simple_delay_slots): Likewise.
3286 (fill_slots_from_thread): Likewise.
3287 * resource.c (mark_referenced_resources): Likewise.
3288 (mark_set_resources, find_dead_or_set_registers): Likewise.
3290 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3292 * substring-locations.h (format_warning_va): Formatting fix for
3294 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3296 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3297 * substring-locations.c: Include intl.h.
3298 (format_warning_va): Turned into small wrapper around
3299 format_warning_n_va, renamed to ...
3300 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3301 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3303 (format_warning_at_substring_n): New function.
3304 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3305 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3306 format_warning_at_substring with just a shorter name instead of
3307 const function pointer.
3308 (fmtwarn_n): New function.
3309 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3310 appropriate, get rid of all the fmtstr temporaries, move conditionals
3311 with G_() wrapped string literals directly into fmtwarn arguments,
3312 cast dir.len to (int), formatting fixes.
3314 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3316 * doc/invoke.texi: Remove "Cilk Plus" references.
3318 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3319 Richard Biener <rguenther@suse.de>
3322 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3323 for error or warning attributes if CALL_FROM_THUNK_P is set.
3326 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3329 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3330 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3333 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3334 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3336 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3337 instead of -1U in last predictors element's probability member.
3339 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3342 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3343 arguments if they are comparable.
3345 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3347 PR tree-optimization/84634
3348 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3349 masks and masked_loop_p with a single loop_masks, making sure it's
3350 null for bb vectorization.
3352 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3354 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3355 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3356 loop->force_vectorize to check whether there is no alias.
3358 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3361 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3363 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3365 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3366 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3367 instead of a loop around prev_real_insn.
3368 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3372 * config/i386/i386.c (ix86_print_operand): Use conditional
3373 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3376 2018-03-02 Richard Biener <rguenther@suse.de>
3378 PR tree-optimization/84427
3379 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3380 (bitmap_set_subtract_values): Rewrite to handle multiple
3383 (prune_clobbered_mems): Likewise.
3384 (phi_translate): Take edge instead of pred/phiblock.
3385 (phi_translate_1): Likewise.
3386 (phi_translate_set): Likewise. Insert all translated
3387 exprs for a value into the set, keeping possibly multiple
3388 expressions per value.
3389 (compute_antic_aux): Adjust for phi_translate changes.
3390 When intersecting union the expressions and prune those
3391 not in the final value set, keeping possibly multiple
3392 expressions per value. Do not use value-insertion
3393 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3394 all expressions. Add verification that the value-sets
3395 only shrink during iteration.
3396 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3397 (do_pre_regular_insertion): Likewise.
3398 (do_pre_partial_partial_insertion): Likewise.
3400 2018-03-02 Richard Biener <rguenther@suse.de>
3403 * config/darwin.c (saved_debug_info_level): New static global.
3404 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3405 (darwin_asm_lto_end): Restore debug info generation settings.
3407 2018-03-01 Martin Liska <mliska@suse.cz>
3410 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3413 2018-03-01 Richard Biener <rguenther@suse.de>
3416 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3417 type annotation with LTO when debug was disabled at compile-time.
3419 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3421 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3423 (mips_final_postscan_insn): Likewise.
3425 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3427 PR rtl-optimization/84528
3428 * alias.c (init_alias_target): Add commentary.
3429 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3430 a unique base value if the frame pointer is not eliminated
3431 to the stack pointer.
3433 2018-03-01 Tom de Vries <tom@codesourcery.com>
3435 PR rtl-optimization/83327
3436 * lra-int.h (hard_regs_spilled_into): Declare.
3437 * lra.c (hard_regs_spilled_into): Define.
3438 (init_reg_info): Init hard_regs_spilled_into.
3439 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3440 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3441 (process_bb_lives): Handle hard_regs_spilled_into.
3442 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3443 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3445 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3447 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3448 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3449 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3450 * config/rs6000/aix72.h: New file.
3452 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3454 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3455 instead of warning_at with conditional singular and plural messages
3459 * stor-layout.c (update_alignment_for_field): For
3460 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3461 && !DECL_PACKED (field), do the alignment update, just use
3462 only desired_align instead of MAX (type_align, desired_align)
3464 (place_field): Don't do known_align < desired_align handling
3465 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3466 is non-NULL, instead do it after rli->prev_field handling and
3467 only if not within a bitfield word. For DECL_PACKED (field)
3468 use type_align of BITS_PER_UNIT.
3470 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3472 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3473 superfluous parentheses and trailing spaces.
3475 2018-02-28 Richard Biener <rguenther@suse.de>
3477 PR tree-optimization/84584
3478 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3479 SCoPs with fake exit edge.
3481 2018-02-28 Martin Liska <mliska@suse.cz>
3484 * timevar.c (timer::print): Fix format to properly print 100%
3487 2018-02-28 Richard Biener <rguenther@suse.de>
3490 * genmatch.c (capture_info::walk_match): Do not mark
3491 captured expressions without operands as expr_p given
3492 they act more like predicates and should be subject to
3493 "lost tail" side-effect preserving.
3495 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3497 PR rtl-optimization/81611
3498 * auto-inc-dec.c (attempt_change): Move dead note from
3499 mem_insn if it's the next use of regno
3500 (find_address): Take address use of reg holding
3501 non-incremented value. Add parm to limit search to the named
3503 (merge_in_block): Attempt to use a mem insn that is the next
3504 use of the original regno.
3506 2018-02-27 Martin Sebor <msebor@redhat.com>
3509 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3510 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3512 2018-02-27 Martin Sebor <msebor@redhat.com>
3514 PR translation/84207
3515 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3516 n argument to unsigned HOST_WIDE_INT.
3517 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3518 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3519 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3520 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3522 2018-02-27 Richard Biener <rguenther@suse.de>
3524 PR tree-optimization/84512
3525 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3526 Do not use the estimate returned from record_stmt_cost for
3527 the scalar iteration cost but sum properly using add_stmt_cost.
3529 2018-02-27 Richard Biener <rguenther@suse.de>
3531 PR tree-optimization/84466
3532 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3533 Adjust last change to less strictly validate use operands.
3535 2018-02-27 Martin Liska <mliska@suse.cz>
3537 PR gcov-profile/84548
3538 * gcov.c (process_file): Allow partial overlap and consider it
3539 also as group functions.
3540 (output_lines): Properly calculate range of lines for a group.
3542 2018-02-27 Martin Liska <mliska@suse.cz>
3544 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3545 'ggc' suffixes. Change first column width.
3546 (timer::print): Fix formatting of the column.
3548 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3550 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3551 preserve inline entry blocks for the sake of debug inline
3552 entry point markers alone.
3553 (remove_unused_locals): Suggest in comments a better place to
3554 force the preservation of inline entry blocks that are
3555 otherwise unused, but do not preserve them.
3557 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3559 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3561 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3564 * config/i386/constraints.md (Bs): Replace
3565 ix86_indirect_branch_register with
3566 TARGET_INDIRECT_BRANCH_REGISTER.
3568 * config/i386/i386.md (indirect_jump): Likewise.
3569 (tablejump): Likewise.
3570 (*sibcall_memory): Likewise.
3571 (*sibcall_value_memory): Likewise.
3572 Peepholes of indirect call and jump via memory: Likewise.
3573 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3574 (*sibcall_value_GOT_32): Likewise.
3575 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3576 (GOT_memory_operand): Likewise.
3577 (call_insn_operand): Likewise.
3578 (sibcall_insn_operand): Likewise.
3579 (GOT32_symbol_operand): Likewise.
3580 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3582 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3584 PR rtl-optimization/83496
3585 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3586 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3587 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3588 redundant insn, if any.
3589 (relax_delay_slots): Likewise.
3590 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3592 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3594 PR tree-optimization/83965
3595 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3596 that grouped statements are part of a reduction chain. Return
3597 true if the statement is not marked as a reduction itself but
3599 (vect_recog_dot_prod_pattern): Don't check whether the statement
3600 is part of a group here.
3601 (vect_recog_sad_pattern): Likewise.
3602 (vect_recog_widen_sum_pattern): Likewise.
3604 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3607 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3609 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3612 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3614 (ix86_output_indirect_function_return): New prototype.
3615 (ix86_split_simple_return_pop_internal): Likewise.
3616 * config/i386/i386.c (indirect_return_via_cx): New.
3617 (indirect_return_via_cx_bnd): Likewise.
3618 (indirect_thunk_name): Handle return va CX_REG.
3619 (output_indirect_thunk_function): Create alias for
3620 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3621 (ix86_output_indirect_jmp): Remove the bool argument.
3622 (ix86_output_indirect_function_return): New function.
3623 (ix86_split_simple_return_pop_internal): Likewise.
3624 * config/i386/i386.md (*indirect_jump): Don't pass false
3625 to ix86_output_indirect_jmp.
3626 (*tablejump_1): Likewise.
3627 (simple_return_pop_internal): Change it to define_insn_and_split.
3628 Call ix86_split_simple_return_pop_internal to split it for
3630 (simple_return_indirect_internal): Call
3631 ix86_output_indirect_function_return instead of
3632 ix86_output_indirect_jmp.
3634 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3637 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3638 memset and value initialization afterwards.
3640 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3642 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3644 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3647 * common/config/aarch64/aarch64-common.c
3648 (aarch_option_optimization_table[]): Switch
3649 off fomit-frame-pointer
3651 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3652 Chung-Ju Wu <jasonwucj@gmail.com>
3654 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3656 (store_multiple): Ditto.
3658 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3660 * config.gcc: Add --with-cpu support for nds32 target.
3661 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3662 * config/nds32/nds32.opt: Add -mcpu= option.
3664 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3666 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3667 isel=yes): Warn for these deprecated options.
3669 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3671 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3672 ISA_2_5_MASKS_EMBEDDED.
3674 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3676 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3677 p->max as pointers rather than using iterative_hash_expr.
3679 2018-02-23 Carl Love <cel@us.ibm.com>
3681 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3682 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3684 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3685 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3686 P8V_BUILTIN_VEC_VUNSIGNED2.
3688 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3691 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3692 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3694 * lra-constraints.c (curr_insn_transform): Set up
3695 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3697 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3698 * lra-eliminations.c (spill_pseudos): Ditto.
3699 (process_insn_for_elimination): Ditto.
3700 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3701 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3702 LRA_NON_CLOBBERED_ALT.
3704 2018-02-22 Martin Sebor <msebor@redhat.com>
3706 PR tree-optimization/84480
3707 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3708 to maybe_diag_stxncpy_trunc. Call it.
3709 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3710 from gimple_fold_builtin_strcpy. Print inlining stack.
3711 (handle_builtin_stxncpy): Print inlining stack.
3712 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3714 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3717 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3718 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3719 and -fcheck-pointer-bounds are used together.
3720 (indirect_thunk_prefix): New enum.
3721 (indirect_thunk_need_prefix): New function.
3722 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3723 "_nt" instead of "_bnd" for NOTRACK prefix.
3724 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3725 (output_indirect_thunk_function): Likewise.
3727 (ix86_code_end): Update output_indirect_thunk_function calls.
3728 (ix86_output_indirect_branch_via_reg): Replace
3729 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3730 (ix86_output_indirect_branch_via_push): Likewise.
3731 (ix86_output_function_return): Likewise.
3732 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3733 incompatible with -fcf-protection=branch and
3734 -fcheck-pointer-bounds.
3736 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3739 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3740 Change gcc_assert call to output_operand_lossage.
3742 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3744 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3746 2018-02-22 DJ Delorie <dj@redhat.com>
3747 Sebastian Perta <sebastian.perta@renesas.com>
3748 Oleg Endo <olegendo@gcc.gnu.org>
3750 * config/rx/rx.c (rx_rtx_costs): New function.
3751 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3753 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3755 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3757 2018-02-22 Martin Liska <mliska@suse.cz>
3760 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3761 Add "native" as a possible value.
3763 2018-02-22 Martin Liska <mliska@suse.cz>
3766 * config/i386/i386.c (ix86_option_override_internal):
3767 Add "native" as a possible value for -march and -mtune.
3769 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3772 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3773 to all type variants.
3775 PR tree-optimization/84503
3776 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3777 width as info->bitpos + info->bitsize - start.
3778 (merged_store_group::merge_overlapping): Simplify width computation.
3779 (check_no_overlap): New function.
3780 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3781 start + width and last_order of the group, fail if check_no_overlap
3783 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3784 to group if check_no_overlap fails.
3786 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3788 * config/rs6000/altivec.md: Delete contraint arguments to
3789 define_expand, define_split, and define_peephole2, and in
3790 define_insn_and_split if always unused.
3791 * config/rs6000/darwin.md: Ditto.
3792 * config/rs6000/dfp.md: Ditto.
3793 * config/rs6000/rs6000.md: Ditto.
3794 * config/rs6000/sync.md: Ditto.
3795 * config/rs6000/vector.md: Ditto.
3796 * config/rs6000/vsx.md: Ditto.
3798 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3800 * config/rs6000/altivec.md: Write output control strings as braced
3801 blocks instead of double-quoted strings.
3802 * config/rs6000/darwin.md: Ditto.
3803 * config/rs6000/rs6000.md: Ditto.
3804 * config/rs6000/vector.md: Ditto.
3805 * config/rs6000/vsx.md: Ditto.
3807 2018-02-21 Jason Merrill <jason@redhat.com>
3809 PR c++/84314 - ICE with templates and fastcall attribute.
3810 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3812 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3814 * ipa-cp.c (determine_versionability): Fix comment typos.
3816 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3819 * ipa-cp.c (determine_versionability): Do not version functions caling
3822 2018-02-21 Martin Liska <mliska@suse.cz>
3825 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3826 Add "native" as a possible value.
3827 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3828 the macro when native cpu detection is available.
3830 2018-02-21 Martin Liska <mliska@suse.cz>
3833 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3834 Add "native" as a possible value.
3835 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3836 when native cpu detection is available.
3838 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3839 Martin Sebor <msebor@redhat.com>
3841 PR tree-optimization/84478
3842 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3844 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3845 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3846 support which is conservatively correct, for 2 only stay conservative
3847 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3848 argument to the 2 argument get_range_strlen, adjust 6 arg
3849 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3851 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3852 (gimple_fold_builtin_strlen): Pass true as last argument to
3855 2018-02-20 Martin Sebor <msebor@redhat.com>
3858 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3859 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3860 (builtin_memref::builtin_memref): Factor out parts into
3861 set_base_and_offset and call it.
3863 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3866 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3867 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3868 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3869 search at the associated MODE_INT.
3871 2018-02-20 Jeff Law <law@redhat.com>
3874 PR tree-optimization/81592
3876 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3877 for range data rather than using global data.
3878 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3879 range data rather than using global data.
3880 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3881 pass it to children as needed.
3882 (struct directive::fmtresult): Similarly.
3883 (struct directive::set_width): Similarly.
3884 (struct directive::set_precision): Similarly.
3885 (format_integer, format_directive, parse_directive): Similarly.
3886 (format_none): Accept unnamed vr_values parameter.
3887 (format_percent, format_floating, format_character): Similarly.
3888 (format_string, format_plain): Similarly.
3889 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3890 the EVRP range analyzer for range data rather than using global data.
3891 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3892 gimple-ssa-evrp-analyze.h
3893 (class sprintf_dom_walker): Add after_dom_children member function.
3894 Add evrp_range_analyzer member.
3895 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3896 range analyzer as needed.
3897 (sprintf_dom_walker::after_dom_children): New member function.
3898 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3900 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3901 (evrp_range_analyzer::pop_to_marker): Likewise.
3903 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3905 PR tree-optimization/84419
3906 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3907 with the required type if its current type is compatible but
3910 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3913 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3914 after vectorization.
3916 2018-02-20 Martin Liska <mliska@suse.cz>
3919 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3920 possible values if we don't have a hint.
3922 2018-02-20 Martin Liska <mliska@suse.cz>
3926 * final.c (shorten_branches): Build align_tab array with one
3928 * opts.c (finish_options): Add alignment option limit check.
3929 (MAX_CODE_ALIGN): Likewise.
3930 (MAX_CODE_ALIGN_VALUE): Likewise.
3931 * doc/invoke.texi: Document maximum allowed option value for
3932 all -falign-* options.
3934 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3937 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3938 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3939 * var-tracking.c (emit_note_insn_var_location): Remove all references
3940 to NOTE_INSN_CALL_ARG_LOCATION.
3941 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3942 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3943 Use copy_rtx_if_shared.
3944 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3945 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3946 (dwarf2out_var_location): Remove handling of
3947 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3949 * final.c (final_scan_insn): Remove all references to
3950 NOTE_INSN_CALL_ARG_LOCATION.
3951 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3952 before dumping final insns.
3953 * except.c (emit_note_eh_region_end): Remove all references to
3954 NOTE_INSN_CALL_ARG_LOCATION.
3955 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3956 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3957 * config/arc/arc.c (hwloop_optimize): Likewise.
3958 * config/arm/arm.c (create_fix_barrier): Likewise.
3959 * config/s390/s390.c (s390_chunkify_start): Likewise.
3960 * config/sh/sh.c (find_barrier): Likewise.
3961 * config/i386/i386.c (rest_of_insert_endbranch,
3962 ix86_seh_fixup_eh_fallthru): Likewise.
3963 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3964 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3965 * config/frv/frv.c (frv_function_prologue): Likewise.
3966 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3968 (note_outside_basic_block_p): Remove all references to
3969 NOTE_INSN_CALL_ARG_LOCATION.
3970 * gengtype.c (adjust_field_rtx_def): Likewise.
3971 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3973 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3974 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3977 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3980 PR tree-optimization/84452
3981 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3982 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3985 2018-02-19 Martin Liska <mliska@suse.cz>
3988 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3990 2018-02-19 Martin Liska <mliska@suse.cz>
3991 Richard Sandiford <richard.sandiford@linaro.org>
3993 PR tree-optimization/82491
3994 * gimple-fold.c (get_base_constructor): Make earlier bail out
3997 2018-02-19 Carl Love <cel@us.ibm.com>
3999 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4000 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4002 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4003 P8V_BUILTIN_VEC_NEG.
4005 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4007 * config/rl78/rl78.md (movdf): New define expand.
4009 2018-02-19 Martin Liska <mliska@suse.cz>
4012 * doc/invoke.texi: Fix typo.
4013 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4015 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4017 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4018 handle rs6000_single_float and rs6000_double_float specially for
4021 2018-02-16 Jeff Law <law@redhat.com>
4023 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4026 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4029 * ipa-inline.c (inline_small_functions): Fix a typo.
4031 2018-02-16 Nathan Sidwell <nathan@acm.org>
4033 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4035 2018-02-16 Carl Love <cel@us.ibm.com>
4037 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4038 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4039 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4040 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4041 expansion to P8V_BUILTIN_VEC_FLOAT2.
4043 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4045 PR rtl-optimization/70023
4046 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4047 src_regno into account.
4049 2018-02-16 Carl Love <cel@us.ibm.com>
4051 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4052 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4053 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4054 * config/rs6000/rs6000.c: Remove case statements for
4055 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4056 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4057 and P9V_BUILTIN_VEC_VINSERT4B.
4058 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4059 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4060 * config/rs6000/vsx.md:
4061 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4064 2018-02-16 Carl Love <cel@us.ibm.com>
4066 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4068 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4070 * config/rs6000/rs6000-c.c: Add the definitions for
4071 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4072 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4073 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4074 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4075 definition for insert4b and define insn *insert3b_internal.
4076 * doc/extend.texi: Add documentation for vec_extract4b.
4078 2018-02-16 Nathan Sidwell <nathan@acm.org>
4080 * doc/extend.texi (Backwards Compatibility): Mention friend
4081 injection. Note for-scope is deprecated.
4082 * doc/invoke.texi (-ffriend-injection): Deprecate.
4084 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4086 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4087 that moved to I2, also allow destinations that are a paradoxical
4088 subreg (instead of a normal reg).
4090 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4093 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4096 2018-02-16 Richard Biener <rguenther@suse.de>
4098 PR tree-optimization/84037
4099 PR tree-optimization/84016
4101 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4102 Adjust vec_construct for the fact we need additional higher latency
4103 128bit inserts for AVX256 and AVX512 vector builds.
4104 (ix86_add_stmt_cost): Scale vector construction cost for
4107 2018-02-16 Richard Biener <rguenther@suse.de>
4109 PR tree-optimization/84417
4110 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4111 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4112 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4114 2018-02-16 Martin Liska <mliska@suse.cz>
4117 * internal-fn.def (ASAN_CHECK): Set proper flags.
4118 (ASAN_MARK): Likewise.
4120 2018-02-16 Julia Koval <julia.koval@intel.com>
4122 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4123 from PTA_CANNONLAKE.
4125 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4128 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4129 Use ++iter rather than iter++ for std::list iterators.
4130 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4131 defer deleting them until all nodes in the forest are processed. Do
4132 free even leaf nodes. Change to_process into auto_vec.
4135 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4136 * vec.h (vec_default_construct): Use memset instead of placement new
4137 if BROKEN_VALUE_INITIALIZATION is defined.
4138 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4139 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4142 PR rtl-optimization/83723
4143 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4144 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4145 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4147 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4149 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4151 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4153 PR rtl-optimization/81443
4154 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4155 from inner REGs to paradoxical SUBREGs.
4157 2018-02-16 Richard Biener <rguenther@suse.de>
4159 PR tree-optimization/84399
4160 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4161 For operands we can analyze at their definition make sure we can
4162 analyze them at each use as well.
4164 2018-02-16 Richard Biener <rguenther@suse.de>
4166 PR tree-optimization/84190
4167 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4168 volatile accesses if the decl isn't volatile.
4170 2018-02-15 Jason Merrill <jason@redhat.com>
4172 PR c++/84314 - ICE with templates and fastcall attribute.
4173 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4174 TYPE_CANONICAL on an existing type.
4176 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4178 PR tree-optimization/84383
4179 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4180 dstoff nor call operand_equal_p if dstbase is NULL.
4182 PR tree-optimization/84334
4183 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4184 also a CONSTANT_CLASS_P, punt.
4186 2018-02-14 Jim Wilson <jimw@sifive.com>
4188 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4189 first SMALL_OPERAND check. New local min_second_step. Move assert
4190 to where locals are set. Add TARGET_RVC support.
4191 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4193 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4195 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4197 2018-02-14 Martin Sebor <msebor@redhat.com>
4199 PR tree-optimization/83698
4200 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4201 arrays constrain the offset range to their bounds.
4202 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4203 (builtin_access::overlap): Avoid setting the size of overlap if it's
4205 (maybe_diag_overlap): Also consider arrays when deciding what values
4206 of offsets to include in diagnostics.
4208 2018-02-14 Martin Sebor <msebor@redhat.com>
4211 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4212 that correspond to the kind of a declaration.
4214 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4217 * config/pa/pa.md: Load address of PIC label using the linkage table
4218 if the label is nonlocal.
4220 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4222 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4223 warning message if user requests -maltivec=be.
4224 * doc/invoke.texi: Document deprecation of -maltivec=be.
4226 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4229 * config/rs6000/rs6000-c.c: Update definitions for
4230 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4231 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4233 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4236 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4237 add _get_ssp intrinsics. Remove argument from
4238 __builtin_ia32_rdssp[d|q].
4239 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4240 * config/i386/i386-builtin.def: Remove argument from
4241 __builtin_ia32_rdssp[d|q].
4242 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4243 ix86_expand_special_args_builtin for _rdssp[d|q].
4244 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4245 Clear register before usage.
4246 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4247 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4249 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4251 PR tree-optimization/84357
4252 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4253 operand 1 of an ARRAY_REF too.
4255 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4258 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4259 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4261 (set_of_reg): New struct.
4262 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4263 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4264 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4266 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4267 Split into bitclr, bitset, bitinvert patterns if appropriate.
4268 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4269 use rx_fuse_in_memory_bitop.
4270 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4271 to named insn, correct maximum insn length.
4273 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4276 * machmode.def: Define a complex mode for PARTIAL_INT.
4277 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4279 * doc/rtl.texi: Document CSPImode.
4280 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4282 (msp430_hard_regno_nregs_with_padding): Likewise.
4284 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4287 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4289 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4291 PR rtl-optimization/84169
4292 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4293 we generated a parallel as new i3 and we split that to new i2 and i3
4294 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4295 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4296 those to i2, not i1. Partially rewrite this scan code.
4298 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4301 * stor-layout.c (place_field): For variable length fields, adjust
4302 offset_align afterwards not just based on the field's alignment,
4303 but also on the size.
4306 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4307 of exps and logs in the use_exp2 case.
4309 2018-02-13 Jeff Law <law@redhat.com>
4311 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4314 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4317 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4321 * common.opt (gas-loc-support, gas-locview-support): New.
4322 (ginline-points, ginternal-reset-location-views): New.
4323 * doc/invoke.texi: Document them. Use @itemx where intended.
4324 (gvariable-location-views): Adjust.
4325 * target.def (reset_location_view): New.
4326 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4327 (TARGET_RESET_LOCATION_VIEW): New.
4328 * doc/tm.texi: Rebuilt.
4329 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4330 (dwarf2out_default_as_locview_support): New.
4331 (output_asm_line_debug_info): Use option variables.
4332 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4333 (output_loc_list): Likewise.
4334 (add_high_low_attributes): Check option variables.
4335 Don't output entry view attribute in strict mode.
4336 (gen_inlined_subroutine_die): Check option variables.
4337 (dwarf2out_inline_entry): Likewise.
4338 (init_sections_and_labels): Likewise.
4339 (dwarf2out_early_finish): Likewise.
4340 (maybe_reset_location_view): New, from...
4341 (dwarf2out_var_location): ... here. Call it.
4342 * debug.h (dwarf2out_default_as_loc_support): Declare.
4343 (dwarf2out_default_as_locview_support): Declare.
4344 * hooks.c (hook_int_rtx_insn_0): New.
4345 * hooks.h (hook_int_rtx_insn_0): Declare.
4346 * toplev.c (process_options): Take -gas-loc-support and
4347 -gas-locview-support from dwarf2out. Enable
4348 -gvariable-location-views by default only with locview
4349 assembler support. Enable -ginternal-reset-location-views by
4350 default only if the target defines the corresponding hook.
4351 Enable -ginline-points by default if location views are
4352 enabled; force it disabled if statement frontiers are
4354 * tree-inline.c (expand_call_inline): Check option variables.
4355 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4357 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4359 PR tree-optimization/84321
4360 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4361 handling. Also check whether the anti-range contains any values
4362 that satisfy the mask; switch to a VR_RANGE if not.
4364 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4367 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4369 2018-02-13 Martin Jambor <mjambor@suse.cz>
4372 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4373 of call statements, also set location of a load to a temporary.
4375 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4377 * config/rl78/rl78.c (add_vector_labels): New function.
4378 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4379 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4380 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4381 which checks that no arguments are passed.
4382 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4383 * doc/extend.texi: Documentation for the new attribute.
4385 2018-02-13 Andreas Schwab <schwab@suse.de>
4387 * config/riscv/linux.h (CPP_SPEC): Define.
4389 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4392 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4393 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4394 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4395 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4396 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4397 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4398 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4399 temporarily for AES and PCLMUL builtins.
4401 PR tree-optimization/84339
4402 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4403 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4407 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4408 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4409 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4411 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4413 * omp-simd-clone.h: New file.
4414 * omp-simd-clone.c: Include omp-simd-clone.h.
4415 (expand_simd_clones): No longer static.
4416 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4417 cgraph.h and omp-simd-clone.h.
4418 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4419 (vect_recog_widen_shift_pattern): Formatting fix.
4420 (vect_pattern_recog_1): Don't check optab for calls.
4423 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4424 operands[2] into a REG before using gen_lowpart on it.
4426 2018-02-12 Jeff Law <law@redhat.com>
4429 * config/sh/sh.c (find_barrier): Consider a sibling call
4432 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4433 successfully back substituting a reg.
4435 2018-02-12 Richard Biener <rguenther@suse.de>
4437 PR tree-optimization/84037
4438 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4439 parameter, move visited init to caller.
4440 (vect_slp_analyze_operations): Separate cost from validity
4441 check, initialize visited once for all instances.
4442 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4444 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4446 (vect_model_promotion_demotion_cost): Likewise.
4447 (vectorizable_bswap): Guard cost modeling with !slp_node
4448 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4450 (vectorizable_call): Likewise.
4451 (vectorizable_conversion): Likewise.
4452 (vectorizable_assignment): Likewise.
4453 (vectorizable_shift): Likewise.
4454 (vectorizable_operation): Likewise.
4455 (vectorizable_store): Likewise.
4456 (vectorizable_load): Likewise.
4457 (vectorizable_condition): Likewise.
4458 (vectorizable_comparison): Likewise.
4460 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4463 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4464 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4465 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4467 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4470 * params.def (inline-min-speedup): Increase from 8 to 15.
4471 (max-inline-insns-auto): Decrease from 40 to 30.
4472 * ipa-split.c (consider_split): Add some buffer for function to
4473 be considered inlining candidate.
4474 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4477 2018-02-12 Richard Biener <rguenther@suse.de>
4479 PR tree-optimization/84037
4480 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4481 matched stmts if we cannot swap the non-matched ones.
4483 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4485 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4486 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4487 _mm_maskz_scalef_round_ss): New intrinsics.
4488 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4489 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4490 __builtin_ia32_scalefss_round): Remove.
4491 (__builtin_ia32_scalefsd_mask_round,
4492 __builtin_ia32_scalefss_mask_round): New intrinsics.
4493 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4494 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4495 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4496 "<round_constraint>")): Changed to ...
4497 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4498 "<round_scalar_constraint>")): ... this.
4499 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4500 %0, %1, %2<round_op3>}"): Changed to ...
4501 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4502 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4503 %2<round_scalar_mask_op3>}"): ... this.
4504 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4506 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4508 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4509 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4510 (_mm_maskz_sqrt_round_ss): New intrinsics.
4511 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4512 (__builtin_ia32_sqrtsd_mask_round)
4513 (__builtin_ia32_sqrtss_mask_round): New builtins.
4514 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4515 (__builtin_ia32_sqrtss_round): Remove.
4516 (__builtin_ia32_sqrtsd_mask_round)
4517 (__builtin_ia32_sqrtss_mask_round): New builtins.
4518 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4519 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4520 ((match_operand:VF_128 1 "vector_operand"
4521 "xBm,<round_constraint>")): Changed to ...
4522 ((match_operand:VF_128 1 "vector_operand"
4523 "xBm,<round_scalar_constraint>")): ... this.
4524 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4525 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4526 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4527 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4528 %<iptr>1<round_scalar_mask_op3>}): ... this.
4529 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4530 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4532 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4535 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4536 Cast vec_cmpeq result to correct type.
4537 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4538 Cast vec_cmpgt result to correct type.
4540 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4542 * final.c (final_scan_insn_1): Renamed from...
4543 (final_scan_insn): ... this. New wrapper, to recover
4544 seen from the outermost call in recursive ones.
4545 * config/sparc/sparc.c (output_return): Drop seen from call.
4546 (output_sibcall): Likewise.
4547 * config/visium/visium.c (output_branch): Likewise.
4549 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4551 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4554 2018-02-10 Alan Modra <amodra@gmail.com>
4557 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4558 Specify LR as an input.
4560 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4563 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4564 remove_member_access_dummy_vars): New functions.
4565 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4566 lower_omp_1, execute_lower_omp): Use them.
4568 PR rtl-optimization/84308
4569 * shrink-wrap.c (spread_components): Release todo vector.
4571 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4573 PR rtl-optimization/57193
4574 * ira-color.c (struct allocno_color_data): Add member
4575 conflict_allocno_hard_prefs.
4576 (update_conflict_allocno_hard_prefs): New.
4577 (bucket_allocno_compare_func): Add a preference based on
4578 conflict_allocno_hard_prefs.
4579 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4580 (color_allocnos): Remove a dead code. Initiate
4581 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4583 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4586 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4587 constraint from =wa to wa. Avoid a subreg on the output operand,
4588 instead use a pseudo and subreg it in a move.
4589 (p9_xxbrd_<mode>): Changed to ...
4590 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4591 (p9_xxbrd_v2df): New expander.
4592 (p9_xxbrw_<mode>): Changed to ...
4593 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4594 (p9_xxbrw_v4sf): New expander.
4596 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4598 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4600 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4603 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4604 multiply in 32-bit mode.
4605 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4606 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4609 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4611 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4612 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4613 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4614 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4616 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4619 * dwarf2out.c (is_trivial_indirect_ref): New function.
4620 (dwarf2out_late_global_decl): Do not generate a location
4621 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4622 and that are not defined in the current unit.
4624 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4626 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4627 instead of a libcall for UNORDERED.
4629 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4632 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4633 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4635 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4638 * config/s390/s390.c (s390_set_current_function): Invoke
4639 s390_indirect_branch_settings also if fndecl didn't change.
4641 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4643 * config/rs6000/rs6000.md (blockage): Set length to zero.
4645 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4647 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4649 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4652 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4653 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4657 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4658 PARALLEL incoming that failed vt_get_decl_and_offset check.
4661 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4662 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4664 (get_variable_section): For decls in named .bss* sections pass true as
4665 second argument to bss_initializer_p.
4667 2018-02-09 Marek Polacek <polacek@redhat.com>
4668 Jakub Jelinek <jakub@redhat.com>
4671 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4672 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4673 Sync some changes from cxx_fold_indirect_ref.
4675 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4677 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4679 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4680 (BLOCK_INLINE_ENTRY_LABEL): New.
4681 (dwarf2out_var_location): Disregard inline entry markers.
4682 (inline_entry_data): New struct.
4683 (inline_entry_data_hasher): New hashtable type.
4684 (inline_entry_data_hasher::hash): New.
4685 (inline_entry_data_hasher::equal): New.
4686 (inline_entry_data_table): New variable.
4687 (add_high_low_attributes): Add DW_AT_entry_pc and
4688 DW_AT_GNU_entry_view attributes if a pending entry is found
4689 in inline_entry_data_table. Add old entry_pc attribute only
4690 if debug nonbinding markers are disabled.
4691 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4692 markers are enabled.
4693 (block_within_block_p, dwarf2out_inline_entry): New.
4694 (dwarf2out_finish): Check that no entries remained in
4695 inline_entry_data_table.
4696 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4697 (final_scan_insn, notice_source_line): Likewise.
4698 (rest_of_clean_state): Skip inline entry markers.
4699 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4701 * gimple.c (gimple_build_debug_inline_entry): New.
4702 * gimple.h (enum gimple_debug_subcode): Add
4703 GIMPLE_DEBUG_INLINE_ENTRY.
4704 (gimple_build_debug_inline_entry): Declare.
4705 (gimple_debug_inline_entry_p): New.
4706 (gimple_debug_nonbind_marker_p): Adjust.
4707 * insn-notes.def (INLINE_ENTRY): New.
4708 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4709 inline entry marker notes.
4710 (print_insn): Likewise.
4711 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4712 (INSN_DEBUG_MARKER_KIND): Likewise.
4713 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4714 * tree-inline.c (expand_call_inline): Build and insert
4715 debug_inline_entry stmt.
4716 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4717 inline entry blocks early, if nonbind markers are enabled.
4718 (dump_scope_block): Dump fragment info.
4719 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4720 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4721 (gimple_build_debug_inline_entry): New.
4722 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4723 Enable/disable inline entry points too.
4724 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4725 (DEBUG_INSN): Describe inline entry markers.
4727 * common.opt (gvariable-location-views): New.
4728 (gvariable-location-views=incompat5): New.
4729 * config.in: Rebuilt.
4730 * configure: Rebuilt.
4731 * configure.ac: Test assembler for view support.
4732 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4733 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4734 * dwarf2out.c (var_loc_view): New typedef.
4735 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4736 (dwarf2out_locviews_in_attribute): New.
4737 (dwarf2out_locviews_in_loclist): New.
4738 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4739 (enum dw_line_info_opcode): Add LI_adv_address.
4740 (struct dw_line_info_table): Add view.
4741 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4742 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4743 (zero_view_p): New variable.
4744 (ZERO_VIEW_P): New macro.
4745 (output_asm_line_debug_info): New.
4746 (struct var_loc_node): Add view.
4747 (add_AT_view_list, AT_loc_list): New.
4748 (add_var_loc_to_decl): Add view param. Test it against last.
4749 (new_loc_list): Add view params. Record them.
4750 (AT_loc_list_ptr): Handle loc and view lists.
4751 (view_list_to_loc_list_val_node): New.
4752 (print_dw_val): Handle dw_val_class_view_list.
4753 (size_of_die): Likewise.
4754 (value_format): Likewise.
4755 (loc_list_has_views): New.
4756 (gen_llsym): Set vl_symbol too.
4757 (maybe_gen_llsym, skip_loc_list_entry): New.
4758 (dwarf2out_maybe_output_loclist_view_pair): New.
4759 (output_loc_list): Output view list or entries too.
4760 (output_view_list_offset): New.
4761 (output_die): Handle dw_val_class_view_list.
4762 (output_dwarf_version): New.
4763 (output_compilation_unit_header): Use it.
4764 (output_skeleton_debug_sections): Likewise.
4765 (output_rnglists, output_line_info): Likewise.
4766 (output_pubnames, output_aranges): Update version comments.
4767 (output_one_line_info_table): Output view numbers in asm comments.
4768 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4769 Call maybe_gen_llsym.
4770 (loc_list_from_tree_1): Adjust.
4771 (add_AT_location_description): Create view list attribute if
4772 needed, check it's absent otherwise.
4773 (convert_cfa_to_fb_loc_list): Adjust.
4774 (maybe_emit_file): Call output_asm_line_debug_info for test.
4775 (dwarf2out_var_location): Reset views as needed. Precompute
4776 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4777 attribute. Set view.
4778 (new_line_info_table): Reset next view.
4779 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4780 (dwarf2out_source_line): Likewise. Output view resets and labels to
4781 the assembler, or select appropriate line info opcodes.
4782 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4783 (optimize_string_length): Catch it. Adjust.
4784 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4785 dw_val_class_view_list, and remove it if no longer needed.
4786 (hash_loc_list): Hash view numbers.
4787 (loc_list_hasher::equal): Compare them.
4788 (optimize_location_lists): Check whether a view list symbol is
4789 needed, and whether the locview attribute is present, and
4790 whether they match. Remove the locview attribute if no longer
4792 (index_location_lists): Call skip_loc_list_entry for test.
4793 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4794 Use output_dwarf_version.
4795 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4796 (struct dw_val_node): Add val_view_list.
4797 * final.c (SEEN_NEXT_VIEW): New.
4798 (set_next_view_needed): New.
4799 (clear_next_view_needed): New.
4800 (maybe_output_next_view): New.
4801 (final_start_function): Rename to...
4802 (final_start_function_1): ... this. Take pointer to FIRST,
4803 add SEEN parameter. Emit param bindings in the initial view.
4804 (final_start_function): Reintroduce SEEN-less interface.
4805 (final): Rename to...
4806 (final_1): ... this. Take SEEN parameter. Output final pending
4807 next view at the end.
4808 (final): Reintroduce seen-less interface.
4809 (final_scan_insn): Output pending next view before switching
4810 sections or ending a block. Mark the next view as needed when
4811 outputting variable locations. Notify debug backend of section
4812 changes, and of location view changes.
4813 (rest_of_handle_final): Adjust.
4814 * toplev.c (process_options): Autodetect value for debug variable
4815 location views option. Warn on incompat5 without -gdwarf-5.
4816 * doc/invoke.texi (gvariable-location-views): New.
4817 (gvariable-location-views=incompat5): New.
4818 (gno-variable-location-views): New.
4820 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4822 PR tree-optimization/84136
4823 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4824 that the result of find_edge is non-NULL.
4826 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4829 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4830 storing integer register in SImode. Fix cost of 256 and 512
4831 byte aligned SSE register store.
4833 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4835 * config/i386/i386.c (ix86_multiplication_cost): Fix
4836 multiplication cost for TARGET_AVX512DQ.
4838 2018-02-08 Marek Polacek <polacek@redhat.com>
4840 PR tree-optimization/84238
4841 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4844 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4846 PR tree-optimization/84265
4847 * tree-vect-stmts.c (vectorizable_store): Don't treat
4848 VMAT_CONTIGUOUS accesses as grouped.
4849 (vectorizable_load): Likewise.
4851 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4853 PR tree-optimization/81635
4854 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4855 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4856 (test_round_for_mask): New functions.
4857 (wide_int_cc_tests): Call test_round_for_mask.
4858 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4859 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4860 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4861 range returned by get_range_info.
4863 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4866 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4867 * symtab.c: Include builtins.h
4868 (symtab_node::output_to_lto_symbol_table_p): Move here
4869 from lto-streamer-out.c:output_symbol_p.
4870 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4871 (output_symbol_p): Move all logic to symtab.c
4872 (produce_symtab): Update.
4874 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4876 * config/s390/s390-opts.h (enum indirect_branch): Define.
4877 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4878 (s390_indirect_branch_via_thunk)
4879 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4880 (enum s390_indirect_branch_type): Define.
4881 * config/s390/s390.c (struct s390_frame_layout, struct
4882 machine_function): Remove.
4883 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4884 (indirect_branch_table_label_no, indirect_branch_table_name):
4886 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4887 (enum s390_indirect_branch_option): Define.
4888 (s390_return_addr_from_memory): New function.
4889 (s390_handle_string_attribute): New function.
4890 (s390_attribute_table): Add new attribute handler.
4891 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4892 (s390_indirect_branch_via_thunk): New function.
4893 (s390_indirect_branch_via_inline_thunk): New function.
4894 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4895 sibling call optimization for non z10 compiles.
4896 (s390_emit_call): Force indirect branch target to be a single
4897 register. Add r1 clobber for non-z10 compiles.
4898 (s390_emit_epilogue): Emit return jump via return_use expander.
4899 (s390_reorg): Handle JUMP_INSNs as execute targets.
4900 (s390_option_override_internal): Perform validity checks for the
4901 new command line options.
4902 (s390_indirect_branch_attrvalue): New function.
4903 (s390_indirect_branch_settings): New function.
4904 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4905 (s390_output_indirect_thunk_function): New function.
4906 (s390_code_end): Implement target hook.
4907 (s390_case_values_threshold): Implement target hook.
4908 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4910 * config/s390/s390.h (struct s390_frame_layout)
4911 (struct machine_function): Move here from s390.c.
4912 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4913 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4914 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4915 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4916 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4917 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4918 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4919 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4920 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4921 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4922 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4923 (mnemonic attribute): Add values which aren't recognized
4925 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4926 pattern for branch conversion. Fix mnemonic attribute.
4927 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4928 indirect branch via thunk if requested.
4929 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4930 ("*indirect_jump"): Disable for branch conversion using out of
4932 ("indirect_jump_via_thunk<mode>_z10")
4933 ("indirect_jump_via_thunk<mode>")
4934 ("indirect_jump_via_inlinethunk<mode>_z10")
4935 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4936 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4937 ("casesi_jump_via_inlinethunk<mode>_z10")
4938 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4939 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4940 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4941 ("*indirect2_jump"): Disable for branch conversion.
4942 ("casesi_jump"): Turn into expander and expand patterns for branch
4944 ("return_use"): New expander.
4945 ("*return"): Emit return via thunk and rename it to ...
4946 ("*return<mode>"): ... this one.
4947 * config/s390/s390.opt: Add new options and and enum for the
4950 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4952 * lra-constraints.c (match_reload): Unconditionally use
4953 gen_lowpart_SUBREG, rather than selecting between that
4954 and equivalent gen_rtx_SUBREG code.
4956 2018-02-08 Richard Biener <rguenther@suse.de>
4958 PR tree-optimization/84233
4959 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4960 changed flag instead of boguously re-using phi_inserted.
4962 2018-02-08 Martin Jambor <mjambor@suse.cz>
4964 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4965 static local variables.
4967 2018-02-08 Richard Biener <rguenther@suse.de>
4969 PR tree-optimization/84278
4970 * tree-vect-stmts.c (vectorizable_store): When looking for
4971 smaller vector types to perform grouped strided loads/stores
4972 make sure the mode is supported by the target.
4973 (vectorizable_load): Likewise.
4975 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4977 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4978 Increase LDP/STP opportunities by adding adjacent callee-saves.
4980 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4982 PR rtl-optimization/84068
4983 PR rtl-optimization/83459
4984 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4986 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4988 PR tree-optimization/84224
4989 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4990 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4993 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4996 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4997 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4998 count, remove one USE.
5000 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5002 * doc/install.texi (Configuration): Document the
5003 --with-long-double-format={ibm,ieee} PowerPC configuration
5007 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5008 Convert from define_expand to be define_insn_and_split. Rework
5009 float/double/_Float128 conversions to QI/HI/SImode to work with
5010 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5011 conversions to QI/HImode types did a store and then a load to
5012 truncate the value. For conversions to VSX registers, don't split
5013 the insn, instead emit the code directly. Use the code iterator
5014 any_fix to combine signed and unsigned conversions.
5015 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5016 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5017 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5018 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5019 (fix_<mode>di2_hw): Likewise.
5020 (fixuns_<mode>di2_hw): Likewise.
5021 (fix_<mode>si2_hw): Likewise.
5022 (fixuns_<mode>si2_hw): Likewise.
5023 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5024 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5025 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5026 fix<uns>_trunc<SFDF:mode>si2_p8.
5027 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5029 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5030 (fix<uns>_<mode>_mem): Likewise.
5031 (fctiw<u>z_<mode>_mem): Likewise.
5032 (fix<uns>_<mode>_mem): Likewise.
5033 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5034 the register allocator from doing a direct move to the GPRs to do
5035 a store, and instead use the ISA 3.0 store byte/half-word from
5036 vector register instruction. For IEEE 128-bit floating point,
5037 also optimize stores of 32-bit ints.
5038 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5040 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5042 * genextract.c (push_pathstr_operand): New function to support
5044 (walk_rtx): Call push_pathstr_operand.
5045 (print_path): Support [a-zA-Z].
5047 2018-02-07 Richard Biener <rguenther@suse.de>
5049 PR tree-optimization/84037
5050 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5051 (cse_and_gimplify_to_preheader): Declare.
5052 (vect_get_place_in_interleaving_chain): Likewise.
5053 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5055 (_loop_vec_info::~_loop_vec_info): Delete it.
5056 (cse_and_gimplify_to_preheader): New function.
5057 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5058 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5059 (vectorizable_load): Likewise. For grouped stores always base
5060 the IV on the first element.
5061 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5062 condition before gimplifying.
5064 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5066 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5067 *DIV_EXPR and *MOD_EXPR.
5069 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5072 * config/i386/i386.c (ix86_option_override_internal): Mask out
5073 the CF_SET bit when checking -fcf-protection.
5075 2018-02-07 Tom de Vries <tom@codesourcery.com>
5078 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5081 2018-02-07 Richard Biener <rguenther@suse.de>
5083 PR tree-optimization/84204
5084 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5087 PR tree-optimization/84205
5088 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5089 special-case isl_ast_op_zdiv_r.
5091 PR tree-optimization/84223
5092 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5093 Only add conditions from within the region.
5094 (gather_bbs::after_dom_children): Adjust.
5096 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5099 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5100 * config/avr/avr.md: Only post-reload split REG-REG moves if
5101 either register is GENERAL_REG_P.
5103 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5105 PR tree-optimization/84235
5106 * tree-ssa-scopedtables.c
5107 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5108 if the subtraction is performed in floating point type where NaNs are
5109 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5110 build 1. Formatting fix.
5112 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5115 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5116 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5117 and skip it regardless of bb boundaries. Use CALL_P macro,
5118 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5119 unnecessarily, formatting fix.
5121 2018-02-06 Michael Collison <michael.collison@arm.com>
5123 * config/arm/thumb2.md:
5124 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5125 (*thumb_mov_notscc): Ditto.
5127 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5130 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5131 unsigned_fix, not "s".
5133 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5135 * configure.ac (gcc_fn_eh_frame_ro): New function.
5136 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5137 correct .eh_frame permissions.
5138 * configure: Regenerate.
5140 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5142 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5145 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5147 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5148 Display warning message for -mno-speculate-indirect-jumps.
5150 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5152 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5154 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5156 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5158 PR tree-optimization/84225
5159 * tree-eh.c (find_trapping_overflow): Only call
5160 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5162 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5165 * config/i386/i386.c: Reimplement the check of possible options
5166 -mibt/-mshstk conbination. Change error messages.
5167 * doc/invoke.texi: Fix a typo: remove extra '='.
5169 2018-02-06 Marek Polacek <polacek@redhat.com>
5171 PR tree-optimization/84228
5172 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5174 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5177 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5178 emitted arch directives.
5179 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5180 __ARM_FEATURE_COPROC before changing architectures.
5182 2018-02-06 Richard Biener <rguenther@suse.de>
5184 * config/i386/i386.c (print_reg): Fix typo.
5185 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5187 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5189 * configure: Regenerate.
5191 2018-02-05 Martin Sebor <msebor@redhat.com>
5193 PR tree-optimization/83369
5194 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5197 2018-02-05 Martin Liska <mliska@suse.cz>
5199 * doc/invoke.texi: Cherry-pick upstream r323995.
5201 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5203 * ira.c (ira_init_register_move_cost): Adjust comment.
5205 2018-02-05 Martin Liska <mliska@suse.cz>
5207 PR gcov-profile/84137
5208 * doc/gcov.texi: Fix typo in documentation.
5210 2018-02-05 Martin Liska <mliska@suse.cz>
5212 PR gcov-profile/83879
5213 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5214 using dlopen functionality.
5216 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5218 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5219 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5220 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5221 _mm_maskz_range_round_ss): New intrinsics.
5222 (__builtin_ia32_rangesd128_round)
5223 (__builtin_ia32_rangess128_round): Remove.
5224 (__builtin_ia32_rangesd128_mask_round,
5225 __builtin_ia32_rangess128_mask_round): New builtins.
5226 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5227 __builtin_ia32_rangess128_round): Remove.
5228 (__builtin_ia32_rangesd128_mask_round,
5229 __builtin_ia32_rangess128_mask_round): New builtins.
5230 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5231 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5232 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5233 "<round_saeonly_constraint>")): Changed to ...
5234 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5235 "<round_saeonly_scalar_constraint>")): ... this.
5236 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5237 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5238 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5239 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5240 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5242 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5244 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5246 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5247 Remove all values except native, 8540 and 8548.
5249 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5251 * config/i386/i386.c (ix86_output_function_return): Pass
5252 INVALID_REGNUM, instead of -1, as invalid register number to
5253 indirect_thunk_name and output_indirect_thunk.
5255 2018-02-02 Julia Koval <julia.koval@intel.com>
5257 * config.gcc: Add -march=icelake.
5258 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5259 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5260 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5261 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5262 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5263 (processor_target_table): Add icelake.
5264 (ix86_option_override_internal): Handle new PTAs.
5265 (get_builtin_code_for_version): Handle icelake.
5266 (M_INTEL_COREI7_ICELAKE): New.
5267 (fold_builtin_cpu): Handle icelake.
5268 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5269 * doc/invoke.texi: Add -march=icelake.
5271 2018-02-02 Julia Koval <julia.koval@intel.com>
5273 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5274 to wide_int_bitmask.
5275 * wide-int-bitmask.h: New.
5277 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5280 * config/i386/i386.md: Replace Pmode with word_mode in
5281 builtin_setjmp_setup and builtin_longjmp to support x32.
5283 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5287 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5289 (rs6000_supported_cpu_names): New static variable.
5290 (linux_cpu_translation_table): Likewise.
5291 (elf_platform) <cpu>: Define new static variable and use it.
5292 Translate kernel AT_PLATFORM name to canonical name if needed.
5293 Error if platform name is unknown.
5295 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5298 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5300 2018-02-01 Jeff Law <law@redhat.com>
5303 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5304 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5305 the scratch if RELEASE_VIA_POP is false.
5306 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5307 If we have to save a temporary register, decrement SIZE appropriately.
5308 Pass new arguments to release_scratch_register_on_entry.
5309 (ix86_adjust_stack_and_probe): Likewise.
5310 (ix86_emit_probe_stack_range): Pass new arguments to
5311 release_scratch_register_on_entry.
5313 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5315 PR rtl-optimization/84157
5316 * combine.c (change_zero_ext): Use REG_P predicate in
5317 front of HARD_REGISTER_P predicate.
5319 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5321 * config/avr/avr.c (avr_option_override): Move disabling of
5322 -fdelete-null-pointer-checks to...
5323 * common/config/avr/avr-common.c (avr_option_optimization_table):
5326 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5328 PR tree-optimization/81635
5329 * tree-data-ref.c (split_constant_offset_1): For types that
5330 wrap on overflow, try to use range info to prove that wrapping
5333 2018-02-01 Renlin Li <renlin.li@arm.com>
5336 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5338 (aarch64_register_move_cost): Likewise.
5339 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5341 (REG_CLASS_NAMES): Likewise.
5342 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5343 TAILCALL_ADDR_REGS. Remove IP registers.
5344 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5346 2018-02-01 Richard Biener <rguenther@suse.de>
5348 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5349 for specifying RPO order and allow NULL for that.
5350 * domwalk.c (dom_walker::dom_walker): Likewise.
5351 (dom_walker::walk): Handle NULL RPO order.
5352 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5354 (rewrite_update_dom_walker): Likewise.
5355 (mark_def_dom_walker): Likewise.
5357 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5359 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5360 (aarch64_maybe_expand_sve_subreg_move): Declare.
5361 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5362 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5364 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5365 that are semantically a reverse operation.
5366 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5367 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5368 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5370 (aarch64_can_change_mode_class): For big-endian, forbid changes
5371 between two SVE modes if they have different element sizes.
5373 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5375 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5376 the TImode handling for big-endian targets.
5378 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5380 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5381 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5383 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5384 Remove BSWAP handing for big-endian targets and use the form of
5385 LD1RQ appropariate for the mode.
5387 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5389 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5390 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5393 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5396 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5397 check for operands that need to go through aarch64_sve_reload_be.
5399 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5401 PR tree-optimization/81661
5402 PR tree-optimization/84117
5403 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5404 * tree-eh.c: Include gimplify.h.
5405 (find_trapping_overflow, replace_trapping_overflow,
5406 rewrite_to_non_trapping_overflow): New functions.
5407 * tree-vect-loop.c: Include tree-eh.h.
5408 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5409 * tree-data-ref.c: Include tree-eh.h.
5410 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5412 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5414 PR rtl-optimization/84123
5415 * combine.c (change_zero_ext): Check if hard register satisfies
5416 can_change_dest_mode before calling gen_lowpart_SUBREG.
5418 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5421 * ira.c (ira_init_register_move_cost): Remove assert.
5423 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5425 PR rtl-optimization/84071
5426 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5427 * doc/tm.texi: Regenerate.
5429 2018-01-31 Richard Biener <rguenther@suse.de>
5431 PR tree-optimization/84132
5432 * tree-data-ref.c (analyze_miv_subscript): Properly
5433 check whether evolution_function_is_affine_multivariate_p
5434 before calling gcd_of_steps_may_divide_p.
5436 2018-01-31 Julia Koval <julia.koval@intel.com>
5439 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5440 * config/i386/i386.md (rdpid_rex64) New.
5441 (rdpid): Make 32bit only.
5443 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5446 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5447 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5449 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5452 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5454 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5456 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5458 PR rtl-optimization/84071
5459 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5460 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5462 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5464 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5465 (arc_attribute_table): Add 'aux' attribute.
5466 (arc_in_small_data_p): Consider aux like variables.
5467 (arc_is_aux_reg_p): New function.
5468 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5469 (arc_get_aux_arg): New function.
5470 (prepare_move_operands): Handle aux-register access.
5471 (arc_handle_aux_attribute): New function.
5472 * doc/extend.texi (ARC Variable attributes): Add subsection.
5474 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5476 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5477 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5478 (arc_attribute_table): Add 'uncached' attribute.
5479 (arc_print_operand): Print '.di' flag for uncached memory
5481 (arc_in_small_data_p): Do not consider for small data the uncached
5483 (arc_is_uncached_mem_p): New function.
5484 * config/arc/predicates.md (compact_store_memory_operand): Check
5485 for uncached memory accesses.
5486 (nonvol_nonimm_operand): Likewise.
5487 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5489 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5492 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5493 falign-loops=): Add Optimization flag.
5495 2018-01-30 Jeff Law <law@redhat.com>
5498 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5499 INT_REGISTERS_SAVED. Check it prior to calling
5500 get_scratch_register_on_entry.
5501 (ix86_adjust_stack_and_probe): Similarly.
5502 (ix86_emit_probe_stack_range): Similarly.
5503 (ix86_expand_prologue): Corresponding changes.
5505 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5508 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5509 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5511 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5514 * lra-constraints.c (curr_insn_transform): Process AND in the
5517 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5519 PR rtl-optimization/83986
5520 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5521 dependence against last_pending_memory_flush in addition to
5524 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5526 PR tree-optimization/81611
5527 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5530 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5533 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5536 2018-01-30 Richard Biener <rguenther@suse.de>
5537 Jakub Jelinek <jakub@redhat.com>
5539 PR tree-optimization/84111
5540 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5541 inner loops added during recursion, as they don't have up-to-date
5544 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5547 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5548 (can_inline_edge_by_limits_p): ... here.
5549 (can_early_inline_edge_p, check_callers,
5550 update_caller_keys, update_callee_keys, recursive_inlining,
5551 add_new_edges_to_heap, speculation_useful_p,
5552 inline_small_functions,
5553 inline_small_functions, flatten_function,
5554 inline_to_all_callers_1): Update.
5556 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5558 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5561 2018-01-30 Richard Biener <rguenther@suse.de>
5563 PR tree-optimization/83008
5564 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5565 invariant and constant vector uses in stmts when they need
5568 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5571 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5572 * configure: Regenerate.
5574 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5576 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5578 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5579 Use gen_rtx_REG rather than gen_lowpart.
5581 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5583 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5584 rather than 0 when creating partial subregs.
5586 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5588 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5591 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5594 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5595 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5596 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5597 flags. This restores the settings used before the 2017-07-24.
5598 Turning off pre increment/decrement/modify allows IVOPTS to
5599 optimize DF/SF loops where the index is an int.
5601 2018-01-29 Richard Biener <rguenther@suse.de>
5602 Kelvin Nilsen <kelvin@gcc.gnu.org>
5605 * tree-vect-stmts.c (vectorizable_call): Don't call
5606 targetm.vectorize_builtin_md_vectorized_function if callee is
5609 2018-01-22 Carl Love <cel@us.ibm.com>
5611 * doc/extend.tex: Fix typo in second arg in
5612 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5614 2018-01-29 Richard Biener <rguenther@suse.de>
5616 PR tree-optimization/84086
5617 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5618 (flush_ssaname_freelist): When SSA names were released reset
5619 the SCEV hash table.
5621 2018-01-29 Richard Biener <rguenther@suse.de>
5623 PR tree-optimization/84057
5624 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5625 removed paths when removing edges.
5627 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5629 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5630 -mfunction-return=@var{choice}.
5632 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5635 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5637 (layout::print_source_line): Likewise.
5638 (test_get_line_width_without_trailing_whitespace): Add test cases.
5640 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5643 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5646 2018-01-26 Jim Wilson <jimw@sifive.com>
5648 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5650 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5653 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5655 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5656 and CMP + SUB-immediate -> SUBS.
5658 2018-01-26 Martin Sebor <msebor@redhat.com>
5660 PR tree-optimization/83896
5661 * tree-ssa-strlen.c (get_string_len): Rename...
5662 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5663 Avoid assuming length is constant.
5664 (handle_char_store): Use HOST_WIDE_INT for string length.
5666 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5669 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5670 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5672 2018-01-26 Richard Biener <rguenther@suse.de>
5674 PR rtl-optimization/84003
5675 * dse.c (record_store): Only record redundant stores when
5676 the earlier store aliases at least all accesses the later one does.
5678 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5680 PR rtl-optimization/83985
5681 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5682 REG_CFA_RESTORE insns.
5683 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5684 REG_CFA_RESTORE insns here.
5687 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5688 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5690 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5692 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5693 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5694 (arc_init): Likewise.
5695 (arc_override_options): Likewise.
5696 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5698 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5700 * config/arc/arc.h (TARGET_DBNZ): Define.
5701 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5702 properly set the tune attribute.
5703 (dbnz): Use TARGET_DBNZ guard.
5704 * config/arc/arc.opt (mtune): Add core3 option.
5706 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5708 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5709 recognize new pic like addresses.
5710 (arc_delegitimize_address): Clean up.
5712 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5714 * config/arc/arc-arches.def: Option mrf16 valid for all
5716 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5717 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5718 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5719 * config/arc/arc-tables.opt: Regenerate.
5720 * config/arc/arc.c (arc_conditional_register_usage): Handle
5721 reduced register file case.
5722 (arc_file_start): Set must have build attributes.
5723 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5725 * config/arc/arc.opt (mrf16): Add new option.
5726 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5727 * config/arc/genmultilib.awk: Handle new mrf16 option.
5728 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5729 * config/arc/t-multilib: Regenerate.
5730 * doc/invoke.texi (ARC Options): Document mrf16 option.
5732 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5734 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5735 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5736 (arc_attribute_table): Add 'secure_call' attribute.
5737 (arc_print_operand): Print secure call operand.
5738 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5740 (arc_is_secure_call_p): New function. * config/arc/arc.md
5741 (call_i): Add support for sjli instruction.
5742 (call_value_i): Likewise.
5743 * config/arc/constraints.md (Csc): New constraint.
5745 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5746 John Eric Martin <John.Martin@emmicro-us.com>
5748 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5749 * config/arc/arc.c (_arc_jli_section): New struct.
5750 (arc_jli_section): New type.
5751 (rc_jli_sections): New static variable.
5752 (arc_handle_jli_attribute): New function.
5753 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5754 (arc_file_end): New function.
5755 (TARGET_ASM_FILE_END): Define.
5756 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5757 (arc_add_jli_section): New function.
5758 (jli_call_scan): Likewise.
5759 (arc_reorg): Call jli_call_scan.
5760 (arc_output_addsi): Remove 'S' from printing asm operand.
5761 (arc_is_jli_call_p): New function.
5762 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5764 (movhi_insn): Likewise.
5765 (movsi_insn): Likewise.
5766 (movsi_set_cc_insn): Likewise.
5767 (loadqi_update): Likewise.
5768 (load_zeroextendqisi_update): Likewise.
5769 (load_signextendqisi_update): Likewise.
5770 (loadhi_update): Likewise.
5771 (load_zeroextendhisi_update): Likewise.
5772 (load_signextendhisi_update): Likewise.
5773 (loadsi_update): Likewise.
5774 (loadsf_update): Likewise.
5775 (movsicc_insn): Likewise.
5776 (bset_insn): Likewise.
5777 (bxor_insn): Likewise.
5778 (bclr_insn): Likewise.
5779 (bmsk_insn): Likewise.
5780 (bicsi3_insn): Likewise.
5781 (cmpsi_cc_c_insn): Likewise.
5782 (movsi_ne): Likewise.
5783 (movsi_cond_exec): Likewise.
5784 (clrsbsi2): Likewise.
5795 (call_i): Remove 'S' asm letter, add jli instruction.
5796 (call_value_i): Likewise.
5797 * config/arc/arc.op (mjli-always): New option.
5798 * config/arc/constraints.md (Cji): New constraint.
5799 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5801 (subsf3_fpx): Likewise.
5802 (mulsf3_fpx): Likewise.
5803 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5805 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5807 * doc/invoke.texi (ARC): Document mjli-always option.
5809 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5811 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5812 avoid addition with 0 and use incw and decw where possible.
5814 2018-01-26 Richard Biener <rguenther@suse.de>
5816 PR tree-optimization/81082
5817 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5818 association if it requires casting to unsigned.
5819 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5820 from fold_plusminus_mult_expr to catch important cases late when
5821 range info is available.
5823 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5825 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5826 * configure.ac (hidden_linkonce): New test.
5827 * configure: Regenerate.
5828 * config.in: Regenerate.
5830 2018-01-26 Julia Koval <julia.koval@intel.com>
5832 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5833 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5834 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5835 _mm_mask_bitshuffle_epi64_mask): Fix type.
5836 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5837 USI_FTYPE_V4DI_V4DI_USI): Remove.
5838 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5839 __builtin_ia32_vpshufbitqmb256_mask,
5840 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5841 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5842 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5844 2018-01-26 Alan Modra <amodra@gmail.com>
5847 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5848 UNSPEC_VBPERMQ. Sort other unspecs.
5850 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5852 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5854 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5857 * predict.c (drop_profile): Do not push/pop cfun; update also
5859 (handle_missing_profiles): Fix logic looking for zero profiles.
5861 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5864 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5865 on functions with #pragma omp declare simd or functions with simd
5867 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5868 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5869 Remove trailing \n from warning_at calls.
5871 2018-01-25 Tom de Vries <tom@codesourcery.com>
5874 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5875 for neutered workers.
5877 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5880 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5881 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5883 2018-01-24 Jeff Law <law@redhat.com>
5886 * i386.c (get_probe_interval): Move to earlier point.
5887 (ix86_compute_frame_layout): If -fstack-clash-protection and
5888 the frame is larger than the probe interval, then use pushes
5889 to save registers rather than reg->mem moves.
5890 (ix86_expand_prologue): Remove conditional for int_registers_saved
5893 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5896 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5897 min/max for never referenced object.
5899 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5902 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5904 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5905 attributes from DECL_ATTRIBUTES (decl) without affecting
5906 DECL_ATTRIBUTES (current_function_decl).
5907 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5908 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5910 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5912 PR tree-optimization/83979
5913 * fold-const.c (fold_comparison): Use constant_boolean_node
5914 instead of boolean_{true,false}_node.
5916 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5918 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5921 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5923 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5924 Simplify the clause that sets the length attribute.
5925 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5926 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5927 clause that sets the length attribute.
5928 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5930 2018-01-24 Tom de Vries <tom@codesourcery.com>
5933 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5934 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5935 Add strict parameter.
5936 (prevent_branch_around_nothing): Insert dummy insn between branch to
5937 label and label with no ptx insn inbetween.
5938 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5940 2018-01-24 Tom de Vries <tom@codesourcery.com>
5943 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5944 for neutered threads in warp.
5945 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5947 2018-01-24 Richard Biener <rguenther@suse.de>
5949 PR tree-optimization/83176
5950 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5953 2018-01-24 Richard Biener <rguenther@suse.de>
5955 PR tree-optimization/82819
5956 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5957 code generating pluses that are no-ops in the target precision.
5959 2018-01-24 Richard Biener <rguenther@suse.de>
5962 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5964 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5966 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5967 to merge probabilities.
5968 * predict.c (probably_never_executed): Also mark as cold functions
5969 with global 0 profile and guessed local profile.
5970 * profile-count.c (profile_probability::combine_with_count): New
5972 * profile-count.h (profile_probability::operator*,
5973 profile_probability::operator*=, profile_probability::operator/,
5974 profile_probability::operator/=): Reduce precision to adjusted
5975 and set value to guessed on contradictory divisions.
5976 (profile_probability::combine_with_freq): Remove.
5977 (profile_probability::combine_wiht_count): Declare.
5978 (profile_count::force_nonzero):: Set to adjusted.
5979 (profile_count::probability_in):: Set quality to adjusted.
5980 * tree-ssa-tail-merge.c (replace_block_by): Use
5983 2018-01-23 Andrew Waterman <andrew@sifive.com>
5984 Jim Wilson <jimw@sifive.com>
5986 * config/riscv/riscv.c (riscv_stack_boundary): New.
5987 (riscv_option_override): Set riscv_stack_boundary. Handle
5988 riscv_preferred_stack_boundary_arg.
5989 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5990 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5991 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5992 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5993 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5994 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5996 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5999 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6000 of struct ix86_frame.
6001 (ix86_expand_epilogue): Likewise. Add a local variable for
6002 the reg_save_offset field in struct ix86_frame.
6004 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6006 PR tree-optimization/82604
6007 * tree-loop-distribution.c (enum partition_kind): New enum item
6008 PKIND_PARTIAL_MEMSET.
6009 (partition_builtin_p): Support above new enum item.
6010 (generate_code_for_partition): Ditto.
6011 (compute_access_range): Differentiate cases that equality can be
6012 proven at all loops, the innermost loops or no loops.
6013 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6014 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6015 (finalize_partitions, distribute_loop): Don't fuse partition of
6016 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6017 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6020 2018-01-23 Martin Liska <mliska@suse.cz>
6022 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6023 order to ignore the predictor.
6024 (PRED_POLYMORPHIC_CALL): Likewise.
6025 (PRED_RECURSIVE_CALL): Likewise.
6027 2018-01-23 Martin Liska <mliska@suse.cz>
6029 * tree-profile.c (tree_profiling): Print function header to
6030 aware reader which function we are working on.
6031 * value-prof.c (gimple_find_values_to_profile): Do not print
6032 not interesting value histograms.
6034 2018-01-23 Martin Liska <mliska@suse.cz>
6036 * profile-count.h (enum profile_quality): Add
6037 profile_uninitialized as the first value. Do not number values
6038 as they are zero based.
6039 (profile_count::verify): Update sanity check.
6040 (profile_probability::verify): Likewise.
6042 2018-01-23 Nathan Sidwell <nathan@acm.org>
6044 * doc/invoke.texi (ffor-scope): Deprecate.
6046 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6048 PR tree-optimization/83510
6049 * domwalk.c (set_all_edges_as_executable): New function.
6050 (dom_walker::dom_walker): Convert bool param
6051 "skip_unreachable_blocks" to enum reachability. Move setup of
6052 edge flags to set_all_edges_as_executable and only do it when
6053 reachability is REACHABLE_BLOCKS.
6054 * domwalk.h (enum dom_walker::reachability): New enum.
6055 (dom_walker::dom_walker): Convert bool param
6056 "skip_unreachable_blocks" to enum reachability.
6057 (set_all_edges_as_executable): New decl.
6058 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6059 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6061 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6062 but converting true to REACHABLE_BLOCKS.
6063 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6065 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6066 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6067 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6069 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6070 if check_all_array_refs will be called.
6072 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6074 * tree.c (selftest::test_location_wrappers): Add more test
6077 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6079 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6080 (selftest::test_bit_in_range): Likewise.
6082 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6085 * doc/sourcebuild.texi (vect_float): Say that the selector
6086 only describes the situation when -funsafe-math-optimizations is on.
6087 (vect_float_strict): Document.
6089 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6091 PR tree-optimization/83965
6092 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6093 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6094 instead of checking only for a reduction.
6095 (vect_recog_widen_sum_pattern): Likewise.
6097 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6099 * predict.c (probably_never_executed): Only use precise profile info.
6100 (compute_function_frequency): Skip after inlining hack since we now
6101 have quality checking.
6103 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6105 * profile-count.h (profile_probability::very_unlikely,
6106 profile_probability::unlikely, profile_probability::even): Set
6107 precision to guessed.
6109 2018-01-23 Richard Biener <rguenther@suse.de>
6111 PR tree-optimization/83963
6112 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6113 Properly terminate dominator walk when crossing the exit edge not
6114 when visiting its source block.
6116 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6119 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6120 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6122 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6124 PR tree-optimization/83957
6125 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6126 semicolon after for body surrounded by braces.
6128 PR tree-optimization/83081
6129 * profile-count.h (profile_probability::split): New method.
6130 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6131 Use profile_probability::split.
6132 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6133 when splitting a single conditional jump into 2.
6135 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6137 PR tree-optimization/69452
6138 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6141 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6143 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6144 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6145 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6147 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6149 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6150 * config/rl78/rl78.md (movdi): New define_expand.
6151 * config/rl78/rl78.c (rl78_split_movdi): New function.
6153 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6156 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6158 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6159 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6160 128-bit to produce an UNSPEC move to get the double word with the
6161 signbit and then a shift directly to do signbit.
6162 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6163 implementation with a new version that just does either a direct
6164 move or a regular move. Move memory interface to separate insns.
6165 Move insns so they are next to the expander.
6166 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6167 with signbit move. Split big and little endian case.
6168 (signbit<mode>2_dm_mem_le): Likewise.
6169 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6170 (signbit<mode>2_dm2): Likewise.
6172 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6174 * config/rl78/rl78.md (anddi3): New define_expand.
6176 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6178 * config/rl78/rl78.md (umindi3): New define_expand.
6180 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6182 * config/rl78/rl78.md (smindi3): New define_expand.
6184 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6186 * config/rl78/rl78.md (smaxdi3): New define_expand.
6188 2018-01-22 Carl Love <cel@us.ibm.com>
6190 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6191 LVX_V1TI): Add macro expansion.
6192 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6193 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6194 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6195 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6196 Change check to determine if the instruction is a byte reversing
6197 entry. Fix typo in comment.
6198 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6199 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6200 Add def_builtin calls for new builtins.
6201 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6202 Add define_insn expansion.
6204 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6206 * config/rl78/rl78.md (umaxdi3): New define_expand.
6208 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6210 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6211 for non-QImode registers.
6213 2018-01-22 Richard Biener <rguenther@suse.de>
6215 PR tree-optimization/83963
6216 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6217 including the loop exit block.
6218 (scop_detection::merge_sese): Likewise.
6219 (scop_detection::add_scop): Do it here instead.
6221 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6223 * doc/sourcebuild.texi (arm_softfloat): Document.
6225 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6228 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6229 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6230 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6232 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6233 David Edelsohn <dje.gcc@gmail.com>
6236 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6237 Change "crset eq" to "crset 2".
6238 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6239 (*call_indirect_aix<mode>_nospec): Likewise.
6240 (*call_value_indirect_aix<mode>_nospec): Likewise.
6241 (*call_indirect_elfv2<mode>_nospec): Likewise.
6242 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6243 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6244 change assembly output from . to $.
6245 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6246 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6247 (*tablejump<mode>_internal1_nospec): Likewise.
6249 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6252 * config/sh/sh_optimize_sett_clrt.cc:
6253 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6255 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6257 PR tree-optimization/83940
6258 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6259 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6260 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6261 use it to pass back the definition type.
6262 (vect_check_store_rhs): Likewise rhs_dt_out.
6263 (vect_build_gather_load_calls): Add a mask_dt argument and use
6264 it instead of a call to vect_is_simple_use.
6265 (vectorizable_store): Update calls to vect_check_load_store_mask
6266 and vect_check_store_rhs. Use the dt returned by the latter instead
6267 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6268 instead of calls to vect_is_simple_use. Pass the scalar rather
6269 than the vector operand to vect_is_simple_use when handling
6270 second and subsequent copies of an rhs value.
6271 (vectorizable_load): Update calls to vect_check_load_store_mask
6272 and vect_build_gather_load_calls. Use the cached mask_dt and
6273 gs_info.offset_dt instead of calls to vect_is_simple_use.
6275 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6278 * tree-emutls.c: Include gimplify.h.
6279 (lower_emutls_2): New function.
6280 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6281 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6282 it before further processing.
6285 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6286 UINTVAL (trueop1) instead of INTVAL (op1).
6288 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6292 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6293 INCOMING_FRAME_SP_OFFSET if not defined.
6294 (scan_trace): Add ENTRY argument. If true and
6295 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6296 emit a note to adjust the CFA offset.
6297 (create_cfi_notes): Adjust scan_trace callers.
6298 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6299 INCOMING_FRAME_SP_OFFSET in the CIE.
6300 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6301 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6303 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6304 * doc/tm.texi: Regenerated.
6306 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6308 PR rtl-optimization/83147
6309 * lra-constraints.c (remove_inheritance_pseudos): Use
6310 lra_substitute_pseudo_within_insn.
6312 2018-01-19 Tom de Vries <tom@codesourcery.com>
6313 Cesar Philippidis <cesar@codesourcery.com>
6316 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6318 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6321 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6322 spaces for function labels.
6324 2018-01-19 Martin Liska <mliska@suse.cz>
6326 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6327 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6328 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6329 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6330 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6331 (PRED_CONST_RETURN): Change from 69 to 65.
6332 (PRED_NULL_RETURN): Change from 91 to 71.
6333 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6334 (PRED_LOOP_GUARD): Change from 66 to 73.
6336 2018-01-19 Martin Liska <mliska@suse.cz>
6338 * predict.c (predict_insn_def): Add new assert.
6339 (struct branch_predictor): Change type to signed integer.
6340 (test_prediction_value_range): Amend test to cover
6342 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6343 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6344 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6345 (PRED_LOOP_IV_COMPARE): Likewise.
6346 * predict.h (PROB_UNINITIALIZED): Define new constant.
6348 2018-01-19 Martin Liska <mliska@suse.cz>
6350 * predict.c (dump_prediction): Add new format for
6351 analyze_brprob.py script which is enabled with -details
6353 * profile-count.h (precise_p): New function.
6355 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6357 PR tree-optimization/83922
6358 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6359 there are no statements that need masking.
6360 (vect_active_double_reduction_p): New function.
6361 (vect_analyze_loop_operations): Use it when handling phis that
6362 are not in the loop header.
6364 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6366 PR tree-optimization/83914
6367 * tree-vect-loop.c (vectorizable_induction): Don't convert
6368 init_expr or apply the peeling adjustment for inductions
6369 that are nested within the vectorized loop.
6371 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6373 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6376 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6380 * function.h (gimplify_parameters): Add gimple_seq * argument.
6381 * function.c: Include gimple.h and options.h.
6382 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6383 for the added local temporaries if needed.
6384 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6385 if there are any parameter cleanups, wrap whole body into a
6386 try/finally with the cleanups.
6388 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6391 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6392 Use GET_MODE_CLASS for scalar floating point.
6394 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6398 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6399 Fix call of call_cgraph_insertion_hooks.
6401 2018-01-18 Martin Sebor <msebor@redhat.com>
6403 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6405 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6408 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6411 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6414 * common.opt: (-ffile-prefix-map): New option.
6415 * opts.c (common_handle_option): Defer it.
6416 * opts-global.c (handle_common_deferred_options): Handle it.
6417 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6418 * file-prefix-map.h: New file.
6419 (remap_debug_filename, add_debug_prefix_map): ...here.
6420 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6421 * final.c (debug_prefix_map, add_debug_prefix_map
6422 remap_debug_filename): Move to...
6423 * file-prefix-map.c: New file.
6424 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6425 generalize, get rid of alloca(), use strrchr() instead of strchr().
6426 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6427 Implement in terms of add_prefix_map().
6428 (remap_macro_filename, remap_debug_filename): Implement in term of
6430 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6431 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6432 * dbxout.c: Include file-prefix-map.h.
6433 * varasm.c: Likewise.
6434 * vmsdbgout.c: Likewise.
6435 * xcoffout.c: Likewise.
6436 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6437 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6438 * doc/invoke.texi (-ffile-prefix-map): Document.
6439 (-fdebug-prefix-map): Update description.
6441 2018-01-18 Martin Liska <mliska@suse.cz>
6443 * config/i386/i386.c (indirect_thunk_name): Document that also
6445 (output_indirect_thunk): Document why both instructions
6446 (pause and lfence) are generated.
6448 2018-01-18 Richard Biener <rguenther@suse.de>
6450 PR tree-optimization/83887
6451 * graphite-scop-detection.c
6452 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6453 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6454 (scop_detection::merge_sese): Re-implement with a flood-fill
6455 algorithm that properly finds a SESE region if it exists.
6457 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6460 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6461 pointer_diff optimizations use view_convert instead of convert.
6463 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6465 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6466 Generate different code for -mno-speculate-indirect-jumps.
6467 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6468 (*call_indirect_aix<mode>): Disable for
6469 -mno-speculate-indirect-jumps.
6470 (*call_indirect_aix<mode>_nospec): New define_insn.
6471 (*call_value_indirect_aix<mode>): Disable for
6472 -mno-speculate-indirect-jumps.
6473 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6474 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6475 -mno-speculate-indirect-jumps.
6476 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6478 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6480 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6481 long double type, set the flags for noting the default long double
6482 type, even if we don't pass or return a long double type.
6484 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6487 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6490 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6492 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6493 support for merge[hl].
6494 (fold_mergehl_helper): New helper function.
6495 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6496 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6497 (altivec_vmrglw_direct): Add xxmrglw insn.
6499 2018-01-17 Andrew Waterman <andrew@sifive.com>
6501 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6502 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6504 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6507 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6508 call the lto_location_cache before reading the
6509 DECL_SOURCE_LOCATION of the types.
6511 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6512 Richard Sandiford <richard.sandiford@linaro.org>
6514 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6515 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6516 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6517 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6518 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6520 * config/aarch64/constraints.md (aarch64_movti_operand):
6522 * config/aarch64/predicates.md (Uti): Add new constraint.
6524 2018-01-17 Carl Love <cel@us.ibm.com>
6526 * config/rs6000/vsx.md (define_expand xl_len_r,
6527 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6528 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6530 (define_expand, define_insn): Move the shift left from the
6531 define_insn to the define_expand for lxvl and stxvl instructions.
6532 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6533 and XL_LEN_R definitions to PURE.
6535 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6537 * config/i386/i386.c (indirect_thunk_name): Declare regno
6538 as unsigned int. Compare regno with INVALID_REGNUM.
6539 (output_indirect_thunk): Ditto.
6540 (output_indirect_thunk_function): Ditto.
6541 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6542 in the call to output_indirect_thunk_function.
6544 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6547 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6548 rather than the size of inner_type to determine the stack slot size
6549 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6551 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6554 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6557 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6559 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6560 endian Linux systems to optionally enable multilibs for selecting
6561 the long double type if the user configured an explicit type.
6562 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6563 have no long double multilibs if not defined.
6564 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6565 warn if the user used -mabi={ieee,ibm}longdouble and we built
6566 multilibs for long double.
6567 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6568 appropriate multilib option.
6569 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6571 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6572 for building long double multilibs.
6573 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6575 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6577 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6580 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6582 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6585 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6588 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6591 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6593 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6594 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6596 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6598 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6599 different rtl trees depending on TARGET_64BIT.
6600 (rs6000_gen_lvx): Likewise.
6602 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6604 * config/visium/visium.md (nop): Tweak comment.
6605 (hazard_nop): Likewise.
6607 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6609 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6610 -mspeculate-indirect-jumps.
6611 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6612 for -mno-speculate-indirect-jumps.
6613 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6614 (*call_value_indirect_elfv2<mode>): Disable for
6615 -mno-speculate-indirect-jumps.
6616 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6617 (indirect_jump): Emit different RTL for
6618 -mno-speculate-indirect-jumps.
6619 (*indirect_jump<mode>): Disable for
6620 -mno-speculate-indirect-jumps.
6621 (*indirect_jump<mode>_nospec): New define_insn.
6622 (tablejump): Emit different RTL for
6623 -mno-speculate-indirect-jumps.
6624 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6625 (tablejumpsi_nospec): New define_expand.
6626 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6627 (tablejumpdi_nospec): New define_expand.
6628 (*tablejump<mode>_internal1): Disable for
6629 -mno-speculate-indirect-jumps.
6630 (*tablejump<mode>_internal1_nospec): New define_insn.
6631 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6634 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6636 * caller-save.c (insert_save): Drop unnecessary parameter. All
6639 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6640 Richard Biener <rguenth@suse.de>
6643 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6644 return early, inline manually is_gimple_sizepos. Make sure if we
6645 call gimplify_expr we don't end up with a gimple constant.
6646 * tree.c (variably_modified_type_p): Don't return true for
6647 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6648 * gimplify.h (is_gimple_sizepos): Remove.
6650 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6652 PR tree-optimization/83857
6653 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6654 vectorizable_live_operation for pure SLP statements.
6655 (vectorizable_live_operation): Handle PHIs.
6657 2018-01-16 Richard Biener <rguenther@suse.de>
6659 PR tree-optimization/83867
6660 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6661 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6663 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6666 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6667 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6668 If off is not INTEGER_CST, issue a may not be aligned warning
6669 rather than isn't aligned. Use isn%'t rather than isn't.
6670 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6672 <case MULT_EXPR>: Improve the case when bottom and one of the
6673 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6674 operand, in that case check if the other operand is multiple of
6675 bottom divided by the INTEGER_CST operand.
6677 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6680 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6681 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6682 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6683 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6684 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6685 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6686 (pa_function_arg_size): New function.
6688 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6690 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6691 in a separate statement.
6693 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6695 PR tree-optimization/83847
6696 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6697 group gathers and scatters.
6699 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6701 PR rtl-optimization/86620
6702 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6704 PR rtl-optimization/83213
6705 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6706 to last if both are JUMP_INSNs.
6708 PR tree-optimization/83843
6709 * gimple-ssa-store-merging.c
6710 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6711 store_immediate_info for bswap/nop orig_stores.
6713 2018-01-15 Andrew Waterman <andrew@sifive.com>
6715 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6717 <UDIV>: Increase cost if !TARGET_DIV.
6719 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6721 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6722 (define_attr "cr_logical_3op"): New.
6723 (cceq_ior_compare): Adjust.
6724 (cceq_ior_compare_complement): Adjust.
6725 (*cceq_rev_compare): Adjust.
6726 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6727 (is_cracked_insn): Adjust.
6728 (insn_must_be_first_in_group): Adjust.
6729 * config/rs6000/40x.md: Adjust.
6730 * config/rs6000/440.md: Adjust.
6731 * config/rs6000/476.md: Adjust.
6732 * config/rs6000/601.md: Adjust.
6733 * config/rs6000/603.md: Adjust.
6734 * config/rs6000/6xx.md: Adjust.
6735 * config/rs6000/7450.md: Adjust.
6736 * config/rs6000/7xx.md: Adjust.
6737 * config/rs6000/8540.md: Adjust.
6738 * config/rs6000/cell.md: Adjust.
6739 * config/rs6000/e300c2c3.md: Adjust.
6740 * config/rs6000/e500mc.md: Adjust.
6741 * config/rs6000/e500mc64.md: Adjust.
6742 * config/rs6000/e5500.md: Adjust.
6743 * config/rs6000/e6500.md: Adjust.
6744 * config/rs6000/mpc.md: Adjust.
6745 * config/rs6000/power4.md: Adjust.
6746 * config/rs6000/power5.md: Adjust.
6747 * config/rs6000/power6.md: Adjust.
6748 * config/rs6000/power7.md: Adjust.
6749 * config/rs6000/power8.md: Adjust.
6750 * config/rs6000/power9.md: Adjust.
6751 * config/rs6000/rs64.md: Adjust.
6752 * config/rs6000/titan.md: Adjust.
6754 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6756 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6757 ix86_indirect_branch_register logic.
6759 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6761 * config/i386/constraints.md (Bs): Update
6762 ix86_indirect_branch_register check. Don't check
6763 ix86_indirect_branch_register with GOT_memory_operand.
6765 * config/i386/predicates.md (GOT_memory_operand): Don't check
6766 ix86_indirect_branch_register here.
6767 (GOT32_symbol_operand): Likewise.
6769 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6771 * config/i386/predicates.md (constant_call_address_operand):
6772 Rewrite ix86_indirect_branch_register logic.
6773 (sibcall_insn_operand): Likewise.
6775 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6777 * config/i386/constraints.md (Bs): Replace
6778 ix86_indirect_branch_thunk_register with
6779 ix86_indirect_branch_register.
6781 * config/i386/i386.md (indirect_jump): Likewise.
6782 (tablejump): Likewise.
6783 (*sibcall_memory): Likewise.
6784 (*sibcall_value_memory): Likewise.
6785 Peepholes of indirect call and jump via memory: Likewise.
6786 * config/i386/i386.opt: Likewise.
6787 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6788 (GOT_memory_operand): Likewise.
6789 (call_insn_operand): Likewise.
6790 (sibcall_insn_operand): Likewise.
6791 (GOT32_symbol_operand): Likewise.
6793 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6796 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6797 type rather than type addr's type points to.
6798 (expand_omp_atomic_mutex): Likewise.
6799 (expand_omp_atomic): Likewise.
6801 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6804 * config/i386/i386.c (output_indirect_thunk_function): Use
6805 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6806 for __x86_return_thunk.
6808 2018-01-15 Richard Biener <rguenther@suse.de>
6811 * expmed.c (extract_bit_field_1): Fix typo.
6813 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6816 * config/arm/iterators.md (VF): New mode iterator.
6817 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6818 Remove integer-related logic from pattern.
6819 (neon_vabd<mode>_3): Likewise.
6821 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6824 * common.opt (fstrict-overflow): No longer an alias.
6825 (fwrapv-pointer): New option.
6826 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6827 also for pointer types based on flag_wrapv_pointer.
6828 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6829 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6830 opts->x_flag_wrapv got set.
6831 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6832 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6833 POINTER_TYPE_OVERFLOW_UNDEFINED.
6834 * match.pd: Likewise in address comparison pattern.
6835 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6837 2018-01-15 Richard Biener <rguenther@suse.de>
6840 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6841 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6842 Reset type names to their identifier if their TYPE_DECL doesn't
6843 have linkage (and thus is used for ODR and devirt).
6844 (save_debug_info_for_decl): Remove.
6845 (save_debug_info_for_type): Likewise.
6846 (add_tree_to_fld_list): Adjust.
6847 * tree-pretty-print.c (dump_generic_node): Make dumping of
6848 type names more robust.
6850 2018-01-15 Richard Biener <rguenther@suse.de>
6852 * BASE-VER: Bump to 8.0.1.
6854 2018-01-14 Martin Sebor <msebor@redhat.com>
6857 * builtins.c (check_access): Avoid warning when the no-warning bit
6860 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6862 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6863 * ira-color (allocno_hard_regs_compare): Likewise.
6865 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6868 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6869 Use .pushsection/.popsection.
6871 2018-01-14 Martin Sebor <msebor@redhat.com>
6874 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6876 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6878 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6879 entry from extra_headers.
6880 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6881 extra_headers, make the list bitwise identical to the i?86-*-* one.
6883 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6885 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6886 -mcmodel=large with -mindirect-branch=thunk,
6887 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6888 -mfunction-return=thunk-extern.
6889 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6890 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6891 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6893 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6895 * config/i386/i386.c (print_reg): Print the name of the full
6896 integer register without '%'.
6897 (ix86_print_operand): Handle 'V'.
6898 * doc/extend.texi: Document 'V' modifier.
6900 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6902 * config/i386/constraints.md (Bs): Disallow memory operand for
6903 -mindirect-branch-register.
6905 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6906 (GOT_memory_operand): Likewise.
6907 (call_insn_operand): Likewise.
6908 (sibcall_insn_operand): Likewise.
6909 (GOT32_symbol_operand): Likewise.
6910 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6911 for -mindirect-branch-register.
6912 (tablejump): Likewise.
6913 (*sibcall_memory): Likewise.
6914 (*sibcall_value_memory): Likewise.
6915 Disallow peepholes of indirect call and jump via memory for
6916 -mindirect-branch-register.
6917 (*call_pop): Replace m with Bw.
6918 (*call_value_pop): Likewise.
6919 (*sibcall_pop_memory): Replace m with Bs.
6920 * config/i386/i386.opt (mindirect-branch-register): New option.
6921 * doc/invoke.texi: Document -mindirect-branch-register option.
6923 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6925 * config/i386/i386-protos.h (ix86_output_function_return): New.
6926 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6927 set function_return_type.
6928 (indirect_thunk_name): Add ret_p to indicate thunk for function
6930 (output_indirect_thunk_function): Pass false to
6931 indirect_thunk_name.
6932 (ix86_output_indirect_branch_via_reg): Likewise.
6933 (ix86_output_indirect_branch_via_push): Likewise.
6934 (output_indirect_thunk_function): Create alias for function
6935 return thunk if regno < 0.
6936 (ix86_output_function_return): New function.
6937 (ix86_handle_fndecl_attribute): Handle function_return.
6938 (ix86_attribute_table): Add function_return.
6939 * config/i386/i386.h (machine_function): Add
6940 function_return_type.
6941 * config/i386/i386.md (simple_return_internal): Use
6942 ix86_output_function_return.
6943 (simple_return_internal_long): Likewise.
6944 * config/i386/i386.opt (mfunction-return=): New option.
6945 (indirect_branch): Mention -mfunction-return=.
6946 * doc/extend.texi: Document function_return function attribute.
6947 * doc/invoke.texi: Document -mfunction-return= option.
6949 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6951 * config/i386/i386-opts.h (indirect_branch): New.
6952 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6953 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6954 with local indirect jump when converting indirect call and jump.
6955 (ix86_set_indirect_branch_type): New.
6956 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6957 (indirectlabelno): New.
6958 (indirect_thunk_needed): Likewise.
6959 (indirect_thunk_bnd_needed): Likewise.
6960 (indirect_thunks_used): Likewise.
6961 (indirect_thunks_bnd_used): Likewise.
6962 (INDIRECT_LABEL): Likewise.
6963 (indirect_thunk_name): Likewise.
6964 (output_indirect_thunk): Likewise.
6965 (output_indirect_thunk_function): Likewise.
6966 (ix86_output_indirect_branch_via_reg): Likewise.
6967 (ix86_output_indirect_branch_via_push): Likewise.
6968 (ix86_output_indirect_branch): Likewise.
6969 (ix86_output_indirect_jmp): Likewise.
6970 (ix86_code_end): Call output_indirect_thunk_function if needed.
6971 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6973 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6974 (ix86_attribute_table): Add indirect_branch.
6975 * config/i386/i386.h (machine_function): Add indirect_branch_type
6976 and has_local_indirect_jump.
6977 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6979 (tablejump): Likewise.
6980 (*indirect_jump): Use ix86_output_indirect_jmp.
6981 (*tablejump_1): Likewise.
6982 (simple_return_indirect_internal): Likewise.
6983 * config/i386/i386.opt (mindirect-branch=): New option.
6984 (indirect_branch): New.
6987 (thunk-inline): Likewise.
6988 (thunk-extern): Likewise.
6989 * doc/extend.texi: Document indirect_branch function attribute.
6990 * doc/invoke.texi: Document -mindirect-branch= option.
6992 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6995 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6997 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6999 * ipa-inline.c (want_inline_small_function_p): Return false if
7000 inlining has already failed with CIF_FINAL_ERROR.
7001 (update_caller_keys): Call want_inline_small_function_p before
7003 (update_callee_keys): Likewise.
7005 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7007 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7009 (rs6000_quadword_masked_address_p): Likewise.
7010 (quad_aligned_load_p): Likewise.
7011 (quad_aligned_store_p): Likewise.
7012 (const_load_sequence_p): Add comment to describe the outer-most loop.
7013 (mimic_memory_attributes_and_flags): New function.
7014 (rs6000_gen_stvx): Likewise.
7015 (replace_swapped_aligned_store): Likewise.
7016 (rs6000_gen_lvx): Likewise.
7017 (replace_swapped_aligned_load): Likewise.
7018 (replace_swapped_load_constant): Capitalize argument name in
7019 comment describing this function.
7020 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7021 and stores that access quad-word aligned addresses and replace
7022 with stvx or lvx instructions when appropriate.
7023 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7024 New function prototype.
7025 (rs6000_quadword_masked_address_p): Likewise.
7026 (rs6000_gen_lvx): Likewise.
7027 (rs6000_gen_stvx): Likewise.
7028 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7029 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7030 when memory address is aligned.
7031 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7032 this split to select lvx instruction when memory address is aligned.
7033 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7034 instruction when memory address is aligned.
7035 (*vsx_le_perm_load_v16qi): Likewise.
7036 (four unnamed splitters): Modify to select the stvx instruction
7037 when memory is aligned.
7039 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7041 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7042 which appears in the queue multiple times.
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 * tree-vectorizer.h (vec_lower_bound): New structure.
7049 (_loop_vec_info): Add check_nonzero and lower_bounds.
7050 (LOOP_VINFO_CHECK_NONZERO): New macro.
7051 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7052 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7053 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7054 fields. Make seg_len the distance travelled, not including the
7056 (dr_direction_indicator): Declare.
7057 (dr_zero_step_indicator): Likewise.
7058 (dr_known_forward_stride_p): Likewise.
7059 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7061 (runtime_alias_check_p): Allow runtime alias checks with
7063 (operator ==): Compare access_size and align.
7064 (prune_runtime_alias_test_list): Rework for new distinction between
7065 the access_size and seg_len.
7066 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7068 (get_segment_min_max): New function.
7069 (create_intersect_range_checks): Use it.
7070 (dr_step_indicator): New function.
7071 (dr_direction_indicator): Likewise.
7072 (dr_zero_step_indicator): Likewise.
7073 (dr_known_forward_stride_p): Likewise.
7074 * tree-loop-distribution.c (data_ref_segment_size): Return
7075 DR_STEP * (niters - 1).
7076 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7078 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7079 (vect_preserves_scalar_order_p): New function, split out from...
7080 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7081 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7082 (vect_vfa_access_size): New function.
7083 (vect_vfa_align): Likewise.
7084 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7085 (dump_lower_bound): New function.
7086 (vect_check_lower_bound): Likewise.
7087 (vect_small_gap_p): Likewise.
7088 (vectorizable_with_step_bound_p): Likewise.
7089 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7090 depencies if the vectorization factor is 1. Convert the checks
7091 for nonzero steps into checks on the bounds of DR_STEP. Try using
7092 a bunds check for variable steps if the minimum required step is
7093 relatively small. Update calls to the dr_with_seg_len
7094 constructor and to vect_compile_time_alias.
7095 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7097 (vect_loop_versioning): Call it.
7098 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7100 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7103 Alan Hayward <alan.hayward@arm.com>
7104 David Sherwood <david.sherwood@arm.com>
7106 * doc/sourcebuild.texi (vect_scatter_store): Document.
7107 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7109 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7111 * genopinit.c (main): Add supports_vec_scatter_store and
7112 supports_vec_scatter_store_cached to target_optabs.
7113 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7114 IFN_MASK_SCATTER_STORE.
7115 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7117 * internal-fn.h (internal_store_fn_p): Declare.
7118 (internal_fn_stored_value_index): Likewise.
7119 * internal-fn.c (scatter_store_direct): New macro.
7120 (expand_scatter_store_optab_fn): New function.
7121 (direct_scatter_store_optab_supported_p): New macro.
7122 (internal_store_fn_p): New function.
7123 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7124 IFN_MASK_SCATTER_STORE.
7125 (internal_fn_mask_index): Likewise.
7126 (internal_fn_stored_value_index): New function.
7127 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7129 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7130 * optabs-query.c (supports_vec_scatter_store_p): New function.
7131 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7132 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7133 true for scatter stores.
7134 (vect_gather_scatter_fn_p): Handle scatter stores too.
7135 (vect_check_gather_scatter): Consider using scatter stores if
7136 supports_vec_scatter_store_p.
7137 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7139 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7140 internal_fn_stored_value_index.
7141 (check_load_store_masking): Handle scatter stores too.
7142 (vect_get_store_rhs): Make public.
7143 (vectorizable_call): Use internal_store_fn_p.
7144 (vectorizable_store): Handle scatter store internal functions.
7145 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7146 when deciding whether the end of the group has been reached.
7147 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7148 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7149 (mask_scatter_store<mode>): New insns.
7151 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7152 Alan Hayward <alan.hayward@arm.com>
7153 David Sherwood <david.sherwood@arm.com>
7155 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7156 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7157 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7159 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7160 Use vect_truncate_gather_scatter_offset if we can't treat the
7161 operation as a normal gather load or scatter store.
7162 (get_group_load_store_type): Take the gather_scatter_info
7163 as argument. Try using a gather load or scatter store for
7164 single-element groups.
7165 (get_load_store_type): Update calls to get_group_load_store_type
7166 and vect_use_strided_gather_scatters_p.
7168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7169 Alan Hayward <alan.hayward@arm.com>
7170 David Sherwood <david.sherwood@arm.com>
7172 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7173 optional tree argument.
7174 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7176 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7177 but continue to use the current value as a fallback.
7178 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7179 to compare the updates.
7180 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7181 (get_load_store_type): Use it when handling a strided access.
7182 (vect_get_strided_load_store_ops): New function.
7183 (vect_get_data_ptr_increment): Likewise.
7184 (vectorizable_load): Handle strided gather loads. Always pass
7185 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7187 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7188 Alan Hayward <alan.hayward@arm.com>
7189 David Sherwood <david.sherwood@arm.com>
7191 * doc/md.texi (gather_load@var{m}): Document.
7192 (mask_gather_load@var{m}): Likewise.
7193 * genopinit.c (main): Add supports_vec_gather_load and
7194 supports_vec_gather_load_cached to target_optabs.
7195 * optabs-tree.c (init_tree_optimization_optabs): Use
7196 ggc_cleared_alloc to allocate target_optabs.
7197 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7198 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7200 * internal-fn.h (internal_load_fn_p): Declare.
7201 (internal_gather_scatter_fn_p): Likewise.
7202 (internal_fn_mask_index): Likewise.
7203 (internal_gather_scatter_fn_supported_p): Likewise.
7204 * internal-fn.c (gather_load_direct): New macro.
7205 (expand_gather_load_optab_fn): New function.
7206 (direct_gather_load_optab_supported_p): New macro.
7207 (direct_internal_fn_optab): New function.
7208 (internal_load_fn_p): Likewise.
7209 (internal_gather_scatter_fn_p): Likewise.
7210 (internal_fn_mask_index): Likewise.
7211 (internal_gather_scatter_fn_supported_p): Likewise.
7212 * optabs-query.c (supports_at_least_one_mode_p): New function.
7213 (supports_vec_gather_load_p): Likewise.
7214 * optabs-query.h (supports_vec_gather_load_p): Declare.
7215 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7216 and memory_type field.
7217 (NUM_PATTERNS): Bump to 15.
7218 * tree-vect-data-refs.c: Include internal-fn.h.
7219 (vect_gather_scatter_fn_p): New function.
7220 (vect_describe_gather_scatter_call): Likewise.
7221 (vect_check_gather_scatter): Try using internal functions for
7222 gather loads. Recognize existing calls to a gather load function.
7223 (vect_analyze_data_refs): Consider using gather loads if
7224 supports_vec_gather_load_p.
7225 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7226 (vect_get_gather_scatter_offset_type): Likewise.
7227 (vect_convert_mask_for_vectype): Likewise.
7228 (vect_add_conversion_to_patterm): Likewise.
7229 (vect_try_gather_scatter_pattern): Likewise.
7230 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7231 (vect_vect_recog_func_ptrs): Add it.
7232 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7233 internal_fn_mask_index and internal_gather_scatter_fn_p.
7234 (check_load_store_masking): Take the gather_scatter_info as an
7235 argument and handle gather loads.
7236 (vect_get_gather_scatter_ops): New function.
7237 (vectorizable_call): Check internal_load_fn_p.
7238 (vectorizable_load): Likewise. Handle gather load internal
7240 (vectorizable_store): Update call to check_load_store_masking.
7241 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7242 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7243 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7244 (aarch64_gather_scale_operand_d): New predicates.
7245 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7246 (mask_gather_load<mode>): New insns.
7248 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7249 Alan Hayward <alan.hayward@arm.com>
7250 David Sherwood <david.sherwood@arm.com>
7252 * optabs.def (fold_left_plus_optab): New optab.
7253 * doc/md.texi (fold_left_plus_@var{m}): Document.
7254 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7255 * internal-fn.c (fold_left_direct): Define.
7256 (expand_fold_left_optab_fn): Likewise.
7257 (direct_fold_left_optab_supported_p): Likewise.
7258 * fold-const-call.c (fold_const_fold_left): New function.
7259 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7260 * tree-parloops.c (valid_reduction_p): New function.
7261 (gather_scalar_reductions): Use it.
7262 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7263 (vect_finish_replace_stmt): Declare.
7264 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7265 (needs_fold_left_reduction_p): New function, split out from...
7266 (vect_is_simple_reduction): ...here. Accept reductions that
7267 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7268 (vect_force_simple_reduction): Also store the reduction type in
7269 the assignment's STMT_VINFO_REDUC_TYPE.
7270 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7271 (merge_with_identity): New function.
7272 (vect_expand_fold_left): Likewise.
7273 (vectorize_fold_left_reduction): Likewise.
7274 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7275 scalar phi in place for it. Check for target support and reject
7276 cases that would reassociate the operation. Defer the transform
7277 phase to vectorize_fold_left_reduction.
7278 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7279 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7280 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7282 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7284 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7285 call to ifc_temp_var.
7287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7288 Alan Hayward <alan.hayward@arm.com>
7289 David Sherwood <david.sherwood@arm.com>
7291 * target.def (legitimize_address_displacement): Take the original
7292 offset as a poly_int.
7293 * targhooks.h (default_legitimize_address_displacement): Update
7295 * targhooks.c (default_legitimize_address_displacement): Likewise.
7296 * doc/tm.texi: Regenerate.
7297 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7298 as an argument, moving assert of ad->disp == ad->disp_term to...
7299 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7300 Try calling targetm.legitimize_address_displacement before expanding
7301 the address rather than afterwards, and adjust for the new interface.
7302 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7303 Match the new hook interface. Handle SVE addresses.
7304 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7307 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7309 * Makefile.in (OBJS): Add early-remat.o.
7310 * target.def (select_early_remat_modes): New hook.
7311 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7312 * doc/tm.texi: Regenerate.
7313 * targhooks.h (default_select_early_remat_modes): Declare.
7314 * targhooks.c (default_select_early_remat_modes): New function.
7315 * timevar.def (TV_EARLY_REMAT): New timevar.
7316 * passes.def (pass_early_remat): New pass.
7317 * tree-pass.h (make_pass_early_remat): Declare.
7318 * early-remat.c: New file.
7319 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7321 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7323 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7324 Alan Hayward <alan.hayward@arm.com>
7325 David Sherwood <david.sherwood@arm.com>
7327 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7328 vfm1 with a bound_epilog parameter.
7329 (vect_do_peeling): Update calls accordingly, and move the prologue
7330 call earlier in the function. Treat the base bound_epilog as 0 for
7331 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7332 this base when peeling for gaps.
7333 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7334 with fully-masked loops.
7335 (vect_estimate_min_profitable_iters): Handle the single peeled
7336 iteration in that case.
7338 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7339 Alan Hayward <alan.hayward@arm.com>
7340 David Sherwood <david.sherwood@arm.com>
7342 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7343 single-element interleaving even if the size is not a power of 2.
7344 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7345 accesses for single-element interleaving if the group size is
7348 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7349 Alan Hayward <alan.hayward@arm.com>
7350 David Sherwood <david.sherwood@arm.com>
7352 * doc/md.texi (fold_extract_last_@var{m}): Document.
7353 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7354 * optabs.def (fold_extract_last_optab): New optab.
7355 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7356 * internal-fn.c (fold_extract_direct): New macro.
7357 (expand_fold_extract_optab_fn): Likewise.
7358 (direct_fold_extract_optab_supported_p): Likewise.
7359 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7360 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7361 EXTRACT_LAST_REDUCTION.
7362 (get_initial_def_for_reduction): Do not create an initial vector
7363 for EXTRACT_LAST_REDUCTION reductions.
7364 (vectorizable_reduction): Leave the scalar phi in place for
7365 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7366 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7367 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7368 transform phase to vectorizable_condition.
7369 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7371 (vect_finish_stmt_generation): ...here.
7372 (vect_finish_replace_stmt): New function.
7373 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7374 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7376 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7378 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7379 Alan Hayward <alan.hayward@arm.com>
7380 David Sherwood <david.sherwood@arm.com>
7382 * doc/md.texi (extract_last_@var{m}): Document.
7383 * optabs.def (extract_last_optab): New optab.
7384 * internal-fn.def (EXTRACT_LAST): New internal function.
7385 * internal-fn.c (cond_unary_direct): New macro.
7386 (expand_cond_unary_optab_fn): Likewise.
7387 (direct_cond_unary_optab_supported_p): Likewise.
7388 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7389 loops using EXTRACT_LAST.
7390 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7391 (extract_last_<mode>): ...this optab.
7392 (vec_extract<mode><Vel>): Update accordingly.
7394 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7395 Alan Hayward <alan.hayward@arm.com>
7396 David Sherwood <david.sherwood@arm.com>
7398 * target.def (empty_mask_is_expensive): New hook.
7399 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7400 * doc/tm.texi: Regenerate.
7401 * targhooks.h (default_empty_mask_is_expensive): Declare.
7402 * targhooks.c (default_empty_mask_is_expensive): New function.
7403 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7404 if the target says that empty masks are expensive.
7405 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7407 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7409 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7410 Alan Hayward <alan.hayward@arm.com>
7411 David Sherwood <david.sherwood@arm.com>
7413 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7414 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7415 (vect_use_loop_mask_for_alignment_p): New function.
7416 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7417 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7418 niters_skip argument. Make sure that the first niters_skip elements
7419 of the first iteration are inactive.
7420 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7421 Update call to vect_set_loop_masks_directly.
7422 (get_misalign_in_elems): New function, split out from...
7423 (vect_gen_prolog_loop_niters): ...here.
7424 (vect_update_init_of_dr): Take a code argument that specifies whether
7425 the adjustment should be added or subtracted.
7426 (vect_update_init_of_drs): Likewise.
7427 (vect_prepare_for_masked_peels): New function.
7428 (vect_do_peeling): Skip prologue peeling if we're using a mask
7429 instead. Update call to vect_update_inits_of_drs.
7430 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7432 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7433 alignment. Do not include the number of peeled iterations in
7434 the minimum threshold in that case.
7435 (vectorizable_induction): Adjust the start value down by
7436 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7437 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7438 Take the number of skipped iterations into account when calculating
7440 * tree-vect-stmts.c (vect_gen_while_not): New function.
7442 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7443 Alan Hayward <alan.hayward@arm.com>
7444 David Sherwood <david.sherwood@arm.com>
7446 * doc/sourcebuild.texi (vect_fully_masked): Document.
7447 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7449 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7451 (vect_analyze_loop_2): ...here. Don't check the vectorization
7452 factor against the number of loop iterations if the loop is
7455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7456 Alan Hayward <alan.hayward@arm.com>
7457 David Sherwood <david.sherwood@arm.com>
7459 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7460 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7461 (dump_groups): Update accordingly.
7462 (iv_use::mem_type): New member variable.
7463 (address_p): New function.
7464 (record_use): Add a mem_type argument and initialize the new
7466 (record_group_use): Add a mem_type argument. Use address_p.
7467 Remove obsolete null checks of base_object. Update call to record_use.
7468 (find_interesting_uses_op): Update call to record_group_use.
7469 (find_interesting_uses_cond): Likewise.
7470 (find_interesting_uses_address): Likewise.
7471 (get_mem_type_for_internal_fn): New function.
7472 (find_address_like_use): Likewise.
7473 (find_interesting_uses_stmt): Try find_address_like_use before
7474 calling find_interesting_uses_op.
7475 (addr_offset_valid_p): Use the iv mem_type field as the type
7476 of the addressed memory.
7477 (add_autoinc_candidates): Likewise.
7478 (get_address_cost): Likewise.
7479 (split_small_address_groups_p): Use address_p.
7480 (split_address_groups): Likewise.
7481 (add_iv_candidate_for_use): Likewise.
7482 (autoinc_possible_for_pair): Likewise.
7483 (rewrite_groups): Likewise.
7484 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7485 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7486 (get_alias_ptr_type_for_ptr_address): New function.
7487 (rewrite_use_address): Rewrite address uses in calls that were
7488 identified by find_address_like_use.
7490 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7491 Alan Hayward <alan.hayward@arm.com>
7492 David Sherwood <david.sherwood@arm.com>
7494 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7496 * gimple-expr.h (is_gimple_addressable: Likewise.
7497 * gimple-expr.c (is_gimple_address): Likewise.
7498 * internal-fn.c (expand_call_mem_ref): New function.
7499 (expand_mask_load_optab_fn): Use it.
7500 (expand_mask_store_optab_fn): Likewise.
7502 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7503 Alan Hayward <alan.hayward@arm.com>
7504 David Sherwood <david.sherwood@arm.com>
7506 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7507 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7508 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7509 (cond_umax@var{mode}): Document.
7510 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7511 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7512 (cond_umin_optab, cond_umax_optab): New optabs.
7513 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7514 (COND_IOR, COND_XOR): New internal functions.
7515 * internal-fn.h (get_conditional_internal_fn): Declare.
7516 * internal-fn.c (cond_binary_direct): New macro.
7517 (expand_cond_binary_optab_fn): Likewise.
7518 (direct_cond_binary_optab_supported_p): Likewise.
7519 (get_conditional_internal_fn): New function.
7520 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7521 Cope with reduction statements that are vectorized as calls rather
7523 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7524 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7525 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7526 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7527 (UNSPEC_COND_EOR): New unspecs.
7528 (optab): Add mappings for them.
7529 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7530 (sve_int_op, sve_fp_op): New int attributes.
7532 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7533 Alan Hayward <alan.hayward@arm.com>
7534 David Sherwood <david.sherwood@arm.com>
7536 * optabs.def (while_ult_optab): New optab.
7537 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7538 * internal-fn.def (WHILE_ULT): New internal function.
7539 * internal-fn.h (direct_internal_fn_supported_p): New override
7540 that takes two types as argument.
7541 * internal-fn.c (while_direct): New macro.
7542 (expand_while_optab_fn): New function.
7543 (convert_optab_supported_p): Likewise.
7544 (direct_while_optab_supported_p): New macro.
7545 * wide-int.h (wi::udiv_ceil): New function.
7546 * tree-vectorizer.h (rgroup_masks): New structure.
7547 (vec_loop_masks): New typedef.
7548 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7550 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7551 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7552 (vect_max_vf): New function.
7553 (slpeel_make_loop_iterate_ntimes): Delete.
7554 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7555 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7556 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7557 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7558 internal-fn.h, stor-layout.h and optabs-query.h.
7559 (vect_set_loop_mask): New function.
7560 (add_preheader_seq): Likewise.
7561 (add_header_seq): Likewise.
7562 (interleave_supported_p): Likewise.
7563 (vect_maybe_permute_loop_masks): Likewise.
7564 (vect_set_loop_masks_directly): Likewise.
7565 (vect_set_loop_condition_masked): Likewise.
7566 (vect_set_loop_condition_unmasked): New function, split out from
7567 slpeel_make_loop_iterate_ntimes.
7568 (slpeel_make_loop_iterate_ntimes): Rename to..
7569 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7570 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7571 (vect_do_peeling): Update call accordingly.
7572 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7574 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7575 mask_compare_type, can_fully_mask_p and fully_masked_p.
7576 (release_vec_loop_masks): New function.
7577 (_loop_vec_info): Use it to free the loop masks.
7578 (can_produce_all_loop_masks_p): New function.
7579 (vect_get_max_nscalars_per_iter): Likewise.
7580 (vect_verify_full_masking): Likewise.
7581 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7582 retries, and free the mask rgroups before retrying. Check loop-wide
7583 reasons for disallowing fully-masked loops. Make the final decision
7584 about whether use a fully-masked loop or not.
7585 (vect_estimate_min_profitable_iters): Do not assume that peeling
7586 for the number of iterations will be needed for fully-masked loops.
7587 (vectorizable_reduction): Disable fully-masked loops.
7588 (vectorizable_live_operation): Likewise.
7589 (vect_halve_mask_nunits): New function.
7590 (vect_double_mask_nunits): Likewise.
7591 (vect_record_loop_mask): Likewise.
7592 (vect_get_loop_mask): Likewise.
7593 (vect_transform_loop): Handle the case in which the final loop
7594 iteration might handle a partial vector. Call vect_set_loop_condition
7595 instead of slpeel_make_loop_iterate_ntimes.
7596 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7597 (check_load_store_masking): New function.
7598 (prepare_load_store_mask): Likewise.
7599 (vectorizable_store): Handle fully-masked loops.
7600 (vectorizable_load): Likewise.
7601 (supportable_widening_operation): Use vect_halve_mask_nunits for
7603 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7604 (vect_gen_while): New function.
7605 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7606 (aarch64_uqdec<mode>): New insn.
7608 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7609 Alan Hayward <alan.hayward@arm.com>
7610 David Sherwood <david.sherwood@arm.com>
7612 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7613 (reduc_xor_scal_optab): New optabs.
7614 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7615 (reduc_xor_scal_@var{m}): Document.
7616 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7617 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7619 * fold-const-call.c (fold_const_call): Handle them.
7620 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7621 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7622 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7623 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7624 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7625 (UNSPEC_XORV): New unspecs.
7626 (optab): Add entries for them.
7627 (BITWISEV): New int iterator.
7628 (bit_reduc_op): New int attributes.
7630 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7631 Alan Hayward <alan.hayward@arm.com>
7632 David Sherwood <david.sherwood@arm.com>
7634 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7635 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7636 * optabs.def (vec_shl_insert_optab): New optab.
7637 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7638 (duplicate_and_interleave): Likewise.
7639 * tree-vect-loop.c: Include internal-fn.h.
7640 (neutral_op_for_slp_reduction): New function, split out from
7641 get_initial_defs_for_reduction.
7642 (get_initial_def_for_reduction): Handle option 2 for variable-length
7643 vectors by loading the neutral value into a vector and then shifting
7644 the initial value into element 0.
7645 (get_initial_defs_for_reduction): Replace the code argument with
7646 the neutral value calculated by neutral_op_for_slp_reduction.
7647 Use gimple_build_vector for constant-length vectors.
7648 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7649 but the first group_size elements have a neutral value.
7650 Use duplicate_and_interleave otherwise.
7651 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7652 Update call to get_initial_defs_for_reduction. Handle SLP
7653 reductions for variable-length vectors by creating one vector
7654 result for each scalar result, with the elements associated
7655 with other scalar results stubbed out with the neutral value.
7656 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7657 Require IFN_VEC_SHL_INSERT for double reductions on
7658 variable-length vectors, or SLP reductions that have
7659 a neutral value. Require can_duplicate_and_interleave_p
7660 support for variable-length unchained SLP reductions if there
7661 is no neutral value, such as for MIN/MAX reductions. Also require
7662 the number of vector elements to be a multiple of the number of
7663 SLP statements when doing variable-length unchained SLP reductions.
7664 Update call to vect_create_epilog_for_reduction.
7665 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7666 and remove initial values.
7667 (duplicate_and_interleave): Make public.
7668 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7669 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7672 Alan Hayward <alan.hayward@arm.com>
7673 David Sherwood <david.sherwood@arm.com>
7675 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7676 (can_duplicate_and_interleave_p): New function.
7677 (vect_get_and_check_slp_defs): Take the vector of statements
7678 rather than just the current one. Remove excess parentheses.
7679 Restriction rejectinon of vect_constant_def and vect_external_def
7680 for variable-length vectors to boolean types, or types for which
7681 can_duplicate_and_interleave_p is false.
7682 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7683 (duplicate_and_interleave): New function.
7684 (vect_get_constant_vectors): Use gimple_build_vector for
7685 constant-length vectors and suitable variable-length constant
7686 vectors. Use duplicate_and_interleave for other variable-length
7687 vectors. Don't defer the update when inserting new statements.
7689 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7690 Alan Hayward <alan.hayward@arm.com>
7691 David Sherwood <david.sherwood@arm.com>
7693 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7694 min_profitable_iters doesn't go negative.
7696 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7697 Alan Hayward <alan.hayward@arm.com>
7698 David Sherwood <david.sherwood@arm.com>
7700 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7701 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7702 * optabs.def (vec_mask_load_lanes_optab): New optab.
7703 (vec_mask_store_lanes_optab): Likewise.
7704 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7705 (MASK_STORE_LANES): Likewise.
7706 * internal-fn.c (mask_load_lanes_direct): New macro.
7707 (mask_store_lanes_direct): Likewise.
7708 (expand_mask_load_optab_fn): Handle masked operations.
7709 (expand_mask_load_lanes_optab_fn): New macro.
7710 (expand_mask_store_optab_fn): Handle masked operations.
7711 (expand_mask_store_lanes_optab_fn): New macro.
7712 (direct_mask_load_lanes_optab_supported_p): Likewise.
7713 (direct_mask_store_lanes_optab_supported_p): Likewise.
7714 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7716 (vect_load_lanes_supported): Likewise.
7717 * tree-vect-data-refs.c (strip_conversion): New function.
7718 (can_group_stmts_p): Likewise.
7719 (vect_analyze_data_ref_accesses): Use it instead of checking
7720 for a pair of assignments.
7721 (vect_store_lanes_supported): Take a masked_p parameter.
7722 (vect_load_lanes_supported): Likewise.
7723 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7724 vect_store_lanes_supported and vect_load_lanes_supported.
7725 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7726 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7727 parameter. Don't allow gaps for masked accesses.
7728 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7729 and vect_load_lanes_supported.
7730 (get_load_store_type): Take a masked_p parameter and update
7731 call to get_group_load_store_type.
7732 (vectorizable_store): Update call to get_load_store_type.
7733 Handle IFN_MASK_STORE_LANES.
7734 (vectorizable_load): Update call to get_load_store_type.
7735 Handle IFN_MASK_LOAD_LANES.
7737 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7738 Alan Hayward <alan.hayward@arm.com>
7739 David Sherwood <david.sherwood@arm.com>
7741 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7743 * config/aarch64/aarch64-protos.h
7744 (aarch64_sve_struct_memory_operand_p): Declare.
7745 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7746 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7747 (VPRED, vpred): Handle SVE structure modes.
7748 * config/aarch64/constraints.md (Utx): New constraint.
7749 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7750 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7751 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7752 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7753 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7754 structure modes. Split into pieces after RA.
7755 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7756 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7758 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7759 SVE structure modes.
7760 (aarch64_classify_address): Likewise.
7761 (sizetochar): Move earlier in file.
7762 (aarch64_print_operand): Handle SVE register lists.
7763 (aarch64_array_mode): New function.
7764 (aarch64_sve_struct_memory_operand_p): Likewise.
7765 (TARGET_ARRAY_MODE): Redefine.
7767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7768 Alan Hayward <alan.hayward@arm.com>
7769 David Sherwood <david.sherwood@arm.com>
7771 * target.def (array_mode): New target hook.
7772 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7773 * doc/tm.texi: Regenerate.
7774 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7775 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7776 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7778 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7781 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7782 Alan Hayward <alan.hayward@arm.com>
7783 David Sherwood <david.sherwood@arm.com>
7785 * fold-const.c (fold_binary_loc): Check the argument types
7786 rather than the result type when testing for a vector operation.
7788 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7790 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7791 * doc/tm.texi: Regenerate.
7793 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7794 Alan Hayward <alan.hayward@arm.com>
7795 David Sherwood <david.sherwood@arm.com>
7797 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7798 (sve): Document new AArch64 extension.
7799 * doc/md.texi (w): Extend the description of the AArch64
7800 constraint to include SVE vectors.
7801 (Upl, Upa): Document new AArch64 predicate constraints.
7802 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7804 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7805 (msve-vector-bits=): New option.
7806 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7807 SVE when these are disabled.
7808 (sve): New extension.
7809 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7810 modes. Adjust their number of units based on aarch64_sve_vg.
7811 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7812 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7813 aarch64_addr_query_type.
7814 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7815 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7816 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7817 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7818 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7819 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7820 (aarch64_simd_imm_zero_p): Delete.
7821 (aarch64_check_zero_based_sve_index_immediate): Declare.
7822 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7823 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7824 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7825 (aarch64_sve_float_mul_immediate_p): Likewise.
7826 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7828 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7829 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7830 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7831 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7832 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7833 (aarch64_regmode_natural_size): Likewise.
7834 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7835 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7837 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7838 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7839 for VG and the SVE predicate registers.
7840 (V_ALIASES): Add a "z"-prefixed alias.
7841 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7842 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7843 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7844 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7845 (REG_CLASS_NAMES): Add entries for them.
7846 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7847 and the predicate registers.
7848 (aarch64_sve_vg): Declare.
7849 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7850 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7851 (REGMODE_NATURAL_SIZE): Define.
7852 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7854 * config/aarch64/aarch64.c: Include cfgrtl.h.
7855 (simd_immediate_info): Add a constructor for series vectors,
7856 and an associated step field.
7857 (aarch64_sve_vg): New variable.
7858 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7859 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7860 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7861 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7862 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7863 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7864 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7865 (aarch64_get_mask_mode): New functions.
7866 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7867 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7868 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7869 predicate modes and predicate registers. Explicitly restrict
7870 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7871 to store a vector mode if it is recognized by
7872 aarch64_classify_vector_mode.
7873 (aarch64_regmode_natural_size): New function.
7874 (aarch64_hard_regno_caller_save_mode): Return the original mode
7876 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7877 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7878 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7879 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7881 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7882 does not overlap dest if the function is frame-related. Handle
7884 (aarch64_split_add_offset): New function.
7885 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7886 them aarch64_add_offset.
7887 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7888 and update call to aarch64_sub_sp.
7889 (aarch64_add_cfa_expression): New function.
7890 (aarch64_expand_prologue): Pass extra temporary registers to the
7891 functions above. Handle the case in which we need to emit new
7892 DW_CFA_expressions for registers that were originally saved
7893 relative to the stack pointer, but now have to be expressed
7894 relative to the frame pointer.
7895 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7897 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7898 IP0 and IP1 values for SVE frames.
7899 (aarch64_expand_vec_series): New function.
7900 (aarch64_expand_sve_widened_duplicate): Likewise.
7901 (aarch64_expand_sve_const_vector): Likewise.
7902 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7903 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7904 into the register, rather than emitting a SET directly.
7905 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7906 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7907 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7908 (offset_9bit_signed_scaled_p): New functions.
7909 (aarch64_replicate_bitmask_imm): New function.
7910 (aarch64_bitmask_imm): Use it.
7911 (aarch64_cannot_force_const_mem): Reject expressions involving
7912 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7913 (aarch64_classify_index): Handle SVE indices, by requiring
7914 a plain register index with a scale that matches the element size.
7915 (aarch64_classify_address): Handle SVE addresses. Assert that
7916 the mode of the address is VOIDmode or an integer mode.
7917 Update call to aarch64_classify_symbol.
7918 (aarch64_classify_symbolic_expression): Update call to
7919 aarch64_classify_symbol.
7920 (aarch64_const_vec_all_in_range_p): New function.
7921 (aarch64_print_vector_float_operand): Likewise.
7922 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7923 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7924 and the FP immediates 1.0 and 0.5.
7925 (aarch64_print_address_internal): Handle SVE addresses.
7926 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7927 (aarch64_regno_regclass): Handle predicate registers.
7928 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7930 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7931 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7932 (aarch64_convert_sve_vector_bits): New function.
7933 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7934 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7936 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7937 Handle SVE vector and predicate modes. Accept VL-based constants
7938 that need only one temporary register, and VL offsets that require
7939 no temporary registers.
7940 (aarch64_conditional_register_usage): Mark the predicate registers
7941 as fixed if SVE isn't available.
7942 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7943 Return true for SVE vector and predicate modes.
7944 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7945 rather than an unsigned int. Handle SVE modes.
7946 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7948 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7950 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7951 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7952 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7953 (aarch64_sve_float_mul_immediate_p): New functions.
7954 (aarch64_sve_valid_immediate): New function.
7955 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7956 Explicitly reject structure modes. Check for INDEX constants.
7957 Handle PTRUE and PFALSE constants.
7958 (aarch64_check_zero_based_sve_index_immediate): New function.
7959 (aarch64_simd_imm_zero_p): Delete.
7960 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7961 vector modes. Accept constants in the range of CNT[BHWD].
7962 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7963 ask for an Advanced SIMD mode.
7964 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7965 (aarch64_simd_vector_alignment): Handle SVE predicates.
7966 (aarch64_vectorize_preferred_vector_alignment): New function.
7967 (aarch64_simd_vector_alignment_reachable): Use it instead of
7969 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7970 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7972 (MAX_VECT_LEN): Delete.
7973 (expand_vec_perm_d): Add a vec_flags field.
7974 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7975 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7976 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7978 (aarch64_evpc_rev): Rename to...
7979 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7980 (aarch64_evpc_rev_global): New function.
7981 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7982 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7984 (aarch64_evpc_sve_tbl): New function.
7985 (aarch64_expand_vec_perm_const_1): Update after rename of
7986 aarch64_evpc_rev. Handle SVE permutes too, trying
7987 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7988 than aarch64_evpc_tbl.
7989 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7990 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7991 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7992 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7993 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7994 (aarch64_expand_sve_vcond): New functions.
7995 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7996 of aarch64_vector_mode_p.
7997 (aarch64_dwarf_poly_indeterminate_value): New function.
7998 (aarch64_compute_pressure_classes): Likewise.
7999 (aarch64_can_change_mode_class): Likewise.
8000 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8001 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8002 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8003 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8004 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8005 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8006 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8007 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8009 (Dn, Dl, Dr): Accept const as well as const_vector.
8010 (Dz): Likewise. Compare against CONST0_RTX.
8011 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8012 of "vector" where appropriate.
8013 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8014 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8015 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8016 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8017 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8018 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8019 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8020 (v_int_equiv): Extend to SVE modes.
8021 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8023 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8024 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8025 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8026 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8027 (SVE_COND_FP_CMP): New int iterators.
8028 (perm_hilo): Handle the new unpack unspecs.
8029 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8031 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8032 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8033 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8034 (aarch64_equality_operator, aarch64_constant_vector_operand)
8035 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8036 (aarch64_sve_nonimmediate_operand): Likewise.
8037 (aarch64_sve_general_operand): Likewise.
8038 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8039 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8040 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8041 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8042 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8043 (aarch64_sve_float_arith_immediate): Likewise.
8044 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8045 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8046 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8047 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8048 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8049 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8050 (aarch64_sve_float_arith_operand): Likewise.
8051 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8052 (aarch64_sve_float_mul_operand): Likewise.
8053 (aarch64_sve_vec_perm_operand): Likewise.
8054 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8055 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8056 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8057 as well as const_vector.
8058 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8059 in file. Use CONST0_RTX and CONSTM1_RTX.
8060 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8061 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8062 Use aarch64_simd_imm_zero.
8063 * config/aarch64/aarch64-sve.md: New file.
8064 * config/aarch64/aarch64.md: Include it.
8065 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8066 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8067 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8068 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8069 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8070 (sve): New attribute.
8071 (enabled): Disable instructions with the sve attribute unless
8073 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8074 aarch64_expand_mov_immediate.
8075 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8076 CNT[BHSD] immediates.
8077 (movti): Split CONST_POLY_INT moves into two halves.
8078 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8079 Split additions that need a temporary here if the destination
8080 is the stack pointer.
8081 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8082 (*add<mode>3_poly_1): New instruction.
8083 (set_clobber_cc): New expander.
8085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8087 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8088 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8089 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8090 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8091 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8092 Change innermode from fixed_mode_size to machine_mode.
8093 (simplify_subreg): Update call accordingly. Handle a constant-sized
8094 subreg of a variable-length CONST_VECTOR.
8096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8097 Alan Hayward <alan.hayward@arm.com>
8098 David Sherwood <david.sherwood@arm.com>
8100 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8101 (add_offset_to_base): New function, split out from...
8102 (create_mem_ref): ...here. When handling a scale other than 1,
8103 check first whether the address is valid without the offset.
8104 Add it into the base if so, leaving the index and scale as-is.
8106 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8109 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8110 fold_for_warn before checking if arg2 is INTEGER_CST.
8112 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8114 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8115 (store_multiple_operation): Delete.
8116 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8117 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8118 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8119 guarded by TARGET_STRING.
8120 (rs6000_output_load_multiple): Delete.
8121 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8122 OPTION_MASK_STRING / TARGET_STRING handling.
8123 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8124 (const rs6000_opt_masks) <"string">: Change mask to 0.
8125 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8126 (MASK_STRING): Delete.
8127 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8129 (load_multiple): Delete.
8136 (store_multiple): Delete.
8143 (movmemsi_8reg): Delete.
8144 (corresponding unnamed define_insn): Delete.
8145 (movmemsi_6reg): Delete.
8146 (corresponding unnamed define_insn): Delete.
8147 (movmemsi_4reg): Delete.
8148 (corresponding unnamed define_insn): Delete.
8149 (movmemsi_2reg): Delete.
8150 (corresponding unnamed define_insn): Delete.
8151 (movmemsi_1reg): Delete.
8152 (corresponding unnamed define_insn): Delete.
8153 * config/rs6000/rs6000.opt (mno-string): New.
8154 (mstring): Replace by deprecation warning stub.
8155 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8157 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8159 * regrename.c (regrename_do_replace): If replacing the same
8160 reg multiple times, try to reuse last created gen_raw_REG.
8163 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8164 main to workaround a bug in GDB.
8166 2018-01-12 Tom de Vries <tom@codesourcery.com>
8169 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8171 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8173 PR rtl-optimization/80481
8174 * ira-color.c (get_cap_member): New function.
8175 (allocnos_conflict_by_live_ranges_p): Use it.
8176 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8177 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8179 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8182 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8183 (*saddl_se_1): Ditto.
8185 (*ssubl_se_1): Ditto.
8187 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8189 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8190 rather than wi::to_widest for DR_INITs.
8191 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8192 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8193 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8195 (vect_analyze_group_access_1): Note that here.
8197 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8199 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8200 polynomial type sizes.
8202 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8204 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8205 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8206 (gimple_add_tmp_var): Likewise.
8208 2018-01-12 Martin Liska <mliska@suse.cz>
8210 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8211 (gimple_alloc_sizes): Likewise.
8212 (dump_gimple_statistics): Use PRIu64 in printf format.
8213 * gimple.h: Change uint64_t to int.
8215 2018-01-12 Martin Liska <mliska@suse.cz>
8217 * tree-core.h: Use uint64_t instead of int.
8218 * tree.c (tree_node_counts): Likewise.
8219 (tree_node_sizes): Likewise.
8220 (dump_tree_statistics): Use PRIu64 in printf format.
8222 2018-01-12 Martin Liska <mliska@suse.cz>
8224 * Makefile.in: As qsort_chk is implemented in vec.c, add
8225 vec.o to linkage of gencfn-macros.
8226 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8227 passing the info to record_node_allocation_statistics.
8228 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8230 * ggc-common.c (struct ggc_usage): Add operator== and use
8231 it in operator< and compare function.
8232 * mem-stats.h (struct mem_usage): Likewise.
8233 * vec.c (struct vec_usage): Remove operator< and compare
8234 function. Can be simply inherited.
8236 2018-01-12 Martin Jambor <mjambor@suse.cz>
8239 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8240 * tree-ssa-math-opts.c: Include domwalk.h.
8241 (convert_mult_to_fma_1): New function.
8242 (fma_transformation_info): New type.
8243 (fma_deferring_state): Likewise.
8244 (cancel_fma_deferring): New function.
8245 (result_of_phi): Likewise.
8246 (last_fma_candidate_feeds_initial_phi): Likewise.
8247 (convert_mult_to_fma): Added deferring logic, split actual
8248 transformation to convert_mult_to_fma_1.
8249 (math_opts_dom_walker): New type.
8250 (math_opts_dom_walker::after_dom_children): New method, body moved
8251 here from pass_optimize_widening_mul::execute, added deferring logic
8253 (pass_optimize_widening_mul::execute): Moved most of code to
8254 math_opts_dom_walker::after_dom_children.
8255 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8256 * config/i386/i386.c (ix86_option_override_internal): Added
8257 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8259 2018-01-12 Richard Biener <rguenther@suse.de>
8262 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8263 inline instance vars.
8265 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8268 * config/rx/rx.c (rx_is_restricted_memory_address):
8271 2018-01-12 Richard Biener <rguenther@suse.de>
8273 PR tree-optimization/80846
8274 * target.def (split_reduction): New target hook.
8275 * targhooks.c (default_split_reduction): New function.
8276 * targhooks.h (default_split_reduction): Declare.
8277 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8278 target requests first reduce vectors by combining low and high
8280 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8281 (get_vectype_for_scalar_type_and_size): Export.
8282 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8283 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8284 * doc/tm.texi: Regenerate.
8285 * config/i386/i386.c (ix86_split_reduction): Implement
8286 TARGET_VECTORIZE_SPLIT_REDUCTION.
8288 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8291 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8292 in PIC mode except for TARGET_VXWORKS_RTP.
8293 * config/sparc/sparc.c: Include cfgrtl.h.
8294 (TARGET_INIT_PIC_REG): Define.
8295 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8296 (sparc_pic_register_p): New predicate.
8297 (sparc_legitimate_address_p): Use it.
8298 (sparc_legitimize_pic_address): Likewise.
8299 (sparc_delegitimize_address): Likewise.
8300 (sparc_mode_dependent_address_p): Likewise.
8301 (gen_load_pcrel_sym): Remove 4th parameter.
8302 (load_got_register): Adjust call to above. Remove obsolete stuff.
8303 (sparc_expand_prologue): Do not call load_got_register here.
8304 (sparc_flat_expand_prologue): Likewise.
8305 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8306 (sparc_use_pseudo_pic_reg): New function.
8307 (sparc_init_pic_reg): Likewise.
8308 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8309 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8311 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8313 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8314 Add item for branch_cost.
8316 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8318 PR rtl-optimization/83565
8319 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8320 not extend the result to a larger mode for rotate operations.
8321 (num_sign_bit_copies1): Likewise.
8323 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8326 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8328 Use values-Xc.o for -pedantic.
8329 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8331 2018-01-12 Martin Liska <mliska@suse.cz>
8334 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8336 (possible_polymorphic_call_targets): Use it.
8337 (ipa_devirt): Likewise.
8339 2018-01-12 Martin Liska <mliska@suse.cz>
8341 * profile-count.h (enum profile_quality): Use 0 as invalid
8342 enum value of profile_quality.
8344 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8346 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8347 -mext-string options.
8349 2018-01-12 Richard Biener <rguenther@suse.de>
8351 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8352 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8353 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8355 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8357 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8359 * configure.ac (--with-long-double-format): Add support for the
8360 configuration option to change the default long double format on
8362 * config.gcc (powerpc*-linux*-*): Likewise.
8363 * configure: Regenerate.
8364 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8365 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8366 used without modification.
8368 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8370 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8371 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8372 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8373 MISC_BUILTIN_SPEC_BARRIER.
8374 (rs6000_init_builtins): Likewise.
8375 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8377 (speculation_barrier): New define_insn.
8378 * doc/extend.texi: Document __builtin_speculation_barrier.
8380 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8383 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8384 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8385 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8387 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8388 integral modes instead of "ss" and "sd".
8389 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8390 vectors with 32-bit and 64-bit elements.
8391 (vecdupssescalarmodesuffix): New mode attribute.
8392 (vec_dup<mode>): Use it.
8394 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8397 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8398 frame if argument is passed on stack.
8400 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8403 * ree.c (combine_reaching_defs): Optimize also
8404 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8405 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8407 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8410 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8412 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8415 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8416 after they are computed.
8418 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8420 PR tree-optimization/83695
8421 * gimple-loop-linterchange.cc
8422 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8423 reset cached scev information after interchange.
8424 (pass_linterchange::execute): Remove call to scev_reset_htab.
8426 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8428 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8429 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8430 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8431 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8432 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8433 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8434 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8435 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8436 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8437 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8438 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8439 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8440 (V_lane_reg): Likewise.
8441 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8443 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8444 (vfmal_lane_low<mode>_intrinsic,
8445 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8446 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8447 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8448 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8449 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8450 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8452 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8454 * config/arm/arm-cpus.in (fp16fml): New feature.
8455 (ALL_SIMD): Add fp16fml.
8456 (armv8.2-a): Add fp16fml as an option.
8457 (armv8.3-a): Likewise.
8458 (armv8.4-a): Add fp16fml as part of fp16.
8459 * config/arm/arm.h (TARGET_FP16FML): Define.
8460 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8462 * config/arm/arm-modes.def (V2HF): Define.
8463 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8464 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8465 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8466 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8467 vfmsl_low, vfmsl_high): New set of builtins.
8468 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8469 (vfml_op): New code attribute.
8470 (VFMLHALVES): New int iterator.
8471 (VFML, VFMLSEL): New mode attributes.
8472 (V_reg): Define mapping for V2HF.
8473 (V_hi, V_lo): New mode attributes.
8474 (VF_constraint): Likewise.
8475 (vfml_half, vfml_half_selector): New int attributes.
8476 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8478 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8479 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8481 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8482 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8483 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8484 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8486 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8487 Document new effective target and option set.
8489 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8491 * config/arm/arm-cpus.in (armv8_4): New feature.
8492 (ARMv8_4a): New fgroup.
8493 (armv8.4-a): New arch.
8494 * config/arm/arm-tables.opt: Regenerate.
8495 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8496 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8497 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8498 Add matching rules for -march=armv8.4-a and extensions.
8499 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8501 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8504 * config/rx/rx.md (BW): New mode attribute.
8505 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8507 2018-01-11 Richard Biener <rguenther@suse.de>
8509 PR tree-optimization/83435
8510 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8511 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8512 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8514 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8515 Alan Hayward <alan.hayward@arm.com>
8516 David Sherwood <david.sherwood@arm.com>
8518 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8520 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8521 (aarch64_print_address_internal): Use it to check for a zero offset.
8523 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8524 Alan Hayward <alan.hayward@arm.com>
8525 David Sherwood <david.sherwood@arm.com>
8527 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8528 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8529 Return a poly_int64 rather than a HOST_WIDE_INT.
8530 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8531 rather than a HOST_WIDE_INT.
8532 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8533 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8534 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8535 final_offset from HOST_WIDE_INT to poly_int64.
8536 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8537 to_constant when getting the number of units in an Advanced SIMD
8539 (aarch64_builtin_vectorized_function): Check for a constant number
8541 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8543 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8544 attribute instead of GET_MODE_NUNITS.
8545 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8546 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8547 GET_MODE_SIZE for fixed-size registers.
8548 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8549 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8550 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8551 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8552 (aarch64_print_operand, aarch64_print_address_internal)
8553 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8554 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8555 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8556 Handle polynomial GET_MODE_SIZE.
8557 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8558 wider than SImode without modification.
8559 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8560 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8561 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8562 passing and returning SVE modes.
8563 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8564 rather than GEN_INT.
8565 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8566 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8567 (aarch64_allocate_and_probe_stack_space): Likewise.
8568 (aarch64_layout_frame): Cope with polynomial offsets.
8569 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8570 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8572 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8573 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8574 poly_int64 rather than a HOST_WIDE_INT.
8575 (aarch64_get_separate_components, aarch64_process_components)
8576 (aarch64_expand_prologue, aarch64_expand_epilogue)
8577 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8578 (aarch64_anchor_offset): New function, split out from...
8579 (aarch64_legitimize_address): ...here.
8580 (aarch64_builtin_vectorization_cost): Handle polynomial
8581 TYPE_VECTOR_SUBPARTS.
8582 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8584 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8585 number of elements from the PARALLEL rather than the mode.
8586 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8587 rather than GET_MODE_BITSIZE.
8588 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8589 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8590 (aarch64_expand_vec_perm_const_1): Handle polynomial
8591 d->perm.length () and d->perm elements.
8592 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8593 Apply to_constant to d->perm elements.
8594 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8595 polynomial CONST_VECTOR_NUNITS.
8596 (aarch64_move_pointer): Take amount as a poly_int64 rather
8598 (aarch64_progress_pointer): Avoid temporary variable.
8599 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8600 the mode attribute instead of GET_MODE.
8602 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8603 Alan Hayward <alan.hayward@arm.com>
8604 David Sherwood <david.sherwood@arm.com>
8606 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8607 x exists before using it.
8608 (aarch64_add_constant_internal): Rename to...
8609 (aarch64_add_offset_1): ...this. Replace regnum with separate
8610 src and dest rtxes. Handle the case in which they're different,
8611 including when the offset is zero. Replace scratchreg with an rtx.
8612 Use 2 additions if there is no spare register into which we can
8613 move a 16-bit constant.
8614 (aarch64_add_constant): Delete.
8615 (aarch64_add_offset): Replace reg with separate src and dest
8616 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8617 Use aarch64_add_offset_1.
8618 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8619 an rtx rather than an int. Take the delta as a poly_int64
8620 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8621 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8622 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8623 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8624 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8626 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8627 aarch64_add_constant.
8629 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8631 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8632 Use scalar_float_mode.
8634 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8636 * config/aarch64/aarch64-simd.md
8637 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8638 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8639 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8640 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8641 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8642 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8643 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8644 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8645 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8646 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8648 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8651 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8652 targ_options->x_arm_arch_string is non NULL.
8654 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8656 * config/aarch64/aarch64.h
8657 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8659 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8662 * expmed.c (emit_store_flag_force): Swap if const op0
8663 and change VOIDmode to mode of op0.
8665 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8667 PR rtl-optimization/83761
8668 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8669 than bytes to mode_for_size.
8671 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8674 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8675 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8678 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8681 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8682 when in layout mode.
8683 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8684 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8687 2018-01-10 Michael Collison <michael.collison@arm.com>
8689 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8690 * config/aarch64/aarch64-option-extension.def: Add
8691 AARCH64_OPT_EXTENSION of 'fp16fml'.
8692 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8693 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8694 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8695 * config/aarch64/constraints.md (Ui7): New constraint.
8696 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8697 (VFMLA_SEL_W): Ditto.
8700 (VFMLA16_LOW): New int iterator.
8701 (VFMLA16_HIGH): Ditto.
8702 (UNSPEC_FMLAL): New unspec.
8703 (UNSPEC_FMLSL): Ditto.
8704 (UNSPEC_FMLAL2): Ditto.
8705 (UNSPEC_FMLSL2): Ditto.
8706 (f16mac): New code attribute.
8707 * config/aarch64/aarch64-simd-builtins.def
8708 (aarch64_fmlal_lowv2sf): Ditto.
8709 (aarch64_fmlsl_lowv2sf): Ditto.
8710 (aarch64_fmlalq_lowv4sf): Ditto.
8711 (aarch64_fmlslq_lowv4sf): Ditto.
8712 (aarch64_fmlal_highv2sf): Ditto.
8713 (aarch64_fmlsl_highv2sf): Ditto.
8714 (aarch64_fmlalq_highv4sf): Ditto.
8715 (aarch64_fmlslq_highv4sf): Ditto.
8716 (aarch64_fmlal_lane_lowv2sf): Ditto.
8717 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8718 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8719 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8720 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8721 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8722 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8723 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8724 (aarch64_fmlal_lane_highv2sf): Ditto.
8725 (aarch64_fmlsl_lane_highv2sf): Ditto.
8726 (aarch64_fmlal_laneq_highv2sf): Ditto.
8727 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8728 (aarch64_fmlalq_lane_highv4sf): Ditto.
8729 (aarch64_fmlsl_lane_highv4sf): Ditto.
8730 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8731 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8732 * config/aarch64/aarch64-simd.md:
8733 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8734 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8735 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8736 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8737 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8738 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8739 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8740 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8741 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8742 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8743 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8744 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8745 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8746 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8747 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8748 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8749 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8750 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8751 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8752 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8753 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8754 (vfmlsl_low_u32): Ditto.
8755 (vfmlalq_low_u32): Ditto.
8756 (vfmlslq_low_u32): Ditto.
8757 (vfmlal_high_u32): Ditto.
8758 (vfmlsl_high_u32): Ditto.
8759 (vfmlalq_high_u32): Ditto.
8760 (vfmlslq_high_u32): Ditto.
8761 (vfmlal_lane_low_u32): Ditto.
8762 (vfmlsl_lane_low_u32): Ditto.
8763 (vfmlal_laneq_low_u32): Ditto.
8764 (vfmlsl_laneq_low_u32): Ditto.
8765 (vfmlalq_lane_low_u32): Ditto.
8766 (vfmlslq_lane_low_u32): Ditto.
8767 (vfmlalq_laneq_low_u32): Ditto.
8768 (vfmlslq_laneq_low_u32): Ditto.
8769 (vfmlal_lane_high_u32): Ditto.
8770 (vfmlsl_lane_high_u32): Ditto.
8771 (vfmlal_laneq_high_u32): Ditto.
8772 (vfmlsl_laneq_high_u32): Ditto.
8773 (vfmlalq_lane_high_u32): Ditto.
8774 (vfmlslq_lane_high_u32): Ditto.
8775 (vfmlalq_laneq_high_u32): Ditto.
8776 (vfmlslq_laneq_high_u32): Ditto.
8777 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8778 (AARCH64_FL_FOR_ARCH8_4): New.
8779 (AARCH64_ISA_F16FML): New ISA flag.
8780 (TARGET_F16FML): New feature flag for fp16fml.
8781 (doc/invoke.texi): Document new fp16fml option.
8783 2018-01-10 Michael Collison <michael.collison@arm.com>
8785 * config/aarch64/aarch64-builtins.c:
8786 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8787 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8788 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8789 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8790 (AARCH64_ISA_SHA3): New ISA flag.
8791 (TARGET_SHA3): New feature flag for sha3.
8792 * config/aarch64/iterators.md (sha512_op): New int attribute.
8793 (CRYPTO_SHA512): New int iterator.
8794 (UNSPEC_SHA512H): New unspec.
8795 (UNSPEC_SHA512H2): Ditto.
8796 (UNSPEC_SHA512SU0): Ditto.
8797 (UNSPEC_SHA512SU1): Ditto.
8798 * config/aarch64/aarch64-simd-builtins.def
8799 (aarch64_crypto_sha512hqv2di): New builtin.
8800 (aarch64_crypto_sha512h2qv2di): Ditto.
8801 (aarch64_crypto_sha512su0qv2di): Ditto.
8802 (aarch64_crypto_sha512su1qv2di): Ditto.
8803 (aarch64_eor3qv8hi): Ditto.
8804 (aarch64_rax1qv2di): Ditto.
8805 (aarch64_xarqv2di): Ditto.
8806 (aarch64_bcaxqv8hi): Ditto.
8807 * config/aarch64/aarch64-simd.md:
8808 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8809 (aarch64_crypto_sha512su0qv2di): Ditto.
8810 (aarch64_crypto_sha512su1qv2di): Ditto.
8811 (aarch64_eor3qv8hi): Ditto.
8812 (aarch64_rax1qv2di): Ditto.
8813 (aarch64_xarqv2di): Ditto.
8814 (aarch64_bcaxqv8hi): Ditto.
8815 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8816 (vsha512h2q_u64): Ditto.
8817 (vsha512su0q_u64): Ditto.
8818 (vsha512su1q_u64): Ditto.
8819 (veor3q_u16): Ditto.
8820 (vrax1q_u64): Ditto.
8822 (vbcaxq_u16): Ditto.
8823 * config/arm/types.md (crypto_sha512): New type attribute.
8824 (crypto_sha3): Ditto.
8825 (doc/invoke.texi): Document new sha3 option.
8827 2018-01-10 Michael Collison <michael.collison@arm.com>
8829 * config/aarch64/aarch64-builtins.c:
8830 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8831 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8832 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8833 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8834 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8835 (AARCH64_ISA_SM4): New ISA flag.
8836 (TARGET_SM4): New feature flag for sm4.
8837 * config/aarch64/aarch64-simd-builtins.def
8838 (aarch64_sm3ss1qv4si): Ditto.
8839 (aarch64_sm3tt1aq4si): Ditto.
8840 (aarch64_sm3tt1bq4si): Ditto.
8841 (aarch64_sm3tt2aq4si): Ditto.
8842 (aarch64_sm3tt2bq4si): Ditto.
8843 (aarch64_sm3partw1qv4si): Ditto.
8844 (aarch64_sm3partw2qv4si): Ditto.
8845 (aarch64_sm4eqv4si): Ditto.
8846 (aarch64_sm4ekeyqv4si): Ditto.
8847 * config/aarch64/aarch64-simd.md:
8848 (aarch64_sm3ss1qv4si): Ditto.
8849 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8850 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8851 (aarch64_sm4eqv4si): Ditto.
8852 (aarch64_sm4ekeyqv4si): Ditto.
8853 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8854 (sm3part_op): Ditto.
8855 (CRYPTO_SM3TT): Ditto.
8856 (CRYPTO_SM3PART): Ditto.
8857 (UNSPEC_SM3SS1): New unspec.
8858 (UNSPEC_SM3TT1A): Ditto.
8859 (UNSPEC_SM3TT1B): Ditto.
8860 (UNSPEC_SM3TT2A): Ditto.
8861 (UNSPEC_SM3TT2B): Ditto.
8862 (UNSPEC_SM3PARTW1): Ditto.
8863 (UNSPEC_SM3PARTW2): Ditto.
8864 (UNSPEC_SM4E): Ditto.
8865 (UNSPEC_SM4EKEY): Ditto.
8866 * config/aarch64/constraints.md (Ui2): New constraint.
8867 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8868 * config/arm/types.md (crypto_sm3): New type attribute.
8869 (crypto_sm4): Ditto.
8870 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8871 (vsm3tt1aq_u32): Ditto.
8872 (vsm3tt1bq_u32): Ditto.
8873 (vsm3tt2aq_u32): Ditto.
8874 (vsm3tt2bq_u32): Ditto.
8875 (vsm3partw1q_u32): Ditto.
8876 (vsm3partw2q_u32): Ditto.
8877 (vsm4eq_u32): Ditto.
8878 (vsm4ekeyq_u32): Ditto.
8879 (doc/invoke.texi): Document new sm4 option.
8881 2018-01-10 Michael Collison <michael.collison@arm.com>
8883 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8884 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8885 (AARCH64_FL_FOR_ARCH8_4): New.
8886 (AARCH64_FL_V8_4): New flag.
8887 (doc/invoke.texi): Document new armv8.4-a option.
8889 2018-01-10 Michael Collison <michael.collison@arm.com>
8891 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8892 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8893 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8894 * config/aarch64/aarch64-option-extension.def: Add
8895 AARCH64_OPT_EXTENSION of 'sha2'.
8896 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8897 (crypto): Disable sha2 and aes if crypto disabled.
8898 (crypto): Enable aes and sha2 if enabled.
8899 (simd): Disable sha2 and aes if simd disabled.
8900 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8902 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8903 (TARGET_SHA2): New feature flag for sha2.
8904 (TARGET_AES): New feature flag for aes.
8905 * config/aarch64/aarch64-simd.md:
8906 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8907 conditional on TARGET_AES.
8908 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8909 (aarch64_crypto_sha1hsi): Make pattern conditional
8911 (aarch64_crypto_sha1hv4si): Ditto.
8912 (aarch64_be_crypto_sha1hv4si): Ditto.
8913 (aarch64_crypto_sha1su1v4si): Ditto.
8914 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8915 (aarch64_crypto_sha1su0v4si): Ditto.
8916 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8917 (aarch64_crypto_sha256su0v4si): Ditto.
8918 (aarch64_crypto_sha256su1v4si): Ditto.
8919 (doc/invoke.texi): Document new aes and sha2 options.
8921 2018-01-10 Martin Sebor <msebor@redhat.com>
8923 PR tree-optimization/83781
8924 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8927 2018-01-11 Martin Sebor <msebor@gmail.com>
8928 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8930 PR tree-optimization/83501
8931 PR tree-optimization/81703
8933 * tree-ssa-strlen.c (get_string_cst): Rename...
8934 (get_string_len): ...to this. Handle global constants.
8935 (handle_char_store): Adjust.
8937 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8938 Jim Wilson <jimw@sifive.com>
8940 * config/riscv/riscv-protos.h (riscv_output_return): New.
8941 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8942 (riscv_attribute_table, riscv_output_return),
8943 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8944 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8945 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8947 (riscv_expand_prologue): Add early return for naked function.
8948 (riscv_expand_epilogue): Likewise.
8949 (riscv_function_ok_for_sibcall): Return false for naked function.
8950 (riscv_set_current_function): New.
8951 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8952 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8953 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8954 * doc/extend.texi (RISC-V Function Attributes): New.
8956 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8958 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8959 check for 128-bit long double before checking TCmode.
8960 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8961 128-bit long doubles before checking TFmode or TCmode.
8962 (FLOAT128_IBM_P): Likewise.
8964 2018-01-10 Martin Sebor <msebor@redhat.com>
8966 PR tree-optimization/83671
8967 * builtins.c (c_strlen): Unconditionally return zero for the empty
8969 Use -Warray-bounds for warnings.
8970 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8971 for non-constant array indices with COMPONENT_REF, arrays of
8972 arrays, and pointers to arrays.
8973 (gimple_fold_builtin_strlen): Determine and set length range for
8974 non-constant character arrays.
8976 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8979 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8982 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8984 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8986 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8989 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8990 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8991 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8992 indexed_or_indirect_operand predicate.
8993 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8994 (*vsx_le_perm_load_v8hi): Likewise.
8995 (*vsx_le_perm_load_v16qi): Likewise.
8996 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8997 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8998 (*vsx_le_perm_store_v8hi): Likewise.
8999 (*vsx_le_perm_store_v16qi): Likewise.
9000 (eight unnamed splitters): Likewise.
9002 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9004 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9005 * config/rs6000/emmintrin.h: Likewise.
9006 * config/rs6000/mmintrin.h: Likewise.
9007 * config/rs6000/xmmintrin.h: Likewise.
9009 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9012 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9014 * tree.c (tree_nop_conversion): Return true for location wrapper
9016 (maybe_wrap_with_location): New function.
9017 (selftest::check_strip_nops): New function.
9018 (selftest::test_location_wrappers): New function.
9019 (selftest::tree_c_tests): Call it.
9020 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9021 (maybe_wrap_with_location): New decl.
9022 (EXPR_LOCATION_WRAPPER_P): New macro.
9023 (location_wrapper_p): New inline function.
9024 (tree_strip_any_location_wrapper): New inline function.
9026 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9029 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9030 stack_realign_offset for the largest alignment of stack slot
9032 (ix86_find_max_used_stack_alignment): New function.
9033 (ix86_finalize_stack_frame_flags): Use it. Set
9034 max_used_stack_alignment if we don't realign stack.
9035 * config/i386/i386.h (machine_function): Add
9036 max_used_stack_alignment.
9038 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9040 * config/arm/arm.opt (-mbranch-cost): New option.
9041 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9044 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9047 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9048 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9050 2018-01-10 Richard Biener <rguenther@suse.de>
9053 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9054 early out so it also covers the case where we have a non-NULL
9057 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9059 PR tree-optimization/83753
9060 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9061 for non-strided grouped accesses if the number of elements is 1.
9063 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9066 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9067 * i386.h (TARGET_USE_GATHER): Define.
9068 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9070 2018-01-10 Martin Liska <mliska@suse.cz>
9073 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9074 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9076 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9077 CLEANUP_NO_PARTITIONING is not set.
9079 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9081 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9082 for vectors, as a partial revert of r254296.
9083 * rtl.h (const_vec_p): Delete.
9084 (const_vec_duplicate_p): Don't test for vector CONSTs.
9085 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9086 * expmed.c (make_tree): Likewise.
9089 * common.md (E, F): Use CONSTANT_P instead of checking for
9091 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9092 checking for CONST_VECTOR.
9094 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9097 * predict.c (force_edge_cold): Handle in more sane way edges
9100 2018-01-09 Carl Love <cel@us.ibm.com>
9102 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9104 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9105 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9106 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9107 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9108 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9109 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9110 * config/rs6000/rs6000-protos.h: Add extern defition for
9111 rs6000_generate_float2_double_code.
9112 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9114 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9115 (float2_v2df): Add define_expand.
9117 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9120 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9121 op_mode in the force_to_mode call.
9123 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9125 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9126 instead of checking each element individually.
9127 (aarch64_evpc_uzp): Likewise.
9128 (aarch64_evpc_zip): Likewise.
9129 (aarch64_evpc_ext): Likewise.
9130 (aarch64_evpc_rev): Likewise.
9131 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9132 instead of checking each element individually. Return true without
9134 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9135 whether all selected elements come from the same input, instead of
9136 checking each element individually. Remove calls to gen_rtx_REG,
9137 start_sequence and end_sequence and instead assert that no rtl is
9140 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9142 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9143 order of HIGH and CONST checks.
9145 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9147 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9148 if the destination isn't an SSA_NAME.
9150 2018-01-09 Richard Biener <rguenther@suse.de>
9152 PR tree-optimization/83668
9153 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9155 (canonicalize_loop_form): ... here, renamed from ...
9156 (canonicalize_loop_closed_ssa_form): ... this and amended to
9157 swap successor edges for loop exit blocks to make us use
9158 the RPO order we need for initial schedule generation.
9160 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9162 PR tree-optimization/64811
9163 * match.pd: When optimizing comparisons with Inf, avoid
9164 introducing or losing exceptions from comparisons with NaN.
9166 2018-01-09 Martin Liska <mliska@suse.cz>
9169 * asan.c (shadow_mem_size): Add gcc_assert.
9171 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9173 Don't save registers in main().
9176 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9177 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9178 * config/avr/avr.c (avr_set_current_function): Don't error if
9179 naked, OS_task or OS_main are specified at the same time.
9180 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9182 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9184 * common/config/avr/avr-common.c (avr_option_optimization_table):
9185 Switch on -mmain-is-OS_task for optimizing compilations.
9187 2018-01-09 Richard Biener <rguenther@suse.de>
9189 PR tree-optimization/83572
9190 * graphite.c: Include cfganal.h.
9191 (graphite_transform_loops): Connect infinite loops to exit
9192 and remove fake edges at the end.
9194 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9196 * ipa-inline.c (edge_badness): Revert accidental checkin.
9198 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9201 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9202 symbols; not inline clones.
9204 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9207 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9208 hard registers. Formatting fixes.
9210 PR preprocessor/83722
9211 * gcc.c (try_generate_repro): Pass
9212 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9213 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9216 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9217 Kito Cheng <kito.cheng@gmail.com>
9219 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9220 (riscv_leaf_function_p): Delete.
9221 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9223 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9225 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9227 (do_ifelse): New function.
9228 (do_isel): New function.
9229 (do_sub3): New function.
9230 (do_add3): New function.
9231 (do_load_mask_compare): New function.
9232 (do_overlap_load_compare): New function.
9233 (expand_compare_loop): New function.
9234 (expand_block_compare): Call expand_compare_loop() when appropriate.
9235 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9237 (-mblock-compare-inline-loop-limit): New option.
9239 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9242 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9243 Reverse order of second and third operands in first alternative.
9244 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9245 of first and second elements in UNSPEC_VPERMR vector.
9246 (altivec_expand_vec_perm_le): Likewise.
9248 2018-01-08 Jeff Law <law@redhat.com>
9250 PR rtl-optimizatin/81308
9251 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9252 (process_switch): If group_case_labels makes a change, then set
9254 (pass_convert_switch::execute): If a switch is converted, then
9255 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9257 PR rtl-optimization/81308
9258 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9261 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9263 PR target/83663 - Revert r255946
9264 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9265 generation for cases where splatting a value is not useful.
9266 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9267 across a vec_duplicate and a paradoxical subreg forming a vector
9268 mode to a vec_concat.
9270 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9272 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9273 -march=armv8.3-a variants.
9274 * config/arm/t-multilib: Likewise.
9275 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9277 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9279 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9281 (cceq_ior_compare_complement): Give it a name so I can use it, and
9282 change boolean_or_operator predicate to boolean_operator so it can
9283 be used to generate a crand.
9284 (eqne): New code iterator.
9285 (bd/bd_neg): New code_attrs.
9286 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9287 a single define_insn.
9288 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9289 decrement (bdnzt/bdnzf/bdzt/bdzf).
9290 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9291 with the new names of the branch decrement patterns, and added the
9292 names of the branch decrement conditional patterns.
9294 2018-01-08 Richard Biener <rguenther@suse.de>
9296 PR tree-optimization/83563
9297 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9300 2018-01-08 Richard Biener <rguenther@suse.de>
9303 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9305 2018-01-08 Richard Biener <rguenther@suse.de>
9307 PR tree-optimization/83685
9308 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9309 references to abnormals.
9311 2018-01-08 Richard Biener <rguenther@suse.de>
9314 * dwarf2out.c (output_indirect_strings): Handle empty
9315 skeleton_debug_str_hash.
9316 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9318 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9320 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9321 (emit_store_direct): Likewise.
9322 (arc_trampoline_adjust_address): Likewise.
9323 (arc_asm_trampoline_template): New function.
9324 (arc_initialize_trampoline): Use asm_trampoline_template.
9325 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9326 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9327 * config/arc/arc.md (flush_icache): Delete pattern.
9329 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9331 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9332 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9335 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9338 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9339 by not USED_FOR_TARGET.
9340 (make_pass_resolve_sw_modes): Likewise.
9342 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9344 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9347 2018-01-08 Richard Biener <rguenther@suse.de>
9350 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9352 2018-01-08 Richard Biener <rguenther@suse.de>
9355 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9357 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9360 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9361 basic blocks with a small number of successors.
9362 (convert_control_dep_chain_into_preds): Improve handling of
9364 (dump_predicates): Split apart into...
9365 (dump_pred_chain): ...here...
9366 (dump_pred_info): ...and here.
9367 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9368 (can_chain_union_be_invalidated_p): Improve check for invalidation
9370 (uninit_uses_cannot_happen): Avoid unnecessary if
9371 convert_control_dep_chain_into_preds yielded nothing.
9373 2018-01-06 Martin Sebor <msebor@redhat.com>
9375 PR tree-optimization/83640
9376 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9377 subtracting negative offset from size.
9378 (builtin_access::overlap): Adjust offset bounds of the access to fall
9379 within the size of the object if possible.
9381 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9383 PR rtl-optimization/83699
9384 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9385 extract_bit_field_as_subreg to cases in which the extracted
9386 value is also a vector.
9388 * lra-constraints.c (process_alt_operands): Test for the equivalence
9389 substitutions when detecting a possible reload cycle.
9391 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9394 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9395 by default if flag_selective_schedling{,2}. Formatting fixes.
9397 PR rtl-optimization/83682
9398 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9399 if it has non-VECTOR_MODE element mode.
9400 (vec_duplicate_p): Likewise.
9403 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9404 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9406 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9409 * config/i386/i386-builtin.def
9410 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9411 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9412 Require also OPTION_MASK_ISA_AVX512F in addition to
9413 OPTION_MASK_ISA_GFNI.
9414 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9415 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9416 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9417 to OPTION_MASK_ISA_GFNI.
9418 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9419 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9420 OPTION_MASK_ISA_AVX512BW.
9421 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9422 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9423 addition to OPTION_MASK_ISA_GFNI.
9424 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9425 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9426 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9427 to OPTION_MASK_ISA_GFNI.
9428 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9429 a requirement for all ISAs rather than any of them with a few
9431 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9433 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9434 bitmasks to be enabled with 3 exceptions, instead of requiring any
9435 enabled ISA with lots of exceptions.
9436 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9437 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9438 Change avx512bw in isa attribute to avx512f.
9439 * config/i386/sgxintrin.h: Add license boilerplate.
9440 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9441 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9442 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9443 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9445 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9446 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9447 temporarily sse2 rather than sse if not enabled already.
9450 * config/i386/sse.md (VI248_VLBW): Rename to ...
9451 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9452 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9453 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9454 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9455 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9456 mode iterator instead of VI248_VLBW.
9458 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9460 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9461 (record_modified): Skip clobbers; add debug output.
9462 (param_change_prob): Use sreal frequencies.
9464 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9466 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9467 punt for user-aligned variables.
9469 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9471 * tree-chrec.c (chrec_contains_symbols): Return true for
9474 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9477 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9478 of (x|y) == x for BICS pattern.
9480 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9482 PR tree-optimization/83605
9483 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9484 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9487 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9489 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9490 * config/epiphany/rtems.h: New file.
9492 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9493 Uros Bizjak <ubizjak@gmail.com>
9496 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9497 QIreg_operand instead of register_operand predicate.
9498 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9499 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9500 comments instead of -fmitigate[-_]rop.
9502 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9505 * cgraphunit.c (symbol_table::compile): Switch to text_section
9506 before calling assembly_start debug hook.
9507 * run-rtl-passes.c (run_rtl_passes): Likewise.
9510 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9512 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9513 range_int_cst_p rather than !symbolic_range_p before calling
9514 extract_range_from_multiplicative_op_1.
9516 2018-01-04 Jeff Law <law@redhat.com>
9518 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9519 redundant test in assertion.
9521 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9523 * doc/rtl.texi: Document machine_mode wrapper classes.
9525 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9527 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9530 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9532 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9533 the VEC_PERM_EXPR fold to fail.
9535 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9538 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9539 to switched_sections.
9541 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9544 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9547 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9550 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9551 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9553 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9556 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9557 is BLKmode and bitpos not zero or mode change is needed.
9559 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9562 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9565 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9568 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9569 instead of MULT rtx. Update all corresponding splitters.
9571 (*ssub<modesuffix>): Ditto.
9573 (*cmp_sadd_di): Update split patterns.
9574 (*cmp_sadd_si): Ditto.
9575 (*cmp_sadd_sidi): Ditto.
9576 (*cmp_ssub_di): Ditto.
9577 (*cmp_ssub_si): Ditto.
9578 (*cmp_ssub_sidi): Ditto.
9579 * config/alpha/predicates.md (const23_operand): New predicate.
9580 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9581 Look for ASHIFT, not MULT inner operand.
9582 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9584 2018-01-04 Martin Liska <mliska@suse.cz>
9586 PR gcov-profile/83669
9587 * gcov.c (output_intermediate_file): Add version to intermediate
9589 * doc/gcov.texi: Document new field 'version' in intermediate
9590 file format. Fix location of '-k' option of gcov command.
9592 2018-01-04 Martin Liska <mliska@suse.cz>
9595 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9597 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9599 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9601 2018-01-03 Martin Sebor <msebor@redhat.com>
9603 PR tree-optimization/83655
9604 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9605 checking calls with invalid arguments.
9607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9609 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9610 (vectorizable_mask_load_store): Delete.
9611 (vectorizable_call): Return false for masked loads and stores.
9612 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9613 instead of gimple_assign_rhs1.
9614 (vectorizable_load): Handle IFN_MASK_LOAD.
9615 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9619 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9621 (vectorizable_mask_load_store): ...here.
9622 (vectorizable_load): ...and here.
9624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9626 * tree-vect-stmts.c (vect_build_all_ones_mask)
9627 (vect_build_zero_merge_argument): New functions, split out from...
9628 (vectorizable_load): ...here.
9630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9632 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9634 (vectorizable_mask_load_store): ...here.
9635 (vectorizable_store): ...and here.
9637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9639 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9641 (vectorizable_mask_load_store): ...here.
9643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9645 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9646 (vect_model_store_cost): Take a vec_load_store_type instead of a
9648 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9649 (vect_model_store_cost): Take a vec_load_store_type instead of a
9651 (vectorizable_mask_load_store): Update accordingly.
9652 (vectorizable_store): Likewise.
9653 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9657 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9658 IFN_MASK_LOAD calls here rather than...
9659 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9662 Alan Hayward <alan.hayward@arm.com>
9663 David Sherwood <david.sherwood@arm.com>
9665 * expmed.c (extract_bit_field_1): For vector extracts,
9666 fall back to extract_bit_field_as_subreg if vec_extract
9669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9670 Alan Hayward <alan.hayward@arm.com>
9671 David Sherwood <david.sherwood@arm.com>
9673 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9674 they are variable or constant sized.
9675 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9676 slots for constant-sized data.
9678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9679 Alan Hayward <alan.hayward@arm.com>
9680 David Sherwood <david.sherwood@arm.com>
9682 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9683 handling COND_EXPRs with boolean comparisons, try to find a better
9684 basis for the mask type than the boolean itself.
9686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9688 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9689 is calculated and how it can be overridden.
9690 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9691 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9693 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9697 Alan Hayward <alan.hayward@arm.com>
9698 David Sherwood <david.sherwood@arm.com>
9700 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9701 Remove the mode argument.
9702 (aarch64_simd_valid_immediate): Remove the mode and inverse
9704 * config/aarch64/iterators.md (bitsize): New iterator.
9705 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9706 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9707 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9708 aarch64_simd_valid_immediate.
9709 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9710 (aarch64_reg_or_bic_imm): Likewise.
9711 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9712 with an insn_type enum and msl with a modifier_type enum.
9713 Replace element_width with a scalar_mode. Change the shift
9714 to unsigned int. Add constructors for scalar_float_mode and
9715 scalar_int_mode elements.
9716 (aarch64_vect_float_const_representable_p): Delete.
9717 (aarch64_can_const_movi_rtx_p)
9718 (aarch64_simd_scalar_immediate_valid_for_move)
9719 (aarch64_simd_make_constant): Update call to
9720 aarch64_simd_valid_immediate.
9721 (aarch64_advsimd_valid_immediate_hs): New function.
9722 (aarch64_advsimd_valid_immediate): Likewise.
9723 (aarch64_simd_valid_immediate): Remove mode and inverse
9724 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9725 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9726 and aarch64_float_const_representable_p on the result.
9727 (aarch64_output_simd_mov_immediate): Remove mode argument.
9728 Update call to aarch64_simd_valid_immediate and use of
9729 simd_immediate_info.
9730 (aarch64_output_scalar_simd_mov_immediate): Update call
9733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9734 Alan Hayward <alan.hayward@arm.com>
9735 David Sherwood <david.sherwood@arm.com>
9737 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9738 (mode_nunits): Likewise CONST_MODE_NUNITS.
9739 * machmode.def (ADJUST_NUNITS): Document.
9740 * genmodes.c (mode_data::need_nunits_adj): New field.
9741 (blank_mode): Update accordingly.
9742 (adj_nunits): New variable.
9743 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9745 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9746 listed in adj_nunits.
9747 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9748 listed in adj_nunits. Don't emit case statements for such modes.
9749 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9750 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9751 nothing if adj_nunits is nonnull.
9752 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9753 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9754 (emit_mode_fbit): Update use of print_maybe_const_decl.
9755 (emit_move_size): Likewise. Treat the array as non-const
9757 (emit_mode_adjustments): Handle adj_nunits.
9759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9761 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9762 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9763 (VECTOR_MODES): Use it.
9764 (make_vector_modes): Take the prefix as an argument.
9766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9767 Alan Hayward <alan.hayward@arm.com>
9768 David Sherwood <david.sherwood@arm.com>
9770 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9771 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9772 for MODE_VECTOR_BOOL.
9773 * machmode.def (VECTOR_BOOL_MODE): Document.
9774 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9775 (make_vector_bool_mode): New function.
9776 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9778 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9779 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9781 * stor-layout.c (int_mode_for_mode): Likewise.
9782 * tree.c (build_vector_type_for_mode): Likewise.
9783 * varasm.c (output_constant_pool_2): Likewise.
9784 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9785 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9786 for MODE_VECTOR_BOOL.
9787 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9788 of mode class checks.
9789 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9790 instead of a list of mode class checks.
9791 (expand_vector_scalar_condition): Likewise.
9792 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9795 Alan Hayward <alan.hayward@arm.com>
9796 David Sherwood <david.sherwood@arm.com>
9798 * machmode.h (mode_size): Change from unsigned short to
9800 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9801 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9802 or if measurement_type is not polynomial.
9803 (fixed_size_mode::includes_p): Check for constant-sized modes.
9804 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9805 return a poly_uint16 rather than an unsigned short.
9806 (emit_mode_size): Change the type of mode_size from unsigned short
9807 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9808 (emit_mode_adjustments): Cope with polynomial vector sizes.
9809 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9811 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9813 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9814 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9815 * caller-save.c (setup_save_areas): Likewise.
9816 (replace_reg_with_saved_mem): Likewise.
9817 * calls.c (emit_library_call_value_1): Likewise.
9818 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9819 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9820 (gen_lowpart_for_combine): Likewise.
9821 * convert.c (convert_to_integer_1): Likewise.
9822 * cse.c (equiv_constant, cse_insn): Likewise.
9823 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9824 (cselib_subst_to_values): Likewise.
9825 * dce.c (word_dce_process_block): Likewise.
9826 * df-problems.c (df_word_lr_mark_ref): Likewise.
9827 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9828 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9829 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9830 (rtl_for_decl_location): Likewise.
9831 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9832 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9833 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9834 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9835 (expand_expr_real_1): Likewise.
9836 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9837 (pad_below): Likewise.
9838 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9839 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9840 * ira.c (get_subreg_tracking_sizes): Likewise.
9841 * ira-build.c (ira_create_allocno_objects): Likewise.
9842 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9843 (ira_sort_regnos_for_alter_reg): Likewise.
9844 * ira-costs.c (record_operand_costs): Likewise.
9845 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9846 (resolve_simple_move): Likewise.
9847 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9848 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9849 (lra_constraints): Likewise.
9850 (CONST_POOL_OK_P): Reject variable-sized modes.
9851 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9852 (add_pseudo_to_slot, lra_spill): Likewise.
9853 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9854 * optabs-query.c (get_best_extraction_insn): Likewise.
9855 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9856 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9857 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9858 * recog.c (offsettable_address_addr_space_p): Likewise.
9859 * regcprop.c (maybe_mode_change): Likewise.
9860 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9861 * regrename.c (build_def_use): Likewise.
9862 * regstat.c (dump_reg_info): Likewise.
9863 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9864 (find_reloads, find_reloads_subreg_address): Likewise.
9865 * reload1.c (eliminate_regs_1): Likewise.
9866 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9867 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9868 (simplify_binary_operation_1, simplify_subreg): Likewise.
9869 * targhooks.c (default_function_arg_padding): Likewise.
9870 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9871 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9872 (verify_gimple_assign_ternary): Likewise.
9873 * tree-inline.c (estimate_move_cost): Likewise.
9874 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9875 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9876 (get_address_cost_ainc): Likewise.
9877 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9878 (vect_supportable_dr_alignment): Likewise.
9879 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9880 (vectorizable_reduction): Likewise.
9881 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9882 (vectorizable_operation, vectorizable_load): Likewise.
9883 * tree.c (build_same_sized_truth_vector_type): Likewise.
9884 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9885 * var-tracking.c (emit_note_insn_var_location): Likewise.
9886 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9887 (ADDR_VEC_ALIGN): Likewise.
9889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9890 Alan Hayward <alan.hayward@arm.com>
9891 David Sherwood <david.sherwood@arm.com>
9893 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9895 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9896 or if measurement_type is polynomial.
9897 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9898 * combine.c (make_extraction): Likewise.
9899 * dse.c (find_shift_sequence): Likewise.
9900 * dwarf2out.c (mem_loc_descriptor): Likewise.
9901 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9902 (extract_bit_field, extract_low_bits): Likewise.
9903 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9904 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9905 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9906 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9907 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9908 * reload.c (find_reloads): Likewise.
9909 * reload1.c (alter_reg): Likewise.
9910 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9911 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9912 * tree-if-conv.c (predicate_mem_writes): Likewise.
9913 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9914 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9915 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9916 * valtrack.c (dead_debug_insert_temp): Likewise.
9917 * varasm.c (mergeable_constant_section): Likewise.
9918 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9921 Alan Hayward <alan.hayward@arm.com>
9922 David Sherwood <david.sherwood@arm.com>
9924 * expr.c (expand_assignment): Cope with polynomial mode sizes
9925 when assigning to a CONCAT.
9927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9928 Alan Hayward <alan.hayward@arm.com>
9929 David Sherwood <david.sherwood@arm.com>
9931 * machmode.h (mode_precision): Change from unsigned short to
9933 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9935 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9936 or if measurement_type is not polynomial.
9937 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9938 in which the mode is already known to be a scalar_int_mode.
9939 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9940 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9942 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9943 for GET_MODE_PRECISION.
9944 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9945 for GET_MODE_PRECISION.
9946 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9948 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9949 (expand_field_assignment, make_extraction): Likewise.
9950 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9951 (get_last_value): Likewise.
9952 * convert.c (convert_to_integer_1): Likewise.
9953 * cse.c (cse_insn): Likewise.
9954 * expr.c (expand_expr_real_1): Likewise.
9955 * lra-constraints.c (simplify_operand_subreg): Likewise.
9956 * optabs-query.c (can_atomic_load_p): Likewise.
9957 * optabs.c (expand_atomic_load): Likewise.
9958 (expand_atomic_store): Likewise.
9959 * ree.c (combine_reaching_defs): Likewise.
9960 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9961 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9962 * tree.h (type_has_mode_precision_p): Likewise.
9963 * ubsan.c (instrument_si_overflow): Likewise.
9965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9966 Alan Hayward <alan.hayward@arm.com>
9967 David Sherwood <david.sherwood@arm.com>
9969 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9970 polynomial numbers of units.
9971 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9972 (valid_vector_subparts_p): New function.
9973 (build_vector_type): Remove temporary shim and take the number
9974 of units as a poly_uint64 rather than an int.
9975 (build_opaque_vector_type): Take the number of units as a
9976 poly_uint64 rather than an int.
9977 * tree.c (build_vector_from_ctor): Handle polynomial
9978 TYPE_VECTOR_SUBPARTS.
9979 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9980 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9981 (build_vector_from_val): If the number of units is variable,
9982 use build_vec_duplicate_cst for constant operands and
9983 VEC_DUPLICATE_EXPR otherwise.
9984 (make_vector_type): Remove temporary is_constant ().
9985 (build_vector_type, build_opaque_vector_type): Take the number of
9986 units as a poly_uint64 rather than an int.
9987 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9989 * cfgexpand.c (expand_debug_expr): Likewise.
9990 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9991 (store_constructor, expand_expr_real_1): Likewise.
9992 (const_scalar_mask_from_tree): Likewise.
9993 * fold-const-call.c (fold_const_reduction): Likewise.
9994 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9995 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9996 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9997 (fold_relational_const): Likewise.
9998 (native_interpret_vector): Likewise. Change the size from an
9999 int to an unsigned int.
10000 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10001 TYPE_VECTOR_SUBPARTS.
10002 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10003 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10004 duplicating a non-constant operand into a variable-length vector.
10005 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10006 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10007 * ipa-icf.c (sem_variable::equals): Likewise.
10008 * match.pd: Likewise.
10009 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10010 * print-tree.c (print_node): Likewise.
10011 * stor-layout.c (layout_type): Likewise.
10012 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10013 * tree-cfg.c (verify_gimple_comparison): Likewise.
10014 (verify_gimple_assign_binary): Likewise.
10015 (verify_gimple_assign_ternary): Likewise.
10016 (verify_gimple_assign_single): Likewise.
10017 * tree-pretty-print.c (dump_generic_node): Likewise.
10018 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10019 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10020 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10021 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10022 (vect_shift_permute_load_chain): Likewise.
10023 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10024 (expand_vector_condition, optimize_vector_constructor): Likewise.
10025 (lower_vec_perm, get_compute_type): Likewise.
10026 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10027 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10028 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10029 (vect_recog_mask_conversion_pattern): Likewise.
10030 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10031 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10032 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10033 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10034 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10035 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10036 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10037 (supportable_widening_operation): Likewise.
10038 (supportable_narrowing_operation): Likewise.
10039 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10041 * varasm.c (output_constant): Likewise.
10043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10044 Alan Hayward <alan.hayward@arm.com>
10045 David Sherwood <david.sherwood@arm.com>
10047 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10048 so that both the length == 3 and length != 3 cases set up their
10049 own permute vectors. Add comments explaining why we know the
10050 number of elements is constant.
10051 (vect_permute_load_chain): Likewise.
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 * machmode.h (mode_nunits): Change from unsigned char to
10059 (ONLY_FIXED_SIZE_MODES): New macro.
10060 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10061 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10062 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10064 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10065 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10066 or if measurement_type is not polynomial.
10067 * genmodes.c (ZERO_COEFFS): New macro.
10068 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10070 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10071 Use ZERO_COEFFS when emitting initializers.
10072 * data-streamer.h (bp_pack_poly_value): New function.
10073 (bp_unpack_poly_value): Likewise.
10074 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10075 for GET_MODE_NUNITS.
10076 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10077 for GET_MODE_NUNITS.
10078 * tree.c (make_vector_type): Remove temporary shim and make
10079 the real function take the number of units as a poly_uint64
10080 rather than an int.
10081 (build_vector_type_for_mode): Handle polynomial nunits.
10082 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10083 * emit-rtl.c (const_vec_series_p_1): Likewise.
10084 (gen_rtx_CONST_VECTOR): Likewise.
10085 * fold-const.c (test_vec_duplicate_folding): Likewise.
10086 * genrecog.c (validate_pattern): Likewise.
10087 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10088 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10089 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10090 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10091 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10092 * rtlanal.c (subreg_get_info): Likewise.
10093 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10094 (vect_grouped_load_supported): Likewise.
10095 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10096 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10097 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10098 (simplify_const_unary_operation, simplify_binary_operation_1)
10099 (simplify_const_binary_operation, simplify_ternary_operation)
10100 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10101 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10102 instead of CONST_VECTOR_NUNITS.
10103 * varasm.c (output_constant_pool_2): Likewise.
10104 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10105 explicit-encoded elements in the XVEC for variable-length vectors.
10107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10109 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10112 Alan Hayward <alan.hayward@arm.com>
10113 David Sherwood <david.sherwood@arm.com>
10115 * coretypes.h (fixed_size_mode): Declare.
10116 (fixed_size_mode_pod): New typedef.
10117 * builtins.h (target_builtins::x_apply_args_mode)
10118 (target_builtins::x_apply_result_mode): Change type to
10119 fixed_size_mode_pod.
10120 * builtins.c (apply_args_size, apply_result_size, result_vector)
10121 (expand_builtin_apply_args_1, expand_builtin_apply)
10122 (expand_builtin_return): Update accordingly.
10124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10126 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10127 * cselib.c (cselib_hash_rtx): Likewise.
10128 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10129 CONST_VECTOR encoding.
10131 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10132 Jeff Law <law@redhat.com>
10135 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10136 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10137 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10138 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10141 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10142 explicitly probe *sp in a noreturn function if there were any callee
10143 register saves or frame pointer is needed.
10145 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10148 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10149 BLKmode for ternary, binary or unary expressions.
10152 * var-tracking.c (delete_vta_debug_insn): New inline function.
10153 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10154 insns from get_insns () to NULL instead of each bb separately.
10155 Use delete_vta_debug_insn. No longer static.
10156 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10157 delete_vta_debug_insns callers.
10158 * rtl.h (delete_vta_debug_insns): Declare.
10159 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10160 instead of variable_tracking_main.
10162 2018-01-03 Martin Sebor <msebor@redhat.com>
10164 PR tree-optimization/83603
10165 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10166 arguments past the endof the argument list in functions declared
10167 without a prototype.
10168 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10169 Avoid checking when arguments are null.
10171 2018-01-03 Martin Sebor <msebor@redhat.com>
10174 * doc/extend.texi (attribute const): Fix a typo.
10175 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10176 issuing -Wsuggest-attribute for void functions.
10178 2018-01-03 Martin Sebor <msebor@redhat.com>
10180 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10181 offset_int::from instead of wide_int::to_shwi.
10182 (maybe_diag_overlap): Remove assertion.
10183 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10184 * gimple-ssa-sprintf.c (format_directive): Same.
10185 (parse_directive): Same.
10186 (sprintf_dom_walker::compute_format_length): Same.
10187 (try_substitute_return_value): Same.
10189 2018-01-03 Jeff Law <law@redhat.com>
10191 PR middle-end/83654
10192 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10193 non-constant residual for zero at runtime and avoid probing in
10194 that case. Reorganize code for trailing problem to mirror handling
10197 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10199 PR tree-optimization/83501
10200 * tree-ssa-strlen.c (get_string_cst): New.
10201 (handle_char_store): Call get_string_cst.
10203 2018-01-03 Martin Liska <mliska@suse.cz>
10205 PR tree-optimization/83593
10206 * tree-ssa-strlen.c: Include tree-cfg.h.
10207 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10208 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10209 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10211 (strlen_dom_walker::before_dom_children): Call
10212 gimple_purge_dead_eh_edges. Dump tranformation with details
10214 (strlen_dom_walker::before_dom_children): Update call by adding
10215 new argument cleanup_eh.
10216 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10218 2018-01-03 Martin Liska <mliska@suse.cz>
10221 * cif-code.def (VARIADIC_THUNK): New enum value.
10222 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10225 2018-01-03 Jan Beulich <jbeulich@suse.com>
10227 * sse.md (mov<mode>_internal): Tighten condition for when to use
10228 vmovdqu<ssescalarsize> for TI and OI modes.
10230 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10232 Update copyright years.
10234 2018-01-03 Martin Liska <mliska@suse.cz>
10237 * ipa-visibility.c (function_and_variable_visibility): Skip
10238 functions with noipa attribure.
10240 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10242 * gcc.c (process_command): Update copyright notice dates.
10243 * gcov-dump.c (print_version): Ditto.
10244 * gcov.c (print_version): Ditto.
10245 * gcov-tool.c (print_version): Ditto.
10246 * gengtype.c (create_file): Ditto.
10247 * doc/cpp.texi: Bump @copying's copyright year.
10248 * doc/cppinternals.texi: Ditto.
10249 * doc/gcc.texi: Ditto.
10250 * doc/gccint.texi: Ditto.
10251 * doc/gcov.texi: Ditto.
10252 * doc/install.texi: Ditto.
10253 * doc/invoke.texi: Ditto.
10255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10257 * vector-builder.h (vector_builder::m_full_nelts): Change from
10258 unsigned int to poly_uint64.
10259 (vector_builder::full_nelts): Update prototype accordingly.
10260 (vector_builder::new_vector): Likewise.
10261 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10262 (vector_builder::operator ==): Likewise.
10263 (vector_builder::finalize): Likewise.
10264 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10265 Take the number of elements as a poly_uint64 rather than an
10267 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10268 from unsigned int to poly_uint64.
10269 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10270 (vec_perm_indices::new_vector): Likewise.
10271 (vec_perm_indices::length): Likewise.
10272 (vec_perm_indices::nelts_per_input): Likewise.
10273 (vec_perm_indices::input_nelts): Likewise.
10274 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10275 number of elements per input as a poly_uint64 rather than an
10276 unsigned int. Use the original encoding for variable-length
10277 vectors, rather than clamping each individual element.
10278 For the second and subsequent elements in each pattern,
10279 clamp the step and base before clamping their sum.
10280 (vec_perm_indices::series_p): Handle polynomial element counts.
10281 (vec_perm_indices::all_in_range_p): Likewise.
10282 (vec_perm_indices_to_tree): Likewise.
10283 (vec_perm_indices_to_rtx): Likewise.
10284 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10285 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10286 (tree_vector_builder::new_binary_operation): Handle polynomial
10287 element counts. Return false if we need to know the number
10288 of elements at compile time.
10289 * fold-const.c (fold_vec_perm): Punt if the number of elements
10290 isn't known at compile time.
10292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10294 * vec-perm-indices.h (vec_perm_builder): Change element type
10295 from HOST_WIDE_INT to poly_int64.
10296 (vec_perm_indices::element_type): Update accordingly.
10297 (vec_perm_indices::clamp): Handle polynomial element_types.
10298 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10299 (vec_perm_indices::all_in_range_p): Likewise.
10300 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10302 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10303 polynomial vec_perm_indices element types.
10304 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10305 * fold-const.c (fold_vec_perm): Likewise.
10306 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10307 * tree-vect-generic.c (lower_vec_perm): Likewise.
10308 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10309 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10310 element type to HOST_WIDE_INT.
10312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10313 Alan Hayward <alan.hayward@arm.com>
10314 David Sherwood <david.sherwood@arm.com>
10316 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10317 rather than an int. Use plus_constant.
10318 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10319 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10322 Alan Hayward <alan.hayward@arm.com>
10323 David Sherwood <david.sherwood@arm.com>
10325 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10326 a HOST_WIDE_INT to a poly_int64.
10328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10329 Alan Hayward <alan.hayward@arm.com>
10330 David Sherwood <david.sherwood@arm.com>
10332 * calls.c (load_register_parameters): Cope with polynomial
10333 mode sizes. Require a constant size for BLKmode parameters
10334 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10335 forces a parameter to be padded at the lsb end in order to
10336 fill a complete number of words, require the parameter size
10337 to be ordered wrt UNITS_PER_WORD.
10339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10340 Alan Hayward <alan.hayward@arm.com>
10341 David Sherwood <david.sherwood@arm.com>
10343 * reload1.c (spill_stack_slot_width): Change element type
10344 from unsigned int to poly_uint64_pod.
10345 (alter_reg): Treat mode sizes as polynomial.
10347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10348 Alan Hayward <alan.hayward@arm.com>
10349 David Sherwood <david.sherwood@arm.com>
10351 * reload.c (complex_word_subreg_p): New function.
10352 (reload_inner_reg_of_subreg, push_reload): Use it.
10354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10355 Alan Hayward <alan.hayward@arm.com>
10356 David Sherwood <david.sherwood@arm.com>
10358 * lra-constraints.c (process_alt_operands): Reject matched
10359 operands whose sizes aren't ordered.
10360 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10367 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10374 Use plus_constant instead of gen_rtx_PLUS.
10376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10377 Alan Hayward <alan.hayward@arm.com>
10378 David Sherwood <david.sherwood@arm.com>
10380 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10381 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10382 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10383 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10384 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10385 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10386 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10387 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10388 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10389 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10391 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10392 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10393 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10394 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10395 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10396 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10397 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10398 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10399 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10400 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10402 * expr.c (emit_move_resolve_push): Treat the input and result
10403 of PUSH_ROUNDING as a poly_int64.
10404 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10405 (emit_push_insn): Likewise.
10406 * lra-eliminations.c (mark_not_eliminable): Likewise.
10407 * recog.c (push_operand): Likewise.
10408 * reload1.c (elimination_effects): Likewise.
10409 * rtlanal.c (nonzero_bits1): Likewise.
10410 * calls.c (store_one_arg): Likewise. Require the padding to be
10411 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10418 Use plus_constant instead of gen_rtx_PLUS.
10420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10421 Alan Hayward <alan.hayward@arm.com>
10422 David Sherwood <david.sherwood@arm.com>
10424 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10425 rather than an int.
10427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10428 Alan Hayward <alan.hayward@arm.com>
10429 David Sherwood <david.sherwood@arm.com>
10431 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10432 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10433 via stack temporaries. Treat the mode size as polynomial too.
10435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10436 Alan Hayward <alan.hayward@arm.com>
10437 David Sherwood <david.sherwood@arm.com>
10439 * expr.c (expand_expr_real_2): When handling conversions involving
10440 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10441 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10442 as a poly_uint64 too.
10444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10445 Alan Hayward <alan.hayward@arm.com>
10446 David Sherwood <david.sherwood@arm.com>
10448 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10451 Alan Hayward <alan.hayward@arm.com>
10452 David Sherwood <david.sherwood@arm.com>
10454 * combine.c (can_change_dest_mode): Handle polynomial
10455 REGMODE_NATURAL_SIZE.
10456 * expmed.c (store_bit_field_1): Likewise.
10457 * expr.c (store_constructor): Likewise.
10458 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10459 and polynomial REGMODE_NATURAL_SIZE.
10460 (gen_lowpart_common): Likewise.
10461 * reginfo.c (record_subregs_of_mode): Likewise.
10462 * rtlanal.c (read_modify_subreg_p): Likewise.
10464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10465 Alan Hayward <alan.hayward@arm.com>
10466 David Sherwood <david.sherwood@arm.com>
10468 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10469 numbers of elements.
10471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10472 Alan Hayward <alan.hayward@arm.com>
10473 David Sherwood <david.sherwood@arm.com>
10475 * match.pd: Cope with polynomial numbers of vector elements.
10477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10478 Alan Hayward <alan.hayward@arm.com>
10479 David Sherwood <david.sherwood@arm.com>
10481 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10482 in a POINTER_PLUS_EXPR.
10484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10485 Alan Hayward <alan.hayward@arm.com>
10486 David Sherwood <david.sherwood@arm.com>
10488 * omp-simd-clone.c (simd_clone_subparts): New function.
10489 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10490 (ipa_simd_modify_function_body): Likewise.
10492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10493 Alan Hayward <alan.hayward@arm.com>
10494 David Sherwood <david.sherwood@arm.com>
10496 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10497 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10498 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10499 (expand_vector_condition, vector_element): Likewise.
10500 (subparts_gt): New function.
10501 (get_compute_type): Use subparts_gt.
10502 (count_type_subparts): Delete.
10503 (expand_vector_operations_1): Use subparts_gt instead of
10504 count_type_subparts.
10506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10507 Alan Hayward <alan.hayward@arm.com>
10508 David Sherwood <david.sherwood@arm.com>
10510 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10511 (vect_compile_time_alias): ...this new function. Do the calculation
10512 on poly_ints rather than trees.
10513 (vect_prune_runtime_alias_test_list): Update call accordingly.
10515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10516 Alan Hayward <alan.hayward@arm.com>
10517 David Sherwood <david.sherwood@arm.com>
10519 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10521 (vect_schedule_slp_instance): Likewise.
10523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10524 Alan Hayward <alan.hayward@arm.com>
10525 David Sherwood <david.sherwood@arm.com>
10527 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10528 constant and extern definitions for variable-length vectors.
10529 (vect_get_constant_vectors): Note that the number of units
10530 is known to be constant.
10532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10533 Alan Hayward <alan.hayward@arm.com>
10534 David Sherwood <david.sherwood@arm.com>
10536 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10537 of units as polynomial. Choose between WIDE and NARROW based
10540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10541 Alan Hayward <alan.hayward@arm.com>
10542 David Sherwood <david.sherwood@arm.com>
10544 * tree-vect-stmts.c (simd_clone_subparts): New function.
10545 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10548 Alan Hayward <alan.hayward@arm.com>
10549 David Sherwood <david.sherwood@arm.com>
10551 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10552 vectors as polynomial. Use build_index_vector for
10553 IFN_GOMP_SIMD_LANE.
10555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10556 Alan Hayward <alan.hayward@arm.com>
10557 David Sherwood <david.sherwood@arm.com>
10559 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10560 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10561 for variable-length vectors.
10562 (vectorizable_mask_load_store): Treat the number of units as
10563 polynomial, asserting that it is constant if the condition has
10564 already been enforced.
10565 (vectorizable_store, vectorizable_load): Likewise.
10567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10568 Alan Hayward <alan.hayward@arm.com>
10569 David Sherwood <david.sherwood@arm.com>
10571 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10572 of units as polynomial. Punt if we can't tell at compile time
10573 which vector contains the final result.
10575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10576 Alan Hayward <alan.hayward@arm.com>
10577 David Sherwood <david.sherwood@arm.com>
10579 * tree-vect-loop.c (vectorizable_induction): Treat the number
10580 of units as polynomial. Punt on SLP inductions. Use an integer
10581 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10582 cast of such a series for variable-length floating-point
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 * tree.h (build_index_vector): Declare.
10590 * tree.c (build_index_vector): New function.
10591 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10592 of units as polynomial, forcibly converting it to a constant if
10593 vectorizable_reduction has already enforced the condition.
10594 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10595 to create a {1,2,3,...} vector.
10596 (vectorizable_reduction): Treat the number of units as polynomial.
10597 Choose vectype_in based on the largest scalar element size rather
10598 than the smallest number of units. Enforce the restrictions
10601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10602 Alan Hayward <alan.hayward@arm.com>
10603 David Sherwood <david.sherwood@arm.com>
10605 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10606 number of units as polynomial.
10608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10609 Alan Hayward <alan.hayward@arm.com>
10610 David Sherwood <david.sherwood@arm.com>
10612 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10613 * target.def (autovectorize_vector_sizes): Return the vector sizes
10614 by pointer, using vector_sizes rather than a bitmask.
10615 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10616 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10617 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10619 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10620 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10621 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10622 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10623 * omp-general.c (omp_max_vf): Likewise.
10624 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10625 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10626 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10627 * tree-vect-slp.c (vect_slp_bb): Likewise.
10628 * doc/tm.texi: Regenerate.
10629 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10631 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10632 the vector size as a poly_uint64 rather than an unsigned int.
10633 (current_vector_size): Change from an unsigned int to a poly_uint64.
10634 (get_vectype_for_scalar_type): Update accordingly.
10635 * tree.h (build_truth_vector_type): Take the size and number of
10636 units as a poly_uint64 rather than an unsigned int.
10637 (build_vector_type): Add a temporary overload that takes
10638 the number of units as a poly_uint64 rather than an unsigned int.
10639 * tree.c (make_vector_type): Likewise.
10640 (build_truth_vector_type): Take the number of units as a poly_uint64
10641 rather than an unsigned int.
10643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10644 Alan Hayward <alan.hayward@arm.com>
10645 David Sherwood <david.sherwood@arm.com>
10647 * target.def (get_mask_mode): Take the number of units and length
10648 as poly_uint64s rather than unsigned ints.
10649 * targhooks.h (default_get_mask_mode): Update accordingly.
10650 * targhooks.c (default_get_mask_mode): Likewise.
10651 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10652 * doc/tm.texi: Regenerate.
10654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10655 Alan Hayward <alan.hayward@arm.com>
10656 David Sherwood <david.sherwood@arm.com>
10658 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10659 * omp-general.c (omp_max_vf): Likewise.
10660 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10661 (expand_omp_simd): Handle polynomial safelen.
10662 * omp-low.c (omplow_simd_context): Add a default constructor.
10663 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10664 (lower_rec_simd_input_clauses): Update accordingly.
10665 (lower_rec_input_clauses): Likewise.
10667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10668 Alan Hayward <alan.hayward@arm.com>
10669 David Sherwood <david.sherwood@arm.com>
10671 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10672 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10673 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10674 (vect_analyze_slp_cost): Likewise.
10675 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10676 (vect_model_load_cost): Likewise.
10678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10679 Alan Hayward <alan.hayward@arm.com>
10680 David Sherwood <david.sherwood@arm.com>
10682 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10683 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10684 from an unsigned int * to a poly_uint64_pod *.
10685 (calculate_unrolling_factor): New function.
10686 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10689 Alan Hayward <alan.hayward@arm.com>
10690 David Sherwood <david.sherwood@arm.com>
10692 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10693 from an unsigned int to a poly_uint64.
10694 (_loop_vec_info::slp_unrolling_factor): Likewise.
10695 (_loop_vec_info::vectorization_factor): Change from an int
10697 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10698 (vect_get_num_vectors): New function.
10699 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10700 (vect_get_num_copies): Use vect_get_num_vectors.
10701 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10702 to an unsigned int *.
10703 (vect_analyze_data_refs): Change min_vf from an int * to a
10705 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10706 than an unsigned HOST_WIDE_INT.
10707 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10708 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10709 to an unsigned int *.
10710 (vect_analyze_data_ref_dependences): Likewise.
10711 (vect_compute_data_ref_alignment): Handle polynomial vf.
10712 (vect_enhance_data_refs_alignment): Likewise.
10713 (vect_prune_runtime_alias_test_list): Likewise.
10714 (vect_shift_permute_load_chain): Likewise.
10715 (vect_supportable_dr_alignment): Likewise.
10716 (dependence_distance_ge_vf): Take the vectorization factor as a
10717 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10718 (vect_analyze_data_refs): Change min_vf from an int * to a
10720 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10721 vfm1 as a poly_uint64 rather than an int. Make the same change
10722 for the returned bound_scalar.
10723 (vect_gen_vector_loop_niters): Handle polynomial vf.
10724 (vect_do_peeling): Likewise. Update call to
10725 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10726 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10728 * tree-vect-loop.c (vect_determine_vectorization_factor)
10729 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10730 (vect_get_known_peeling_cost): Likewise.
10731 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10732 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10733 (vect_transform_loop): Likewise. Use the lowest possible VF when
10734 updating the upper bounds of the loop.
10735 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10736 rather than an int.
10737 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10738 polynomial unroll factors.
10739 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10740 (vect_make_slp_decision): Likewise.
10741 (vect_supported_load_permutation_p): Likewise, and polynomial
10743 (vect_analyze_slp_cost): Handle polynomial vf.
10744 (vect_slp_analyze_node_operations): Likewise.
10745 (vect_slp_analyze_bb_1): Likewise.
10746 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10747 than an unsigned HOST_WIDE_INT.
10748 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10749 (vectorizable_load): Handle polynomial vf.
10750 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10752 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10755 Alan Hayward <alan.hayward@arm.com>
10756 David Sherwood <david.sherwood@arm.com>
10758 * match.pd: Handle bit operations involving three constants
10759 and try to fold one pair.
10761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10763 * tree-vect-loop-manip.c: Include gimple-fold.h.
10764 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10765 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10766 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10767 Add a path that uses a step of VF instead of 1, but disable it
10769 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10770 and niters_no_overflow parameters. Update calls to
10771 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10772 Create a new SSA name if the latter choses to use a ste other
10773 than zero, and return it via niters_vector_mult_vf_var.
10774 * tree-vect-loop.c (vect_transform_loop): Update calls to
10775 vect_do_peeling, vect_gen_vector_loop_niters and
10776 slpeel_make_loop_iterate_ntimes.
10777 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10778 (vect_gen_vector_loop_niters): Update declarations after above changes.
10780 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10782 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10783 128-bit round to integer instructions.
10784 (ceil<mode>2): Likewise.
10785 (btrunc<mode>2): Likewise.
10786 (round<mode>2): Likewise.
10788 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10790 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10791 unaligned VSX load/store on P8/P9.
10792 (expand_block_clear): Allow the use of unaligned VSX
10793 load/store on P8/P9.
10795 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10797 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10799 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10800 swap associated with both a load and a store.
10802 2018-01-02 Andrew Waterman <andrew@sifive.com>
10804 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10805 * config/riscv/riscv.md (clear_cache): Use it.
10807 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10809 * web.c: Remove out-of-date comment.
10811 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10813 * expr.c (fixup_args_size_notes): Check that any existing
10814 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10815 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10816 (emit_single_push_insn): ...here.
10818 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10820 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10821 (const_vector_encoded_nelts): New function.
10822 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10823 (const_vector_int_elt, const_vector_elt): Declare.
10824 * emit-rtl.c (const_vector_int_elt_1): New function.
10825 (const_vector_elt): Likewise.
10826 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10827 of CONST_VECTOR_ELT.
10829 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10831 * expr.c: Include rtx-vector-builder.h.
10832 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10833 directly on the tree encoding.
10834 (const_vector_from_tree): Likewise.
10835 * optabs.c: Include rtx-vector-builder.h.
10836 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10837 sequence of "u" values.
10838 * vec-perm-indices.c: Include rtx-vector-builder.h.
10839 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10840 directly on the vec_perm_indices encoding.
10842 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10844 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10845 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10846 * rtx-vector-builder.h: New file.
10847 * rtx-vector-builder.c: Likewise.
10848 * rtl.h (rtx_def::u2): Add a const_vector field.
10849 (CONST_VECTOR_NPATTERNS): New macro.
10850 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10851 (CONST_VECTOR_DUPLICATE_P): Likewise.
10852 (CONST_VECTOR_STEPPED_P): Likewise.
10853 (CONST_VECTOR_ENCODED_ELT): Likewise.
10854 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10855 (unwrap_const_vec_duplicate): Likewise.
10856 (const_vec_series_p): Check for a non-duplicated vector encoding.
10857 Say that the function only returns true for integer vectors.
10858 * emit-rtl.c: Include rtx-vector-builder.h.
10859 (gen_const_vec_duplicate_1): Delete.
10860 (gen_const_vector): Call gen_const_vec_duplicate instead of
10861 gen_const_vec_duplicate_1.
10862 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10863 (gen_const_vec_duplicate): Use rtx_vector_builder.
10864 (gen_const_vec_series): Likewise.
10865 (gen_rtx_CONST_VECTOR): Likewise.
10866 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10867 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10868 Build a new vector rather than modifying a CONST_VECTOR in-place.
10869 (handle_special_swappables): Update call accordingly.
10870 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10871 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10872 Build a new vector rather than modifying a CONST_VECTOR in-place.
10873 (handle_special_swappables): Update call accordingly.
10875 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10877 * simplify-rtx.c (simplify_const_binary_operation): Use
10878 CONST_VECTOR_ELT instead of XVECEXP.
10880 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10882 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10883 the selector elements to be different from the data elements
10884 if the selector is a VECTOR_CST.
10885 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10886 ssizetype for the selector.
10888 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10890 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10891 before testing each element individually.
10892 * tree-vect-generic.c (lower_vec_perm): Likewise.
10894 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10896 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10897 * selftest-run-tests.c (selftest::run_tests): Call it.
10898 * vector-builder.h (vector_builder::operator ==): New function.
10899 (vector_builder::operator !=): Likewise.
10900 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10901 (vec_perm_indices::all_from_input_p): New function.
10902 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10903 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10904 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10905 instead of reading the VECTOR_CST directly. Detect whether both
10906 vector inputs are the same before constructing the vec_perm_indices,
10907 and update the number of inputs argument accordingly. Use the
10908 utility functions added above. Only construct sel2 if we need to.
10910 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10912 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10913 the broadcast of the low byte.
10914 (expand_mult_highpart): Use an explicit encoding for the permutes.
10915 * optabs-query.c (can_mult_highpart_p): Likewise.
10916 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10917 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10918 (vectorizable_bswap): Likewise.
10919 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10920 explicit encoding for the power-of-2 permutes.
10921 (vect_permute_store_chain): Likewise.
10922 (vect_grouped_load_supported): Likewise.
10923 (vect_permute_load_chain): Likewise.
10925 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10927 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10928 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10929 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10930 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10931 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10932 (vect_gen_perm_mask_any): Likewise.
10934 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10936 * int-vector-builder.h: New file.
10937 * vec-perm-indices.h: Include int-vector-builder.h.
10938 (vec_perm_indices): Redefine as an int_vector_builder.
10939 (auto_vec_perm_indices): Delete.
10940 (vec_perm_builder): Redefine as a stand-alone class.
10941 (vec_perm_indices::vec_perm_indices): New function.
10942 (vec_perm_indices::clamp): Likewise.
10943 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10944 (vec_perm_indices::new_vector): New function.
10945 (vec_perm_indices::new_expanded_vector): Update for new
10946 vec_perm_indices class.
10947 (vec_perm_indices::rotate_inputs): New function.
10948 (vec_perm_indices::all_in_range_p): Operate directly on the
10949 encoded form, without computing elided elements.
10950 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10951 encoding. Update for new vec_perm_indices class.
10952 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10953 the given vec_perm_builder.
10954 (expand_vec_perm_var): Update vec_perm_builder constructor.
10955 (expand_mult_highpart): Use vec_perm_builder instead of
10956 auto_vec_perm_indices.
10957 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10958 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10959 or double series encoding as appropriate.
10960 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10961 vec_perm_indices instead of auto_vec_perm_indices.
10962 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10963 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10964 (vect_permute_store_chain): Likewise.
10965 (vect_grouped_load_supported): Likewise.
10966 (vect_permute_load_chain): Likewise.
10967 (vect_shift_permute_load_chain): Likewise.
10968 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10969 (vect_transform_slp_perm_load): Likewise.
10970 (vect_schedule_slp_instance): Likewise.
10971 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10972 (vectorizable_mask_load_store): Likewise.
10973 (vectorizable_bswap): Likewise.
10974 (vectorizable_store): Likewise.
10975 (vectorizable_load): Likewise.
10976 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10977 vec_perm_indices instead of auto_vec_perm_indices. Use
10978 tree_to_vec_perm_builder to read the vector from a tree.
10979 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10980 vec_perm_builder instead of a vec_perm_indices.
10981 (have_whole_vector_shift): Use vec_perm_builder and
10982 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10983 truncation to calc_vec_perm_mask_for_shift.
10984 (vect_create_epilog_for_reduction): Likewise.
10985 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10986 from auto_vec_perm_indices to vec_perm_indices.
10987 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10988 instead of changing individual elements.
10989 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10990 the vector in d.perm.
10991 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10992 from auto_vec_perm_indices to vec_perm_indices.
10993 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10994 instead of changing individual elements.
10995 (arm_vectorize_vec_perm_const): Use new_vector to install
10996 the vector in d.perm.
10997 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10998 Update vec_perm_builder constructor.
10999 (rs6000_expand_interleave): Likewise.
11000 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11001 (rs6000_expand_interleave): Likewise.
11003 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11005 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11006 to qimode could truncate the indices.
11007 * optabs.c (expand_vec_perm_var): Likewise.
11009 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11011 * Makefile.in (OBJS): Add vec-perm-indices.o.
11012 * vec-perm-indices.h: New file.
11013 * vec-perm-indices.c: Likewise.
11014 * target.h (vec_perm_indices): Replace with a forward class
11016 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11017 * optabs.h: Include vec-perm-indices.h.
11018 (expand_vec_perm): Delete.
11019 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11020 (expand_vec_perm_const): Declare.
11021 * target.def (vec_perm_const_ok): Replace with...
11022 (vec_perm_const): ...this new hook.
11023 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11024 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11025 * doc/tm.texi: Regenerate.
11026 * optabs.def (vec_perm_const): Delete.
11027 * doc/md.texi (vec_perm_const): Likewise.
11028 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11029 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11030 expand_vec_perm for constant permutation vectors. Assert that
11031 the mode of variable permutation vectors is the integer equivalent
11032 of the mode that is being permuted.
11033 * optabs-query.h (selector_fits_mode_p): Declare.
11034 * optabs-query.c: Include vec-perm-indices.h.
11035 (selector_fits_mode_p): New function.
11036 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11037 is defined, instead of checking whether the vec_perm_const_optab
11038 exists. Use targetm.vectorize.vec_perm_const instead of
11039 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11040 fit in the vector mode before using a variable permute.
11041 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11042 vec_perm_indices instead of an rtx.
11043 (expand_vec_perm): Replace with...
11044 (expand_vec_perm_const): ...this new function. Take the selector
11045 as a vec_perm_indices rather than an rtx. Also take the mode of
11046 the selector. Update call to shift_amt_for_vec_perm_mask.
11047 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11048 Use vec_perm_indices::new_expanded_vector to expand the original
11049 selector into bytes. Check whether the indices fit in the vector
11050 mode before using a variable permute.
11051 (expand_vec_perm_var): Make global.
11052 (expand_mult_highpart): Use expand_vec_perm_const.
11053 * fold-const.c: Includes vec-perm-indices.h.
11054 * tree-ssa-forwprop.c: Likewise.
11055 * tree-vect-data-refs.c: Likewise.
11056 * tree-vect-generic.c: Likewise.
11057 * tree-vect-loop.c: Likewise.
11058 * tree-vect-slp.c: Likewise.
11059 * tree-vect-stmts.c: Likewise.
11060 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11062 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11063 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11064 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11065 (aarch64_vectorize_vec_perm_const): ...this new function.
11066 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11067 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11068 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11069 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11070 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11071 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11072 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11074 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11075 check for NEON modes.
11076 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11077 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11078 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11079 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11081 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11082 the old VEC_PERM_CONST conditions.
11083 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11084 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11085 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11086 (ia64_vectorize_vec_perm_const_ok): Merge into...
11087 (ia64_vectorize_vec_perm_const): ...this new function.
11088 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11089 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11090 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11091 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11092 * config/mips/mips.c (mips_expand_vec_perm_const)
11093 (mips_vectorize_vec_perm_const_ok): Merge into...
11094 (mips_vectorize_vec_perm_const): ...this new function.
11095 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11096 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11097 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11098 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11099 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11100 (rs6000_expand_vec_perm_const): Delete.
11101 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11103 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11104 (altivec_expand_vec_perm_const_le): Take each operand individually.
11105 Operate on constant selectors rather than rtxes.
11106 (altivec_expand_vec_perm_const): Likewise. Update call to
11107 altivec_expand_vec_perm_const_le.
11108 (rs6000_expand_vec_perm_const): Delete.
11109 (rs6000_vectorize_vec_perm_const_ok): Delete.
11110 (rs6000_vectorize_vec_perm_const): New function.
11111 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11112 an element count and rtx array.
11113 (rs6000_expand_extract_even): Update call accordingly.
11114 (rs6000_expand_interleave): Likewise.
11115 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11116 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11117 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11118 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11119 (rs6000_expand_vec_perm_const): Delete.
11120 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11121 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11122 (altivec_expand_vec_perm_const_le): Take each operand individually.
11123 Operate on constant selectors rather than rtxes.
11124 (altivec_expand_vec_perm_const): Likewise. Update call to
11125 altivec_expand_vec_perm_const_le.
11126 (rs6000_expand_vec_perm_const): Delete.
11127 (rs6000_vectorize_vec_perm_const_ok): Delete.
11128 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11129 reference to the SPE evmerge intructions.
11130 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11131 an element count and rtx array.
11132 (rs6000_expand_extract_even): Update call accordingly.
11133 (rs6000_expand_interleave): Likewise.
11134 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11135 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11137 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11139 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11141 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11142 vector mode and that that mode matches the mode of the data
11144 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11145 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11146 directly using expand_vec_perm_1 when forcing selectors into
11148 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11150 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11152 * optabs-query.h (can_vec_perm_p): Delete.
11153 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11154 * optabs-query.c (can_vec_perm_p): Split into...
11155 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11156 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11157 particular selector is valid.
11158 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11159 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11160 (vect_grouped_load_supported): Likewise.
11161 (vect_shift_permute_load_chain): Likewise.
11162 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11163 (vect_transform_slp_perm_load): Likewise.
11164 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11165 (vectorizable_bswap): Likewise.
11166 (vect_gen_perm_mask_checked): Likewise.
11167 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11168 implementations of variable permutation vectors into account
11169 when deciding which selector to use.
11170 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11171 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11172 with a false third argument.
11173 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11174 to test whether the constant selector is valid and can_vec_perm_var_p
11175 to test whether a variable selector is valid.
11177 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11179 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11180 * optabs-query.c (can_vec_perm_p): Likewise.
11181 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11182 instead of vec_perm_indices.
11183 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11184 (vect_gen_perm_mask_checked): Likewise,
11185 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11186 (vect_gen_perm_mask_checked): Likewise,
11188 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11190 * optabs-query.h (qimode_for_vec_perm): Declare.
11191 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11192 (qimode_for_vec_perm): ...this new function.
11193 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11195 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11197 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11198 does not have a conditional at the top.
11200 2018-01-02 Richard Biener <rguenther@suse.de>
11202 * ipa-inline.c (big_speedup_p): Fix expression.
11204 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11207 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11210 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11214 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11215 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11216 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11217 cond_taken_branch_cost 3->4.
11219 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11221 PR tree-optimization/83581
11222 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11223 TODO_cleanup_cfg if any changes have been made.
11225 PR middle-end/83608
11226 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11227 convert_modes if target mode has the right side, but different mode
11230 PR middle-end/83609
11231 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11232 last argument when extracting from CONCAT. If either from_real or
11233 from_imag is NULL, use expansion through memory. If result is not
11234 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11235 the parts directly to inner mode, if even that fails, use expansion
11238 PR middle-end/83623
11239 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11240 check for bswap in mode rather than HImode and use that in expand_unop
11243 Copyright (C) 2018 Free Software Foundation, Inc.
11245 Copying and distribution of this file, with or without modification,
11246 are permitted in any medium without royalty provided the copyright
11247 notice and this notice are preserved.