1 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3 * config/nds32/nds32-doubleword.md: New define_split pattern for
4 illegal register number.
6 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
10 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
14 2018-04-25 Richard Biener <rguenther@suse.de>
16 * lto-streamer.h (LTO_major_version): Bump to 8.
18 2018-04-25 Jakub Jelinek <jakub@redhat.com>
20 * BASE-VER: Set to 9.0.0.
22 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
24 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
25 in __abskf2 and __powikf2.
27 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30 * config/aarch64/constraints.md (Usg, Usj): New constraints.
31 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
32 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
33 Use the above on operand 2. Reindent.
34 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
36 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
39 * common/config/i386/i386-common.c (ix86_handle_option): Don't
41 * config/i386/i386.opt (mcet): Removed.
42 * doc/install.texi: Remove -mcet documentation.
43 * doc/invoke.texi: Likewise.
45 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
48 * doc/install.texi: Remove -mcet from bootstrap-cet.
50 2018-04-24 Jakub Jelinek <jakub@redhat.com>
53 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
54 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
58 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
59 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
60 containing a CONST_VECTOR.
62 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
64 * doc/install.texi: Update newlib dependency for nvptx.
66 2018-04-24 Jakub Jelinek <jakub@redhat.com>
69 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
70 instead of INTVAL when shifting x left.
72 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
74 PR tree-optimization/85478
75 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
76 vect_grouped_store_supported for single element vectors.
78 2018-04-24 Richard Biener <rguenther@suse.de>
81 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
82 load cost increase to the case of non-constant step.
84 2018-04-24 Jakub Jelinek <jakub@redhat.com>
87 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
88 destination if any_malformed_asm.
90 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
93 * expr.c (store_field): In the bitfield case, if the value comes from
94 a function call and is returned in registers by means of a PARALLEL,
95 do not change the mode of the temporary unless BLKmode and VOIDmode.
97 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
99 PR rtl-optimization/85423
100 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
101 dependencies to debug insns when the previous insn is non-debug.
103 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
105 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
106 enums into a single definition.
107 (fls): Fix predicates and printing.
110 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
112 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
113 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
114 and short u6 immediate.
115 (check_if_valid_sleep_operand): Remove.
116 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
118 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
120 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
121 flag_always_save_lp condition.
122 * config/nds32/nds32.opt (malways-save-lp): New option.
124 2018-04-22 Shiva Chen <shiva0217@gmail.com>
126 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
127 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
128 * config/nds32/nds32.h
129 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
130 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
132 2018-04-22 Shiva Chen <shiva0217@gmail.com>
134 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
135 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
137 2018-04-22 Shiva Chen <shiva0217@gmail.com>
138 Chung-Ju Wu <jasonwucj@gmail.com>
140 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
142 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
143 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
145 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
147 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
149 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
151 * config/nds32/nds32-protos.h (nds32_data_alignment,
152 nds32_local_alignment): Declare.
153 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
154 nds32_local_alignment): New functions.
155 (TARGET_CONSTANT_ALIGNMENT): Define.
156 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
158 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
160 * config/nds32/nds32.c
161 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
162 (TARGET_MODES_TIEABLE_P): Likewise.
164 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
166 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
169 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
170 Chung-Ju Wu <jasonwucj@gmail.com>
172 * config/nds32/constants.md (unspec_volatile_element): Add enum values
173 for unaligned access.
174 * config/nds32/nds32-intrinsic.c: Implementation of expanding
176 * config/nds32/nds32-intrinsic.md: Likewise.
177 * config/nds32/nds32_intrinsic.h: Likewise.
178 * config/nds32/nds32.h (nds32_builtins): Likewise.
179 * config/nds32/nds32.opt (munaligned-access): New option.
180 * config/nds32/nds32.c (nds32_asm_file_start): Display
181 flag_unaligned_access status.
183 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
185 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
186 -mno-relax is present.
187 * config/riscv/linux.h (LINK_SPEC): Ditto.
189 2018-04-20 Martin Sebor <msebor@redhat.com>
192 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
194 (gimple_fold_builtin_stxcpy_chk): Same.
195 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
197 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
200 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
201 __powikf2 when long double is IEEE 128-bit.
203 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
205 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
206 step to make sure stack always aligned.
208 2018-04-20 Carl Love <cel@us.ibm.com>
211 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
214 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
215 Tom de Vries <tom@codesourcery.com>
218 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
219 Emit insns for calls too.
220 (nvptx_find_par): Always look for worker-level predecessor insn.
221 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
223 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
224 (nvptx_process_pars): Propagate frames for calls.
226 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
229 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
231 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
232 (ix86_handle_option): Don't handle OPT_mibt.
233 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
235 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
237 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
238 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
239 (ix86_target_macros): Define __CET__ with flag_cf_protection
241 * config/i386/i386.c (isa2_opts): Remove -mibt.
242 * config/i386/i386.h (TARGET_IBT): Removed.
243 (TARGET_IBT_P): Likewise.
244 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
245 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
246 * config/i386/i386.opt (mcet): Update help message.
249 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
250 -mcet as an alias for -mshstk.
252 2018-04-20 Richard Biener <rguenther@suse.de>
255 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
256 complexity by forcing a single use of the multiply operand.
258 2018-04-20 Martin Jambor <mjambor@suse.cz>
261 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
262 recursion dependency to only apply to non-clones.
264 2018-04-20 Martin Jambor <mjambor@suse.cz>
267 * ipa-cp.c (create_specialized_node): Check that clones of
268 self-recursive edges exist during IPA-CP.
270 2018-04-19 Toon Moene <toon@moene.org>
272 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
275 2018-04-19 Jakub Jelinek <jakub@redhat.com>
277 PR tree-optimization/85467
278 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
279 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
280 VECTOR_CST element to type.
282 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
285 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
286 * config/i386/i386.md (builtin_setjmp_setup): Removed.
287 (builtin_longjmp): Likewise.
288 (save_stack_nonlocal): New pattern.
289 (restore_stack_nonlocal): Likewise.
291 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
294 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
295 Replace ASM_OUTPUT_LABEL with fprintf.
297 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
300 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
301 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
302 * config/i386/i386-c.c (ix86_target_macros_internal): Also
303 define __IBT__ and __SHSTK__ for -fcf-protection.
304 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
306 (ix86_trampoline_init): Likewise.
307 (x86_output_mi_thunk): Likewise.
308 (ix86_notrack_prefixed_insn_p): Likewise.
309 (ix86_option_override_internal): Don't disallow -fcf-protection.
310 * config/i386/i386.md (rdssp<mode>): Also enable for
312 (incssp<mode>): Likewise.
313 (nop_endbr): Likewise.
314 * config/i386/i386.opt (mcet): Change help message to built-in
318 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
319 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
320 enable CET built-in functions.
322 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
324 * common/config/i386/i386-common.c
325 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
326 OPTION_MASK_ISA_MOVDIRI_UNSET,
327 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
328 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
329 * config.gcc (movdirintrin.h): New header.
330 * config/i386/cpuid.h (bit_MOVDIRI,
331 bit_MOVDIR64B): New bits.
332 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
334 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
335 (VOID, PVOID, PCVOID)): New function types.
336 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
337 __builtin_ia32_directstoreu_u64,
338 __builtin_ia32_movdir64b): New builtins.
339 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
340 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
342 (ix86_valid_target_attribute_inner_p): Ditto.
343 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
344 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
345 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
346 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
347 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
348 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
349 (movdiri<mode>, movdir64b_<mode>): New.
350 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
351 * config/i386/immintrin.h: Include movdirintrin.h.
352 * config/i386/movdirintrin.h: New file.
353 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
355 2018-04-19 Richard Biener <rguenther@suse.de>
358 * cfg.c (clear_bb_flags): When loop state says we have
359 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
361 2018-04-19 Richard Biener <rguenther@suse.de>
363 PR tree-optimization/84737
364 * tree-vect-data-refs.c (vect_copy_ref_info): New function
365 copying restrict info.
366 (vect_setup_realignment): Use it.
367 * tree-vectorizer.h (vect_copy_ref_info): Declare.
368 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
369 the first DR to all generated stores.
370 (vectorizable_load): Likewise for loads.
372 2018-04-19 Jakub Jelinek <jakub@redhat.com>
374 PR tree-optimization/85446
375 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
376 the integral and pointer types to have the same precision.
378 * doc/install.texi: Document --disable-cet being the default and
381 2018-04-18 Martin Liska <mliska@suse.cz>
383 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
386 2018-04-18 Martin Liska <mliska@suse.cz>
389 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
392 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
393 arguments if they are comparable.
395 2018-04-18 Martin Liska <mliska@suse.cz>
398 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
401 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
404 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
407 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
408 ENDBR after calling __morestack.
410 2018-04-18 David Malcolm <dmalcolm@redhat.com>
413 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
414 by using gcc_base_ver to generate a gcc_driver_version, and use
415 it when generating GCC_DRIVER_NAME.
416 * configure: Regenerate.
418 2018-04-18 Jakub Jelinek <jakub@redhat.com>
421 * config.gcc: Obsolete powerpc*-*-*spe*.
423 2018-04-17 Jakub Jelinek <jakub@redhat.com>
426 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
427 (stabstr_D): Change type of unum from unsigned int to
428 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
431 2018-04-17 Jim Wilson <jimw@sifive.com>
434 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
435 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
436 Set arg_pointer_offset after using pretend_args_size.
438 2018-04-17 Jakub Jelinek <jakub@redhat.com>
440 PR rtl-optimization/85431
441 * dse.c (record_store): Ignore zero width stores.
444 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
445 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
446 __builtin_stack_restore rather than after it.
447 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
448 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
449 argument instead of virtual_dynamic_stack_rtx.
451 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
453 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
455 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
456 Add note to error message to explain internal mapping of overloaded
457 built-in function name to non-overloaded built-in function name.
458 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
461 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
464 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
465 where the inputs overlap with the output.
467 2018-04-17 Jakub Jelinek <jakub@redhat.com>
469 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
470 (=v, v) alternative and explicit "memory" attribute.
471 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
472 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
474 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
475 "sselog1" type instead of "sselog".
476 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
477 "sselog". Remove explicit "memory" attribute.
478 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
479 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
481 (vec_extract_hi_v32hi): Merge all alternatives into one, use
482 "sselog1" type instead of "sselog". Remove explicit "memory"
484 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
485 use "sselog1" type instead of "sselog". Remove explicit "memory"
487 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
488 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
490 (vec_extract_hi_v64qi): Merge all alternatives into one, use
491 "sselog1" type instead of "sselog". Remove explicit "memory"
493 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
494 use "sselog1" type instead of "sselog". Remove explicit "memory"
498 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
501 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
504 2018-04-17 Martin Jambor <mjambor@suse.cz>
507 * ipa-cp.c (create_specialized_node): Call
508 expand_all_artificial_thunks if necessary.
510 2018-04-17 Martin Liska <mliska@suse.cz>
513 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
514 in message, remote space in between '_G' and '('.
516 2018-04-17 Jakub Jelinek <jakub@redhat.com>
519 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
520 avx512f_vmcmp<mode>3<round_saeonly_name>,
521 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
522 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
523 avx512f_rndscale<mode><round_saeonly_name>,
524 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
525 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
526 Use %<iptr>2 instead of %2 for -masm=intel.
527 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
528 avx512f_vcvttss2usi<round_saeonly_name>,
529 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
531 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
532 avx512f_vcvttsd2usi<round_saeonly_name>,
533 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
534 Use %q1 instead of %1 for -masm=intel.
535 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
536 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
537 of %3 for -masm=intel.
538 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
540 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
542 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
544 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
545 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
547 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
549 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
550 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
551 %g1 and one with %0 and %1.
552 (avx512er_vmrcp28<mode><round_saeonly_name>,
553 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
555 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
556 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
557 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
558 of %0 and %{%4%} for -masm=intel.
559 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
560 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
561 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
562 order of %0 and %{%5%}%{z%} for -masm=intel.
564 2018-04-17 Jan Hubicka <jh@suse.cz>
567 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
569 2018-04-17 Martin Liska <mliska@suse.cz>
572 * multiple_target.c (create_dispatcher_calls): Set apostrophes
573 for target_clone error message. Make default implementation
574 clone to be a local declaration.
575 (separate_attrs): Add new argument and check for an empty
577 (expand_target_clones): Handle it.
578 (ipa_target_clone): Make redirection just for target_clones
581 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
582 Tom de Vries <tom@codesourcery.com>
585 * omp-expand.c (expand_oacc_for): Add dummy false branch for
586 tiled basic blocks without omp continue statements.
588 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
591 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
592 vec_extract expression as having side effects to make sure it gets
595 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
598 * config/i386/i386.c (get_builtin_code_for_version): Check
601 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
604 * gcc/config.gcc: Support "skylake".
605 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
607 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
608 (processor_target_table): Add "skylake".
609 (ix86_option_override_internal): Add "skylake".
610 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
611 PROCESSOR_CANNONLAKE.
612 (get_builtin_code_for_version): Fix priority for
613 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
614 PROCESSOR_SKYLAKE-AVX512.
615 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
616 (processor_type): Add PROCESSOR_SKYLAKE.
618 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
619 Jason Merrill <jason@redhat.com>
622 * convert.c (convert_to_integer_1): Use direct recursion for
623 enumeral types and types with a precision less than the number
624 of bits in their mode.
626 2018-04-16 Julia Koval <julia.koval@intel.com>
629 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
630 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
632 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
635 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
636 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
637 and -mno-direct-move.
639 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
642 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
643 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
644 Ensure negative shifts result in {0}.
646 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
648 PR rtl-optimization/79916
649 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
650 regs (if any) to define how to gnerate SD moves when LRA is in
653 2018-04-13 Jakub Jelinek <jakub@redhat.com>
655 PR rtl-optimization/85393
656 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
657 * except.c (expand_dw2_landing_pad_for_region): Make static.
658 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
659 a label and unconditional jump to old_bb, rather than
660 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
663 PR rtl-optimization/85376
664 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
665 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
666 instead of a specific value.
668 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
669 Bin Cheng <bin.cheng@arm.com>
671 PR tree-optimization/82965
672 PR tree-optimization/83991
673 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
674 by_profile_only parameter.
675 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
676 information if the loop was predicted to iterate too many times.
677 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
679 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
682 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
685 2018-04-13 Martin Liska <mliska@suse.cz>
686 Jakub Jelinek <jakub@redhat.com>
689 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
690 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
691 * builtins.c (expand_builtin_memory_copy_args): Use
692 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
693 handle dest_addr == pc_rtx.
695 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
698 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
699 asked to not generate direct moves.
700 (fix_trunc<mode>si2_stfiwx): Similar.
701 (fix_trunc<mode>si2_internal): Similar.
703 2018-04-12 Jakub Jelinek <jakub@redhat.com>
706 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
707 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
708 lookup if dest in some wider mode is known to be const0_rtx and
709 if so, record permanent equivalence for it to be ZERO_EXTEND of
710 the narrower mode destination.
712 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
714 * lto-streamer-out.c (output_function): Revert 259346.
715 * omp-expand.c (expand_oacc_for): Likewise.
717 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
719 PR rtl-optimization/85354
720 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
721 * sel-sched.c (sel_global_init): ... here.
723 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
726 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
727 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
728 mode for PE-COFF targets.
729 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
730 (i386_pe_asm_lto_end): Likewise.
731 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
732 (TARGET_ASM_LTO_END): Likewise.
733 * config/i386/winnt.c (saved_debug_info_level): New static variable.
734 (i386_pe_asm_lto_start): New function.
735 (i386_pe_asm_lto_end): Likewise.
737 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
738 Richard Biener <rguenther@suse.de>
741 * lto-streamer-out.c (output_function): Fix CFG loop state before
743 * omp-expand.c (expand_oacc_for): Handle calls to internal
744 functions like regular functions.
746 2018-04-12 Richard Biener <rguenther@suse.de>
749 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
750 for the early LTO debug to properly generate references to it
751 during DIE emission. Do not re-use that for the skeleton for
753 (dwarf2out_early_finish): Likewise.
755 2018-04-12 Jakub Jelinek <jakub@redhat.com>
759 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
760 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
761 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
762 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
763 and output is a reg, avoid creating invalid lowpart subreg, but
764 instead split into a 512-bit move. Don't split if not AVX512VL,
765 input is xmm16+ reg and output is a mem.
766 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
767 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
768 xmm16+ reg and output is a mem.
770 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
772 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
773 also for flag_dwarf2_cfi_asm.
775 2018-04-12 Jakub Jelinek <jakub@redhat.com>
777 PR rtl-optimization/85342
778 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
779 a bool scalar var inside of the loop instead. Don't try to update
780 recog_data.operand after failed apply_change_group.
782 2018-04-12 Tom de Vries <tom@codesourcery.com>
785 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
786 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
787 array with flexible array member as array without given dimension.
788 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
789 argument for undefined param to true.
791 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
794 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
795 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
796 from PowerPC section.
797 * config/rs6000/sysv4.opt (mcall-): Improve help text.
798 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
799 help text that is too long.
800 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
801 help text that is too long.
802 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
803 help text that is too long.
805 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
807 * config/alpha/alpha.md (stack_probe_internal): Rename
808 from "probe_stack". Update all callers.
810 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
812 PR rtl-optimization/84566
813 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
814 sched_macro_fuse_insns.
816 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
819 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
820 (compute_block_dependences): ... from here.
822 2018-04-11 Jakub Jelinek <jakub@redhat.com>
824 PR tree-optimization/85331
825 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
826 from int to HOST_WIDE_INT.
828 2018-04-11 Martin Jambor <mjambor@suse.cz>
831 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
832 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
833 not the same as the source val.
834 (cgraph_edge_brings_value_p): New parameter.
835 (gather_edges_for_value): Pass destination value to
836 cgraph_edge_brings_value_p.
837 (perhaps_add_new_callers): Likewise.
838 (get_info_about_necessary_edges): Likewise and exclude values brought
839 only by self-recursive edges.
840 (create_specialized_node): Redirect only clones of self-calling edges.
841 (+self_recursive_pass_through_p): New function.
842 (find_more_scalar_values_for_callers_subset): Use it.
843 (find_aggregate_values_for_callers_subset): Likewise.
844 (known_aggs_to_agg_replacement_list): Removed.
845 (decide_whether_version_node): Re-calculate known constants for all
846 remaining context clones.
848 2018-04-11 Richard Biener <rguenther@suse.de>
851 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
852 from early DWARF output.
853 (dwarf2out_early_finish): Output line info unconditionally into
854 early DWARF and add reference to it.
856 2018-04-11 Jakub Jelinek <jakub@redhat.com>
859 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
860 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
861 other than V2DFmode using iptr mode attribute.
862 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
864 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
866 PR rtl-optimization/84659
867 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
869 2018-04-11 Jakub Jelinek <jakub@redhat.com>
872 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
874 (output_loc_list): Pass address of a dummy size variable even in the
875 locview handling loop.
876 (index_location_lists): Add comment on why skip_loc_list_entry can't
879 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
882 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
885 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
888 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
889 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
890 and -mstring-compare-inline-limit.
892 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
895 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
896 for stack clash protection in a register whenever we need it to be in
899 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
901 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
902 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
904 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
907 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
909 (mlong-double-): Ditto.
910 * config/rs6000/sysv4.opt (msdata=): Ditto.
913 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
915 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
916 erroneous entries for
917 "vector int vec_ldl (int, long int *)", and
918 "vector unsigned int vec_ldl (int, unsigned long int *)".
919 Add comments and entries for
920 "vector bool char vec_ldl (int, bool char *)",
921 "vector bool short vec_ldl (int, bool short *)",
922 "vector bool int vec_ldl (int, bool int *)",
923 "vector bool long long vec_ldl (int, bool long long *)",
924 "vector pixel vec_ldl (int, pixel *)",
925 "vector long long vec_ldl (int, long long *)",
926 "vector unsigned long long vec_ldl (int, unsigned long long *)".
927 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
928 type tree bool_long_long_type_node and correct definition of
929 bool_V2DI_type_node to make reference to this new type tree.
930 (rs6000_mangle_type): Replace erroneous reference to
931 bool_long_type_node with bool_long_long_type_node.
932 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
933 comments to emphasize sign distinctions for char and int types and
934 replace RS6000_BTI_bool_long constant with
935 RS6000_BTI_bool_long_long constant. Also add comment to restrict
936 use of RS6000_BTI_pixel.
937 (bool_long_type_node): Remove this macro definition.
938 (bool_long_long_type_node): New macro definition
940 2018-04-10 Jakub Jelinek <jakub@redhat.com>
942 PR rtl-optimization/85300
943 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
944 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
945 simplify_unary_operation fails.
947 2018-04-10 Martin Liska <mliska@suse.cz>
949 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
950 cgraph_edge and ipa_ref.
952 2018-04-10 Jakub Jelinek <jakub@redhat.com>
957 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
958 computation of the VEC_MERGE selector from mask.
959 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
960 Fix decoding of the VEC_MERGE selector into mask.
962 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
964 PR tree-optimization/85286
965 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
967 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
969 * final.c (final_1): Set insn_last_address as well as
970 insn_current_address.
972 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
975 * explow.c (emit_stack_probe): Call validize_mem on memory location
976 before passing it to gen_probe_stack. Create address operand and
977 legitimize it for the probe_stack_address case.
979 2018-04-09 Jan Hubicka <jh@suse.cz>
982 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
983 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
984 * tree.c (free_lang_data_in_type): Fix handling of binfos;
986 (free_lang_data): Rebuild type inheritance graph.
988 2018-04-09 Martin Sebor <msebor@redhat.com>
990 * invoke.texi (-finline-small-functions): Mention other optimization
992 (-findirect-inlining, -fpartial-inlining): Same.
993 (-finline-functions-called-once): Same.
994 (-freorder-blocks-and-partition): Same.
996 2018-04-09 Jan Hubicka <jh@suse.cz>
999 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1000 jumps; choose last target that matches the criteria (i.e.
1001 no partition changes for non-crossing jumps).
1002 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1003 support for redirecting crossing jumps to non-crossing.
1005 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1007 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1008 also for naked functions.
1010 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1012 * config/arc/arc.md (add_shift): New pattern.
1013 (add_shift2): Likewise.
1014 (sub_shift): Likewise.
1015 (sub_shift_cmp0_noout): Likewise.
1016 (compare_si_ashiftsi): Likewise.
1017 (xbfu_cmp0_noout): New combine pattern.
1018 (xbfu_cmp0"): Likewise.
1019 (movsi_set_cc_insn): Place the predicable variant first.
1020 (commutative_binary_cmp0_noout): Remove clobber.
1021 (commutative_binary_cmp0): New pattern.
1022 (noncommutative_binary_cmp0): Likewise.
1023 (noncommutative_binary_cmp0_noout): Likewise.
1024 (noncommutative_binary_comparison_result_used): Removed.
1025 (rsub_cmp0): New pattern.
1026 (rsub_cmp0_noout): Likewise.
1027 (extzvsi): Changed, keep only meaningful variants.
1028 (SQH, SEZ): New iterators.
1029 (SQH_postfix): New mode attribute.
1030 (SEZ_prefix): New code attribute.
1031 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1032 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1033 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1035 (noncommutative_operator): Check the availability of barrel
1038 2018-04-09 Richard Biener <rguenther@suse.de>
1040 PR tree-optimization/85284
1041 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1042 Only use the niter constraining form of simple_iv when the exit
1045 2018-04-09 Tom de Vries <tom@codesourcery.com>
1048 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1049 (define_expand "*memory_barrier"): New define_expand.
1050 (define_insn "memory_barrier"): New insn.
1052 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1054 PR rtl-optimization/80463
1055 PR rtl-optimization/83972
1056 PR rtl-optimization/83480
1058 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1059 correct producer for the insn.
1060 (tidy_control_flow): Fixup seqnos in case of debug insns.
1062 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1064 PR rtl-optimization/83913
1066 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1067 different sched-times when merging exprs.
1069 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1071 PR rtl-optimization/83962
1073 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1074 tidy_fallthru_edge and tidy_control_flow.
1076 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1078 PR rtl-optimization/83530
1080 * sel-sched.c (force_next_insn): New global variable.
1081 (remove_insn_for_debug): When force_next_insn is true, also leave only
1082 next insn in the ready list.
1083 (sel_sched_region): When the region wasn't scheduled, make another pass
1084 over it with force_next_insn set to 1.
1086 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1088 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1090 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1091 for interrupt control.
1092 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1093 functions for interrupt control.
1094 * config/nds32/nds32-intrinsic.md: Likewise.
1095 * config/nds32/nds32_intrinsic.h: Likewise.
1096 * config/nds32/nds32.h (nds32_builtins): Likewise.
1098 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1100 * config/nds32/nds32.c (nds32_init_machine_status,
1101 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1102 strict_aligned_p field.
1103 (nds32_expand_to_rtl_hook): New function.
1104 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1105 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1107 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1108 Chung-Ju Wu <jasonwucj@gmail.com>
1110 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1111 * config/nds32/nds32-n7.md: New file.
1112 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1113 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1115 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1116 * config/nds32/nds32.md (pipeline_model): Add n7.
1117 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1118 * config/nds32/pipelines.md: Include n7 settings.
1120 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1121 Chung-Ju Wu <jasonwucj@gmail.com>
1123 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1124 * config/nds32/nds32-e8.md: New file.
1125 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1126 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1128 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1129 * config/nds32/nds32.md (pipeline_model): Add e8.
1130 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1131 * config/nds32/pipelines.md: Include e8 settings.
1133 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1134 Chung-Ju Wu <jasonwucj@gmail.com>
1136 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1137 * config/nds32/nds32-n8.md: New file.
1138 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1139 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1141 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1142 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1143 * config/nds32/nds32.md (pipeline_model): Add n8.
1144 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1145 * config/nds32/pipelines.md: Include n8 settings.
1147 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1148 Chung-Ju Wu <jasonwucj@gmail.com>
1150 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1151 * config/nds32/nds32-n9-2r1w.md: New file.
1152 * config/nds32/nds32-n9-3r2w.md: New file.
1153 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1154 nds32_register_ports): New or modify for cpu n9.
1155 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1157 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1158 * config/nds32/nds32-utils.c: New file.
1159 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1160 TARGET_MUL_SLOW): Define.
1161 * config/nds32/nds32.md (pipeline_model): New attribute.
1162 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1163 New options that support cpu n9.
1164 * config/nds32/pipelines.md: Include n9 settings.
1165 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1167 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1169 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1170 information if necessary.
1171 (output_cond_branch_compare_zero): Likewise.
1172 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1173 (nds32_target_alignment): Refine for alignment.
1174 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1175 (FUNCTION_BOUNDARY): Modify.
1176 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1178 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1180 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1182 * config/nds32/constants.md (unspec_volatile_element): Add values for
1183 TLB operation and data prefetch.
1184 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1185 functions for TLB operation and data prefetch.
1186 * config/nds32/nds32-intrinsic.md: Likewise.
1187 * config/nds32/nds32_intrinsic.h: Likewise.
1188 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1189 (nds32_print_operand): Likewise.
1190 * config/nds32/nds32.h (nds32_builtins): Likewise.
1192 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1193 Andrew Pinski <pinsika@gcc.gnu.org>
1196 * match.pd: Use constant_boolean_node of correct type instead of
1197 boolean_true_node or boolean_false_node for simplifying
1198 pointer comparisons to zero.
1200 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1202 PR tree-optimization/80021
1203 * tree.c (verify_type_variant): Make error call in verify_variant_match
1204 translatable and remove final full stop.
1206 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1208 * config/nds32/constants.md (unspec_volatile_element): Add
1209 UNSPEC_VOLATILE_EH_RETURN.
1210 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1211 nds32_output_stack_pop): Support dwarf exception handling process.
1212 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1213 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1214 exception handling process.
1215 (nds32_compute_stack_frame): Likewise.
1216 (nds32_return_addr_rtx): Likewise.
1217 (nds32_initial_elimination_offset): Likewise.
1218 (nds32_expand_prologue): Likewise.
1219 (nds32_expand_epilogue): Likewise.
1220 (nds32_dynamic_chain_address): New function.
1221 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1223 (DYNAMIC_CHAIN_ADDRESS): Define.
1224 (EH_RETURN_DATA_REGNO): Define.
1225 (EH_RETURN_STACKADJ_RTX): Define.
1226 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1227 patterns for dwarf exception handling.
1229 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1231 * config/nds32/nds32.h: Clean up obsolete macros.
1233 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1235 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1236 Add enum values for particular instructions.
1237 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1238 particular intrinsic functions.
1239 * config/nds32/nds32-intrinsic.md: Likewise.
1240 * config/nds32/nds32_intrinsic.h: Likewise.
1241 * config/nds32/nds32.h (nds32_builtins): Likewise.
1242 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1243 (btst, ave): New patterns for particular instructions.
1245 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1247 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1248 Add enum values for atomic load/store and memory sync.
1249 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1251 * config/nds32/nds32-intrinsic.md: Likewise.
1252 * config/nds32/nds32_intrinsic.h: Likewise.
1253 * config/nds32/nds32.h (nds32_builtins): Likewise.
1255 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1257 PR tree-optimization/85257
1258 * fold-const.c (native_encode_vector): If not all elts could fit
1259 and off is -1, return 0 rather than offset.
1260 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1261 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1262 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1263 adjust buffer in native_interpret_expr call.
1265 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1267 * config/nds32/constants.md (unspec_volatile_element): Add cache
1268 control enum values.
1269 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1270 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1271 * config/nds32/nds32.c (nds32_cctl_names): New.
1272 (nds32_print_operand): Handle cache control register names.
1273 * config/nds32/nds32.h (nds32_builtins): New enum values.
1274 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1276 * config/nds32/nds32.md (type): Add mmu.
1277 * config/nds32/pipelines.md (simple_insn): Add mmu.
1279 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1281 * config/nds32/nds32.md (type): Remove call.
1282 * config/nds32/pipelines.md (simple_insn): Likewise.
1284 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1286 * config/nds32/constants.md (unspec_volatile_element): Add
1287 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1288 UNSPEC_VOLATILE_FMFCFG.
1289 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1290 description for fmfcfg and fmfcsr.
1291 (bdesc_1arg): Add fmtcsr.
1292 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1293 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1294 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1295 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1296 unspec_fmfcfg): New patterns.
1297 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1298 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1299 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1300 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1301 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1302 __nds32__fmfcfg): Define.
1304 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1306 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1307 intrinsic register names.
1308 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1309 intrinsic register enum values and macros.
1311 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1313 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1314 for load/store addressing form.
1315 (nds32_print_operand_address): Likewise.
1317 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1320 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1321 based on LABEL_REF. Remove useless assertion.
1322 (pic_address_needs_scratch): Fix formatting.
1323 (sparc_legitimize_pic_address): Minor tweaks.
1324 (sparc_delegitimize_address): Adjust assertion accordingly.
1325 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1326 into symbolic_operand.
1327 (movsi_high_pic_label_ref): Likewise.
1328 (movsi_lo_sum_pic_label_ref): Likewise.
1329 (movdi_pic_label_ref): Likewise.
1330 (movdi_high_pic_label_ref): Likewise.
1331 (movdi_lo_sum_pic_label_ref): Likewise.
1333 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1335 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1336 custom LIB_SPEC setup.
1338 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1339 Kito Cheng <kito.cheng@gmail.com>
1341 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1342 * config/riscv/freebsd.h: New.
1344 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1346 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1347 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1350 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1351 Kito Cheng <kito.cheng@gmail.com>
1353 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1354 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1355 * config/nds32/nds32-protos.h (nds32_output_call,
1356 nds32_output_return): Declare.
1357 * config/nds32/nds32.md: Refine all the call and return patterns.
1359 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1362 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1363 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1365 PR rtl-optimization/84872
1366 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1367 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1370 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1372 * expr.c (copy_blkmode_to_reg): Revert 254862.
1373 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1375 2018-04-06 Richard Biener <rguenther@suse.de>
1378 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1379 after seeing a component reference with an adjacent field. Treat
1380 refs to arrays at struct end of external decls similar to
1381 refs to unconstrained commons.
1383 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1386 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1387 look through SAVE_EXPRs with non-side-effects argument. Adjust
1389 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1392 2018-04-06 Richard Biener <rguenther@suse.de>
1395 * alias.c (find_base_term): New wrapper around find_base_term
1396 unwinding CSELIB_VAL_PTR changes.
1397 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1400 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1402 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1404 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1405 constant definitions.
1406 ("nop"): lr 0,0 -> nopr r0
1407 ("nop_lr0", "nop_lr1"): New insn definitions.
1409 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1411 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1412 NDS32_V3PUSH_AVAILABLE_P macro.
1414 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1415 Chung-Ju Wu <jasonwucj@gmail.com>
1417 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1418 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1419 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1420 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1421 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1422 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1423 * config/nds32/constraints.md: New constraints and checking for hard
1424 float configuration.
1425 * config/nds32/iterators.md: New mode iterator and attribute for hard
1426 float configuration.
1427 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1429 * config/nds32/nds32-fpu.md: New file.
1430 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1431 deal with hard float code generation.
1432 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1434 (abi_type, float_reg_number): New enum type.
1435 * config/nds32/nds32-predicates.c: New predicates for hard float.
1436 * config/nds32/nds32-protos.h: Declare functions for hard float.
1437 * config/nds32/nds32.c: Implementation for hard float configuration.
1438 * config/nds32/nds32.h: Definitions for hard float configuration.
1439 * config/nds32/nds32.md: Include hard float machine description and
1440 modify patterns for hard float configuration.
1441 * config/nds32/nds32.opt: New options for hard float configuration.
1442 * config/nds32/predicates.md: New predicates for hard float
1445 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1447 * common/config/nds32/nds32-common.c
1448 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1450 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1453 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1454 CONSTRUCTOR_ELT (ctor, ...)->value.
1456 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1459 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1461 2018-04-05 Tom de Vries <tom@codesourcery.com>
1464 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1467 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1468 Kito Cheng <kito.cheng@gmail.com>
1470 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1471 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1472 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1473 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1475 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1476 Kito Cheng <kito.cheng@gmail.com>
1478 * config/nds32/constraints.md (Ufe): New memory constraint.
1479 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1480 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1481 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1483 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1484 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1486 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1488 * config/nds32/nds32.md: Use optimize_size in the condition for
1489 alu-shift instructions.
1491 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1493 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1495 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1497 * config/nds32/nds32.md (negsi2): Refine pattern.
1499 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1500 Chung-Ju Wu <jasonwucj@gmail.com>
1502 * config/nds32/iterators.md (shift_rotate): New code iterator.
1503 (shift): New code attribute.
1504 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1505 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1506 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1507 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1508 bit-wise operations.
1509 (andsi3, *andsi3): Ditto.
1510 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1511 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1512 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1513 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1514 nds32_ior_operand, nds32_xor_operand): New predicates.
1516 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1518 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1519 (addsi3, subsi3): ... this.
1521 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1523 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1525 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1527 * config/nds32/nds32.md: Adjust indention.
1529 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1531 * config/nds32/nds32.md (feature): New attribute.
1533 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1535 * config/nds32/nds32.md (subtype): New attribute.
1537 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1540 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1541 expansion to perform a bitwise AND of the argument followed by a
1542 boolean negation of the result.
1544 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1546 PR rtl-optimization/84878
1547 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1548 the basic block. Assert the use reference is not artificial and that
1549 it has an associated insn.
1551 2018-04-04 Michael Matz <matz@suse.de>
1553 * builtins.c (compute_objsize): Pass correct operand
1554 to array_at_struct_end_p.
1556 2018-04-04 Richard Biener <rguenther@suse.de>
1559 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1560 from contexts for DINFO_LEVEL_TERSE and below.
1562 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1564 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1565 resiter_operand condition.
1566 * config/nds32/nds32.md (*move<mode>): Ditto.
1568 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1569 Monk Chiang <sh.chiang04@gmail.com>
1571 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1573 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1575 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1577 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1578 Kito Cheng <kito.cheng@gmail.com>
1580 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1581 nds32_cond_code_str, output_cond_branch,
1582 output_cond_branch_compare_zero, nds32_expand_cbranch,
1583 nds32_expand_cstore, nds32_expand_movcc,
1584 nds32_output_cbranchsi4_equality_zero,
1585 nds32_output_cbranchsi4_equality_reg,
1586 nds32_output_cbranchsi4_equality_reg_or_const_int,
1587 nds32_output_cbranchsi4_greater_less_zero: New functions.
1588 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1589 nds32_expand_cstore, nds32_expand_movcc,
1590 nds32_output_cbranchsi4_equality_zero,
1591 nds32_output_cbranchsi4_equality_reg,
1592 nds32_output_cbranchsi4_equality_reg_or_const_int,
1593 nds32_output_cbranchsi4_greater_less_zero): Declare.
1594 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1595 nds32_rimm11s_operand): New predicates.
1596 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1597 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1600 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1602 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1603 * config/nds32/nds32.md: Ditto.
1604 * config/nds32/pipelines.md: Ditto.
1606 2018-04-04 Richard Biener <rguenther@suse.de>
1608 PR tree-optimization/85168
1609 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1610 propagating abnormals.
1612 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1614 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1616 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1617 Kito Cheng <kito.cheng@gmail.com>
1619 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1620 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1621 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1622 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1623 * config/nds32/nds32.md (sibcall_internal): New.
1624 (sibcall_register): Remove.
1625 (sibcall_immediate): Remove.
1626 (sibcall_value_internal): New.
1627 (sibcall_value_register): Remove.
1628 (sibcall_value_immediate): Remove.
1629 * config/nds32/predicates.md (nds32_general_register_operand): New.
1630 (nds32_call_address_operand): New.
1632 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1634 PR rtl-optimization/85167
1635 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1636 bb_defs if *split_p, instead preinitialize it to NULL.
1638 PR tree-optimization/85156
1639 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1640 evaluating the argument multiple times.
1642 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1644 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1646 (_mm_cvtpd_ps): Likewise.
1647 (_mm_cvttpd_epi32): Likewise.
1648 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1649 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1650 vector, pixel, and bool following altivec.h include.
1652 2018-04-03 Martin Sebor <msebor@redhat.com>
1654 * doc/extend.texi (Common Function Attributes): Clarify.
1655 (const attribute): Likewise.
1656 (pure attribute): Likewise.
1658 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1661 * config/i386/i386.c (ix86_expand_vector_set): Use
1662 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1664 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1666 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1667 instructions when changing rounding bits to preserve precision bits
1668 in the x87 control word.
1670 2018-04-03 Martin Liska <mliska@suse.cz>
1672 PR tree-optimization/82491
1673 * rtl.h (strip_offset_and_add): Replace += suboffset with
1674 poly_uint64 () + suboffset.
1676 2018-03-29 Martin Liska <mliska@suse.cz>
1677 Martin Jambor <mjambor@suse.cz>
1680 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1681 param_type is not an integral or pointer type.
1683 2018-04-03 Richard Biener <rguenther@suse.de>
1685 * sese.h (recompute_all_dominators): Remove.
1687 2018-04-02 Martin Sebor <msebor@redhat.com>
1689 * doc/invoke.texi (-Wrestrict): Fix typos.
1691 2018-04-02 Jim Wilson <jimw@sifive.com>
1693 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1694 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1695 (<optab>di3, <optab>si3_extend): Likewise.
1696 (<optab>si3_mask, <optab>si3_mask_1): New.
1697 (<optab>di3_mask, <optab>di3_mask_1): New.
1698 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1699 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1700 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1702 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1704 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1707 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1709 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1710 (nds32_canonicalize_comparison): New function.
1712 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1713 Kito Cheng <kito.cheng@gmail.com>
1714 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1716 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1717 * config/nds32/constants.md (unspec_volatile_element): Add
1718 UNSPEC_VOLATILE_RELAX_GROUP.
1719 * config/nds32/nds32-relax-opt.c: New file.
1720 * config/nds32/nds32-predicates.c
1721 (nds32_symbol_load_store_p): New function.
1722 * config/nds32/nds32-protos.h
1723 (nds32_symbol_load_store_p): Declare function.
1724 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1725 * config/nds32/nds32.c
1726 (nds32_register_pass): New function to register pass.
1727 (nds32_register_passes): New function to register passes.
1728 * config/nds32/nds32.md (relax_group): New pattern.
1729 * config/nds32/nds32.opt (mrelax-hint): New option.
1730 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1732 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1734 * config/nds32/t-nds32: Modify files dependency.
1736 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1738 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1739 (PROFILE_HOOK): Define its implementation.
1741 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1743 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1744 type and 32-bit size.
1746 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1749 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1750 (V_128_256): New mode iterator.
1751 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1752 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1753 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1755 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1756 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1758 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1761 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1762 NaN inputs correctly.
1764 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1767 * config/rs6000/vsx.md (??r): New mode attribute.
1768 (*vsx_mov<mode>_64bit): Use it.
1769 (*vsx_mov<mode>_32bit): Likewise.
1771 2018-03-30 Martin Sebor <msebor@redhat.com>
1773 PR tree-optimization/84818
1774 * builtins.c (check_access): Use warning_n.
1776 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1779 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1781 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1784 2018-03-30 Julia Koval <julia.koval@intel.com>
1787 * x86-tune.def (movx, partial_reg_dependency): Enable for
1790 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1793 * lra-constraints.c (process_alt_operands): Move setting
1794 this_alternative_matches below.
1796 2018-03-29 Martin Liska <mliska@suse.cz>
1799 * doc/invoke.texi: Document how LTO works with debug info.
1800 Describe auto-load support of binutils. Mention 'x86-64'
1801 as valid option value of -march option.
1803 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1805 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1808 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1809 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1810 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1813 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1816 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1817 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1818 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1819 for RS6000_BTM_POWERPC64.
1820 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1821 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1822 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1827 2018-03-28 Carl Love <cel@us.ibm.com>
1830 2017-09-27 Carl Love <cel@us.ibm.com>
1832 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1833 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1834 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1837 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1839 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1840 instead of __vector bool.
1841 (_mm_max_pu8): Likewise.
1842 (_mm_min_pi16): Likewise.
1844 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1847 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1848 (DIVWEUO): Likewise.
1850 (DIVDEUO): Likewise.
1851 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1852 DIVWEUO and DIVDEUO.
1853 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1854 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1855 (div_extend): Likewise.
1856 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1858 (__builtin_divweuo): Likewise.
1859 (__builtin_divdeo): Likewise.
1860 (__builtin_divdeuo): Likewise.
1862 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1865 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1866 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1868 PR tree-optimization/82004
1869 * gimple-match-head.c (optimize_pow_to_exp): New function.
1870 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1871 Don't fold to exp if optimize_pow_to_exp is false.
1873 2018-03-28 Martin Liska <mliska@suse.cz>
1876 * calls.c (initialize_argument_information): Fix trailing space.
1877 * common.opt: Fix typo and provide better explanation for
1878 -fsanitize-coverage option.
1879 * config/i386/i386.opt: Fix typo.
1881 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1882 Martin Liska <mliska@suse.cz>
1885 * gimplify.c (asan_poison_variable): Don't do the check for
1886 gimplify_omp_ctxp here.
1887 (gimplify_decl_expr): Do it here.
1888 (gimplify_target_expr): Likewise.
1890 2018-03-28 Martin Liska <mliska@suse.cz>
1893 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1894 chkp_type_bounds_count if MPX is not enabled.
1896 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1898 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1900 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1903 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1904 function to create the function decl for complex long double
1905 multiply and divide for -mabi=ieeelongdouble.
1906 (init_float128_ieee): Call it.
1908 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1911 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1912 -fcf-protection=branch -mibt.
1913 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1915 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1918 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1920 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1923 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1924 extern array declarations.
1926 2018-03-27 Richard Biener <rguenther@suse.de>
1929 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1930 explicit single_use checks.
1932 2018-03-27 Richard Biener <rguenther@suse.de>
1934 PR tree-optimization/85082
1935 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1938 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1940 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1941 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1942 Turn on fasynchronous-unwind-tables and funwind-tables.
1944 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1947 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1948 (*bmi_blsr_<mode>_ccz): Ditto.
1950 2018-03-26 Tom de Vries <tom@codesourcery.com>
1952 PR tree-optimization/85063
1953 * omp-general.c (offloading_function_p): New function. Factor out
1955 * omp-offload.c (pass_omp_target_link::gate): ... here.
1956 * omp-general.h (offloading_function_p): Declare.
1957 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1958 with attribute omp declare target for offloading functions.
1960 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1962 PR tree-optimization/84005
1963 * tree-data-ref.h (get_base_for_alignment): Declare.
1964 * tree-data-ref.c (get_base_for_alignment_1): New function.
1965 (get_base_for_alignment): Likewise.
1966 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1967 get_base_for_alignment to find a suitable base object, instead
1968 of always using drb->base_address.
1970 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1973 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1974 known size by default.
1976 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1979 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1980 and non BLKmode operands.
1982 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1985 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1986 Clean up attributes.
1988 2018-03-23 Richard Biener <rguenther@suse.de>
1991 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1992 we are going to emit early debug for LTO.
1994 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1997 * function.c (match_asm_constraints_1): Don't optimize if input
1998 doesn't satisfy general_operand predicate for output's mode.
2001 * alias.c (write_dependence_p): Don't require for x_canonicalized
2002 non-VOIDmode if x has VOIDmode.
2005 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2006 just don't try to optimize it rather than assert it never happens.
2008 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2010 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2011 macro expansions for definition of ST_INTERNAL_<mode> and
2012 LD_INTERNAL_<mode> builtins.
2013 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2015 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2017 (altivec_expand_st_builtin): Likewise.
2018 (altivec_expand_builtin): Remove calls to deleted functions.
2019 (rs6000_address_for_altivec): Delete this function.
2020 * config/rs6000/vector.md: Remove expands for
2021 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2023 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2026 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2027 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2028 re-computing once computed.
2029 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2030 (arm_init_machine_status): Initialize
2031 machine->static_chain_stack_bytes.
2033 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2036 * doc/extend.texi: Add four new prototypes for vec_ld.
2037 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2038 definitions for more logical presentation.
2039 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2040 entries for V1TI variants of __builtin_altivec_ld builtin.
2041 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2042 handling of V1TI variant of LVX icode pattern.
2043 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2044 (rs6000_gimple_fold_builtin): Likewise.
2045 (altivec_init_builtins): Add code to define
2046 __builtin_altivec_lvx_v1ti function.
2048 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2051 * function.c (match_asm_constraints_1): Don't do the optimization
2052 if input isn't a REG, SUBREG, MEM or constant.
2054 2018-03-22 Tom de Vries <tom@codesourcery.com>
2056 PR tree-optimization/84956
2057 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2058 bb_has_abnormal_pred.
2060 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2063 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2064 DECL_INITIAL (decl) to decl at the end.
2065 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2068 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2070 * doc/extend.texi (__builtin_tgmath): Document when complex
2071 integer types are treated as _Complex _Float64.
2073 2018-03-21 Tom de Vries <tom@codesourcery.com>
2075 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2077 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2079 PR tree-optimization/84960
2080 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2081 if it is ENTRY block, move them into single succ of ENTRY in that case.
2083 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2085 PR tree-optimization/84811
2086 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2087 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2088 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2089 (known_subrange_p): Update accordingly. Cast each value involved
2090 in the size comparison, rather than casting the result of the
2093 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2095 PR tree-optimization/84982
2096 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2097 by flipping the least significant bit rather than all bits from
2098 bitpos to bitpos + bitsize - 1.
2100 2018-03-21 Nathan Sidwell <nathan@acm.org>
2102 * doc/extend.texi (Deprecated Features): Remove mention of
2103 long-deleted deprecations.
2105 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2108 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2109 * configure: Regenerate.
2111 2018-03-21 Tom de Vries <tom@codesourcery.com>
2113 PR tree-optimization/83126
2114 * tree-parloops.c (num_phis): New function.
2115 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2117 2018-03-21 Nathan Sidwell <nathan@acm.org>
2119 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2120 mention anon-struct/union members and trailing attributes.
2122 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2124 PR tree-optimization/84969
2125 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2126 builtin memset partitions if they set different rhs values.
2128 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2130 PR rtl-optimization/84989
2131 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2132 VEC_DUPLICATE with scalar result mode.
2134 2018-03-21 Martin Liska <mliska@suse.cz>
2137 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2138 not intended return statement.
2140 2018-03-21 Martin Liska <mliska@suse.cz>
2143 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2144 (chkp_find_bound_slots_1): Limit number of iterations.
2146 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2149 * Minor grammar fixes for x86 options.
2151 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2154 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2155 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2157 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2160 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2161 (altivec_lvx_<mode>_1op): Likewise.
2162 (altivec_stvx_<mode>_2op): Likewise.
2163 (altivec_stvx_<mode>_1op): Likewise.
2164 (altivec_lvx_<VM2:mode>): New define_expand.
2165 (altivec_stvx_<VM2:mode>): Likewise.
2166 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2167 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2168 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2169 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2170 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2171 (rs6000_gen_lvx): Likewise.
2172 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2173 (altivec_expand_stv_builtin): Likewise.
2174 (altivec_expand_builtin): Likewise.
2175 * config/rs6000/vector.md: Likewise.
2177 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2180 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2183 2018-03-20 Richard Biener <rguenther@suse.de>
2186 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2187 sign-conversions as zero, fall back to standard scalar_stmt
2190 2018-03-20 Martin Liska <mliska@suse.cz>
2193 * predict.c (rebuild_frequencies): Handle case when we have
2194 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2196 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2199 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2200 flag_section_anchors.
2201 * varasm.c (use_blocks_for_decl_p): Remove hack for
2202 dw2_force_const_mem.
2205 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2207 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2208 be created, use lowpart_subreg of operands[0] rather than operands[0]
2210 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2211 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2212 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2213 and n constraint instead of aarch64_shift_imm_di and Usd.
2214 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2215 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2217 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2220 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2221 to favor GPR over NEON registers.
2222 (<shift>di3_neon): Likewise.
2224 2018-03-20 Tom de Vries <tom@codesourcery.com>
2227 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2228 (nvptx_process_pars): Emit bar.sync asap and alap.
2230 2018-03-20 Tom de Vries <tom@codesourcery.com>
2233 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2234 seen_label if seen_label is already set.
2236 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2239 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2240 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2241 Use 1U instead of 1. Formatting fixes.
2244 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2245 instead of TREE_TYPE (s1) for the return value.
2247 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2249 PR tree-optimization/84946
2250 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2251 bitsize + bitsize in poly_uint64 rather than poly_int64.
2254 * dwarf2asm.c: Include fold-const.c.
2255 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2256 of decl rather than decl itself.
2258 PR rtl-optimization/84643
2259 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2261 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2264 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2265 calling assemble_variable.
2267 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2270 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2271 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2273 2018-03-19 Jim Wilson <jimw@sifive.com>
2276 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2277 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2278 (riscv_first_stack_step): Likewise.
2279 (riscv_option_override): Use STACK_BOUNDARY instead of
2281 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2283 (BIGGEST_ALIGNMENT): Set to 128.
2284 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2285 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2288 2018-03-19 Richard Biener <rguenther@suse.de>
2290 PR tree-optimization/84933
2291 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2292 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2294 2018-03-19 Richard Biener <rguenther@suse.de>
2296 PR tree-optimization/84859
2297 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2298 (cond_if_else_store_replacement): Perform sinking operation on
2299 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2300 Generalize what a BB with a single eligible store is.
2302 2018-03-19 Richard Biener <rguenther@suse.de>
2304 PR tree-optimization/84929
2305 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2306 chrec_is_positive against non-chrec arg.
2308 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2311 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2313 2018-03-18 Martin Liska <mliska@suse.cz>
2315 PR rtl-optimization/84635
2316 * regrename.c (build_def_use): Use matches_mode only when
2319 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2321 PR tree-optimization/84913
2322 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2323 vectorize chains of COND_EXPRs.
2325 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2327 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2329 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2331 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2333 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2335 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2337 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2338 Kito Cheng <kito.cheng@gmail.com>
2340 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2341 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2342 (nds32_adjust_reg_alloc_order): New function.
2343 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2345 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2347 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2348 nds32_print_operand, nds32_print_operand_address): Use
2349 HOST_WIDE_INT_PRINT_DEC instead.
2351 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2353 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2355 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2358 * config/i386/i386.c (initial_ix86_tune_features,
2359 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2361 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2362 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2363 rather than 1u << ix86_tune. Formatting fix.
2364 (ix86_option_override_internal): Change ix86_arch_mask from
2365 unsigned int to unsigned HOST_WIDE_INT, initialize to
2366 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2367 (ix86_function_specific_restore): Likewise.
2369 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2372 * postreload.c (reload_combine_recognize_pattern): Perform
2373 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2374 truncate_int_for_mode the result for the destination's mode.
2377 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2378 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2380 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2383 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2386 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2387 regno_allocno_class_array and sorted_pseudos.
2388 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2389 insns where regno is used.
2391 2018-03-16 Martin Liska <mliska@suse.cz>
2394 * multiple_target.c (create_dispatcher_calls): Redirect
2395 reference in the symbol table.
2397 2018-03-16 Martin Liska <mliska@suse.cz>
2400 * multiple_target.c (create_dispatcher_calls): Redirect also
2403 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2407 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2408 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2409 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2411 2018-03-16 Julia Koval <julia.koval@intel.com>
2413 * doc/invoke.texi (Skylake Server): Add CLWB.
2414 Cannonlake): Remove CLWB.
2416 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2418 PR tree-optimization/84841
2419 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2421 (FLOAT_ONE_CONST_TYPE): Define.
2422 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2423 (sort_by_operand_rank): Put entries with higher constant_type last
2424 rather than first to match comments.
2426 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2428 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2431 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2434 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2435 check and use address of target always.
2437 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2440 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2441 (indirect_thunk_bnd_needed): Likewise.
2442 (indirect_thunks_used): Likewise.
2443 (indirect_thunks_bnd_used): Likewise.
2444 (indirect_return_needed): New.
2445 (indirect_return_bnd_needed): Likewise.
2446 (output_indirect_thunk_function): Add a bool argument for
2448 (output_indirect_thunk_function): Don't generate alias for
2449 function return thunk.
2450 (ix86_code_end): Call output_indirect_thunk_function to generate
2451 function return thunks.
2452 (ix86_output_function_return): Set indirect_return_bnd_needed
2453 and indirect_return_needed instead of indirect_thunk_bnd_needed
2454 and indirect_thunk_needed.
2456 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2458 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2459 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2460 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2462 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2463 Paul Hua <paul.hua.gm@gmail.com>
2466 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2468 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2470 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2471 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2474 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2477 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2478 instead of GET_MODE_SIZE when comparing Units.
2480 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2483 * varasm.c (hash_section): Return an unchangeble hash value
2484 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2485 Return !aarch64_can_use_per_function_literal_pools_p ().
2487 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2490 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2491 rather than address of cmode as last argument to prepare_cmp_insn.
2493 2018-03-15 Julia Koval <julia.koval@intel.com>
2495 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2496 F_AVX512VNNI, F_AVX512BITALG): New.
2498 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2501 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2502 insn for floating-point loads and stores.
2504 2018-03-14 Carl Love <cel@us.ibm.com>
2506 * config/rs6000/rs6000-c.c: Add macro definitions for
2507 ALTIVEC_BUILTIN_VEC_PERMXOR.
2508 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2509 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2510 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2511 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2513 * config/doc/extend.texi: Add prototypes for vec_permxor.
2515 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2518 * diagnostic-show-locus.c (class layout_point): Convert m_line
2519 from int to linenum_type.
2520 (line_span::comparator): Use linenum "compare" function when
2521 comparing line numbers.
2522 (test_line_span): New function.
2523 (layout_range::contains_point): Convert param "row" from int to
2525 (layout_range::intersects_line_p): Likewise.
2526 (layout::will_show_line_p): Likewise.
2527 (layout::print_source_line): Likewise.
2528 (layout::should_print_annotation_line_p): Likewise.
2529 (layout::print_annotation_line): Likewise.
2530 (layout::print_leading_fixits): Likewise.
2531 (layout::annotation_line_showed_range_p): Likewise.
2532 (struct line_corrections): Likewise for field m_row.
2533 (line_corrections::line_corrections): Likewise for param "row".
2534 (layout::print_trailing_fixits): Likewise.
2535 (layout::get_state_at_point): Likewise.
2536 (layout::get_x_bound_for_row): Likewise.
2537 (layout::print_line): Likewise.
2538 (diagnostic_show_locus): Likewise for locals "last_line" and
2540 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2541 * input.c (selftest::test_linenum_comparisons): New function.
2542 (selftest::input_c_tests): Call it.
2543 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2544 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2545 * selftest.h (ASSERT_GT): New macro.
2546 (ASSERT_GT_AT): New macro.
2547 (ASSERT_LT): New macro.
2548 (ASSERT_LT_AT): New macro.
2550 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2552 PR rtl-optimization/84780
2553 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2555 2018-03-14 Martin Liska <mliska@suse.cz>
2557 * tree.c (record_node_allocation_statistics): Use
2558 get_stats_node_kind.
2559 (get_stats_node_kind): New function extracted from
2560 record_node_allocation_statistics.
2561 (free_node): Use get_stats_node_kind.
2563 2018-03-14 Richard Biener <rguenther@suse.de>
2565 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2566 that the value-set of ANTIC_IN doesn't grow.
2569 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2571 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2572 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2574 2018-03-14 Julia Koval <julia.koval@intel.com>
2576 * config.gcc (icelake-client, icelake-server): New.
2578 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2579 (initial_ix86_arch_features): Ditto.
2580 (PTA_SKYLAKE): Add SGX.
2581 (PTA_ICELAKE): Remove.
2582 (PTA_ICELAKE_CLIENT): New.
2583 (PTA_ICELAKE_SERVER): New.
2584 (ix86_option_override_internal): Split up icelake on icelake client and
2586 (get_builtin_code_for_version): Ditto.
2587 (fold_builtin_cpu): Ditto.
2588 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2589 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2590 * config/i386/i386.h (processor_type): Ditto.
2591 * doc/invoke.texi: Ditto.
2593 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2596 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2597 INTEGER_CST offset, add it together with bitpos / 8 and
2598 sign extend based on POINTER_SIZE.
2602 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2605 * config/i386/constraints.md (Yc): New register constraint.
2606 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2607 Use Yc constraint for alternative 2 of operand 0. Remove
2608 preferred_for_speed attribute.
2610 2018-03-14 Richard Biener <rguenther@suse.de>
2612 PR tree-optimization/84830
2613 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2614 with the old one to avoid oscillations.
2616 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2619 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2621 (assign_by_spills): Return a flag of reload assignment failure.
2622 Do not process the reload assignment failures. Do not spill other
2623 reload pseudos if they has the same reg class. Update n if
2625 (lra_assign): Add a return arg. Set up from the result of
2626 assign_by_spills call.
2627 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2628 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2629 usage_insns if it is not NULL.
2630 (spill_hard_reg_in_range): New function.
2631 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2632 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2633 function prototypes.
2634 (lra_assign): Change prototype.
2635 * lra.c (lra): Add code to deal with fails by splitting hard reg
2638 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2640 * config/riscv/riscv.opt (mrelax): New option.
2641 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2642 "norelax" when riscv_mrelax is disabled.
2643 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2645 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2648 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2649 reassociation for int modes.
2651 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2653 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2654 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2656 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2657 * config/aarch64/aarch64-sve.md
2658 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2659 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2660 (*extend<mode><Vwide>2): Rename to...
2661 (aarch64_sve_extend<mode><Vwide>2): ...this.
2662 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2663 renaming the old pattern to...
2664 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2666 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2667 define_expand, renaming the old pattern to...
2668 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2669 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2670 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2671 account when deciding which SVE instruction the optab should use.
2672 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2674 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2676 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2677 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2678 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2679 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2680 (tlsdesc_small_<mode>): Turn a define_expand and use
2681 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2682 (tlsdesc_small_advsimd_<mode>): ...this.
2683 (tlsdesc_small_sve_<mode>): New pattern.
2685 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2687 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2688 (UNSPEC_UMUL_HIGHPART): New constants.
2689 (MUL_HIGHPART): New int iteraor.
2690 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2691 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2693 (*<su>mul<mode>3_highpart): New define_insn.
2695 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2698 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2701 2018-03-13 Martin Liska <mliska@suse.cz>
2704 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2706 (sem_item_optimizer::~sem_item_optimizer): Release it.
2707 (sem_item_optimizer::merge_classes): Register variable aliases.
2708 (sem_item_optimizer::fixup_pt_set): New function.
2709 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2710 * ipa-icf.h: Declare new variables and functions.
2712 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2715 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2716 integer_pow2p@2 and test integer_pow2p in condition.
2717 (A < 0 ? C : 0): Similarly for @1.
2720 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2721 characters starting at p contain '\0' character, don't look beyond
2725 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2726 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2729 * reg-stack.c (change_stack): Change update_end var from int to
2730 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2731 also call set_block_for_insn on the newly added insns and rescan.
2734 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2735 on the last operand.
2738 * tree.c (stabilize_reference_1): Return save_expr (e) for
2739 STATEMENT_LIST even if it doesn't have side-effects.
2741 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2743 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2745 2018-03-12 Renlin Li <renlin.li@arm.com>
2747 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2748 aarch64_output_scalar_simd_mov_immediate.
2750 2018-03-12 Martin Sebor <msebor@redhat.com>
2752 PR tree-optimization/83456
2753 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2754 for perfectly overlapping calls to memcpy.
2755 (gimple_fold_builtin_memory_chk): Same.
2756 (gimple_fold_builtin_strcpy): Handle no-warning.
2757 (gimple_fold_builtin_stxcpy_chk): Same.
2758 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2760 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2762 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2763 parameter. Use it for SFmode.
2764 (rs6000_function_arg_advance_1): Adjust.
2765 (rs6000_function_arg): Adjust.
2766 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2768 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2770 PR rtl-optimization/84169
2771 PR rtl-optimization/84780
2772 * combine.c (can_combine_p): Check for a 2-insn combination whether
2773 the destination register is used between the two insns, too.
2775 2018-03-12 Richard Biener <rguenther@suse.de>
2777 PR tree-optimization/84803
2778 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2779 for refs DR analysis didn't process.
2781 2018-03-12 Richard Biener <rguenther@suse.de>
2783 PR tree-optimization/84777
2784 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2785 force-vectorize loops ignore whether we are optimizing for size.
2787 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2789 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2790 (TARGET_MD_ASM_ADJUST): Define.
2792 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2793 Kito Cheng <kito.cheng@gmail.com>
2794 Chung-Ju Wu <jasonwucj@gmail.com>
2796 * config/nds32/nds32.c (nds32_compute_stack_frame,
2797 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2798 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2799 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2800 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2801 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2802 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2803 * config/nds32/nds32.md (prologue, epilogue): Use macro
2804 NDS32_V3PUSH_AVAILABLE_P to do checking.
2806 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2809 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2810 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2811 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2812 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2813 addition of most attributes on !orig_type_die or the attribute not
2814 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2816 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2817 Chung-Ju Wu <jasonwucj@gmail.com>
2819 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2821 * config/nds32/nds32.opt (mvh): New option.
2823 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2824 Chung-Ju Wu <jasonwucj@gmail.com>
2826 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2828 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2829 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2832 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2833 Chung-Ju Wu <jasonwucj@gmail.com>
2835 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2837 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2838 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2840 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2841 Kito Cheng <kito.cheng@gmail.com>
2842 Chung-Ju Wu <jasonwucj@gmail.com>
2844 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2845 UNSPEC_FFMISM and UNSPEC_FLMISM.
2846 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2847 for ffb, ffmism and flmism.
2848 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2849 (unspec_ffmism): Ditto.
2850 (unspec_flmism): Ditto.
2851 (nds32_expand_builtin_impl): Check if string extension is available.
2852 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2853 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2855 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2858 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2861 * lra-assigns.c (assign_by_spills): Return a flag of reload
2862 assignment failure. Do not process the reload assignment
2863 failures. Do not spill other reload pseudos if they has the same
2865 (lra_assign): Add a return arg. Set up from the result of
2866 assign_by_spills call.
2867 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2868 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2869 usage_insns if it is not NULL.
2870 (spill_hard_reg_in_range): New function.
2871 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2872 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2873 function prototypes.
2874 (lra_assign): Change prototype.
2875 * lra.c (lra): Add code to deal with fails by splitting hard reg
2878 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2881 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2883 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2886 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2887 (dw_val_node): Add val_symbolic_view.
2888 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2889 (symview_upper_bound): New.
2890 (new_line_info_table): Initialize symviews_since_reset.
2891 (dwarf2out_source_line): Count symviews_since_reset and set
2892 symview_upper_bound.
2893 (dw_val_equal_p): Handle symview.
2894 (add_AT_symview): New.
2895 (print_dw_val): Handle symview.
2896 (attr_checksum, attr_checksum_ordered): Likewise.
2897 (same_dw_val_p, size_of_die): Likewise.
2898 (value_format, output_die): Likewise.
2899 (add_high_low_attributes): Use add_AT_symview for entry_view.
2900 (dwarf2out_finish): Reset symview_upper_bound, clear
2903 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2906 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2907 Add strict argument and use it.
2908 (rs6000_split_multireg_move): Update for new strict argument.
2909 (mem_operand_gpr): Disallow all non-offsettable addresses.
2910 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2912 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2915 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2916 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2917 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2920 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2921 decl, use remap_type if we want to use the type.
2923 2018-03-09 Martin Sebor <msebor@redhat.com>
2925 PR tree-optimization/84526
2926 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2928 (builtin_access::generic_overlap): Be prepared to handle non-array
2931 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2933 PR rtl-optimization/84682
2934 * lra-constraints.c (process_address_1): Check is_address flag
2935 for address constraints.
2936 (process_alt_operands): Likewise.
2937 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2938 preprocess_constraints.
2939 * recog.h (preprocess_constraints): Add oploc parameter.
2941 * recog.c (preprocess_constraints): Test address_operand for
2942 CT_ADDRESS constraints.
2944 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2947 * lra-assigns.c (assign_by_spills): Return a flag of reload
2948 assignment failure. Do not process the reload assignment
2949 failures. Do not spill other reload pseudos if they has the same
2951 (lra_assign): Add a return arg. Set up from the result of
2952 assign_by_spills call.
2953 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2954 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2955 usage_insns if it is not NULL.
2956 (spill_hard_reg_in_range): New function.
2957 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2958 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2959 function prototypes.
2960 (lra_assign): Change prototype.
2961 * lra.c (lra): Add code to deal with fails by splitting hard reg
2964 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2967 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2968 Accept complain bool parameter. Only emit errors if it is true.
2969 (arm_parse_cpu_option_name): Likewise.
2970 (arm_target_thumb_only): Adjust callers of the above.
2971 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2972 prototype to take a default true bool parameter.
2973 (arm_parse_arch_option_name): Likewise.
2975 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2976 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2980 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2981 * configure: Regenerate.
2982 * configure.ac ("linker --version-script option"): New.
2983 ("linker soname option"): New.
2985 2018-03-09 Richard Biener <rguenther@suse.de>
2987 PR tree-optimization/84775
2988 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2989 immediate uses of predicate stmts and mark them modified.
2992 PR tree-optimization/84178
2993 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2995 (version_loop_for_if_conversion): Delay update_ssa call.
2996 (tree_if_conversion): Delay update_ssa until after predicate
2999 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3002 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3003 when the function accesses prior frames.
3005 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3008 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3009 gen_llsym, otherwise call maybe_gen_llsym.
3012 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3013 has ',' character inside of it.
3015 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3018 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3019 as clobbering CC_REGNUM.
3021 2018-03-08 Richard Biener <rguenther@suse.de>
3024 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3025 (follow_copies_to_constant): Do not follow SSA names registered
3028 2018-03-08 Richard Biener <rguenther@suse.de>
3030 PR tree-optimization/84178
3031 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3033 (version_loop_for_if_conversion): Delay update_ssa call.
3034 (tree_if_conversion): Delay update_ssa until after predicate
3037 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3039 PR tree-optimization/84178
3040 * tree-if-conv.c (release_bb_predicate): Remove the
3041 the assertion that the stmts have NULL use_ops.
3042 Discard the statements, asserting that they haven't
3043 yet been added to a BB.
3045 2018-03-08 Richard Biener <rguenther@suse.de>
3047 PR tree-optimization/84746
3048 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3049 (phi_translate): Pass in destination ANTIC_OUT set.
3050 (phi_translate_1): Likewise. For a simplified result lookup
3051 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3052 (phi_translate_set): Adjust.
3053 (do_pre_regular_insertion): Likewise.
3054 (do_pre_partial_partial_insertion): Likewise.
3056 2018-03-08 Martin Liska <mliska@suse.cz>
3058 PR gcov-profile/84735
3059 * doc/gcov.texi: Document usage of profile files.
3060 * gcov-io.h: Document changes in the format.
3062 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3066 * dwarf2out.c (struct dw_line_info_table): Update comments for
3068 (FORCE_RESET_NEXT_VIEW): New.
3069 (FORCE_RESETTING_VIEW_P): New.
3070 (RESETTING_VIEW_P): Check for -1 too.
3071 (ZERO_VIEW_P): Likewise.
3072 (new_line_info_table): Force-reset next view.
3073 (dwarf2out_begin_function): Likewise.
3074 (dwarf2out_source_line): Simplify zero_view_p initialization.
3075 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3076 view directly. Omit view when omitting .loc at line 0.
3078 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3080 PR tree-optimization/84740
3081 * tree-switch-conversion.c (process_switch): Call build_constructors
3082 only if info.phi_count is non-zero.
3084 PR tree-optimization/84739
3085 * tree-tailcall.c (find_tail_calls): Check call arguments against
3086 DECL_ARGUMENTS (current_function_decl) rather than
3087 DECL_ARGUMENTS (func) when checking for tail recursion.
3089 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3091 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3092 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3093 Volker Reichelt's entry and add entries for people that perform
3094 GCC fuzzy testing and report numerous bugs.
3096 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3099 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3100 readonly data in sdata, if that is disabled.
3101 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3102 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3103 -mreadonly-in-sdata option.
3105 2018-03-07 Martin Sebor <msebor@redhat.com>
3107 PR tree-optimization/84468
3108 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3109 basic block when looking for nul assignment.
3111 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3114 * except.h (output_function_exception_table): Adjust prototype.
3115 * except.c (output_function_exception_table): Remove FNNAME parameter
3116 and add SECTION parameter. Ouput one part of the table at a time.
3117 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3118 the first part of the exception table and emit unwind directives.
3119 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3120 (i386_pe_seh_cold_init): Likewise.
3121 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3122 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3123 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3124 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3125 * config/i386/winnt.c: Include except.h.
3126 (struct seh_frame_state): Add reg_offset, after_prologue and
3127 in_cold_section fields.
3128 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3129 (i386_pe_seh_cold_init): New function.
3130 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3131 to seh->in_cold_section.
3132 (seh_emit_push): Record the offset of the push.
3133 (seh_emit_save): Record the offet of the save.
3134 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3135 Test seh->after_prologue to disregard the epilogue.
3136 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3137 (i386_pe_end_cold_function): New function.
3139 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3142 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3143 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3146 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3148 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3149 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3152 * multiple_target.c: Include tree-inline.h and intl.h.
3153 (expand_target_clones): Diagnose and fail if node->definition and
3154 !tree_versionable_function_p (node->decl).
3156 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3158 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3160 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3161 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3162 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3164 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3167 * combine.c (try_combine): Use reg_or_subregno instead of handling
3168 just paradoxical SUBREGs and REGs.
3170 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3172 * config/arc/arc.c (arc_finalize_pic): Remove function.
3173 (arc_must_save_register): We use single base PIC register, remove
3174 checks to save/restore the PIC register.
3175 (arc_expand_prologue): Likewise.
3176 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3178 (arc_verify_short): Likewise.
3179 (arc_attr_type): Likewise.
3180 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3181 (walk_stores): Likewise.
3182 (arc_address_cost): Make it static.
3183 (arc_verify_short): Likewise.
3184 (branch_dest): Likewise.
3185 (arc_attr_type): Likewise.
3186 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3187 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3188 (arc_final_prescan_insn): Remove inserting the nops due to
3189 hardware hazards. It is done in reorg step.
3190 (insn_length_variant_t): Remove.
3191 (insn_length_parameters_t): Likewise.
3192 (arc_insn_length_parameters): Likewise.
3193 (arc_get_insn_variants): Likewise.
3194 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3196 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3199 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3202 PR tree-optimization/84687
3203 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3205 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3207 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3209 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3210 Rename to ppc_speculation_barrier.
3211 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3212 __builtin_ppc_speculation_barrier.
3214 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3217 * combine.c (combine_simplify_rtx): Don't try to simplify if
3218 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3221 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3223 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3224 to 32 bytes when compiling for POWER9.
3226 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3229 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3230 regparm >= 3 with no arg reg available also for calls with
3231 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3234 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3236 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3238 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3241 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3243 2018-03-05 Richard Biener <rguenther@suse.de>
3245 PR tree-optimization/84486
3246 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3247 When inserting a __builtin_assume_aligned call set the LHS
3248 SSA name alignment info accordingly.
3250 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3252 PR tree-optimization/84114
3253 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3254 Avoid reassociation of FLOAT_MODE addition.
3256 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3258 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3259 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3260 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3261 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3262 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3263 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3264 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3266 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3267 __builtin_ia32_wbinvd): New builtins.
3268 (SPECIAL_ARGS2): New.
3269 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3270 (SPECIAL_ARGS2): New.
3271 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3272 (ix86_valid_target_attribute_inner_p): Ditto.
3273 (ix86_init_mmx_sse_builtins): Add special_args2.
3274 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3275 TARGET_WBNOINVD_P): New.
3276 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3277 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3278 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3279 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3280 * config/i386/pconfigintrin.h: New file.
3281 * config/i386/wbnoinvdintrin.h: Ditto.
3282 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3283 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3285 2018-03-05 Richard Biener <rguenther@suse.de>
3287 PR tree-optimization/84670
3288 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3290 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3291 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3292 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3293 doesn't grow if all successors (recursively) were visited at least
3296 2018-03-05 Richard Biener <rguenther@suse.de>
3298 PR tree-optimization/84650
3299 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3300 if executed in the loop pipeline.
3302 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3304 * doc/configfiles.texi (Configuration Files): Move info about
3305 conditionalizing $target-protos.h to...
3306 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3307 differs from $target-protos.h.
3309 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3310 Chung-Ju Wu <jasonwucj@gmail.com>
3312 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3313 * config/nds32/nds32-multiple.md (setmemsi): Define.
3314 * config/nds32/nds32-memory-manipulation.c
3315 (nds32_gen_dup_4_byte_to_word_value): New.
3316 (emit_setmem_word_loop): New.
3317 (emit_setmem_byte_loop): New.
3318 (nds32_expand_setmem_loop): New.
3319 (nds32_expand_setmem_loop_v3m): New.
3320 (nds32_expand_setmem_unroll): New.
3321 (nds32_expand_setmem): New.
3323 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3324 Chung-Ju Wu <jasonwucj@gmail.com>
3326 * config/nds32/nds32-memory-manipulation.c
3327 (nds32_emit_load_store): New.
3328 (nds32_emit_post_inc_load_store): New.
3329 (nds32_emit_mem_move): New.
3330 (nds32_emit_mem_move_block): New.
3331 (nds32_expand_movmemsi_loop_unknown_size): New.
3332 (nds32_expand_movmemsi_loop_known_size): New.
3333 (nds32_expand_movmemsi_loop): New.
3334 (nds32_expand_movmemsi_unroll): New.
3335 (nds32_expand_movmemqi): Rename ...
3336 (nds32_expand_movmemsi): ... to this.
3337 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3338 (movmemsi): ... to this.
3339 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3340 (nds32_expand_movmemsi): ... to this.
3342 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3343 Monk Chiang <sh.chiang04@gmail.com>
3344 Chung-Ju Wu <jasonwucj@gmail.com>
3346 * config/nds32/nds32-protos.h
3347 (nds32_expand_load_multiple): New arguments.
3348 (nds32_expand_store_multiple): Ditto.
3349 (nds32_valid_multiple_load_store): Rename ...
3350 (nds32_valid_multiple_load_store_p): ... to this.
3351 * config/nds32/nds32-memory-manipulation.c
3352 (nds32_expand_load_multiple): Refine implementation.
3353 (nds32_expand_store_multiple): Ditto.
3354 * config/nds32/nds32-multiple.md
3355 (load_multiple): Update nds32_expand_load_multiple interface.
3356 (store_multiple): Update nds32_expand_store_multiple interface.
3357 * config/nds32/nds32-predicates.c
3358 (nds32_valid_multiple_load_store): Rename ...
3359 (nds32_valid_multiple_load_store_p): ... to this and refine
3361 * config/nds32/predicates.md
3362 (nds32_load_multiple_and_update_address_operation): New predicate.
3363 (nds32_store_multiple_and_update_address_operation): New predicate.
3365 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3366 Chung-Ju Wu <jasonwucj@gmail.com>
3368 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3369 (combo): New attribute.
3370 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3372 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3374 * config/nds32/nds32.opt: Change -mcmodel= default value.
3376 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3377 Monk Chiang <sh.chiang04@gmail.com>
3378 Chung-Ju Wu <jasonwucj@gmail.com>
3380 * config/nds32/constants.md (unspec_element): New enum.
3381 * config/nds32/constraints.md (Umw): New constraint.
3382 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3383 * config/nds32/nds32-intrinsic.md: Likewise.
3384 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3385 (nds32_valid_smw_lwm_base_p): New.
3386 (nds32_output_smw_single_word): New.
3387 (nds32_output_lmw_single_word): New.
3388 (nds32_expand_unaligned_load): New.
3389 (nds32_expand_unaligned_store): New.
3390 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3391 (nds32_output_smw_single_word): Declare.
3392 (nds32_output_lmw_single_word): Declare.
3393 (nds32_expand_unaligned_load): Declare.
3394 (nds32_expand_unaligned_store): Declare.
3395 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3396 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3397 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3398 NDS32_BUILTIN_UASTORE_DW.
3399 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3402 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3403 Kito Cheng <kito.cheng@gmail.com>
3404 Chung-Ju Wu <jasonwucj@gmail.com>
3406 * config/nds32/nds32-intrinsic.c
3407 (nds32_expand_builtin_null_ftype_reg): Delete.
3408 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3409 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3410 (nds32_read_argument): New.
3411 (nds32_legitimize_target): Ditto.
3412 (nds32_legitimize_argument): Ditto.
3413 (nds32_check_constant_argument): Ditto.
3414 (nds32_expand_unop_builtin): Ditto.
3415 (nds32_expand_unopimm_builtin): Ditto.
3416 (nds32_expand_binop_builtin): Ditto.
3417 (nds32_builtin_decl_impl): Ditto.
3418 (builtin_description): Ditto.
3419 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3420 (nds32_init_builtins_impl): Ditto.
3421 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3422 (nds32_builtin_decl): New.
3423 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3424 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3426 2018-03-02 Jeff Law <law@redhat.com>
3428 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3429 (redundant_insn, fill_simple_delay_slots): Likewise.
3430 (fill_slots_from_thread): Likewise.
3431 * resource.c (mark_referenced_resources): Likewise.
3432 (mark_set_resources, find_dead_or_set_registers): Likewise.
3434 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3436 * substring-locations.h (format_warning_va): Formatting fix for
3438 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3440 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3441 * substring-locations.c: Include intl.h.
3442 (format_warning_va): Turned into small wrapper around
3443 format_warning_n_va, renamed to ...
3444 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3445 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3447 (format_warning_at_substring_n): New function.
3448 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3449 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3450 format_warning_at_substring with just a shorter name instead of
3451 const function pointer.
3452 (fmtwarn_n): New function.
3453 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3454 appropriate, get rid of all the fmtstr temporaries, move conditionals
3455 with G_() wrapped string literals directly into fmtwarn arguments,
3456 cast dir.len to (int), formatting fixes.
3458 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3460 * doc/invoke.texi: Remove "Cilk Plus" references.
3462 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3463 Richard Biener <rguenther@suse.de>
3466 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3467 for error or warning attributes if CALL_FROM_THUNK_P is set.
3470 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3473 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3474 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3477 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3478 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3480 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3481 instead of -1U in last predictors element's probability member.
3483 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3486 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3487 arguments if they are comparable.
3489 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3491 PR tree-optimization/84634
3492 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3493 masks and masked_loop_p with a single loop_masks, making sure it's
3494 null for bb vectorization.
3496 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3498 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3499 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3500 loop->force_vectorize to check whether there is no alias.
3502 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3505 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3507 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3509 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3510 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3511 instead of a loop around prev_real_insn.
3512 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3516 * config/i386/i386.c (ix86_print_operand): Use conditional
3517 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3520 2018-03-02 Richard Biener <rguenther@suse.de>
3522 PR tree-optimization/84427
3523 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3524 (bitmap_set_subtract_values): Rewrite to handle multiple
3527 (prune_clobbered_mems): Likewise.
3528 (phi_translate): Take edge instead of pred/phiblock.
3529 (phi_translate_1): Likewise.
3530 (phi_translate_set): Likewise. Insert all translated
3531 exprs for a value into the set, keeping possibly multiple
3532 expressions per value.
3533 (compute_antic_aux): Adjust for phi_translate changes.
3534 When intersecting union the expressions and prune those
3535 not in the final value set, keeping possibly multiple
3536 expressions per value. Do not use value-insertion
3537 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3538 all expressions. Add verification that the value-sets
3539 only shrink during iteration.
3540 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3541 (do_pre_regular_insertion): Likewise.
3542 (do_pre_partial_partial_insertion): Likewise.
3544 2018-03-02 Richard Biener <rguenther@suse.de>
3547 * config/darwin.c (saved_debug_info_level): New static global.
3548 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3549 (darwin_asm_lto_end): Restore debug info generation settings.
3551 2018-03-01 Martin Liska <mliska@suse.cz>
3554 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3557 2018-03-01 Richard Biener <rguenther@suse.de>
3560 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3561 type annotation with LTO when debug was disabled at compile-time.
3563 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3565 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3567 (mips_final_postscan_insn): Likewise.
3569 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3571 PR rtl-optimization/84528
3572 * alias.c (init_alias_target): Add commentary.
3573 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3574 a unique base value if the frame pointer is not eliminated
3575 to the stack pointer.
3577 2018-03-01 Tom de Vries <tom@codesourcery.com>
3579 PR rtl-optimization/83327
3580 * lra-int.h (hard_regs_spilled_into): Declare.
3581 * lra.c (hard_regs_spilled_into): Define.
3582 (init_reg_info): Init hard_regs_spilled_into.
3583 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3584 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3585 (process_bb_lives): Handle hard_regs_spilled_into.
3586 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3587 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3589 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3591 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3592 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3593 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3594 * config/rs6000/aix72.h: New file.
3596 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3598 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3599 instead of warning_at with conditional singular and plural messages
3603 * stor-layout.c (update_alignment_for_field): For
3604 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3605 && !DECL_PACKED (field), do the alignment update, just use
3606 only desired_align instead of MAX (type_align, desired_align)
3608 (place_field): Don't do known_align < desired_align handling
3609 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3610 is non-NULL, instead do it after rli->prev_field handling and
3611 only if not within a bitfield word. For DECL_PACKED (field)
3612 use type_align of BITS_PER_UNIT.
3614 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3616 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3617 superfluous parentheses and trailing spaces.
3619 2018-02-28 Richard Biener <rguenther@suse.de>
3621 PR tree-optimization/84584
3622 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3623 SCoPs with fake exit edge.
3625 2018-02-28 Martin Liska <mliska@suse.cz>
3628 * timevar.c (timer::print): Fix format to properly print 100%
3631 2018-02-28 Richard Biener <rguenther@suse.de>
3634 * genmatch.c (capture_info::walk_match): Do not mark
3635 captured expressions without operands as expr_p given
3636 they act more like predicates and should be subject to
3637 "lost tail" side-effect preserving.
3639 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3641 PR rtl-optimization/81611
3642 * auto-inc-dec.c (attempt_change): Move dead note from
3643 mem_insn if it's the next use of regno
3644 (find_address): Take address use of reg holding
3645 non-incremented value. Add parm to limit search to the named
3647 (merge_in_block): Attempt to use a mem insn that is the next
3648 use of the original regno.
3650 2018-02-27 Martin Sebor <msebor@redhat.com>
3653 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3654 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3656 2018-02-27 Martin Sebor <msebor@redhat.com>
3658 PR translation/84207
3659 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3660 n argument to unsigned HOST_WIDE_INT.
3661 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3662 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3663 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3664 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3666 2018-02-27 Richard Biener <rguenther@suse.de>
3668 PR tree-optimization/84512
3669 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3670 Do not use the estimate returned from record_stmt_cost for
3671 the scalar iteration cost but sum properly using add_stmt_cost.
3673 2018-02-27 Richard Biener <rguenther@suse.de>
3675 PR tree-optimization/84466
3676 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3677 Adjust last change to less strictly validate use operands.
3679 2018-02-27 Martin Liska <mliska@suse.cz>
3681 PR gcov-profile/84548
3682 * gcov.c (process_file): Allow partial overlap and consider it
3683 also as group functions.
3684 (output_lines): Properly calculate range of lines for a group.
3686 2018-02-27 Martin Liska <mliska@suse.cz>
3688 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3689 'ggc' suffixes. Change first column width.
3690 (timer::print): Fix formatting of the column.
3692 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3694 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3695 preserve inline entry blocks for the sake of debug inline
3696 entry point markers alone.
3697 (remove_unused_locals): Suggest in comments a better place to
3698 force the preservation of inline entry blocks that are
3699 otherwise unused, but do not preserve them.
3701 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3703 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3705 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3708 * config/i386/constraints.md (Bs): Replace
3709 ix86_indirect_branch_register with
3710 TARGET_INDIRECT_BRANCH_REGISTER.
3712 * config/i386/i386.md (indirect_jump): Likewise.
3713 (tablejump): Likewise.
3714 (*sibcall_memory): Likewise.
3715 (*sibcall_value_memory): Likewise.
3716 Peepholes of indirect call and jump via memory: Likewise.
3717 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3718 (*sibcall_value_GOT_32): Likewise.
3719 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3720 (GOT_memory_operand): Likewise.
3721 (call_insn_operand): Likewise.
3722 (sibcall_insn_operand): Likewise.
3723 (GOT32_symbol_operand): Likewise.
3724 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3726 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3728 PR rtl-optimization/83496
3729 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3730 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3731 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3732 redundant insn, if any.
3733 (relax_delay_slots): Likewise.
3734 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3736 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3738 PR tree-optimization/83965
3739 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3740 that grouped statements are part of a reduction chain. Return
3741 true if the statement is not marked as a reduction itself but
3743 (vect_recog_dot_prod_pattern): Don't check whether the statement
3744 is part of a group here.
3745 (vect_recog_sad_pattern): Likewise.
3746 (vect_recog_widen_sum_pattern): Likewise.
3748 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3751 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3753 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3756 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3758 (ix86_output_indirect_function_return): New prototype.
3759 (ix86_split_simple_return_pop_internal): Likewise.
3760 * config/i386/i386.c (indirect_return_via_cx): New.
3761 (indirect_return_via_cx_bnd): Likewise.
3762 (indirect_thunk_name): Handle return va CX_REG.
3763 (output_indirect_thunk_function): Create alias for
3764 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3765 (ix86_output_indirect_jmp): Remove the bool argument.
3766 (ix86_output_indirect_function_return): New function.
3767 (ix86_split_simple_return_pop_internal): Likewise.
3768 * config/i386/i386.md (*indirect_jump): Don't pass false
3769 to ix86_output_indirect_jmp.
3770 (*tablejump_1): Likewise.
3771 (simple_return_pop_internal): Change it to define_insn_and_split.
3772 Call ix86_split_simple_return_pop_internal to split it for
3774 (simple_return_indirect_internal): Call
3775 ix86_output_indirect_function_return instead of
3776 ix86_output_indirect_jmp.
3778 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3781 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3782 memset and value initialization afterwards.
3784 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3786 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3788 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3791 * common/config/aarch64/aarch64-common.c
3792 (aarch_option_optimization_table[]): Switch
3793 off fomit-frame-pointer
3795 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3796 Chung-Ju Wu <jasonwucj@gmail.com>
3798 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3800 (store_multiple): Ditto.
3802 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3804 * config.gcc: Add --with-cpu support for nds32 target.
3805 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3806 * config/nds32/nds32.opt: Add -mcpu= option.
3808 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3810 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3811 isel=yes): Warn for these deprecated options.
3813 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3815 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3816 ISA_2_5_MASKS_EMBEDDED.
3818 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3820 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3821 p->max as pointers rather than using iterative_hash_expr.
3823 2018-02-23 Carl Love <cel@us.ibm.com>
3825 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3826 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3828 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3829 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3830 P8V_BUILTIN_VEC_VUNSIGNED2.
3832 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3835 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3836 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3838 * lra-constraints.c (curr_insn_transform): Set up
3839 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3841 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3842 * lra-eliminations.c (spill_pseudos): Ditto.
3843 (process_insn_for_elimination): Ditto.
3844 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3845 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3846 LRA_NON_CLOBBERED_ALT.
3848 2018-02-22 Martin Sebor <msebor@redhat.com>
3850 PR tree-optimization/84480
3851 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3852 to maybe_diag_stxncpy_trunc. Call it.
3853 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3854 from gimple_fold_builtin_strcpy. Print inlining stack.
3855 (handle_builtin_stxncpy): Print inlining stack.
3856 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3858 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3861 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3862 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3863 and -fcheck-pointer-bounds are used together.
3864 (indirect_thunk_prefix): New enum.
3865 (indirect_thunk_need_prefix): New function.
3866 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3867 "_nt" instead of "_bnd" for NOTRACK prefix.
3868 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3869 (output_indirect_thunk_function): Likewise.
3871 (ix86_code_end): Update output_indirect_thunk_function calls.
3872 (ix86_output_indirect_branch_via_reg): Replace
3873 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3874 (ix86_output_indirect_branch_via_push): Likewise.
3875 (ix86_output_function_return): Likewise.
3876 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3877 incompatible with -fcf-protection=branch and
3878 -fcheck-pointer-bounds.
3880 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3883 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3884 Change gcc_assert call to output_operand_lossage.
3886 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3888 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3890 2018-02-22 DJ Delorie <dj@redhat.com>
3891 Sebastian Perta <sebastian.perta@renesas.com>
3892 Oleg Endo <olegendo@gcc.gnu.org>
3894 * config/rx/rx.c (rx_rtx_costs): New function.
3895 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3897 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3899 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3901 2018-02-22 Martin Liska <mliska@suse.cz>
3904 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3905 Add "native" as a possible value.
3907 2018-02-22 Martin Liska <mliska@suse.cz>
3910 * config/i386/i386.c (ix86_option_override_internal):
3911 Add "native" as a possible value for -march and -mtune.
3913 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3916 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3917 to all type variants.
3919 PR tree-optimization/84503
3920 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3921 width as info->bitpos + info->bitsize - start.
3922 (merged_store_group::merge_overlapping): Simplify width computation.
3923 (check_no_overlap): New function.
3924 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3925 start + width and last_order of the group, fail if check_no_overlap
3927 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3928 to group if check_no_overlap fails.
3930 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3932 * config/rs6000/altivec.md: Delete contraint arguments to
3933 define_expand, define_split, and define_peephole2, and in
3934 define_insn_and_split if always unused.
3935 * config/rs6000/darwin.md: Ditto.
3936 * config/rs6000/dfp.md: Ditto.
3937 * config/rs6000/rs6000.md: Ditto.
3938 * config/rs6000/sync.md: Ditto.
3939 * config/rs6000/vector.md: Ditto.
3940 * config/rs6000/vsx.md: Ditto.
3942 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3944 * config/rs6000/altivec.md: Write output control strings as braced
3945 blocks instead of double-quoted strings.
3946 * config/rs6000/darwin.md: Ditto.
3947 * config/rs6000/rs6000.md: Ditto.
3948 * config/rs6000/vector.md: Ditto.
3949 * config/rs6000/vsx.md: Ditto.
3951 2018-02-21 Jason Merrill <jason@redhat.com>
3953 PR c++/84314 - ICE with templates and fastcall attribute.
3954 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3956 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3958 * ipa-cp.c (determine_versionability): Fix comment typos.
3960 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3963 * ipa-cp.c (determine_versionability): Do not version functions caling
3966 2018-02-21 Martin Liska <mliska@suse.cz>
3969 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3970 Add "native" as a possible value.
3971 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3972 the macro when native cpu detection is available.
3974 2018-02-21 Martin Liska <mliska@suse.cz>
3977 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3978 Add "native" as a possible value.
3979 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3980 when native cpu detection is available.
3982 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3983 Martin Sebor <msebor@redhat.com>
3985 PR tree-optimization/84478
3986 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3988 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3989 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3990 support which is conservatively correct, for 2 only stay conservative
3991 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3992 argument to the 2 argument get_range_strlen, adjust 6 arg
3993 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3995 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3996 (gimple_fold_builtin_strlen): Pass true as last argument to
3999 2018-02-20 Martin Sebor <msebor@redhat.com>
4002 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4003 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4004 (builtin_memref::builtin_memref): Factor out parts into
4005 set_base_and_offset and call it.
4007 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4010 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4011 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4012 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4013 search at the associated MODE_INT.
4015 2018-02-20 Jeff Law <law@redhat.com>
4018 PR tree-optimization/81592
4020 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4021 for range data rather than using global data.
4022 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4023 range data rather than using global data.
4024 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4025 pass it to children as needed.
4026 (struct directive::fmtresult): Similarly.
4027 (struct directive::set_width): Similarly.
4028 (struct directive::set_precision): Similarly.
4029 (format_integer, format_directive, parse_directive): Similarly.
4030 (format_none): Accept unnamed vr_values parameter.
4031 (format_percent, format_floating, format_character): Similarly.
4032 (format_string, format_plain): Similarly.
4033 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4034 the EVRP range analyzer for range data rather than using global data.
4035 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4036 gimple-ssa-evrp-analyze.h
4037 (class sprintf_dom_walker): Add after_dom_children member function.
4038 Add evrp_range_analyzer member.
4039 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4040 range analyzer as needed.
4041 (sprintf_dom_walker::after_dom_children): New member function.
4042 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4044 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4045 (evrp_range_analyzer::pop_to_marker): Likewise.
4047 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4049 PR tree-optimization/84419
4050 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4051 with the required type if its current type is compatible but
4054 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4057 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4058 after vectorization.
4060 2018-02-20 Martin Liska <mliska@suse.cz>
4063 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4064 possible values if we don't have a hint.
4066 2018-02-20 Martin Liska <mliska@suse.cz>
4070 * final.c (shorten_branches): Build align_tab array with one
4072 * opts.c (finish_options): Add alignment option limit check.
4073 (MAX_CODE_ALIGN): Likewise.
4074 (MAX_CODE_ALIGN_VALUE): Likewise.
4075 * doc/invoke.texi: Document maximum allowed option value for
4076 all -falign-* options.
4078 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4081 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4082 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4083 * var-tracking.c (emit_note_insn_var_location): Remove all references
4084 to NOTE_INSN_CALL_ARG_LOCATION.
4085 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4086 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4087 Use copy_rtx_if_shared.
4088 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4089 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4090 (dwarf2out_var_location): Remove handling of
4091 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4093 * final.c (final_scan_insn): Remove all references to
4094 NOTE_INSN_CALL_ARG_LOCATION.
4095 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4096 before dumping final insns.
4097 * except.c (emit_note_eh_region_end): Remove all references to
4098 NOTE_INSN_CALL_ARG_LOCATION.
4099 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4100 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4101 * config/arc/arc.c (hwloop_optimize): Likewise.
4102 * config/arm/arm.c (create_fix_barrier): Likewise.
4103 * config/s390/s390.c (s390_chunkify_start): Likewise.
4104 * config/sh/sh.c (find_barrier): Likewise.
4105 * config/i386/i386.c (rest_of_insert_endbranch,
4106 ix86_seh_fixup_eh_fallthru): Likewise.
4107 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4108 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4109 * config/frv/frv.c (frv_function_prologue): Likewise.
4110 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4112 (note_outside_basic_block_p): Remove all references to
4113 NOTE_INSN_CALL_ARG_LOCATION.
4114 * gengtype.c (adjust_field_rtx_def): Likewise.
4115 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4117 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4118 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4121 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4124 PR tree-optimization/84452
4125 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4126 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4129 2018-02-19 Martin Liska <mliska@suse.cz>
4132 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4134 2018-02-19 Martin Liska <mliska@suse.cz>
4135 Richard Sandiford <richard.sandiford@linaro.org>
4137 PR tree-optimization/82491
4138 * gimple-fold.c (get_base_constructor): Make earlier bail out
4141 2018-02-19 Carl Love <cel@us.ibm.com>
4143 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4144 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4146 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4147 P8V_BUILTIN_VEC_NEG.
4149 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4151 * config/rl78/rl78.md (movdf): New define expand.
4153 2018-02-19 Martin Liska <mliska@suse.cz>
4156 * doc/invoke.texi: Fix typo.
4157 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4159 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4161 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4162 handle rs6000_single_float and rs6000_double_float specially for
4165 2018-02-16 Jeff Law <law@redhat.com>
4167 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4170 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4173 * ipa-inline.c (inline_small_functions): Fix a typo.
4175 2018-02-16 Nathan Sidwell <nathan@acm.org>
4177 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4179 2018-02-16 Carl Love <cel@us.ibm.com>
4181 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4182 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4183 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4184 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4185 expansion to P8V_BUILTIN_VEC_FLOAT2.
4187 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4189 PR rtl-optimization/70023
4190 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4191 src_regno into account.
4193 2018-02-16 Carl Love <cel@us.ibm.com>
4195 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4196 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4197 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4198 * config/rs6000/rs6000.c: Remove case statements for
4199 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4200 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4201 and P9V_BUILTIN_VEC_VINSERT4B.
4202 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4203 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4204 * config/rs6000/vsx.md:
4205 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4208 2018-02-16 Carl Love <cel@us.ibm.com>
4210 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4212 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4214 * config/rs6000/rs6000-c.c: Add the definitions for
4215 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4216 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4217 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4218 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4219 definition for insert4b and define insn *insert3b_internal.
4220 * doc/extend.texi: Add documentation for vec_extract4b.
4222 2018-02-16 Nathan Sidwell <nathan@acm.org>
4224 * doc/extend.texi (Backwards Compatibility): Mention friend
4225 injection. Note for-scope is deprecated.
4226 * doc/invoke.texi (-ffriend-injection): Deprecate.
4228 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4230 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4231 that moved to I2, also allow destinations that are a paradoxical
4232 subreg (instead of a normal reg).
4234 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4237 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4240 2018-02-16 Richard Biener <rguenther@suse.de>
4242 PR tree-optimization/84037
4243 PR tree-optimization/84016
4245 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4246 Adjust vec_construct for the fact we need additional higher latency
4247 128bit inserts for AVX256 and AVX512 vector builds.
4248 (ix86_add_stmt_cost): Scale vector construction cost for
4251 2018-02-16 Richard Biener <rguenther@suse.de>
4253 PR tree-optimization/84417
4254 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4255 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4256 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4258 2018-02-16 Martin Liska <mliska@suse.cz>
4261 * internal-fn.def (ASAN_CHECK): Set proper flags.
4262 (ASAN_MARK): Likewise.
4264 2018-02-16 Julia Koval <julia.koval@intel.com>
4266 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4267 from PTA_CANNONLAKE.
4269 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4272 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4273 Use ++iter rather than iter++ for std::list iterators.
4274 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4275 defer deleting them until all nodes in the forest are processed. Do
4276 free even leaf nodes. Change to_process into auto_vec.
4279 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4280 * vec.h (vec_default_construct): Use memset instead of placement new
4281 if BROKEN_VALUE_INITIALIZATION is defined.
4282 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4283 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4286 PR rtl-optimization/83723
4287 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4288 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4289 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4291 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4293 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4295 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4297 PR rtl-optimization/81443
4298 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4299 from inner REGs to paradoxical SUBREGs.
4301 2018-02-16 Richard Biener <rguenther@suse.de>
4303 PR tree-optimization/84399
4304 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4305 For operands we can analyze at their definition make sure we can
4306 analyze them at each use as well.
4308 2018-02-16 Richard Biener <rguenther@suse.de>
4310 PR tree-optimization/84190
4311 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4312 volatile accesses if the decl isn't volatile.
4314 2018-02-15 Jason Merrill <jason@redhat.com>
4316 PR c++/84314 - ICE with templates and fastcall attribute.
4317 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4318 TYPE_CANONICAL on an existing type.
4320 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4322 PR tree-optimization/84383
4323 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4324 dstoff nor call operand_equal_p if dstbase is NULL.
4326 PR tree-optimization/84334
4327 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4328 also a CONSTANT_CLASS_P, punt.
4330 2018-02-14 Jim Wilson <jimw@sifive.com>
4332 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4333 first SMALL_OPERAND check. New local min_second_step. Move assert
4334 to where locals are set. Add TARGET_RVC support.
4335 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4337 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4339 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4341 2018-02-14 Martin Sebor <msebor@redhat.com>
4343 PR tree-optimization/83698
4344 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4345 arrays constrain the offset range to their bounds.
4346 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4347 (builtin_access::overlap): Avoid setting the size of overlap if it's
4349 (maybe_diag_overlap): Also consider arrays when deciding what values
4350 of offsets to include in diagnostics.
4352 2018-02-14 Martin Sebor <msebor@redhat.com>
4355 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4356 that correspond to the kind of a declaration.
4358 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4361 * config/pa/pa.md: Load address of PIC label using the linkage table
4362 if the label is nonlocal.
4364 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4366 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4367 warning message if user requests -maltivec=be.
4368 * doc/invoke.texi: Document deprecation of -maltivec=be.
4370 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4373 * config/rs6000/rs6000-c.c: Update definitions for
4374 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4375 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4377 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4380 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4381 add _get_ssp intrinsics. Remove argument from
4382 __builtin_ia32_rdssp[d|q].
4383 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4384 * config/i386/i386-builtin.def: Remove argument from
4385 __builtin_ia32_rdssp[d|q].
4386 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4387 ix86_expand_special_args_builtin for _rdssp[d|q].
4388 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4389 Clear register before usage.
4390 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4391 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4393 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4395 PR tree-optimization/84357
4396 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4397 operand 1 of an ARRAY_REF too.
4399 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4402 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4403 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4405 (set_of_reg): New struct.
4406 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4407 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4408 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4410 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4411 Split into bitclr, bitset, bitinvert patterns if appropriate.
4412 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4413 use rx_fuse_in_memory_bitop.
4414 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4415 to named insn, correct maximum insn length.
4417 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4420 * machmode.def: Define a complex mode for PARTIAL_INT.
4421 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4423 * doc/rtl.texi: Document CSPImode.
4424 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4426 (msp430_hard_regno_nregs_with_padding): Likewise.
4428 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4431 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4433 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4435 PR rtl-optimization/84169
4436 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4437 we generated a parallel as new i3 and we split that to new i2 and i3
4438 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4439 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4440 those to i2, not i1. Partially rewrite this scan code.
4442 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4445 * stor-layout.c (place_field): For variable length fields, adjust
4446 offset_align afterwards not just based on the field's alignment,
4447 but also on the size.
4450 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4451 of exps and logs in the use_exp2 case.
4453 2018-02-13 Jeff Law <law@redhat.com>
4455 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4458 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4461 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4465 * common.opt (gas-loc-support, gas-locview-support): New.
4466 (ginline-points, ginternal-reset-location-views): New.
4467 * doc/invoke.texi: Document them. Use @itemx where intended.
4468 (gvariable-location-views): Adjust.
4469 * target.def (reset_location_view): New.
4470 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4471 (TARGET_RESET_LOCATION_VIEW): New.
4472 * doc/tm.texi: Rebuilt.
4473 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4474 (dwarf2out_default_as_locview_support): New.
4475 (output_asm_line_debug_info): Use option variables.
4476 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4477 (output_loc_list): Likewise.
4478 (add_high_low_attributes): Check option variables.
4479 Don't output entry view attribute in strict mode.
4480 (gen_inlined_subroutine_die): Check option variables.
4481 (dwarf2out_inline_entry): Likewise.
4482 (init_sections_and_labels): Likewise.
4483 (dwarf2out_early_finish): Likewise.
4484 (maybe_reset_location_view): New, from...
4485 (dwarf2out_var_location): ... here. Call it.
4486 * debug.h (dwarf2out_default_as_loc_support): Declare.
4487 (dwarf2out_default_as_locview_support): Declare.
4488 * hooks.c (hook_int_rtx_insn_0): New.
4489 * hooks.h (hook_int_rtx_insn_0): Declare.
4490 * toplev.c (process_options): Take -gas-loc-support and
4491 -gas-locview-support from dwarf2out. Enable
4492 -gvariable-location-views by default only with locview
4493 assembler support. Enable -ginternal-reset-location-views by
4494 default only if the target defines the corresponding hook.
4495 Enable -ginline-points by default if location views are
4496 enabled; force it disabled if statement frontiers are
4498 * tree-inline.c (expand_call_inline): Check option variables.
4499 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4501 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4503 PR tree-optimization/84321
4504 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4505 handling. Also check whether the anti-range contains any values
4506 that satisfy the mask; switch to a VR_RANGE if not.
4508 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4511 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4513 2018-02-13 Martin Jambor <mjambor@suse.cz>
4516 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4517 of call statements, also set location of a load to a temporary.
4519 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4521 * config/rl78/rl78.c (add_vector_labels): New function.
4522 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4523 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4524 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4525 which checks that no arguments are passed.
4526 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4527 * doc/extend.texi: Documentation for the new attribute.
4529 2018-02-13 Andreas Schwab <schwab@suse.de>
4531 * config/riscv/linux.h (CPP_SPEC): Define.
4533 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4536 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4537 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4538 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4539 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4540 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4541 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4542 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4543 temporarily for AES and PCLMUL builtins.
4545 PR tree-optimization/84339
4546 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4547 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4551 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4552 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4553 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4555 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4557 * omp-simd-clone.h: New file.
4558 * omp-simd-clone.c: Include omp-simd-clone.h.
4559 (expand_simd_clones): No longer static.
4560 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4561 cgraph.h and omp-simd-clone.h.
4562 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4563 (vect_recog_widen_shift_pattern): Formatting fix.
4564 (vect_pattern_recog_1): Don't check optab for calls.
4567 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4568 operands[2] into a REG before using gen_lowpart on it.
4570 2018-02-12 Jeff Law <law@redhat.com>
4573 * config/sh/sh.c (find_barrier): Consider a sibling call
4576 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4577 successfully back substituting a reg.
4579 2018-02-12 Richard Biener <rguenther@suse.de>
4581 PR tree-optimization/84037
4582 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4583 parameter, move visited init to caller.
4584 (vect_slp_analyze_operations): Separate cost from validity
4585 check, initialize visited once for all instances.
4586 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4588 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4590 (vect_model_promotion_demotion_cost): Likewise.
4591 (vectorizable_bswap): Guard cost modeling with !slp_node
4592 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4594 (vectorizable_call): Likewise.
4595 (vectorizable_conversion): Likewise.
4596 (vectorizable_assignment): Likewise.
4597 (vectorizable_shift): Likewise.
4598 (vectorizable_operation): Likewise.
4599 (vectorizable_store): Likewise.
4600 (vectorizable_load): Likewise.
4601 (vectorizable_condition): Likewise.
4602 (vectorizable_comparison): Likewise.
4604 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4607 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4608 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4609 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4611 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4614 * params.def (inline-min-speedup): Increase from 8 to 15.
4615 (max-inline-insns-auto): Decrease from 40 to 30.
4616 * ipa-split.c (consider_split): Add some buffer for function to
4617 be considered inlining candidate.
4618 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4621 2018-02-12 Richard Biener <rguenther@suse.de>
4623 PR tree-optimization/84037
4624 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4625 matched stmts if we cannot swap the non-matched ones.
4627 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4629 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4630 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4631 _mm_maskz_scalef_round_ss): New intrinsics.
4632 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4633 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4634 __builtin_ia32_scalefss_round): Remove.
4635 (__builtin_ia32_scalefsd_mask_round,
4636 __builtin_ia32_scalefss_mask_round): New intrinsics.
4637 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4638 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4639 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4640 "<round_constraint>")): Changed to ...
4641 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4642 "<round_scalar_constraint>")): ... this.
4643 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4644 %0, %1, %2<round_op3>}"): Changed to ...
4645 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4646 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4647 %2<round_scalar_mask_op3>}"): ... this.
4648 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4650 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4652 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4653 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4654 (_mm_maskz_sqrt_round_ss): New intrinsics.
4655 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4656 (__builtin_ia32_sqrtsd_mask_round)
4657 (__builtin_ia32_sqrtss_mask_round): New builtins.
4658 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4659 (__builtin_ia32_sqrtss_round): Remove.
4660 (__builtin_ia32_sqrtsd_mask_round)
4661 (__builtin_ia32_sqrtss_mask_round): New builtins.
4662 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4663 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4664 ((match_operand:VF_128 1 "vector_operand"
4665 "xBm,<round_constraint>")): Changed to ...
4666 ((match_operand:VF_128 1 "vector_operand"
4667 "xBm,<round_scalar_constraint>")): ... this.
4668 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4669 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4670 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4671 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4672 %<iptr>1<round_scalar_mask_op3>}): ... this.
4673 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4674 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4676 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4679 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4680 Cast vec_cmpeq result to correct type.
4681 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4682 Cast vec_cmpgt result to correct type.
4684 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4686 * final.c (final_scan_insn_1): Renamed from...
4687 (final_scan_insn): ... this. New wrapper, to recover
4688 seen from the outermost call in recursive ones.
4689 * config/sparc/sparc.c (output_return): Drop seen from call.
4690 (output_sibcall): Likewise.
4691 * config/visium/visium.c (output_branch): Likewise.
4693 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4695 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4698 2018-02-10 Alan Modra <amodra@gmail.com>
4701 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4702 Specify LR as an input.
4704 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4707 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4708 remove_member_access_dummy_vars): New functions.
4709 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4710 lower_omp_1, execute_lower_omp): Use them.
4712 PR rtl-optimization/84308
4713 * shrink-wrap.c (spread_components): Release todo vector.
4715 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4717 PR rtl-optimization/57193
4718 * ira-color.c (struct allocno_color_data): Add member
4719 conflict_allocno_hard_prefs.
4720 (update_conflict_allocno_hard_prefs): New.
4721 (bucket_allocno_compare_func): Add a preference based on
4722 conflict_allocno_hard_prefs.
4723 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4724 (color_allocnos): Remove a dead code. Initiate
4725 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4727 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4730 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4731 constraint from =wa to wa. Avoid a subreg on the output operand,
4732 instead use a pseudo and subreg it in a move.
4733 (p9_xxbrd_<mode>): Changed to ...
4734 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4735 (p9_xxbrd_v2df): New expander.
4736 (p9_xxbrw_<mode>): Changed to ...
4737 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4738 (p9_xxbrw_v4sf): New expander.
4740 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4742 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4744 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4747 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4748 multiply in 32-bit mode.
4749 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4750 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4753 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4755 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4756 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4757 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4758 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4760 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4763 * dwarf2out.c (is_trivial_indirect_ref): New function.
4764 (dwarf2out_late_global_decl): Do not generate a location
4765 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4766 and that are not defined in the current unit.
4768 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4770 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4771 instead of a libcall for UNORDERED.
4773 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4776 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4777 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4779 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4782 * config/s390/s390.c (s390_set_current_function): Invoke
4783 s390_indirect_branch_settings also if fndecl didn't change.
4785 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4787 * config/rs6000/rs6000.md (blockage): Set length to zero.
4789 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4791 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4793 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4796 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4797 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4801 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4802 PARALLEL incoming that failed vt_get_decl_and_offset check.
4805 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4806 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4808 (get_variable_section): For decls in named .bss* sections pass true as
4809 second argument to bss_initializer_p.
4811 2018-02-09 Marek Polacek <polacek@redhat.com>
4812 Jakub Jelinek <jakub@redhat.com>
4815 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4816 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4817 Sync some changes from cxx_fold_indirect_ref.
4819 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4821 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4823 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4824 (BLOCK_INLINE_ENTRY_LABEL): New.
4825 (dwarf2out_var_location): Disregard inline entry markers.
4826 (inline_entry_data): New struct.
4827 (inline_entry_data_hasher): New hashtable type.
4828 (inline_entry_data_hasher::hash): New.
4829 (inline_entry_data_hasher::equal): New.
4830 (inline_entry_data_table): New variable.
4831 (add_high_low_attributes): Add DW_AT_entry_pc and
4832 DW_AT_GNU_entry_view attributes if a pending entry is found
4833 in inline_entry_data_table. Add old entry_pc attribute only
4834 if debug nonbinding markers are disabled.
4835 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4836 markers are enabled.
4837 (block_within_block_p, dwarf2out_inline_entry): New.
4838 (dwarf2out_finish): Check that no entries remained in
4839 inline_entry_data_table.
4840 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4841 (final_scan_insn, notice_source_line): Likewise.
4842 (rest_of_clean_state): Skip inline entry markers.
4843 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4845 * gimple.c (gimple_build_debug_inline_entry): New.
4846 * gimple.h (enum gimple_debug_subcode): Add
4847 GIMPLE_DEBUG_INLINE_ENTRY.
4848 (gimple_build_debug_inline_entry): Declare.
4849 (gimple_debug_inline_entry_p): New.
4850 (gimple_debug_nonbind_marker_p): Adjust.
4851 * insn-notes.def (INLINE_ENTRY): New.
4852 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4853 inline entry marker notes.
4854 (print_insn): Likewise.
4855 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4856 (INSN_DEBUG_MARKER_KIND): Likewise.
4857 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4858 * tree-inline.c (expand_call_inline): Build and insert
4859 debug_inline_entry stmt.
4860 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4861 inline entry blocks early, if nonbind markers are enabled.
4862 (dump_scope_block): Dump fragment info.
4863 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4864 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4865 (gimple_build_debug_inline_entry): New.
4866 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4867 Enable/disable inline entry points too.
4868 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4869 (DEBUG_INSN): Describe inline entry markers.
4871 * common.opt (gvariable-location-views): New.
4872 (gvariable-location-views=incompat5): New.
4873 * config.in: Rebuilt.
4874 * configure: Rebuilt.
4875 * configure.ac: Test assembler for view support.
4876 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4877 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4878 * dwarf2out.c (var_loc_view): New typedef.
4879 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4880 (dwarf2out_locviews_in_attribute): New.
4881 (dwarf2out_locviews_in_loclist): New.
4882 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4883 (enum dw_line_info_opcode): Add LI_adv_address.
4884 (struct dw_line_info_table): Add view.
4885 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4886 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4887 (zero_view_p): New variable.
4888 (ZERO_VIEW_P): New macro.
4889 (output_asm_line_debug_info): New.
4890 (struct var_loc_node): Add view.
4891 (add_AT_view_list, AT_loc_list): New.
4892 (add_var_loc_to_decl): Add view param. Test it against last.
4893 (new_loc_list): Add view params. Record them.
4894 (AT_loc_list_ptr): Handle loc and view lists.
4895 (view_list_to_loc_list_val_node): New.
4896 (print_dw_val): Handle dw_val_class_view_list.
4897 (size_of_die): Likewise.
4898 (value_format): Likewise.
4899 (loc_list_has_views): New.
4900 (gen_llsym): Set vl_symbol too.
4901 (maybe_gen_llsym, skip_loc_list_entry): New.
4902 (dwarf2out_maybe_output_loclist_view_pair): New.
4903 (output_loc_list): Output view list or entries too.
4904 (output_view_list_offset): New.
4905 (output_die): Handle dw_val_class_view_list.
4906 (output_dwarf_version): New.
4907 (output_compilation_unit_header): Use it.
4908 (output_skeleton_debug_sections): Likewise.
4909 (output_rnglists, output_line_info): Likewise.
4910 (output_pubnames, output_aranges): Update version comments.
4911 (output_one_line_info_table): Output view numbers in asm comments.
4912 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4913 Call maybe_gen_llsym.
4914 (loc_list_from_tree_1): Adjust.
4915 (add_AT_location_description): Create view list attribute if
4916 needed, check it's absent otherwise.
4917 (convert_cfa_to_fb_loc_list): Adjust.
4918 (maybe_emit_file): Call output_asm_line_debug_info for test.
4919 (dwarf2out_var_location): Reset views as needed. Precompute
4920 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4921 attribute. Set view.
4922 (new_line_info_table): Reset next view.
4923 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4924 (dwarf2out_source_line): Likewise. Output view resets and labels to
4925 the assembler, or select appropriate line info opcodes.
4926 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4927 (optimize_string_length): Catch it. Adjust.
4928 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4929 dw_val_class_view_list, and remove it if no longer needed.
4930 (hash_loc_list): Hash view numbers.
4931 (loc_list_hasher::equal): Compare them.
4932 (optimize_location_lists): Check whether a view list symbol is
4933 needed, and whether the locview attribute is present, and
4934 whether they match. Remove the locview attribute if no longer
4936 (index_location_lists): Call skip_loc_list_entry for test.
4937 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4938 Use output_dwarf_version.
4939 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4940 (struct dw_val_node): Add val_view_list.
4941 * final.c (SEEN_NEXT_VIEW): New.
4942 (set_next_view_needed): New.
4943 (clear_next_view_needed): New.
4944 (maybe_output_next_view): New.
4945 (final_start_function): Rename to...
4946 (final_start_function_1): ... this. Take pointer to FIRST,
4947 add SEEN parameter. Emit param bindings in the initial view.
4948 (final_start_function): Reintroduce SEEN-less interface.
4949 (final): Rename to...
4950 (final_1): ... this. Take SEEN parameter. Output final pending
4951 next view at the end.
4952 (final): Reintroduce seen-less interface.
4953 (final_scan_insn): Output pending next view before switching
4954 sections or ending a block. Mark the next view as needed when
4955 outputting variable locations. Notify debug backend of section
4956 changes, and of location view changes.
4957 (rest_of_handle_final): Adjust.
4958 * toplev.c (process_options): Autodetect value for debug variable
4959 location views option. Warn on incompat5 without -gdwarf-5.
4960 * doc/invoke.texi (gvariable-location-views): New.
4961 (gvariable-location-views=incompat5): New.
4962 (gno-variable-location-views): New.
4964 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4966 PR tree-optimization/84136
4967 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4968 that the result of find_edge is non-NULL.
4970 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4973 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4974 storing integer register in SImode. Fix cost of 256 and 512
4975 byte aligned SSE register store.
4977 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4979 * config/i386/i386.c (ix86_multiplication_cost): Fix
4980 multiplication cost for TARGET_AVX512DQ.
4982 2018-02-08 Marek Polacek <polacek@redhat.com>
4984 PR tree-optimization/84238
4985 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4988 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4990 PR tree-optimization/84265
4991 * tree-vect-stmts.c (vectorizable_store): Don't treat
4992 VMAT_CONTIGUOUS accesses as grouped.
4993 (vectorizable_load): Likewise.
4995 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4997 PR tree-optimization/81635
4998 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4999 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5000 (test_round_for_mask): New functions.
5001 (wide_int_cc_tests): Call test_round_for_mask.
5002 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5003 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5004 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5005 range returned by get_range_info.
5007 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5010 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5011 * symtab.c: Include builtins.h
5012 (symtab_node::output_to_lto_symbol_table_p): Move here
5013 from lto-streamer-out.c:output_symbol_p.
5014 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5015 (output_symbol_p): Move all logic to symtab.c
5016 (produce_symtab): Update.
5018 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5020 * config/s390/s390-opts.h (enum indirect_branch): Define.
5021 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5022 (s390_indirect_branch_via_thunk)
5023 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5024 (enum s390_indirect_branch_type): Define.
5025 * config/s390/s390.c (struct s390_frame_layout, struct
5026 machine_function): Remove.
5027 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5028 (indirect_branch_table_label_no, indirect_branch_table_name):
5030 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5031 (enum s390_indirect_branch_option): Define.
5032 (s390_return_addr_from_memory): New function.
5033 (s390_handle_string_attribute): New function.
5034 (s390_attribute_table): Add new attribute handler.
5035 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5036 (s390_indirect_branch_via_thunk): New function.
5037 (s390_indirect_branch_via_inline_thunk): New function.
5038 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5039 sibling call optimization for non z10 compiles.
5040 (s390_emit_call): Force indirect branch target to be a single
5041 register. Add r1 clobber for non-z10 compiles.
5042 (s390_emit_epilogue): Emit return jump via return_use expander.
5043 (s390_reorg): Handle JUMP_INSNs as execute targets.
5044 (s390_option_override_internal): Perform validity checks for the
5045 new command line options.
5046 (s390_indirect_branch_attrvalue): New function.
5047 (s390_indirect_branch_settings): New function.
5048 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5049 (s390_output_indirect_thunk_function): New function.
5050 (s390_code_end): Implement target hook.
5051 (s390_case_values_threshold): Implement target hook.
5052 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5054 * config/s390/s390.h (struct s390_frame_layout)
5055 (struct machine_function): Move here from s390.c.
5056 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5057 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5058 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5059 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5060 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5061 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5062 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5063 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5064 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5065 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5066 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5067 (mnemonic attribute): Add values which aren't recognized
5069 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5070 pattern for branch conversion. Fix mnemonic attribute.
5071 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5072 indirect branch via thunk if requested.
5073 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5074 ("*indirect_jump"): Disable for branch conversion using out of
5076 ("indirect_jump_via_thunk<mode>_z10")
5077 ("indirect_jump_via_thunk<mode>")
5078 ("indirect_jump_via_inlinethunk<mode>_z10")
5079 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5080 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5081 ("casesi_jump_via_inlinethunk<mode>_z10")
5082 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5083 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5084 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5085 ("*indirect2_jump"): Disable for branch conversion.
5086 ("casesi_jump"): Turn into expander and expand patterns for branch
5088 ("return_use"): New expander.
5089 ("*return"): Emit return via thunk and rename it to ...
5090 ("*return<mode>"): ... this one.
5091 * config/s390/s390.opt: Add new options and and enum for the
5094 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5096 * lra-constraints.c (match_reload): Unconditionally use
5097 gen_lowpart_SUBREG, rather than selecting between that
5098 and equivalent gen_rtx_SUBREG code.
5100 2018-02-08 Richard Biener <rguenther@suse.de>
5102 PR tree-optimization/84233
5103 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5104 changed flag instead of boguously re-using phi_inserted.
5106 2018-02-08 Martin Jambor <mjambor@suse.cz>
5108 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5109 static local variables.
5111 2018-02-08 Richard Biener <rguenther@suse.de>
5113 PR tree-optimization/84278
5114 * tree-vect-stmts.c (vectorizable_store): When looking for
5115 smaller vector types to perform grouped strided loads/stores
5116 make sure the mode is supported by the target.
5117 (vectorizable_load): Likewise.
5119 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5121 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5122 Increase LDP/STP opportunities by adding adjacent callee-saves.
5124 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5126 PR rtl-optimization/84068
5127 PR rtl-optimization/83459
5128 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5130 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5132 PR tree-optimization/84224
5133 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5134 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5137 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5140 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5141 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5142 count, remove one USE.
5144 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5146 * doc/install.texi (Configuration): Document the
5147 --with-long-double-format={ibm,ieee} PowerPC configuration
5151 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5152 Convert from define_expand to be define_insn_and_split. Rework
5153 float/double/_Float128 conversions to QI/HI/SImode to work with
5154 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5155 conversions to QI/HImode types did a store and then a load to
5156 truncate the value. For conversions to VSX registers, don't split
5157 the insn, instead emit the code directly. Use the code iterator
5158 any_fix to combine signed and unsigned conversions.
5159 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5160 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5161 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5162 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5163 (fix_<mode>di2_hw): Likewise.
5164 (fixuns_<mode>di2_hw): Likewise.
5165 (fix_<mode>si2_hw): Likewise.
5166 (fixuns_<mode>si2_hw): Likewise.
5167 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5168 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5169 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5170 fix<uns>_trunc<SFDF:mode>si2_p8.
5171 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5173 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5174 (fix<uns>_<mode>_mem): Likewise.
5175 (fctiw<u>z_<mode>_mem): Likewise.
5176 (fix<uns>_<mode>_mem): Likewise.
5177 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5178 the register allocator from doing a direct move to the GPRs to do
5179 a store, and instead use the ISA 3.0 store byte/half-word from
5180 vector register instruction. For IEEE 128-bit floating point,
5181 also optimize stores of 32-bit ints.
5182 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5184 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5186 * genextract.c (push_pathstr_operand): New function to support
5188 (walk_rtx): Call push_pathstr_operand.
5189 (print_path): Support [a-zA-Z].
5191 2018-02-07 Richard Biener <rguenther@suse.de>
5193 PR tree-optimization/84037
5194 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5195 (cse_and_gimplify_to_preheader): Declare.
5196 (vect_get_place_in_interleaving_chain): Likewise.
5197 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5199 (_loop_vec_info::~_loop_vec_info): Delete it.
5200 (cse_and_gimplify_to_preheader): New function.
5201 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5202 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5203 (vectorizable_load): Likewise. For grouped stores always base
5204 the IV on the first element.
5205 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5206 condition before gimplifying.
5208 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5210 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5211 *DIV_EXPR and *MOD_EXPR.
5213 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5216 * config/i386/i386.c (ix86_option_override_internal): Mask out
5217 the CF_SET bit when checking -fcf-protection.
5219 2018-02-07 Tom de Vries <tom@codesourcery.com>
5222 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5225 2018-02-07 Richard Biener <rguenther@suse.de>
5227 PR tree-optimization/84204
5228 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5231 PR tree-optimization/84205
5232 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5233 special-case isl_ast_op_zdiv_r.
5235 PR tree-optimization/84223
5236 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5237 Only add conditions from within the region.
5238 (gather_bbs::after_dom_children): Adjust.
5240 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5243 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5244 * config/avr/avr.md: Only post-reload split REG-REG moves if
5245 either register is GENERAL_REG_P.
5247 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5249 PR tree-optimization/84235
5250 * tree-ssa-scopedtables.c
5251 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5252 if the subtraction is performed in floating point type where NaNs are
5253 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5254 build 1. Formatting fix.
5256 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5259 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5260 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5261 and skip it regardless of bb boundaries. Use CALL_P macro,
5262 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5263 unnecessarily, formatting fix.
5265 2018-02-06 Michael Collison <michael.collison@arm.com>
5267 * config/arm/thumb2.md:
5268 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5269 (*thumb_mov_notscc): Ditto.
5271 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5274 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5275 unsigned_fix, not "s".
5277 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5279 * configure.ac (gcc_fn_eh_frame_ro): New function.
5280 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5281 correct .eh_frame permissions.
5282 * configure: Regenerate.
5284 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5286 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5289 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5291 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5292 Display warning message for -mno-speculate-indirect-jumps.
5294 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5296 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5298 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5300 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5302 PR tree-optimization/84225
5303 * tree-eh.c (find_trapping_overflow): Only call
5304 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5306 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5309 * config/i386/i386.c: Reimplement the check of possible options
5310 -mibt/-mshstk conbination. Change error messages.
5311 * doc/invoke.texi: Fix a typo: remove extra '='.
5313 2018-02-06 Marek Polacek <polacek@redhat.com>
5315 PR tree-optimization/84228
5316 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5318 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5321 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5322 emitted arch directives.
5323 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5324 __ARM_FEATURE_COPROC before changing architectures.
5326 2018-02-06 Richard Biener <rguenther@suse.de>
5328 * config/i386/i386.c (print_reg): Fix typo.
5329 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5331 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5333 * configure: Regenerate.
5335 2018-02-05 Martin Sebor <msebor@redhat.com>
5337 PR tree-optimization/83369
5338 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5341 2018-02-05 Martin Liska <mliska@suse.cz>
5343 * doc/invoke.texi: Cherry-pick upstream r323995.
5345 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5347 * ira.c (ira_init_register_move_cost): Adjust comment.
5349 2018-02-05 Martin Liska <mliska@suse.cz>
5351 PR gcov-profile/84137
5352 * doc/gcov.texi: Fix typo in documentation.
5354 2018-02-05 Martin Liska <mliska@suse.cz>
5356 PR gcov-profile/83879
5357 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5358 using dlopen functionality.
5360 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5362 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5363 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5364 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5365 _mm_maskz_range_round_ss): New intrinsics.
5366 (__builtin_ia32_rangesd128_round)
5367 (__builtin_ia32_rangess128_round): Remove.
5368 (__builtin_ia32_rangesd128_mask_round,
5369 __builtin_ia32_rangess128_mask_round): New builtins.
5370 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5371 __builtin_ia32_rangess128_round): Remove.
5372 (__builtin_ia32_rangesd128_mask_round,
5373 __builtin_ia32_rangess128_mask_round): New builtins.
5374 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5375 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5376 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5377 "<round_saeonly_constraint>")): Changed to ...
5378 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5379 "<round_saeonly_scalar_constraint>")): ... this.
5380 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5381 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5382 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5383 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5384 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5386 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5388 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5390 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5391 Remove all values except native, 8540 and 8548.
5393 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5395 * config/i386/i386.c (ix86_output_function_return): Pass
5396 INVALID_REGNUM, instead of -1, as invalid register number to
5397 indirect_thunk_name and output_indirect_thunk.
5399 2018-02-02 Julia Koval <julia.koval@intel.com>
5401 * config.gcc: Add -march=icelake.
5402 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5403 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5404 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5405 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5406 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5407 (processor_target_table): Add icelake.
5408 (ix86_option_override_internal): Handle new PTAs.
5409 (get_builtin_code_for_version): Handle icelake.
5410 (M_INTEL_COREI7_ICELAKE): New.
5411 (fold_builtin_cpu): Handle icelake.
5412 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5413 * doc/invoke.texi: Add -march=icelake.
5415 2018-02-02 Julia Koval <julia.koval@intel.com>
5417 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5418 to wide_int_bitmask.
5419 * wide-int-bitmask.h: New.
5421 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5424 * config/i386/i386.md: Replace Pmode with word_mode in
5425 builtin_setjmp_setup and builtin_longjmp to support x32.
5427 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5431 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5433 (rs6000_supported_cpu_names): New static variable.
5434 (linux_cpu_translation_table): Likewise.
5435 (elf_platform) <cpu>: Define new static variable and use it.
5436 Translate kernel AT_PLATFORM name to canonical name if needed.
5437 Error if platform name is unknown.
5439 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5442 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5444 2018-02-01 Jeff Law <law@redhat.com>
5447 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5448 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5449 the scratch if RELEASE_VIA_POP is false.
5450 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5451 If we have to save a temporary register, decrement SIZE appropriately.
5452 Pass new arguments to release_scratch_register_on_entry.
5453 (ix86_adjust_stack_and_probe): Likewise.
5454 (ix86_emit_probe_stack_range): Pass new arguments to
5455 release_scratch_register_on_entry.
5457 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5459 PR rtl-optimization/84157
5460 * combine.c (change_zero_ext): Use REG_P predicate in
5461 front of HARD_REGISTER_P predicate.
5463 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5465 * config/avr/avr.c (avr_option_override): Move disabling of
5466 -fdelete-null-pointer-checks to...
5467 * common/config/avr/avr-common.c (avr_option_optimization_table):
5470 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5472 PR tree-optimization/81635
5473 * tree-data-ref.c (split_constant_offset_1): For types that
5474 wrap on overflow, try to use range info to prove that wrapping
5477 2018-02-01 Renlin Li <renlin.li@arm.com>
5480 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5482 (aarch64_register_move_cost): Likewise.
5483 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5485 (REG_CLASS_NAMES): Likewise.
5486 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5487 TAILCALL_ADDR_REGS. Remove IP registers.
5488 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5490 2018-02-01 Richard Biener <rguenther@suse.de>
5492 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5493 for specifying RPO order and allow NULL for that.
5494 * domwalk.c (dom_walker::dom_walker): Likewise.
5495 (dom_walker::walk): Handle NULL RPO order.
5496 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5498 (rewrite_update_dom_walker): Likewise.
5499 (mark_def_dom_walker): Likewise.
5501 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5503 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5504 (aarch64_maybe_expand_sve_subreg_move): Declare.
5505 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5506 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5508 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5509 that are semantically a reverse operation.
5510 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5511 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5512 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5514 (aarch64_can_change_mode_class): For big-endian, forbid changes
5515 between two SVE modes if they have different element sizes.
5517 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5519 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5520 the TImode handling for big-endian targets.
5522 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5524 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5525 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5527 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5528 Remove BSWAP handing for big-endian targets and use the form of
5529 LD1RQ appropariate for the mode.
5531 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5533 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5534 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5537 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5540 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5541 check for operands that need to go through aarch64_sve_reload_be.
5543 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5545 PR tree-optimization/81661
5546 PR tree-optimization/84117
5547 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5548 * tree-eh.c: Include gimplify.h.
5549 (find_trapping_overflow, replace_trapping_overflow,
5550 rewrite_to_non_trapping_overflow): New functions.
5551 * tree-vect-loop.c: Include tree-eh.h.
5552 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5553 * tree-data-ref.c: Include tree-eh.h.
5554 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5556 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5558 PR rtl-optimization/84123
5559 * combine.c (change_zero_ext): Check if hard register satisfies
5560 can_change_dest_mode before calling gen_lowpart_SUBREG.
5562 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5565 * ira.c (ira_init_register_move_cost): Remove assert.
5567 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5569 PR rtl-optimization/84071
5570 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5571 * doc/tm.texi: Regenerate.
5573 2018-01-31 Richard Biener <rguenther@suse.de>
5575 PR tree-optimization/84132
5576 * tree-data-ref.c (analyze_miv_subscript): Properly
5577 check whether evolution_function_is_affine_multivariate_p
5578 before calling gcd_of_steps_may_divide_p.
5580 2018-01-31 Julia Koval <julia.koval@intel.com>
5583 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5584 * config/i386/i386.md (rdpid_rex64) New.
5585 (rdpid): Make 32bit only.
5587 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5590 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5591 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5593 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5596 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5598 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5600 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5602 PR rtl-optimization/84071
5603 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5604 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5606 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5608 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5609 (arc_attribute_table): Add 'aux' attribute.
5610 (arc_in_small_data_p): Consider aux like variables.
5611 (arc_is_aux_reg_p): New function.
5612 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5613 (arc_get_aux_arg): New function.
5614 (prepare_move_operands): Handle aux-register access.
5615 (arc_handle_aux_attribute): New function.
5616 * doc/extend.texi (ARC Variable attributes): Add subsection.
5618 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5620 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5621 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5622 (arc_attribute_table): Add 'uncached' attribute.
5623 (arc_print_operand): Print '.di' flag for uncached memory
5625 (arc_in_small_data_p): Do not consider for small data the uncached
5627 (arc_is_uncached_mem_p): New function.
5628 * config/arc/predicates.md (compact_store_memory_operand): Check
5629 for uncached memory accesses.
5630 (nonvol_nonimm_operand): Likewise.
5631 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5633 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5636 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5637 falign-loops=): Add Optimization flag.
5639 2018-01-30 Jeff Law <law@redhat.com>
5642 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5643 INT_REGISTERS_SAVED. Check it prior to calling
5644 get_scratch_register_on_entry.
5645 (ix86_adjust_stack_and_probe): Similarly.
5646 (ix86_emit_probe_stack_range): Similarly.
5647 (ix86_expand_prologue): Corresponding changes.
5649 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5652 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5653 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5655 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5658 * lra-constraints.c (curr_insn_transform): Process AND in the
5661 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5663 PR rtl-optimization/83986
5664 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5665 dependence against last_pending_memory_flush in addition to
5668 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5670 PR tree-optimization/81611
5671 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5674 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5677 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5680 2018-01-30 Richard Biener <rguenther@suse.de>
5681 Jakub Jelinek <jakub@redhat.com>
5683 PR tree-optimization/84111
5684 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5685 inner loops added during recursion, as they don't have up-to-date
5688 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5691 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5692 (can_inline_edge_by_limits_p): ... here.
5693 (can_early_inline_edge_p, check_callers,
5694 update_caller_keys, update_callee_keys, recursive_inlining,
5695 add_new_edges_to_heap, speculation_useful_p,
5696 inline_small_functions,
5697 inline_small_functions, flatten_function,
5698 inline_to_all_callers_1): Update.
5700 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5702 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5705 2018-01-30 Richard Biener <rguenther@suse.de>
5707 PR tree-optimization/83008
5708 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5709 invariant and constant vector uses in stmts when they need
5712 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5715 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5716 * configure: Regenerate.
5718 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5720 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5722 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5723 Use gen_rtx_REG rather than gen_lowpart.
5725 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5727 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5728 rather than 0 when creating partial subregs.
5730 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5732 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5735 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5738 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5739 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5740 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5741 flags. This restores the settings used before the 2017-07-24.
5742 Turning off pre increment/decrement/modify allows IVOPTS to
5743 optimize DF/SF loops where the index is an int.
5745 2018-01-29 Richard Biener <rguenther@suse.de>
5746 Kelvin Nilsen <kelvin@gcc.gnu.org>
5749 * tree-vect-stmts.c (vectorizable_call): Don't call
5750 targetm.vectorize_builtin_md_vectorized_function if callee is
5753 2018-01-22 Carl Love <cel@us.ibm.com>
5755 * doc/extend.tex: Fix typo in second arg in
5756 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5758 2018-01-29 Richard Biener <rguenther@suse.de>
5760 PR tree-optimization/84086
5761 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5762 (flush_ssaname_freelist): When SSA names were released reset
5763 the SCEV hash table.
5765 2018-01-29 Richard Biener <rguenther@suse.de>
5767 PR tree-optimization/84057
5768 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5769 removed paths when removing edges.
5771 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5773 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5774 -mfunction-return=@var{choice}.
5776 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5779 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5781 (layout::print_source_line): Likewise.
5782 (test_get_line_width_without_trailing_whitespace): Add test cases.
5784 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5787 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5790 2018-01-26 Jim Wilson <jimw@sifive.com>
5792 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5794 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5797 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5799 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5800 and CMP + SUB-immediate -> SUBS.
5802 2018-01-26 Martin Sebor <msebor@redhat.com>
5804 PR tree-optimization/83896
5805 * tree-ssa-strlen.c (get_string_len): Rename...
5806 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5807 Avoid assuming length is constant.
5808 (handle_char_store): Use HOST_WIDE_INT for string length.
5810 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5813 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5814 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5816 2018-01-26 Richard Biener <rguenther@suse.de>
5818 PR rtl-optimization/84003
5819 * dse.c (record_store): Only record redundant stores when
5820 the earlier store aliases at least all accesses the later one does.
5822 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5824 PR rtl-optimization/83985
5825 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5826 REG_CFA_RESTORE insns.
5827 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5828 REG_CFA_RESTORE insns here.
5831 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5832 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5834 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5836 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5837 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5838 (arc_init): Likewise.
5839 (arc_override_options): Likewise.
5840 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5842 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5844 * config/arc/arc.h (TARGET_DBNZ): Define.
5845 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5846 properly set the tune attribute.
5847 (dbnz): Use TARGET_DBNZ guard.
5848 * config/arc/arc.opt (mtune): Add core3 option.
5850 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5852 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5853 recognize new pic like addresses.
5854 (arc_delegitimize_address): Clean up.
5856 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5858 * config/arc/arc-arches.def: Option mrf16 valid for all
5860 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5861 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5862 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5863 * config/arc/arc-tables.opt: Regenerate.
5864 * config/arc/arc.c (arc_conditional_register_usage): Handle
5865 reduced register file case.
5866 (arc_file_start): Set must have build attributes.
5867 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5869 * config/arc/arc.opt (mrf16): Add new option.
5870 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5871 * config/arc/genmultilib.awk: Handle new mrf16 option.
5872 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5873 * config/arc/t-multilib: Regenerate.
5874 * doc/invoke.texi (ARC Options): Document mrf16 option.
5876 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5878 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5879 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5880 (arc_attribute_table): Add 'secure_call' attribute.
5881 (arc_print_operand): Print secure call operand.
5882 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5884 (arc_is_secure_call_p): New function. * config/arc/arc.md
5885 (call_i): Add support for sjli instruction.
5886 (call_value_i): Likewise.
5887 * config/arc/constraints.md (Csc): New constraint.
5889 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5890 John Eric Martin <John.Martin@emmicro-us.com>
5892 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5893 * config/arc/arc.c (_arc_jli_section): New struct.
5894 (arc_jli_section): New type.
5895 (rc_jli_sections): New static variable.
5896 (arc_handle_jli_attribute): New function.
5897 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5898 (arc_file_end): New function.
5899 (TARGET_ASM_FILE_END): Define.
5900 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5901 (arc_add_jli_section): New function.
5902 (jli_call_scan): Likewise.
5903 (arc_reorg): Call jli_call_scan.
5904 (arc_output_addsi): Remove 'S' from printing asm operand.
5905 (arc_is_jli_call_p): New function.
5906 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5908 (movhi_insn): Likewise.
5909 (movsi_insn): Likewise.
5910 (movsi_set_cc_insn): Likewise.
5911 (loadqi_update): Likewise.
5912 (load_zeroextendqisi_update): Likewise.
5913 (load_signextendqisi_update): Likewise.
5914 (loadhi_update): Likewise.
5915 (load_zeroextendhisi_update): Likewise.
5916 (load_signextendhisi_update): Likewise.
5917 (loadsi_update): Likewise.
5918 (loadsf_update): Likewise.
5919 (movsicc_insn): Likewise.
5920 (bset_insn): Likewise.
5921 (bxor_insn): Likewise.
5922 (bclr_insn): Likewise.
5923 (bmsk_insn): Likewise.
5924 (bicsi3_insn): Likewise.
5925 (cmpsi_cc_c_insn): Likewise.
5926 (movsi_ne): Likewise.
5927 (movsi_cond_exec): Likewise.
5928 (clrsbsi2): Likewise.
5939 (call_i): Remove 'S' asm letter, add jli instruction.
5940 (call_value_i): Likewise.
5941 * config/arc/arc.op (mjli-always): New option.
5942 * config/arc/constraints.md (Cji): New constraint.
5943 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5945 (subsf3_fpx): Likewise.
5946 (mulsf3_fpx): Likewise.
5947 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5949 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5951 * doc/invoke.texi (ARC): Document mjli-always option.
5953 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5955 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5956 avoid addition with 0 and use incw and decw where possible.
5958 2018-01-26 Richard Biener <rguenther@suse.de>
5960 PR tree-optimization/81082
5961 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5962 association if it requires casting to unsigned.
5963 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5964 from fold_plusminus_mult_expr to catch important cases late when
5965 range info is available.
5967 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5969 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5970 * configure.ac (hidden_linkonce): New test.
5971 * configure: Regenerate.
5972 * config.in: Regenerate.
5974 2018-01-26 Julia Koval <julia.koval@intel.com>
5976 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5977 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5978 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5979 _mm_mask_bitshuffle_epi64_mask): Fix type.
5980 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5981 USI_FTYPE_V4DI_V4DI_USI): Remove.
5982 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5983 __builtin_ia32_vpshufbitqmb256_mask,
5984 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5985 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5986 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5988 2018-01-26 Alan Modra <amodra@gmail.com>
5991 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5992 UNSPEC_VBPERMQ. Sort other unspecs.
5994 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5996 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5998 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6001 * predict.c (drop_profile): Do not push/pop cfun; update also
6003 (handle_missing_profiles): Fix logic looking for zero profiles.
6005 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6008 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6009 on functions with #pragma omp declare simd or functions with simd
6011 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6012 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6013 Remove trailing \n from warning_at calls.
6015 2018-01-25 Tom de Vries <tom@codesourcery.com>
6018 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6019 for neutered workers.
6021 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6024 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6025 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6027 2018-01-24 Jeff Law <law@redhat.com>
6030 * i386.c (get_probe_interval): Move to earlier point.
6031 (ix86_compute_frame_layout): If -fstack-clash-protection and
6032 the frame is larger than the probe interval, then use pushes
6033 to save registers rather than reg->mem moves.
6034 (ix86_expand_prologue): Remove conditional for int_registers_saved
6037 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6040 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6041 min/max for never referenced object.
6043 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6046 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6048 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6049 attributes from DECL_ATTRIBUTES (decl) without affecting
6050 DECL_ATTRIBUTES (current_function_decl).
6051 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6052 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6054 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6056 PR tree-optimization/83979
6057 * fold-const.c (fold_comparison): Use constant_boolean_node
6058 instead of boolean_{true,false}_node.
6060 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6062 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6065 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6067 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6068 Simplify the clause that sets the length attribute.
6069 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6070 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6071 clause that sets the length attribute.
6072 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6074 2018-01-24 Tom de Vries <tom@codesourcery.com>
6077 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6078 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6079 Add strict parameter.
6080 (prevent_branch_around_nothing): Insert dummy insn between branch to
6081 label and label with no ptx insn inbetween.
6082 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6084 2018-01-24 Tom de Vries <tom@codesourcery.com>
6087 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6088 for neutered threads in warp.
6089 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6091 2018-01-24 Richard Biener <rguenther@suse.de>
6093 PR tree-optimization/83176
6094 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6097 2018-01-24 Richard Biener <rguenther@suse.de>
6099 PR tree-optimization/82819
6100 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6101 code generating pluses that are no-ops in the target precision.
6103 2018-01-24 Richard Biener <rguenther@suse.de>
6106 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6108 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6110 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6111 to merge probabilities.
6112 * predict.c (probably_never_executed): Also mark as cold functions
6113 with global 0 profile and guessed local profile.
6114 * profile-count.c (profile_probability::combine_with_count): New
6116 * profile-count.h (profile_probability::operator*,
6117 profile_probability::operator*=, profile_probability::operator/,
6118 profile_probability::operator/=): Reduce precision to adjusted
6119 and set value to guessed on contradictory divisions.
6120 (profile_probability::combine_with_freq): Remove.
6121 (profile_probability::combine_wiht_count): Declare.
6122 (profile_count::force_nonzero):: Set to adjusted.
6123 (profile_count::probability_in):: Set quality to adjusted.
6124 * tree-ssa-tail-merge.c (replace_block_by): Use
6127 2018-01-23 Andrew Waterman <andrew@sifive.com>
6128 Jim Wilson <jimw@sifive.com>
6130 * config/riscv/riscv.c (riscv_stack_boundary): New.
6131 (riscv_option_override): Set riscv_stack_boundary. Handle
6132 riscv_preferred_stack_boundary_arg.
6133 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6134 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6135 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6136 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6137 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6138 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6140 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6143 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6144 of struct ix86_frame.
6145 (ix86_expand_epilogue): Likewise. Add a local variable for
6146 the reg_save_offset field in struct ix86_frame.
6148 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6150 PR tree-optimization/82604
6151 * tree-loop-distribution.c (enum partition_kind): New enum item
6152 PKIND_PARTIAL_MEMSET.
6153 (partition_builtin_p): Support above new enum item.
6154 (generate_code_for_partition): Ditto.
6155 (compute_access_range): Differentiate cases that equality can be
6156 proven at all loops, the innermost loops or no loops.
6157 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6158 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6159 (finalize_partitions, distribute_loop): Don't fuse partition of
6160 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6161 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6164 2018-01-23 Martin Liska <mliska@suse.cz>
6166 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6167 order to ignore the predictor.
6168 (PRED_POLYMORPHIC_CALL): Likewise.
6169 (PRED_RECURSIVE_CALL): Likewise.
6171 2018-01-23 Martin Liska <mliska@suse.cz>
6173 * tree-profile.c (tree_profiling): Print function header to
6174 aware reader which function we are working on.
6175 * value-prof.c (gimple_find_values_to_profile): Do not print
6176 not interesting value histograms.
6178 2018-01-23 Martin Liska <mliska@suse.cz>
6180 * profile-count.h (enum profile_quality): Add
6181 profile_uninitialized as the first value. Do not number values
6182 as they are zero based.
6183 (profile_count::verify): Update sanity check.
6184 (profile_probability::verify): Likewise.
6186 2018-01-23 Nathan Sidwell <nathan@acm.org>
6188 * doc/invoke.texi (ffor-scope): Deprecate.
6190 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6192 PR tree-optimization/83510
6193 * domwalk.c (set_all_edges_as_executable): New function.
6194 (dom_walker::dom_walker): Convert bool param
6195 "skip_unreachable_blocks" to enum reachability. Move setup of
6196 edge flags to set_all_edges_as_executable and only do it when
6197 reachability is REACHABLE_BLOCKS.
6198 * domwalk.h (enum dom_walker::reachability): New enum.
6199 (dom_walker::dom_walker): Convert bool param
6200 "skip_unreachable_blocks" to enum reachability.
6201 (set_all_edges_as_executable): New decl.
6202 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6203 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6205 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6206 but converting true to REACHABLE_BLOCKS.
6207 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6209 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6210 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6211 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6213 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6214 if check_all_array_refs will be called.
6216 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6218 * tree.c (selftest::test_location_wrappers): Add more test
6221 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6223 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6224 (selftest::test_bit_in_range): Likewise.
6226 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6229 * doc/sourcebuild.texi (vect_float): Say that the selector
6230 only describes the situation when -funsafe-math-optimizations is on.
6231 (vect_float_strict): Document.
6233 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6235 PR tree-optimization/83965
6236 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6237 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6238 instead of checking only for a reduction.
6239 (vect_recog_widen_sum_pattern): Likewise.
6241 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6243 * predict.c (probably_never_executed): Only use precise profile info.
6244 (compute_function_frequency): Skip after inlining hack since we now
6245 have quality checking.
6247 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6249 * profile-count.h (profile_probability::very_unlikely,
6250 profile_probability::unlikely, profile_probability::even): Set
6251 precision to guessed.
6253 2018-01-23 Richard Biener <rguenther@suse.de>
6255 PR tree-optimization/83963
6256 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6257 Properly terminate dominator walk when crossing the exit edge not
6258 when visiting its source block.
6260 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6263 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6264 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6266 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6268 PR tree-optimization/83957
6269 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6270 semicolon after for body surrounded by braces.
6272 PR tree-optimization/83081
6273 * profile-count.h (profile_probability::split): New method.
6274 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6275 Use profile_probability::split.
6276 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6277 when splitting a single conditional jump into 2.
6279 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6281 PR tree-optimization/69452
6282 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6285 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6287 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6288 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6289 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6291 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6293 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6294 * config/rl78/rl78.md (movdi): New define_expand.
6295 * config/rl78/rl78.c (rl78_split_movdi): New function.
6297 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6300 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6302 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6303 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6304 128-bit to produce an UNSPEC move to get the double word with the
6305 signbit and then a shift directly to do signbit.
6306 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6307 implementation with a new version that just does either a direct
6308 move or a regular move. Move memory interface to separate insns.
6309 Move insns so they are next to the expander.
6310 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6311 with signbit move. Split big and little endian case.
6312 (signbit<mode>2_dm_mem_le): Likewise.
6313 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6314 (signbit<mode>2_dm2): Likewise.
6316 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6318 * config/rl78/rl78.md (anddi3): New define_expand.
6320 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6322 * config/rl78/rl78.md (umindi3): New define_expand.
6324 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6326 * config/rl78/rl78.md (smindi3): New define_expand.
6328 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6330 * config/rl78/rl78.md (smaxdi3): New define_expand.
6332 2018-01-22 Carl Love <cel@us.ibm.com>
6334 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6335 LVX_V1TI): Add macro expansion.
6336 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6337 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6338 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6339 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6340 Change check to determine if the instruction is a byte reversing
6341 entry. Fix typo in comment.
6342 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6343 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6344 Add def_builtin calls for new builtins.
6345 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6346 Add define_insn expansion.
6348 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6350 * config/rl78/rl78.md (umaxdi3): New define_expand.
6352 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6354 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6355 for non-QImode registers.
6357 2018-01-22 Richard Biener <rguenther@suse.de>
6359 PR tree-optimization/83963
6360 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6361 including the loop exit block.
6362 (scop_detection::merge_sese): Likewise.
6363 (scop_detection::add_scop): Do it here instead.
6365 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6367 * doc/sourcebuild.texi (arm_softfloat): Document.
6369 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6372 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6373 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6374 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6376 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6377 David Edelsohn <dje.gcc@gmail.com>
6380 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6381 Change "crset eq" to "crset 2".
6382 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6383 (*call_indirect_aix<mode>_nospec): Likewise.
6384 (*call_value_indirect_aix<mode>_nospec): Likewise.
6385 (*call_indirect_elfv2<mode>_nospec): Likewise.
6386 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6387 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6388 change assembly output from . to $.
6389 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6390 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6391 (*tablejump<mode>_internal1_nospec): Likewise.
6393 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6396 * config/sh/sh_optimize_sett_clrt.cc:
6397 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6399 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6401 PR tree-optimization/83940
6402 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6403 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6404 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6405 use it to pass back the definition type.
6406 (vect_check_store_rhs): Likewise rhs_dt_out.
6407 (vect_build_gather_load_calls): Add a mask_dt argument and use
6408 it instead of a call to vect_is_simple_use.
6409 (vectorizable_store): Update calls to vect_check_load_store_mask
6410 and vect_check_store_rhs. Use the dt returned by the latter instead
6411 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6412 instead of calls to vect_is_simple_use. Pass the scalar rather
6413 than the vector operand to vect_is_simple_use when handling
6414 second and subsequent copies of an rhs value.
6415 (vectorizable_load): Update calls to vect_check_load_store_mask
6416 and vect_build_gather_load_calls. Use the cached mask_dt and
6417 gs_info.offset_dt instead of calls to vect_is_simple_use.
6419 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6422 * tree-emutls.c: Include gimplify.h.
6423 (lower_emutls_2): New function.
6424 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6425 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6426 it before further processing.
6429 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6430 UINTVAL (trueop1) instead of INTVAL (op1).
6432 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6436 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6437 INCOMING_FRAME_SP_OFFSET if not defined.
6438 (scan_trace): Add ENTRY argument. If true and
6439 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6440 emit a note to adjust the CFA offset.
6441 (create_cfi_notes): Adjust scan_trace callers.
6442 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6443 INCOMING_FRAME_SP_OFFSET in the CIE.
6444 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6445 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6447 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6448 * doc/tm.texi: Regenerated.
6450 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6452 PR rtl-optimization/83147
6453 * lra-constraints.c (remove_inheritance_pseudos): Use
6454 lra_substitute_pseudo_within_insn.
6456 2018-01-19 Tom de Vries <tom@codesourcery.com>
6457 Cesar Philippidis <cesar@codesourcery.com>
6460 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6462 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6465 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6466 spaces for function labels.
6468 2018-01-19 Martin Liska <mliska@suse.cz>
6470 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6471 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6472 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6473 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6474 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6475 (PRED_CONST_RETURN): Change from 69 to 65.
6476 (PRED_NULL_RETURN): Change from 91 to 71.
6477 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6478 (PRED_LOOP_GUARD): Change from 66 to 73.
6480 2018-01-19 Martin Liska <mliska@suse.cz>
6482 * predict.c (predict_insn_def): Add new assert.
6483 (struct branch_predictor): Change type to signed integer.
6484 (test_prediction_value_range): Amend test to cover
6486 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6487 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6488 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6489 (PRED_LOOP_IV_COMPARE): Likewise.
6490 * predict.h (PROB_UNINITIALIZED): Define new constant.
6492 2018-01-19 Martin Liska <mliska@suse.cz>
6494 * predict.c (dump_prediction): Add new format for
6495 analyze_brprob.py script which is enabled with -details
6497 * profile-count.h (precise_p): New function.
6499 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6501 PR tree-optimization/83922
6502 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6503 there are no statements that need masking.
6504 (vect_active_double_reduction_p): New function.
6505 (vect_analyze_loop_operations): Use it when handling phis that
6506 are not in the loop header.
6508 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6510 PR tree-optimization/83914
6511 * tree-vect-loop.c (vectorizable_induction): Don't convert
6512 init_expr or apply the peeling adjustment for inductions
6513 that are nested within the vectorized loop.
6515 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6517 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6520 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6524 * function.h (gimplify_parameters): Add gimple_seq * argument.
6525 * function.c: Include gimple.h and options.h.
6526 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6527 for the added local temporaries if needed.
6528 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6529 if there are any parameter cleanups, wrap whole body into a
6530 try/finally with the cleanups.
6532 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6535 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6536 Use GET_MODE_CLASS for scalar floating point.
6538 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6542 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6543 Fix call of call_cgraph_insertion_hooks.
6545 2018-01-18 Martin Sebor <msebor@redhat.com>
6547 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6549 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6552 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6555 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6558 * common.opt: (-ffile-prefix-map): New option.
6559 * opts.c (common_handle_option): Defer it.
6560 * opts-global.c (handle_common_deferred_options): Handle it.
6561 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6562 * file-prefix-map.h: New file.
6563 (remap_debug_filename, add_debug_prefix_map): ...here.
6564 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6565 * final.c (debug_prefix_map, add_debug_prefix_map
6566 remap_debug_filename): Move to...
6567 * file-prefix-map.c: New file.
6568 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6569 generalize, get rid of alloca(), use strrchr() instead of strchr().
6570 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6571 Implement in terms of add_prefix_map().
6572 (remap_macro_filename, remap_debug_filename): Implement in term of
6574 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6575 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6576 * dbxout.c: Include file-prefix-map.h.
6577 * varasm.c: Likewise.
6578 * vmsdbgout.c: Likewise.
6579 * xcoffout.c: Likewise.
6580 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6581 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6582 * doc/invoke.texi (-ffile-prefix-map): Document.
6583 (-fdebug-prefix-map): Update description.
6585 2018-01-18 Martin Liska <mliska@suse.cz>
6587 * config/i386/i386.c (indirect_thunk_name): Document that also
6589 (output_indirect_thunk): Document why both instructions
6590 (pause and lfence) are generated.
6592 2018-01-18 Richard Biener <rguenther@suse.de>
6594 PR tree-optimization/83887
6595 * graphite-scop-detection.c
6596 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6597 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6598 (scop_detection::merge_sese): Re-implement with a flood-fill
6599 algorithm that properly finds a SESE region if it exists.
6601 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6604 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6605 pointer_diff optimizations use view_convert instead of convert.
6607 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6609 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6610 Generate different code for -mno-speculate-indirect-jumps.
6611 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6612 (*call_indirect_aix<mode>): Disable for
6613 -mno-speculate-indirect-jumps.
6614 (*call_indirect_aix<mode>_nospec): New define_insn.
6615 (*call_value_indirect_aix<mode>): Disable for
6616 -mno-speculate-indirect-jumps.
6617 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6618 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6619 -mno-speculate-indirect-jumps.
6620 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6622 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6624 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6625 long double type, set the flags for noting the default long double
6626 type, even if we don't pass or return a long double type.
6628 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6631 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6634 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6636 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6637 support for merge[hl].
6638 (fold_mergehl_helper): New helper function.
6639 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6640 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6641 (altivec_vmrglw_direct): Add xxmrglw insn.
6643 2018-01-17 Andrew Waterman <andrew@sifive.com>
6645 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6646 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6648 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6651 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6652 call the lto_location_cache before reading the
6653 DECL_SOURCE_LOCATION of the types.
6655 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6656 Richard Sandiford <richard.sandiford@linaro.org>
6658 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6659 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6660 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6661 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6662 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6664 * config/aarch64/constraints.md (aarch64_movti_operand):
6666 * config/aarch64/predicates.md (Uti): Add new constraint.
6668 2018-01-17 Carl Love <cel@us.ibm.com>
6670 * config/rs6000/vsx.md (define_expand xl_len_r,
6671 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6672 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6674 (define_expand, define_insn): Move the shift left from the
6675 define_insn to the define_expand for lxvl and stxvl instructions.
6676 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6677 and XL_LEN_R definitions to PURE.
6679 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6681 * config/i386/i386.c (indirect_thunk_name): Declare regno
6682 as unsigned int. Compare regno with INVALID_REGNUM.
6683 (output_indirect_thunk): Ditto.
6684 (output_indirect_thunk_function): Ditto.
6685 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6686 in the call to output_indirect_thunk_function.
6688 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6691 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6692 rather than the size of inner_type to determine the stack slot size
6693 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6695 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6698 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6701 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6703 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6704 endian Linux systems to optionally enable multilibs for selecting
6705 the long double type if the user configured an explicit type.
6706 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6707 have no long double multilibs if not defined.
6708 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6709 warn if the user used -mabi={ieee,ibm}longdouble and we built
6710 multilibs for long double.
6711 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6712 appropriate multilib option.
6713 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6715 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6716 for building long double multilibs.
6717 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6719 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6721 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6724 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6726 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6729 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6732 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6735 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6737 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6738 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6740 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6742 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6743 different rtl trees depending on TARGET_64BIT.
6744 (rs6000_gen_lvx): Likewise.
6746 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6748 * config/visium/visium.md (nop): Tweak comment.
6749 (hazard_nop): Likewise.
6751 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6753 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6754 -mspeculate-indirect-jumps.
6755 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6756 for -mno-speculate-indirect-jumps.
6757 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6758 (*call_value_indirect_elfv2<mode>): Disable for
6759 -mno-speculate-indirect-jumps.
6760 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6761 (indirect_jump): Emit different RTL for
6762 -mno-speculate-indirect-jumps.
6763 (*indirect_jump<mode>): Disable for
6764 -mno-speculate-indirect-jumps.
6765 (*indirect_jump<mode>_nospec): New define_insn.
6766 (tablejump): Emit different RTL for
6767 -mno-speculate-indirect-jumps.
6768 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6769 (tablejumpsi_nospec): New define_expand.
6770 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6771 (tablejumpdi_nospec): New define_expand.
6772 (*tablejump<mode>_internal1): Disable for
6773 -mno-speculate-indirect-jumps.
6774 (*tablejump<mode>_internal1_nospec): New define_insn.
6775 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6778 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6780 * caller-save.c (insert_save): Drop unnecessary parameter. All
6783 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6784 Richard Biener <rguenth@suse.de>
6787 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6788 return early, inline manually is_gimple_sizepos. Make sure if we
6789 call gimplify_expr we don't end up with a gimple constant.
6790 * tree.c (variably_modified_type_p): Don't return true for
6791 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6792 * gimplify.h (is_gimple_sizepos): Remove.
6794 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6796 PR tree-optimization/83857
6797 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6798 vectorizable_live_operation for pure SLP statements.
6799 (vectorizable_live_operation): Handle PHIs.
6801 2018-01-16 Richard Biener <rguenther@suse.de>
6803 PR tree-optimization/83867
6804 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6805 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6807 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6810 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6811 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6812 If off is not INTEGER_CST, issue a may not be aligned warning
6813 rather than isn't aligned. Use isn%'t rather than isn't.
6814 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6816 <case MULT_EXPR>: Improve the case when bottom and one of the
6817 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6818 operand, in that case check if the other operand is multiple of
6819 bottom divided by the INTEGER_CST operand.
6821 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6824 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6825 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6826 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6827 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6828 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6829 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6830 (pa_function_arg_size): New function.
6832 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6834 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6835 in a separate statement.
6837 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6839 PR tree-optimization/83847
6840 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6841 group gathers and scatters.
6843 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6845 PR rtl-optimization/86620
6846 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6848 PR rtl-optimization/83213
6849 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6850 to last if both are JUMP_INSNs.
6852 PR tree-optimization/83843
6853 * gimple-ssa-store-merging.c
6854 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6855 store_immediate_info for bswap/nop orig_stores.
6857 2018-01-15 Andrew Waterman <andrew@sifive.com>
6859 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6861 <UDIV>: Increase cost if !TARGET_DIV.
6863 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6865 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6866 (define_attr "cr_logical_3op"): New.
6867 (cceq_ior_compare): Adjust.
6868 (cceq_ior_compare_complement): Adjust.
6869 (*cceq_rev_compare): Adjust.
6870 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6871 (is_cracked_insn): Adjust.
6872 (insn_must_be_first_in_group): Adjust.
6873 * config/rs6000/40x.md: Adjust.
6874 * config/rs6000/440.md: Adjust.
6875 * config/rs6000/476.md: Adjust.
6876 * config/rs6000/601.md: Adjust.
6877 * config/rs6000/603.md: Adjust.
6878 * config/rs6000/6xx.md: Adjust.
6879 * config/rs6000/7450.md: Adjust.
6880 * config/rs6000/7xx.md: Adjust.
6881 * config/rs6000/8540.md: Adjust.
6882 * config/rs6000/cell.md: Adjust.
6883 * config/rs6000/e300c2c3.md: Adjust.
6884 * config/rs6000/e500mc.md: Adjust.
6885 * config/rs6000/e500mc64.md: Adjust.
6886 * config/rs6000/e5500.md: Adjust.
6887 * config/rs6000/e6500.md: Adjust.
6888 * config/rs6000/mpc.md: Adjust.
6889 * config/rs6000/power4.md: Adjust.
6890 * config/rs6000/power5.md: Adjust.
6891 * config/rs6000/power6.md: Adjust.
6892 * config/rs6000/power7.md: Adjust.
6893 * config/rs6000/power8.md: Adjust.
6894 * config/rs6000/power9.md: Adjust.
6895 * config/rs6000/rs64.md: Adjust.
6896 * config/rs6000/titan.md: Adjust.
6898 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6900 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6901 ix86_indirect_branch_register logic.
6903 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6905 * config/i386/constraints.md (Bs): Update
6906 ix86_indirect_branch_register check. Don't check
6907 ix86_indirect_branch_register with GOT_memory_operand.
6909 * config/i386/predicates.md (GOT_memory_operand): Don't check
6910 ix86_indirect_branch_register here.
6911 (GOT32_symbol_operand): Likewise.
6913 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6915 * config/i386/predicates.md (constant_call_address_operand):
6916 Rewrite ix86_indirect_branch_register logic.
6917 (sibcall_insn_operand): Likewise.
6919 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6921 * config/i386/constraints.md (Bs): Replace
6922 ix86_indirect_branch_thunk_register with
6923 ix86_indirect_branch_register.
6925 * config/i386/i386.md (indirect_jump): Likewise.
6926 (tablejump): Likewise.
6927 (*sibcall_memory): Likewise.
6928 (*sibcall_value_memory): Likewise.
6929 Peepholes of indirect call and jump via memory: Likewise.
6930 * config/i386/i386.opt: Likewise.
6931 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6932 (GOT_memory_operand): Likewise.
6933 (call_insn_operand): Likewise.
6934 (sibcall_insn_operand): Likewise.
6935 (GOT32_symbol_operand): Likewise.
6937 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6940 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6941 type rather than type addr's type points to.
6942 (expand_omp_atomic_mutex): Likewise.
6943 (expand_omp_atomic): Likewise.
6945 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6948 * config/i386/i386.c (output_indirect_thunk_function): Use
6949 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6950 for __x86_return_thunk.
6952 2018-01-15 Richard Biener <rguenther@suse.de>
6955 * expmed.c (extract_bit_field_1): Fix typo.
6957 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6960 * config/arm/iterators.md (VF): New mode iterator.
6961 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6962 Remove integer-related logic from pattern.
6963 (neon_vabd<mode>_3): Likewise.
6965 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6968 * common.opt (fstrict-overflow): No longer an alias.
6969 (fwrapv-pointer): New option.
6970 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6971 also for pointer types based on flag_wrapv_pointer.
6972 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6973 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6974 opts->x_flag_wrapv got set.
6975 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6976 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6977 POINTER_TYPE_OVERFLOW_UNDEFINED.
6978 * match.pd: Likewise in address comparison pattern.
6979 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6981 2018-01-15 Richard Biener <rguenther@suse.de>
6984 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6985 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6986 Reset type names to their identifier if their TYPE_DECL doesn't
6987 have linkage (and thus is used for ODR and devirt).
6988 (save_debug_info_for_decl): Remove.
6989 (save_debug_info_for_type): Likewise.
6990 (add_tree_to_fld_list): Adjust.
6991 * tree-pretty-print.c (dump_generic_node): Make dumping of
6992 type names more robust.
6994 2018-01-15 Richard Biener <rguenther@suse.de>
6996 * BASE-VER: Bump to 8.0.1.
6998 2018-01-14 Martin Sebor <msebor@redhat.com>
7001 * builtins.c (check_access): Avoid warning when the no-warning bit
7004 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7006 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7007 * ira-color (allocno_hard_regs_compare): Likewise.
7009 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7012 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7013 Use .pushsection/.popsection.
7015 2018-01-14 Martin Sebor <msebor@redhat.com>
7018 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7020 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7022 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7023 entry from extra_headers.
7024 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7025 extra_headers, make the list bitwise identical to the i?86-*-* one.
7027 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7029 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7030 -mcmodel=large with -mindirect-branch=thunk,
7031 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7032 -mfunction-return=thunk-extern.
7033 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7034 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7035 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7037 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7039 * config/i386/i386.c (print_reg): Print the name of the full
7040 integer register without '%'.
7041 (ix86_print_operand): Handle 'V'.
7042 * doc/extend.texi: Document 'V' modifier.
7044 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7046 * config/i386/constraints.md (Bs): Disallow memory operand for
7047 -mindirect-branch-register.
7049 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7050 (GOT_memory_operand): Likewise.
7051 (call_insn_operand): Likewise.
7052 (sibcall_insn_operand): Likewise.
7053 (GOT32_symbol_operand): Likewise.
7054 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7055 for -mindirect-branch-register.
7056 (tablejump): Likewise.
7057 (*sibcall_memory): Likewise.
7058 (*sibcall_value_memory): Likewise.
7059 Disallow peepholes of indirect call and jump via memory for
7060 -mindirect-branch-register.
7061 (*call_pop): Replace m with Bw.
7062 (*call_value_pop): Likewise.
7063 (*sibcall_pop_memory): Replace m with Bs.
7064 * config/i386/i386.opt (mindirect-branch-register): New option.
7065 * doc/invoke.texi: Document -mindirect-branch-register option.
7067 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7069 * config/i386/i386-protos.h (ix86_output_function_return): New.
7070 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7071 set function_return_type.
7072 (indirect_thunk_name): Add ret_p to indicate thunk for function
7074 (output_indirect_thunk_function): Pass false to
7075 indirect_thunk_name.
7076 (ix86_output_indirect_branch_via_reg): Likewise.
7077 (ix86_output_indirect_branch_via_push): Likewise.
7078 (output_indirect_thunk_function): Create alias for function
7079 return thunk if regno < 0.
7080 (ix86_output_function_return): New function.
7081 (ix86_handle_fndecl_attribute): Handle function_return.
7082 (ix86_attribute_table): Add function_return.
7083 * config/i386/i386.h (machine_function): Add
7084 function_return_type.
7085 * config/i386/i386.md (simple_return_internal): Use
7086 ix86_output_function_return.
7087 (simple_return_internal_long): Likewise.
7088 * config/i386/i386.opt (mfunction-return=): New option.
7089 (indirect_branch): Mention -mfunction-return=.
7090 * doc/extend.texi: Document function_return function attribute.
7091 * doc/invoke.texi: Document -mfunction-return= option.
7093 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7095 * config/i386/i386-opts.h (indirect_branch): New.
7096 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7097 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7098 with local indirect jump when converting indirect call and jump.
7099 (ix86_set_indirect_branch_type): New.
7100 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7101 (indirectlabelno): New.
7102 (indirect_thunk_needed): Likewise.
7103 (indirect_thunk_bnd_needed): Likewise.
7104 (indirect_thunks_used): Likewise.
7105 (indirect_thunks_bnd_used): Likewise.
7106 (INDIRECT_LABEL): Likewise.
7107 (indirect_thunk_name): Likewise.
7108 (output_indirect_thunk): Likewise.
7109 (output_indirect_thunk_function): Likewise.
7110 (ix86_output_indirect_branch_via_reg): Likewise.
7111 (ix86_output_indirect_branch_via_push): Likewise.
7112 (ix86_output_indirect_branch): Likewise.
7113 (ix86_output_indirect_jmp): Likewise.
7114 (ix86_code_end): Call output_indirect_thunk_function if needed.
7115 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7117 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7118 (ix86_attribute_table): Add indirect_branch.
7119 * config/i386/i386.h (machine_function): Add indirect_branch_type
7120 and has_local_indirect_jump.
7121 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7123 (tablejump): Likewise.
7124 (*indirect_jump): Use ix86_output_indirect_jmp.
7125 (*tablejump_1): Likewise.
7126 (simple_return_indirect_internal): Likewise.
7127 * config/i386/i386.opt (mindirect-branch=): New option.
7128 (indirect_branch): New.
7131 (thunk-inline): Likewise.
7132 (thunk-extern): Likewise.
7133 * doc/extend.texi: Document indirect_branch function attribute.
7134 * doc/invoke.texi: Document -mindirect-branch= option.
7136 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7139 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7141 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7143 * ipa-inline.c (want_inline_small_function_p): Return false if
7144 inlining has already failed with CIF_FINAL_ERROR.
7145 (update_caller_keys): Call want_inline_small_function_p before
7147 (update_callee_keys): Likewise.
7149 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7151 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7153 (rs6000_quadword_masked_address_p): Likewise.
7154 (quad_aligned_load_p): Likewise.
7155 (quad_aligned_store_p): Likewise.
7156 (const_load_sequence_p): Add comment to describe the outer-most loop.
7157 (mimic_memory_attributes_and_flags): New function.
7158 (rs6000_gen_stvx): Likewise.
7159 (replace_swapped_aligned_store): Likewise.
7160 (rs6000_gen_lvx): Likewise.
7161 (replace_swapped_aligned_load): Likewise.
7162 (replace_swapped_load_constant): Capitalize argument name in
7163 comment describing this function.
7164 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7165 and stores that access quad-word aligned addresses and replace
7166 with stvx or lvx instructions when appropriate.
7167 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7168 New function prototype.
7169 (rs6000_quadword_masked_address_p): Likewise.
7170 (rs6000_gen_lvx): Likewise.
7171 (rs6000_gen_stvx): Likewise.
7172 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7173 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7174 when memory address is aligned.
7175 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7176 this split to select lvx instruction when memory address is aligned.
7177 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7178 instruction when memory address is aligned.
7179 (*vsx_le_perm_load_v16qi): Likewise.
7180 (four unnamed splitters): Modify to select the stvx instruction
7181 when memory is aligned.
7183 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7185 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7186 which appears in the queue multiple times.
7188 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7189 Alan Hayward <alan.hayward@arm.com>
7190 David Sherwood <david.sherwood@arm.com>
7192 * tree-vectorizer.h (vec_lower_bound): New structure.
7193 (_loop_vec_info): Add check_nonzero and lower_bounds.
7194 (LOOP_VINFO_CHECK_NONZERO): New macro.
7195 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7196 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7197 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7198 fields. Make seg_len the distance travelled, not including the
7200 (dr_direction_indicator): Declare.
7201 (dr_zero_step_indicator): Likewise.
7202 (dr_known_forward_stride_p): Likewise.
7203 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7205 (runtime_alias_check_p): Allow runtime alias checks with
7207 (operator ==): Compare access_size and align.
7208 (prune_runtime_alias_test_list): Rework for new distinction between
7209 the access_size and seg_len.
7210 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7212 (get_segment_min_max): New function.
7213 (create_intersect_range_checks): Use it.
7214 (dr_step_indicator): New function.
7215 (dr_direction_indicator): Likewise.
7216 (dr_zero_step_indicator): Likewise.
7217 (dr_known_forward_stride_p): Likewise.
7218 * tree-loop-distribution.c (data_ref_segment_size): Return
7219 DR_STEP * (niters - 1).
7220 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7222 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7223 (vect_preserves_scalar_order_p): New function, split out from...
7224 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7225 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7226 (vect_vfa_access_size): New function.
7227 (vect_vfa_align): Likewise.
7228 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7229 (dump_lower_bound): New function.
7230 (vect_check_lower_bound): Likewise.
7231 (vect_small_gap_p): Likewise.
7232 (vectorizable_with_step_bound_p): Likewise.
7233 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7234 depencies if the vectorization factor is 1. Convert the checks
7235 for nonzero steps into checks on the bounds of DR_STEP. Try using
7236 a bunds check for variable steps if the minimum required step is
7237 relatively small. Update calls to the dr_with_seg_len
7238 constructor and to vect_compile_time_alias.
7239 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7241 (vect_loop_versioning): Call it.
7242 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7244 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7246 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7247 Alan Hayward <alan.hayward@arm.com>
7248 David Sherwood <david.sherwood@arm.com>
7250 * doc/sourcebuild.texi (vect_scatter_store): Document.
7251 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7253 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7255 * genopinit.c (main): Add supports_vec_scatter_store and
7256 supports_vec_scatter_store_cached to target_optabs.
7257 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7258 IFN_MASK_SCATTER_STORE.
7259 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7261 * internal-fn.h (internal_store_fn_p): Declare.
7262 (internal_fn_stored_value_index): Likewise.
7263 * internal-fn.c (scatter_store_direct): New macro.
7264 (expand_scatter_store_optab_fn): New function.
7265 (direct_scatter_store_optab_supported_p): New macro.
7266 (internal_store_fn_p): New function.
7267 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7268 IFN_MASK_SCATTER_STORE.
7269 (internal_fn_mask_index): Likewise.
7270 (internal_fn_stored_value_index): New function.
7271 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7273 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7274 * optabs-query.c (supports_vec_scatter_store_p): New function.
7275 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7276 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7277 true for scatter stores.
7278 (vect_gather_scatter_fn_p): Handle scatter stores too.
7279 (vect_check_gather_scatter): Consider using scatter stores if
7280 supports_vec_scatter_store_p.
7281 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7283 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7284 internal_fn_stored_value_index.
7285 (check_load_store_masking): Handle scatter stores too.
7286 (vect_get_store_rhs): Make public.
7287 (vectorizable_call): Use internal_store_fn_p.
7288 (vectorizable_store): Handle scatter store internal functions.
7289 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7290 when deciding whether the end of the group has been reached.
7291 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7292 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7293 (mask_scatter_store<mode>): New insns.
7295 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7296 Alan Hayward <alan.hayward@arm.com>
7297 David Sherwood <david.sherwood@arm.com>
7299 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7300 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7301 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7303 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7304 Use vect_truncate_gather_scatter_offset if we can't treat the
7305 operation as a normal gather load or scatter store.
7306 (get_group_load_store_type): Take the gather_scatter_info
7307 as argument. Try using a gather load or scatter store for
7308 single-element groups.
7309 (get_load_store_type): Update calls to get_group_load_store_type
7310 and vect_use_strided_gather_scatters_p.
7312 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7313 Alan Hayward <alan.hayward@arm.com>
7314 David Sherwood <david.sherwood@arm.com>
7316 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7317 optional tree argument.
7318 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7320 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7321 but continue to use the current value as a fallback.
7322 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7323 to compare the updates.
7324 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7325 (get_load_store_type): Use it when handling a strided access.
7326 (vect_get_strided_load_store_ops): New function.
7327 (vect_get_data_ptr_increment): Likewise.
7328 (vectorizable_load): Handle strided gather loads. Always pass
7329 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7332 Alan Hayward <alan.hayward@arm.com>
7333 David Sherwood <david.sherwood@arm.com>
7335 * doc/md.texi (gather_load@var{m}): Document.
7336 (mask_gather_load@var{m}): Likewise.
7337 * genopinit.c (main): Add supports_vec_gather_load and
7338 supports_vec_gather_load_cached to target_optabs.
7339 * optabs-tree.c (init_tree_optimization_optabs): Use
7340 ggc_cleared_alloc to allocate target_optabs.
7341 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7342 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7344 * internal-fn.h (internal_load_fn_p): Declare.
7345 (internal_gather_scatter_fn_p): Likewise.
7346 (internal_fn_mask_index): Likewise.
7347 (internal_gather_scatter_fn_supported_p): Likewise.
7348 * internal-fn.c (gather_load_direct): New macro.
7349 (expand_gather_load_optab_fn): New function.
7350 (direct_gather_load_optab_supported_p): New macro.
7351 (direct_internal_fn_optab): New function.
7352 (internal_load_fn_p): Likewise.
7353 (internal_gather_scatter_fn_p): Likewise.
7354 (internal_fn_mask_index): Likewise.
7355 (internal_gather_scatter_fn_supported_p): Likewise.
7356 * optabs-query.c (supports_at_least_one_mode_p): New function.
7357 (supports_vec_gather_load_p): Likewise.
7358 * optabs-query.h (supports_vec_gather_load_p): Declare.
7359 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7360 and memory_type field.
7361 (NUM_PATTERNS): Bump to 15.
7362 * tree-vect-data-refs.c: Include internal-fn.h.
7363 (vect_gather_scatter_fn_p): New function.
7364 (vect_describe_gather_scatter_call): Likewise.
7365 (vect_check_gather_scatter): Try using internal functions for
7366 gather loads. Recognize existing calls to a gather load function.
7367 (vect_analyze_data_refs): Consider using gather loads if
7368 supports_vec_gather_load_p.
7369 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7370 (vect_get_gather_scatter_offset_type): Likewise.
7371 (vect_convert_mask_for_vectype): Likewise.
7372 (vect_add_conversion_to_patterm): Likewise.
7373 (vect_try_gather_scatter_pattern): Likewise.
7374 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7375 (vect_vect_recog_func_ptrs): Add it.
7376 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7377 internal_fn_mask_index and internal_gather_scatter_fn_p.
7378 (check_load_store_masking): Take the gather_scatter_info as an
7379 argument and handle gather loads.
7380 (vect_get_gather_scatter_ops): New function.
7381 (vectorizable_call): Check internal_load_fn_p.
7382 (vectorizable_load): Likewise. Handle gather load internal
7384 (vectorizable_store): Update call to check_load_store_masking.
7385 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7386 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7387 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7388 (aarch64_gather_scale_operand_d): New predicates.
7389 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7390 (mask_gather_load<mode>): New insns.
7392 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7393 Alan Hayward <alan.hayward@arm.com>
7394 David Sherwood <david.sherwood@arm.com>
7396 * optabs.def (fold_left_plus_optab): New optab.
7397 * doc/md.texi (fold_left_plus_@var{m}): Document.
7398 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7399 * internal-fn.c (fold_left_direct): Define.
7400 (expand_fold_left_optab_fn): Likewise.
7401 (direct_fold_left_optab_supported_p): Likewise.
7402 * fold-const-call.c (fold_const_fold_left): New function.
7403 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7404 * tree-parloops.c (valid_reduction_p): New function.
7405 (gather_scalar_reductions): Use it.
7406 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7407 (vect_finish_replace_stmt): Declare.
7408 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7409 (needs_fold_left_reduction_p): New function, split out from...
7410 (vect_is_simple_reduction): ...here. Accept reductions that
7411 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7412 (vect_force_simple_reduction): Also store the reduction type in
7413 the assignment's STMT_VINFO_REDUC_TYPE.
7414 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7415 (merge_with_identity): New function.
7416 (vect_expand_fold_left): Likewise.
7417 (vectorize_fold_left_reduction): Likewise.
7418 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7419 scalar phi in place for it. Check for target support and reject
7420 cases that would reassociate the operation. Defer the transform
7421 phase to vectorize_fold_left_reduction.
7422 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7423 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7424 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7426 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7428 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7429 call to ifc_temp_var.
7431 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7432 Alan Hayward <alan.hayward@arm.com>
7433 David Sherwood <david.sherwood@arm.com>
7435 * target.def (legitimize_address_displacement): Take the original
7436 offset as a poly_int.
7437 * targhooks.h (default_legitimize_address_displacement): Update
7439 * targhooks.c (default_legitimize_address_displacement): Likewise.
7440 * doc/tm.texi: Regenerate.
7441 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7442 as an argument, moving assert of ad->disp == ad->disp_term to...
7443 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7444 Try calling targetm.legitimize_address_displacement before expanding
7445 the address rather than afterwards, and adjust for the new interface.
7446 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7447 Match the new hook interface. Handle SVE addresses.
7448 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7451 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7453 * Makefile.in (OBJS): Add early-remat.o.
7454 * target.def (select_early_remat_modes): New hook.
7455 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7456 * doc/tm.texi: Regenerate.
7457 * targhooks.h (default_select_early_remat_modes): Declare.
7458 * targhooks.c (default_select_early_remat_modes): New function.
7459 * timevar.def (TV_EARLY_REMAT): New timevar.
7460 * passes.def (pass_early_remat): New pass.
7461 * tree-pass.h (make_pass_early_remat): Declare.
7462 * early-remat.c: New file.
7463 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7465 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7468 Alan Hayward <alan.hayward@arm.com>
7469 David Sherwood <david.sherwood@arm.com>
7471 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7472 vfm1 with a bound_epilog parameter.
7473 (vect_do_peeling): Update calls accordingly, and move the prologue
7474 call earlier in the function. Treat the base bound_epilog as 0 for
7475 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7476 this base when peeling for gaps.
7477 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7478 with fully-masked loops.
7479 (vect_estimate_min_profitable_iters): Handle the single peeled
7480 iteration in that case.
7482 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7483 Alan Hayward <alan.hayward@arm.com>
7484 David Sherwood <david.sherwood@arm.com>
7486 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7487 single-element interleaving even if the size is not a power of 2.
7488 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7489 accesses for single-element interleaving if the group size is
7492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7493 Alan Hayward <alan.hayward@arm.com>
7494 David Sherwood <david.sherwood@arm.com>
7496 * doc/md.texi (fold_extract_last_@var{m}): Document.
7497 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7498 * optabs.def (fold_extract_last_optab): New optab.
7499 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7500 * internal-fn.c (fold_extract_direct): New macro.
7501 (expand_fold_extract_optab_fn): Likewise.
7502 (direct_fold_extract_optab_supported_p): Likewise.
7503 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7504 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7505 EXTRACT_LAST_REDUCTION.
7506 (get_initial_def_for_reduction): Do not create an initial vector
7507 for EXTRACT_LAST_REDUCTION reductions.
7508 (vectorizable_reduction): Leave the scalar phi in place for
7509 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7510 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7511 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7512 transform phase to vectorizable_condition.
7513 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7515 (vect_finish_stmt_generation): ...here.
7516 (vect_finish_replace_stmt): New function.
7517 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7518 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7520 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7523 Alan Hayward <alan.hayward@arm.com>
7524 David Sherwood <david.sherwood@arm.com>
7526 * doc/md.texi (extract_last_@var{m}): Document.
7527 * optabs.def (extract_last_optab): New optab.
7528 * internal-fn.def (EXTRACT_LAST): New internal function.
7529 * internal-fn.c (cond_unary_direct): New macro.
7530 (expand_cond_unary_optab_fn): Likewise.
7531 (direct_cond_unary_optab_supported_p): Likewise.
7532 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7533 loops using EXTRACT_LAST.
7534 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7535 (extract_last_<mode>): ...this optab.
7536 (vec_extract<mode><Vel>): Update accordingly.
7538 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7539 Alan Hayward <alan.hayward@arm.com>
7540 David Sherwood <david.sherwood@arm.com>
7542 * target.def (empty_mask_is_expensive): New hook.
7543 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7544 * doc/tm.texi: Regenerate.
7545 * targhooks.h (default_empty_mask_is_expensive): Declare.
7546 * targhooks.c (default_empty_mask_is_expensive): New function.
7547 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7548 if the target says that empty masks are expensive.
7549 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7551 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7554 Alan Hayward <alan.hayward@arm.com>
7555 David Sherwood <david.sherwood@arm.com>
7557 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7558 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7559 (vect_use_loop_mask_for_alignment_p): New function.
7560 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7561 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7562 niters_skip argument. Make sure that the first niters_skip elements
7563 of the first iteration are inactive.
7564 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7565 Update call to vect_set_loop_masks_directly.
7566 (get_misalign_in_elems): New function, split out from...
7567 (vect_gen_prolog_loop_niters): ...here.
7568 (vect_update_init_of_dr): Take a code argument that specifies whether
7569 the adjustment should be added or subtracted.
7570 (vect_update_init_of_drs): Likewise.
7571 (vect_prepare_for_masked_peels): New function.
7572 (vect_do_peeling): Skip prologue peeling if we're using a mask
7573 instead. Update call to vect_update_inits_of_drs.
7574 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7576 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7577 alignment. Do not include the number of peeled iterations in
7578 the minimum threshold in that case.
7579 (vectorizable_induction): Adjust the start value down by
7580 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7581 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7582 Take the number of skipped iterations into account when calculating
7584 * tree-vect-stmts.c (vect_gen_while_not): New function.
7586 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7587 Alan Hayward <alan.hayward@arm.com>
7588 David Sherwood <david.sherwood@arm.com>
7590 * doc/sourcebuild.texi (vect_fully_masked): Document.
7591 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7593 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7595 (vect_analyze_loop_2): ...here. Don't check the vectorization
7596 factor against the number of loop iterations if the loop is
7599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7600 Alan Hayward <alan.hayward@arm.com>
7601 David Sherwood <david.sherwood@arm.com>
7603 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7604 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7605 (dump_groups): Update accordingly.
7606 (iv_use::mem_type): New member variable.
7607 (address_p): New function.
7608 (record_use): Add a mem_type argument and initialize the new
7610 (record_group_use): Add a mem_type argument. Use address_p.
7611 Remove obsolete null checks of base_object. Update call to record_use.
7612 (find_interesting_uses_op): Update call to record_group_use.
7613 (find_interesting_uses_cond): Likewise.
7614 (find_interesting_uses_address): Likewise.
7615 (get_mem_type_for_internal_fn): New function.
7616 (find_address_like_use): Likewise.
7617 (find_interesting_uses_stmt): Try find_address_like_use before
7618 calling find_interesting_uses_op.
7619 (addr_offset_valid_p): Use the iv mem_type field as the type
7620 of the addressed memory.
7621 (add_autoinc_candidates): Likewise.
7622 (get_address_cost): Likewise.
7623 (split_small_address_groups_p): Use address_p.
7624 (split_address_groups): Likewise.
7625 (add_iv_candidate_for_use): Likewise.
7626 (autoinc_possible_for_pair): Likewise.
7627 (rewrite_groups): Likewise.
7628 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7629 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7630 (get_alias_ptr_type_for_ptr_address): New function.
7631 (rewrite_use_address): Rewrite address uses in calls that were
7632 identified by find_address_like_use.
7634 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7635 Alan Hayward <alan.hayward@arm.com>
7636 David Sherwood <david.sherwood@arm.com>
7638 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7640 * gimple-expr.h (is_gimple_addressable: Likewise.
7641 * gimple-expr.c (is_gimple_address): Likewise.
7642 * internal-fn.c (expand_call_mem_ref): New function.
7643 (expand_mask_load_optab_fn): Use it.
7644 (expand_mask_store_optab_fn): Likewise.
7646 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7647 Alan Hayward <alan.hayward@arm.com>
7648 David Sherwood <david.sherwood@arm.com>
7650 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7651 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7652 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7653 (cond_umax@var{mode}): Document.
7654 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7655 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7656 (cond_umin_optab, cond_umax_optab): New optabs.
7657 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7658 (COND_IOR, COND_XOR): New internal functions.
7659 * internal-fn.h (get_conditional_internal_fn): Declare.
7660 * internal-fn.c (cond_binary_direct): New macro.
7661 (expand_cond_binary_optab_fn): Likewise.
7662 (direct_cond_binary_optab_supported_p): Likewise.
7663 (get_conditional_internal_fn): New function.
7664 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7665 Cope with reduction statements that are vectorized as calls rather
7667 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7668 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7669 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7670 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7671 (UNSPEC_COND_EOR): New unspecs.
7672 (optab): Add mappings for them.
7673 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7674 (sve_int_op, sve_fp_op): New int attributes.
7676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7677 Alan Hayward <alan.hayward@arm.com>
7678 David Sherwood <david.sherwood@arm.com>
7680 * optabs.def (while_ult_optab): New optab.
7681 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7682 * internal-fn.def (WHILE_ULT): New internal function.
7683 * internal-fn.h (direct_internal_fn_supported_p): New override
7684 that takes two types as argument.
7685 * internal-fn.c (while_direct): New macro.
7686 (expand_while_optab_fn): New function.
7687 (convert_optab_supported_p): Likewise.
7688 (direct_while_optab_supported_p): New macro.
7689 * wide-int.h (wi::udiv_ceil): New function.
7690 * tree-vectorizer.h (rgroup_masks): New structure.
7691 (vec_loop_masks): New typedef.
7692 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7694 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7695 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7696 (vect_max_vf): New function.
7697 (slpeel_make_loop_iterate_ntimes): Delete.
7698 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7699 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7700 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7701 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7702 internal-fn.h, stor-layout.h and optabs-query.h.
7703 (vect_set_loop_mask): New function.
7704 (add_preheader_seq): Likewise.
7705 (add_header_seq): Likewise.
7706 (interleave_supported_p): Likewise.
7707 (vect_maybe_permute_loop_masks): Likewise.
7708 (vect_set_loop_masks_directly): Likewise.
7709 (vect_set_loop_condition_masked): Likewise.
7710 (vect_set_loop_condition_unmasked): New function, split out from
7711 slpeel_make_loop_iterate_ntimes.
7712 (slpeel_make_loop_iterate_ntimes): Rename to..
7713 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7714 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7715 (vect_do_peeling): Update call accordingly.
7716 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7718 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7719 mask_compare_type, can_fully_mask_p and fully_masked_p.
7720 (release_vec_loop_masks): New function.
7721 (_loop_vec_info): Use it to free the loop masks.
7722 (can_produce_all_loop_masks_p): New function.
7723 (vect_get_max_nscalars_per_iter): Likewise.
7724 (vect_verify_full_masking): Likewise.
7725 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7726 retries, and free the mask rgroups before retrying. Check loop-wide
7727 reasons for disallowing fully-masked loops. Make the final decision
7728 about whether use a fully-masked loop or not.
7729 (vect_estimate_min_profitable_iters): Do not assume that peeling
7730 for the number of iterations will be needed for fully-masked loops.
7731 (vectorizable_reduction): Disable fully-masked loops.
7732 (vectorizable_live_operation): Likewise.
7733 (vect_halve_mask_nunits): New function.
7734 (vect_double_mask_nunits): Likewise.
7735 (vect_record_loop_mask): Likewise.
7736 (vect_get_loop_mask): Likewise.
7737 (vect_transform_loop): Handle the case in which the final loop
7738 iteration might handle a partial vector. Call vect_set_loop_condition
7739 instead of slpeel_make_loop_iterate_ntimes.
7740 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7741 (check_load_store_masking): New function.
7742 (prepare_load_store_mask): Likewise.
7743 (vectorizable_store): Handle fully-masked loops.
7744 (vectorizable_load): Likewise.
7745 (supportable_widening_operation): Use vect_halve_mask_nunits for
7747 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7748 (vect_gen_while): New function.
7749 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7750 (aarch64_uqdec<mode>): New insn.
7752 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7753 Alan Hayward <alan.hayward@arm.com>
7754 David Sherwood <david.sherwood@arm.com>
7756 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7757 (reduc_xor_scal_optab): New optabs.
7758 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7759 (reduc_xor_scal_@var{m}): Document.
7760 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7761 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7763 * fold-const-call.c (fold_const_call): Handle them.
7764 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7765 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7766 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7767 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7768 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7769 (UNSPEC_XORV): New unspecs.
7770 (optab): Add entries for them.
7771 (BITWISEV): New int iterator.
7772 (bit_reduc_op): New int attributes.
7774 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7775 Alan Hayward <alan.hayward@arm.com>
7776 David Sherwood <david.sherwood@arm.com>
7778 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7779 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7780 * optabs.def (vec_shl_insert_optab): New optab.
7781 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7782 (duplicate_and_interleave): Likewise.
7783 * tree-vect-loop.c: Include internal-fn.h.
7784 (neutral_op_for_slp_reduction): New function, split out from
7785 get_initial_defs_for_reduction.
7786 (get_initial_def_for_reduction): Handle option 2 for variable-length
7787 vectors by loading the neutral value into a vector and then shifting
7788 the initial value into element 0.
7789 (get_initial_defs_for_reduction): Replace the code argument with
7790 the neutral value calculated by neutral_op_for_slp_reduction.
7791 Use gimple_build_vector for constant-length vectors.
7792 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7793 but the first group_size elements have a neutral value.
7794 Use duplicate_and_interleave otherwise.
7795 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7796 Update call to get_initial_defs_for_reduction. Handle SLP
7797 reductions for variable-length vectors by creating one vector
7798 result for each scalar result, with the elements associated
7799 with other scalar results stubbed out with the neutral value.
7800 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7801 Require IFN_VEC_SHL_INSERT for double reductions on
7802 variable-length vectors, or SLP reductions that have
7803 a neutral value. Require can_duplicate_and_interleave_p
7804 support for variable-length unchained SLP reductions if there
7805 is no neutral value, such as for MIN/MAX reductions. Also require
7806 the number of vector elements to be a multiple of the number of
7807 SLP statements when doing variable-length unchained SLP reductions.
7808 Update call to vect_create_epilog_for_reduction.
7809 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7810 and remove initial values.
7811 (duplicate_and_interleave): Make public.
7812 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7813 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7815 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7816 Alan Hayward <alan.hayward@arm.com>
7817 David Sherwood <david.sherwood@arm.com>
7819 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7820 (can_duplicate_and_interleave_p): New function.
7821 (vect_get_and_check_slp_defs): Take the vector of statements
7822 rather than just the current one. Remove excess parentheses.
7823 Restriction rejectinon of vect_constant_def and vect_external_def
7824 for variable-length vectors to boolean types, or types for which
7825 can_duplicate_and_interleave_p is false.
7826 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7827 (duplicate_and_interleave): New function.
7828 (vect_get_constant_vectors): Use gimple_build_vector for
7829 constant-length vectors and suitable variable-length constant
7830 vectors. Use duplicate_and_interleave for other variable-length
7831 vectors. Don't defer the update when inserting new statements.
7833 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7834 Alan Hayward <alan.hayward@arm.com>
7835 David Sherwood <david.sherwood@arm.com>
7837 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7838 min_profitable_iters doesn't go negative.
7840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7841 Alan Hayward <alan.hayward@arm.com>
7842 David Sherwood <david.sherwood@arm.com>
7844 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7845 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7846 * optabs.def (vec_mask_load_lanes_optab): New optab.
7847 (vec_mask_store_lanes_optab): Likewise.
7848 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7849 (MASK_STORE_LANES): Likewise.
7850 * internal-fn.c (mask_load_lanes_direct): New macro.
7851 (mask_store_lanes_direct): Likewise.
7852 (expand_mask_load_optab_fn): Handle masked operations.
7853 (expand_mask_load_lanes_optab_fn): New macro.
7854 (expand_mask_store_optab_fn): Handle masked operations.
7855 (expand_mask_store_lanes_optab_fn): New macro.
7856 (direct_mask_load_lanes_optab_supported_p): Likewise.
7857 (direct_mask_store_lanes_optab_supported_p): Likewise.
7858 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7860 (vect_load_lanes_supported): Likewise.
7861 * tree-vect-data-refs.c (strip_conversion): New function.
7862 (can_group_stmts_p): Likewise.
7863 (vect_analyze_data_ref_accesses): Use it instead of checking
7864 for a pair of assignments.
7865 (vect_store_lanes_supported): Take a masked_p parameter.
7866 (vect_load_lanes_supported): Likewise.
7867 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7868 vect_store_lanes_supported and vect_load_lanes_supported.
7869 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7870 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7871 parameter. Don't allow gaps for masked accesses.
7872 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7873 and vect_load_lanes_supported.
7874 (get_load_store_type): Take a masked_p parameter and update
7875 call to get_group_load_store_type.
7876 (vectorizable_store): Update call to get_load_store_type.
7877 Handle IFN_MASK_STORE_LANES.
7878 (vectorizable_load): Update call to get_load_store_type.
7879 Handle IFN_MASK_LOAD_LANES.
7881 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7882 Alan Hayward <alan.hayward@arm.com>
7883 David Sherwood <david.sherwood@arm.com>
7885 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7887 * config/aarch64/aarch64-protos.h
7888 (aarch64_sve_struct_memory_operand_p): Declare.
7889 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7890 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7891 (VPRED, vpred): Handle SVE structure modes.
7892 * config/aarch64/constraints.md (Utx): New constraint.
7893 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7894 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7895 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7896 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7897 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7898 structure modes. Split into pieces after RA.
7899 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7900 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7902 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7903 SVE structure modes.
7904 (aarch64_classify_address): Likewise.
7905 (sizetochar): Move earlier in file.
7906 (aarch64_print_operand): Handle SVE register lists.
7907 (aarch64_array_mode): New function.
7908 (aarch64_sve_struct_memory_operand_p): Likewise.
7909 (TARGET_ARRAY_MODE): Redefine.
7911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7912 Alan Hayward <alan.hayward@arm.com>
7913 David Sherwood <david.sherwood@arm.com>
7915 * target.def (array_mode): New target hook.
7916 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7917 * doc/tm.texi: Regenerate.
7918 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7919 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7920 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7922 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7925 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7926 Alan Hayward <alan.hayward@arm.com>
7927 David Sherwood <david.sherwood@arm.com>
7929 * fold-const.c (fold_binary_loc): Check the argument types
7930 rather than the result type when testing for a vector operation.
7932 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7934 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7935 * doc/tm.texi: Regenerate.
7937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7938 Alan Hayward <alan.hayward@arm.com>
7939 David Sherwood <david.sherwood@arm.com>
7941 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7942 (sve): Document new AArch64 extension.
7943 * doc/md.texi (w): Extend the description of the AArch64
7944 constraint to include SVE vectors.
7945 (Upl, Upa): Document new AArch64 predicate constraints.
7946 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7948 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7949 (msve-vector-bits=): New option.
7950 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7951 SVE when these are disabled.
7952 (sve): New extension.
7953 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7954 modes. Adjust their number of units based on aarch64_sve_vg.
7955 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7956 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7957 aarch64_addr_query_type.
7958 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7959 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7960 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7961 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7962 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7963 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7964 (aarch64_simd_imm_zero_p): Delete.
7965 (aarch64_check_zero_based_sve_index_immediate): Declare.
7966 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7967 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7968 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7969 (aarch64_sve_float_mul_immediate_p): Likewise.
7970 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7972 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7973 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7974 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7975 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7976 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7977 (aarch64_regmode_natural_size): Likewise.
7978 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7979 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7981 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7982 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7983 for VG and the SVE predicate registers.
7984 (V_ALIASES): Add a "z"-prefixed alias.
7985 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7986 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7987 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7988 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7989 (REG_CLASS_NAMES): Add entries for them.
7990 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7991 and the predicate registers.
7992 (aarch64_sve_vg): Declare.
7993 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7994 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7995 (REGMODE_NATURAL_SIZE): Define.
7996 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7998 * config/aarch64/aarch64.c: Include cfgrtl.h.
7999 (simd_immediate_info): Add a constructor for series vectors,
8000 and an associated step field.
8001 (aarch64_sve_vg): New variable.
8002 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8003 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8004 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8005 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8006 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8007 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8008 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8009 (aarch64_get_mask_mode): New functions.
8010 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8011 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8012 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8013 predicate modes and predicate registers. Explicitly restrict
8014 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8015 to store a vector mode if it is recognized by
8016 aarch64_classify_vector_mode.
8017 (aarch64_regmode_natural_size): New function.
8018 (aarch64_hard_regno_caller_save_mode): Return the original mode
8020 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8021 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8022 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8023 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8025 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8026 does not overlap dest if the function is frame-related. Handle
8028 (aarch64_split_add_offset): New function.
8029 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8030 them aarch64_add_offset.
8031 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8032 and update call to aarch64_sub_sp.
8033 (aarch64_add_cfa_expression): New function.
8034 (aarch64_expand_prologue): Pass extra temporary registers to the
8035 functions above. Handle the case in which we need to emit new
8036 DW_CFA_expressions for registers that were originally saved
8037 relative to the stack pointer, but now have to be expressed
8038 relative to the frame pointer.
8039 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8041 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8042 IP0 and IP1 values for SVE frames.
8043 (aarch64_expand_vec_series): New function.
8044 (aarch64_expand_sve_widened_duplicate): Likewise.
8045 (aarch64_expand_sve_const_vector): Likewise.
8046 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8047 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8048 into the register, rather than emitting a SET directly.
8049 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8050 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8051 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8052 (offset_9bit_signed_scaled_p): New functions.
8053 (aarch64_replicate_bitmask_imm): New function.
8054 (aarch64_bitmask_imm): Use it.
8055 (aarch64_cannot_force_const_mem): Reject expressions involving
8056 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8057 (aarch64_classify_index): Handle SVE indices, by requiring
8058 a plain register index with a scale that matches the element size.
8059 (aarch64_classify_address): Handle SVE addresses. Assert that
8060 the mode of the address is VOIDmode or an integer mode.
8061 Update call to aarch64_classify_symbol.
8062 (aarch64_classify_symbolic_expression): Update call to
8063 aarch64_classify_symbol.
8064 (aarch64_const_vec_all_in_range_p): New function.
8065 (aarch64_print_vector_float_operand): Likewise.
8066 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8067 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8068 and the FP immediates 1.0 and 0.5.
8069 (aarch64_print_address_internal): Handle SVE addresses.
8070 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8071 (aarch64_regno_regclass): Handle predicate registers.
8072 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8074 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8075 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8076 (aarch64_convert_sve_vector_bits): New function.
8077 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8078 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8080 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8081 Handle SVE vector and predicate modes. Accept VL-based constants
8082 that need only one temporary register, and VL offsets that require
8083 no temporary registers.
8084 (aarch64_conditional_register_usage): Mark the predicate registers
8085 as fixed if SVE isn't available.
8086 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8087 Return true for SVE vector and predicate modes.
8088 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8089 rather than an unsigned int. Handle SVE modes.
8090 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8092 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8094 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8095 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8096 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8097 (aarch64_sve_float_mul_immediate_p): New functions.
8098 (aarch64_sve_valid_immediate): New function.
8099 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8100 Explicitly reject structure modes. Check for INDEX constants.
8101 Handle PTRUE and PFALSE constants.
8102 (aarch64_check_zero_based_sve_index_immediate): New function.
8103 (aarch64_simd_imm_zero_p): Delete.
8104 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8105 vector modes. Accept constants in the range of CNT[BHWD].
8106 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8107 ask for an Advanced SIMD mode.
8108 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8109 (aarch64_simd_vector_alignment): Handle SVE predicates.
8110 (aarch64_vectorize_preferred_vector_alignment): New function.
8111 (aarch64_simd_vector_alignment_reachable): Use it instead of
8113 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8114 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8116 (MAX_VECT_LEN): Delete.
8117 (expand_vec_perm_d): Add a vec_flags field.
8118 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8119 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8120 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8122 (aarch64_evpc_rev): Rename to...
8123 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8124 (aarch64_evpc_rev_global): New function.
8125 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8126 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8128 (aarch64_evpc_sve_tbl): New function.
8129 (aarch64_expand_vec_perm_const_1): Update after rename of
8130 aarch64_evpc_rev. Handle SVE permutes too, trying
8131 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8132 than aarch64_evpc_tbl.
8133 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8134 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8135 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8136 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8137 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8138 (aarch64_expand_sve_vcond): New functions.
8139 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8140 of aarch64_vector_mode_p.
8141 (aarch64_dwarf_poly_indeterminate_value): New function.
8142 (aarch64_compute_pressure_classes): Likewise.
8143 (aarch64_can_change_mode_class): Likewise.
8144 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8145 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8146 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8147 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8148 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8149 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8150 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8151 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8153 (Dn, Dl, Dr): Accept const as well as const_vector.
8154 (Dz): Likewise. Compare against CONST0_RTX.
8155 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8156 of "vector" where appropriate.
8157 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8158 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8159 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8160 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8161 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8162 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8163 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8164 (v_int_equiv): Extend to SVE modes.
8165 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8167 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8168 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8169 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8170 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8171 (SVE_COND_FP_CMP): New int iterators.
8172 (perm_hilo): Handle the new unpack unspecs.
8173 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8175 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8176 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8177 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8178 (aarch64_equality_operator, aarch64_constant_vector_operand)
8179 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8180 (aarch64_sve_nonimmediate_operand): Likewise.
8181 (aarch64_sve_general_operand): Likewise.
8182 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8183 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8184 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8185 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8186 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8187 (aarch64_sve_float_arith_immediate): Likewise.
8188 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8189 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8190 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8191 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8192 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8193 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8194 (aarch64_sve_float_arith_operand): Likewise.
8195 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8196 (aarch64_sve_float_mul_operand): Likewise.
8197 (aarch64_sve_vec_perm_operand): Likewise.
8198 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8199 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8200 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8201 as well as const_vector.
8202 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8203 in file. Use CONST0_RTX and CONSTM1_RTX.
8204 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8205 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8206 Use aarch64_simd_imm_zero.
8207 * config/aarch64/aarch64-sve.md: New file.
8208 * config/aarch64/aarch64.md: Include it.
8209 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8210 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8211 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8212 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8213 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8214 (sve): New attribute.
8215 (enabled): Disable instructions with the sve attribute unless
8217 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8218 aarch64_expand_mov_immediate.
8219 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8220 CNT[BHSD] immediates.
8221 (movti): Split CONST_POLY_INT moves into two halves.
8222 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8223 Split additions that need a temporary here if the destination
8224 is the stack pointer.
8225 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8226 (*add<mode>3_poly_1): New instruction.
8227 (set_clobber_cc): New expander.
8229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8231 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8232 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8233 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8234 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8235 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8236 Change innermode from fixed_mode_size to machine_mode.
8237 (simplify_subreg): Update call accordingly. Handle a constant-sized
8238 subreg of a variable-length CONST_VECTOR.
8240 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8241 Alan Hayward <alan.hayward@arm.com>
8242 David Sherwood <david.sherwood@arm.com>
8244 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8245 (add_offset_to_base): New function, split out from...
8246 (create_mem_ref): ...here. When handling a scale other than 1,
8247 check first whether the address is valid without the offset.
8248 Add it into the base if so, leaving the index and scale as-is.
8250 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8253 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8254 fold_for_warn before checking if arg2 is INTEGER_CST.
8256 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8258 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8259 (store_multiple_operation): Delete.
8260 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8261 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8262 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8263 guarded by TARGET_STRING.
8264 (rs6000_output_load_multiple): Delete.
8265 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8266 OPTION_MASK_STRING / TARGET_STRING handling.
8267 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8268 (const rs6000_opt_masks) <"string">: Change mask to 0.
8269 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8270 (MASK_STRING): Delete.
8271 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8273 (load_multiple): Delete.
8280 (store_multiple): Delete.
8287 (movmemsi_8reg): Delete.
8288 (corresponding unnamed define_insn): Delete.
8289 (movmemsi_6reg): Delete.
8290 (corresponding unnamed define_insn): Delete.
8291 (movmemsi_4reg): Delete.
8292 (corresponding unnamed define_insn): Delete.
8293 (movmemsi_2reg): Delete.
8294 (corresponding unnamed define_insn): Delete.
8295 (movmemsi_1reg): Delete.
8296 (corresponding unnamed define_insn): Delete.
8297 * config/rs6000/rs6000.opt (mno-string): New.
8298 (mstring): Replace by deprecation warning stub.
8299 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8301 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8303 * regrename.c (regrename_do_replace): If replacing the same
8304 reg multiple times, try to reuse last created gen_raw_REG.
8307 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8308 main to workaround a bug in GDB.
8310 2018-01-12 Tom de Vries <tom@codesourcery.com>
8313 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8315 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8317 PR rtl-optimization/80481
8318 * ira-color.c (get_cap_member): New function.
8319 (allocnos_conflict_by_live_ranges_p): Use it.
8320 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8321 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8323 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8326 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8327 (*saddl_se_1): Ditto.
8329 (*ssubl_se_1): Ditto.
8331 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8333 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8334 rather than wi::to_widest for DR_INITs.
8335 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8336 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8337 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8339 (vect_analyze_group_access_1): Note that here.
8341 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8343 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8344 polynomial type sizes.
8346 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8348 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8349 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8350 (gimple_add_tmp_var): Likewise.
8352 2018-01-12 Martin Liska <mliska@suse.cz>
8354 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8355 (gimple_alloc_sizes): Likewise.
8356 (dump_gimple_statistics): Use PRIu64 in printf format.
8357 * gimple.h: Change uint64_t to int.
8359 2018-01-12 Martin Liska <mliska@suse.cz>
8361 * tree-core.h: Use uint64_t instead of int.
8362 * tree.c (tree_node_counts): Likewise.
8363 (tree_node_sizes): Likewise.
8364 (dump_tree_statistics): Use PRIu64 in printf format.
8366 2018-01-12 Martin Liska <mliska@suse.cz>
8368 * Makefile.in: As qsort_chk is implemented in vec.c, add
8369 vec.o to linkage of gencfn-macros.
8370 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8371 passing the info to record_node_allocation_statistics.
8372 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8374 * ggc-common.c (struct ggc_usage): Add operator== and use
8375 it in operator< and compare function.
8376 * mem-stats.h (struct mem_usage): Likewise.
8377 * vec.c (struct vec_usage): Remove operator< and compare
8378 function. Can be simply inherited.
8380 2018-01-12 Martin Jambor <mjambor@suse.cz>
8383 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8384 * tree-ssa-math-opts.c: Include domwalk.h.
8385 (convert_mult_to_fma_1): New function.
8386 (fma_transformation_info): New type.
8387 (fma_deferring_state): Likewise.
8388 (cancel_fma_deferring): New function.
8389 (result_of_phi): Likewise.
8390 (last_fma_candidate_feeds_initial_phi): Likewise.
8391 (convert_mult_to_fma): Added deferring logic, split actual
8392 transformation to convert_mult_to_fma_1.
8393 (math_opts_dom_walker): New type.
8394 (math_opts_dom_walker::after_dom_children): New method, body moved
8395 here from pass_optimize_widening_mul::execute, added deferring logic
8397 (pass_optimize_widening_mul::execute): Moved most of code to
8398 math_opts_dom_walker::after_dom_children.
8399 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8400 * config/i386/i386.c (ix86_option_override_internal): Added
8401 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8403 2018-01-12 Richard Biener <rguenther@suse.de>
8406 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8407 inline instance vars.
8409 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8412 * config/rx/rx.c (rx_is_restricted_memory_address):
8415 2018-01-12 Richard Biener <rguenther@suse.de>
8417 PR tree-optimization/80846
8418 * target.def (split_reduction): New target hook.
8419 * targhooks.c (default_split_reduction): New function.
8420 * targhooks.h (default_split_reduction): Declare.
8421 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8422 target requests first reduce vectors by combining low and high
8424 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8425 (get_vectype_for_scalar_type_and_size): Export.
8426 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8427 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8428 * doc/tm.texi: Regenerate.
8429 * config/i386/i386.c (ix86_split_reduction): Implement
8430 TARGET_VECTORIZE_SPLIT_REDUCTION.
8432 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8435 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8436 in PIC mode except for TARGET_VXWORKS_RTP.
8437 * config/sparc/sparc.c: Include cfgrtl.h.
8438 (TARGET_INIT_PIC_REG): Define.
8439 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8440 (sparc_pic_register_p): New predicate.
8441 (sparc_legitimate_address_p): Use it.
8442 (sparc_legitimize_pic_address): Likewise.
8443 (sparc_delegitimize_address): Likewise.
8444 (sparc_mode_dependent_address_p): Likewise.
8445 (gen_load_pcrel_sym): Remove 4th parameter.
8446 (load_got_register): Adjust call to above. Remove obsolete stuff.
8447 (sparc_expand_prologue): Do not call load_got_register here.
8448 (sparc_flat_expand_prologue): Likewise.
8449 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8450 (sparc_use_pseudo_pic_reg): New function.
8451 (sparc_init_pic_reg): Likewise.
8452 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8453 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8455 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8457 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8458 Add item for branch_cost.
8460 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8462 PR rtl-optimization/83565
8463 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8464 not extend the result to a larger mode for rotate operations.
8465 (num_sign_bit_copies1): Likewise.
8467 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8470 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8472 Use values-Xc.o for -pedantic.
8473 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8475 2018-01-12 Martin Liska <mliska@suse.cz>
8478 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8480 (possible_polymorphic_call_targets): Use it.
8481 (ipa_devirt): Likewise.
8483 2018-01-12 Martin Liska <mliska@suse.cz>
8485 * profile-count.h (enum profile_quality): Use 0 as invalid
8486 enum value of profile_quality.
8488 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8490 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8491 -mext-string options.
8493 2018-01-12 Richard Biener <rguenther@suse.de>
8495 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8496 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8497 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8499 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8501 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8503 * configure.ac (--with-long-double-format): Add support for the
8504 configuration option to change the default long double format on
8506 * config.gcc (powerpc*-linux*-*): Likewise.
8507 * configure: Regenerate.
8508 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8509 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8510 used without modification.
8512 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8514 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8515 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8516 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8517 MISC_BUILTIN_SPEC_BARRIER.
8518 (rs6000_init_builtins): Likewise.
8519 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8521 (speculation_barrier): New define_insn.
8522 * doc/extend.texi: Document __builtin_speculation_barrier.
8524 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8527 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8528 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8529 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8531 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8532 integral modes instead of "ss" and "sd".
8533 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8534 vectors with 32-bit and 64-bit elements.
8535 (vecdupssescalarmodesuffix): New mode attribute.
8536 (vec_dup<mode>): Use it.
8538 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8541 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8542 frame if argument is passed on stack.
8544 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8547 * ree.c (combine_reaching_defs): Optimize also
8548 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8549 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8551 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8554 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8556 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8559 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8560 after they are computed.
8562 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8564 PR tree-optimization/83695
8565 * gimple-loop-linterchange.cc
8566 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8567 reset cached scev information after interchange.
8568 (pass_linterchange::execute): Remove call to scev_reset_htab.
8570 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8572 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8573 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8574 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8575 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8576 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8577 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8578 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8579 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8580 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8581 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8582 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8583 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8584 (V_lane_reg): Likewise.
8585 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8587 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8588 (vfmal_lane_low<mode>_intrinsic,
8589 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8590 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8591 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8592 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8593 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8594 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8596 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8598 * config/arm/arm-cpus.in (fp16fml): New feature.
8599 (ALL_SIMD): Add fp16fml.
8600 (armv8.2-a): Add fp16fml as an option.
8601 (armv8.3-a): Likewise.
8602 (armv8.4-a): Add fp16fml as part of fp16.
8603 * config/arm/arm.h (TARGET_FP16FML): Define.
8604 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8606 * config/arm/arm-modes.def (V2HF): Define.
8607 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8608 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8609 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8610 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8611 vfmsl_low, vfmsl_high): New set of builtins.
8612 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8613 (vfml_op): New code attribute.
8614 (VFMLHALVES): New int iterator.
8615 (VFML, VFMLSEL): New mode attributes.
8616 (V_reg): Define mapping for V2HF.
8617 (V_hi, V_lo): New mode attributes.
8618 (VF_constraint): Likewise.
8619 (vfml_half, vfml_half_selector): New int attributes.
8620 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8622 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8623 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8625 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8626 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8627 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8628 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8630 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8631 Document new effective target and option set.
8633 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8635 * config/arm/arm-cpus.in (armv8_4): New feature.
8636 (ARMv8_4a): New fgroup.
8637 (armv8.4-a): New arch.
8638 * config/arm/arm-tables.opt: Regenerate.
8639 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8640 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8641 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8642 Add matching rules for -march=armv8.4-a and extensions.
8643 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8645 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8648 * config/rx/rx.md (BW): New mode attribute.
8649 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8651 2018-01-11 Richard Biener <rguenther@suse.de>
8653 PR tree-optimization/83435
8654 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8655 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8656 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8658 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8659 Alan Hayward <alan.hayward@arm.com>
8660 David Sherwood <david.sherwood@arm.com>
8662 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8664 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8665 (aarch64_print_address_internal): Use it to check for a zero offset.
8667 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8668 Alan Hayward <alan.hayward@arm.com>
8669 David Sherwood <david.sherwood@arm.com>
8671 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8672 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8673 Return a poly_int64 rather than a HOST_WIDE_INT.
8674 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8675 rather than a HOST_WIDE_INT.
8676 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8677 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8678 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8679 final_offset from HOST_WIDE_INT to poly_int64.
8680 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8681 to_constant when getting the number of units in an Advanced SIMD
8683 (aarch64_builtin_vectorized_function): Check for a constant number
8685 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8687 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8688 attribute instead of GET_MODE_NUNITS.
8689 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8690 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8691 GET_MODE_SIZE for fixed-size registers.
8692 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8693 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8694 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8695 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8696 (aarch64_print_operand, aarch64_print_address_internal)
8697 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8698 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8699 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8700 Handle polynomial GET_MODE_SIZE.
8701 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8702 wider than SImode without modification.
8703 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8704 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8705 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8706 passing and returning SVE modes.
8707 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8708 rather than GEN_INT.
8709 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8710 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8711 (aarch64_allocate_and_probe_stack_space): Likewise.
8712 (aarch64_layout_frame): Cope with polynomial offsets.
8713 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8714 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8716 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8717 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8718 poly_int64 rather than a HOST_WIDE_INT.
8719 (aarch64_get_separate_components, aarch64_process_components)
8720 (aarch64_expand_prologue, aarch64_expand_epilogue)
8721 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8722 (aarch64_anchor_offset): New function, split out from...
8723 (aarch64_legitimize_address): ...here.
8724 (aarch64_builtin_vectorization_cost): Handle polynomial
8725 TYPE_VECTOR_SUBPARTS.
8726 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8728 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8729 number of elements from the PARALLEL rather than the mode.
8730 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8731 rather than GET_MODE_BITSIZE.
8732 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8733 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8734 (aarch64_expand_vec_perm_const_1): Handle polynomial
8735 d->perm.length () and d->perm elements.
8736 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8737 Apply to_constant to d->perm elements.
8738 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8739 polynomial CONST_VECTOR_NUNITS.
8740 (aarch64_move_pointer): Take amount as a poly_int64 rather
8742 (aarch64_progress_pointer): Avoid temporary variable.
8743 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8744 the mode attribute instead of GET_MODE.
8746 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8747 Alan Hayward <alan.hayward@arm.com>
8748 David Sherwood <david.sherwood@arm.com>
8750 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8751 x exists before using it.
8752 (aarch64_add_constant_internal): Rename to...
8753 (aarch64_add_offset_1): ...this. Replace regnum with separate
8754 src and dest rtxes. Handle the case in which they're different,
8755 including when the offset is zero. Replace scratchreg with an rtx.
8756 Use 2 additions if there is no spare register into which we can
8757 move a 16-bit constant.
8758 (aarch64_add_constant): Delete.
8759 (aarch64_add_offset): Replace reg with separate src and dest
8760 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8761 Use aarch64_add_offset_1.
8762 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8763 an rtx rather than an int. Take the delta as a poly_int64
8764 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8765 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8766 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8767 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8768 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8770 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8771 aarch64_add_constant.
8773 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8775 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8776 Use scalar_float_mode.
8778 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8780 * config/aarch64/aarch64-simd.md
8781 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8782 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8783 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8784 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8785 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8786 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8787 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8788 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8789 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8790 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8792 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8795 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8796 targ_options->x_arm_arch_string is non NULL.
8798 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8800 * config/aarch64/aarch64.h
8801 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8803 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8806 * expmed.c (emit_store_flag_force): Swap if const op0
8807 and change VOIDmode to mode of op0.
8809 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8811 PR rtl-optimization/83761
8812 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8813 than bytes to mode_for_size.
8815 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8818 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8819 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8822 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8825 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8826 when in layout mode.
8827 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8828 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8831 2018-01-10 Michael Collison <michael.collison@arm.com>
8833 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8834 * config/aarch64/aarch64-option-extension.def: Add
8835 AARCH64_OPT_EXTENSION of 'fp16fml'.
8836 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8837 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8838 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8839 * config/aarch64/constraints.md (Ui7): New constraint.
8840 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8841 (VFMLA_SEL_W): Ditto.
8844 (VFMLA16_LOW): New int iterator.
8845 (VFMLA16_HIGH): Ditto.
8846 (UNSPEC_FMLAL): New unspec.
8847 (UNSPEC_FMLSL): Ditto.
8848 (UNSPEC_FMLAL2): Ditto.
8849 (UNSPEC_FMLSL2): Ditto.
8850 (f16mac): New code attribute.
8851 * config/aarch64/aarch64-simd-builtins.def
8852 (aarch64_fmlal_lowv2sf): Ditto.
8853 (aarch64_fmlsl_lowv2sf): Ditto.
8854 (aarch64_fmlalq_lowv4sf): Ditto.
8855 (aarch64_fmlslq_lowv4sf): Ditto.
8856 (aarch64_fmlal_highv2sf): Ditto.
8857 (aarch64_fmlsl_highv2sf): Ditto.
8858 (aarch64_fmlalq_highv4sf): Ditto.
8859 (aarch64_fmlslq_highv4sf): Ditto.
8860 (aarch64_fmlal_lane_lowv2sf): Ditto.
8861 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8862 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8863 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8864 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8865 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8866 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8867 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8868 (aarch64_fmlal_lane_highv2sf): Ditto.
8869 (aarch64_fmlsl_lane_highv2sf): Ditto.
8870 (aarch64_fmlal_laneq_highv2sf): Ditto.
8871 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8872 (aarch64_fmlalq_lane_highv4sf): Ditto.
8873 (aarch64_fmlsl_lane_highv4sf): Ditto.
8874 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8875 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8876 * config/aarch64/aarch64-simd.md:
8877 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8878 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8879 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8880 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8881 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8882 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8883 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8884 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8885 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8886 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8887 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8888 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8889 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8890 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8891 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8892 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8893 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8894 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8895 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8896 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8897 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8898 (vfmlsl_low_u32): Ditto.
8899 (vfmlalq_low_u32): Ditto.
8900 (vfmlslq_low_u32): Ditto.
8901 (vfmlal_high_u32): Ditto.
8902 (vfmlsl_high_u32): Ditto.
8903 (vfmlalq_high_u32): Ditto.
8904 (vfmlslq_high_u32): Ditto.
8905 (vfmlal_lane_low_u32): Ditto.
8906 (vfmlsl_lane_low_u32): Ditto.
8907 (vfmlal_laneq_low_u32): Ditto.
8908 (vfmlsl_laneq_low_u32): Ditto.
8909 (vfmlalq_lane_low_u32): Ditto.
8910 (vfmlslq_lane_low_u32): Ditto.
8911 (vfmlalq_laneq_low_u32): Ditto.
8912 (vfmlslq_laneq_low_u32): Ditto.
8913 (vfmlal_lane_high_u32): Ditto.
8914 (vfmlsl_lane_high_u32): Ditto.
8915 (vfmlal_laneq_high_u32): Ditto.
8916 (vfmlsl_laneq_high_u32): Ditto.
8917 (vfmlalq_lane_high_u32): Ditto.
8918 (vfmlslq_lane_high_u32): Ditto.
8919 (vfmlalq_laneq_high_u32): Ditto.
8920 (vfmlslq_laneq_high_u32): Ditto.
8921 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8922 (AARCH64_FL_FOR_ARCH8_4): New.
8923 (AARCH64_ISA_F16FML): New ISA flag.
8924 (TARGET_F16FML): New feature flag for fp16fml.
8925 (doc/invoke.texi): Document new fp16fml option.
8927 2018-01-10 Michael Collison <michael.collison@arm.com>
8929 * config/aarch64/aarch64-builtins.c:
8930 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8931 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8932 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8933 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8934 (AARCH64_ISA_SHA3): New ISA flag.
8935 (TARGET_SHA3): New feature flag for sha3.
8936 * config/aarch64/iterators.md (sha512_op): New int attribute.
8937 (CRYPTO_SHA512): New int iterator.
8938 (UNSPEC_SHA512H): New unspec.
8939 (UNSPEC_SHA512H2): Ditto.
8940 (UNSPEC_SHA512SU0): Ditto.
8941 (UNSPEC_SHA512SU1): Ditto.
8942 * config/aarch64/aarch64-simd-builtins.def
8943 (aarch64_crypto_sha512hqv2di): New builtin.
8944 (aarch64_crypto_sha512h2qv2di): Ditto.
8945 (aarch64_crypto_sha512su0qv2di): Ditto.
8946 (aarch64_crypto_sha512su1qv2di): Ditto.
8947 (aarch64_eor3qv8hi): Ditto.
8948 (aarch64_rax1qv2di): Ditto.
8949 (aarch64_xarqv2di): Ditto.
8950 (aarch64_bcaxqv8hi): Ditto.
8951 * config/aarch64/aarch64-simd.md:
8952 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8953 (aarch64_crypto_sha512su0qv2di): Ditto.
8954 (aarch64_crypto_sha512su1qv2di): Ditto.
8955 (aarch64_eor3qv8hi): Ditto.
8956 (aarch64_rax1qv2di): Ditto.
8957 (aarch64_xarqv2di): Ditto.
8958 (aarch64_bcaxqv8hi): Ditto.
8959 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8960 (vsha512h2q_u64): Ditto.
8961 (vsha512su0q_u64): Ditto.
8962 (vsha512su1q_u64): Ditto.
8963 (veor3q_u16): Ditto.
8964 (vrax1q_u64): Ditto.
8966 (vbcaxq_u16): Ditto.
8967 * config/arm/types.md (crypto_sha512): New type attribute.
8968 (crypto_sha3): Ditto.
8969 (doc/invoke.texi): Document new sha3 option.
8971 2018-01-10 Michael Collison <michael.collison@arm.com>
8973 * config/aarch64/aarch64-builtins.c:
8974 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8975 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8976 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8977 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8978 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8979 (AARCH64_ISA_SM4): New ISA flag.
8980 (TARGET_SM4): New feature flag for sm4.
8981 * config/aarch64/aarch64-simd-builtins.def
8982 (aarch64_sm3ss1qv4si): Ditto.
8983 (aarch64_sm3tt1aq4si): Ditto.
8984 (aarch64_sm3tt1bq4si): Ditto.
8985 (aarch64_sm3tt2aq4si): Ditto.
8986 (aarch64_sm3tt2bq4si): Ditto.
8987 (aarch64_sm3partw1qv4si): Ditto.
8988 (aarch64_sm3partw2qv4si): Ditto.
8989 (aarch64_sm4eqv4si): Ditto.
8990 (aarch64_sm4ekeyqv4si): Ditto.
8991 * config/aarch64/aarch64-simd.md:
8992 (aarch64_sm3ss1qv4si): Ditto.
8993 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8994 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8995 (aarch64_sm4eqv4si): Ditto.
8996 (aarch64_sm4ekeyqv4si): Ditto.
8997 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8998 (sm3part_op): Ditto.
8999 (CRYPTO_SM3TT): Ditto.
9000 (CRYPTO_SM3PART): Ditto.
9001 (UNSPEC_SM3SS1): New unspec.
9002 (UNSPEC_SM3TT1A): Ditto.
9003 (UNSPEC_SM3TT1B): Ditto.
9004 (UNSPEC_SM3TT2A): Ditto.
9005 (UNSPEC_SM3TT2B): Ditto.
9006 (UNSPEC_SM3PARTW1): Ditto.
9007 (UNSPEC_SM3PARTW2): Ditto.
9008 (UNSPEC_SM4E): Ditto.
9009 (UNSPEC_SM4EKEY): Ditto.
9010 * config/aarch64/constraints.md (Ui2): New constraint.
9011 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9012 * config/arm/types.md (crypto_sm3): New type attribute.
9013 (crypto_sm4): Ditto.
9014 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9015 (vsm3tt1aq_u32): Ditto.
9016 (vsm3tt1bq_u32): Ditto.
9017 (vsm3tt2aq_u32): Ditto.
9018 (vsm3tt2bq_u32): Ditto.
9019 (vsm3partw1q_u32): Ditto.
9020 (vsm3partw2q_u32): Ditto.
9021 (vsm4eq_u32): Ditto.
9022 (vsm4ekeyq_u32): Ditto.
9023 (doc/invoke.texi): Document new sm4 option.
9025 2018-01-10 Michael Collison <michael.collison@arm.com>
9027 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9028 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9029 (AARCH64_FL_FOR_ARCH8_4): New.
9030 (AARCH64_FL_V8_4): New flag.
9031 (doc/invoke.texi): Document new armv8.4-a option.
9033 2018-01-10 Michael Collison <michael.collison@arm.com>
9035 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9036 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9037 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9038 * config/aarch64/aarch64-option-extension.def: Add
9039 AARCH64_OPT_EXTENSION of 'sha2'.
9040 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9041 (crypto): Disable sha2 and aes if crypto disabled.
9042 (crypto): Enable aes and sha2 if enabled.
9043 (simd): Disable sha2 and aes if simd disabled.
9044 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9046 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9047 (TARGET_SHA2): New feature flag for sha2.
9048 (TARGET_AES): New feature flag for aes.
9049 * config/aarch64/aarch64-simd.md:
9050 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9051 conditional on TARGET_AES.
9052 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9053 (aarch64_crypto_sha1hsi): Make pattern conditional
9055 (aarch64_crypto_sha1hv4si): Ditto.
9056 (aarch64_be_crypto_sha1hv4si): Ditto.
9057 (aarch64_crypto_sha1su1v4si): Ditto.
9058 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9059 (aarch64_crypto_sha1su0v4si): Ditto.
9060 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9061 (aarch64_crypto_sha256su0v4si): Ditto.
9062 (aarch64_crypto_sha256su1v4si): Ditto.
9063 (doc/invoke.texi): Document new aes and sha2 options.
9065 2018-01-10 Martin Sebor <msebor@redhat.com>
9067 PR tree-optimization/83781
9068 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9071 2018-01-11 Martin Sebor <msebor@gmail.com>
9072 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9074 PR tree-optimization/83501
9075 PR tree-optimization/81703
9077 * tree-ssa-strlen.c (get_string_cst): Rename...
9078 (get_string_len): ...to this. Handle global constants.
9079 (handle_char_store): Adjust.
9081 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9082 Jim Wilson <jimw@sifive.com>
9084 * config/riscv/riscv-protos.h (riscv_output_return): New.
9085 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9086 (riscv_attribute_table, riscv_output_return),
9087 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9088 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9089 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9091 (riscv_expand_prologue): Add early return for naked function.
9092 (riscv_expand_epilogue): Likewise.
9093 (riscv_function_ok_for_sibcall): Return false for naked function.
9094 (riscv_set_current_function): New.
9095 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9096 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9097 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9098 * doc/extend.texi (RISC-V Function Attributes): New.
9100 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9102 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9103 check for 128-bit long double before checking TCmode.
9104 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9105 128-bit long doubles before checking TFmode or TCmode.
9106 (FLOAT128_IBM_P): Likewise.
9108 2018-01-10 Martin Sebor <msebor@redhat.com>
9110 PR tree-optimization/83671
9111 * builtins.c (c_strlen): Unconditionally return zero for the empty
9113 Use -Warray-bounds for warnings.
9114 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9115 for non-constant array indices with COMPONENT_REF, arrays of
9116 arrays, and pointers to arrays.
9117 (gimple_fold_builtin_strlen): Determine and set length range for
9118 non-constant character arrays.
9120 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9123 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9126 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9128 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9130 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9133 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9134 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9135 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9136 indexed_or_indirect_operand predicate.
9137 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9138 (*vsx_le_perm_load_v8hi): Likewise.
9139 (*vsx_le_perm_load_v16qi): Likewise.
9140 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9141 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9142 (*vsx_le_perm_store_v8hi): Likewise.
9143 (*vsx_le_perm_store_v16qi): Likewise.
9144 (eight unnamed splitters): Likewise.
9146 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9148 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9149 * config/rs6000/emmintrin.h: Likewise.
9150 * config/rs6000/mmintrin.h: Likewise.
9151 * config/rs6000/xmmintrin.h: Likewise.
9153 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9156 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9158 * tree.c (tree_nop_conversion): Return true for location wrapper
9160 (maybe_wrap_with_location): New function.
9161 (selftest::check_strip_nops): New function.
9162 (selftest::test_location_wrappers): New function.
9163 (selftest::tree_c_tests): Call it.
9164 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9165 (maybe_wrap_with_location): New decl.
9166 (EXPR_LOCATION_WRAPPER_P): New macro.
9167 (location_wrapper_p): New inline function.
9168 (tree_strip_any_location_wrapper): New inline function.
9170 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9173 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9174 stack_realign_offset for the largest alignment of stack slot
9176 (ix86_find_max_used_stack_alignment): New function.
9177 (ix86_finalize_stack_frame_flags): Use it. Set
9178 max_used_stack_alignment if we don't realign stack.
9179 * config/i386/i386.h (machine_function): Add
9180 max_used_stack_alignment.
9182 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9184 * config/arm/arm.opt (-mbranch-cost): New option.
9185 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9188 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9191 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9192 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9194 2018-01-10 Richard Biener <rguenther@suse.de>
9197 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9198 early out so it also covers the case where we have a non-NULL
9201 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9203 PR tree-optimization/83753
9204 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9205 for non-strided grouped accesses if the number of elements is 1.
9207 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9210 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9211 * i386.h (TARGET_USE_GATHER): Define.
9212 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9214 2018-01-10 Martin Liska <mliska@suse.cz>
9217 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9218 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9220 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9221 CLEANUP_NO_PARTITIONING is not set.
9223 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9225 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9226 for vectors, as a partial revert of r254296.
9227 * rtl.h (const_vec_p): Delete.
9228 (const_vec_duplicate_p): Don't test for vector CONSTs.
9229 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9230 * expmed.c (make_tree): Likewise.
9233 * common.md (E, F): Use CONSTANT_P instead of checking for
9235 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9236 checking for CONST_VECTOR.
9238 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9241 * predict.c (force_edge_cold): Handle in more sane way edges
9244 2018-01-09 Carl Love <cel@us.ibm.com>
9246 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9248 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9249 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9250 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9251 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9252 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9253 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9254 * config/rs6000/rs6000-protos.h: Add extern defition for
9255 rs6000_generate_float2_double_code.
9256 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9258 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9259 (float2_v2df): Add define_expand.
9261 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9264 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9265 op_mode in the force_to_mode call.
9267 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9269 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9270 instead of checking each element individually.
9271 (aarch64_evpc_uzp): Likewise.
9272 (aarch64_evpc_zip): Likewise.
9273 (aarch64_evpc_ext): Likewise.
9274 (aarch64_evpc_rev): Likewise.
9275 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9276 instead of checking each element individually. Return true without
9278 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9279 whether all selected elements come from the same input, instead of
9280 checking each element individually. Remove calls to gen_rtx_REG,
9281 start_sequence and end_sequence and instead assert that no rtl is
9284 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9286 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9287 order of HIGH and CONST checks.
9289 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9291 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9292 if the destination isn't an SSA_NAME.
9294 2018-01-09 Richard Biener <rguenther@suse.de>
9296 PR tree-optimization/83668
9297 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9299 (canonicalize_loop_form): ... here, renamed from ...
9300 (canonicalize_loop_closed_ssa_form): ... this and amended to
9301 swap successor edges for loop exit blocks to make us use
9302 the RPO order we need for initial schedule generation.
9304 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9306 PR tree-optimization/64811
9307 * match.pd: When optimizing comparisons with Inf, avoid
9308 introducing or losing exceptions from comparisons with NaN.
9310 2018-01-09 Martin Liska <mliska@suse.cz>
9313 * asan.c (shadow_mem_size): Add gcc_assert.
9315 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9317 Don't save registers in main().
9320 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9321 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9322 * config/avr/avr.c (avr_set_current_function): Don't error if
9323 naked, OS_task or OS_main are specified at the same time.
9324 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9326 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9328 * common/config/avr/avr-common.c (avr_option_optimization_table):
9329 Switch on -mmain-is-OS_task for optimizing compilations.
9331 2018-01-09 Richard Biener <rguenther@suse.de>
9333 PR tree-optimization/83572
9334 * graphite.c: Include cfganal.h.
9335 (graphite_transform_loops): Connect infinite loops to exit
9336 and remove fake edges at the end.
9338 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9340 * ipa-inline.c (edge_badness): Revert accidental checkin.
9342 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9345 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9346 symbols; not inline clones.
9348 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9351 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9352 hard registers. Formatting fixes.
9354 PR preprocessor/83722
9355 * gcc.c (try_generate_repro): Pass
9356 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9357 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9360 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9361 Kito Cheng <kito.cheng@gmail.com>
9363 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9364 (riscv_leaf_function_p): Delete.
9365 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9367 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9369 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9371 (do_ifelse): New function.
9372 (do_isel): New function.
9373 (do_sub3): New function.
9374 (do_add3): New function.
9375 (do_load_mask_compare): New function.
9376 (do_overlap_load_compare): New function.
9377 (expand_compare_loop): New function.
9378 (expand_block_compare): Call expand_compare_loop() when appropriate.
9379 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9381 (-mblock-compare-inline-loop-limit): New option.
9383 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9386 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9387 Reverse order of second and third operands in first alternative.
9388 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9389 of first and second elements in UNSPEC_VPERMR vector.
9390 (altivec_expand_vec_perm_le): Likewise.
9392 2018-01-08 Jeff Law <law@redhat.com>
9394 PR rtl-optimizatin/81308
9395 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9396 (process_switch): If group_case_labels makes a change, then set
9398 (pass_convert_switch::execute): If a switch is converted, then
9399 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9401 PR rtl-optimization/81308
9402 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9405 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9407 PR target/83663 - Revert r255946
9408 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9409 generation for cases where splatting a value is not useful.
9410 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9411 across a vec_duplicate and a paradoxical subreg forming a vector
9412 mode to a vec_concat.
9414 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9416 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9417 -march=armv8.3-a variants.
9418 * config/arm/t-multilib: Likewise.
9419 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9421 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9423 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9425 (cceq_ior_compare_complement): Give it a name so I can use it, and
9426 change boolean_or_operator predicate to boolean_operator so it can
9427 be used to generate a crand.
9428 (eqne): New code iterator.
9429 (bd/bd_neg): New code_attrs.
9430 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9431 a single define_insn.
9432 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9433 decrement (bdnzt/bdnzf/bdzt/bdzf).
9434 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9435 with the new names of the branch decrement patterns, and added the
9436 names of the branch decrement conditional patterns.
9438 2018-01-08 Richard Biener <rguenther@suse.de>
9440 PR tree-optimization/83563
9441 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9444 2018-01-08 Richard Biener <rguenther@suse.de>
9447 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9449 2018-01-08 Richard Biener <rguenther@suse.de>
9451 PR tree-optimization/83685
9452 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9453 references to abnormals.
9455 2018-01-08 Richard Biener <rguenther@suse.de>
9458 * dwarf2out.c (output_indirect_strings): Handle empty
9459 skeleton_debug_str_hash.
9460 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9462 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9464 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9465 (emit_store_direct): Likewise.
9466 (arc_trampoline_adjust_address): Likewise.
9467 (arc_asm_trampoline_template): New function.
9468 (arc_initialize_trampoline): Use asm_trampoline_template.
9469 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9470 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9471 * config/arc/arc.md (flush_icache): Delete pattern.
9473 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9475 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9476 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9479 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9482 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9483 by not USED_FOR_TARGET.
9484 (make_pass_resolve_sw_modes): Likewise.
9486 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9488 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9491 2018-01-08 Richard Biener <rguenther@suse.de>
9494 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9496 2018-01-08 Richard Biener <rguenther@suse.de>
9499 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9501 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9504 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9505 basic blocks with a small number of successors.
9506 (convert_control_dep_chain_into_preds): Improve handling of
9508 (dump_predicates): Split apart into...
9509 (dump_pred_chain): ...here...
9510 (dump_pred_info): ...and here.
9511 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9512 (can_chain_union_be_invalidated_p): Improve check for invalidation
9514 (uninit_uses_cannot_happen): Avoid unnecessary if
9515 convert_control_dep_chain_into_preds yielded nothing.
9517 2018-01-06 Martin Sebor <msebor@redhat.com>
9519 PR tree-optimization/83640
9520 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9521 subtracting negative offset from size.
9522 (builtin_access::overlap): Adjust offset bounds of the access to fall
9523 within the size of the object if possible.
9525 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9527 PR rtl-optimization/83699
9528 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9529 extract_bit_field_as_subreg to cases in which the extracted
9530 value is also a vector.
9532 * lra-constraints.c (process_alt_operands): Test for the equivalence
9533 substitutions when detecting a possible reload cycle.
9535 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9538 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9539 by default if flag_selective_schedling{,2}. Formatting fixes.
9541 PR rtl-optimization/83682
9542 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9543 if it has non-VECTOR_MODE element mode.
9544 (vec_duplicate_p): Likewise.
9547 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9548 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9550 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9553 * config/i386/i386-builtin.def
9554 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9555 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9556 Require also OPTION_MASK_ISA_AVX512F in addition to
9557 OPTION_MASK_ISA_GFNI.
9558 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9559 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9560 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9561 to OPTION_MASK_ISA_GFNI.
9562 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9563 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9564 OPTION_MASK_ISA_AVX512BW.
9565 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9566 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9567 addition to OPTION_MASK_ISA_GFNI.
9568 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9569 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9570 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9571 to OPTION_MASK_ISA_GFNI.
9572 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9573 a requirement for all ISAs rather than any of them with a few
9575 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9577 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9578 bitmasks to be enabled with 3 exceptions, instead of requiring any
9579 enabled ISA with lots of exceptions.
9580 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9581 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9582 Change avx512bw in isa attribute to avx512f.
9583 * config/i386/sgxintrin.h: Add license boilerplate.
9584 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9585 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9586 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9587 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9589 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9590 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9591 temporarily sse2 rather than sse if not enabled already.
9594 * config/i386/sse.md (VI248_VLBW): Rename to ...
9595 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9596 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9597 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9598 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9599 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9600 mode iterator instead of VI248_VLBW.
9602 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9604 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9605 (record_modified): Skip clobbers; add debug output.
9606 (param_change_prob): Use sreal frequencies.
9608 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9610 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9611 punt for user-aligned variables.
9613 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9615 * tree-chrec.c (chrec_contains_symbols): Return true for
9618 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9621 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9622 of (x|y) == x for BICS pattern.
9624 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9626 PR tree-optimization/83605
9627 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9628 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9631 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9633 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9634 * config/epiphany/rtems.h: New file.
9636 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9637 Uros Bizjak <ubizjak@gmail.com>
9640 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9641 QIreg_operand instead of register_operand predicate.
9642 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9643 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9644 comments instead of -fmitigate[-_]rop.
9646 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9649 * cgraphunit.c (symbol_table::compile): Switch to text_section
9650 before calling assembly_start debug hook.
9651 * run-rtl-passes.c (run_rtl_passes): Likewise.
9654 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9656 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9657 range_int_cst_p rather than !symbolic_range_p before calling
9658 extract_range_from_multiplicative_op_1.
9660 2018-01-04 Jeff Law <law@redhat.com>
9662 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9663 redundant test in assertion.
9665 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9667 * doc/rtl.texi: Document machine_mode wrapper classes.
9669 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9671 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9674 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9676 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9677 the VEC_PERM_EXPR fold to fail.
9679 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9682 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9683 to switched_sections.
9685 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9688 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9691 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9694 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9695 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9697 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9700 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9701 is BLKmode and bitpos not zero or mode change is needed.
9703 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9706 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9709 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9712 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9713 instead of MULT rtx. Update all corresponding splitters.
9715 (*ssub<modesuffix>): Ditto.
9717 (*cmp_sadd_di): Update split patterns.
9718 (*cmp_sadd_si): Ditto.
9719 (*cmp_sadd_sidi): Ditto.
9720 (*cmp_ssub_di): Ditto.
9721 (*cmp_ssub_si): Ditto.
9722 (*cmp_ssub_sidi): Ditto.
9723 * config/alpha/predicates.md (const23_operand): New predicate.
9724 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9725 Look for ASHIFT, not MULT inner operand.
9726 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9728 2018-01-04 Martin Liska <mliska@suse.cz>
9730 PR gcov-profile/83669
9731 * gcov.c (output_intermediate_file): Add version to intermediate
9733 * doc/gcov.texi: Document new field 'version' in intermediate
9734 file format. Fix location of '-k' option of gcov command.
9736 2018-01-04 Martin Liska <mliska@suse.cz>
9739 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9741 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9743 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9745 2018-01-03 Martin Sebor <msebor@redhat.com>
9747 PR tree-optimization/83655
9748 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9749 checking calls with invalid arguments.
9751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9753 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9754 (vectorizable_mask_load_store): Delete.
9755 (vectorizable_call): Return false for masked loads and stores.
9756 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9757 instead of gimple_assign_rhs1.
9758 (vectorizable_load): Handle IFN_MASK_LOAD.
9759 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9763 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9765 (vectorizable_mask_load_store): ...here.
9766 (vectorizable_load): ...and here.
9768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9770 * tree-vect-stmts.c (vect_build_all_ones_mask)
9771 (vect_build_zero_merge_argument): New functions, split out from...
9772 (vectorizable_load): ...here.
9774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9776 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9778 (vectorizable_mask_load_store): ...here.
9779 (vectorizable_store): ...and here.
9781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9783 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9785 (vectorizable_mask_load_store): ...here.
9787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9789 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9790 (vect_model_store_cost): Take a vec_load_store_type instead of a
9792 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9793 (vect_model_store_cost): Take a vec_load_store_type instead of a
9795 (vectorizable_mask_load_store): Update accordingly.
9796 (vectorizable_store): Likewise.
9797 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9801 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9802 IFN_MASK_LOAD calls here rather than...
9803 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9806 Alan Hayward <alan.hayward@arm.com>
9807 David Sherwood <david.sherwood@arm.com>
9809 * expmed.c (extract_bit_field_1): For vector extracts,
9810 fall back to extract_bit_field_as_subreg if vec_extract
9813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9814 Alan Hayward <alan.hayward@arm.com>
9815 David Sherwood <david.sherwood@arm.com>
9817 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9818 they are variable or constant sized.
9819 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9820 slots for constant-sized data.
9822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9823 Alan Hayward <alan.hayward@arm.com>
9824 David Sherwood <david.sherwood@arm.com>
9826 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9827 handling COND_EXPRs with boolean comparisons, try to find a better
9828 basis for the mask type than the boolean itself.
9830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9832 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9833 is calculated and how it can be overridden.
9834 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9835 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9837 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9841 Alan Hayward <alan.hayward@arm.com>
9842 David Sherwood <david.sherwood@arm.com>
9844 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9845 Remove the mode argument.
9846 (aarch64_simd_valid_immediate): Remove the mode and inverse
9848 * config/aarch64/iterators.md (bitsize): New iterator.
9849 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9850 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9851 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9852 aarch64_simd_valid_immediate.
9853 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9854 (aarch64_reg_or_bic_imm): Likewise.
9855 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9856 with an insn_type enum and msl with a modifier_type enum.
9857 Replace element_width with a scalar_mode. Change the shift
9858 to unsigned int. Add constructors for scalar_float_mode and
9859 scalar_int_mode elements.
9860 (aarch64_vect_float_const_representable_p): Delete.
9861 (aarch64_can_const_movi_rtx_p)
9862 (aarch64_simd_scalar_immediate_valid_for_move)
9863 (aarch64_simd_make_constant): Update call to
9864 aarch64_simd_valid_immediate.
9865 (aarch64_advsimd_valid_immediate_hs): New function.
9866 (aarch64_advsimd_valid_immediate): Likewise.
9867 (aarch64_simd_valid_immediate): Remove mode and inverse
9868 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9869 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9870 and aarch64_float_const_representable_p on the result.
9871 (aarch64_output_simd_mov_immediate): Remove mode argument.
9872 Update call to aarch64_simd_valid_immediate and use of
9873 simd_immediate_info.
9874 (aarch64_output_scalar_simd_mov_immediate): Update call
9877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9878 Alan Hayward <alan.hayward@arm.com>
9879 David Sherwood <david.sherwood@arm.com>
9881 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9882 (mode_nunits): Likewise CONST_MODE_NUNITS.
9883 * machmode.def (ADJUST_NUNITS): Document.
9884 * genmodes.c (mode_data::need_nunits_adj): New field.
9885 (blank_mode): Update accordingly.
9886 (adj_nunits): New variable.
9887 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9889 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9890 listed in adj_nunits.
9891 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9892 listed in adj_nunits. Don't emit case statements for such modes.
9893 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9894 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9895 nothing if adj_nunits is nonnull.
9896 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9897 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9898 (emit_mode_fbit): Update use of print_maybe_const_decl.
9899 (emit_move_size): Likewise. Treat the array as non-const
9901 (emit_mode_adjustments): Handle adj_nunits.
9903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9905 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9906 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9907 (VECTOR_MODES): Use it.
9908 (make_vector_modes): Take the prefix as an argument.
9910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9911 Alan Hayward <alan.hayward@arm.com>
9912 David Sherwood <david.sherwood@arm.com>
9914 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9915 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9916 for MODE_VECTOR_BOOL.
9917 * machmode.def (VECTOR_BOOL_MODE): Document.
9918 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9919 (make_vector_bool_mode): New function.
9920 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9922 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9923 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9925 * stor-layout.c (int_mode_for_mode): Likewise.
9926 * tree.c (build_vector_type_for_mode): Likewise.
9927 * varasm.c (output_constant_pool_2): Likewise.
9928 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9929 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9930 for MODE_VECTOR_BOOL.
9931 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9932 of mode class checks.
9933 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9934 instead of a list of mode class checks.
9935 (expand_vector_scalar_condition): Likewise.
9936 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9939 Alan Hayward <alan.hayward@arm.com>
9940 David Sherwood <david.sherwood@arm.com>
9942 * machmode.h (mode_size): Change from unsigned short to
9944 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9945 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9946 or if measurement_type is not polynomial.
9947 (fixed_size_mode::includes_p): Check for constant-sized modes.
9948 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9949 return a poly_uint16 rather than an unsigned short.
9950 (emit_mode_size): Change the type of mode_size from unsigned short
9951 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9952 (emit_mode_adjustments): Cope with polynomial vector sizes.
9953 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9955 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9957 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9958 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9959 * caller-save.c (setup_save_areas): Likewise.
9960 (replace_reg_with_saved_mem): Likewise.
9961 * calls.c (emit_library_call_value_1): Likewise.
9962 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9963 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9964 (gen_lowpart_for_combine): Likewise.
9965 * convert.c (convert_to_integer_1): Likewise.
9966 * cse.c (equiv_constant, cse_insn): Likewise.
9967 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9968 (cselib_subst_to_values): Likewise.
9969 * dce.c (word_dce_process_block): Likewise.
9970 * df-problems.c (df_word_lr_mark_ref): Likewise.
9971 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9972 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9973 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9974 (rtl_for_decl_location): Likewise.
9975 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9976 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9977 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9978 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9979 (expand_expr_real_1): Likewise.
9980 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9981 (pad_below): Likewise.
9982 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9983 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9984 * ira.c (get_subreg_tracking_sizes): Likewise.
9985 * ira-build.c (ira_create_allocno_objects): Likewise.
9986 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9987 (ira_sort_regnos_for_alter_reg): Likewise.
9988 * ira-costs.c (record_operand_costs): Likewise.
9989 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9990 (resolve_simple_move): Likewise.
9991 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9992 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9993 (lra_constraints): Likewise.
9994 (CONST_POOL_OK_P): Reject variable-sized modes.
9995 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9996 (add_pseudo_to_slot, lra_spill): Likewise.
9997 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9998 * optabs-query.c (get_best_extraction_insn): Likewise.
9999 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10000 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10001 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10002 * recog.c (offsettable_address_addr_space_p): Likewise.
10003 * regcprop.c (maybe_mode_change): Likewise.
10004 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10005 * regrename.c (build_def_use): Likewise.
10006 * regstat.c (dump_reg_info): Likewise.
10007 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10008 (find_reloads, find_reloads_subreg_address): Likewise.
10009 * reload1.c (eliminate_regs_1): Likewise.
10010 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10011 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10012 (simplify_binary_operation_1, simplify_subreg): Likewise.
10013 * targhooks.c (default_function_arg_padding): Likewise.
10014 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10015 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10016 (verify_gimple_assign_ternary): Likewise.
10017 * tree-inline.c (estimate_move_cost): Likewise.
10018 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10019 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10020 (get_address_cost_ainc): Likewise.
10021 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10022 (vect_supportable_dr_alignment): Likewise.
10023 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10024 (vectorizable_reduction): Likewise.
10025 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10026 (vectorizable_operation, vectorizable_load): Likewise.
10027 * tree.c (build_same_sized_truth_vector_type): Likewise.
10028 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10029 * var-tracking.c (emit_note_insn_var_location): Likewise.
10030 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10031 (ADDR_VEC_ALIGN): Likewise.
10033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10034 Alan Hayward <alan.hayward@arm.com>
10035 David Sherwood <david.sherwood@arm.com>
10037 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10039 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10040 or if measurement_type is polynomial.
10041 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10042 * combine.c (make_extraction): Likewise.
10043 * dse.c (find_shift_sequence): Likewise.
10044 * dwarf2out.c (mem_loc_descriptor): Likewise.
10045 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10046 (extract_bit_field, extract_low_bits): Likewise.
10047 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10048 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10049 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10050 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10051 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10052 * reload.c (find_reloads): Likewise.
10053 * reload1.c (alter_reg): Likewise.
10054 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10055 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10056 * tree-if-conv.c (predicate_mem_writes): Likewise.
10057 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10058 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10059 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10060 * valtrack.c (dead_debug_insert_temp): Likewise.
10061 * varasm.c (mergeable_constant_section): Likewise.
10062 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10065 Alan Hayward <alan.hayward@arm.com>
10066 David Sherwood <david.sherwood@arm.com>
10068 * expr.c (expand_assignment): Cope with polynomial mode sizes
10069 when assigning to a CONCAT.
10071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10072 Alan Hayward <alan.hayward@arm.com>
10073 David Sherwood <david.sherwood@arm.com>
10075 * machmode.h (mode_precision): Change from unsigned short to
10077 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10079 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10080 or if measurement_type is not polynomial.
10081 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10082 in which the mode is already known to be a scalar_int_mode.
10083 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10084 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10086 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10087 for GET_MODE_PRECISION.
10088 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10089 for GET_MODE_PRECISION.
10090 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10092 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10093 (expand_field_assignment, make_extraction): Likewise.
10094 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10095 (get_last_value): Likewise.
10096 * convert.c (convert_to_integer_1): Likewise.
10097 * cse.c (cse_insn): Likewise.
10098 * expr.c (expand_expr_real_1): Likewise.
10099 * lra-constraints.c (simplify_operand_subreg): Likewise.
10100 * optabs-query.c (can_atomic_load_p): Likewise.
10101 * optabs.c (expand_atomic_load): Likewise.
10102 (expand_atomic_store): Likewise.
10103 * ree.c (combine_reaching_defs): Likewise.
10104 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10105 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10106 * tree.h (type_has_mode_precision_p): Likewise.
10107 * ubsan.c (instrument_si_overflow): Likewise.
10109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10110 Alan Hayward <alan.hayward@arm.com>
10111 David Sherwood <david.sherwood@arm.com>
10113 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10114 polynomial numbers of units.
10115 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10116 (valid_vector_subparts_p): New function.
10117 (build_vector_type): Remove temporary shim and take the number
10118 of units as a poly_uint64 rather than an int.
10119 (build_opaque_vector_type): Take the number of units as a
10120 poly_uint64 rather than an int.
10121 * tree.c (build_vector_from_ctor): Handle polynomial
10122 TYPE_VECTOR_SUBPARTS.
10123 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10124 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10125 (build_vector_from_val): If the number of units is variable,
10126 use build_vec_duplicate_cst for constant operands and
10127 VEC_DUPLICATE_EXPR otherwise.
10128 (make_vector_type): Remove temporary is_constant ().
10129 (build_vector_type, build_opaque_vector_type): Take the number of
10130 units as a poly_uint64 rather than an int.
10131 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10133 * cfgexpand.c (expand_debug_expr): Likewise.
10134 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10135 (store_constructor, expand_expr_real_1): Likewise.
10136 (const_scalar_mask_from_tree): Likewise.
10137 * fold-const-call.c (fold_const_reduction): Likewise.
10138 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10139 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10140 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10141 (fold_relational_const): Likewise.
10142 (native_interpret_vector): Likewise. Change the size from an
10143 int to an unsigned int.
10144 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10145 TYPE_VECTOR_SUBPARTS.
10146 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10147 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10148 duplicating a non-constant operand into a variable-length vector.
10149 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10150 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10151 * ipa-icf.c (sem_variable::equals): Likewise.
10152 * match.pd: Likewise.
10153 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10154 * print-tree.c (print_node): Likewise.
10155 * stor-layout.c (layout_type): Likewise.
10156 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10157 * tree-cfg.c (verify_gimple_comparison): Likewise.
10158 (verify_gimple_assign_binary): Likewise.
10159 (verify_gimple_assign_ternary): Likewise.
10160 (verify_gimple_assign_single): Likewise.
10161 * tree-pretty-print.c (dump_generic_node): Likewise.
10162 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10163 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10164 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10165 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10166 (vect_shift_permute_load_chain): Likewise.
10167 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10168 (expand_vector_condition, optimize_vector_constructor): Likewise.
10169 (lower_vec_perm, get_compute_type): Likewise.
10170 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10171 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10172 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10173 (vect_recog_mask_conversion_pattern): Likewise.
10174 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10175 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10176 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10177 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10178 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10179 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10180 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10181 (supportable_widening_operation): Likewise.
10182 (supportable_narrowing_operation): Likewise.
10183 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10185 * varasm.c (output_constant): Likewise.
10187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10188 Alan Hayward <alan.hayward@arm.com>
10189 David Sherwood <david.sherwood@arm.com>
10191 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10192 so that both the length == 3 and length != 3 cases set up their
10193 own permute vectors. Add comments explaining why we know the
10194 number of elements is constant.
10195 (vect_permute_load_chain): Likewise.
10197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10198 Alan Hayward <alan.hayward@arm.com>
10199 David Sherwood <david.sherwood@arm.com>
10201 * machmode.h (mode_nunits): Change from unsigned char to
10203 (ONLY_FIXED_SIZE_MODES): New macro.
10204 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10205 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10206 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10208 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10209 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10210 or if measurement_type is not polynomial.
10211 * genmodes.c (ZERO_COEFFS): New macro.
10212 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10214 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10215 Use ZERO_COEFFS when emitting initializers.
10216 * data-streamer.h (bp_pack_poly_value): New function.
10217 (bp_unpack_poly_value): Likewise.
10218 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10219 for GET_MODE_NUNITS.
10220 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10221 for GET_MODE_NUNITS.
10222 * tree.c (make_vector_type): Remove temporary shim and make
10223 the real function take the number of units as a poly_uint64
10224 rather than an int.
10225 (build_vector_type_for_mode): Handle polynomial nunits.
10226 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10227 * emit-rtl.c (const_vec_series_p_1): Likewise.
10228 (gen_rtx_CONST_VECTOR): Likewise.
10229 * fold-const.c (test_vec_duplicate_folding): Likewise.
10230 * genrecog.c (validate_pattern): Likewise.
10231 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10232 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10233 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10234 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10235 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10236 * rtlanal.c (subreg_get_info): Likewise.
10237 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10238 (vect_grouped_load_supported): Likewise.
10239 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10240 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10241 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10242 (simplify_const_unary_operation, simplify_binary_operation_1)
10243 (simplify_const_binary_operation, simplify_ternary_operation)
10244 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10245 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10246 instead of CONST_VECTOR_NUNITS.
10247 * varasm.c (output_constant_pool_2): Likewise.
10248 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10249 explicit-encoded elements in the XVEC for variable-length vectors.
10251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10253 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10256 Alan Hayward <alan.hayward@arm.com>
10257 David Sherwood <david.sherwood@arm.com>
10259 * coretypes.h (fixed_size_mode): Declare.
10260 (fixed_size_mode_pod): New typedef.
10261 * builtins.h (target_builtins::x_apply_args_mode)
10262 (target_builtins::x_apply_result_mode): Change type to
10263 fixed_size_mode_pod.
10264 * builtins.c (apply_args_size, apply_result_size, result_vector)
10265 (expand_builtin_apply_args_1, expand_builtin_apply)
10266 (expand_builtin_return): Update accordingly.
10268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10270 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10271 * cselib.c (cselib_hash_rtx): Likewise.
10272 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10273 CONST_VECTOR encoding.
10275 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10276 Jeff Law <law@redhat.com>
10279 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10280 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10281 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10282 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10285 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10286 explicitly probe *sp in a noreturn function if there were any callee
10287 register saves or frame pointer is needed.
10289 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10292 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10293 BLKmode for ternary, binary or unary expressions.
10296 * var-tracking.c (delete_vta_debug_insn): New inline function.
10297 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10298 insns from get_insns () to NULL instead of each bb separately.
10299 Use delete_vta_debug_insn. No longer static.
10300 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10301 delete_vta_debug_insns callers.
10302 * rtl.h (delete_vta_debug_insns): Declare.
10303 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10304 instead of variable_tracking_main.
10306 2018-01-03 Martin Sebor <msebor@redhat.com>
10308 PR tree-optimization/83603
10309 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10310 arguments past the endof the argument list in functions declared
10311 without a prototype.
10312 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10313 Avoid checking when arguments are null.
10315 2018-01-03 Martin Sebor <msebor@redhat.com>
10318 * doc/extend.texi (attribute const): Fix a typo.
10319 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10320 issuing -Wsuggest-attribute for void functions.
10322 2018-01-03 Martin Sebor <msebor@redhat.com>
10324 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10325 offset_int::from instead of wide_int::to_shwi.
10326 (maybe_diag_overlap): Remove assertion.
10327 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10328 * gimple-ssa-sprintf.c (format_directive): Same.
10329 (parse_directive): Same.
10330 (sprintf_dom_walker::compute_format_length): Same.
10331 (try_substitute_return_value): Same.
10333 2018-01-03 Jeff Law <law@redhat.com>
10335 PR middle-end/83654
10336 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10337 non-constant residual for zero at runtime and avoid probing in
10338 that case. Reorganize code for trailing problem to mirror handling
10341 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10343 PR tree-optimization/83501
10344 * tree-ssa-strlen.c (get_string_cst): New.
10345 (handle_char_store): Call get_string_cst.
10347 2018-01-03 Martin Liska <mliska@suse.cz>
10349 PR tree-optimization/83593
10350 * tree-ssa-strlen.c: Include tree-cfg.h.
10351 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10352 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10353 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10355 (strlen_dom_walker::before_dom_children): Call
10356 gimple_purge_dead_eh_edges. Dump tranformation with details
10358 (strlen_dom_walker::before_dom_children): Update call by adding
10359 new argument cleanup_eh.
10360 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10362 2018-01-03 Martin Liska <mliska@suse.cz>
10365 * cif-code.def (VARIADIC_THUNK): New enum value.
10366 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10369 2018-01-03 Jan Beulich <jbeulich@suse.com>
10371 * sse.md (mov<mode>_internal): Tighten condition for when to use
10372 vmovdqu<ssescalarsize> for TI and OI modes.
10374 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10376 Update copyright years.
10378 2018-01-03 Martin Liska <mliska@suse.cz>
10381 * ipa-visibility.c (function_and_variable_visibility): Skip
10382 functions with noipa attribure.
10384 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10386 * gcc.c (process_command): Update copyright notice dates.
10387 * gcov-dump.c (print_version): Ditto.
10388 * gcov.c (print_version): Ditto.
10389 * gcov-tool.c (print_version): Ditto.
10390 * gengtype.c (create_file): Ditto.
10391 * doc/cpp.texi: Bump @copying's copyright year.
10392 * doc/cppinternals.texi: Ditto.
10393 * doc/gcc.texi: Ditto.
10394 * doc/gccint.texi: Ditto.
10395 * doc/gcov.texi: Ditto.
10396 * doc/install.texi: Ditto.
10397 * doc/invoke.texi: Ditto.
10399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10401 * vector-builder.h (vector_builder::m_full_nelts): Change from
10402 unsigned int to poly_uint64.
10403 (vector_builder::full_nelts): Update prototype accordingly.
10404 (vector_builder::new_vector): Likewise.
10405 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10406 (vector_builder::operator ==): Likewise.
10407 (vector_builder::finalize): Likewise.
10408 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10409 Take the number of elements as a poly_uint64 rather than an
10411 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10412 from unsigned int to poly_uint64.
10413 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10414 (vec_perm_indices::new_vector): Likewise.
10415 (vec_perm_indices::length): Likewise.
10416 (vec_perm_indices::nelts_per_input): Likewise.
10417 (vec_perm_indices::input_nelts): Likewise.
10418 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10419 number of elements per input as a poly_uint64 rather than an
10420 unsigned int. Use the original encoding for variable-length
10421 vectors, rather than clamping each individual element.
10422 For the second and subsequent elements in each pattern,
10423 clamp the step and base before clamping their sum.
10424 (vec_perm_indices::series_p): Handle polynomial element counts.
10425 (vec_perm_indices::all_in_range_p): Likewise.
10426 (vec_perm_indices_to_tree): Likewise.
10427 (vec_perm_indices_to_rtx): Likewise.
10428 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10429 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10430 (tree_vector_builder::new_binary_operation): Handle polynomial
10431 element counts. Return false if we need to know the number
10432 of elements at compile time.
10433 * fold-const.c (fold_vec_perm): Punt if the number of elements
10434 isn't known at compile time.
10436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10438 * vec-perm-indices.h (vec_perm_builder): Change element type
10439 from HOST_WIDE_INT to poly_int64.
10440 (vec_perm_indices::element_type): Update accordingly.
10441 (vec_perm_indices::clamp): Handle polynomial element_types.
10442 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10443 (vec_perm_indices::all_in_range_p): Likewise.
10444 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10446 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10447 polynomial vec_perm_indices element types.
10448 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10449 * fold-const.c (fold_vec_perm): Likewise.
10450 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10451 * tree-vect-generic.c (lower_vec_perm): Likewise.
10452 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10453 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10454 element type to HOST_WIDE_INT.
10456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10457 Alan Hayward <alan.hayward@arm.com>
10458 David Sherwood <david.sherwood@arm.com>
10460 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10461 rather than an int. Use plus_constant.
10462 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10463 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10466 Alan Hayward <alan.hayward@arm.com>
10467 David Sherwood <david.sherwood@arm.com>
10469 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10470 a HOST_WIDE_INT to a poly_int64.
10472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10473 Alan Hayward <alan.hayward@arm.com>
10474 David Sherwood <david.sherwood@arm.com>
10476 * calls.c (load_register_parameters): Cope with polynomial
10477 mode sizes. Require a constant size for BLKmode parameters
10478 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10479 forces a parameter to be padded at the lsb end in order to
10480 fill a complete number of words, require the parameter size
10481 to be ordered wrt UNITS_PER_WORD.
10483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10484 Alan Hayward <alan.hayward@arm.com>
10485 David Sherwood <david.sherwood@arm.com>
10487 * reload1.c (spill_stack_slot_width): Change element type
10488 from unsigned int to poly_uint64_pod.
10489 (alter_reg): Treat mode sizes as polynomial.
10491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10492 Alan Hayward <alan.hayward@arm.com>
10493 David Sherwood <david.sherwood@arm.com>
10495 * reload.c (complex_word_subreg_p): New function.
10496 (reload_inner_reg_of_subreg, push_reload): Use it.
10498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10499 Alan Hayward <alan.hayward@arm.com>
10500 David Sherwood <david.sherwood@arm.com>
10502 * lra-constraints.c (process_alt_operands): Reject matched
10503 operands whose sizes aren't ordered.
10504 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10511 that the mode size is in the set {1, 2, 4, 8, 16}.
10513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10514 Alan Hayward <alan.hayward@arm.com>
10515 David Sherwood <david.sherwood@arm.com>
10517 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10518 Use plus_constant instead of gen_rtx_PLUS.
10520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10521 Alan Hayward <alan.hayward@arm.com>
10522 David Sherwood <david.sherwood@arm.com>
10524 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10525 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10526 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10527 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10528 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10529 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10530 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10531 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10532 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10533 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10535 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10536 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10537 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10538 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10539 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10540 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10541 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10542 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10543 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10544 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10546 * expr.c (emit_move_resolve_push): Treat the input and result
10547 of PUSH_ROUNDING as a poly_int64.
10548 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10549 (emit_push_insn): Likewise.
10550 * lra-eliminations.c (mark_not_eliminable): Likewise.
10551 * recog.c (push_operand): Likewise.
10552 * reload1.c (elimination_effects): Likewise.
10553 * rtlanal.c (nonzero_bits1): Likewise.
10554 * calls.c (store_one_arg): Likewise. Require the padding to be
10555 known at compile time.
10557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10558 Alan Hayward <alan.hayward@arm.com>
10559 David Sherwood <david.sherwood@arm.com>
10561 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10562 Use plus_constant instead of gen_rtx_PLUS.
10564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10565 Alan Hayward <alan.hayward@arm.com>
10566 David Sherwood <david.sherwood@arm.com>
10568 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10569 rather than an int.
10571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10572 Alan Hayward <alan.hayward@arm.com>
10573 David Sherwood <david.sherwood@arm.com>
10575 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10576 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10577 via stack temporaries. Treat the mode size as polynomial too.
10579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10580 Alan Hayward <alan.hayward@arm.com>
10581 David Sherwood <david.sherwood@arm.com>
10583 * expr.c (expand_expr_real_2): When handling conversions involving
10584 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10585 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10586 as a poly_uint64 too.
10588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10589 Alan Hayward <alan.hayward@arm.com>
10590 David Sherwood <david.sherwood@arm.com>
10592 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10595 Alan Hayward <alan.hayward@arm.com>
10596 David Sherwood <david.sherwood@arm.com>
10598 * combine.c (can_change_dest_mode): Handle polynomial
10599 REGMODE_NATURAL_SIZE.
10600 * expmed.c (store_bit_field_1): Likewise.
10601 * expr.c (store_constructor): Likewise.
10602 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10603 and polynomial REGMODE_NATURAL_SIZE.
10604 (gen_lowpart_common): Likewise.
10605 * reginfo.c (record_subregs_of_mode): Likewise.
10606 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10613 numbers of elements.
10615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10616 Alan Hayward <alan.hayward@arm.com>
10617 David Sherwood <david.sherwood@arm.com>
10619 * match.pd: Cope with polynomial numbers of vector elements.
10621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10622 Alan Hayward <alan.hayward@arm.com>
10623 David Sherwood <david.sherwood@arm.com>
10625 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10626 in a POINTER_PLUS_EXPR.
10628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10629 Alan Hayward <alan.hayward@arm.com>
10630 David Sherwood <david.sherwood@arm.com>
10632 * omp-simd-clone.c (simd_clone_subparts): New function.
10633 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10634 (ipa_simd_modify_function_body): Likewise.
10636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10637 Alan Hayward <alan.hayward@arm.com>
10638 David Sherwood <david.sherwood@arm.com>
10640 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10641 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10642 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10643 (expand_vector_condition, vector_element): Likewise.
10644 (subparts_gt): New function.
10645 (get_compute_type): Use subparts_gt.
10646 (count_type_subparts): Delete.
10647 (expand_vector_operations_1): Use subparts_gt instead of
10648 count_type_subparts.
10650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10651 Alan Hayward <alan.hayward@arm.com>
10652 David Sherwood <david.sherwood@arm.com>
10654 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10655 (vect_compile_time_alias): ...this new function. Do the calculation
10656 on poly_ints rather than trees.
10657 (vect_prune_runtime_alias_test_list): Update call accordingly.
10659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10660 Alan Hayward <alan.hayward@arm.com>
10661 David Sherwood <david.sherwood@arm.com>
10663 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10665 (vect_schedule_slp_instance): 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-vect-slp.c (vect_get_and_check_slp_defs): Reject
10672 constant and extern definitions for variable-length vectors.
10673 (vect_get_constant_vectors): Note that the number of units
10674 is known to be constant.
10676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10677 Alan Hayward <alan.hayward@arm.com>
10678 David Sherwood <david.sherwood@arm.com>
10680 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10681 of units as polynomial. Choose between WIDE and NARROW based
10684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10685 Alan Hayward <alan.hayward@arm.com>
10686 David Sherwood <david.sherwood@arm.com>
10688 * tree-vect-stmts.c (simd_clone_subparts): New function.
10689 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10692 Alan Hayward <alan.hayward@arm.com>
10693 David Sherwood <david.sherwood@arm.com>
10695 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10696 vectors as polynomial. Use build_index_vector for
10697 IFN_GOMP_SIMD_LANE.
10699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10700 Alan Hayward <alan.hayward@arm.com>
10701 David Sherwood <david.sherwood@arm.com>
10703 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10704 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10705 for variable-length vectors.
10706 (vectorizable_mask_load_store): Treat the number of units as
10707 polynomial, asserting that it is constant if the condition has
10708 already been enforced.
10709 (vectorizable_store, vectorizable_load): Likewise.
10711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10712 Alan Hayward <alan.hayward@arm.com>
10713 David Sherwood <david.sherwood@arm.com>
10715 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10716 of units as polynomial. Punt if we can't tell at compile time
10717 which vector contains the final result.
10719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10720 Alan Hayward <alan.hayward@arm.com>
10721 David Sherwood <david.sherwood@arm.com>
10723 * tree-vect-loop.c (vectorizable_induction): Treat the number
10724 of units as polynomial. Punt on SLP inductions. Use an integer
10725 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10726 cast of such a series for variable-length floating-point
10729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10730 Alan Hayward <alan.hayward@arm.com>
10731 David Sherwood <david.sherwood@arm.com>
10733 * tree.h (build_index_vector): Declare.
10734 * tree.c (build_index_vector): New function.
10735 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10736 of units as polynomial, forcibly converting it to a constant if
10737 vectorizable_reduction has already enforced the condition.
10738 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10739 to create a {1,2,3,...} vector.
10740 (vectorizable_reduction): Treat the number of units as polynomial.
10741 Choose vectype_in based on the largest scalar element size rather
10742 than the smallest number of units. Enforce the restrictions
10745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10746 Alan Hayward <alan.hayward@arm.com>
10747 David Sherwood <david.sherwood@arm.com>
10749 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10750 number of units as polynomial.
10752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10753 Alan Hayward <alan.hayward@arm.com>
10754 David Sherwood <david.sherwood@arm.com>
10756 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10757 * target.def (autovectorize_vector_sizes): Return the vector sizes
10758 by pointer, using vector_sizes rather than a bitmask.
10759 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10760 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10761 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10763 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10764 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10765 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10766 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10767 * omp-general.c (omp_max_vf): Likewise.
10768 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10769 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10770 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10771 * tree-vect-slp.c (vect_slp_bb): Likewise.
10772 * doc/tm.texi: Regenerate.
10773 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10775 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10776 the vector size as a poly_uint64 rather than an unsigned int.
10777 (current_vector_size): Change from an unsigned int to a poly_uint64.
10778 (get_vectype_for_scalar_type): Update accordingly.
10779 * tree.h (build_truth_vector_type): Take the size and number of
10780 units as a poly_uint64 rather than an unsigned int.
10781 (build_vector_type): Add a temporary overload that takes
10782 the number of units as a poly_uint64 rather than an unsigned int.
10783 * tree.c (make_vector_type): Likewise.
10784 (build_truth_vector_type): Take the number of units as a poly_uint64
10785 rather than an unsigned int.
10787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10788 Alan Hayward <alan.hayward@arm.com>
10789 David Sherwood <david.sherwood@arm.com>
10791 * target.def (get_mask_mode): Take the number of units and length
10792 as poly_uint64s rather than unsigned ints.
10793 * targhooks.h (default_get_mask_mode): Update accordingly.
10794 * targhooks.c (default_get_mask_mode): Likewise.
10795 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10796 * doc/tm.texi: Regenerate.
10798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10799 Alan Hayward <alan.hayward@arm.com>
10800 David Sherwood <david.sherwood@arm.com>
10802 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10803 * omp-general.c (omp_max_vf): Likewise.
10804 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10805 (expand_omp_simd): Handle polynomial safelen.
10806 * omp-low.c (omplow_simd_context): Add a default constructor.
10807 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10808 (lower_rec_simd_input_clauses): Update accordingly.
10809 (lower_rec_input_clauses): Likewise.
10811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10812 Alan Hayward <alan.hayward@arm.com>
10813 David Sherwood <david.sherwood@arm.com>
10815 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10816 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10817 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10818 (vect_analyze_slp_cost): Likewise.
10819 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10820 (vect_model_load_cost): Likewise.
10822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10823 Alan Hayward <alan.hayward@arm.com>
10824 David Sherwood <david.sherwood@arm.com>
10826 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10827 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10828 from an unsigned int * to a poly_uint64_pod *.
10829 (calculate_unrolling_factor): New function.
10830 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10833 Alan Hayward <alan.hayward@arm.com>
10834 David Sherwood <david.sherwood@arm.com>
10836 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10837 from an unsigned int to a poly_uint64.
10838 (_loop_vec_info::slp_unrolling_factor): Likewise.
10839 (_loop_vec_info::vectorization_factor): Change from an int
10841 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10842 (vect_get_num_vectors): New function.
10843 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10844 (vect_get_num_copies): Use vect_get_num_vectors.
10845 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10846 to an unsigned int *.
10847 (vect_analyze_data_refs): Change min_vf from an int * to a
10849 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10850 than an unsigned HOST_WIDE_INT.
10851 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10852 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10853 to an unsigned int *.
10854 (vect_analyze_data_ref_dependences): Likewise.
10855 (vect_compute_data_ref_alignment): Handle polynomial vf.
10856 (vect_enhance_data_refs_alignment): Likewise.
10857 (vect_prune_runtime_alias_test_list): Likewise.
10858 (vect_shift_permute_load_chain): Likewise.
10859 (vect_supportable_dr_alignment): Likewise.
10860 (dependence_distance_ge_vf): Take the vectorization factor as a
10861 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10862 (vect_analyze_data_refs): Change min_vf from an int * to a
10864 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10865 vfm1 as a poly_uint64 rather than an int. Make the same change
10866 for the returned bound_scalar.
10867 (vect_gen_vector_loop_niters): Handle polynomial vf.
10868 (vect_do_peeling): Likewise. Update call to
10869 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10870 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10872 * tree-vect-loop.c (vect_determine_vectorization_factor)
10873 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10874 (vect_get_known_peeling_cost): Likewise.
10875 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10876 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10877 (vect_transform_loop): Likewise. Use the lowest possible VF when
10878 updating the upper bounds of the loop.
10879 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10880 rather than an int.
10881 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10882 polynomial unroll factors.
10883 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10884 (vect_make_slp_decision): Likewise.
10885 (vect_supported_load_permutation_p): Likewise, and polynomial
10887 (vect_analyze_slp_cost): Handle polynomial vf.
10888 (vect_slp_analyze_node_operations): Likewise.
10889 (vect_slp_analyze_bb_1): Likewise.
10890 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10891 than an unsigned HOST_WIDE_INT.
10892 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10893 (vectorizable_load): Handle polynomial vf.
10894 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10896 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10899 Alan Hayward <alan.hayward@arm.com>
10900 David Sherwood <david.sherwood@arm.com>
10902 * match.pd: Handle bit operations involving three constants
10903 and try to fold one pair.
10905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10907 * tree-vect-loop-manip.c: Include gimple-fold.h.
10908 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10909 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10910 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10911 Add a path that uses a step of VF instead of 1, but disable it
10913 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10914 and niters_no_overflow parameters. Update calls to
10915 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10916 Create a new SSA name if the latter choses to use a ste other
10917 than zero, and return it via niters_vector_mult_vf_var.
10918 * tree-vect-loop.c (vect_transform_loop): Update calls to
10919 vect_do_peeling, vect_gen_vector_loop_niters and
10920 slpeel_make_loop_iterate_ntimes.
10921 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10922 (vect_gen_vector_loop_niters): Update declarations after above changes.
10924 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10926 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10927 128-bit round to integer instructions.
10928 (ceil<mode>2): Likewise.
10929 (btrunc<mode>2): Likewise.
10930 (round<mode>2): Likewise.
10932 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10934 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10935 unaligned VSX load/store on P8/P9.
10936 (expand_block_clear): Allow the use of unaligned VSX
10937 load/store on P8/P9.
10939 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10941 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10943 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10944 swap associated with both a load and a store.
10946 2018-01-02 Andrew Waterman <andrew@sifive.com>
10948 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10949 * config/riscv/riscv.md (clear_cache): Use it.
10951 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10953 * web.c: Remove out-of-date comment.
10955 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10957 * expr.c (fixup_args_size_notes): Check that any existing
10958 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10959 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10960 (emit_single_push_insn): ...here.
10962 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10964 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10965 (const_vector_encoded_nelts): New function.
10966 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10967 (const_vector_int_elt, const_vector_elt): Declare.
10968 * emit-rtl.c (const_vector_int_elt_1): New function.
10969 (const_vector_elt): Likewise.
10970 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10971 of CONST_VECTOR_ELT.
10973 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10975 * expr.c: Include rtx-vector-builder.h.
10976 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10977 directly on the tree encoding.
10978 (const_vector_from_tree): Likewise.
10979 * optabs.c: Include rtx-vector-builder.h.
10980 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10981 sequence of "u" values.
10982 * vec-perm-indices.c: Include rtx-vector-builder.h.
10983 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10984 directly on the vec_perm_indices encoding.
10986 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10988 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10989 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10990 * rtx-vector-builder.h: New file.
10991 * rtx-vector-builder.c: Likewise.
10992 * rtl.h (rtx_def::u2): Add a const_vector field.
10993 (CONST_VECTOR_NPATTERNS): New macro.
10994 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10995 (CONST_VECTOR_DUPLICATE_P): Likewise.
10996 (CONST_VECTOR_STEPPED_P): Likewise.
10997 (CONST_VECTOR_ENCODED_ELT): Likewise.
10998 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10999 (unwrap_const_vec_duplicate): Likewise.
11000 (const_vec_series_p): Check for a non-duplicated vector encoding.
11001 Say that the function only returns true for integer vectors.
11002 * emit-rtl.c: Include rtx-vector-builder.h.
11003 (gen_const_vec_duplicate_1): Delete.
11004 (gen_const_vector): Call gen_const_vec_duplicate instead of
11005 gen_const_vec_duplicate_1.
11006 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11007 (gen_const_vec_duplicate): Use rtx_vector_builder.
11008 (gen_const_vec_series): Likewise.
11009 (gen_rtx_CONST_VECTOR): Likewise.
11010 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11011 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11012 Build a new vector rather than modifying a CONST_VECTOR in-place.
11013 (handle_special_swappables): Update call accordingly.
11014 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11015 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11016 Build a new vector rather than modifying a CONST_VECTOR in-place.
11017 (handle_special_swappables): Update call accordingly.
11019 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11021 * simplify-rtx.c (simplify_const_binary_operation): Use
11022 CONST_VECTOR_ELT instead of XVECEXP.
11024 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11026 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11027 the selector elements to be different from the data elements
11028 if the selector is a VECTOR_CST.
11029 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11030 ssizetype for the selector.
11032 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11034 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11035 before testing each element individually.
11036 * tree-vect-generic.c (lower_vec_perm): Likewise.
11038 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11040 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11041 * selftest-run-tests.c (selftest::run_tests): Call it.
11042 * vector-builder.h (vector_builder::operator ==): New function.
11043 (vector_builder::operator !=): Likewise.
11044 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11045 (vec_perm_indices::all_from_input_p): New function.
11046 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11047 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11048 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11049 instead of reading the VECTOR_CST directly. Detect whether both
11050 vector inputs are the same before constructing the vec_perm_indices,
11051 and update the number of inputs argument accordingly. Use the
11052 utility functions added above. Only construct sel2 if we need to.
11054 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11056 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11057 the broadcast of the low byte.
11058 (expand_mult_highpart): Use an explicit encoding for the permutes.
11059 * optabs-query.c (can_mult_highpart_p): Likewise.
11060 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11061 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11062 (vectorizable_bswap): Likewise.
11063 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11064 explicit encoding for the power-of-2 permutes.
11065 (vect_permute_store_chain): Likewise.
11066 (vect_grouped_load_supported): Likewise.
11067 (vect_permute_load_chain): Likewise.
11069 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11071 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11072 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11073 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11074 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11075 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11076 (vect_gen_perm_mask_any): Likewise.
11078 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11080 * int-vector-builder.h: New file.
11081 * vec-perm-indices.h: Include int-vector-builder.h.
11082 (vec_perm_indices): Redefine as an int_vector_builder.
11083 (auto_vec_perm_indices): Delete.
11084 (vec_perm_builder): Redefine as a stand-alone class.
11085 (vec_perm_indices::vec_perm_indices): New function.
11086 (vec_perm_indices::clamp): Likewise.
11087 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11088 (vec_perm_indices::new_vector): New function.
11089 (vec_perm_indices::new_expanded_vector): Update for new
11090 vec_perm_indices class.
11091 (vec_perm_indices::rotate_inputs): New function.
11092 (vec_perm_indices::all_in_range_p): Operate directly on the
11093 encoded form, without computing elided elements.
11094 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11095 encoding. Update for new vec_perm_indices class.
11096 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11097 the given vec_perm_builder.
11098 (expand_vec_perm_var): Update vec_perm_builder constructor.
11099 (expand_mult_highpart): Use vec_perm_builder instead of
11100 auto_vec_perm_indices.
11101 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11102 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11103 or double series encoding as appropriate.
11104 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11105 vec_perm_indices instead of auto_vec_perm_indices.
11106 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11107 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11108 (vect_permute_store_chain): Likewise.
11109 (vect_grouped_load_supported): Likewise.
11110 (vect_permute_load_chain): Likewise.
11111 (vect_shift_permute_load_chain): Likewise.
11112 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11113 (vect_transform_slp_perm_load): Likewise.
11114 (vect_schedule_slp_instance): Likewise.
11115 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11116 (vectorizable_mask_load_store): Likewise.
11117 (vectorizable_bswap): Likewise.
11118 (vectorizable_store): Likewise.
11119 (vectorizable_load): Likewise.
11120 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11121 vec_perm_indices instead of auto_vec_perm_indices. Use
11122 tree_to_vec_perm_builder to read the vector from a tree.
11123 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11124 vec_perm_builder instead of a vec_perm_indices.
11125 (have_whole_vector_shift): Use vec_perm_builder and
11126 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11127 truncation to calc_vec_perm_mask_for_shift.
11128 (vect_create_epilog_for_reduction): Likewise.
11129 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11130 from auto_vec_perm_indices to vec_perm_indices.
11131 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11132 instead of changing individual elements.
11133 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11134 the vector in d.perm.
11135 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11136 from auto_vec_perm_indices to vec_perm_indices.
11137 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11138 instead of changing individual elements.
11139 (arm_vectorize_vec_perm_const): Use new_vector to install
11140 the vector in d.perm.
11141 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11142 Update vec_perm_builder constructor.
11143 (rs6000_expand_interleave): Likewise.
11144 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11145 (rs6000_expand_interleave): Likewise.
11147 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11149 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11150 to qimode could truncate the indices.
11151 * optabs.c (expand_vec_perm_var): Likewise.
11153 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11155 * Makefile.in (OBJS): Add vec-perm-indices.o.
11156 * vec-perm-indices.h: New file.
11157 * vec-perm-indices.c: Likewise.
11158 * target.h (vec_perm_indices): Replace with a forward class
11160 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11161 * optabs.h: Include vec-perm-indices.h.
11162 (expand_vec_perm): Delete.
11163 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11164 (expand_vec_perm_const): Declare.
11165 * target.def (vec_perm_const_ok): Replace with...
11166 (vec_perm_const): ...this new hook.
11167 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11168 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11169 * doc/tm.texi: Regenerate.
11170 * optabs.def (vec_perm_const): Delete.
11171 * doc/md.texi (vec_perm_const): Likewise.
11172 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11173 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11174 expand_vec_perm for constant permutation vectors. Assert that
11175 the mode of variable permutation vectors is the integer equivalent
11176 of the mode that is being permuted.
11177 * optabs-query.h (selector_fits_mode_p): Declare.
11178 * optabs-query.c: Include vec-perm-indices.h.
11179 (selector_fits_mode_p): New function.
11180 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11181 is defined, instead of checking whether the vec_perm_const_optab
11182 exists. Use targetm.vectorize.vec_perm_const instead of
11183 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11184 fit in the vector mode before using a variable permute.
11185 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11186 vec_perm_indices instead of an rtx.
11187 (expand_vec_perm): Replace with...
11188 (expand_vec_perm_const): ...this new function. Take the selector
11189 as a vec_perm_indices rather than an rtx. Also take the mode of
11190 the selector. Update call to shift_amt_for_vec_perm_mask.
11191 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11192 Use vec_perm_indices::new_expanded_vector to expand the original
11193 selector into bytes. Check whether the indices fit in the vector
11194 mode before using a variable permute.
11195 (expand_vec_perm_var): Make global.
11196 (expand_mult_highpart): Use expand_vec_perm_const.
11197 * fold-const.c: Includes vec-perm-indices.h.
11198 * tree-ssa-forwprop.c: Likewise.
11199 * tree-vect-data-refs.c: Likewise.
11200 * tree-vect-generic.c: Likewise.
11201 * tree-vect-loop.c: Likewise.
11202 * tree-vect-slp.c: Likewise.
11203 * tree-vect-stmts.c: Likewise.
11204 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11206 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11207 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11208 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11209 (aarch64_vectorize_vec_perm_const): ...this new function.
11210 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11211 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11212 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11213 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11214 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11215 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11216 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11218 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11219 check for NEON modes.
11220 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11221 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11222 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11223 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11225 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11226 the old VEC_PERM_CONST conditions.
11227 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11228 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11229 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11230 (ia64_vectorize_vec_perm_const_ok): Merge into...
11231 (ia64_vectorize_vec_perm_const): ...this new function.
11232 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11233 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11234 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11235 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11236 * config/mips/mips.c (mips_expand_vec_perm_const)
11237 (mips_vectorize_vec_perm_const_ok): Merge into...
11238 (mips_vectorize_vec_perm_const): ...this new function.
11239 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11240 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11241 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11242 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11243 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11244 (rs6000_expand_vec_perm_const): Delete.
11245 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11247 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11248 (altivec_expand_vec_perm_const_le): Take each operand individually.
11249 Operate on constant selectors rather than rtxes.
11250 (altivec_expand_vec_perm_const): Likewise. Update call to
11251 altivec_expand_vec_perm_const_le.
11252 (rs6000_expand_vec_perm_const): Delete.
11253 (rs6000_vectorize_vec_perm_const_ok): Delete.
11254 (rs6000_vectorize_vec_perm_const): New function.
11255 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11256 an element count and rtx array.
11257 (rs6000_expand_extract_even): Update call accordingly.
11258 (rs6000_expand_interleave): Likewise.
11259 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11260 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11261 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11262 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11263 (rs6000_expand_vec_perm_const): Delete.
11264 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11265 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11266 (altivec_expand_vec_perm_const_le): Take each operand individually.
11267 Operate on constant selectors rather than rtxes.
11268 (altivec_expand_vec_perm_const): Likewise. Update call to
11269 altivec_expand_vec_perm_const_le.
11270 (rs6000_expand_vec_perm_const): Delete.
11271 (rs6000_vectorize_vec_perm_const_ok): Delete.
11272 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11273 reference to the SPE evmerge intructions.
11274 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11275 an element count and rtx array.
11276 (rs6000_expand_extract_even): Update call accordingly.
11277 (rs6000_expand_interleave): Likewise.
11278 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11279 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11281 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11283 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11285 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11286 vector mode and that that mode matches the mode of the data
11288 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11289 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11290 directly using expand_vec_perm_1 when forcing selectors into
11292 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11294 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11296 * optabs-query.h (can_vec_perm_p): Delete.
11297 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11298 * optabs-query.c (can_vec_perm_p): Split into...
11299 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11300 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11301 particular selector is valid.
11302 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11303 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11304 (vect_grouped_load_supported): Likewise.
11305 (vect_shift_permute_load_chain): Likewise.
11306 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11307 (vect_transform_slp_perm_load): Likewise.
11308 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11309 (vectorizable_bswap): Likewise.
11310 (vect_gen_perm_mask_checked): Likewise.
11311 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11312 implementations of variable permutation vectors into account
11313 when deciding which selector to use.
11314 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11315 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11316 with a false third argument.
11317 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11318 to test whether the constant selector is valid and can_vec_perm_var_p
11319 to test whether a variable selector is valid.
11321 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11323 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11324 * optabs-query.c (can_vec_perm_p): Likewise.
11325 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11326 instead of vec_perm_indices.
11327 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11328 (vect_gen_perm_mask_checked): Likewise,
11329 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11330 (vect_gen_perm_mask_checked): Likewise,
11332 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11334 * optabs-query.h (qimode_for_vec_perm): Declare.
11335 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11336 (qimode_for_vec_perm): ...this new function.
11337 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11339 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11341 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11342 does not have a conditional at the top.
11344 2018-01-02 Richard Biener <rguenther@suse.de>
11346 * ipa-inline.c (big_speedup_p): Fix expression.
11348 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11351 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11354 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11358 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11359 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11360 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11361 cond_taken_branch_cost 3->4.
11363 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11365 PR tree-optimization/83581
11366 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11367 TODO_cleanup_cfg if any changes have been made.
11369 PR middle-end/83608
11370 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11371 convert_modes if target mode has the right side, but different mode
11374 PR middle-end/83609
11375 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11376 last argument when extracting from CONCAT. If either from_real or
11377 from_imag is NULL, use expansion through memory. If result is not
11378 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11379 the parts directly to inner mode, if even that fails, use expansion
11382 PR middle-end/83623
11383 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11384 check for bswap in mode rather than HImode and use that in expand_unop
11387 Copyright (C) 2018 Free Software Foundation, Inc.
11389 Copying and distribution of this file, with or without modification,
11390 are permitted in any medium without royalty provided the copyright
11391 notice and this notice are preserved.