1 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3 * config/nds32/constants.md (unspec_volatile_element): Add values for
4 TLB operation and data prefetch.
5 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6 functions for TLB operation and data prefetch.
7 * config/nds32/nds32-intrinsic.md: Likewise.
8 * config/nds32/nds32_intrinsic.h: Likewise.
9 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
10 (nds32_print_operand): Likewise.
11 * config/nds32/nds32.h (nds32_builtins): Likewise.
13 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
14 Andrew Pinski <pinsika@gcc.gnu.org>
17 * match.pd: Use constant_boolean_node of correct type instead of
18 boolean_true_node or boolean_false_node for simplifying
19 pointer comparisons to zero.
21 2018-04-07 Jakub Jelinek <jakub@redhat.com>
23 PR tree-optimization/80021
24 * tree.c (verify_type_variant): Make error call in verify_variant_match
25 translatable and remove final full stop.
27 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
29 * config/nds32/constants.md (unspec_volatile_element): Add
30 UNSPEC_VOLATILE_EH_RETURN.
31 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
32 nds32_output_stack_pop): Support dwarf exception handling process.
33 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
34 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
35 exception handling process.
36 (nds32_compute_stack_frame): Likewise.
37 (nds32_return_addr_rtx): Likewise.
38 (nds32_initial_elimination_offset): Likewise.
39 (nds32_expand_prologue): Likewise.
40 (nds32_expand_epilogue): Likewise.
41 (nds32_dynamic_chain_address): New function.
42 * config/nds32/nds32.h (machine_function): Add fields for dwarf
44 (DYNAMIC_CHAIN_ADDRESS): Define.
45 (EH_RETURN_DATA_REGNO): Define.
46 (EH_RETURN_STACKADJ_RTX): Define.
47 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
48 patterns for dwarf exception handling.
50 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
52 * config/nds32/nds32.h: Clean up obsolete macros.
54 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
56 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
57 Add enum values for particular instructions.
58 * config/nds32/nds32-intrinsic.c: Implementation of expanding
59 particular intrinsic functions.
60 * config/nds32/nds32-intrinsic.md: Likewise.
61 * config/nds32/nds32_intrinsic.h: Likewise.
62 * config/nds32/nds32.h (nds32_builtins): Likewise.
63 * config/nds32/nds32.md (type): Add pbsad and pbsada.
64 (btst, ave): New patterns for particular instructions.
66 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
68 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
69 Add enum values for atomic load/store and memory sync.
70 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
72 * config/nds32/nds32-intrinsic.md: Likewise.
73 * config/nds32/nds32_intrinsic.h: Likewise.
74 * config/nds32/nds32.h (nds32_builtins): Likewise.
76 2018-04-07 Jakub Jelinek <jakub@redhat.com>
78 PR tree-optimization/85257
79 * fold-const.c (native_encode_vector): If not all elts could fit
80 and off is -1, return 0 rather than offset.
81 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
82 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
83 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
84 adjust buffer in native_interpret_expr call.
86 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
88 * config/nds32/constants.md (unspec_volatile_element): Add cache
90 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
91 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
92 * config/nds32/nds32.c (nds32_cctl_names): New.
93 (nds32_print_operand): Handle cache control register names.
94 * config/nds32/nds32.h (nds32_builtins): New enum values.
95 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
97 * config/nds32/nds32.md (type): Add mmu.
98 * config/nds32/pipelines.md (simple_insn): Add mmu.
100 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
102 * config/nds32/nds32.md (type): Remove call.
103 * config/nds32/pipelines.md (simple_insn): Likewise.
105 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
107 * config/nds32/constants.md (unspec_volatile_element): Add
108 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
109 UNSPEC_VOLATILE_FMFCFG.
110 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
111 description for fmfcfg and fmfcsr.
112 (bdesc_1arg): Add fmtcsr.
113 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
114 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
115 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
116 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
117 unspec_fmfcfg): New patterns.
118 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
119 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
120 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
121 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
122 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
123 __nds32__fmfcfg): Define.
125 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
127 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
128 intrinsic register names.
129 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
130 intrinsic register enum values and macros.
132 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
134 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
135 for load/store addressing form.
136 (nds32_print_operand_address): Likewise.
138 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
141 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
142 based on LABEL_REF. Remove useless assertion.
143 (pic_address_needs_scratch): Fix formatting.
144 (sparc_legitimize_pic_address): Minor tweaks.
145 (sparc_delegitimize_address): Adjust assertion accordingly.
146 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
147 into symbolic_operand.
148 (movsi_high_pic_label_ref): Likewise.
149 (movsi_lo_sum_pic_label_ref): Likewise.
150 (movdi_pic_label_ref): Likewise.
151 (movdi_high_pic_label_ref): Likewise.
152 (movdi_lo_sum_pic_label_ref): Likewise.
154 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
156 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
157 custom LIB_SPEC setup.
159 2018-04-06 Ruslan Bukin <br@bsdpad.com>
160 Kito Cheng <kito.cheng@gmail.com>
162 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
163 * config/riscv/freebsd.h: New.
165 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
167 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
168 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
171 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
172 Kito Cheng <kito.cheng@gmail.com>
174 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
175 nds32_output_call, nds32_symbol_binds_local_p): New functions.
176 * config/nds32/nds32-protos.h (nds32_output_call,
177 nds32_output_return): Declare.
178 * config/nds32/nds32.md: Refine all the call and return patterns.
180 2018-04-06 Jakub Jelinek <jakub@redhat.com>
183 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
184 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
186 PR rtl-optimization/84872
187 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
188 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
191 2018-04-06 Tamar Christina <tamar.christina@arm.com>
193 * expr.c (copy_blkmode_to_reg): Revert 254862.
194 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
196 2018-04-06 Richard Biener <rguenther@suse.de>
199 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
200 after seeing a component reference with an adjacent field. Treat
201 refs to arrays at struct end of external decls similar to
202 refs to unconstrained commons.
204 2018-04-06 Jakub Jelinek <jakub@redhat.com>
207 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
208 look through SAVE_EXPRs with non-side-effects argument. Adjust
210 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
213 2018-04-06 Richard Biener <rguenther@suse.de>
216 * alias.c (find_base_term): New wrapper around find_base_term
217 unwinding CSELIB_VAL_PTR changes.
218 (find_base_term): Do not restore CSELIB_VAL_PTR during the
221 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
223 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
225 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
226 constant definitions.
227 ("nop"): lr 0,0 -> nopr r0
228 ("nop_lr0", "nop_lr1"): New insn definitions.
230 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
232 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
233 NDS32_V3PUSH_AVAILABLE_P macro.
235 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
236 Chung-Ju Wu <jasonwucj@gmail.com>
238 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
239 (nds32*-*-*): Add float and fpu_config into supported_defaults.
240 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
241 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
242 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
243 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
244 * config/nds32/constraints.md: New constraints and checking for hard
246 * config/nds32/iterators.md: New mode iterator and attribute for hard
248 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
250 * config/nds32/nds32-fpu.md: New file.
251 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
252 deal with hard float code generation.
253 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
255 (abi_type, float_reg_number): New enum type.
256 * config/nds32/nds32-predicates.c: New predicates for hard float.
257 * config/nds32/nds32-protos.h: Declare functions for hard float.
258 * config/nds32/nds32.c: Implementation for hard float configuration.
259 * config/nds32/nds32.h: Definitions for hard float configuration.
260 * config/nds32/nds32.md: Include hard float machine description and
261 modify patterns for hard float configuration.
262 * config/nds32/nds32.opt: New options for hard float configuration.
263 * config/nds32/predicates.md: New predicates for hard float
266 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
268 * common/config/nds32/nds32-common.c
269 (nds32_option_optimization_table): Enable -mreleax-hint by default.
271 2018-04-05 Jakub Jelinek <jakub@redhat.com>
274 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
275 CONSTRUCTOR_ELT (ctor, ...)->value.
277 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
280 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
282 2018-04-05 Tom de Vries <tom@codesourcery.com>
285 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
288 2018-04-05 Shiva Chen <shiva0217@gmail.com>
289 Kito Cheng <kito.cheng@gmail.com>
291 * config/nds32/constraints.md (U33): Fine-tune checking condition.
292 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
293 * config/nds32/nds32.h (nds32_16bit_address_type): Add
294 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
296 2018-04-05 Shiva Chen <shiva0217@gmail.com>
297 Kito Cheng <kito.cheng@gmail.com>
299 * config/nds32/constraints.md (Ufe): New memory constraint.
300 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
301 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
302 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
304 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
305 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
307 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
309 * config/nds32/nds32.md: Use optimize_size in the condition for
310 alu-shift instructions.
312 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
314 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
316 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
318 * config/nds32/nds32.md (negsi2): Refine pattern.
320 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
321 Chung-Ju Wu <jasonwucj@gmail.com>
323 * config/nds32/iterators.md (shift_rotate): New code iterator.
324 (shift): New code attribute.
325 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
326 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
327 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
328 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
330 (andsi3, *andsi3): Ditto.
331 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
332 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
333 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
334 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
335 nds32_ior_operand, nds32_xor_operand): New predicates.
337 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
339 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
340 (addsi3, subsi3): ... this.
342 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
344 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
346 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
348 * config/nds32/nds32.md: Adjust indention.
350 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
352 * config/nds32/nds32.md (feature): New attribute.
354 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
356 * config/nds32/nds32.md (subtype): New attribute.
358 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
361 * config/arm/arm-builtins.c (arm_expand_builtin): Change
362 expansion to perform a bitwise AND of the argument followed by a
363 boolean negation of the result.
365 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
367 PR rtl-optimization/84878
368 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
369 the basic block. Assert the use reference is not artificial and that
370 it has an associated insn.
372 2018-04-04 Michael Matz <matz@suse.de>
374 * builtins.c (compute_objsize): Pass correct operand
375 to array_at_struct_end_p.
377 2018-04-04 Richard Biener <rguenther@suse.de>
380 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
381 from contexts for DINFO_LEVEL_TERSE and below.
383 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
385 * config/nds32/nds32-doubleword.md (move_<mode>): Require
386 resiter_operand condition.
387 * config/nds32/nds32.md (*move<mode>): Ditto.
389 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
390 Monk Chiang <sh.chiang04@gmail.com>
392 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
394 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
396 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
398 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
399 Kito Cheng <kito.cheng@gmail.com>
401 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
402 nds32_cond_code_str, output_cond_branch,
403 output_cond_branch_compare_zero, nds32_expand_cbranch,
404 nds32_expand_cstore, nds32_expand_movcc,
405 nds32_output_cbranchsi4_equality_zero,
406 nds32_output_cbranchsi4_equality_reg,
407 nds32_output_cbranchsi4_equality_reg_or_const_int,
408 nds32_output_cbranchsi4_greater_less_zero: New functions.
409 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
410 nds32_expand_cstore, nds32_expand_movcc,
411 nds32_output_cbranchsi4_equality_zero,
412 nds32_output_cbranchsi4_equality_reg,
413 nds32_output_cbranchsi4_equality_reg_or_const_int,
414 nds32_output_cbranchsi4_greater_less_zero): Declare.
415 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
416 nds32_rimm11s_operand): New predicates.
417 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
418 * config/nds32/nds32.md: Rewrite all the branch and conditional move
421 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
423 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
424 * config/nds32/nds32.md: Ditto.
425 * config/nds32/pipelines.md: Ditto.
427 2018-04-04 Richard Biener <rguenther@suse.de>
429 PR tree-optimization/85168
430 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
431 propagating abnormals.
433 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
435 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
437 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
438 Kito Cheng <kito.cheng@gmail.com>
440 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
441 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
442 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
443 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
444 * config/nds32/nds32.md (sibcall_internal): New.
445 (sibcall_register): Remove.
446 (sibcall_immediate): Remove.
447 (sibcall_value_internal): New.
448 (sibcall_value_register): Remove.
449 (sibcall_value_immediate): Remove.
450 * config/nds32/predicates.md (nds32_general_register_operand): New.
451 (nds32_call_address_operand): New.
453 2018-04-03 Jakub Jelinek <jakub@redhat.com>
455 PR rtl-optimization/85167
456 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
457 bb_defs if *split_p, instead preinitialize it to NULL.
459 PR tree-optimization/85156
460 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
461 evaluating the argument multiple times.
463 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
465 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
467 (_mm_cvtpd_ps): Likewise.
468 (_mm_cvttpd_epi32): Likewise.
469 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
470 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
471 vector, pixel, and bool following altivec.h include.
473 2018-04-03 Martin Sebor <msebor@redhat.com>
475 * doc/extend.texi (Common Function Attributes): Clarify.
476 (const attribute): Likewise.
477 (pure attribute): Likewise.
479 2018-04-03 Jakub Jelinek <jakub@redhat.com>
482 * config/i386/i386.c (ix86_expand_vector_set): Use
483 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
485 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
487 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
488 instructions when changing rounding bits to preserve precision bits
489 in the x87 control word.
491 2018-04-03 Martin Liska <mliska@suse.cz>
493 PR tree-optimization/82491
494 * rtl.h (strip_offset_and_add): Replace += suboffset with
495 poly_uint64 () + suboffset.
497 2018-03-29 Martin Liska <mliska@suse.cz>
498 Martin Jambor <mjambor@suse.cz>
501 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
502 param_type is not an integral or pointer type.
504 2018-04-03 Richard Biener <rguenther@suse.de>
506 * sese.h (recompute_all_dominators): Remove.
508 2018-04-02 Martin Sebor <msebor@redhat.com>
510 * doc/invoke.texi (-Wrestrict): Fix typos.
512 2018-04-02 Jim Wilson <jimw@sifive.com>
514 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
515 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
516 (<optab>di3, <optab>si3_extend): Likewise.
517 (<optab>si3_mask, <optab>si3_mask_1): New.
518 (<optab>di3_mask, <optab>di3_mask_1): New.
519 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
520 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
521 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
523 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
525 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
528 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
530 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
531 (nds32_canonicalize_comparison): New function.
533 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
534 Kito Cheng <kito.cheng@gmail.com>
535 Kuan-Lin Chen <kuanlinchentw@gmail.com>
537 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
538 * config/nds32/constants.md (unspec_volatile_element): Add
539 UNSPEC_VOLATILE_RELAX_GROUP.
540 * config/nds32/nds32-relax-opt.c: New file.
541 * config/nds32/nds32-predicates.c
542 (nds32_symbol_load_store_p): New function.
543 * config/nds32/nds32-protos.h
544 (nds32_symbol_load_store_p): Declare function.
545 (make_pass_nds32_relax_opt): Declare new rtl pass function.
546 * config/nds32/nds32.c
547 (nds32_register_pass): New function to register pass.
548 (nds32_register_passes): New function to register passes.
549 * config/nds32/nds32.md (relax_group): New pattern.
550 * config/nds32/nds32.opt (mrelax-hint): New option.
551 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
553 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
555 * config/nds32/t-nds32: Modify files dependency.
557 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
559 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
560 (PROFILE_HOOK): Define its implementation.
562 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
564 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
565 type and 32-bit size.
567 2018-04-01 Jakub Jelinek <jakub@redhat.com>
570 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
571 (V_128_256): New mode iterator.
572 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
573 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
574 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
576 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
577 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
579 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
582 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
583 NaN inputs correctly.
585 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
588 * config/rs6000/vsx.md (??r): New mode attribute.
589 (*vsx_mov<mode>_64bit): Use it.
590 (*vsx_mov<mode>_32bit): Likewise.
592 2018-03-30 Martin Sebor <msebor@redhat.com>
594 PR tree-optimization/84818
595 * builtins.c (check_access): Use warning_n.
597 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
600 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
602 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
605 2018-03-30 Julia Koval <julia.koval@intel.com>
608 * x86-tune.def (movx, partial_reg_dependency): Enable for
611 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
614 * lra-constraints.c (process_alt_operands): Move setting
615 this_alternative_matches below.
617 2018-03-29 Martin Liska <mliska@suse.cz>
620 * doc/invoke.texi: Document how LTO works with debug info.
621 Describe auto-load support of binutils. Mention 'x86-64'
622 as valid option value of -march option.
624 2018-03-29 Jakub Jelinek <jakub@redhat.com>
626 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
629 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
630 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
631 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
634 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
637 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
638 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
639 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
640 for RS6000_BTM_POWERPC64.
641 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
642 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
643 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
648 2018-03-28 Carl Love <cel@us.ibm.com>
651 2017-09-27 Carl Love <cel@us.ibm.com>
653 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
654 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
655 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
658 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
660 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
661 instead of __vector bool.
662 (_mm_max_pu8): Likewise.
663 (_mm_min_pi16): Likewise.
665 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
668 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
672 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
674 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
675 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
676 (div_extend): Likewise.
677 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
679 (__builtin_divweuo): Likewise.
680 (__builtin_divdeo): Likewise.
681 (__builtin_divdeuo): Likewise.
683 2018-03-28 Jakub Jelinek <jakub@redhat.com>
686 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
687 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
689 PR tree-optimization/82004
690 * gimple-match-head.c (optimize_pow_to_exp): New function.
691 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
692 Don't fold to exp if optimize_pow_to_exp is false.
694 2018-03-28 Martin Liska <mliska@suse.cz>
697 * calls.c (initialize_argument_information): Fix trailing space.
698 * common.opt: Fix typo and provide better explanation for
699 -fsanitize-coverage option.
700 * config/i386/i386.opt: Fix typo.
702 2018-03-28 Jakub Jelinek <jakub@redhat.com>
703 Martin Liska <mliska@suse.cz>
706 * gimplify.c (asan_poison_variable): Don't do the check for
707 gimplify_omp_ctxp here.
708 (gimplify_decl_expr): Do it here.
709 (gimplify_target_expr): Likewise.
711 2018-03-28 Martin Liska <mliska@suse.cz>
714 * config/i386/i386.c (ix86_function_arg_advance): Do not call
715 chkp_type_bounds_count if MPX is not enabled.
717 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
719 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
721 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
724 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
725 function to create the function decl for complex long double
726 multiply and divide for -mabi=ieeelongdouble.
727 (init_float128_ieee): Call it.
729 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
732 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
733 -fcf-protection=branch -mibt.
734 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
736 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
739 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
741 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
744 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
745 extern array declarations.
747 2018-03-27 Richard Biener <rguenther@suse.de>
750 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
751 explicit single_use checks.
753 2018-03-27 Richard Biener <rguenther@suse.de>
755 PR tree-optimization/85082
756 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
759 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
761 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
762 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
763 Turn on fasynchronous-unwind-tables and funwind-tables.
765 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
768 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
769 (*bmi_blsr_<mode>_ccz): Ditto.
771 2018-03-26 Tom de Vries <tom@codesourcery.com>
773 PR tree-optimization/85063
774 * omp-general.c (offloading_function_p): New function. Factor out
776 * omp-offload.c (pass_omp_target_link::gate): ... here.
777 * omp-general.h (offloading_function_p): Declare.
778 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
779 with attribute omp declare target for offloading functions.
781 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
783 PR tree-optimization/84005
784 * tree-data-ref.h (get_base_for_alignment): Declare.
785 * tree-data-ref.c (get_base_for_alignment_1): New function.
786 (get_base_for_alignment): Likewise.
787 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
788 get_base_for_alignment to find a suitable base object, instead
789 of always using drb->base_address.
791 2018-03-23 Jakub Jelinek <jakub@redhat.com>
794 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
795 known size by default.
797 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
800 * lra-constraints.c (process_alt_operands): Don't match BLKmode
801 and non BLKmode operands.
803 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
806 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
809 2018-03-23 Richard Biener <rguenther@suse.de>
812 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
813 we are going to emit early debug for LTO.
815 2018-03-23 Jakub Jelinek <jakub@redhat.com>
818 * function.c (match_asm_constraints_1): Don't optimize if input
819 doesn't satisfy general_operand predicate for output's mode.
822 * alias.c (write_dependence_p): Don't require for x_canonicalized
823 non-VOIDmode if x has VOIDmode.
826 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
827 just don't try to optimize it rather than assert it never happens.
829 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
831 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
832 macro expansions for definition of ST_INTERNAL_<mode> and
833 LD_INTERNAL_<mode> builtins.
834 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
836 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
838 (altivec_expand_st_builtin): Likewise.
839 (altivec_expand_builtin): Remove calls to deleted functions.
840 (rs6000_address_for_altivec): Delete this function.
841 * config/rs6000/vector.md: Remove expands for
842 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
844 2018-03-22 Sudakshina Das <sudi.das@arm.com>
847 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
848 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
849 re-computing once computed.
850 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
851 (arm_init_machine_status): Initialize
852 machine->static_chain_stack_bytes.
854 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
857 * doc/extend.texi: Add four new prototypes for vec_ld.
858 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
859 definitions for more logical presentation.
860 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
861 entries for V1TI variants of __builtin_altivec_ld builtin.
862 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
863 handling of V1TI variant of LVX icode pattern.
864 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
865 (rs6000_gimple_fold_builtin): Likewise.
866 (altivec_init_builtins): Add code to define
867 __builtin_altivec_lvx_v1ti function.
869 2018-03-22 Jakub Jelinek <jakub@redhat.com>
872 * function.c (match_asm_constraints_1): Don't do the optimization
873 if input isn't a REG, SUBREG, MEM or constant.
875 2018-03-22 Tom de Vries <tom@codesourcery.com>
877 PR tree-optimization/84956
878 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
879 bb_has_abnormal_pred.
881 2018-03-22 Jakub Jelinek <jakub@redhat.com>
884 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
885 DECL_INITIAL (decl) to decl at the end.
886 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
889 2018-03-21 Joseph Myers <joseph@codesourcery.com>
891 * doc/extend.texi (__builtin_tgmath): Document when complex
892 integer types are treated as _Complex _Float64.
894 2018-03-21 Tom de Vries <tom@codesourcery.com>
896 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
898 2018-03-21 Jakub Jelinek <jakub@redhat.com>
900 PR tree-optimization/84960
901 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
902 if it is ENTRY block, move them into single succ of ENTRY in that case.
904 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
906 PR tree-optimization/84811
907 * poly-int.h (poly_span_traits): Remove the T3 parameter and
908 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
909 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
910 (known_subrange_p): Update accordingly. Cast each value involved
911 in the size comparison, rather than casting the result of the
914 2018-03-21 Jakub Jelinek <jakub@redhat.com>
916 PR tree-optimization/84982
917 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
918 by flipping the least significant bit rather than all bits from
919 bitpos to bitpos + bitsize - 1.
921 2018-03-21 Nathan Sidwell <nathan@acm.org>
923 * doc/extend.texi (Deprecated Features): Remove mention of
924 long-deleted deprecations.
926 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
929 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
930 * configure: Regenerate.
932 2018-03-21 Tom de Vries <tom@codesourcery.com>
934 PR tree-optimization/83126
935 * tree-parloops.c (num_phis): New function.
936 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
938 2018-03-21 Nathan Sidwell <nathan@acm.org>
940 * doc/extend.texi (Deprecated Features): Update deprecated flags,
941 mention anon-struct/union members and trailing attributes.
943 2018-03-21 Bin Cheng <bin.cheng@arm.com>
945 PR tree-optimization/84969
946 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
947 builtin memset partitions if they set different rhs values.
949 2018-03-21 Jakub Jelinek <jakub@redhat.com>
951 PR rtl-optimization/84989
952 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
953 VEC_DUPLICATE with scalar result mode.
955 2018-03-21 Martin Liska <mliska@suse.cz>
958 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
959 not intended return statement.
961 2018-03-21 Martin Liska <mliska@suse.cz>
964 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
965 (chkp_find_bound_slots_1): Limit number of iterations.
967 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
970 * Minor grammar fixes for x86 options.
972 2018-03-20 Jakub Jelinek <jakub@redhat.com>
975 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
976 holding REG_CFA_RESTORE notes, instead turn them into a USE.
978 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
981 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
982 (altivec_lvx_<mode>_1op): Likewise.
983 (altivec_stvx_<mode>_2op): Likewise.
984 (altivec_stvx_<mode>_1op): Likewise.
985 (altivec_lvx_<VM2:mode>): New define_expand.
986 (altivec_stvx_<VM2:mode>): Likewise.
987 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
988 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
989 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
990 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
991 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
992 (rs6000_gen_lvx): Likewise.
993 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
994 (altivec_expand_stv_builtin): Likewise.
995 (altivec_expand_builtin): Likewise.
996 * config/rs6000/vector.md: Likewise.
998 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1001 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1004 2018-03-20 Richard Biener <rguenther@suse.de>
1007 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1008 sign-conversions as zero, fall back to standard scalar_stmt
1011 2018-03-20 Martin Liska <mliska@suse.cz>
1014 * predict.c (rebuild_frequencies): Handle case when we have
1015 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1017 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1020 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1021 flag_section_anchors.
1022 * varasm.c (use_blocks_for_decl_p): Remove hack for
1023 dw2_force_const_mem.
1026 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1028 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1029 be created, use lowpart_subreg of operands[0] rather than operands[0]
1031 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1032 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1033 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1034 and n constraint instead of aarch64_shift_imm_di and Usd.
1035 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1036 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1038 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1041 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1042 to favor GPR over NEON registers.
1043 (<shift>di3_neon): Likewise.
1045 2018-03-20 Tom de Vries <tom@codesourcery.com>
1048 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1049 (nvptx_process_pars): Emit bar.sync asap and alap.
1051 2018-03-20 Tom de Vries <tom@codesourcery.com>
1054 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1055 seen_label if seen_label is already set.
1057 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1060 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1061 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1062 Use 1U instead of 1. Formatting fixes.
1065 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1066 instead of TREE_TYPE (s1) for the return value.
1068 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1070 PR tree-optimization/84946
1071 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1072 bitsize + bitsize in poly_uint64 rather than poly_int64.
1075 * dwarf2asm.c: Include fold-const.c.
1076 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1077 of decl rather than decl itself.
1079 PR rtl-optimization/84643
1080 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1082 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1085 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1086 calling assemble_variable.
1088 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1091 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1092 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1094 2018-03-19 Jim Wilson <jimw@sifive.com>
1097 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1098 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1099 (riscv_first_stack_step): Likewise.
1100 (riscv_option_override): Use STACK_BOUNDARY instead of
1102 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1104 (BIGGEST_ALIGNMENT): Set to 128.
1105 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1106 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1109 2018-03-19 Richard Biener <rguenther@suse.de>
1111 PR tree-optimization/84933
1112 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1113 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1115 2018-03-19 Richard Biener <rguenther@suse.de>
1117 PR tree-optimization/84859
1118 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1119 (cond_if_else_store_replacement): Perform sinking operation on
1120 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1121 Generalize what a BB with a single eligible store is.
1123 2018-03-19 Richard Biener <rguenther@suse.de>
1125 PR tree-optimization/84929
1126 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1127 chrec_is_positive against non-chrec arg.
1129 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1132 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1134 2018-03-18 Martin Liska <mliska@suse.cz>
1136 PR rtl-optimization/84635
1137 * regrename.c (build_def_use): Use matches_mode only when
1140 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1142 PR tree-optimization/84913
1143 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1144 vectorize chains of COND_EXPRs.
1146 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1148 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1150 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1152 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1154 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1156 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1158 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1159 Kito Cheng <kito.cheng@gmail.com>
1161 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1162 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1163 (nds32_adjust_reg_alloc_order): New function.
1164 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1166 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1168 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1169 nds32_print_operand, nds32_print_operand_address): Use
1170 HOST_WIDE_INT_PRINT_DEC instead.
1172 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1174 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1176 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1179 * config/i386/i386.c (initial_ix86_tune_features,
1180 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1182 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1183 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1184 rather than 1u << ix86_tune. Formatting fix.
1185 (ix86_option_override_internal): Change ix86_arch_mask from
1186 unsigned int to unsigned HOST_WIDE_INT, initialize to
1187 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1188 (ix86_function_specific_restore): Likewise.
1190 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1193 * postreload.c (reload_combine_recognize_pattern): Perform
1194 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1195 truncate_int_for_mode the result for the destination's mode.
1198 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1199 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1201 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1204 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1207 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1208 regno_allocno_class_array and sorted_pseudos.
1209 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1210 insns where regno is used.
1212 2018-03-16 Martin Liska <mliska@suse.cz>
1215 * multiple_target.c (create_dispatcher_calls): Redirect
1216 reference in the symbol table.
1218 2018-03-16 Martin Liska <mliska@suse.cz>
1221 * multiple_target.c (create_dispatcher_calls): Redirect also
1224 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1228 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1229 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1230 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1232 2018-03-16 Julia Koval <julia.koval@intel.com>
1234 * doc/invoke.texi (Skylake Server): Add CLWB.
1235 Cannonlake): Remove CLWB.
1237 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1239 PR tree-optimization/84841
1240 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1242 (FLOAT_ONE_CONST_TYPE): Define.
1243 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1244 (sort_by_operand_rank): Put entries with higher constant_type last
1245 rather than first to match comments.
1247 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1249 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1252 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1255 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1256 check and use address of target always.
1258 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1261 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1262 (indirect_thunk_bnd_needed): Likewise.
1263 (indirect_thunks_used): Likewise.
1264 (indirect_thunks_bnd_used): Likewise.
1265 (indirect_return_needed): New.
1266 (indirect_return_bnd_needed): Likewise.
1267 (output_indirect_thunk_function): Add a bool argument for
1269 (output_indirect_thunk_function): Don't generate alias for
1270 function return thunk.
1271 (ix86_code_end): Call output_indirect_thunk_function to generate
1272 function return thunks.
1273 (ix86_output_function_return): Set indirect_return_bnd_needed
1274 and indirect_return_needed instead of indirect_thunk_bnd_needed
1275 and indirect_thunk_needed.
1277 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1279 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1280 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1281 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1283 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1284 Paul Hua <paul.hua.gm@gmail.com>
1287 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1289 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1291 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1292 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1295 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1298 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1299 instead of GET_MODE_SIZE when comparing Units.
1301 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1304 * varasm.c (hash_section): Return an unchangeble hash value
1305 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1306 Return !aarch64_can_use_per_function_literal_pools_p ().
1308 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1311 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1312 rather than address of cmode as last argument to prepare_cmp_insn.
1314 2018-03-15 Julia Koval <julia.koval@intel.com>
1316 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1317 F_AVX512VNNI, F_AVX512BITALG): New.
1319 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1322 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1323 insn for floating-point loads and stores.
1325 2018-03-14 Carl Love <cel@us.ibm.com>
1327 * config/rs6000/rs6000-c.c: Add macro definitions for
1328 ALTIVEC_BUILTIN_VEC_PERMXOR.
1329 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1330 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1331 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1332 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1334 * config/doc/extend.texi: Add prototypes for vec_permxor.
1336 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1339 * diagnostic-show-locus.c (class layout_point): Convert m_line
1340 from int to linenum_type.
1341 (line_span::comparator): Use linenum "compare" function when
1342 comparing line numbers.
1343 (test_line_span): New function.
1344 (layout_range::contains_point): Convert param "row" from int to
1346 (layout_range::intersects_line_p): Likewise.
1347 (layout::will_show_line_p): Likewise.
1348 (layout::print_source_line): Likewise.
1349 (layout::should_print_annotation_line_p): Likewise.
1350 (layout::print_annotation_line): Likewise.
1351 (layout::print_leading_fixits): Likewise.
1352 (layout::annotation_line_showed_range_p): Likewise.
1353 (struct line_corrections): Likewise for field m_row.
1354 (line_corrections::line_corrections): Likewise for param "row".
1355 (layout::print_trailing_fixits): Likewise.
1356 (layout::get_state_at_point): Likewise.
1357 (layout::get_x_bound_for_row): Likewise.
1358 (layout::print_line): Likewise.
1359 (diagnostic_show_locus): Likewise for locals "last_line" and
1361 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1362 * input.c (selftest::test_linenum_comparisons): New function.
1363 (selftest::input_c_tests): Call it.
1364 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1365 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1366 * selftest.h (ASSERT_GT): New macro.
1367 (ASSERT_GT_AT): New macro.
1368 (ASSERT_LT): New macro.
1369 (ASSERT_LT_AT): New macro.
1371 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1373 PR rtl-optimization/84780
1374 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1376 2018-03-14 Martin Liska <mliska@suse.cz>
1378 * tree.c (record_node_allocation_statistics): Use
1379 get_stats_node_kind.
1380 (get_stats_node_kind): New function extracted from
1381 record_node_allocation_statistics.
1382 (free_node): Use get_stats_node_kind.
1384 2018-03-14 Richard Biener <rguenther@suse.de>
1386 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1387 that the value-set of ANTIC_IN doesn't grow.
1390 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1392 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1393 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1395 2018-03-14 Julia Koval <julia.koval@intel.com>
1397 * config.gcc (icelake-client, icelake-server): New.
1399 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1400 (initial_ix86_arch_features): Ditto.
1401 (PTA_SKYLAKE): Add SGX.
1402 (PTA_ICELAKE): Remove.
1403 (PTA_ICELAKE_CLIENT): New.
1404 (PTA_ICELAKE_SERVER): New.
1405 (ix86_option_override_internal): Split up icelake on icelake client and
1407 (get_builtin_code_for_version): Ditto.
1408 (fold_builtin_cpu): Ditto.
1409 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1410 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1411 * config/i386/i386.h (processor_type): Ditto.
1412 * doc/invoke.texi: Ditto.
1414 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1417 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1418 INTEGER_CST offset, add it together with bitpos / 8 and
1419 sign extend based on POINTER_SIZE.
1423 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1426 * config/i386/constraints.md (Yc): New register constraint.
1427 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1428 Use Yc constraint for alternative 2 of operand 0. Remove
1429 preferred_for_speed attribute.
1431 2018-03-14 Richard Biener <rguenther@suse.de>
1433 PR tree-optimization/84830
1434 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1435 with the old one to avoid oscillations.
1437 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1440 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1442 (assign_by_spills): Return a flag of reload assignment failure.
1443 Do not process the reload assignment failures. Do not spill other
1444 reload pseudos if they has the same reg class. Update n if
1446 (lra_assign): Add a return arg. Set up from the result of
1447 assign_by_spills call.
1448 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1449 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1450 usage_insns if it is not NULL.
1451 (spill_hard_reg_in_range): New function.
1452 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1453 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1454 function prototypes.
1455 (lra_assign): Change prototype.
1456 * lra.c (lra): Add code to deal with fails by splitting hard reg
1459 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1461 * config/riscv/riscv.opt (mrelax): New option.
1462 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1463 "norelax" when riscv_mrelax is disabled.
1464 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1466 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1469 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1470 reassociation for int modes.
1472 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1474 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1475 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1477 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1478 * config/aarch64/aarch64-sve.md
1479 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1480 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1481 (*extend<mode><Vwide>2): Rename to...
1482 (aarch64_sve_extend<mode><Vwide>2): ...this.
1483 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1484 renaming the old pattern to...
1485 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1487 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1488 define_expand, renaming the old pattern to...
1489 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1490 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1491 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1492 account when deciding which SVE instruction the optab should use.
1493 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1495 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1497 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1498 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1499 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1500 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1501 (tlsdesc_small_<mode>): Turn a define_expand and use
1502 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1503 (tlsdesc_small_advsimd_<mode>): ...this.
1504 (tlsdesc_small_sve_<mode>): New pattern.
1506 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1508 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1509 (UNSPEC_UMUL_HIGHPART): New constants.
1510 (MUL_HIGHPART): New int iteraor.
1511 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1512 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1514 (*<su>mul<mode>3_highpart): New define_insn.
1516 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1519 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1522 2018-03-13 Martin Liska <mliska@suse.cz>
1525 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1527 (sem_item_optimizer::~sem_item_optimizer): Release it.
1528 (sem_item_optimizer::merge_classes): Register variable aliases.
1529 (sem_item_optimizer::fixup_pt_set): New function.
1530 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1531 * ipa-icf.h: Declare new variables and functions.
1533 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1536 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1537 integer_pow2p@2 and test integer_pow2p in condition.
1538 (A < 0 ? C : 0): Similarly for @1.
1541 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1542 characters starting at p contain '\0' character, don't look beyond
1546 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1547 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1550 * reg-stack.c (change_stack): Change update_end var from int to
1551 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1552 also call set_block_for_insn on the newly added insns and rescan.
1555 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1556 on the last operand.
1559 * tree.c (stabilize_reference_1): Return save_expr (e) for
1560 STATEMENT_LIST even if it doesn't have side-effects.
1562 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1564 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1566 2018-03-12 Renlin Li <renlin.li@arm.com>
1568 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1569 aarch64_output_scalar_simd_mov_immediate.
1571 2018-03-12 Martin Sebor <msebor@redhat.com>
1573 PR tree-optimization/83456
1574 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1575 for perfectly overlapping calls to memcpy.
1576 (gimple_fold_builtin_memory_chk): Same.
1577 (gimple_fold_builtin_strcpy): Handle no-warning.
1578 (gimple_fold_builtin_stxcpy_chk): Same.
1579 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1581 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1583 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1584 parameter. Use it for SFmode.
1585 (rs6000_function_arg_advance_1): Adjust.
1586 (rs6000_function_arg): Adjust.
1587 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1589 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1591 PR rtl-optimization/84169
1592 PR rtl-optimization/84780
1593 * combine.c (can_combine_p): Check for a 2-insn combination whether
1594 the destination register is used between the two insns, too.
1596 2018-03-12 Richard Biener <rguenther@suse.de>
1598 PR tree-optimization/84803
1599 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1600 for refs DR analysis didn't process.
1602 2018-03-12 Richard Biener <rguenther@suse.de>
1604 PR tree-optimization/84777
1605 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1606 force-vectorize loops ignore whether we are optimizing for size.
1608 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1610 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1611 (TARGET_MD_ASM_ADJUST): Define.
1613 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1614 Kito Cheng <kito.cheng@gmail.com>
1615 Chung-Ju Wu <jasonwucj@gmail.com>
1617 * config/nds32/nds32.c (nds32_compute_stack_frame,
1618 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1619 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1620 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1621 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1622 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1623 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1624 * config/nds32/nds32.md (prologue, epilogue): Use macro
1625 NDS32_V3PUSH_AVAILABLE_P to do checking.
1627 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1630 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1631 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1632 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1633 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1634 addition of most attributes on !orig_type_die or the attribute not
1635 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1637 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1638 Chung-Ju Wu <jasonwucj@gmail.com>
1640 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1642 * config/nds32/nds32.opt (mvh): New option.
1644 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1645 Chung-Ju Wu <jasonwucj@gmail.com>
1647 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1649 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1650 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1653 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1654 Chung-Ju Wu <jasonwucj@gmail.com>
1656 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1658 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1659 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1661 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1662 Kito Cheng <kito.cheng@gmail.com>
1663 Chung-Ju Wu <jasonwucj@gmail.com>
1665 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1666 UNSPEC_FFMISM and UNSPEC_FLMISM.
1667 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1668 for ffb, ffmism and flmism.
1669 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1670 (unspec_ffmism): Ditto.
1671 (unspec_flmism): Ditto.
1672 (nds32_expand_builtin_impl): Check if string extension is available.
1673 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1674 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1676 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1679 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1682 * lra-assigns.c (assign_by_spills): Return a flag of reload
1683 assignment failure. Do not process the reload assignment
1684 failures. Do not spill other reload pseudos if they has the same
1686 (lra_assign): Add a return arg. Set up from the result of
1687 assign_by_spills call.
1688 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1689 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1690 usage_insns if it is not NULL.
1691 (spill_hard_reg_in_range): New function.
1692 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1693 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1694 function prototypes.
1695 (lra_assign): Change prototype.
1696 * lra.c (lra): Add code to deal with fails by splitting hard reg
1699 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1702 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1704 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1707 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1708 (dw_val_node): Add val_symbolic_view.
1709 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1710 (symview_upper_bound): New.
1711 (new_line_info_table): Initialize symviews_since_reset.
1712 (dwarf2out_source_line): Count symviews_since_reset and set
1713 symview_upper_bound.
1714 (dw_val_equal_p): Handle symview.
1715 (add_AT_symview): New.
1716 (print_dw_val): Handle symview.
1717 (attr_checksum, attr_checksum_ordered): Likewise.
1718 (same_dw_val_p, size_of_die): Likewise.
1719 (value_format, output_die): Likewise.
1720 (add_high_low_attributes): Use add_AT_symview for entry_view.
1721 (dwarf2out_finish): Reset symview_upper_bound, clear
1724 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1727 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1728 Add strict argument and use it.
1729 (rs6000_split_multireg_move): Update for new strict argument.
1730 (mem_operand_gpr): Disallow all non-offsettable addresses.
1731 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1733 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1736 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1737 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1738 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1741 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1742 decl, use remap_type if we want to use the type.
1744 2018-03-09 Martin Sebor <msebor@redhat.com>
1746 PR tree-optimization/84526
1747 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1749 (builtin_access::generic_overlap): Be prepared to handle non-array
1752 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1754 PR rtl-optimization/84682
1755 * lra-constraints.c (process_address_1): Check is_address flag
1756 for address constraints.
1757 (process_alt_operands): Likewise.
1758 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1759 preprocess_constraints.
1760 * recog.h (preprocess_constraints): Add oploc parameter.
1762 * recog.c (preprocess_constraints): Test address_operand for
1763 CT_ADDRESS constraints.
1765 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1768 * lra-assigns.c (assign_by_spills): Return a flag of reload
1769 assignment failure. Do not process the reload assignment
1770 failures. Do not spill other reload pseudos if they has the same
1772 (lra_assign): Add a return arg. Set up from the result of
1773 assign_by_spills call.
1774 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1775 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1776 usage_insns if it is not NULL.
1777 (spill_hard_reg_in_range): New function.
1778 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1779 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1780 function prototypes.
1781 (lra_assign): Change prototype.
1782 * lra.c (lra): Add code to deal with fails by splitting hard reg
1785 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1788 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1789 Accept complain bool parameter. Only emit errors if it is true.
1790 (arm_parse_cpu_option_name): Likewise.
1791 (arm_target_thumb_only): Adjust callers of the above.
1792 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1793 prototype to take a default true bool parameter.
1794 (arm_parse_arch_option_name): Likewise.
1796 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1797 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1801 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1802 * configure: Regenerate.
1803 * configure.ac ("linker --version-script option"): New.
1804 ("linker soname option"): New.
1806 2018-03-09 Richard Biener <rguenther@suse.de>
1808 PR tree-optimization/84775
1809 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1810 immediate uses of predicate stmts and mark them modified.
1813 PR tree-optimization/84178
1814 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1816 (version_loop_for_if_conversion): Delay update_ssa call.
1817 (tree_if_conversion): Delay update_ssa until after predicate
1820 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1823 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1824 when the function accesses prior frames.
1826 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1829 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1830 gen_llsym, otherwise call maybe_gen_llsym.
1833 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1834 has ',' character inside of it.
1836 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1839 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1840 as clobbering CC_REGNUM.
1842 2018-03-08 Richard Biener <rguenther@suse.de>
1845 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1846 (follow_copies_to_constant): Do not follow SSA names registered
1849 2018-03-08 Richard Biener <rguenther@suse.de>
1851 PR tree-optimization/84178
1852 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1854 (version_loop_for_if_conversion): Delay update_ssa call.
1855 (tree_if_conversion): Delay update_ssa until after predicate
1858 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1860 PR tree-optimization/84178
1861 * tree-if-conv.c (release_bb_predicate): Remove the
1862 the assertion that the stmts have NULL use_ops.
1863 Discard the statements, asserting that they haven't
1864 yet been added to a BB.
1866 2018-03-08 Richard Biener <rguenther@suse.de>
1868 PR tree-optimization/84746
1869 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1870 (phi_translate): Pass in destination ANTIC_OUT set.
1871 (phi_translate_1): Likewise. For a simplified result lookup
1872 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1873 (phi_translate_set): Adjust.
1874 (do_pre_regular_insertion): Likewise.
1875 (do_pre_partial_partial_insertion): Likewise.
1877 2018-03-08 Martin Liska <mliska@suse.cz>
1879 PR gcov-profile/84735
1880 * doc/gcov.texi: Document usage of profile files.
1881 * gcov-io.h: Document changes in the format.
1883 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1887 * dwarf2out.c (struct dw_line_info_table): Update comments for
1889 (FORCE_RESET_NEXT_VIEW): New.
1890 (FORCE_RESETTING_VIEW_P): New.
1891 (RESETTING_VIEW_P): Check for -1 too.
1892 (ZERO_VIEW_P): Likewise.
1893 (new_line_info_table): Force-reset next view.
1894 (dwarf2out_begin_function): Likewise.
1895 (dwarf2out_source_line): Simplify zero_view_p initialization.
1896 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1897 view directly. Omit view when omitting .loc at line 0.
1899 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1901 PR tree-optimization/84740
1902 * tree-switch-conversion.c (process_switch): Call build_constructors
1903 only if info.phi_count is non-zero.
1905 PR tree-optimization/84739
1906 * tree-tailcall.c (find_tail_calls): Check call arguments against
1907 DECL_ARGUMENTS (current_function_decl) rather than
1908 DECL_ARGUMENTS (func) when checking for tail recursion.
1910 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1912 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1913 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1914 Volker Reichelt's entry and add entries for people that perform
1915 GCC fuzzy testing and report numerous bugs.
1917 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
1920 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
1921 readonly data in sdata, if that is disabled.
1922 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
1923 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1924 -mreadonly-in-sdata option.
1926 2018-03-07 Martin Sebor <msebor@redhat.com>
1928 PR tree-optimization/84468
1929 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
1930 basic block when looking for nul assignment.
1932 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
1935 * except.h (output_function_exception_table): Adjust prototype.
1936 * except.c (output_function_exception_table): Remove FNNAME parameter
1937 and add SECTION parameter. Ouput one part of the table at a time.
1938 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
1939 the first part of the exception table and emit unwind directives.
1940 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
1941 (i386_pe_seh_cold_init): Likewise.
1942 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
1943 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
1944 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
1945 (ix86_output_call_insn): Emit a nop in one more case for SEH.
1946 * config/i386/winnt.c: Include except.h.
1947 (struct seh_frame_state): Add reg_offset, after_prologue and
1948 in_cold_section fields.
1949 (i386_pe_seh_end_prologue): Set seh->after_prologue.
1950 (i386_pe_seh_cold_init): New function.
1951 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
1952 to seh->in_cold_section.
1953 (seh_emit_push): Record the offset of the push.
1954 (seh_emit_save): Record the offet of the save.
1955 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
1956 Test seh->after_prologue to disregard the epilogue.
1957 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
1958 (i386_pe_end_cold_function): New function.
1960 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1963 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
1964 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
1967 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
1969 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
1970 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
1973 * multiple_target.c: Include tree-inline.h and intl.h.
1974 (expand_target_clones): Diagnose and fail if node->definition and
1975 !tree_versionable_function_p (node->decl).
1977 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
1979 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
1981 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
1982 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
1983 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
1985 2018-03-06 Jakub Jelinek <jakub@redhat.com>
1988 * combine.c (try_combine): Use reg_or_subregno instead of handling
1989 just paradoxical SUBREGs and REGs.
1991 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
1993 * config/arc/arc.c (arc_finalize_pic): Remove function.
1994 (arc_must_save_register): We use single base PIC register, remove
1995 checks to save/restore the PIC register.
1996 (arc_expand_prologue): Likewise.
1997 * config/arc/arc-protos.h (arc_set_default_type_attributes):
1999 (arc_verify_short): Likewise.
2000 (arc_attr_type): Likewise.
2001 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2002 (walk_stores): Likewise.
2003 (arc_address_cost): Make it static.
2004 (arc_verify_short): Likewise.
2005 (branch_dest): Likewise.
2006 (arc_attr_type): Likewise.
2007 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2008 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2009 (arc_final_prescan_insn): Remove inserting the nops due to
2010 hardware hazards. It is done in reorg step.
2011 (insn_length_variant_t): Remove.
2012 (insn_length_parameters_t): Likewise.
2013 (arc_insn_length_parameters): Likewise.
2014 (arc_get_insn_variants): Likewise.
2015 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2017 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2020 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2023 PR tree-optimization/84687
2024 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2026 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2028 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2030 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2031 Rename to ppc_speculation_barrier.
2032 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2033 __builtin_ppc_speculation_barrier.
2035 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2038 * combine.c (combine_simplify_rtx): Don't try to simplify if
2039 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2042 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2044 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2045 to 32 bytes when compiling for POWER9.
2047 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2050 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2051 regparm >= 3 with no arg reg available also for calls with
2052 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2055 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2057 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2059 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2062 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2064 2018-03-05 Richard Biener <rguenther@suse.de>
2066 PR tree-optimization/84486
2067 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2068 When inserting a __builtin_assume_aligned call set the LHS
2069 SSA name alignment info accordingly.
2071 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2073 PR tree-optimization/84114
2074 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2075 Avoid reassociation of FLOAT_MODE addition.
2077 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2079 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2080 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2081 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2082 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2083 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2084 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2085 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2087 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2088 __builtin_ia32_wbinvd): New builtins.
2089 (SPECIAL_ARGS2): New.
2090 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2091 (SPECIAL_ARGS2): New.
2092 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2093 (ix86_valid_target_attribute_inner_p): Ditto.
2094 (ix86_init_mmx_sse_builtins): Add special_args2.
2095 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2096 TARGET_WBNOINVD_P): New.
2097 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2098 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2099 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2100 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2101 * config/i386/pconfigintrin.h: New file.
2102 * config/i386/wbnoinvdintrin.h: Ditto.
2103 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2104 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2106 2018-03-05 Richard Biener <rguenther@suse.de>
2108 PR tree-optimization/84670
2109 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2111 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2112 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2113 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2114 doesn't grow if all successors (recursively) were visited at least
2117 2018-03-05 Richard Biener <rguenther@suse.de>
2119 PR tree-optimization/84650
2120 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2121 if executed in the loop pipeline.
2123 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2125 * doc/configfiles.texi (Configuration Files): Move info about
2126 conditionalizing $target-protos.h to...
2127 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2128 differs from $target-protos.h.
2130 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2131 Chung-Ju Wu <jasonwucj@gmail.com>
2133 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2134 * config/nds32/nds32-multiple.md (setmemsi): Define.
2135 * config/nds32/nds32-memory-manipulation.c
2136 (nds32_gen_dup_4_byte_to_word_value): New.
2137 (emit_setmem_word_loop): New.
2138 (emit_setmem_byte_loop): New.
2139 (nds32_expand_setmem_loop): New.
2140 (nds32_expand_setmem_loop_v3m): New.
2141 (nds32_expand_setmem_unroll): New.
2142 (nds32_expand_setmem): New.
2144 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2145 Chung-Ju Wu <jasonwucj@gmail.com>
2147 * config/nds32/nds32-memory-manipulation.c
2148 (nds32_emit_load_store): New.
2149 (nds32_emit_post_inc_load_store): New.
2150 (nds32_emit_mem_move): New.
2151 (nds32_emit_mem_move_block): New.
2152 (nds32_expand_movmemsi_loop_unknown_size): New.
2153 (nds32_expand_movmemsi_loop_known_size): New.
2154 (nds32_expand_movmemsi_loop): New.
2155 (nds32_expand_movmemsi_unroll): New.
2156 (nds32_expand_movmemqi): Rename ...
2157 (nds32_expand_movmemsi): ... to this.
2158 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2159 (movmemsi): ... to this.
2160 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2161 (nds32_expand_movmemsi): ... to this.
2163 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2164 Monk Chiang <sh.chiang04@gmail.com>
2165 Chung-Ju Wu <jasonwucj@gmail.com>
2167 * config/nds32/nds32-protos.h
2168 (nds32_expand_load_multiple): New arguments.
2169 (nds32_expand_store_multiple): Ditto.
2170 (nds32_valid_multiple_load_store): Rename ...
2171 (nds32_valid_multiple_load_store_p): ... to this.
2172 * config/nds32/nds32-memory-manipulation.c
2173 (nds32_expand_load_multiple): Refine implementation.
2174 (nds32_expand_store_multiple): Ditto.
2175 * config/nds32/nds32-multiple.md
2176 (load_multiple): Update nds32_expand_load_multiple interface.
2177 (store_multiple): Update nds32_expand_store_multiple interface.
2178 * config/nds32/nds32-predicates.c
2179 (nds32_valid_multiple_load_store): Rename ...
2180 (nds32_valid_multiple_load_store_p): ... to this and refine
2182 * config/nds32/predicates.md
2183 (nds32_load_multiple_and_update_address_operation): New predicate.
2184 (nds32_store_multiple_and_update_address_operation): New predicate.
2186 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2187 Chung-Ju Wu <jasonwucj@gmail.com>
2189 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2190 (combo): New attribute.
2191 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2193 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2195 * config/nds32/nds32.opt: Change -mcmodel= default value.
2197 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2198 Monk Chiang <sh.chiang04@gmail.com>
2199 Chung-Ju Wu <jasonwucj@gmail.com>
2201 * config/nds32/constants.md (unspec_element): New enum.
2202 * config/nds32/constraints.md (Umw): New constraint.
2203 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2204 * config/nds32/nds32-intrinsic.md: Likewise.
2205 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2206 (nds32_valid_smw_lwm_base_p): New.
2207 (nds32_output_smw_single_word): New.
2208 (nds32_output_lmw_single_word): New.
2209 (nds32_expand_unaligned_load): New.
2210 (nds32_expand_unaligned_store): New.
2211 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2212 (nds32_output_smw_single_word): Declare.
2213 (nds32_output_lmw_single_word): Declare.
2214 (nds32_expand_unaligned_load): Declare.
2215 (nds32_expand_unaligned_store): Declare.
2216 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2217 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2218 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2219 NDS32_BUILTIN_UASTORE_DW.
2220 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2223 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2224 Kito Cheng <kito.cheng@gmail.com>
2225 Chung-Ju Wu <jasonwucj@gmail.com>
2227 * config/nds32/nds32-intrinsic.c
2228 (nds32_expand_builtin_null_ftype_reg): Delete.
2229 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2230 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2231 (nds32_read_argument): New.
2232 (nds32_legitimize_target): Ditto.
2233 (nds32_legitimize_argument): Ditto.
2234 (nds32_check_constant_argument): Ditto.
2235 (nds32_expand_unop_builtin): Ditto.
2236 (nds32_expand_unopimm_builtin): Ditto.
2237 (nds32_expand_binop_builtin): Ditto.
2238 (nds32_builtin_decl_impl): Ditto.
2239 (builtin_description): Ditto.
2240 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2241 (nds32_init_builtins_impl): Ditto.
2242 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2243 (nds32_builtin_decl): New.
2244 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2245 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2247 2018-03-02 Jeff Law <law@redhat.com>
2249 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2250 (redundant_insn, fill_simple_delay_slots): Likewise.
2251 (fill_slots_from_thread): Likewise.
2252 * resource.c (mark_referenced_resources): Likewise.
2253 (mark_set_resources, find_dead_or_set_registers): Likewise.
2255 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2257 * substring-locations.h (format_warning_va): Formatting fix for
2259 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2261 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2262 * substring-locations.c: Include intl.h.
2263 (format_warning_va): Turned into small wrapper around
2264 format_warning_n_va, renamed to ...
2265 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2266 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2268 (format_warning_at_substring_n): New function.
2269 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2270 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2271 format_warning_at_substring with just a shorter name instead of
2272 const function pointer.
2273 (fmtwarn_n): New function.
2274 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2275 appropriate, get rid of all the fmtstr temporaries, move conditionals
2276 with G_() wrapped string literals directly into fmtwarn arguments,
2277 cast dir.len to (int), formatting fixes.
2279 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2281 * doc/invoke.texi: Remove "Cilk Plus" references.
2283 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2284 Richard Biener <rguenther@suse.de>
2287 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2288 for error or warning attributes if CALL_FROM_THUNK_P is set.
2291 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2294 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2295 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2298 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2299 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2301 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2302 instead of -1U in last predictors element's probability member.
2304 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2307 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2308 arguments if they are comparable.
2310 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2312 PR tree-optimization/84634
2313 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2314 masks and masked_loop_p with a single loop_masks, making sure it's
2315 null for bb vectorization.
2317 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2319 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2320 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2321 loop->force_vectorize to check whether there is no alias.
2323 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2326 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2328 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2330 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2331 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2332 instead of a loop around prev_real_insn.
2333 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2337 * config/i386/i386.c (ix86_print_operand): Use conditional
2338 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2341 2018-03-02 Richard Biener <rguenther@suse.de>
2343 PR tree-optimization/84427
2344 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2345 (bitmap_set_subtract_values): Rewrite to handle multiple
2348 (prune_clobbered_mems): Likewise.
2349 (phi_translate): Take edge instead of pred/phiblock.
2350 (phi_translate_1): Likewise.
2351 (phi_translate_set): Likewise. Insert all translated
2352 exprs for a value into the set, keeping possibly multiple
2353 expressions per value.
2354 (compute_antic_aux): Adjust for phi_translate changes.
2355 When intersecting union the expressions and prune those
2356 not in the final value set, keeping possibly multiple
2357 expressions per value. Do not use value-insertion
2358 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2359 all expressions. Add verification that the value-sets
2360 only shrink during iteration.
2361 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2362 (do_pre_regular_insertion): Likewise.
2363 (do_pre_partial_partial_insertion): Likewise.
2365 2018-03-02 Richard Biener <rguenther@suse.de>
2368 * config/darwin.c (saved_debug_info_level): New static global.
2369 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2370 (darwin_asm_lto_end): Restore debug info generation settings.
2372 2018-03-01 Martin Liska <mliska@suse.cz>
2375 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2378 2018-03-01 Richard Biener <rguenther@suse.de>
2381 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2382 type annotation with LTO when debug was disabled at compile-time.
2384 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2386 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2388 (mips_final_postscan_insn): Likewise.
2390 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2392 PR rtl-optimization/84528
2393 * alias.c (init_alias_target): Add commentary.
2394 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2395 a unique base value if the frame pointer is not eliminated
2396 to the stack pointer.
2398 2018-03-01 Tom de Vries <tom@codesourcery.com>
2400 PR rtl-optimization/83327
2401 * lra-int.h (hard_regs_spilled_into): Declare.
2402 * lra.c (hard_regs_spilled_into): Define.
2403 (init_reg_info): Init hard_regs_spilled_into.
2404 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2405 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2406 (process_bb_lives): Handle hard_regs_spilled_into.
2407 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2408 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2410 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2412 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2413 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2414 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2415 * config/rs6000/aix72.h: New file.
2417 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2419 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2420 instead of warning_at with conditional singular and plural messages
2424 * stor-layout.c (update_alignment_for_field): For
2425 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2426 && !DECL_PACKED (field), do the alignment update, just use
2427 only desired_align instead of MAX (type_align, desired_align)
2429 (place_field): Don't do known_align < desired_align handling
2430 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2431 is non-NULL, instead do it after rli->prev_field handling and
2432 only if not within a bitfield word. For DECL_PACKED (field)
2433 use type_align of BITS_PER_UNIT.
2435 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2437 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2438 superfluous parentheses and trailing spaces.
2440 2018-02-28 Richard Biener <rguenther@suse.de>
2442 PR tree-optimization/84584
2443 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2444 SCoPs with fake exit edge.
2446 2018-02-28 Martin Liska <mliska@suse.cz>
2449 * timevar.c (timer::print): Fix format to properly print 100%
2452 2018-02-28 Richard Biener <rguenther@suse.de>
2455 * genmatch.c (capture_info::walk_match): Do not mark
2456 captured expressions without operands as expr_p given
2457 they act more like predicates and should be subject to
2458 "lost tail" side-effect preserving.
2460 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2462 PR rtl-optimization/81611
2463 * auto-inc-dec.c (attempt_change): Move dead note from
2464 mem_insn if it's the next use of regno
2465 (find_address): Take address use of reg holding
2466 non-incremented value. Add parm to limit search to the named
2468 (merge_in_block): Attempt to use a mem insn that is the next
2469 use of the original regno.
2471 2018-02-27 Martin Sebor <msebor@redhat.com>
2474 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2475 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2477 2018-02-27 Martin Sebor <msebor@redhat.com>
2479 PR translation/84207
2480 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2481 n argument to unsigned HOST_WIDE_INT.
2482 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2483 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2484 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2485 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2487 2018-02-27 Richard Biener <rguenther@suse.de>
2489 PR tree-optimization/84512
2490 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2491 Do not use the estimate returned from record_stmt_cost for
2492 the scalar iteration cost but sum properly using add_stmt_cost.
2494 2018-02-27 Richard Biener <rguenther@suse.de>
2496 PR tree-optimization/84466
2497 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2498 Adjust last change to less strictly validate use operands.
2500 2018-02-27 Martin Liska <mliska@suse.cz>
2502 PR gcov-profile/84548
2503 * gcov.c (process_file): Allow partial overlap and consider it
2504 also as group functions.
2505 (output_lines): Properly calculate range of lines for a group.
2507 2018-02-27 Martin Liska <mliska@suse.cz>
2509 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2510 'ggc' suffixes. Change first column width.
2511 (timer::print): Fix formatting of the column.
2513 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2515 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2516 preserve inline entry blocks for the sake of debug inline
2517 entry point markers alone.
2518 (remove_unused_locals): Suggest in comments a better place to
2519 force the preservation of inline entry blocks that are
2520 otherwise unused, but do not preserve them.
2522 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2524 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2526 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2529 * config/i386/constraints.md (Bs): Replace
2530 ix86_indirect_branch_register with
2531 TARGET_INDIRECT_BRANCH_REGISTER.
2533 * config/i386/i386.md (indirect_jump): Likewise.
2534 (tablejump): Likewise.
2535 (*sibcall_memory): Likewise.
2536 (*sibcall_value_memory): Likewise.
2537 Peepholes of indirect call and jump via memory: Likewise.
2538 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2539 (*sibcall_value_GOT_32): Likewise.
2540 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2541 (GOT_memory_operand): Likewise.
2542 (call_insn_operand): Likewise.
2543 (sibcall_insn_operand): Likewise.
2544 (GOT32_symbol_operand): Likewise.
2545 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2547 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2549 PR rtl-optimization/83496
2550 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2551 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2552 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2553 redundant insn, if any.
2554 (relax_delay_slots): Likewise.
2555 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2557 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2559 PR tree-optimization/83965
2560 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2561 that grouped statements are part of a reduction chain. Return
2562 true if the statement is not marked as a reduction itself but
2564 (vect_recog_dot_prod_pattern): Don't check whether the statement
2565 is part of a group here.
2566 (vect_recog_sad_pattern): Likewise.
2567 (vect_recog_widen_sum_pattern): Likewise.
2569 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2572 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2574 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2577 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2579 (ix86_output_indirect_function_return): New prototype.
2580 (ix86_split_simple_return_pop_internal): Likewise.
2581 * config/i386/i386.c (indirect_return_via_cx): New.
2582 (indirect_return_via_cx_bnd): Likewise.
2583 (indirect_thunk_name): Handle return va CX_REG.
2584 (output_indirect_thunk_function): Create alias for
2585 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2586 (ix86_output_indirect_jmp): Remove the bool argument.
2587 (ix86_output_indirect_function_return): New function.
2588 (ix86_split_simple_return_pop_internal): Likewise.
2589 * config/i386/i386.md (*indirect_jump): Don't pass false
2590 to ix86_output_indirect_jmp.
2591 (*tablejump_1): Likewise.
2592 (simple_return_pop_internal): Change it to define_insn_and_split.
2593 Call ix86_split_simple_return_pop_internal to split it for
2595 (simple_return_indirect_internal): Call
2596 ix86_output_indirect_function_return instead of
2597 ix86_output_indirect_jmp.
2599 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2602 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2603 memset and value initialization afterwards.
2605 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2607 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2609 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2612 * common/config/aarch64/aarch64-common.c
2613 (aarch_option_optimization_table[]): Switch
2614 off fomit-frame-pointer
2616 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2617 Chung-Ju Wu <jasonwucj@gmail.com>
2619 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2621 (store_multiple): Ditto.
2623 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2625 * config.gcc: Add --with-cpu support for nds32 target.
2626 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2627 * config/nds32/nds32.opt: Add -mcpu= option.
2629 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2631 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2632 isel=yes): Warn for these deprecated options.
2634 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2636 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2637 ISA_2_5_MASKS_EMBEDDED.
2639 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2641 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2642 p->max as pointers rather than using iterative_hash_expr.
2644 2018-02-23 Carl Love <cel@us.ibm.com>
2646 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2647 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2649 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2650 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2651 P8V_BUILTIN_VEC_VUNSIGNED2.
2653 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2656 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2657 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2659 * lra-constraints.c (curr_insn_transform): Set up
2660 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2662 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2663 * lra-eliminations.c (spill_pseudos): Ditto.
2664 (process_insn_for_elimination): Ditto.
2665 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2666 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2667 LRA_NON_CLOBBERED_ALT.
2669 2018-02-22 Martin Sebor <msebor@redhat.com>
2671 PR tree-optimization/84480
2672 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2673 to maybe_diag_stxncpy_trunc. Call it.
2674 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2675 from gimple_fold_builtin_strcpy. Print inlining stack.
2676 (handle_builtin_stxncpy): Print inlining stack.
2677 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2679 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2682 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2683 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2684 and -fcheck-pointer-bounds are used together.
2685 (indirect_thunk_prefix): New enum.
2686 (indirect_thunk_need_prefix): New function.
2687 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2688 "_nt" instead of "_bnd" for NOTRACK prefix.
2689 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2690 (output_indirect_thunk_function): Likewise.
2692 (ix86_code_end): Update output_indirect_thunk_function calls.
2693 (ix86_output_indirect_branch_via_reg): Replace
2694 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2695 (ix86_output_indirect_branch_via_push): Likewise.
2696 (ix86_output_function_return): Likewise.
2697 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2698 incompatible with -fcf-protection=branch and
2699 -fcheck-pointer-bounds.
2701 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2704 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2705 Change gcc_assert call to output_operand_lossage.
2707 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2709 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2711 2018-02-22 DJ Delorie <dj@redhat.com>
2712 Sebastian Perta <sebastian.perta@renesas.com>
2713 Oleg Endo <olegendo@gcc.gnu.org>
2715 * config/rx/rx.c (rx_rtx_costs): New function.
2716 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2718 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2720 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2722 2018-02-22 Martin Liska <mliska@suse.cz>
2725 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2726 Add "native" as a possible value.
2728 2018-02-22 Martin Liska <mliska@suse.cz>
2731 * config/i386/i386.c (ix86_option_override_internal):
2732 Add "native" as a possible value for -march and -mtune.
2734 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2737 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2738 to all type variants.
2740 PR tree-optimization/84503
2741 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2742 width as info->bitpos + info->bitsize - start.
2743 (merged_store_group::merge_overlapping): Simplify width computation.
2744 (check_no_overlap): New function.
2745 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2746 start + width and last_order of the group, fail if check_no_overlap
2748 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2749 to group if check_no_overlap fails.
2751 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2753 * config/rs6000/altivec.md: Delete contraint arguments to
2754 define_expand, define_split, and define_peephole2, and in
2755 define_insn_and_split if always unused.
2756 * config/rs6000/darwin.md: Ditto.
2757 * config/rs6000/dfp.md: Ditto.
2758 * config/rs6000/rs6000.md: Ditto.
2759 * config/rs6000/sync.md: Ditto.
2760 * config/rs6000/vector.md: Ditto.
2761 * config/rs6000/vsx.md: Ditto.
2763 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2765 * config/rs6000/altivec.md: Write output control strings as braced
2766 blocks instead of double-quoted strings.
2767 * config/rs6000/darwin.md: Ditto.
2768 * config/rs6000/rs6000.md: Ditto.
2769 * config/rs6000/vector.md: Ditto.
2770 * config/rs6000/vsx.md: Ditto.
2772 2018-02-21 Jason Merrill <jason@redhat.com>
2774 PR c++/84314 - ICE with templates and fastcall attribute.
2775 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2777 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2779 * ipa-cp.c (determine_versionability): Fix comment typos.
2781 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2784 * ipa-cp.c (determine_versionability): Do not version functions caling
2787 2018-02-21 Martin Liska <mliska@suse.cz>
2790 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2791 Add "native" as a possible value.
2792 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2793 the macro when native cpu detection is available.
2795 2018-02-21 Martin Liska <mliska@suse.cz>
2798 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2799 Add "native" as a possible value.
2800 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2801 when native cpu detection is available.
2803 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2804 Martin Sebor <msebor@redhat.com>
2806 PR tree-optimization/84478
2807 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2809 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2810 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2811 support which is conservatively correct, for 2 only stay conservative
2812 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2813 argument to the 2 argument get_range_strlen, adjust 6 arg
2814 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2816 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2817 (gimple_fold_builtin_strlen): Pass true as last argument to
2820 2018-02-20 Martin Sebor <msebor@redhat.com>
2823 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2824 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2825 (builtin_memref::builtin_memref): Factor out parts into
2826 set_base_and_offset and call it.
2828 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2831 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2832 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2833 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2834 search at the associated MODE_INT.
2836 2018-02-20 Jeff Law <law@redhat.com>
2839 PR tree-optimization/81592
2841 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2842 for range data rather than using global data.
2843 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2844 range data rather than using global data.
2845 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2846 pass it to children as needed.
2847 (struct directive::fmtresult): Similarly.
2848 (struct directive::set_width): Similarly.
2849 (struct directive::set_precision): Similarly.
2850 (format_integer, format_directive, parse_directive): Similarly.
2851 (format_none): Accept unnamed vr_values parameter.
2852 (format_percent, format_floating, format_character): Similarly.
2853 (format_string, format_plain): Similarly.
2854 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2855 the EVRP range analyzer for range data rather than using global data.
2856 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2857 gimple-ssa-evrp-analyze.h
2858 (class sprintf_dom_walker): Add after_dom_children member function.
2859 Add evrp_range_analyzer member.
2860 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2861 range analyzer as needed.
2862 (sprintf_dom_walker::after_dom_children): New member function.
2863 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2865 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2866 (evrp_range_analyzer::pop_to_marker): Likewise.
2868 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2870 PR tree-optimization/84419
2871 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2872 with the required type if its current type is compatible but
2875 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2878 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2879 after vectorization.
2881 2018-02-20 Martin Liska <mliska@suse.cz>
2884 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2885 possible values if we don't have a hint.
2887 2018-02-20 Martin Liska <mliska@suse.cz>
2891 * final.c (shorten_branches): Build align_tab array with one
2893 * opts.c (finish_options): Add alignment option limit check.
2894 (MAX_CODE_ALIGN): Likewise.
2895 (MAX_CODE_ALIGN_VALUE): Likewise.
2896 * doc/invoke.texi: Document maximum allowed option value for
2897 all -falign-* options.
2899 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2902 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2903 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2904 * var-tracking.c (emit_note_insn_var_location): Remove all references
2905 to NOTE_INSN_CALL_ARG_LOCATION.
2906 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2907 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2908 Use copy_rtx_if_shared.
2909 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2910 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2911 (dwarf2out_var_location): Remove handling of
2912 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2914 * final.c (final_scan_insn): Remove all references to
2915 NOTE_INSN_CALL_ARG_LOCATION.
2916 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2917 before dumping final insns.
2918 * except.c (emit_note_eh_region_end): Remove all references to
2919 NOTE_INSN_CALL_ARG_LOCATION.
2920 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
2921 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
2922 * config/arc/arc.c (hwloop_optimize): Likewise.
2923 * config/arm/arm.c (create_fix_barrier): Likewise.
2924 * config/s390/s390.c (s390_chunkify_start): Likewise.
2925 * config/sh/sh.c (find_barrier): Likewise.
2926 * config/i386/i386.c (rest_of_insert_endbranch,
2927 ix86_seh_fixup_eh_fallthru): Likewise.
2928 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
2929 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
2930 * config/frv/frv.c (frv_function_prologue): Likewise.
2931 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
2933 (note_outside_basic_block_p): Remove all references to
2934 NOTE_INSN_CALL_ARG_LOCATION.
2935 * gengtype.c (adjust_field_rtx_def): Likewise.
2936 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
2938 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
2939 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
2942 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
2945 PR tree-optimization/84452
2946 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
2947 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
2950 2018-02-19 Martin Liska <mliska@suse.cz>
2953 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
2955 2018-02-19 Martin Liska <mliska@suse.cz>
2956 Richard Sandiford <richard.sandiford@linaro.org>
2958 PR tree-optimization/82491
2959 * gimple-fold.c (get_base_constructor): Make earlier bail out
2962 2018-02-19 Carl Love <cel@us.ibm.com>
2964 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
2965 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
2967 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
2968 P8V_BUILTIN_VEC_NEG.
2970 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
2972 * config/rl78/rl78.md (movdf): New define expand.
2974 2018-02-19 Martin Liska <mliska@suse.cz>
2977 * doc/invoke.texi: Fix typo.
2978 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
2980 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
2982 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2983 handle rs6000_single_float and rs6000_double_float specially for
2986 2018-02-16 Jeff Law <law@redhat.com>
2988 * config/rx/rx.c (add_pop_cfi_notes): New function.;
2991 2018-02-16 Jakub Jelinek <jakub@redhat.com>
2994 * ipa-inline.c (inline_small_functions): Fix a typo.
2996 2018-02-16 Nathan Sidwell <nathan@acm.org>
2998 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3000 2018-02-16 Carl Love <cel@us.ibm.com>
3002 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3003 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3004 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3005 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3006 expansion to P8V_BUILTIN_VEC_FLOAT2.
3008 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3010 PR rtl-optimization/70023
3011 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3012 src_regno into account.
3014 2018-02-16 Carl Love <cel@us.ibm.com>
3016 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3017 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3018 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3019 * config/rs6000/rs6000.c: Remove case statements for
3020 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3021 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3022 and P9V_BUILTIN_VEC_VINSERT4B.
3023 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3024 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3025 * config/rs6000/vsx.md:
3026 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3029 2018-02-16 Carl Love <cel@us.ibm.com>
3031 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3033 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3035 * config/rs6000/rs6000-c.c: Add the definitions for
3036 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3037 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3038 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3039 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3040 definition for insert4b and define insn *insert3b_internal.
3041 * doc/extend.texi: Add documentation for vec_extract4b.
3043 2018-02-16 Nathan Sidwell <nathan@acm.org>
3045 * doc/extend.texi (Backwards Compatibility): Mention friend
3046 injection. Note for-scope is deprecated.
3047 * doc/invoke.texi (-ffriend-injection): Deprecate.
3049 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3051 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3052 that moved to I2, also allow destinations that are a paradoxical
3053 subreg (instead of a normal reg).
3055 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3058 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3061 2018-02-16 Richard Biener <rguenther@suse.de>
3063 PR tree-optimization/84037
3064 PR tree-optimization/84016
3066 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3067 Adjust vec_construct for the fact we need additional higher latency
3068 128bit inserts for AVX256 and AVX512 vector builds.
3069 (ix86_add_stmt_cost): Scale vector construction cost for
3072 2018-02-16 Richard Biener <rguenther@suse.de>
3074 PR tree-optimization/84417
3075 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3076 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3077 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3079 2018-02-16 Martin Liska <mliska@suse.cz>
3082 * internal-fn.def (ASAN_CHECK): Set proper flags.
3083 (ASAN_MARK): Likewise.
3085 2018-02-16 Julia Koval <julia.koval@intel.com>
3087 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3088 from PTA_CANNONLAKE.
3090 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3093 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3094 Use ++iter rather than iter++ for std::list iterators.
3095 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3096 defer deleting them until all nodes in the forest are processed. Do
3097 free even leaf nodes. Change to_process into auto_vec.
3100 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3101 * vec.h (vec_default_construct): Use memset instead of placement new
3102 if BROKEN_VALUE_INITIALIZATION is defined.
3103 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3104 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3107 PR rtl-optimization/83723
3108 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3109 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3110 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3112 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3114 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3116 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3118 PR rtl-optimization/81443
3119 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3120 from inner REGs to paradoxical SUBREGs.
3122 2018-02-16 Richard Biener <rguenther@suse.de>
3124 PR tree-optimization/84399
3125 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3126 For operands we can analyze at their definition make sure we can
3127 analyze them at each use as well.
3129 2018-02-16 Richard Biener <rguenther@suse.de>
3131 PR tree-optimization/84190
3132 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3133 volatile accesses if the decl isn't volatile.
3135 2018-02-15 Jason Merrill <jason@redhat.com>
3137 PR c++/84314 - ICE with templates and fastcall attribute.
3138 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3139 TYPE_CANONICAL on an existing type.
3141 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3143 PR tree-optimization/84383
3144 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3145 dstoff nor call operand_equal_p if dstbase is NULL.
3147 PR tree-optimization/84334
3148 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3149 also a CONSTANT_CLASS_P, punt.
3151 2018-02-14 Jim Wilson <jimw@sifive.com>
3153 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3154 first SMALL_OPERAND check. New local min_second_step. Move assert
3155 to where locals are set. Add TARGET_RVC support.
3156 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3158 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3160 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3162 2018-02-14 Martin Sebor <msebor@redhat.com>
3164 PR tree-optimization/83698
3165 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3166 arrays constrain the offset range to their bounds.
3167 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3168 (builtin_access::overlap): Avoid setting the size of overlap if it's
3170 (maybe_diag_overlap): Also consider arrays when deciding what values
3171 of offsets to include in diagnostics.
3173 2018-02-14 Martin Sebor <msebor@redhat.com>
3176 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3177 that correspond to the kind of a declaration.
3179 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3182 * config/pa/pa.md: Load address of PIC label using the linkage table
3183 if the label is nonlocal.
3185 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3187 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3188 warning message if user requests -maltivec=be.
3189 * doc/invoke.texi: Document deprecation of -maltivec=be.
3191 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3194 * config/rs6000/rs6000-c.c: Update definitions for
3195 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3196 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3198 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3201 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3202 add _get_ssp intrinsics. Remove argument from
3203 __builtin_ia32_rdssp[d|q].
3204 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3205 * config/i386/i386-builtin.def: Remove argument from
3206 __builtin_ia32_rdssp[d|q].
3207 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3208 ix86_expand_special_args_builtin for _rdssp[d|q].
3209 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3210 Clear register before usage.
3211 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3212 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3214 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3216 PR tree-optimization/84357
3217 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3218 operand 1 of an ARRAY_REF too.
3220 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3223 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3224 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3226 (set_of_reg): New struct.
3227 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3228 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3229 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3231 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3232 Split into bitclr, bitset, bitinvert patterns if appropriate.
3233 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3234 use rx_fuse_in_memory_bitop.
3235 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3236 to named insn, correct maximum insn length.
3238 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3241 * machmode.def: Define a complex mode for PARTIAL_INT.
3242 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3244 * doc/rtl.texi: Document CSPImode.
3245 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3247 (msp430_hard_regno_nregs_with_padding): Likewise.
3249 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3252 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3254 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3256 PR rtl-optimization/84169
3257 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3258 we generated a parallel as new i3 and we split that to new i2 and i3
3259 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3260 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3261 those to i2, not i1. Partially rewrite this scan code.
3263 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3266 * stor-layout.c (place_field): For variable length fields, adjust
3267 offset_align afterwards not just based on the field's alignment,
3268 but also on the size.
3271 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3272 of exps and logs in the use_exp2 case.
3274 2018-02-13 Jeff Law <law@redhat.com>
3276 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3279 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3282 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3286 * common.opt (gas-loc-support, gas-locview-support): New.
3287 (ginline-points, ginternal-reset-location-views): New.
3288 * doc/invoke.texi: Document them. Use @itemx where intended.
3289 (gvariable-location-views): Adjust.
3290 * target.def (reset_location_view): New.
3291 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3292 (TARGET_RESET_LOCATION_VIEW): New.
3293 * doc/tm.texi: Rebuilt.
3294 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3295 (dwarf2out_default_as_locview_support): New.
3296 (output_asm_line_debug_info): Use option variables.
3297 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3298 (output_loc_list): Likewise.
3299 (add_high_low_attributes): Check option variables.
3300 Don't output entry view attribute in strict mode.
3301 (gen_inlined_subroutine_die): Check option variables.
3302 (dwarf2out_inline_entry): Likewise.
3303 (init_sections_and_labels): Likewise.
3304 (dwarf2out_early_finish): Likewise.
3305 (maybe_reset_location_view): New, from...
3306 (dwarf2out_var_location): ... here. Call it.
3307 * debug.h (dwarf2out_default_as_loc_support): Declare.
3308 (dwarf2out_default_as_locview_support): Declare.
3309 * hooks.c (hook_int_rtx_insn_0): New.
3310 * hooks.h (hook_int_rtx_insn_0): Declare.
3311 * toplev.c (process_options): Take -gas-loc-support and
3312 -gas-locview-support from dwarf2out. Enable
3313 -gvariable-location-views by default only with locview
3314 assembler support. Enable -ginternal-reset-location-views by
3315 default only if the target defines the corresponding hook.
3316 Enable -ginline-points by default if location views are
3317 enabled; force it disabled if statement frontiers are
3319 * tree-inline.c (expand_call_inline): Check option variables.
3320 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3322 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3324 PR tree-optimization/84321
3325 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3326 handling. Also check whether the anti-range contains any values
3327 that satisfy the mask; switch to a VR_RANGE if not.
3329 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3332 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3334 2018-02-13 Martin Jambor <mjambor@suse.cz>
3337 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3338 of call statements, also set location of a load to a temporary.
3340 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3342 * config/rl78/rl78.c (add_vector_labels): New function.
3343 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3344 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3345 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3346 which checks that no arguments are passed.
3347 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3348 * doc/extend.texi: Documentation for the new attribute.
3350 2018-02-13 Andreas Schwab <schwab@suse.de>
3352 * config/riscv/linux.h (CPP_SPEC): Define.
3354 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3357 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3358 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3359 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3360 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3361 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3362 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3363 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3364 temporarily for AES and PCLMUL builtins.
3366 PR tree-optimization/84339
3367 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3368 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3372 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3373 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3374 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3376 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3378 * omp-simd-clone.h: New file.
3379 * omp-simd-clone.c: Include omp-simd-clone.h.
3380 (expand_simd_clones): No longer static.
3381 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3382 cgraph.h and omp-simd-clone.h.
3383 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3384 (vect_recog_widen_shift_pattern): Formatting fix.
3385 (vect_pattern_recog_1): Don't check optab for calls.
3388 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3389 operands[2] into a REG before using gen_lowpart on it.
3391 2018-02-12 Jeff Law <law@redhat.com>
3394 * config/sh/sh.c (find_barrier): Consider a sibling call
3397 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3398 successfully back substituting a reg.
3400 2018-02-12 Richard Biener <rguenther@suse.de>
3402 PR tree-optimization/84037
3403 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3404 parameter, move visited init to caller.
3405 (vect_slp_analyze_operations): Separate cost from validity
3406 check, initialize visited once for all instances.
3407 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3409 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3411 (vect_model_promotion_demotion_cost): Likewise.
3412 (vectorizable_bswap): Guard cost modeling with !slp_node
3413 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3415 (vectorizable_call): Likewise.
3416 (vectorizable_conversion): Likewise.
3417 (vectorizable_assignment): Likewise.
3418 (vectorizable_shift): Likewise.
3419 (vectorizable_operation): Likewise.
3420 (vectorizable_store): Likewise.
3421 (vectorizable_load): Likewise.
3422 (vectorizable_condition): Likewise.
3423 (vectorizable_comparison): Likewise.
3425 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3428 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3429 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3430 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3432 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3435 * params.def (inline-min-speedup): Increase from 8 to 15.
3436 (max-inline-insns-auto): Decrease from 40 to 30.
3437 * ipa-split.c (consider_split): Add some buffer for function to
3438 be considered inlining candidate.
3439 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3442 2018-02-12 Richard Biener <rguenther@suse.de>
3444 PR tree-optimization/84037
3445 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3446 matched stmts if we cannot swap the non-matched ones.
3448 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3450 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3451 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3452 _mm_maskz_scalef_round_ss): New intrinsics.
3453 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3454 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3455 __builtin_ia32_scalefss_round): Remove.
3456 (__builtin_ia32_scalefsd_mask_round,
3457 __builtin_ia32_scalefss_mask_round): New intrinsics.
3458 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3459 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3460 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3461 "<round_constraint>")): Changed to ...
3462 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3463 "<round_scalar_constraint>")): ... this.
3464 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3465 %0, %1, %2<round_op3>}"): Changed to ...
3466 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3467 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3468 %2<round_scalar_mask_op3>}"): ... this.
3469 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3471 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3473 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3474 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3475 (_mm_maskz_sqrt_round_ss): New intrinsics.
3476 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3477 (__builtin_ia32_sqrtsd_mask_round)
3478 (__builtin_ia32_sqrtss_mask_round): New builtins.
3479 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3480 (__builtin_ia32_sqrtss_round): Remove.
3481 (__builtin_ia32_sqrtsd_mask_round)
3482 (__builtin_ia32_sqrtss_mask_round): New builtins.
3483 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3484 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3485 ((match_operand:VF_128 1 "vector_operand"
3486 "xBm,<round_constraint>")): Changed to ...
3487 ((match_operand:VF_128 1 "vector_operand"
3488 "xBm,<round_scalar_constraint>")): ... this.
3489 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3490 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3491 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3492 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3493 %<iptr>1<round_scalar_mask_op3>}): ... this.
3494 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3495 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3497 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3500 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3501 Cast vec_cmpeq result to correct type.
3502 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3503 Cast vec_cmpgt result to correct type.
3505 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3507 * final.c (final_scan_insn_1): Renamed from...
3508 (final_scan_insn): ... this. New wrapper, to recover
3509 seen from the outermost call in recursive ones.
3510 * config/sparc/sparc.c (output_return): Drop seen from call.
3511 (output_sibcall): Likewise.
3512 * config/visium/visium.c (output_branch): Likewise.
3514 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3516 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3519 2018-02-10 Alan Modra <amodra@gmail.com>
3522 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3523 Specify LR as an input.
3525 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3528 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3529 remove_member_access_dummy_vars): New functions.
3530 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3531 lower_omp_1, execute_lower_omp): Use them.
3533 PR rtl-optimization/84308
3534 * shrink-wrap.c (spread_components): Release todo vector.
3536 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3538 PR rtl-optimization/57193
3539 * ira-color.c (struct allocno_color_data): Add member
3540 conflict_allocno_hard_prefs.
3541 (update_conflict_allocno_hard_prefs): New.
3542 (bucket_allocno_compare_func): Add a preference based on
3543 conflict_allocno_hard_prefs.
3544 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3545 (color_allocnos): Remove a dead code. Initiate
3546 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3548 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3551 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3552 constraint from =wa to wa. Avoid a subreg on the output operand,
3553 instead use a pseudo and subreg it in a move.
3554 (p9_xxbrd_<mode>): Changed to ...
3555 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3556 (p9_xxbrd_v2df): New expander.
3557 (p9_xxbrw_<mode>): Changed to ...
3558 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3559 (p9_xxbrw_v4sf): New expander.
3561 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3563 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3565 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3568 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3569 multiply in 32-bit mode.
3570 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3571 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3574 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3576 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3577 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3578 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3579 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3581 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3584 * dwarf2out.c (is_trivial_indirect_ref): New function.
3585 (dwarf2out_late_global_decl): Do not generate a location
3586 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3587 and that are not defined in the current unit.
3589 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3591 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3592 instead of a libcall for UNORDERED.
3594 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3597 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3598 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3600 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3603 * config/s390/s390.c (s390_set_current_function): Invoke
3604 s390_indirect_branch_settings also if fndecl didn't change.
3606 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3608 * config/rs6000/rs6000.md (blockage): Set length to zero.
3610 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3612 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3614 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3617 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3618 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3622 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3623 PARALLEL incoming that failed vt_get_decl_and_offset check.
3626 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3627 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3629 (get_variable_section): For decls in named .bss* sections pass true as
3630 second argument to bss_initializer_p.
3632 2018-02-09 Marek Polacek <polacek@redhat.com>
3633 Jakub Jelinek <jakub@redhat.com>
3636 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3637 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3638 Sync some changes from cxx_fold_indirect_ref.
3640 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3642 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3644 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3645 (BLOCK_INLINE_ENTRY_LABEL): New.
3646 (dwarf2out_var_location): Disregard inline entry markers.
3647 (inline_entry_data): New struct.
3648 (inline_entry_data_hasher): New hashtable type.
3649 (inline_entry_data_hasher::hash): New.
3650 (inline_entry_data_hasher::equal): New.
3651 (inline_entry_data_table): New variable.
3652 (add_high_low_attributes): Add DW_AT_entry_pc and
3653 DW_AT_GNU_entry_view attributes if a pending entry is found
3654 in inline_entry_data_table. Add old entry_pc attribute only
3655 if debug nonbinding markers are disabled.
3656 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3657 markers are enabled.
3658 (block_within_block_p, dwarf2out_inline_entry): New.
3659 (dwarf2out_finish): Check that no entries remained in
3660 inline_entry_data_table.
3661 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3662 (final_scan_insn, notice_source_line): Likewise.
3663 (rest_of_clean_state): Skip inline entry markers.
3664 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3666 * gimple.c (gimple_build_debug_inline_entry): New.
3667 * gimple.h (enum gimple_debug_subcode): Add
3668 GIMPLE_DEBUG_INLINE_ENTRY.
3669 (gimple_build_debug_inline_entry): Declare.
3670 (gimple_debug_inline_entry_p): New.
3671 (gimple_debug_nonbind_marker_p): Adjust.
3672 * insn-notes.def (INLINE_ENTRY): New.
3673 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3674 inline entry marker notes.
3675 (print_insn): Likewise.
3676 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3677 (INSN_DEBUG_MARKER_KIND): Likewise.
3678 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3679 * tree-inline.c (expand_call_inline): Build and insert
3680 debug_inline_entry stmt.
3681 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3682 inline entry blocks early, if nonbind markers are enabled.
3683 (dump_scope_block): Dump fragment info.
3684 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3685 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3686 (gimple_build_debug_inline_entry): New.
3687 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3688 Enable/disable inline entry points too.
3689 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3690 (DEBUG_INSN): Describe inline entry markers.
3692 * common.opt (gvariable-location-views): New.
3693 (gvariable-location-views=incompat5): New.
3694 * config.in: Rebuilt.
3695 * configure: Rebuilt.
3696 * configure.ac: Test assembler for view support.
3697 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3698 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3699 * dwarf2out.c (var_loc_view): New typedef.
3700 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3701 (dwarf2out_locviews_in_attribute): New.
3702 (dwarf2out_locviews_in_loclist): New.
3703 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3704 (enum dw_line_info_opcode): Add LI_adv_address.
3705 (struct dw_line_info_table): Add view.
3706 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3707 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3708 (zero_view_p): New variable.
3709 (ZERO_VIEW_P): New macro.
3710 (output_asm_line_debug_info): New.
3711 (struct var_loc_node): Add view.
3712 (add_AT_view_list, AT_loc_list): New.
3713 (add_var_loc_to_decl): Add view param. Test it against last.
3714 (new_loc_list): Add view params. Record them.
3715 (AT_loc_list_ptr): Handle loc and view lists.
3716 (view_list_to_loc_list_val_node): New.
3717 (print_dw_val): Handle dw_val_class_view_list.
3718 (size_of_die): Likewise.
3719 (value_format): Likewise.
3720 (loc_list_has_views): New.
3721 (gen_llsym): Set vl_symbol too.
3722 (maybe_gen_llsym, skip_loc_list_entry): New.
3723 (dwarf2out_maybe_output_loclist_view_pair): New.
3724 (output_loc_list): Output view list or entries too.
3725 (output_view_list_offset): New.
3726 (output_die): Handle dw_val_class_view_list.
3727 (output_dwarf_version): New.
3728 (output_compilation_unit_header): Use it.
3729 (output_skeleton_debug_sections): Likewise.
3730 (output_rnglists, output_line_info): Likewise.
3731 (output_pubnames, output_aranges): Update version comments.
3732 (output_one_line_info_table): Output view numbers in asm comments.
3733 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3734 Call maybe_gen_llsym.
3735 (loc_list_from_tree_1): Adjust.
3736 (add_AT_location_description): Create view list attribute if
3737 needed, check it's absent otherwise.
3738 (convert_cfa_to_fb_loc_list): Adjust.
3739 (maybe_emit_file): Call output_asm_line_debug_info for test.
3740 (dwarf2out_var_location): Reset views as needed. Precompute
3741 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3742 attribute. Set view.
3743 (new_line_info_table): Reset next view.
3744 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3745 (dwarf2out_source_line): Likewise. Output view resets and labels to
3746 the assembler, or select appropriate line info opcodes.
3747 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3748 (optimize_string_length): Catch it. Adjust.
3749 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3750 dw_val_class_view_list, and remove it if no longer needed.
3751 (hash_loc_list): Hash view numbers.
3752 (loc_list_hasher::equal): Compare them.
3753 (optimize_location_lists): Check whether a view list symbol is
3754 needed, and whether the locview attribute is present, and
3755 whether they match. Remove the locview attribute if no longer
3757 (index_location_lists): Call skip_loc_list_entry for test.
3758 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3759 Use output_dwarf_version.
3760 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3761 (struct dw_val_node): Add val_view_list.
3762 * final.c (SEEN_NEXT_VIEW): New.
3763 (set_next_view_needed): New.
3764 (clear_next_view_needed): New.
3765 (maybe_output_next_view): New.
3766 (final_start_function): Rename to...
3767 (final_start_function_1): ... this. Take pointer to FIRST,
3768 add SEEN parameter. Emit param bindings in the initial view.
3769 (final_start_function): Reintroduce SEEN-less interface.
3770 (final): Rename to...
3771 (final_1): ... this. Take SEEN parameter. Output final pending
3772 next view at the end.
3773 (final): Reintroduce seen-less interface.
3774 (final_scan_insn): Output pending next view before switching
3775 sections or ending a block. Mark the next view as needed when
3776 outputting variable locations. Notify debug backend of section
3777 changes, and of location view changes.
3778 (rest_of_handle_final): Adjust.
3779 * toplev.c (process_options): Autodetect value for debug variable
3780 location views option. Warn on incompat5 without -gdwarf-5.
3781 * doc/invoke.texi (gvariable-location-views): New.
3782 (gvariable-location-views=incompat5): New.
3783 (gno-variable-location-views): New.
3785 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3787 PR tree-optimization/84136
3788 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3789 that the result of find_edge is non-NULL.
3791 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3794 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3795 storing integer register in SImode. Fix cost of 256 and 512
3796 byte aligned SSE register store.
3798 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3800 * config/i386/i386.c (ix86_multiplication_cost): Fix
3801 multiplication cost for TARGET_AVX512DQ.
3803 2018-02-08 Marek Polacek <polacek@redhat.com>
3805 PR tree-optimization/84238
3806 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3809 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3811 PR tree-optimization/84265
3812 * tree-vect-stmts.c (vectorizable_store): Don't treat
3813 VMAT_CONTIGUOUS accesses as grouped.
3814 (vectorizable_load): Likewise.
3816 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3818 PR tree-optimization/81635
3819 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3820 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3821 (test_round_for_mask): New functions.
3822 (wide_int_cc_tests): Call test_round_for_mask.
3823 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3824 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3825 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3826 range returned by get_range_info.
3828 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3831 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3832 * symtab.c: Include builtins.h
3833 (symtab_node::output_to_lto_symbol_table_p): Move here
3834 from lto-streamer-out.c:output_symbol_p.
3835 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3836 (output_symbol_p): Move all logic to symtab.c
3837 (produce_symtab): Update.
3839 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3841 * config/s390/s390-opts.h (enum indirect_branch): Define.
3842 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3843 (s390_indirect_branch_via_thunk)
3844 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3845 (enum s390_indirect_branch_type): Define.
3846 * config/s390/s390.c (struct s390_frame_layout, struct
3847 machine_function): Remove.
3848 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3849 (indirect_branch_table_label_no, indirect_branch_table_name):
3851 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3852 (enum s390_indirect_branch_option): Define.
3853 (s390_return_addr_from_memory): New function.
3854 (s390_handle_string_attribute): New function.
3855 (s390_attribute_table): Add new attribute handler.
3856 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3857 (s390_indirect_branch_via_thunk): New function.
3858 (s390_indirect_branch_via_inline_thunk): New function.
3859 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3860 sibling call optimization for non z10 compiles.
3861 (s390_emit_call): Force indirect branch target to be a single
3862 register. Add r1 clobber for non-z10 compiles.
3863 (s390_emit_epilogue): Emit return jump via return_use expander.
3864 (s390_reorg): Handle JUMP_INSNs as execute targets.
3865 (s390_option_override_internal): Perform validity checks for the
3866 new command line options.
3867 (s390_indirect_branch_attrvalue): New function.
3868 (s390_indirect_branch_settings): New function.
3869 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3870 (s390_output_indirect_thunk_function): New function.
3871 (s390_code_end): Implement target hook.
3872 (s390_case_values_threshold): Implement target hook.
3873 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3875 * config/s390/s390.h (struct s390_frame_layout)
3876 (struct machine_function): Move here from s390.c.
3877 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3878 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3879 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3880 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3881 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3882 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3883 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3884 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3885 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3886 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3887 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3888 (mnemonic attribute): Add values which aren't recognized
3890 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3891 pattern for branch conversion. Fix mnemonic attribute.
3892 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3893 indirect branch via thunk if requested.
3894 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3895 ("*indirect_jump"): Disable for branch conversion using out of
3897 ("indirect_jump_via_thunk<mode>_z10")
3898 ("indirect_jump_via_thunk<mode>")
3899 ("indirect_jump_via_inlinethunk<mode>_z10")
3900 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3901 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3902 ("casesi_jump_via_inlinethunk<mode>_z10")
3903 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3904 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3905 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3906 ("*indirect2_jump"): Disable for branch conversion.
3907 ("casesi_jump"): Turn into expander and expand patterns for branch
3909 ("return_use"): New expander.
3910 ("*return"): Emit return via thunk and rename it to ...
3911 ("*return<mode>"): ... this one.
3912 * config/s390/s390.opt: Add new options and and enum for the
3915 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3917 * lra-constraints.c (match_reload): Unconditionally use
3918 gen_lowpart_SUBREG, rather than selecting between that
3919 and equivalent gen_rtx_SUBREG code.
3921 2018-02-08 Richard Biener <rguenther@suse.de>
3923 PR tree-optimization/84233
3924 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
3925 changed flag instead of boguously re-using phi_inserted.
3927 2018-02-08 Martin Jambor <mjambor@suse.cz>
3929 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
3930 static local variables.
3932 2018-02-08 Richard Biener <rguenther@suse.de>
3934 PR tree-optimization/84278
3935 * tree-vect-stmts.c (vectorizable_store): When looking for
3936 smaller vector types to perform grouped strided loads/stores
3937 make sure the mode is supported by the target.
3938 (vectorizable_load): Likewise.
3940 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3942 * config/aarch64/aarch64.c (aarch64_components_for_bb):
3943 Increase LDP/STP opportunities by adding adjacent callee-saves.
3945 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
3947 PR rtl-optimization/84068
3948 PR rtl-optimization/83459
3949 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
3951 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
3953 PR tree-optimization/84224
3954 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
3955 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
3958 2018-02-07 Iain Sandoe <iain@codesourcery.com>
3961 * config/rs6000/altivec.md (*restore_world): Remove LR use.
3962 * config/rs6000/predicates.md (restore_world_operation): Adjust op
3963 count, remove one USE.
3965 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
3967 * doc/install.texi (Configuration): Document the
3968 --with-long-double-format={ibm,ieee} PowerPC configuration
3972 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
3973 Convert from define_expand to be define_insn_and_split. Rework
3974 float/double/_Float128 conversions to QI/HI/SImode to work with
3975 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
3976 conversions to QI/HImode types did a store and then a load to
3977 truncate the value. For conversions to VSX registers, don't split
3978 the insn, instead emit the code directly. Use the code iterator
3979 any_fix to combine signed and unsigned conversions.
3980 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
3981 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
3982 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3983 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
3984 (fix_<mode>di2_hw): Likewise.
3985 (fixuns_<mode>di2_hw): Likewise.
3986 (fix_<mode>si2_hw): Likewise.
3987 (fixuns_<mode>si2_hw): Likewise.
3988 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
3989 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
3990 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
3991 fix<uns>_trunc<SFDF:mode>si2_p8.
3992 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
3994 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
3995 (fix<uns>_<mode>_mem): Likewise.
3996 (fctiw<u>z_<mode>_mem): Likewise.
3997 (fix<uns>_<mode>_mem): Likewise.
3998 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
3999 the register allocator from doing a direct move to the GPRs to do
4000 a store, and instead use the ISA 3.0 store byte/half-word from
4001 vector register instruction. For IEEE 128-bit floating point,
4002 also optimize stores of 32-bit ints.
4003 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4005 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4007 * genextract.c (push_pathstr_operand): New function to support
4009 (walk_rtx): Call push_pathstr_operand.
4010 (print_path): Support [a-zA-Z].
4012 2018-02-07 Richard Biener <rguenther@suse.de>
4014 PR tree-optimization/84037
4015 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4016 (cse_and_gimplify_to_preheader): Declare.
4017 (vect_get_place_in_interleaving_chain): Likewise.
4018 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4020 (_loop_vec_info::~_loop_vec_info): Delete it.
4021 (cse_and_gimplify_to_preheader): New function.
4022 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4023 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4024 (vectorizable_load): Likewise. For grouped stores always base
4025 the IV on the first element.
4026 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4027 condition before gimplifying.
4029 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4031 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4032 *DIV_EXPR and *MOD_EXPR.
4034 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4037 * config/i386/i386.c (ix86_option_override_internal): Mask out
4038 the CF_SET bit when checking -fcf-protection.
4040 2018-02-07 Tom de Vries <tom@codesourcery.com>
4043 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4046 2018-02-07 Richard Biener <rguenther@suse.de>
4048 PR tree-optimization/84204
4049 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4052 PR tree-optimization/84205
4053 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4054 special-case isl_ast_op_zdiv_r.
4056 PR tree-optimization/84223
4057 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4058 Only add conditions from within the region.
4059 (gather_bbs::after_dom_children): Adjust.
4061 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4064 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4065 * config/avr/avr.md: Only post-reload split REG-REG moves if
4066 either register is GENERAL_REG_P.
4068 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4070 PR tree-optimization/84235
4071 * tree-ssa-scopedtables.c
4072 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4073 if the subtraction is performed in floating point type where NaNs are
4074 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4075 build 1. Formatting fix.
4077 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4080 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4081 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4082 and skip it regardless of bb boundaries. Use CALL_P macro,
4083 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4084 unnecessarily, formatting fix.
4086 2018-02-06 Michael Collison <michael.collison@arm.com>
4088 * config/arm/thumb2.md:
4089 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4090 (*thumb_mov_notscc): Ditto.
4092 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4095 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4096 unsigned_fix, not "s".
4098 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4100 * configure.ac (gcc_fn_eh_frame_ro): New function.
4101 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4102 correct .eh_frame permissions.
4103 * configure: Regenerate.
4105 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4107 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4110 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4112 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4113 Display warning message for -mno-speculate-indirect-jumps.
4115 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4117 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4119 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4121 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4123 PR tree-optimization/84225
4124 * tree-eh.c (find_trapping_overflow): Only call
4125 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4127 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4130 * config/i386/i386.c: Reimplement the check of possible options
4131 -mibt/-mshstk conbination. Change error messages.
4132 * doc/invoke.texi: Fix a typo: remove extra '='.
4134 2018-02-06 Marek Polacek <polacek@redhat.com>
4136 PR tree-optimization/84228
4137 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4139 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4142 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4143 emitted arch directives.
4144 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4145 __ARM_FEATURE_COPROC before changing architectures.
4147 2018-02-06 Richard Biener <rguenther@suse.de>
4149 * config/i386/i386.c (print_reg): Fix typo.
4150 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4152 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4154 * configure: Regenerate.
4156 2018-02-05 Martin Sebor <msebor@redhat.com>
4158 PR tree-optimization/83369
4159 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4162 2018-02-05 Martin Liska <mliska@suse.cz>
4164 * doc/invoke.texi: Cherry-pick upstream r323995.
4166 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4168 * ira.c (ira_init_register_move_cost): Adjust comment.
4170 2018-02-05 Martin Liska <mliska@suse.cz>
4172 PR gcov-profile/84137
4173 * doc/gcov.texi: Fix typo in documentation.
4175 2018-02-05 Martin Liska <mliska@suse.cz>
4177 PR gcov-profile/83879
4178 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4179 using dlopen functionality.
4181 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4183 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4184 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4185 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4186 _mm_maskz_range_round_ss): New intrinsics.
4187 (__builtin_ia32_rangesd128_round)
4188 (__builtin_ia32_rangess128_round): Remove.
4189 (__builtin_ia32_rangesd128_mask_round,
4190 __builtin_ia32_rangess128_mask_round): New builtins.
4191 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4192 __builtin_ia32_rangess128_round): Remove.
4193 (__builtin_ia32_rangesd128_mask_round,
4194 __builtin_ia32_rangess128_mask_round): New builtins.
4195 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4196 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4197 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4198 "<round_saeonly_constraint>")): Changed to ...
4199 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4200 "<round_saeonly_scalar_constraint>")): ... this.
4201 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4202 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4203 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4204 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4205 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4207 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4209 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4211 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4212 Remove all values except native, 8540 and 8548.
4214 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4216 * config/i386/i386.c (ix86_output_function_return): Pass
4217 INVALID_REGNUM, instead of -1, as invalid register number to
4218 indirect_thunk_name and output_indirect_thunk.
4220 2018-02-02 Julia Koval <julia.koval@intel.com>
4222 * config.gcc: Add -march=icelake.
4223 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4224 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4225 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4226 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4227 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4228 (processor_target_table): Add icelake.
4229 (ix86_option_override_internal): Handle new PTAs.
4230 (get_builtin_code_for_version): Handle icelake.
4231 (M_INTEL_COREI7_ICELAKE): New.
4232 (fold_builtin_cpu): Handle icelake.
4233 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4234 * doc/invoke.texi: Add -march=icelake.
4236 2018-02-02 Julia Koval <julia.koval@intel.com>
4238 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4239 to wide_int_bitmask.
4240 * wide-int-bitmask.h: New.
4242 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4245 * config/i386/i386.md: Replace Pmode with word_mode in
4246 builtin_setjmp_setup and builtin_longjmp to support x32.
4248 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4252 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4254 (rs6000_supported_cpu_names): New static variable.
4255 (linux_cpu_translation_table): Likewise.
4256 (elf_platform) <cpu>: Define new static variable and use it.
4257 Translate kernel AT_PLATFORM name to canonical name if needed.
4258 Error if platform name is unknown.
4260 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4263 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4265 2018-02-01 Jeff Law <law@redhat.com>
4268 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4269 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4270 the scratch if RELEASE_VIA_POP is false.
4271 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4272 If we have to save a temporary register, decrement SIZE appropriately.
4273 Pass new arguments to release_scratch_register_on_entry.
4274 (ix86_adjust_stack_and_probe): Likewise.
4275 (ix86_emit_probe_stack_range): Pass new arguments to
4276 release_scratch_register_on_entry.
4278 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4280 PR rtl-optimization/84157
4281 * combine.c (change_zero_ext): Use REG_P predicate in
4282 front of HARD_REGISTER_P predicate.
4284 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4286 * config/avr/avr.c (avr_option_override): Move disabling of
4287 -fdelete-null-pointer-checks to...
4288 * common/config/avr/avr-common.c (avr_option_optimization_table):
4291 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4293 PR tree-optimization/81635
4294 * tree-data-ref.c (split_constant_offset_1): For types that
4295 wrap on overflow, try to use range info to prove that wrapping
4298 2018-02-01 Renlin Li <renlin.li@arm.com>
4301 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4303 (aarch64_register_move_cost): Likewise.
4304 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4306 (REG_CLASS_NAMES): Likewise.
4307 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4308 TAILCALL_ADDR_REGS. Remove IP registers.
4309 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4311 2018-02-01 Richard Biener <rguenther@suse.de>
4313 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4314 for specifying RPO order and allow NULL for that.
4315 * domwalk.c (dom_walker::dom_walker): Likewise.
4316 (dom_walker::walk): Handle NULL RPO order.
4317 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4319 (rewrite_update_dom_walker): Likewise.
4320 (mark_def_dom_walker): Likewise.
4322 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4324 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4325 (aarch64_maybe_expand_sve_subreg_move): Declare.
4326 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4327 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4329 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4330 that are semantically a reverse operation.
4331 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4332 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4333 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4335 (aarch64_can_change_mode_class): For big-endian, forbid changes
4336 between two SVE modes if they have different element sizes.
4338 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4340 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4341 the TImode handling for big-endian targets.
4343 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4345 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4346 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4348 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4349 Remove BSWAP handing for big-endian targets and use the form of
4350 LD1RQ appropariate for the mode.
4352 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4354 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4355 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4358 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4361 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4362 check for operands that need to go through aarch64_sve_reload_be.
4364 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4366 PR tree-optimization/81661
4367 PR tree-optimization/84117
4368 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4369 * tree-eh.c: Include gimplify.h.
4370 (find_trapping_overflow, replace_trapping_overflow,
4371 rewrite_to_non_trapping_overflow): New functions.
4372 * tree-vect-loop.c: Include tree-eh.h.
4373 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4374 * tree-data-ref.c: Include tree-eh.h.
4375 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4377 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4379 PR rtl-optimization/84123
4380 * combine.c (change_zero_ext): Check if hard register satisfies
4381 can_change_dest_mode before calling gen_lowpart_SUBREG.
4383 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4386 * ira.c (ira_init_register_move_cost): Remove assert.
4388 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4390 PR rtl-optimization/84071
4391 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4392 * doc/tm.texi: Regenerate.
4394 2018-01-31 Richard Biener <rguenther@suse.de>
4396 PR tree-optimization/84132
4397 * tree-data-ref.c (analyze_miv_subscript): Properly
4398 check whether evolution_function_is_affine_multivariate_p
4399 before calling gcd_of_steps_may_divide_p.
4401 2018-01-31 Julia Koval <julia.koval@intel.com>
4404 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4405 * config/i386/i386.md (rdpid_rex64) New.
4406 (rdpid): Make 32bit only.
4408 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4411 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4412 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4414 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4417 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4419 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4421 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4423 PR rtl-optimization/84071
4424 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4425 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4427 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4429 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4430 (arc_attribute_table): Add 'aux' attribute.
4431 (arc_in_small_data_p): Consider aux like variables.
4432 (arc_is_aux_reg_p): New function.
4433 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4434 (arc_get_aux_arg): New function.
4435 (prepare_move_operands): Handle aux-register access.
4436 (arc_handle_aux_attribute): New function.
4437 * doc/extend.texi (ARC Variable attributes): Add subsection.
4439 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4441 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4442 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4443 (arc_attribute_table): Add 'uncached' attribute.
4444 (arc_print_operand): Print '.di' flag for uncached memory
4446 (arc_in_small_data_p): Do not consider for small data the uncached
4448 (arc_is_uncached_mem_p): New function.
4449 * config/arc/predicates.md (compact_store_memory_operand): Check
4450 for uncached memory accesses.
4451 (nonvol_nonimm_operand): Likewise.
4452 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4454 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4457 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4458 falign-loops=): Add Optimization flag.
4460 2018-01-30 Jeff Law <law@redhat.com>
4463 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4464 INT_REGISTERS_SAVED. Check it prior to calling
4465 get_scratch_register_on_entry.
4466 (ix86_adjust_stack_and_probe): Similarly.
4467 (ix86_emit_probe_stack_range): Similarly.
4468 (ix86_expand_prologue): Corresponding changes.
4470 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4473 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4474 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4476 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4479 * lra-constraints.c (curr_insn_transform): Process AND in the
4482 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4484 PR rtl-optimization/83986
4485 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4486 dependence against last_pending_memory_flush in addition to
4489 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4491 PR tree-optimization/81611
4492 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4495 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4498 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4501 2018-01-30 Richard Biener <rguenther@suse.de>
4502 Jakub Jelinek <jakub@redhat.com>
4504 PR tree-optimization/84111
4505 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4506 inner loops added during recursion, as they don't have up-to-date
4509 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4512 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4513 (can_inline_edge_by_limits_p): ... here.
4514 (can_early_inline_edge_p, check_callers,
4515 update_caller_keys, update_callee_keys, recursive_inlining,
4516 add_new_edges_to_heap, speculation_useful_p,
4517 inline_small_functions,
4518 inline_small_functions, flatten_function,
4519 inline_to_all_callers_1): Update.
4521 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4523 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4526 2018-01-30 Richard Biener <rguenther@suse.de>
4528 PR tree-optimization/83008
4529 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4530 invariant and constant vector uses in stmts when they need
4533 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4536 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4537 * configure: Regenerate.
4539 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4541 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4543 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4544 Use gen_rtx_REG rather than gen_lowpart.
4546 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4548 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4549 rather than 0 when creating partial subregs.
4551 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4553 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4556 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4559 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4560 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4561 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4562 flags. This restores the settings used before the 2017-07-24.
4563 Turning off pre increment/decrement/modify allows IVOPTS to
4564 optimize DF/SF loops where the index is an int.
4566 2018-01-29 Richard Biener <rguenther@suse.de>
4567 Kelvin Nilsen <kelvin@gcc.gnu.org>
4570 * tree-vect-stmts.c (vectorizable_call): Don't call
4571 targetm.vectorize_builtin_md_vectorized_function if callee is
4574 2018-01-22 Carl Love <cel@us.ibm.com>
4576 * doc/extend.tex: Fix typo in second arg in
4577 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4579 2018-01-29 Richard Biener <rguenther@suse.de>
4581 PR tree-optimization/84086
4582 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4583 (flush_ssaname_freelist): When SSA names were released reset
4584 the SCEV hash table.
4586 2018-01-29 Richard Biener <rguenther@suse.de>
4588 PR tree-optimization/84057
4589 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4590 removed paths when removing edges.
4592 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4594 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4595 -mfunction-return=@var{choice}.
4597 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4600 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4602 (layout::print_source_line): Likewise.
4603 (test_get_line_width_without_trailing_whitespace): Add test cases.
4605 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4608 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4611 2018-01-26 Jim Wilson <jimw@sifive.com>
4613 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4615 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4618 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4620 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4621 and CMP + SUB-immediate -> SUBS.
4623 2018-01-26 Martin Sebor <msebor@redhat.com>
4625 PR tree-optimization/83896
4626 * tree-ssa-strlen.c (get_string_len): Rename...
4627 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4628 Avoid assuming length is constant.
4629 (handle_char_store): Use HOST_WIDE_INT for string length.
4631 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4634 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4635 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4637 2018-01-26 Richard Biener <rguenther@suse.de>
4639 PR rtl-optimization/84003
4640 * dse.c (record_store): Only record redundant stores when
4641 the earlier store aliases at least all accesses the later one does.
4643 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4645 PR rtl-optimization/83985
4646 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4647 REG_CFA_RESTORE insns.
4648 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4649 REG_CFA_RESTORE insns here.
4652 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4653 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4655 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4657 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4658 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4659 (arc_init): Likewise.
4660 (arc_override_options): Likewise.
4661 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4663 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4665 * config/arc/arc.h (TARGET_DBNZ): Define.
4666 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4667 properly set the tune attribute.
4668 (dbnz): Use TARGET_DBNZ guard.
4669 * config/arc/arc.opt (mtune): Add core3 option.
4671 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4673 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4674 recognize new pic like addresses.
4675 (arc_delegitimize_address): Clean up.
4677 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4679 * config/arc/arc-arches.def: Option mrf16 valid for all
4681 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4682 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4683 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4684 * config/arc/arc-tables.opt: Regenerate.
4685 * config/arc/arc.c (arc_conditional_register_usage): Handle
4686 reduced register file case.
4687 (arc_file_start): Set must have build attributes.
4688 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4690 * config/arc/arc.opt (mrf16): Add new option.
4691 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4692 * config/arc/genmultilib.awk: Handle new mrf16 option.
4693 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4694 * config/arc/t-multilib: Regenerate.
4695 * doc/invoke.texi (ARC Options): Document mrf16 option.
4697 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4699 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4700 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4701 (arc_attribute_table): Add 'secure_call' attribute.
4702 (arc_print_operand): Print secure call operand.
4703 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4705 (arc_is_secure_call_p): New function. * config/arc/arc.md
4706 (call_i): Add support for sjli instruction.
4707 (call_value_i): Likewise.
4708 * config/arc/constraints.md (Csc): New constraint.
4710 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4711 John Eric Martin <John.Martin@emmicro-us.com>
4713 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4714 * config/arc/arc.c (_arc_jli_section): New struct.
4715 (arc_jli_section): New type.
4716 (rc_jli_sections): New static variable.
4717 (arc_handle_jli_attribute): New function.
4718 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4719 (arc_file_end): New function.
4720 (TARGET_ASM_FILE_END): Define.
4721 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4722 (arc_add_jli_section): New function.
4723 (jli_call_scan): Likewise.
4724 (arc_reorg): Call jli_call_scan.
4725 (arc_output_addsi): Remove 'S' from printing asm operand.
4726 (arc_is_jli_call_p): New function.
4727 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4729 (movhi_insn): Likewise.
4730 (movsi_insn): Likewise.
4731 (movsi_set_cc_insn): Likewise.
4732 (loadqi_update): Likewise.
4733 (load_zeroextendqisi_update): Likewise.
4734 (load_signextendqisi_update): Likewise.
4735 (loadhi_update): Likewise.
4736 (load_zeroextendhisi_update): Likewise.
4737 (load_signextendhisi_update): Likewise.
4738 (loadsi_update): Likewise.
4739 (loadsf_update): Likewise.
4740 (movsicc_insn): Likewise.
4741 (bset_insn): Likewise.
4742 (bxor_insn): Likewise.
4743 (bclr_insn): Likewise.
4744 (bmsk_insn): Likewise.
4745 (bicsi3_insn): Likewise.
4746 (cmpsi_cc_c_insn): Likewise.
4747 (movsi_ne): Likewise.
4748 (movsi_cond_exec): Likewise.
4749 (clrsbsi2): Likewise.
4760 (call_i): Remove 'S' asm letter, add jli instruction.
4761 (call_value_i): Likewise.
4762 * config/arc/arc.op (mjli-always): New option.
4763 * config/arc/constraints.md (Cji): New constraint.
4764 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4766 (subsf3_fpx): Likewise.
4767 (mulsf3_fpx): Likewise.
4768 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4770 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4772 * doc/invoke.texi (ARC): Document mjli-always option.
4774 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4776 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4777 avoid addition with 0 and use incw and decw where possible.
4779 2018-01-26 Richard Biener <rguenther@suse.de>
4781 PR tree-optimization/81082
4782 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4783 association if it requires casting to unsigned.
4784 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4785 from fold_plusminus_mult_expr to catch important cases late when
4786 range info is available.
4788 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4790 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4791 * configure.ac (hidden_linkonce): New test.
4792 * configure: Regenerate.
4793 * config.in: Regenerate.
4795 2018-01-26 Julia Koval <julia.koval@intel.com>
4797 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4798 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4799 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4800 _mm_mask_bitshuffle_epi64_mask): Fix type.
4801 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4802 USI_FTYPE_V4DI_V4DI_USI): Remove.
4803 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4804 __builtin_ia32_vpshufbitqmb256_mask,
4805 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4806 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4807 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4809 2018-01-26 Alan Modra <amodra@gmail.com>
4812 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4813 UNSPEC_VBPERMQ. Sort other unspecs.
4815 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4817 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4819 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4822 * predict.c (drop_profile): Do not push/pop cfun; update also
4824 (handle_missing_profiles): Fix logic looking for zero profiles.
4826 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4829 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4830 on functions with #pragma omp declare simd or functions with simd
4832 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4833 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4834 Remove trailing \n from warning_at calls.
4836 2018-01-25 Tom de Vries <tom@codesourcery.com>
4839 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4840 for neutered workers.
4842 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4845 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4846 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4848 2018-01-24 Jeff Law <law@redhat.com>
4851 * i386.c (get_probe_interval): Move to earlier point.
4852 (ix86_compute_frame_layout): If -fstack-clash-protection and
4853 the frame is larger than the probe interval, then use pushes
4854 to save registers rather than reg->mem moves.
4855 (ix86_expand_prologue): Remove conditional for int_registers_saved
4858 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4861 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4862 min/max for never referenced object.
4864 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4867 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4869 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4870 attributes from DECL_ATTRIBUTES (decl) without affecting
4871 DECL_ATTRIBUTES (current_function_decl).
4872 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4873 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4875 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4877 PR tree-optimization/83979
4878 * fold-const.c (fold_comparison): Use constant_boolean_node
4879 instead of boolean_{true,false}_node.
4881 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4883 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4886 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4888 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4889 Simplify the clause that sets the length attribute.
4890 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4891 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4892 clause that sets the length attribute.
4893 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4895 2018-01-24 Tom de Vries <tom@codesourcery.com>
4898 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4899 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4900 Add strict parameter.
4901 (prevent_branch_around_nothing): Insert dummy insn between branch to
4902 label and label with no ptx insn inbetween.
4903 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4905 2018-01-24 Tom de Vries <tom@codesourcery.com>
4908 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4909 for neutered threads in warp.
4910 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4912 2018-01-24 Richard Biener <rguenther@suse.de>
4914 PR tree-optimization/83176
4915 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
4918 2018-01-24 Richard Biener <rguenther@suse.de>
4920 PR tree-optimization/82819
4921 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
4922 code generating pluses that are no-ops in the target precision.
4924 2018-01-24 Richard Biener <rguenther@suse.de>
4927 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
4929 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
4931 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
4932 to merge probabilities.
4933 * predict.c (probably_never_executed): Also mark as cold functions
4934 with global 0 profile and guessed local profile.
4935 * profile-count.c (profile_probability::combine_with_count): New
4937 * profile-count.h (profile_probability::operator*,
4938 profile_probability::operator*=, profile_probability::operator/,
4939 profile_probability::operator/=): Reduce precision to adjusted
4940 and set value to guessed on contradictory divisions.
4941 (profile_probability::combine_with_freq): Remove.
4942 (profile_probability::combine_wiht_count): Declare.
4943 (profile_count::force_nonzero):: Set to adjusted.
4944 (profile_count::probability_in):: Set quality to adjusted.
4945 * tree-ssa-tail-merge.c (replace_block_by): Use
4948 2018-01-23 Andrew Waterman <andrew@sifive.com>
4949 Jim Wilson <jimw@sifive.com>
4951 * config/riscv/riscv.c (riscv_stack_boundary): New.
4952 (riscv_option_override): Set riscv_stack_boundary. Handle
4953 riscv_preferred_stack_boundary_arg.
4954 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
4955 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
4956 (STACK_BOUNDARY): Set to riscv_stack_boundary.
4957 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
4958 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
4959 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
4961 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
4964 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
4965 of struct ix86_frame.
4966 (ix86_expand_epilogue): Likewise. Add a local variable for
4967 the reg_save_offset field in struct ix86_frame.
4969 2018-01-23 Bin Cheng <bin.cheng@arm.com>
4971 PR tree-optimization/82604
4972 * tree-loop-distribution.c (enum partition_kind): New enum item
4973 PKIND_PARTIAL_MEMSET.
4974 (partition_builtin_p): Support above new enum item.
4975 (generate_code_for_partition): Ditto.
4976 (compute_access_range): Differentiate cases that equality can be
4977 proven at all loops, the innermost loops or no loops.
4978 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
4979 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
4980 (finalize_partitions, distribute_loop): Don't fuse partition of
4981 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
4982 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
4985 2018-01-23 Martin Liska <mliska@suse.cz>
4987 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
4988 order to ignore the predictor.
4989 (PRED_POLYMORPHIC_CALL): Likewise.
4990 (PRED_RECURSIVE_CALL): Likewise.
4992 2018-01-23 Martin Liska <mliska@suse.cz>
4994 * tree-profile.c (tree_profiling): Print function header to
4995 aware reader which function we are working on.
4996 * value-prof.c (gimple_find_values_to_profile): Do not print
4997 not interesting value histograms.
4999 2018-01-23 Martin Liska <mliska@suse.cz>
5001 * profile-count.h (enum profile_quality): Add
5002 profile_uninitialized as the first value. Do not number values
5003 as they are zero based.
5004 (profile_count::verify): Update sanity check.
5005 (profile_probability::verify): Likewise.
5007 2018-01-23 Nathan Sidwell <nathan@acm.org>
5009 * doc/invoke.texi (ffor-scope): Deprecate.
5011 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5013 PR tree-optimization/83510
5014 * domwalk.c (set_all_edges_as_executable): New function.
5015 (dom_walker::dom_walker): Convert bool param
5016 "skip_unreachable_blocks" to enum reachability. Move setup of
5017 edge flags to set_all_edges_as_executable and only do it when
5018 reachability is REACHABLE_BLOCKS.
5019 * domwalk.h (enum dom_walker::reachability): New enum.
5020 (dom_walker::dom_walker): Convert bool param
5021 "skip_unreachable_blocks" to enum reachability.
5022 (set_all_edges_as_executable): New decl.
5023 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5024 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5026 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5027 but converting true to REACHABLE_BLOCKS.
5028 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5030 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5031 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5032 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5034 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5035 if check_all_array_refs will be called.
5037 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5039 * tree.c (selftest::test_location_wrappers): Add more test
5042 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5044 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5045 (selftest::test_bit_in_range): Likewise.
5047 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5050 * doc/sourcebuild.texi (vect_float): Say that the selector
5051 only describes the situation when -funsafe-math-optimizations is on.
5052 (vect_float_strict): Document.
5054 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5056 PR tree-optimization/83965
5057 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5058 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5059 instead of checking only for a reduction.
5060 (vect_recog_widen_sum_pattern): Likewise.
5062 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5064 * predict.c (probably_never_executed): Only use precise profile info.
5065 (compute_function_frequency): Skip after inlining hack since we now
5066 have quality checking.
5068 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5070 * profile-count.h (profile_probability::very_unlikely,
5071 profile_probability::unlikely, profile_probability::even): Set
5072 precision to guessed.
5074 2018-01-23 Richard Biener <rguenther@suse.de>
5076 PR tree-optimization/83963
5077 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5078 Properly terminate dominator walk when crossing the exit edge not
5079 when visiting its source block.
5081 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5084 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5085 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5087 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5089 PR tree-optimization/83957
5090 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5091 semicolon after for body surrounded by braces.
5093 PR tree-optimization/83081
5094 * profile-count.h (profile_probability::split): New method.
5095 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5096 Use profile_probability::split.
5097 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5098 when splitting a single conditional jump into 2.
5100 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5102 PR tree-optimization/69452
5103 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5106 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5108 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5109 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5110 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5112 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5114 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5115 * config/rl78/rl78.md (movdi): New define_expand.
5116 * config/rl78/rl78.c (rl78_split_movdi): New function.
5118 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5121 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5123 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5124 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5125 128-bit to produce an UNSPEC move to get the double word with the
5126 signbit and then a shift directly to do signbit.
5127 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5128 implementation with a new version that just does either a direct
5129 move or a regular move. Move memory interface to separate insns.
5130 Move insns so they are next to the expander.
5131 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5132 with signbit move. Split big and little endian case.
5133 (signbit<mode>2_dm_mem_le): Likewise.
5134 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5135 (signbit<mode>2_dm2): Likewise.
5137 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5139 * config/rl78/rl78.md (anddi3): New define_expand.
5141 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5143 * config/rl78/rl78.md (umindi3): New define_expand.
5145 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5147 * config/rl78/rl78.md (smindi3): New define_expand.
5149 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5151 * config/rl78/rl78.md (smaxdi3): New define_expand.
5153 2018-01-22 Carl Love <cel@us.ibm.com>
5155 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5156 LVX_V1TI): Add macro expansion.
5157 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5158 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5159 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5160 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5161 Change check to determine if the instruction is a byte reversing
5162 entry. Fix typo in comment.
5163 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5164 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5165 Add def_builtin calls for new builtins.
5166 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5167 Add define_insn expansion.
5169 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5171 * config/rl78/rl78.md (umaxdi3): New define_expand.
5173 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5175 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5176 for non-QImode registers.
5178 2018-01-22 Richard Biener <rguenther@suse.de>
5180 PR tree-optimization/83963
5181 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5182 including the loop exit block.
5183 (scop_detection::merge_sese): Likewise.
5184 (scop_detection::add_scop): Do it here instead.
5186 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5188 * doc/sourcebuild.texi (arm_softfloat): Document.
5190 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5193 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5194 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5195 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5197 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5198 David Edelsohn <dje.gcc@gmail.com>
5201 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5202 Change "crset eq" to "crset 2".
5203 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5204 (*call_indirect_aix<mode>_nospec): Likewise.
5205 (*call_value_indirect_aix<mode>_nospec): Likewise.
5206 (*call_indirect_elfv2<mode>_nospec): Likewise.
5207 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5208 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5209 change assembly output from . to $.
5210 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5211 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5212 (*tablejump<mode>_internal1_nospec): Likewise.
5214 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5217 * config/sh/sh_optimize_sett_clrt.cc:
5218 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5220 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5222 PR tree-optimization/83940
5223 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5224 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5225 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5226 use it to pass back the definition type.
5227 (vect_check_store_rhs): Likewise rhs_dt_out.
5228 (vect_build_gather_load_calls): Add a mask_dt argument and use
5229 it instead of a call to vect_is_simple_use.
5230 (vectorizable_store): Update calls to vect_check_load_store_mask
5231 and vect_check_store_rhs. Use the dt returned by the latter instead
5232 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5233 instead of calls to vect_is_simple_use. Pass the scalar rather
5234 than the vector operand to vect_is_simple_use when handling
5235 second and subsequent copies of an rhs value.
5236 (vectorizable_load): Update calls to vect_check_load_store_mask
5237 and vect_build_gather_load_calls. Use the cached mask_dt and
5238 gs_info.offset_dt instead of calls to vect_is_simple_use.
5240 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5243 * tree-emutls.c: Include gimplify.h.
5244 (lower_emutls_2): New function.
5245 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5246 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5247 it before further processing.
5250 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5251 UINTVAL (trueop1) instead of INTVAL (op1).
5253 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5257 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5258 INCOMING_FRAME_SP_OFFSET if not defined.
5259 (scan_trace): Add ENTRY argument. If true and
5260 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5261 emit a note to adjust the CFA offset.
5262 (create_cfi_notes): Adjust scan_trace callers.
5263 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5264 INCOMING_FRAME_SP_OFFSET in the CIE.
5265 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5266 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5268 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5269 * doc/tm.texi: Regenerated.
5271 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5273 PR rtl-optimization/83147
5274 * lra-constraints.c (remove_inheritance_pseudos): Use
5275 lra_substitute_pseudo_within_insn.
5277 2018-01-19 Tom de Vries <tom@codesourcery.com>
5278 Cesar Philippidis <cesar@codesourcery.com>
5281 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5283 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5286 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5287 spaces for function labels.
5289 2018-01-19 Martin Liska <mliska@suse.cz>
5291 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5292 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5293 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5294 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5295 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5296 (PRED_CONST_RETURN): Change from 69 to 65.
5297 (PRED_NULL_RETURN): Change from 91 to 71.
5298 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5299 (PRED_LOOP_GUARD): Change from 66 to 73.
5301 2018-01-19 Martin Liska <mliska@suse.cz>
5303 * predict.c (predict_insn_def): Add new assert.
5304 (struct branch_predictor): Change type to signed integer.
5305 (test_prediction_value_range): Amend test to cover
5307 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5308 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5309 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5310 (PRED_LOOP_IV_COMPARE): Likewise.
5311 * predict.h (PROB_UNINITIALIZED): Define new constant.
5313 2018-01-19 Martin Liska <mliska@suse.cz>
5315 * predict.c (dump_prediction): Add new format for
5316 analyze_brprob.py script which is enabled with -details
5318 * profile-count.h (precise_p): New function.
5320 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5322 PR tree-optimization/83922
5323 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5324 there are no statements that need masking.
5325 (vect_active_double_reduction_p): New function.
5326 (vect_analyze_loop_operations): Use it when handling phis that
5327 are not in the loop header.
5329 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5331 PR tree-optimization/83914
5332 * tree-vect-loop.c (vectorizable_induction): Don't convert
5333 init_expr or apply the peeling adjustment for inductions
5334 that are nested within the vectorized loop.
5336 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5338 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5341 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5345 * function.h (gimplify_parameters): Add gimple_seq * argument.
5346 * function.c: Include gimple.h and options.h.
5347 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5348 for the added local temporaries if needed.
5349 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5350 if there are any parameter cleanups, wrap whole body into a
5351 try/finally with the cleanups.
5353 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5356 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5357 Use GET_MODE_CLASS for scalar floating point.
5359 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5363 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5364 Fix call of call_cgraph_insertion_hooks.
5366 2018-01-18 Martin Sebor <msebor@redhat.com>
5368 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5370 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5373 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5376 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5379 * common.opt: (-ffile-prefix-map): New option.
5380 * opts.c (common_handle_option): Defer it.
5381 * opts-global.c (handle_common_deferred_options): Handle it.
5382 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5383 * file-prefix-map.h: New file.
5384 (remap_debug_filename, add_debug_prefix_map): ...here.
5385 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5386 * final.c (debug_prefix_map, add_debug_prefix_map
5387 remap_debug_filename): Move to...
5388 * file-prefix-map.c: New file.
5389 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5390 generalize, get rid of alloca(), use strrchr() instead of strchr().
5391 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5392 Implement in terms of add_prefix_map().
5393 (remap_macro_filename, remap_debug_filename): Implement in term of
5395 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5396 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5397 * dbxout.c: Include file-prefix-map.h.
5398 * varasm.c: Likewise.
5399 * vmsdbgout.c: Likewise.
5400 * xcoffout.c: Likewise.
5401 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5402 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5403 * doc/invoke.texi (-ffile-prefix-map): Document.
5404 (-fdebug-prefix-map): Update description.
5406 2018-01-18 Martin Liska <mliska@suse.cz>
5408 * config/i386/i386.c (indirect_thunk_name): Document that also
5410 (output_indirect_thunk): Document why both instructions
5411 (pause and lfence) are generated.
5413 2018-01-18 Richard Biener <rguenther@suse.de>
5415 PR tree-optimization/83887
5416 * graphite-scop-detection.c
5417 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5418 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5419 (scop_detection::merge_sese): Re-implement with a flood-fill
5420 algorithm that properly finds a SESE region if it exists.
5422 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5425 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5426 pointer_diff optimizations use view_convert instead of convert.
5428 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5430 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5431 Generate different code for -mno-speculate-indirect-jumps.
5432 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5433 (*call_indirect_aix<mode>): Disable for
5434 -mno-speculate-indirect-jumps.
5435 (*call_indirect_aix<mode>_nospec): New define_insn.
5436 (*call_value_indirect_aix<mode>): Disable for
5437 -mno-speculate-indirect-jumps.
5438 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5439 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5440 -mno-speculate-indirect-jumps.
5441 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5443 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5445 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5446 long double type, set the flags for noting the default long double
5447 type, even if we don't pass or return a long double type.
5449 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5452 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5455 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5457 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5458 support for merge[hl].
5459 (fold_mergehl_helper): New helper function.
5460 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5461 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5462 (altivec_vmrglw_direct): Add xxmrglw insn.
5464 2018-01-17 Andrew Waterman <andrew@sifive.com>
5466 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5467 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5469 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5472 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5473 call the lto_location_cache before reading the
5474 DECL_SOURCE_LOCATION of the types.
5476 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5477 Richard Sandiford <richard.sandiford@linaro.org>
5479 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5480 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5481 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5482 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5483 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5485 * config/aarch64/constraints.md (aarch64_movti_operand):
5487 * config/aarch64/predicates.md (Uti): Add new constraint.
5489 2018-01-17 Carl Love <cel@us.ibm.com>
5491 * config/rs6000/vsx.md (define_expand xl_len_r,
5492 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5493 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5495 (define_expand, define_insn): Move the shift left from the
5496 define_insn to the define_expand for lxvl and stxvl instructions.
5497 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5498 and XL_LEN_R definitions to PURE.
5500 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5502 * config/i386/i386.c (indirect_thunk_name): Declare regno
5503 as unsigned int. Compare regno with INVALID_REGNUM.
5504 (output_indirect_thunk): Ditto.
5505 (output_indirect_thunk_function): Ditto.
5506 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5507 in the call to output_indirect_thunk_function.
5509 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5512 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5513 rather than the size of inner_type to determine the stack slot size
5514 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5516 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5519 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5522 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5524 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5525 endian Linux systems to optionally enable multilibs for selecting
5526 the long double type if the user configured an explicit type.
5527 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5528 have no long double multilibs if not defined.
5529 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5530 warn if the user used -mabi={ieee,ibm}longdouble and we built
5531 multilibs for long double.
5532 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5533 appropriate multilib option.
5534 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5536 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5537 for building long double multilibs.
5538 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5540 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5542 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5545 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5547 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5550 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5553 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5556 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5558 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5559 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5561 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5563 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5564 different rtl trees depending on TARGET_64BIT.
5565 (rs6000_gen_lvx): Likewise.
5567 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5569 * config/visium/visium.md (nop): Tweak comment.
5570 (hazard_nop): Likewise.
5572 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5574 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5575 -mspeculate-indirect-jumps.
5576 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5577 for -mno-speculate-indirect-jumps.
5578 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5579 (*call_value_indirect_elfv2<mode>): Disable for
5580 -mno-speculate-indirect-jumps.
5581 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5582 (indirect_jump): Emit different RTL for
5583 -mno-speculate-indirect-jumps.
5584 (*indirect_jump<mode>): Disable for
5585 -mno-speculate-indirect-jumps.
5586 (*indirect_jump<mode>_nospec): New define_insn.
5587 (tablejump): Emit different RTL for
5588 -mno-speculate-indirect-jumps.
5589 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5590 (tablejumpsi_nospec): New define_expand.
5591 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5592 (tablejumpdi_nospec): New define_expand.
5593 (*tablejump<mode>_internal1): Disable for
5594 -mno-speculate-indirect-jumps.
5595 (*tablejump<mode>_internal1_nospec): New define_insn.
5596 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5599 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5601 * caller-save.c (insert_save): Drop unnecessary parameter. All
5604 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5605 Richard Biener <rguenth@suse.de>
5608 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5609 return early, inline manually is_gimple_sizepos. Make sure if we
5610 call gimplify_expr we don't end up with a gimple constant.
5611 * tree.c (variably_modified_type_p): Don't return true for
5612 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5613 * gimplify.h (is_gimple_sizepos): Remove.
5615 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5617 PR tree-optimization/83857
5618 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5619 vectorizable_live_operation for pure SLP statements.
5620 (vectorizable_live_operation): Handle PHIs.
5622 2018-01-16 Richard Biener <rguenther@suse.de>
5624 PR tree-optimization/83867
5625 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5626 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5628 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5631 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5632 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5633 If off is not INTEGER_CST, issue a may not be aligned warning
5634 rather than isn't aligned. Use isn%'t rather than isn't.
5635 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5637 <case MULT_EXPR>: Improve the case when bottom and one of the
5638 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5639 operand, in that case check if the other operand is multiple of
5640 bottom divided by the INTEGER_CST operand.
5642 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5645 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5646 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5647 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5648 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5649 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5650 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5651 (pa_function_arg_size): New function.
5653 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5655 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5656 in a separate statement.
5658 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5660 PR tree-optimization/83847
5661 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5662 group gathers and scatters.
5664 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5666 PR rtl-optimization/86620
5667 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5669 PR rtl-optimization/83213
5670 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5671 to last if both are JUMP_INSNs.
5673 PR tree-optimization/83843
5674 * gimple-ssa-store-merging.c
5675 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5676 store_immediate_info for bswap/nop orig_stores.
5678 2018-01-15 Andrew Waterman <andrew@sifive.com>
5680 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5682 <UDIV>: Increase cost if !TARGET_DIV.
5684 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5686 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5687 (define_attr "cr_logical_3op"): New.
5688 (cceq_ior_compare): Adjust.
5689 (cceq_ior_compare_complement): Adjust.
5690 (*cceq_rev_compare): Adjust.
5691 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5692 (is_cracked_insn): Adjust.
5693 (insn_must_be_first_in_group): Adjust.
5694 * config/rs6000/40x.md: Adjust.
5695 * config/rs6000/440.md: Adjust.
5696 * config/rs6000/476.md: Adjust.
5697 * config/rs6000/601.md: Adjust.
5698 * config/rs6000/603.md: Adjust.
5699 * config/rs6000/6xx.md: Adjust.
5700 * config/rs6000/7450.md: Adjust.
5701 * config/rs6000/7xx.md: Adjust.
5702 * config/rs6000/8540.md: Adjust.
5703 * config/rs6000/cell.md: Adjust.
5704 * config/rs6000/e300c2c3.md: Adjust.
5705 * config/rs6000/e500mc.md: Adjust.
5706 * config/rs6000/e500mc64.md: Adjust.
5707 * config/rs6000/e5500.md: Adjust.
5708 * config/rs6000/e6500.md: Adjust.
5709 * config/rs6000/mpc.md: Adjust.
5710 * config/rs6000/power4.md: Adjust.
5711 * config/rs6000/power5.md: Adjust.
5712 * config/rs6000/power6.md: Adjust.
5713 * config/rs6000/power7.md: Adjust.
5714 * config/rs6000/power8.md: Adjust.
5715 * config/rs6000/power9.md: Adjust.
5716 * config/rs6000/rs64.md: Adjust.
5717 * config/rs6000/titan.md: Adjust.
5719 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5721 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5722 ix86_indirect_branch_register logic.
5724 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5726 * config/i386/constraints.md (Bs): Update
5727 ix86_indirect_branch_register check. Don't check
5728 ix86_indirect_branch_register with GOT_memory_operand.
5730 * config/i386/predicates.md (GOT_memory_operand): Don't check
5731 ix86_indirect_branch_register here.
5732 (GOT32_symbol_operand): Likewise.
5734 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5736 * config/i386/predicates.md (constant_call_address_operand):
5737 Rewrite ix86_indirect_branch_register logic.
5738 (sibcall_insn_operand): Likewise.
5740 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5742 * config/i386/constraints.md (Bs): Replace
5743 ix86_indirect_branch_thunk_register with
5744 ix86_indirect_branch_register.
5746 * config/i386/i386.md (indirect_jump): Likewise.
5747 (tablejump): Likewise.
5748 (*sibcall_memory): Likewise.
5749 (*sibcall_value_memory): Likewise.
5750 Peepholes of indirect call and jump via memory: Likewise.
5751 * config/i386/i386.opt: Likewise.
5752 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5753 (GOT_memory_operand): Likewise.
5754 (call_insn_operand): Likewise.
5755 (sibcall_insn_operand): Likewise.
5756 (GOT32_symbol_operand): Likewise.
5758 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5761 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5762 type rather than type addr's type points to.
5763 (expand_omp_atomic_mutex): Likewise.
5764 (expand_omp_atomic): Likewise.
5766 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5769 * config/i386/i386.c (output_indirect_thunk_function): Use
5770 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5771 for __x86_return_thunk.
5773 2018-01-15 Richard Biener <rguenther@suse.de>
5776 * expmed.c (extract_bit_field_1): Fix typo.
5778 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5781 * config/arm/iterators.md (VF): New mode iterator.
5782 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5783 Remove integer-related logic from pattern.
5784 (neon_vabd<mode>_3): Likewise.
5786 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5789 * common.opt (fstrict-overflow): No longer an alias.
5790 (fwrapv-pointer): New option.
5791 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5792 also for pointer types based on flag_wrapv_pointer.
5793 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5794 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5795 opts->x_flag_wrapv got set.
5796 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5797 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5798 POINTER_TYPE_OVERFLOW_UNDEFINED.
5799 * match.pd: Likewise in address comparison pattern.
5800 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5802 2018-01-15 Richard Biener <rguenther@suse.de>
5805 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5806 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5807 Reset type names to their identifier if their TYPE_DECL doesn't
5808 have linkage (and thus is used for ODR and devirt).
5809 (save_debug_info_for_decl): Remove.
5810 (save_debug_info_for_type): Likewise.
5811 (add_tree_to_fld_list): Adjust.
5812 * tree-pretty-print.c (dump_generic_node): Make dumping of
5813 type names more robust.
5815 2018-01-15 Richard Biener <rguenther@suse.de>
5817 * BASE-VER: Bump to 8.0.1.
5819 2018-01-14 Martin Sebor <msebor@redhat.com>
5822 * builtins.c (check_access): Avoid warning when the no-warning bit
5825 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5827 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5828 * ira-color (allocno_hard_regs_compare): Likewise.
5830 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5833 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5834 Use .pushsection/.popsection.
5836 2018-01-14 Martin Sebor <msebor@redhat.com>
5839 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5841 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5843 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5844 entry from extra_headers.
5845 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5846 extra_headers, make the list bitwise identical to the i?86-*-* one.
5848 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5850 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5851 -mcmodel=large with -mindirect-branch=thunk,
5852 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5853 -mfunction-return=thunk-extern.
5854 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5855 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5856 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5858 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5860 * config/i386/i386.c (print_reg): Print the name of the full
5861 integer register without '%'.
5862 (ix86_print_operand): Handle 'V'.
5863 * doc/extend.texi: Document 'V' modifier.
5865 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5867 * config/i386/constraints.md (Bs): Disallow memory operand for
5868 -mindirect-branch-register.
5870 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5871 (GOT_memory_operand): Likewise.
5872 (call_insn_operand): Likewise.
5873 (sibcall_insn_operand): Likewise.
5874 (GOT32_symbol_operand): Likewise.
5875 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5876 for -mindirect-branch-register.
5877 (tablejump): Likewise.
5878 (*sibcall_memory): Likewise.
5879 (*sibcall_value_memory): Likewise.
5880 Disallow peepholes of indirect call and jump via memory for
5881 -mindirect-branch-register.
5882 (*call_pop): Replace m with Bw.
5883 (*call_value_pop): Likewise.
5884 (*sibcall_pop_memory): Replace m with Bs.
5885 * config/i386/i386.opt (mindirect-branch-register): New option.
5886 * doc/invoke.texi: Document -mindirect-branch-register option.
5888 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5890 * config/i386/i386-protos.h (ix86_output_function_return): New.
5891 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5892 set function_return_type.
5893 (indirect_thunk_name): Add ret_p to indicate thunk for function
5895 (output_indirect_thunk_function): Pass false to
5896 indirect_thunk_name.
5897 (ix86_output_indirect_branch_via_reg): Likewise.
5898 (ix86_output_indirect_branch_via_push): Likewise.
5899 (output_indirect_thunk_function): Create alias for function
5900 return thunk if regno < 0.
5901 (ix86_output_function_return): New function.
5902 (ix86_handle_fndecl_attribute): Handle function_return.
5903 (ix86_attribute_table): Add function_return.
5904 * config/i386/i386.h (machine_function): Add
5905 function_return_type.
5906 * config/i386/i386.md (simple_return_internal): Use
5907 ix86_output_function_return.
5908 (simple_return_internal_long): Likewise.
5909 * config/i386/i386.opt (mfunction-return=): New option.
5910 (indirect_branch): Mention -mfunction-return=.
5911 * doc/extend.texi: Document function_return function attribute.
5912 * doc/invoke.texi: Document -mfunction-return= option.
5914 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5916 * config/i386/i386-opts.h (indirect_branch): New.
5917 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
5918 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
5919 with local indirect jump when converting indirect call and jump.
5920 (ix86_set_indirect_branch_type): New.
5921 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
5922 (indirectlabelno): New.
5923 (indirect_thunk_needed): Likewise.
5924 (indirect_thunk_bnd_needed): Likewise.
5925 (indirect_thunks_used): Likewise.
5926 (indirect_thunks_bnd_used): Likewise.
5927 (INDIRECT_LABEL): Likewise.
5928 (indirect_thunk_name): Likewise.
5929 (output_indirect_thunk): Likewise.
5930 (output_indirect_thunk_function): Likewise.
5931 (ix86_output_indirect_branch_via_reg): Likewise.
5932 (ix86_output_indirect_branch_via_push): Likewise.
5933 (ix86_output_indirect_branch): Likewise.
5934 (ix86_output_indirect_jmp): Likewise.
5935 (ix86_code_end): Call output_indirect_thunk_function if needed.
5936 (ix86_output_call_insn): Call ix86_output_indirect_branch if
5938 (ix86_handle_fndecl_attribute): Handle indirect_branch.
5939 (ix86_attribute_table): Add indirect_branch.
5940 * config/i386/i386.h (machine_function): Add indirect_branch_type
5941 and has_local_indirect_jump.
5942 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
5944 (tablejump): Likewise.
5945 (*indirect_jump): Use ix86_output_indirect_jmp.
5946 (*tablejump_1): Likewise.
5947 (simple_return_indirect_internal): Likewise.
5948 * config/i386/i386.opt (mindirect-branch=): New option.
5949 (indirect_branch): New.
5952 (thunk-inline): Likewise.
5953 (thunk-extern): Likewise.
5954 * doc/extend.texi: Document indirect_branch function attribute.
5955 * doc/invoke.texi: Document -mindirect-branch= option.
5957 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
5960 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
5962 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
5964 * ipa-inline.c (want_inline_small_function_p): Return false if
5965 inlining has already failed with CIF_FINAL_ERROR.
5966 (update_caller_keys): Call want_inline_small_function_p before
5968 (update_callee_keys): Likewise.
5970 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
5972 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
5974 (rs6000_quadword_masked_address_p): Likewise.
5975 (quad_aligned_load_p): Likewise.
5976 (quad_aligned_store_p): Likewise.
5977 (const_load_sequence_p): Add comment to describe the outer-most loop.
5978 (mimic_memory_attributes_and_flags): New function.
5979 (rs6000_gen_stvx): Likewise.
5980 (replace_swapped_aligned_store): Likewise.
5981 (rs6000_gen_lvx): Likewise.
5982 (replace_swapped_aligned_load): Likewise.
5983 (replace_swapped_load_constant): Capitalize argument name in
5984 comment describing this function.
5985 (rs6000_analyze_swaps): Add a third pass to search for vector loads
5986 and stores that access quad-word aligned addresses and replace
5987 with stvx or lvx instructions when appropriate.
5988 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
5989 New function prototype.
5990 (rs6000_quadword_masked_address_p): Likewise.
5991 (rs6000_gen_lvx): Likewise.
5992 (rs6000_gen_stvx): Likewise.
5993 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
5994 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
5995 when memory address is aligned.
5996 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
5997 this split to select lvx instruction when memory address is aligned.
5998 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
5999 instruction when memory address is aligned.
6000 (*vsx_le_perm_load_v16qi): Likewise.
6001 (four unnamed splitters): Modify to select the stvx instruction
6002 when memory is aligned.
6004 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6006 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6007 which appears in the queue multiple times.
6009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6010 Alan Hayward <alan.hayward@arm.com>
6011 David Sherwood <david.sherwood@arm.com>
6013 * tree-vectorizer.h (vec_lower_bound): New structure.
6014 (_loop_vec_info): Add check_nonzero and lower_bounds.
6015 (LOOP_VINFO_CHECK_NONZERO): New macro.
6016 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6017 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6018 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6019 fields. Make seg_len the distance travelled, not including the
6021 (dr_direction_indicator): Declare.
6022 (dr_zero_step_indicator): Likewise.
6023 (dr_known_forward_stride_p): Likewise.
6024 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6026 (runtime_alias_check_p): Allow runtime alias checks with
6028 (operator ==): Compare access_size and align.
6029 (prune_runtime_alias_test_list): Rework for new distinction between
6030 the access_size and seg_len.
6031 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6033 (get_segment_min_max): New function.
6034 (create_intersect_range_checks): Use it.
6035 (dr_step_indicator): New function.
6036 (dr_direction_indicator): Likewise.
6037 (dr_zero_step_indicator): Likewise.
6038 (dr_known_forward_stride_p): Likewise.
6039 * tree-loop-distribution.c (data_ref_segment_size): Return
6040 DR_STEP * (niters - 1).
6041 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6043 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6044 (vect_preserves_scalar_order_p): New function, split out from...
6045 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6046 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6047 (vect_vfa_access_size): New function.
6048 (vect_vfa_align): Likewise.
6049 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6050 (dump_lower_bound): New function.
6051 (vect_check_lower_bound): Likewise.
6052 (vect_small_gap_p): Likewise.
6053 (vectorizable_with_step_bound_p): Likewise.
6054 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6055 depencies if the vectorization factor is 1. Convert the checks
6056 for nonzero steps into checks on the bounds of DR_STEP. Try using
6057 a bunds check for variable steps if the minimum required step is
6058 relatively small. Update calls to the dr_with_seg_len
6059 constructor and to vect_compile_time_alias.
6060 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6062 (vect_loop_versioning): Call it.
6063 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6065 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6067 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6068 Alan Hayward <alan.hayward@arm.com>
6069 David Sherwood <david.sherwood@arm.com>
6071 * doc/sourcebuild.texi (vect_scatter_store): Document.
6072 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6074 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6076 * genopinit.c (main): Add supports_vec_scatter_store and
6077 supports_vec_scatter_store_cached to target_optabs.
6078 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6079 IFN_MASK_SCATTER_STORE.
6080 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6082 * internal-fn.h (internal_store_fn_p): Declare.
6083 (internal_fn_stored_value_index): Likewise.
6084 * internal-fn.c (scatter_store_direct): New macro.
6085 (expand_scatter_store_optab_fn): New function.
6086 (direct_scatter_store_optab_supported_p): New macro.
6087 (internal_store_fn_p): New function.
6088 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6089 IFN_MASK_SCATTER_STORE.
6090 (internal_fn_mask_index): Likewise.
6091 (internal_fn_stored_value_index): New function.
6092 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6094 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6095 * optabs-query.c (supports_vec_scatter_store_p): New function.
6096 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6097 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6098 true for scatter stores.
6099 (vect_gather_scatter_fn_p): Handle scatter stores too.
6100 (vect_check_gather_scatter): Consider using scatter stores if
6101 supports_vec_scatter_store_p.
6102 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6104 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6105 internal_fn_stored_value_index.
6106 (check_load_store_masking): Handle scatter stores too.
6107 (vect_get_store_rhs): Make public.
6108 (vectorizable_call): Use internal_store_fn_p.
6109 (vectorizable_store): Handle scatter store internal functions.
6110 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6111 when deciding whether the end of the group has been reached.
6112 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6113 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6114 (mask_scatter_store<mode>): New insns.
6116 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6117 Alan Hayward <alan.hayward@arm.com>
6118 David Sherwood <david.sherwood@arm.com>
6120 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6121 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6122 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6124 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6125 Use vect_truncate_gather_scatter_offset if we can't treat the
6126 operation as a normal gather load or scatter store.
6127 (get_group_load_store_type): Take the gather_scatter_info
6128 as argument. Try using a gather load or scatter store for
6129 single-element groups.
6130 (get_load_store_type): Update calls to get_group_load_store_type
6131 and vect_use_strided_gather_scatters_p.
6133 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6134 Alan Hayward <alan.hayward@arm.com>
6135 David Sherwood <david.sherwood@arm.com>
6137 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6138 optional tree argument.
6139 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6141 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6142 but continue to use the current value as a fallback.
6143 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6144 to compare the updates.
6145 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6146 (get_load_store_type): Use it when handling a strided access.
6147 (vect_get_strided_load_store_ops): New function.
6148 (vect_get_data_ptr_increment): Likewise.
6149 (vectorizable_load): Handle strided gather loads. Always pass
6150 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6153 Alan Hayward <alan.hayward@arm.com>
6154 David Sherwood <david.sherwood@arm.com>
6156 * doc/md.texi (gather_load@var{m}): Document.
6157 (mask_gather_load@var{m}): Likewise.
6158 * genopinit.c (main): Add supports_vec_gather_load and
6159 supports_vec_gather_load_cached to target_optabs.
6160 * optabs-tree.c (init_tree_optimization_optabs): Use
6161 ggc_cleared_alloc to allocate target_optabs.
6162 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6163 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6165 * internal-fn.h (internal_load_fn_p): Declare.
6166 (internal_gather_scatter_fn_p): Likewise.
6167 (internal_fn_mask_index): Likewise.
6168 (internal_gather_scatter_fn_supported_p): Likewise.
6169 * internal-fn.c (gather_load_direct): New macro.
6170 (expand_gather_load_optab_fn): New function.
6171 (direct_gather_load_optab_supported_p): New macro.
6172 (direct_internal_fn_optab): New function.
6173 (internal_load_fn_p): Likewise.
6174 (internal_gather_scatter_fn_p): Likewise.
6175 (internal_fn_mask_index): Likewise.
6176 (internal_gather_scatter_fn_supported_p): Likewise.
6177 * optabs-query.c (supports_at_least_one_mode_p): New function.
6178 (supports_vec_gather_load_p): Likewise.
6179 * optabs-query.h (supports_vec_gather_load_p): Declare.
6180 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6181 and memory_type field.
6182 (NUM_PATTERNS): Bump to 15.
6183 * tree-vect-data-refs.c: Include internal-fn.h.
6184 (vect_gather_scatter_fn_p): New function.
6185 (vect_describe_gather_scatter_call): Likewise.
6186 (vect_check_gather_scatter): Try using internal functions for
6187 gather loads. Recognize existing calls to a gather load function.
6188 (vect_analyze_data_refs): Consider using gather loads if
6189 supports_vec_gather_load_p.
6190 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6191 (vect_get_gather_scatter_offset_type): Likewise.
6192 (vect_convert_mask_for_vectype): Likewise.
6193 (vect_add_conversion_to_patterm): Likewise.
6194 (vect_try_gather_scatter_pattern): Likewise.
6195 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6196 (vect_vect_recog_func_ptrs): Add it.
6197 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6198 internal_fn_mask_index and internal_gather_scatter_fn_p.
6199 (check_load_store_masking): Take the gather_scatter_info as an
6200 argument and handle gather loads.
6201 (vect_get_gather_scatter_ops): New function.
6202 (vectorizable_call): Check internal_load_fn_p.
6203 (vectorizable_load): Likewise. Handle gather load internal
6205 (vectorizable_store): Update call to check_load_store_masking.
6206 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6207 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6208 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6209 (aarch64_gather_scale_operand_d): New predicates.
6210 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6211 (mask_gather_load<mode>): New insns.
6213 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6214 Alan Hayward <alan.hayward@arm.com>
6215 David Sherwood <david.sherwood@arm.com>
6217 * optabs.def (fold_left_plus_optab): New optab.
6218 * doc/md.texi (fold_left_plus_@var{m}): Document.
6219 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6220 * internal-fn.c (fold_left_direct): Define.
6221 (expand_fold_left_optab_fn): Likewise.
6222 (direct_fold_left_optab_supported_p): Likewise.
6223 * fold-const-call.c (fold_const_fold_left): New function.
6224 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6225 * tree-parloops.c (valid_reduction_p): New function.
6226 (gather_scalar_reductions): Use it.
6227 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6228 (vect_finish_replace_stmt): Declare.
6229 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6230 (needs_fold_left_reduction_p): New function, split out from...
6231 (vect_is_simple_reduction): ...here. Accept reductions that
6232 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6233 (vect_force_simple_reduction): Also store the reduction type in
6234 the assignment's STMT_VINFO_REDUC_TYPE.
6235 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6236 (merge_with_identity): New function.
6237 (vect_expand_fold_left): Likewise.
6238 (vectorize_fold_left_reduction): Likewise.
6239 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6240 scalar phi in place for it. Check for target support and reject
6241 cases that would reassociate the operation. Defer the transform
6242 phase to vectorize_fold_left_reduction.
6243 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6244 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6245 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6247 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6249 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6250 call to ifc_temp_var.
6252 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6253 Alan Hayward <alan.hayward@arm.com>
6254 David Sherwood <david.sherwood@arm.com>
6256 * target.def (legitimize_address_displacement): Take the original
6257 offset as a poly_int.
6258 * targhooks.h (default_legitimize_address_displacement): Update
6260 * targhooks.c (default_legitimize_address_displacement): Likewise.
6261 * doc/tm.texi: Regenerate.
6262 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6263 as an argument, moving assert of ad->disp == ad->disp_term to...
6264 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6265 Try calling targetm.legitimize_address_displacement before expanding
6266 the address rather than afterwards, and adjust for the new interface.
6267 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6268 Match the new hook interface. Handle SVE addresses.
6269 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6272 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6274 * Makefile.in (OBJS): Add early-remat.o.
6275 * target.def (select_early_remat_modes): New hook.
6276 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6277 * doc/tm.texi: Regenerate.
6278 * targhooks.h (default_select_early_remat_modes): Declare.
6279 * targhooks.c (default_select_early_remat_modes): New function.
6280 * timevar.def (TV_EARLY_REMAT): New timevar.
6281 * passes.def (pass_early_remat): New pass.
6282 * tree-pass.h (make_pass_early_remat): Declare.
6283 * early-remat.c: New file.
6284 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6286 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6288 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6289 Alan Hayward <alan.hayward@arm.com>
6290 David Sherwood <david.sherwood@arm.com>
6292 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6293 vfm1 with a bound_epilog parameter.
6294 (vect_do_peeling): Update calls accordingly, and move the prologue
6295 call earlier in the function. Treat the base bound_epilog as 0 for
6296 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6297 this base when peeling for gaps.
6298 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6299 with fully-masked loops.
6300 (vect_estimate_min_profitable_iters): Handle the single peeled
6301 iteration in that case.
6303 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6304 Alan Hayward <alan.hayward@arm.com>
6305 David Sherwood <david.sherwood@arm.com>
6307 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6308 single-element interleaving even if the size is not a power of 2.
6309 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6310 accesses for single-element interleaving if the group size is
6313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6314 Alan Hayward <alan.hayward@arm.com>
6315 David Sherwood <david.sherwood@arm.com>
6317 * doc/md.texi (fold_extract_last_@var{m}): Document.
6318 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6319 * optabs.def (fold_extract_last_optab): New optab.
6320 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6321 * internal-fn.c (fold_extract_direct): New macro.
6322 (expand_fold_extract_optab_fn): Likewise.
6323 (direct_fold_extract_optab_supported_p): Likewise.
6324 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6325 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6326 EXTRACT_LAST_REDUCTION.
6327 (get_initial_def_for_reduction): Do not create an initial vector
6328 for EXTRACT_LAST_REDUCTION reductions.
6329 (vectorizable_reduction): Leave the scalar phi in place for
6330 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6331 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6332 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6333 transform phase to vectorizable_condition.
6334 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6336 (vect_finish_stmt_generation): ...here.
6337 (vect_finish_replace_stmt): New function.
6338 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6339 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6341 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6343 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6344 Alan Hayward <alan.hayward@arm.com>
6345 David Sherwood <david.sherwood@arm.com>
6347 * doc/md.texi (extract_last_@var{m}): Document.
6348 * optabs.def (extract_last_optab): New optab.
6349 * internal-fn.def (EXTRACT_LAST): New internal function.
6350 * internal-fn.c (cond_unary_direct): New macro.
6351 (expand_cond_unary_optab_fn): Likewise.
6352 (direct_cond_unary_optab_supported_p): Likewise.
6353 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6354 loops using EXTRACT_LAST.
6355 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6356 (extract_last_<mode>): ...this optab.
6357 (vec_extract<mode><Vel>): Update accordingly.
6359 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6360 Alan Hayward <alan.hayward@arm.com>
6361 David Sherwood <david.sherwood@arm.com>
6363 * target.def (empty_mask_is_expensive): New hook.
6364 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6365 * doc/tm.texi: Regenerate.
6366 * targhooks.h (default_empty_mask_is_expensive): Declare.
6367 * targhooks.c (default_empty_mask_is_expensive): New function.
6368 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6369 if the target says that empty masks are expensive.
6370 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6372 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6375 Alan Hayward <alan.hayward@arm.com>
6376 David Sherwood <david.sherwood@arm.com>
6378 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6379 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6380 (vect_use_loop_mask_for_alignment_p): New function.
6381 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6382 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6383 niters_skip argument. Make sure that the first niters_skip elements
6384 of the first iteration are inactive.
6385 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6386 Update call to vect_set_loop_masks_directly.
6387 (get_misalign_in_elems): New function, split out from...
6388 (vect_gen_prolog_loop_niters): ...here.
6389 (vect_update_init_of_dr): Take a code argument that specifies whether
6390 the adjustment should be added or subtracted.
6391 (vect_update_init_of_drs): Likewise.
6392 (vect_prepare_for_masked_peels): New function.
6393 (vect_do_peeling): Skip prologue peeling if we're using a mask
6394 instead. Update call to vect_update_inits_of_drs.
6395 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6397 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6398 alignment. Do not include the number of peeled iterations in
6399 the minimum threshold in that case.
6400 (vectorizable_induction): Adjust the start value down by
6401 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6402 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6403 Take the number of skipped iterations into account when calculating
6405 * tree-vect-stmts.c (vect_gen_while_not): New function.
6407 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6408 Alan Hayward <alan.hayward@arm.com>
6409 David Sherwood <david.sherwood@arm.com>
6411 * doc/sourcebuild.texi (vect_fully_masked): Document.
6412 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6414 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6416 (vect_analyze_loop_2): ...here. Don't check the vectorization
6417 factor against the number of loop iterations if the loop is
6420 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6421 Alan Hayward <alan.hayward@arm.com>
6422 David Sherwood <david.sherwood@arm.com>
6424 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6425 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6426 (dump_groups): Update accordingly.
6427 (iv_use::mem_type): New member variable.
6428 (address_p): New function.
6429 (record_use): Add a mem_type argument and initialize the new
6431 (record_group_use): Add a mem_type argument. Use address_p.
6432 Remove obsolete null checks of base_object. Update call to record_use.
6433 (find_interesting_uses_op): Update call to record_group_use.
6434 (find_interesting_uses_cond): Likewise.
6435 (find_interesting_uses_address): Likewise.
6436 (get_mem_type_for_internal_fn): New function.
6437 (find_address_like_use): Likewise.
6438 (find_interesting_uses_stmt): Try find_address_like_use before
6439 calling find_interesting_uses_op.
6440 (addr_offset_valid_p): Use the iv mem_type field as the type
6441 of the addressed memory.
6442 (add_autoinc_candidates): Likewise.
6443 (get_address_cost): Likewise.
6444 (split_small_address_groups_p): Use address_p.
6445 (split_address_groups): Likewise.
6446 (add_iv_candidate_for_use): Likewise.
6447 (autoinc_possible_for_pair): Likewise.
6448 (rewrite_groups): Likewise.
6449 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6450 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6451 (get_alias_ptr_type_for_ptr_address): New function.
6452 (rewrite_use_address): Rewrite address uses in calls that were
6453 identified by find_address_like_use.
6455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6456 Alan Hayward <alan.hayward@arm.com>
6457 David Sherwood <david.sherwood@arm.com>
6459 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6461 * gimple-expr.h (is_gimple_addressable: Likewise.
6462 * gimple-expr.c (is_gimple_address): Likewise.
6463 * internal-fn.c (expand_call_mem_ref): New function.
6464 (expand_mask_load_optab_fn): Use it.
6465 (expand_mask_store_optab_fn): Likewise.
6467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6468 Alan Hayward <alan.hayward@arm.com>
6469 David Sherwood <david.sherwood@arm.com>
6471 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6472 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6473 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6474 (cond_umax@var{mode}): Document.
6475 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6476 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6477 (cond_umin_optab, cond_umax_optab): New optabs.
6478 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6479 (COND_IOR, COND_XOR): New internal functions.
6480 * internal-fn.h (get_conditional_internal_fn): Declare.
6481 * internal-fn.c (cond_binary_direct): New macro.
6482 (expand_cond_binary_optab_fn): Likewise.
6483 (direct_cond_binary_optab_supported_p): Likewise.
6484 (get_conditional_internal_fn): New function.
6485 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6486 Cope with reduction statements that are vectorized as calls rather
6488 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6489 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6490 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6491 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6492 (UNSPEC_COND_EOR): New unspecs.
6493 (optab): Add mappings for them.
6494 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6495 (sve_int_op, sve_fp_op): New int attributes.
6497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6498 Alan Hayward <alan.hayward@arm.com>
6499 David Sherwood <david.sherwood@arm.com>
6501 * optabs.def (while_ult_optab): New optab.
6502 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6503 * internal-fn.def (WHILE_ULT): New internal function.
6504 * internal-fn.h (direct_internal_fn_supported_p): New override
6505 that takes two types as argument.
6506 * internal-fn.c (while_direct): New macro.
6507 (expand_while_optab_fn): New function.
6508 (convert_optab_supported_p): Likewise.
6509 (direct_while_optab_supported_p): New macro.
6510 * wide-int.h (wi::udiv_ceil): New function.
6511 * tree-vectorizer.h (rgroup_masks): New structure.
6512 (vec_loop_masks): New typedef.
6513 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6515 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6516 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6517 (vect_max_vf): New function.
6518 (slpeel_make_loop_iterate_ntimes): Delete.
6519 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6520 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6521 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6522 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6523 internal-fn.h, stor-layout.h and optabs-query.h.
6524 (vect_set_loop_mask): New function.
6525 (add_preheader_seq): Likewise.
6526 (add_header_seq): Likewise.
6527 (interleave_supported_p): Likewise.
6528 (vect_maybe_permute_loop_masks): Likewise.
6529 (vect_set_loop_masks_directly): Likewise.
6530 (vect_set_loop_condition_masked): Likewise.
6531 (vect_set_loop_condition_unmasked): New function, split out from
6532 slpeel_make_loop_iterate_ntimes.
6533 (slpeel_make_loop_iterate_ntimes): Rename to..
6534 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6535 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6536 (vect_do_peeling): Update call accordingly.
6537 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6539 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6540 mask_compare_type, can_fully_mask_p and fully_masked_p.
6541 (release_vec_loop_masks): New function.
6542 (_loop_vec_info): Use it to free the loop masks.
6543 (can_produce_all_loop_masks_p): New function.
6544 (vect_get_max_nscalars_per_iter): Likewise.
6545 (vect_verify_full_masking): Likewise.
6546 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6547 retries, and free the mask rgroups before retrying. Check loop-wide
6548 reasons for disallowing fully-masked loops. Make the final decision
6549 about whether use a fully-masked loop or not.
6550 (vect_estimate_min_profitable_iters): Do not assume that peeling
6551 for the number of iterations will be needed for fully-masked loops.
6552 (vectorizable_reduction): Disable fully-masked loops.
6553 (vectorizable_live_operation): Likewise.
6554 (vect_halve_mask_nunits): New function.
6555 (vect_double_mask_nunits): Likewise.
6556 (vect_record_loop_mask): Likewise.
6557 (vect_get_loop_mask): Likewise.
6558 (vect_transform_loop): Handle the case in which the final loop
6559 iteration might handle a partial vector. Call vect_set_loop_condition
6560 instead of slpeel_make_loop_iterate_ntimes.
6561 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6562 (check_load_store_masking): New function.
6563 (prepare_load_store_mask): Likewise.
6564 (vectorizable_store): Handle fully-masked loops.
6565 (vectorizable_load): Likewise.
6566 (supportable_widening_operation): Use vect_halve_mask_nunits for
6568 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6569 (vect_gen_while): New function.
6570 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6571 (aarch64_uqdec<mode>): New insn.
6573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6574 Alan Hayward <alan.hayward@arm.com>
6575 David Sherwood <david.sherwood@arm.com>
6577 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6578 (reduc_xor_scal_optab): New optabs.
6579 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6580 (reduc_xor_scal_@var{m}): Document.
6581 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6582 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6584 * fold-const-call.c (fold_const_call): Handle them.
6585 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6586 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6587 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6588 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6589 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6590 (UNSPEC_XORV): New unspecs.
6591 (optab): Add entries for them.
6592 (BITWISEV): New int iterator.
6593 (bit_reduc_op): New int attributes.
6595 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6596 Alan Hayward <alan.hayward@arm.com>
6597 David Sherwood <david.sherwood@arm.com>
6599 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6600 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6601 * optabs.def (vec_shl_insert_optab): New optab.
6602 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6603 (duplicate_and_interleave): Likewise.
6604 * tree-vect-loop.c: Include internal-fn.h.
6605 (neutral_op_for_slp_reduction): New function, split out from
6606 get_initial_defs_for_reduction.
6607 (get_initial_def_for_reduction): Handle option 2 for variable-length
6608 vectors by loading the neutral value into a vector and then shifting
6609 the initial value into element 0.
6610 (get_initial_defs_for_reduction): Replace the code argument with
6611 the neutral value calculated by neutral_op_for_slp_reduction.
6612 Use gimple_build_vector for constant-length vectors.
6613 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6614 but the first group_size elements have a neutral value.
6615 Use duplicate_and_interleave otherwise.
6616 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6617 Update call to get_initial_defs_for_reduction. Handle SLP
6618 reductions for variable-length vectors by creating one vector
6619 result for each scalar result, with the elements associated
6620 with other scalar results stubbed out with the neutral value.
6621 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6622 Require IFN_VEC_SHL_INSERT for double reductions on
6623 variable-length vectors, or SLP reductions that have
6624 a neutral value. Require can_duplicate_and_interleave_p
6625 support for variable-length unchained SLP reductions if there
6626 is no neutral value, such as for MIN/MAX reductions. Also require
6627 the number of vector elements to be a multiple of the number of
6628 SLP statements when doing variable-length unchained SLP reductions.
6629 Update call to vect_create_epilog_for_reduction.
6630 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6631 and remove initial values.
6632 (duplicate_and_interleave): Make public.
6633 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6634 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6637 Alan Hayward <alan.hayward@arm.com>
6638 David Sherwood <david.sherwood@arm.com>
6640 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6641 (can_duplicate_and_interleave_p): New function.
6642 (vect_get_and_check_slp_defs): Take the vector of statements
6643 rather than just the current one. Remove excess parentheses.
6644 Restriction rejectinon of vect_constant_def and vect_external_def
6645 for variable-length vectors to boolean types, or types for which
6646 can_duplicate_and_interleave_p is false.
6647 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6648 (duplicate_and_interleave): New function.
6649 (vect_get_constant_vectors): Use gimple_build_vector for
6650 constant-length vectors and suitable variable-length constant
6651 vectors. Use duplicate_and_interleave for other variable-length
6652 vectors. Don't defer the update when inserting new statements.
6654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6655 Alan Hayward <alan.hayward@arm.com>
6656 David Sherwood <david.sherwood@arm.com>
6658 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6659 min_profitable_iters doesn't go negative.
6661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6662 Alan Hayward <alan.hayward@arm.com>
6663 David Sherwood <david.sherwood@arm.com>
6665 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6666 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6667 * optabs.def (vec_mask_load_lanes_optab): New optab.
6668 (vec_mask_store_lanes_optab): Likewise.
6669 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6670 (MASK_STORE_LANES): Likewise.
6671 * internal-fn.c (mask_load_lanes_direct): New macro.
6672 (mask_store_lanes_direct): Likewise.
6673 (expand_mask_load_optab_fn): Handle masked operations.
6674 (expand_mask_load_lanes_optab_fn): New macro.
6675 (expand_mask_store_optab_fn): Handle masked operations.
6676 (expand_mask_store_lanes_optab_fn): New macro.
6677 (direct_mask_load_lanes_optab_supported_p): Likewise.
6678 (direct_mask_store_lanes_optab_supported_p): Likewise.
6679 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6681 (vect_load_lanes_supported): Likewise.
6682 * tree-vect-data-refs.c (strip_conversion): New function.
6683 (can_group_stmts_p): Likewise.
6684 (vect_analyze_data_ref_accesses): Use it instead of checking
6685 for a pair of assignments.
6686 (vect_store_lanes_supported): Take a masked_p parameter.
6687 (vect_load_lanes_supported): Likewise.
6688 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6689 vect_store_lanes_supported and vect_load_lanes_supported.
6690 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6691 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6692 parameter. Don't allow gaps for masked accesses.
6693 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6694 and vect_load_lanes_supported.
6695 (get_load_store_type): Take a masked_p parameter and update
6696 call to get_group_load_store_type.
6697 (vectorizable_store): Update call to get_load_store_type.
6698 Handle IFN_MASK_STORE_LANES.
6699 (vectorizable_load): Update call to get_load_store_type.
6700 Handle IFN_MASK_LOAD_LANES.
6702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6703 Alan Hayward <alan.hayward@arm.com>
6704 David Sherwood <david.sherwood@arm.com>
6706 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6708 * config/aarch64/aarch64-protos.h
6709 (aarch64_sve_struct_memory_operand_p): Declare.
6710 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6711 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6712 (VPRED, vpred): Handle SVE structure modes.
6713 * config/aarch64/constraints.md (Utx): New constraint.
6714 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6715 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6716 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6717 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6718 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6719 structure modes. Split into pieces after RA.
6720 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6721 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6723 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6724 SVE structure modes.
6725 (aarch64_classify_address): Likewise.
6726 (sizetochar): Move earlier in file.
6727 (aarch64_print_operand): Handle SVE register lists.
6728 (aarch64_array_mode): New function.
6729 (aarch64_sve_struct_memory_operand_p): Likewise.
6730 (TARGET_ARRAY_MODE): Redefine.
6732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6733 Alan Hayward <alan.hayward@arm.com>
6734 David Sherwood <david.sherwood@arm.com>
6736 * target.def (array_mode): New target hook.
6737 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6738 * doc/tm.texi: Regenerate.
6739 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6740 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6741 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6743 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6746 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6747 Alan Hayward <alan.hayward@arm.com>
6748 David Sherwood <david.sherwood@arm.com>
6750 * fold-const.c (fold_binary_loc): Check the argument types
6751 rather than the result type when testing for a vector operation.
6753 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6755 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6756 * doc/tm.texi: Regenerate.
6758 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6759 Alan Hayward <alan.hayward@arm.com>
6760 David Sherwood <david.sherwood@arm.com>
6762 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6763 (sve): Document new AArch64 extension.
6764 * doc/md.texi (w): Extend the description of the AArch64
6765 constraint to include SVE vectors.
6766 (Upl, Upa): Document new AArch64 predicate constraints.
6767 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6769 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6770 (msve-vector-bits=): New option.
6771 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6772 SVE when these are disabled.
6773 (sve): New extension.
6774 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6775 modes. Adjust their number of units based on aarch64_sve_vg.
6776 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6777 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6778 aarch64_addr_query_type.
6779 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6780 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6781 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6782 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6783 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6784 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6785 (aarch64_simd_imm_zero_p): Delete.
6786 (aarch64_check_zero_based_sve_index_immediate): Declare.
6787 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6788 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6789 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6790 (aarch64_sve_float_mul_immediate_p): Likewise.
6791 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6793 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6794 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6795 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6796 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6797 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6798 (aarch64_regmode_natural_size): Likewise.
6799 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6800 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6802 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6803 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6804 for VG and the SVE predicate registers.
6805 (V_ALIASES): Add a "z"-prefixed alias.
6806 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6807 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6808 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6809 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6810 (REG_CLASS_NAMES): Add entries for them.
6811 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6812 and the predicate registers.
6813 (aarch64_sve_vg): Declare.
6814 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6815 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6816 (REGMODE_NATURAL_SIZE): Define.
6817 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6819 * config/aarch64/aarch64.c: Include cfgrtl.h.
6820 (simd_immediate_info): Add a constructor for series vectors,
6821 and an associated step field.
6822 (aarch64_sve_vg): New variable.
6823 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6824 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6825 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6826 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6827 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6828 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6829 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6830 (aarch64_get_mask_mode): New functions.
6831 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6832 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6833 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6834 predicate modes and predicate registers. Explicitly restrict
6835 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6836 to store a vector mode if it is recognized by
6837 aarch64_classify_vector_mode.
6838 (aarch64_regmode_natural_size): New function.
6839 (aarch64_hard_regno_caller_save_mode): Return the original mode
6841 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6842 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6843 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6844 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6846 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6847 does not overlap dest if the function is frame-related. Handle
6849 (aarch64_split_add_offset): New function.
6850 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6851 them aarch64_add_offset.
6852 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6853 and update call to aarch64_sub_sp.
6854 (aarch64_add_cfa_expression): New function.
6855 (aarch64_expand_prologue): Pass extra temporary registers to the
6856 functions above. Handle the case in which we need to emit new
6857 DW_CFA_expressions for registers that were originally saved
6858 relative to the stack pointer, but now have to be expressed
6859 relative to the frame pointer.
6860 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6862 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6863 IP0 and IP1 values for SVE frames.
6864 (aarch64_expand_vec_series): New function.
6865 (aarch64_expand_sve_widened_duplicate): Likewise.
6866 (aarch64_expand_sve_const_vector): Likewise.
6867 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6868 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6869 into the register, rather than emitting a SET directly.
6870 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6871 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6872 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6873 (offset_9bit_signed_scaled_p): New functions.
6874 (aarch64_replicate_bitmask_imm): New function.
6875 (aarch64_bitmask_imm): Use it.
6876 (aarch64_cannot_force_const_mem): Reject expressions involving
6877 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6878 (aarch64_classify_index): Handle SVE indices, by requiring
6879 a plain register index with a scale that matches the element size.
6880 (aarch64_classify_address): Handle SVE addresses. Assert that
6881 the mode of the address is VOIDmode or an integer mode.
6882 Update call to aarch64_classify_symbol.
6883 (aarch64_classify_symbolic_expression): Update call to
6884 aarch64_classify_symbol.
6885 (aarch64_const_vec_all_in_range_p): New function.
6886 (aarch64_print_vector_float_operand): Likewise.
6887 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6888 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6889 and the FP immediates 1.0 and 0.5.
6890 (aarch64_print_address_internal): Handle SVE addresses.
6891 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6892 (aarch64_regno_regclass): Handle predicate registers.
6893 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6895 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6896 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6897 (aarch64_convert_sve_vector_bits): New function.
6898 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6899 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6901 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6902 Handle SVE vector and predicate modes. Accept VL-based constants
6903 that need only one temporary register, and VL offsets that require
6904 no temporary registers.
6905 (aarch64_conditional_register_usage): Mark the predicate registers
6906 as fixed if SVE isn't available.
6907 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6908 Return true for SVE vector and predicate modes.
6909 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6910 rather than an unsigned int. Handle SVE modes.
6911 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6913 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6915 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6916 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6917 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6918 (aarch64_sve_float_mul_immediate_p): New functions.
6919 (aarch64_sve_valid_immediate): New function.
6920 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
6921 Explicitly reject structure modes. Check for INDEX constants.
6922 Handle PTRUE and PFALSE constants.
6923 (aarch64_check_zero_based_sve_index_immediate): New function.
6924 (aarch64_simd_imm_zero_p): Delete.
6925 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
6926 vector modes. Accept constants in the range of CNT[BHWD].
6927 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
6928 ask for an Advanced SIMD mode.
6929 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
6930 (aarch64_simd_vector_alignment): Handle SVE predicates.
6931 (aarch64_vectorize_preferred_vector_alignment): New function.
6932 (aarch64_simd_vector_alignment_reachable): Use it instead of
6934 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
6935 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
6937 (MAX_VECT_LEN): Delete.
6938 (expand_vec_perm_d): Add a vec_flags field.
6939 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
6940 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
6941 (aarch64_evpc_ext): Don't apply a big-endian lane correction
6943 (aarch64_evpc_rev): Rename to...
6944 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
6945 (aarch64_evpc_rev_global): New function.
6946 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
6947 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
6949 (aarch64_evpc_sve_tbl): New function.
6950 (aarch64_expand_vec_perm_const_1): Update after rename of
6951 aarch64_evpc_rev. Handle SVE permutes too, trying
6952 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
6953 than aarch64_evpc_tbl.
6954 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
6955 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
6956 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
6957 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
6958 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
6959 (aarch64_expand_sve_vcond): New functions.
6960 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
6961 of aarch64_vector_mode_p.
6962 (aarch64_dwarf_poly_indeterminate_value): New function.
6963 (aarch64_compute_pressure_classes): Likewise.
6964 (aarch64_can_change_mode_class): Likewise.
6965 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
6966 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
6967 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
6968 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
6969 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
6970 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
6971 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
6972 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
6974 (Dn, Dl, Dr): Accept const as well as const_vector.
6975 (Dz): Likewise. Compare against CONST0_RTX.
6976 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
6977 of "vector" where appropriate.
6978 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
6979 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
6980 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
6981 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
6982 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
6983 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
6984 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
6985 (v_int_equiv): Extend to SVE modes.
6986 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
6988 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
6989 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
6990 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
6991 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
6992 (SVE_COND_FP_CMP): New int iterators.
6993 (perm_hilo): Handle the new unpack unspecs.
6994 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
6996 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
6997 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
6998 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
6999 (aarch64_equality_operator, aarch64_constant_vector_operand)
7000 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7001 (aarch64_sve_nonimmediate_operand): Likewise.
7002 (aarch64_sve_general_operand): Likewise.
7003 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7004 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7005 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7006 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7007 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7008 (aarch64_sve_float_arith_immediate): Likewise.
7009 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7010 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7011 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7012 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7013 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7014 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7015 (aarch64_sve_float_arith_operand): Likewise.
7016 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7017 (aarch64_sve_float_mul_operand): Likewise.
7018 (aarch64_sve_vec_perm_operand): Likewise.
7019 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7020 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7021 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7022 as well as const_vector.
7023 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7024 in file. Use CONST0_RTX and CONSTM1_RTX.
7025 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7026 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7027 Use aarch64_simd_imm_zero.
7028 * config/aarch64/aarch64-sve.md: New file.
7029 * config/aarch64/aarch64.md: Include it.
7030 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7031 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7032 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7033 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7034 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7035 (sve): New attribute.
7036 (enabled): Disable instructions with the sve attribute unless
7038 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7039 aarch64_expand_mov_immediate.
7040 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7041 CNT[BHSD] immediates.
7042 (movti): Split CONST_POLY_INT moves into two halves.
7043 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7044 Split additions that need a temporary here if the destination
7045 is the stack pointer.
7046 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7047 (*add<mode>3_poly_1): New instruction.
7048 (set_clobber_cc): New expander.
7050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7052 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7053 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7054 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7055 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7056 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7057 Change innermode from fixed_mode_size to machine_mode.
7058 (simplify_subreg): Update call accordingly. Handle a constant-sized
7059 subreg of a variable-length CONST_VECTOR.
7061 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7062 Alan Hayward <alan.hayward@arm.com>
7063 David Sherwood <david.sherwood@arm.com>
7065 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7066 (add_offset_to_base): New function, split out from...
7067 (create_mem_ref): ...here. When handling a scale other than 1,
7068 check first whether the address is valid without the offset.
7069 Add it into the base if so, leaving the index and scale as-is.
7071 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7074 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7075 fold_for_warn before checking if arg2 is INTEGER_CST.
7077 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7079 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7080 (store_multiple_operation): Delete.
7081 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7082 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7083 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7084 guarded by TARGET_STRING.
7085 (rs6000_output_load_multiple): Delete.
7086 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7087 OPTION_MASK_STRING / TARGET_STRING handling.
7088 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7089 (const rs6000_opt_masks) <"string">: Change mask to 0.
7090 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7091 (MASK_STRING): Delete.
7092 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7094 (load_multiple): Delete.
7101 (store_multiple): Delete.
7108 (movmemsi_8reg): Delete.
7109 (corresponding unnamed define_insn): Delete.
7110 (movmemsi_6reg): Delete.
7111 (corresponding unnamed define_insn): Delete.
7112 (movmemsi_4reg): Delete.
7113 (corresponding unnamed define_insn): Delete.
7114 (movmemsi_2reg): Delete.
7115 (corresponding unnamed define_insn): Delete.
7116 (movmemsi_1reg): Delete.
7117 (corresponding unnamed define_insn): Delete.
7118 * config/rs6000/rs6000.opt (mno-string): New.
7119 (mstring): Replace by deprecation warning stub.
7120 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7122 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7124 * regrename.c (regrename_do_replace): If replacing the same
7125 reg multiple times, try to reuse last created gen_raw_REG.
7128 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7129 main to workaround a bug in GDB.
7131 2018-01-12 Tom de Vries <tom@codesourcery.com>
7134 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7136 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7138 PR rtl-optimization/80481
7139 * ira-color.c (get_cap_member): New function.
7140 (allocnos_conflict_by_live_ranges_p): Use it.
7141 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7142 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7144 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7147 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7148 (*saddl_se_1): Ditto.
7150 (*ssubl_se_1): Ditto.
7152 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7154 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7155 rather than wi::to_widest for DR_INITs.
7156 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7157 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7158 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7160 (vect_analyze_group_access_1): Note that here.
7162 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7164 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7165 polynomial type sizes.
7167 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7169 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7170 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7171 (gimple_add_tmp_var): Likewise.
7173 2018-01-12 Martin Liska <mliska@suse.cz>
7175 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7176 (gimple_alloc_sizes): Likewise.
7177 (dump_gimple_statistics): Use PRIu64 in printf format.
7178 * gimple.h: Change uint64_t to int.
7180 2018-01-12 Martin Liska <mliska@suse.cz>
7182 * tree-core.h: Use uint64_t instead of int.
7183 * tree.c (tree_node_counts): Likewise.
7184 (tree_node_sizes): Likewise.
7185 (dump_tree_statistics): Use PRIu64 in printf format.
7187 2018-01-12 Martin Liska <mliska@suse.cz>
7189 * Makefile.in: As qsort_chk is implemented in vec.c, add
7190 vec.o to linkage of gencfn-macros.
7191 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7192 passing the info to record_node_allocation_statistics.
7193 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7195 * ggc-common.c (struct ggc_usage): Add operator== and use
7196 it in operator< and compare function.
7197 * mem-stats.h (struct mem_usage): Likewise.
7198 * vec.c (struct vec_usage): Remove operator< and compare
7199 function. Can be simply inherited.
7201 2018-01-12 Martin Jambor <mjambor@suse.cz>
7204 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7205 * tree-ssa-math-opts.c: Include domwalk.h.
7206 (convert_mult_to_fma_1): New function.
7207 (fma_transformation_info): New type.
7208 (fma_deferring_state): Likewise.
7209 (cancel_fma_deferring): New function.
7210 (result_of_phi): Likewise.
7211 (last_fma_candidate_feeds_initial_phi): Likewise.
7212 (convert_mult_to_fma): Added deferring logic, split actual
7213 transformation to convert_mult_to_fma_1.
7214 (math_opts_dom_walker): New type.
7215 (math_opts_dom_walker::after_dom_children): New method, body moved
7216 here from pass_optimize_widening_mul::execute, added deferring logic
7218 (pass_optimize_widening_mul::execute): Moved most of code to
7219 math_opts_dom_walker::after_dom_children.
7220 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7221 * config/i386/i386.c (ix86_option_override_internal): Added
7222 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7224 2018-01-12 Richard Biener <rguenther@suse.de>
7227 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7228 inline instance vars.
7230 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7233 * config/rx/rx.c (rx_is_restricted_memory_address):
7236 2018-01-12 Richard Biener <rguenther@suse.de>
7238 PR tree-optimization/80846
7239 * target.def (split_reduction): New target hook.
7240 * targhooks.c (default_split_reduction): New function.
7241 * targhooks.h (default_split_reduction): Declare.
7242 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7243 target requests first reduce vectors by combining low and high
7245 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7246 (get_vectype_for_scalar_type_and_size): Export.
7247 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7248 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7249 * doc/tm.texi: Regenerate.
7250 * config/i386/i386.c (ix86_split_reduction): Implement
7251 TARGET_VECTORIZE_SPLIT_REDUCTION.
7253 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7256 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7257 in PIC mode except for TARGET_VXWORKS_RTP.
7258 * config/sparc/sparc.c: Include cfgrtl.h.
7259 (TARGET_INIT_PIC_REG): Define.
7260 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7261 (sparc_pic_register_p): New predicate.
7262 (sparc_legitimate_address_p): Use it.
7263 (sparc_legitimize_pic_address): Likewise.
7264 (sparc_delegitimize_address): Likewise.
7265 (sparc_mode_dependent_address_p): Likewise.
7266 (gen_load_pcrel_sym): Remove 4th parameter.
7267 (load_got_register): Adjust call to above. Remove obsolete stuff.
7268 (sparc_expand_prologue): Do not call load_got_register here.
7269 (sparc_flat_expand_prologue): Likewise.
7270 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7271 (sparc_use_pseudo_pic_reg): New function.
7272 (sparc_init_pic_reg): Likewise.
7273 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7274 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7276 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7278 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7279 Add item for branch_cost.
7281 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7283 PR rtl-optimization/83565
7284 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7285 not extend the result to a larger mode for rotate operations.
7286 (num_sign_bit_copies1): Likewise.
7288 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7291 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7293 Use values-Xc.o for -pedantic.
7294 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7296 2018-01-12 Martin Liska <mliska@suse.cz>
7299 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7301 (possible_polymorphic_call_targets): Use it.
7302 (ipa_devirt): Likewise.
7304 2018-01-12 Martin Liska <mliska@suse.cz>
7306 * profile-count.h (enum profile_quality): Use 0 as invalid
7307 enum value of profile_quality.
7309 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7311 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7312 -mext-string options.
7314 2018-01-12 Richard Biener <rguenther@suse.de>
7316 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7317 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7318 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7320 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7322 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7324 * configure.ac (--with-long-double-format): Add support for the
7325 configuration option to change the default long double format on
7327 * config.gcc (powerpc*-linux*-*): Likewise.
7328 * configure: Regenerate.
7329 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7330 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7331 used without modification.
7333 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7335 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7336 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7337 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7338 MISC_BUILTIN_SPEC_BARRIER.
7339 (rs6000_init_builtins): Likewise.
7340 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7342 (speculation_barrier): New define_insn.
7343 * doc/extend.texi: Document __builtin_speculation_barrier.
7345 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7348 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7349 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7350 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7352 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7353 integral modes instead of "ss" and "sd".
7354 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7355 vectors with 32-bit and 64-bit elements.
7356 (vecdupssescalarmodesuffix): New mode attribute.
7357 (vec_dup<mode>): Use it.
7359 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7362 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7363 frame if argument is passed on stack.
7365 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7368 * ree.c (combine_reaching_defs): Optimize also
7369 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7370 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7372 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7375 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7377 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7380 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7381 after they are computed.
7383 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7385 PR tree-optimization/83695
7386 * gimple-loop-linterchange.cc
7387 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7388 reset cached scev information after interchange.
7389 (pass_linterchange::execute): Remove call to scev_reset_htab.
7391 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7393 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7394 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7395 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7396 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7397 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7398 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7399 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7400 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7401 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7402 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7403 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7404 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7405 (V_lane_reg): Likewise.
7406 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7408 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7409 (vfmal_lane_low<mode>_intrinsic,
7410 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7411 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7412 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7413 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7414 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7415 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7417 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7419 * config/arm/arm-cpus.in (fp16fml): New feature.
7420 (ALL_SIMD): Add fp16fml.
7421 (armv8.2-a): Add fp16fml as an option.
7422 (armv8.3-a): Likewise.
7423 (armv8.4-a): Add fp16fml as part of fp16.
7424 * config/arm/arm.h (TARGET_FP16FML): Define.
7425 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7427 * config/arm/arm-modes.def (V2HF): Define.
7428 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7429 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7430 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7431 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7432 vfmsl_low, vfmsl_high): New set of builtins.
7433 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7434 (vfml_op): New code attribute.
7435 (VFMLHALVES): New int iterator.
7436 (VFML, VFMLSEL): New mode attributes.
7437 (V_reg): Define mapping for V2HF.
7438 (V_hi, V_lo): New mode attributes.
7439 (VF_constraint): Likewise.
7440 (vfml_half, vfml_half_selector): New int attributes.
7441 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7443 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7444 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7446 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7447 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7448 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7449 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7451 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7452 Document new effective target and option set.
7454 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7456 * config/arm/arm-cpus.in (armv8_4): New feature.
7457 (ARMv8_4a): New fgroup.
7458 (armv8.4-a): New arch.
7459 * config/arm/arm-tables.opt: Regenerate.
7460 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7461 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7462 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7463 Add matching rules for -march=armv8.4-a and extensions.
7464 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7466 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7469 * config/rx/rx.md (BW): New mode attribute.
7470 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7472 2018-01-11 Richard Biener <rguenther@suse.de>
7474 PR tree-optimization/83435
7475 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7476 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7477 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7479 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7480 Alan Hayward <alan.hayward@arm.com>
7481 David Sherwood <david.sherwood@arm.com>
7483 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7485 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7486 (aarch64_print_address_internal): Use it to check for a zero offset.
7488 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7489 Alan Hayward <alan.hayward@arm.com>
7490 David Sherwood <david.sherwood@arm.com>
7492 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7493 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7494 Return a poly_int64 rather than a HOST_WIDE_INT.
7495 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7496 rather than a HOST_WIDE_INT.
7497 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7498 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7499 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7500 final_offset from HOST_WIDE_INT to poly_int64.
7501 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7502 to_constant when getting the number of units in an Advanced SIMD
7504 (aarch64_builtin_vectorized_function): Check for a constant number
7506 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7508 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7509 attribute instead of GET_MODE_NUNITS.
7510 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7511 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7512 GET_MODE_SIZE for fixed-size registers.
7513 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7514 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7515 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7516 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7517 (aarch64_print_operand, aarch64_print_address_internal)
7518 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7519 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7520 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7521 Handle polynomial GET_MODE_SIZE.
7522 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7523 wider than SImode without modification.
7524 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7525 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7526 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7527 passing and returning SVE modes.
7528 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7529 rather than GEN_INT.
7530 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7531 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7532 (aarch64_allocate_and_probe_stack_space): Likewise.
7533 (aarch64_layout_frame): Cope with polynomial offsets.
7534 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7535 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7537 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7538 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7539 poly_int64 rather than a HOST_WIDE_INT.
7540 (aarch64_get_separate_components, aarch64_process_components)
7541 (aarch64_expand_prologue, aarch64_expand_epilogue)
7542 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7543 (aarch64_anchor_offset): New function, split out from...
7544 (aarch64_legitimize_address): ...here.
7545 (aarch64_builtin_vectorization_cost): Handle polynomial
7546 TYPE_VECTOR_SUBPARTS.
7547 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7549 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7550 number of elements from the PARALLEL rather than the mode.
7551 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7552 rather than GET_MODE_BITSIZE.
7553 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7554 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7555 (aarch64_expand_vec_perm_const_1): Handle polynomial
7556 d->perm.length () and d->perm elements.
7557 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7558 Apply to_constant to d->perm elements.
7559 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7560 polynomial CONST_VECTOR_NUNITS.
7561 (aarch64_move_pointer): Take amount as a poly_int64 rather
7563 (aarch64_progress_pointer): Avoid temporary variable.
7564 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7565 the mode attribute instead of GET_MODE.
7567 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7568 Alan Hayward <alan.hayward@arm.com>
7569 David Sherwood <david.sherwood@arm.com>
7571 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7572 x exists before using it.
7573 (aarch64_add_constant_internal): Rename to...
7574 (aarch64_add_offset_1): ...this. Replace regnum with separate
7575 src and dest rtxes. Handle the case in which they're different,
7576 including when the offset is zero. Replace scratchreg with an rtx.
7577 Use 2 additions if there is no spare register into which we can
7578 move a 16-bit constant.
7579 (aarch64_add_constant): Delete.
7580 (aarch64_add_offset): Replace reg with separate src and dest
7581 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7582 Use aarch64_add_offset_1.
7583 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7584 an rtx rather than an int. Take the delta as a poly_int64
7585 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7586 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7587 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7588 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7589 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7591 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7592 aarch64_add_constant.
7594 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7596 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7597 Use scalar_float_mode.
7599 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7601 * config/aarch64/aarch64-simd.md
7602 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7603 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7604 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7605 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7606 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7607 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7608 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7609 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7610 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7611 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7613 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7616 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7617 targ_options->x_arm_arch_string is non NULL.
7619 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7621 * config/aarch64/aarch64.h
7622 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7624 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7627 * expmed.c (emit_store_flag_force): Swap if const op0
7628 and change VOIDmode to mode of op0.
7630 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7632 PR rtl-optimization/83761
7633 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7634 than bytes to mode_for_size.
7636 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7639 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7640 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7643 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7646 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7647 when in layout mode.
7648 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7649 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7652 2018-01-10 Michael Collison <michael.collison@arm.com>
7654 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7655 * config/aarch64/aarch64-option-extension.def: Add
7656 AARCH64_OPT_EXTENSION of 'fp16fml'.
7657 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7658 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7659 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7660 * config/aarch64/constraints.md (Ui7): New constraint.
7661 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7662 (VFMLA_SEL_W): Ditto.
7665 (VFMLA16_LOW): New int iterator.
7666 (VFMLA16_HIGH): Ditto.
7667 (UNSPEC_FMLAL): New unspec.
7668 (UNSPEC_FMLSL): Ditto.
7669 (UNSPEC_FMLAL2): Ditto.
7670 (UNSPEC_FMLSL2): Ditto.
7671 (f16mac): New code attribute.
7672 * config/aarch64/aarch64-simd-builtins.def
7673 (aarch64_fmlal_lowv2sf): Ditto.
7674 (aarch64_fmlsl_lowv2sf): Ditto.
7675 (aarch64_fmlalq_lowv4sf): Ditto.
7676 (aarch64_fmlslq_lowv4sf): Ditto.
7677 (aarch64_fmlal_highv2sf): Ditto.
7678 (aarch64_fmlsl_highv2sf): Ditto.
7679 (aarch64_fmlalq_highv4sf): Ditto.
7680 (aarch64_fmlslq_highv4sf): Ditto.
7681 (aarch64_fmlal_lane_lowv2sf): Ditto.
7682 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7683 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7684 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7685 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7686 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7687 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7688 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7689 (aarch64_fmlal_lane_highv2sf): Ditto.
7690 (aarch64_fmlsl_lane_highv2sf): Ditto.
7691 (aarch64_fmlal_laneq_highv2sf): Ditto.
7692 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7693 (aarch64_fmlalq_lane_highv4sf): Ditto.
7694 (aarch64_fmlsl_lane_highv4sf): Ditto.
7695 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7696 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7697 * config/aarch64/aarch64-simd.md:
7698 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7699 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7700 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7701 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7702 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7703 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7704 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7705 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7706 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7707 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7708 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7709 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7710 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7711 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7712 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7713 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7714 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7715 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7716 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7717 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7718 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7719 (vfmlsl_low_u32): Ditto.
7720 (vfmlalq_low_u32): Ditto.
7721 (vfmlslq_low_u32): Ditto.
7722 (vfmlal_high_u32): Ditto.
7723 (vfmlsl_high_u32): Ditto.
7724 (vfmlalq_high_u32): Ditto.
7725 (vfmlslq_high_u32): Ditto.
7726 (vfmlal_lane_low_u32): Ditto.
7727 (vfmlsl_lane_low_u32): Ditto.
7728 (vfmlal_laneq_low_u32): Ditto.
7729 (vfmlsl_laneq_low_u32): Ditto.
7730 (vfmlalq_lane_low_u32): Ditto.
7731 (vfmlslq_lane_low_u32): Ditto.
7732 (vfmlalq_laneq_low_u32): Ditto.
7733 (vfmlslq_laneq_low_u32): Ditto.
7734 (vfmlal_lane_high_u32): Ditto.
7735 (vfmlsl_lane_high_u32): Ditto.
7736 (vfmlal_laneq_high_u32): Ditto.
7737 (vfmlsl_laneq_high_u32): Ditto.
7738 (vfmlalq_lane_high_u32): Ditto.
7739 (vfmlslq_lane_high_u32): Ditto.
7740 (vfmlalq_laneq_high_u32): Ditto.
7741 (vfmlslq_laneq_high_u32): Ditto.
7742 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7743 (AARCH64_FL_FOR_ARCH8_4): New.
7744 (AARCH64_ISA_F16FML): New ISA flag.
7745 (TARGET_F16FML): New feature flag for fp16fml.
7746 (doc/invoke.texi): Document new fp16fml option.
7748 2018-01-10 Michael Collison <michael.collison@arm.com>
7750 * config/aarch64/aarch64-builtins.c:
7751 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7752 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7753 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7754 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7755 (AARCH64_ISA_SHA3): New ISA flag.
7756 (TARGET_SHA3): New feature flag for sha3.
7757 * config/aarch64/iterators.md (sha512_op): New int attribute.
7758 (CRYPTO_SHA512): New int iterator.
7759 (UNSPEC_SHA512H): New unspec.
7760 (UNSPEC_SHA512H2): Ditto.
7761 (UNSPEC_SHA512SU0): Ditto.
7762 (UNSPEC_SHA512SU1): Ditto.
7763 * config/aarch64/aarch64-simd-builtins.def
7764 (aarch64_crypto_sha512hqv2di): New builtin.
7765 (aarch64_crypto_sha512h2qv2di): Ditto.
7766 (aarch64_crypto_sha512su0qv2di): Ditto.
7767 (aarch64_crypto_sha512su1qv2di): Ditto.
7768 (aarch64_eor3qv8hi): Ditto.
7769 (aarch64_rax1qv2di): Ditto.
7770 (aarch64_xarqv2di): Ditto.
7771 (aarch64_bcaxqv8hi): Ditto.
7772 * config/aarch64/aarch64-simd.md:
7773 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7774 (aarch64_crypto_sha512su0qv2di): Ditto.
7775 (aarch64_crypto_sha512su1qv2di): Ditto.
7776 (aarch64_eor3qv8hi): Ditto.
7777 (aarch64_rax1qv2di): Ditto.
7778 (aarch64_xarqv2di): Ditto.
7779 (aarch64_bcaxqv8hi): Ditto.
7780 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7781 (vsha512h2q_u64): Ditto.
7782 (vsha512su0q_u64): Ditto.
7783 (vsha512su1q_u64): Ditto.
7784 (veor3q_u16): Ditto.
7785 (vrax1q_u64): Ditto.
7787 (vbcaxq_u16): Ditto.
7788 * config/arm/types.md (crypto_sha512): New type attribute.
7789 (crypto_sha3): Ditto.
7790 (doc/invoke.texi): Document new sha3 option.
7792 2018-01-10 Michael Collison <michael.collison@arm.com>
7794 * config/aarch64/aarch64-builtins.c:
7795 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7796 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7797 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7798 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7799 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7800 (AARCH64_ISA_SM4): New ISA flag.
7801 (TARGET_SM4): New feature flag for sm4.
7802 * config/aarch64/aarch64-simd-builtins.def
7803 (aarch64_sm3ss1qv4si): Ditto.
7804 (aarch64_sm3tt1aq4si): Ditto.
7805 (aarch64_sm3tt1bq4si): Ditto.
7806 (aarch64_sm3tt2aq4si): Ditto.
7807 (aarch64_sm3tt2bq4si): Ditto.
7808 (aarch64_sm3partw1qv4si): Ditto.
7809 (aarch64_sm3partw2qv4si): Ditto.
7810 (aarch64_sm4eqv4si): Ditto.
7811 (aarch64_sm4ekeyqv4si): Ditto.
7812 * config/aarch64/aarch64-simd.md:
7813 (aarch64_sm3ss1qv4si): Ditto.
7814 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7815 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7816 (aarch64_sm4eqv4si): Ditto.
7817 (aarch64_sm4ekeyqv4si): Ditto.
7818 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7819 (sm3part_op): Ditto.
7820 (CRYPTO_SM3TT): Ditto.
7821 (CRYPTO_SM3PART): Ditto.
7822 (UNSPEC_SM3SS1): New unspec.
7823 (UNSPEC_SM3TT1A): Ditto.
7824 (UNSPEC_SM3TT1B): Ditto.
7825 (UNSPEC_SM3TT2A): Ditto.
7826 (UNSPEC_SM3TT2B): Ditto.
7827 (UNSPEC_SM3PARTW1): Ditto.
7828 (UNSPEC_SM3PARTW2): Ditto.
7829 (UNSPEC_SM4E): Ditto.
7830 (UNSPEC_SM4EKEY): Ditto.
7831 * config/aarch64/constraints.md (Ui2): New constraint.
7832 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7833 * config/arm/types.md (crypto_sm3): New type attribute.
7834 (crypto_sm4): Ditto.
7835 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7836 (vsm3tt1aq_u32): Ditto.
7837 (vsm3tt1bq_u32): Ditto.
7838 (vsm3tt2aq_u32): Ditto.
7839 (vsm3tt2bq_u32): Ditto.
7840 (vsm3partw1q_u32): Ditto.
7841 (vsm3partw2q_u32): Ditto.
7842 (vsm4eq_u32): Ditto.
7843 (vsm4ekeyq_u32): Ditto.
7844 (doc/invoke.texi): Document new sm4 option.
7846 2018-01-10 Michael Collison <michael.collison@arm.com>
7848 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7849 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7850 (AARCH64_FL_FOR_ARCH8_4): New.
7851 (AARCH64_FL_V8_4): New flag.
7852 (doc/invoke.texi): Document new armv8.4-a option.
7854 2018-01-10 Michael Collison <michael.collison@arm.com>
7856 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7857 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7858 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7859 * config/aarch64/aarch64-option-extension.def: Add
7860 AARCH64_OPT_EXTENSION of 'sha2'.
7861 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7862 (crypto): Disable sha2 and aes if crypto disabled.
7863 (crypto): Enable aes and sha2 if enabled.
7864 (simd): Disable sha2 and aes if simd disabled.
7865 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7867 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7868 (TARGET_SHA2): New feature flag for sha2.
7869 (TARGET_AES): New feature flag for aes.
7870 * config/aarch64/aarch64-simd.md:
7871 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7872 conditional on TARGET_AES.
7873 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7874 (aarch64_crypto_sha1hsi): Make pattern conditional
7876 (aarch64_crypto_sha1hv4si): Ditto.
7877 (aarch64_be_crypto_sha1hv4si): Ditto.
7878 (aarch64_crypto_sha1su1v4si): Ditto.
7879 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7880 (aarch64_crypto_sha1su0v4si): Ditto.
7881 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7882 (aarch64_crypto_sha256su0v4si): Ditto.
7883 (aarch64_crypto_sha256su1v4si): Ditto.
7884 (doc/invoke.texi): Document new aes and sha2 options.
7886 2018-01-10 Martin Sebor <msebor@redhat.com>
7888 PR tree-optimization/83781
7889 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7892 2018-01-11 Martin Sebor <msebor@gmail.com>
7893 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7895 PR tree-optimization/83501
7896 PR tree-optimization/81703
7898 * tree-ssa-strlen.c (get_string_cst): Rename...
7899 (get_string_len): ...to this. Handle global constants.
7900 (handle_char_store): Adjust.
7902 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7903 Jim Wilson <jimw@sifive.com>
7905 * config/riscv/riscv-protos.h (riscv_output_return): New.
7906 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7907 (riscv_attribute_table, riscv_output_return),
7908 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7909 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7910 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7912 (riscv_expand_prologue): Add early return for naked function.
7913 (riscv_expand_epilogue): Likewise.
7914 (riscv_function_ok_for_sibcall): Return false for naked function.
7915 (riscv_set_current_function): New.
7916 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7917 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
7918 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
7919 * doc/extend.texi (RISC-V Function Attributes): New.
7921 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7923 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
7924 check for 128-bit long double before checking TCmode.
7925 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
7926 128-bit long doubles before checking TFmode or TCmode.
7927 (FLOAT128_IBM_P): Likewise.
7929 2018-01-10 Martin Sebor <msebor@redhat.com>
7931 PR tree-optimization/83671
7932 * builtins.c (c_strlen): Unconditionally return zero for the empty
7934 Use -Warray-bounds for warnings.
7935 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
7936 for non-constant array indices with COMPONENT_REF, arrays of
7937 arrays, and pointers to arrays.
7938 (gimple_fold_builtin_strlen): Determine and set length range for
7939 non-constant character arrays.
7941 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
7944 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
7947 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
7949 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
7951 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7954 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
7955 VECTOR_MEM_ALTIVEC_OR_VSX_P.
7956 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
7957 indexed_or_indirect_operand predicate.
7958 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
7959 (*vsx_le_perm_load_v8hi): Likewise.
7960 (*vsx_le_perm_load_v16qi): Likewise.
7961 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
7962 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
7963 (*vsx_le_perm_store_v8hi): Likewise.
7964 (*vsx_le_perm_store_v16qi): Likewise.
7965 (eight unnamed splitters): Likewise.
7967 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
7969 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
7970 * config/rs6000/emmintrin.h: Likewise.
7971 * config/rs6000/mmintrin.h: Likewise.
7972 * config/rs6000/xmmintrin.h: Likewise.
7974 2018-01-10 David Malcolm <dmalcolm@redhat.com>
7977 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
7979 * tree.c (tree_nop_conversion): Return true for location wrapper
7981 (maybe_wrap_with_location): New function.
7982 (selftest::check_strip_nops): New function.
7983 (selftest::test_location_wrappers): New function.
7984 (selftest::tree_c_tests): Call it.
7985 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
7986 (maybe_wrap_with_location): New decl.
7987 (EXPR_LOCATION_WRAPPER_P): New macro.
7988 (location_wrapper_p): New inline function.
7989 (tree_strip_any_location_wrapper): New inline function.
7991 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
7994 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
7995 stack_realign_offset for the largest alignment of stack slot
7997 (ix86_find_max_used_stack_alignment): New function.
7998 (ix86_finalize_stack_frame_flags): Use it. Set
7999 max_used_stack_alignment if we don't realign stack.
8000 * config/i386/i386.h (machine_function): Add
8001 max_used_stack_alignment.
8003 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8005 * config/arm/arm.opt (-mbranch-cost): New option.
8006 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8009 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8012 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8013 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8015 2018-01-10 Richard Biener <rguenther@suse.de>
8018 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8019 early out so it also covers the case where we have a non-NULL
8022 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8024 PR tree-optimization/83753
8025 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8026 for non-strided grouped accesses if the number of elements is 1.
8028 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8031 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8032 * i386.h (TARGET_USE_GATHER): Define.
8033 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8035 2018-01-10 Martin Liska <mliska@suse.cz>
8038 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8039 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8041 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8042 CLEANUP_NO_PARTITIONING is not set.
8044 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8046 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8047 for vectors, as a partial revert of r254296.
8048 * rtl.h (const_vec_p): Delete.
8049 (const_vec_duplicate_p): Don't test for vector CONSTs.
8050 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8051 * expmed.c (make_tree): Likewise.
8054 * common.md (E, F): Use CONSTANT_P instead of checking for
8056 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8057 checking for CONST_VECTOR.
8059 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8062 * predict.c (force_edge_cold): Handle in more sane way edges
8065 2018-01-09 Carl Love <cel@us.ibm.com>
8067 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8069 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8070 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8071 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8072 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8073 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8074 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8075 * config/rs6000/rs6000-protos.h: Add extern defition for
8076 rs6000_generate_float2_double_code.
8077 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8079 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8080 (float2_v2df): Add define_expand.
8082 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8085 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8086 op_mode in the force_to_mode call.
8088 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8090 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8091 instead of checking each element individually.
8092 (aarch64_evpc_uzp): Likewise.
8093 (aarch64_evpc_zip): Likewise.
8094 (aarch64_evpc_ext): Likewise.
8095 (aarch64_evpc_rev): Likewise.
8096 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8097 instead of checking each element individually. Return true without
8099 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8100 whether all selected elements come from the same input, instead of
8101 checking each element individually. Remove calls to gen_rtx_REG,
8102 start_sequence and end_sequence and instead assert that no rtl is
8105 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8107 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8108 order of HIGH and CONST checks.
8110 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8112 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8113 if the destination isn't an SSA_NAME.
8115 2018-01-09 Richard Biener <rguenther@suse.de>
8117 PR tree-optimization/83668
8118 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8120 (canonicalize_loop_form): ... here, renamed from ...
8121 (canonicalize_loop_closed_ssa_form): ... this and amended to
8122 swap successor edges for loop exit blocks to make us use
8123 the RPO order we need for initial schedule generation.
8125 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8127 PR tree-optimization/64811
8128 * match.pd: When optimizing comparisons with Inf, avoid
8129 introducing or losing exceptions from comparisons with NaN.
8131 2018-01-09 Martin Liska <mliska@suse.cz>
8134 * asan.c (shadow_mem_size): Add gcc_assert.
8136 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8138 Don't save registers in main().
8141 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8142 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8143 * config/avr/avr.c (avr_set_current_function): Don't error if
8144 naked, OS_task or OS_main are specified at the same time.
8145 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8147 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8149 * common/config/avr/avr-common.c (avr_option_optimization_table):
8150 Switch on -mmain-is-OS_task for optimizing compilations.
8152 2018-01-09 Richard Biener <rguenther@suse.de>
8154 PR tree-optimization/83572
8155 * graphite.c: Include cfganal.h.
8156 (graphite_transform_loops): Connect infinite loops to exit
8157 and remove fake edges at the end.
8159 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8161 * ipa-inline.c (edge_badness): Revert accidental checkin.
8163 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8166 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8167 symbols; not inline clones.
8169 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8172 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8173 hard registers. Formatting fixes.
8175 PR preprocessor/83722
8176 * gcc.c (try_generate_repro): Pass
8177 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8178 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8181 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8182 Kito Cheng <kito.cheng@gmail.com>
8184 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8185 (riscv_leaf_function_p): Delete.
8186 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8188 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8190 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8192 (do_ifelse): New function.
8193 (do_isel): New function.
8194 (do_sub3): New function.
8195 (do_add3): New function.
8196 (do_load_mask_compare): New function.
8197 (do_overlap_load_compare): New function.
8198 (expand_compare_loop): New function.
8199 (expand_block_compare): Call expand_compare_loop() when appropriate.
8200 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8202 (-mblock-compare-inline-loop-limit): New option.
8204 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8207 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8208 Reverse order of second and third operands in first alternative.
8209 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8210 of first and second elements in UNSPEC_VPERMR vector.
8211 (altivec_expand_vec_perm_le): Likewise.
8213 2018-01-08 Jeff Law <law@redhat.com>
8215 PR rtl-optimizatin/81308
8216 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8217 (process_switch): If group_case_labels makes a change, then set
8219 (pass_convert_switch::execute): If a switch is converted, then
8220 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8222 PR rtl-optimization/81308
8223 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8226 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8228 PR target/83663 - Revert r255946
8229 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8230 generation for cases where splatting a value is not useful.
8231 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8232 across a vec_duplicate and a paradoxical subreg forming a vector
8233 mode to a vec_concat.
8235 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8237 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8238 -march=armv8.3-a variants.
8239 * config/arm/t-multilib: Likewise.
8240 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8242 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8244 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8246 (cceq_ior_compare_complement): Give it a name so I can use it, and
8247 change boolean_or_operator predicate to boolean_operator so it can
8248 be used to generate a crand.
8249 (eqne): New code iterator.
8250 (bd/bd_neg): New code_attrs.
8251 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8252 a single define_insn.
8253 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8254 decrement (bdnzt/bdnzf/bdzt/bdzf).
8255 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8256 with the new names of the branch decrement patterns, and added the
8257 names of the branch decrement conditional patterns.
8259 2018-01-08 Richard Biener <rguenther@suse.de>
8261 PR tree-optimization/83563
8262 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8265 2018-01-08 Richard Biener <rguenther@suse.de>
8268 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8270 2018-01-08 Richard Biener <rguenther@suse.de>
8272 PR tree-optimization/83685
8273 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8274 references to abnormals.
8276 2018-01-08 Richard Biener <rguenther@suse.de>
8279 * dwarf2out.c (output_indirect_strings): Handle empty
8280 skeleton_debug_str_hash.
8281 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8283 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8285 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8286 (emit_store_direct): Likewise.
8287 (arc_trampoline_adjust_address): Likewise.
8288 (arc_asm_trampoline_template): New function.
8289 (arc_initialize_trampoline): Use asm_trampoline_template.
8290 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8291 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8292 * config/arc/arc.md (flush_icache): Delete pattern.
8294 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8296 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8297 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8300 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8303 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8304 by not USED_FOR_TARGET.
8305 (make_pass_resolve_sw_modes): Likewise.
8307 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8309 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8312 2018-01-08 Richard Biener <rguenther@suse.de>
8315 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8317 2018-01-08 Richard Biener <rguenther@suse.de>
8320 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8322 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8325 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8326 basic blocks with a small number of successors.
8327 (convert_control_dep_chain_into_preds): Improve handling of
8329 (dump_predicates): Split apart into...
8330 (dump_pred_chain): ...here...
8331 (dump_pred_info): ...and here.
8332 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8333 (can_chain_union_be_invalidated_p): Improve check for invalidation
8335 (uninit_uses_cannot_happen): Avoid unnecessary if
8336 convert_control_dep_chain_into_preds yielded nothing.
8338 2018-01-06 Martin Sebor <msebor@redhat.com>
8340 PR tree-optimization/83640
8341 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8342 subtracting negative offset from size.
8343 (builtin_access::overlap): Adjust offset bounds of the access to fall
8344 within the size of the object if possible.
8346 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8348 PR rtl-optimization/83699
8349 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8350 extract_bit_field_as_subreg to cases in which the extracted
8351 value is also a vector.
8353 * lra-constraints.c (process_alt_operands): Test for the equivalence
8354 substitutions when detecting a possible reload cycle.
8356 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8359 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8360 by default if flag_selective_schedling{,2}. Formatting fixes.
8362 PR rtl-optimization/83682
8363 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8364 if it has non-VECTOR_MODE element mode.
8365 (vec_duplicate_p): Likewise.
8368 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8369 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8371 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8374 * config/i386/i386-builtin.def
8375 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8376 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8377 Require also OPTION_MASK_ISA_AVX512F in addition to
8378 OPTION_MASK_ISA_GFNI.
8379 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8380 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8381 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8382 to OPTION_MASK_ISA_GFNI.
8383 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8384 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8385 OPTION_MASK_ISA_AVX512BW.
8386 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8387 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8388 addition to OPTION_MASK_ISA_GFNI.
8389 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8390 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8391 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8392 to OPTION_MASK_ISA_GFNI.
8393 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8394 a requirement for all ISAs rather than any of them with a few
8396 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8398 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8399 bitmasks to be enabled with 3 exceptions, instead of requiring any
8400 enabled ISA with lots of exceptions.
8401 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8402 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8403 Change avx512bw in isa attribute to avx512f.
8404 * config/i386/sgxintrin.h: Add license boilerplate.
8405 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8406 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8407 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8408 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8410 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8411 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8412 temporarily sse2 rather than sse if not enabled already.
8415 * config/i386/sse.md (VI248_VLBW): Rename to ...
8416 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8417 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8418 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8419 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8420 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8421 mode iterator instead of VI248_VLBW.
8423 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8425 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8426 (record_modified): Skip clobbers; add debug output.
8427 (param_change_prob): Use sreal frequencies.
8429 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8431 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8432 punt for user-aligned variables.
8434 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8436 * tree-chrec.c (chrec_contains_symbols): Return true for
8439 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8442 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8443 of (x|y) == x for BICS pattern.
8445 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8447 PR tree-optimization/83605
8448 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8449 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8452 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8454 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8455 * config/epiphany/rtems.h: New file.
8457 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8458 Uros Bizjak <ubizjak@gmail.com>
8461 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8462 QIreg_operand instead of register_operand predicate.
8463 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8464 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8465 comments instead of -fmitigate[-_]rop.
8467 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8470 * cgraphunit.c (symbol_table::compile): Switch to text_section
8471 before calling assembly_start debug hook.
8472 * run-rtl-passes.c (run_rtl_passes): Likewise.
8475 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8477 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8478 range_int_cst_p rather than !symbolic_range_p before calling
8479 extract_range_from_multiplicative_op_1.
8481 2018-01-04 Jeff Law <law@redhat.com>
8483 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8484 redundant test in assertion.
8486 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8488 * doc/rtl.texi: Document machine_mode wrapper classes.
8490 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8492 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8495 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8497 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8498 the VEC_PERM_EXPR fold to fail.
8500 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8503 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8504 to switched_sections.
8506 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8509 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8512 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8515 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8516 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8518 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8521 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8522 is BLKmode and bitpos not zero or mode change is needed.
8524 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8527 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8530 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8533 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8534 instead of MULT rtx. Update all corresponding splitters.
8536 (*ssub<modesuffix>): Ditto.
8538 (*cmp_sadd_di): Update split patterns.
8539 (*cmp_sadd_si): Ditto.
8540 (*cmp_sadd_sidi): Ditto.
8541 (*cmp_ssub_di): Ditto.
8542 (*cmp_ssub_si): Ditto.
8543 (*cmp_ssub_sidi): Ditto.
8544 * config/alpha/predicates.md (const23_operand): New predicate.
8545 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8546 Look for ASHIFT, not MULT inner operand.
8547 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8549 2018-01-04 Martin Liska <mliska@suse.cz>
8551 PR gcov-profile/83669
8552 * gcov.c (output_intermediate_file): Add version to intermediate
8554 * doc/gcov.texi: Document new field 'version' in intermediate
8555 file format. Fix location of '-k' option of gcov command.
8557 2018-01-04 Martin Liska <mliska@suse.cz>
8560 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8562 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8564 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8566 2018-01-03 Martin Sebor <msebor@redhat.com>
8568 PR tree-optimization/83655
8569 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8570 checking calls with invalid arguments.
8572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8574 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8575 (vectorizable_mask_load_store): Delete.
8576 (vectorizable_call): Return false for masked loads and stores.
8577 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8578 instead of gimple_assign_rhs1.
8579 (vectorizable_load): Handle IFN_MASK_LOAD.
8580 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8584 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8586 (vectorizable_mask_load_store): ...here.
8587 (vectorizable_load): ...and here.
8589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8591 * tree-vect-stmts.c (vect_build_all_ones_mask)
8592 (vect_build_zero_merge_argument): New functions, split out from...
8593 (vectorizable_load): ...here.
8595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8597 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8599 (vectorizable_mask_load_store): ...here.
8600 (vectorizable_store): ...and here.
8602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8604 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8606 (vectorizable_mask_load_store): ...here.
8608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8610 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8611 (vect_model_store_cost): Take a vec_load_store_type instead of a
8613 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8614 (vect_model_store_cost): Take a vec_load_store_type instead of a
8616 (vectorizable_mask_load_store): Update accordingly.
8617 (vectorizable_store): Likewise.
8618 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8622 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8623 IFN_MASK_LOAD calls here rather than...
8624 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8627 Alan Hayward <alan.hayward@arm.com>
8628 David Sherwood <david.sherwood@arm.com>
8630 * expmed.c (extract_bit_field_1): For vector extracts,
8631 fall back to extract_bit_field_as_subreg if vec_extract
8634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8635 Alan Hayward <alan.hayward@arm.com>
8636 David Sherwood <david.sherwood@arm.com>
8638 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8639 they are variable or constant sized.
8640 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8641 slots for constant-sized data.
8643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8644 Alan Hayward <alan.hayward@arm.com>
8645 David Sherwood <david.sherwood@arm.com>
8647 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8648 handling COND_EXPRs with boolean comparisons, try to find a better
8649 basis for the mask type than the boolean itself.
8651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8653 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8654 is calculated and how it can be overridden.
8655 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8656 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8658 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8662 Alan Hayward <alan.hayward@arm.com>
8663 David Sherwood <david.sherwood@arm.com>
8665 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8666 Remove the mode argument.
8667 (aarch64_simd_valid_immediate): Remove the mode and inverse
8669 * config/aarch64/iterators.md (bitsize): New iterator.
8670 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8671 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8672 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8673 aarch64_simd_valid_immediate.
8674 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8675 (aarch64_reg_or_bic_imm): Likewise.
8676 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8677 with an insn_type enum and msl with a modifier_type enum.
8678 Replace element_width with a scalar_mode. Change the shift
8679 to unsigned int. Add constructors for scalar_float_mode and
8680 scalar_int_mode elements.
8681 (aarch64_vect_float_const_representable_p): Delete.
8682 (aarch64_can_const_movi_rtx_p)
8683 (aarch64_simd_scalar_immediate_valid_for_move)
8684 (aarch64_simd_make_constant): Update call to
8685 aarch64_simd_valid_immediate.
8686 (aarch64_advsimd_valid_immediate_hs): New function.
8687 (aarch64_advsimd_valid_immediate): Likewise.
8688 (aarch64_simd_valid_immediate): Remove mode and inverse
8689 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8690 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8691 and aarch64_float_const_representable_p on the result.
8692 (aarch64_output_simd_mov_immediate): Remove mode argument.
8693 Update call to aarch64_simd_valid_immediate and use of
8694 simd_immediate_info.
8695 (aarch64_output_scalar_simd_mov_immediate): Update call
8698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8699 Alan Hayward <alan.hayward@arm.com>
8700 David Sherwood <david.sherwood@arm.com>
8702 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8703 (mode_nunits): Likewise CONST_MODE_NUNITS.
8704 * machmode.def (ADJUST_NUNITS): Document.
8705 * genmodes.c (mode_data::need_nunits_adj): New field.
8706 (blank_mode): Update accordingly.
8707 (adj_nunits): New variable.
8708 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8710 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8711 listed in adj_nunits.
8712 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8713 listed in adj_nunits. Don't emit case statements for such modes.
8714 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8715 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8716 nothing if adj_nunits is nonnull.
8717 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8718 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8719 (emit_mode_fbit): Update use of print_maybe_const_decl.
8720 (emit_move_size): Likewise. Treat the array as non-const
8722 (emit_mode_adjustments): Handle adj_nunits.
8724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8726 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8727 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8728 (VECTOR_MODES): Use it.
8729 (make_vector_modes): Take the prefix as an argument.
8731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8732 Alan Hayward <alan.hayward@arm.com>
8733 David Sherwood <david.sherwood@arm.com>
8735 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8736 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8737 for MODE_VECTOR_BOOL.
8738 * machmode.def (VECTOR_BOOL_MODE): Document.
8739 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8740 (make_vector_bool_mode): New function.
8741 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8743 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8744 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8746 * stor-layout.c (int_mode_for_mode): Likewise.
8747 * tree.c (build_vector_type_for_mode): Likewise.
8748 * varasm.c (output_constant_pool_2): Likewise.
8749 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8750 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8751 for MODE_VECTOR_BOOL.
8752 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8753 of mode class checks.
8754 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8755 instead of a list of mode class checks.
8756 (expand_vector_scalar_condition): Likewise.
8757 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8760 Alan Hayward <alan.hayward@arm.com>
8761 David Sherwood <david.sherwood@arm.com>
8763 * machmode.h (mode_size): Change from unsigned short to
8765 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8766 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8767 or if measurement_type is not polynomial.
8768 (fixed_size_mode::includes_p): Check for constant-sized modes.
8769 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8770 return a poly_uint16 rather than an unsigned short.
8771 (emit_mode_size): Change the type of mode_size from unsigned short
8772 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8773 (emit_mode_adjustments): Cope with polynomial vector sizes.
8774 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8776 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8778 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8779 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8780 * caller-save.c (setup_save_areas): Likewise.
8781 (replace_reg_with_saved_mem): Likewise.
8782 * calls.c (emit_library_call_value_1): Likewise.
8783 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8784 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8785 (gen_lowpart_for_combine): Likewise.
8786 * convert.c (convert_to_integer_1): Likewise.
8787 * cse.c (equiv_constant, cse_insn): Likewise.
8788 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8789 (cselib_subst_to_values): Likewise.
8790 * dce.c (word_dce_process_block): Likewise.
8791 * df-problems.c (df_word_lr_mark_ref): Likewise.
8792 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8793 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8794 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8795 (rtl_for_decl_location): Likewise.
8796 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8797 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8798 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8799 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8800 (expand_expr_real_1): Likewise.
8801 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8802 (pad_below): Likewise.
8803 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8804 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8805 * ira.c (get_subreg_tracking_sizes): Likewise.
8806 * ira-build.c (ira_create_allocno_objects): Likewise.
8807 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8808 (ira_sort_regnos_for_alter_reg): Likewise.
8809 * ira-costs.c (record_operand_costs): Likewise.
8810 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8811 (resolve_simple_move): Likewise.
8812 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8813 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8814 (lra_constraints): Likewise.
8815 (CONST_POOL_OK_P): Reject variable-sized modes.
8816 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8817 (add_pseudo_to_slot, lra_spill): Likewise.
8818 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8819 * optabs-query.c (get_best_extraction_insn): Likewise.
8820 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8821 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8822 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8823 * recog.c (offsettable_address_addr_space_p): Likewise.
8824 * regcprop.c (maybe_mode_change): Likewise.
8825 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8826 * regrename.c (build_def_use): Likewise.
8827 * regstat.c (dump_reg_info): Likewise.
8828 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8829 (find_reloads, find_reloads_subreg_address): Likewise.
8830 * reload1.c (eliminate_regs_1): Likewise.
8831 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8832 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8833 (simplify_binary_operation_1, simplify_subreg): Likewise.
8834 * targhooks.c (default_function_arg_padding): Likewise.
8835 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8836 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8837 (verify_gimple_assign_ternary): Likewise.
8838 * tree-inline.c (estimate_move_cost): Likewise.
8839 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8840 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8841 (get_address_cost_ainc): Likewise.
8842 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8843 (vect_supportable_dr_alignment): Likewise.
8844 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8845 (vectorizable_reduction): Likewise.
8846 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8847 (vectorizable_operation, vectorizable_load): Likewise.
8848 * tree.c (build_same_sized_truth_vector_type): Likewise.
8849 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8850 * var-tracking.c (emit_note_insn_var_location): Likewise.
8851 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8852 (ADDR_VEC_ALIGN): Likewise.
8854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8855 Alan Hayward <alan.hayward@arm.com>
8856 David Sherwood <david.sherwood@arm.com>
8858 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8860 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8861 or if measurement_type is polynomial.
8862 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8863 * combine.c (make_extraction): Likewise.
8864 * dse.c (find_shift_sequence): Likewise.
8865 * dwarf2out.c (mem_loc_descriptor): Likewise.
8866 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8867 (extract_bit_field, extract_low_bits): Likewise.
8868 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8869 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8870 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8871 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8872 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8873 * reload.c (find_reloads): Likewise.
8874 * reload1.c (alter_reg): Likewise.
8875 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8876 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8877 * tree-if-conv.c (predicate_mem_writes): Likewise.
8878 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8879 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8880 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8881 * valtrack.c (dead_debug_insert_temp): Likewise.
8882 * varasm.c (mergeable_constant_section): Likewise.
8883 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8886 Alan Hayward <alan.hayward@arm.com>
8887 David Sherwood <david.sherwood@arm.com>
8889 * expr.c (expand_assignment): Cope with polynomial mode sizes
8890 when assigning to a CONCAT.
8892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8893 Alan Hayward <alan.hayward@arm.com>
8894 David Sherwood <david.sherwood@arm.com>
8896 * machmode.h (mode_precision): Change from unsigned short to
8898 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8900 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8901 or if measurement_type is not polynomial.
8902 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8903 in which the mode is already known to be a scalar_int_mode.
8904 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8905 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8907 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8908 for GET_MODE_PRECISION.
8909 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8910 for GET_MODE_PRECISION.
8911 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8913 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8914 (expand_field_assignment, make_extraction): Likewise.
8915 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8916 (get_last_value): Likewise.
8917 * convert.c (convert_to_integer_1): Likewise.
8918 * cse.c (cse_insn): Likewise.
8919 * expr.c (expand_expr_real_1): Likewise.
8920 * lra-constraints.c (simplify_operand_subreg): Likewise.
8921 * optabs-query.c (can_atomic_load_p): Likewise.
8922 * optabs.c (expand_atomic_load): Likewise.
8923 (expand_atomic_store): Likewise.
8924 * ree.c (combine_reaching_defs): Likewise.
8925 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
8926 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
8927 * tree.h (type_has_mode_precision_p): Likewise.
8928 * ubsan.c (instrument_si_overflow): Likewise.
8930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8931 Alan Hayward <alan.hayward@arm.com>
8932 David Sherwood <david.sherwood@arm.com>
8934 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
8935 polynomial numbers of units.
8936 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
8937 (valid_vector_subparts_p): New function.
8938 (build_vector_type): Remove temporary shim and take the number
8939 of units as a poly_uint64 rather than an int.
8940 (build_opaque_vector_type): Take the number of units as a
8941 poly_uint64 rather than an int.
8942 * tree.c (build_vector_from_ctor): Handle polynomial
8943 TYPE_VECTOR_SUBPARTS.
8944 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
8945 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
8946 (build_vector_from_val): If the number of units is variable,
8947 use build_vec_duplicate_cst for constant operands and
8948 VEC_DUPLICATE_EXPR otherwise.
8949 (make_vector_type): Remove temporary is_constant ().
8950 (build_vector_type, build_opaque_vector_type): Take the number of
8951 units as a poly_uint64 rather than an int.
8952 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
8954 * cfgexpand.c (expand_debug_expr): Likewise.
8955 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
8956 (store_constructor, expand_expr_real_1): Likewise.
8957 (const_scalar_mask_from_tree): Likewise.
8958 * fold-const-call.c (fold_const_reduction): Likewise.
8959 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
8960 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
8961 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
8962 (fold_relational_const): Likewise.
8963 (native_interpret_vector): Likewise. Change the size from an
8964 int to an unsigned int.
8965 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
8966 TYPE_VECTOR_SUBPARTS.
8967 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
8968 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
8969 duplicating a non-constant operand into a variable-length vector.
8970 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
8971 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
8972 * ipa-icf.c (sem_variable::equals): Likewise.
8973 * match.pd: Likewise.
8974 * omp-simd-clone.c (simd_clone_subparts): Likewise.
8975 * print-tree.c (print_node): Likewise.
8976 * stor-layout.c (layout_type): Likewise.
8977 * targhooks.c (default_builtin_vectorization_cost): Likewise.
8978 * tree-cfg.c (verify_gimple_comparison): Likewise.
8979 (verify_gimple_assign_binary): Likewise.
8980 (verify_gimple_assign_ternary): Likewise.
8981 (verify_gimple_assign_single): Likewise.
8982 * tree-pretty-print.c (dump_generic_node): Likewise.
8983 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8984 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
8985 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
8986 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
8987 (vect_shift_permute_load_chain): Likewise.
8988 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
8989 (expand_vector_condition, optimize_vector_constructor): Likewise.
8990 (lower_vec_perm, get_compute_type): Likewise.
8991 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8992 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
8993 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
8994 (vect_recog_mask_conversion_pattern): Likewise.
8995 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
8996 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
8997 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8998 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
8999 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9000 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9001 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9002 (supportable_widening_operation): Likewise.
9003 (supportable_narrowing_operation): Likewise.
9004 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9006 * varasm.c (output_constant): Likewise.
9008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9009 Alan Hayward <alan.hayward@arm.com>
9010 David Sherwood <david.sherwood@arm.com>
9012 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9013 so that both the length == 3 and length != 3 cases set up their
9014 own permute vectors. Add comments explaining why we know the
9015 number of elements is constant.
9016 (vect_permute_load_chain): Likewise.
9018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9019 Alan Hayward <alan.hayward@arm.com>
9020 David Sherwood <david.sherwood@arm.com>
9022 * machmode.h (mode_nunits): Change from unsigned char to
9024 (ONLY_FIXED_SIZE_MODES): New macro.
9025 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9026 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9027 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9029 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9030 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9031 or if measurement_type is not polynomial.
9032 * genmodes.c (ZERO_COEFFS): New macro.
9033 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9035 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9036 Use ZERO_COEFFS when emitting initializers.
9037 * data-streamer.h (bp_pack_poly_value): New function.
9038 (bp_unpack_poly_value): Likewise.
9039 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9040 for GET_MODE_NUNITS.
9041 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9042 for GET_MODE_NUNITS.
9043 * tree.c (make_vector_type): Remove temporary shim and make
9044 the real function take the number of units as a poly_uint64
9046 (build_vector_type_for_mode): Handle polynomial nunits.
9047 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9048 * emit-rtl.c (const_vec_series_p_1): Likewise.
9049 (gen_rtx_CONST_VECTOR): Likewise.
9050 * fold-const.c (test_vec_duplicate_folding): Likewise.
9051 * genrecog.c (validate_pattern): Likewise.
9052 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9053 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9054 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9055 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9056 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9057 * rtlanal.c (subreg_get_info): Likewise.
9058 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9059 (vect_grouped_load_supported): Likewise.
9060 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9061 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9062 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9063 (simplify_const_unary_operation, simplify_binary_operation_1)
9064 (simplify_const_binary_operation, simplify_ternary_operation)
9065 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9066 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9067 instead of CONST_VECTOR_NUNITS.
9068 * varasm.c (output_constant_pool_2): Likewise.
9069 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9070 explicit-encoded elements in the XVEC for variable-length vectors.
9072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9074 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9077 Alan Hayward <alan.hayward@arm.com>
9078 David Sherwood <david.sherwood@arm.com>
9080 * coretypes.h (fixed_size_mode): Declare.
9081 (fixed_size_mode_pod): New typedef.
9082 * builtins.h (target_builtins::x_apply_args_mode)
9083 (target_builtins::x_apply_result_mode): Change type to
9084 fixed_size_mode_pod.
9085 * builtins.c (apply_args_size, apply_result_size, result_vector)
9086 (expand_builtin_apply_args_1, expand_builtin_apply)
9087 (expand_builtin_return): Update accordingly.
9089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9091 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9092 * cselib.c (cselib_hash_rtx): Likewise.
9093 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9094 CONST_VECTOR encoding.
9096 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9097 Jeff Law <law@redhat.com>
9100 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9101 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9102 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9103 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9106 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9107 explicitly probe *sp in a noreturn function if there were any callee
9108 register saves or frame pointer is needed.
9110 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9113 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9114 BLKmode for ternary, binary or unary expressions.
9117 * var-tracking.c (delete_vta_debug_insn): New inline function.
9118 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9119 insns from get_insns () to NULL instead of each bb separately.
9120 Use delete_vta_debug_insn. No longer static.
9121 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9122 delete_vta_debug_insns callers.
9123 * rtl.h (delete_vta_debug_insns): Declare.
9124 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9125 instead of variable_tracking_main.
9127 2018-01-03 Martin Sebor <msebor@redhat.com>
9129 PR tree-optimization/83603
9130 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9131 arguments past the endof the argument list in functions declared
9132 without a prototype.
9133 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9134 Avoid checking when arguments are null.
9136 2018-01-03 Martin Sebor <msebor@redhat.com>
9139 * doc/extend.texi (attribute const): Fix a typo.
9140 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9141 issuing -Wsuggest-attribute for void functions.
9143 2018-01-03 Martin Sebor <msebor@redhat.com>
9145 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9146 offset_int::from instead of wide_int::to_shwi.
9147 (maybe_diag_overlap): Remove assertion.
9148 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9149 * gimple-ssa-sprintf.c (format_directive): Same.
9150 (parse_directive): Same.
9151 (sprintf_dom_walker::compute_format_length): Same.
9152 (try_substitute_return_value): Same.
9154 2018-01-03 Jeff Law <law@redhat.com>
9157 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9158 non-constant residual for zero at runtime and avoid probing in
9159 that case. Reorganize code for trailing problem to mirror handling
9162 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9164 PR tree-optimization/83501
9165 * tree-ssa-strlen.c (get_string_cst): New.
9166 (handle_char_store): Call get_string_cst.
9168 2018-01-03 Martin Liska <mliska@suse.cz>
9170 PR tree-optimization/83593
9171 * tree-ssa-strlen.c: Include tree-cfg.h.
9172 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9173 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9174 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9176 (strlen_dom_walker::before_dom_children): Call
9177 gimple_purge_dead_eh_edges. Dump tranformation with details
9179 (strlen_dom_walker::before_dom_children): Update call by adding
9180 new argument cleanup_eh.
9181 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9183 2018-01-03 Martin Liska <mliska@suse.cz>
9186 * cif-code.def (VARIADIC_THUNK): New enum value.
9187 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9190 2018-01-03 Jan Beulich <jbeulich@suse.com>
9192 * sse.md (mov<mode>_internal): Tighten condition for when to use
9193 vmovdqu<ssescalarsize> for TI and OI modes.
9195 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9197 Update copyright years.
9199 2018-01-03 Martin Liska <mliska@suse.cz>
9202 * ipa-visibility.c (function_and_variable_visibility): Skip
9203 functions with noipa attribure.
9205 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9207 * gcc.c (process_command): Update copyright notice dates.
9208 * gcov-dump.c (print_version): Ditto.
9209 * gcov.c (print_version): Ditto.
9210 * gcov-tool.c (print_version): Ditto.
9211 * gengtype.c (create_file): Ditto.
9212 * doc/cpp.texi: Bump @copying's copyright year.
9213 * doc/cppinternals.texi: Ditto.
9214 * doc/gcc.texi: Ditto.
9215 * doc/gccint.texi: Ditto.
9216 * doc/gcov.texi: Ditto.
9217 * doc/install.texi: Ditto.
9218 * doc/invoke.texi: Ditto.
9220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9222 * vector-builder.h (vector_builder::m_full_nelts): Change from
9223 unsigned int to poly_uint64.
9224 (vector_builder::full_nelts): Update prototype accordingly.
9225 (vector_builder::new_vector): Likewise.
9226 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9227 (vector_builder::operator ==): Likewise.
9228 (vector_builder::finalize): Likewise.
9229 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9230 Take the number of elements as a poly_uint64 rather than an
9232 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9233 from unsigned int to poly_uint64.
9234 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9235 (vec_perm_indices::new_vector): Likewise.
9236 (vec_perm_indices::length): Likewise.
9237 (vec_perm_indices::nelts_per_input): Likewise.
9238 (vec_perm_indices::input_nelts): Likewise.
9239 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9240 number of elements per input as a poly_uint64 rather than an
9241 unsigned int. Use the original encoding for variable-length
9242 vectors, rather than clamping each individual element.
9243 For the second and subsequent elements in each pattern,
9244 clamp the step and base before clamping their sum.
9245 (vec_perm_indices::series_p): Handle polynomial element counts.
9246 (vec_perm_indices::all_in_range_p): Likewise.
9247 (vec_perm_indices_to_tree): Likewise.
9248 (vec_perm_indices_to_rtx): Likewise.
9249 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9250 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9251 (tree_vector_builder::new_binary_operation): Handle polynomial
9252 element counts. Return false if we need to know the number
9253 of elements at compile time.
9254 * fold-const.c (fold_vec_perm): Punt if the number of elements
9255 isn't known at compile time.
9257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9259 * vec-perm-indices.h (vec_perm_builder): Change element type
9260 from HOST_WIDE_INT to poly_int64.
9261 (vec_perm_indices::element_type): Update accordingly.
9262 (vec_perm_indices::clamp): Handle polynomial element_types.
9263 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9264 (vec_perm_indices::all_in_range_p): Likewise.
9265 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9267 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9268 polynomial vec_perm_indices element types.
9269 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9270 * fold-const.c (fold_vec_perm): Likewise.
9271 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9272 * tree-vect-generic.c (lower_vec_perm): Likewise.
9273 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9274 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9275 element type to HOST_WIDE_INT.
9277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9278 Alan Hayward <alan.hayward@arm.com>
9279 David Sherwood <david.sherwood@arm.com>
9281 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9282 rather than an int. Use plus_constant.
9283 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9284 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9287 Alan Hayward <alan.hayward@arm.com>
9288 David Sherwood <david.sherwood@arm.com>
9290 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9291 a HOST_WIDE_INT to a poly_int64.
9293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9294 Alan Hayward <alan.hayward@arm.com>
9295 David Sherwood <david.sherwood@arm.com>
9297 * calls.c (load_register_parameters): Cope with polynomial
9298 mode sizes. Require a constant size for BLKmode parameters
9299 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9300 forces a parameter to be padded at the lsb end in order to
9301 fill a complete number of words, require the parameter size
9302 to be ordered wrt UNITS_PER_WORD.
9304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9305 Alan Hayward <alan.hayward@arm.com>
9306 David Sherwood <david.sherwood@arm.com>
9308 * reload1.c (spill_stack_slot_width): Change element type
9309 from unsigned int to poly_uint64_pod.
9310 (alter_reg): Treat mode sizes as polynomial.
9312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9313 Alan Hayward <alan.hayward@arm.com>
9314 David Sherwood <david.sherwood@arm.com>
9316 * reload.c (complex_word_subreg_p): New function.
9317 (reload_inner_reg_of_subreg, push_reload): Use it.
9319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9320 Alan Hayward <alan.hayward@arm.com>
9321 David Sherwood <david.sherwood@arm.com>
9323 * lra-constraints.c (process_alt_operands): Reject matched
9324 operands whose sizes aren't ordered.
9325 (match_reload): Refer to this check here.
9327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9328 Alan Hayward <alan.hayward@arm.com>
9329 David Sherwood <david.sherwood@arm.com>
9331 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9332 that the mode size is in the set {1, 2, 4, 8, 16}.
9334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9335 Alan Hayward <alan.hayward@arm.com>
9336 David Sherwood <david.sherwood@arm.com>
9338 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9339 Use plus_constant instead of gen_rtx_PLUS.
9341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9342 Alan Hayward <alan.hayward@arm.com>
9343 David Sherwood <david.sherwood@arm.com>
9345 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9346 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9347 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9348 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9349 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9350 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9351 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9352 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9353 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9354 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9356 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9357 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9358 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9359 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9360 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9361 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9362 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9363 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9364 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9365 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9367 * expr.c (emit_move_resolve_push): Treat the input and result
9368 of PUSH_ROUNDING as a poly_int64.
9369 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9370 (emit_push_insn): Likewise.
9371 * lra-eliminations.c (mark_not_eliminable): Likewise.
9372 * recog.c (push_operand): Likewise.
9373 * reload1.c (elimination_effects): Likewise.
9374 * rtlanal.c (nonzero_bits1): Likewise.
9375 * calls.c (store_one_arg): Likewise. Require the padding to be
9376 known at compile time.
9378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9379 Alan Hayward <alan.hayward@arm.com>
9380 David Sherwood <david.sherwood@arm.com>
9382 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9383 Use plus_constant instead of gen_rtx_PLUS.
9385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9386 Alan Hayward <alan.hayward@arm.com>
9387 David Sherwood <david.sherwood@arm.com>
9389 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9393 Alan Hayward <alan.hayward@arm.com>
9394 David Sherwood <david.sherwood@arm.com>
9396 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9397 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9398 via stack temporaries. Treat the mode size as polynomial too.
9400 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9401 Alan Hayward <alan.hayward@arm.com>
9402 David Sherwood <david.sherwood@arm.com>
9404 * expr.c (expand_expr_real_2): When handling conversions involving
9405 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9406 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9407 as a poly_uint64 too.
9409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9410 Alan Hayward <alan.hayward@arm.com>
9411 David Sherwood <david.sherwood@arm.com>
9413 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9416 Alan Hayward <alan.hayward@arm.com>
9417 David Sherwood <david.sherwood@arm.com>
9419 * combine.c (can_change_dest_mode): Handle polynomial
9420 REGMODE_NATURAL_SIZE.
9421 * expmed.c (store_bit_field_1): Likewise.
9422 * expr.c (store_constructor): Likewise.
9423 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9424 and polynomial REGMODE_NATURAL_SIZE.
9425 (gen_lowpart_common): Likewise.
9426 * reginfo.c (record_subregs_of_mode): Likewise.
9427 * rtlanal.c (read_modify_subreg_p): Likewise.
9429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9430 Alan Hayward <alan.hayward@arm.com>
9431 David Sherwood <david.sherwood@arm.com>
9433 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9434 numbers of elements.
9436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9437 Alan Hayward <alan.hayward@arm.com>
9438 David Sherwood <david.sherwood@arm.com>
9440 * match.pd: Cope with polynomial numbers of vector elements.
9442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9443 Alan Hayward <alan.hayward@arm.com>
9444 David Sherwood <david.sherwood@arm.com>
9446 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9447 in a POINTER_PLUS_EXPR.
9449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9450 Alan Hayward <alan.hayward@arm.com>
9451 David Sherwood <david.sherwood@arm.com>
9453 * omp-simd-clone.c (simd_clone_subparts): New function.
9454 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9455 (ipa_simd_modify_function_body): Likewise.
9457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9458 Alan Hayward <alan.hayward@arm.com>
9459 David Sherwood <david.sherwood@arm.com>
9461 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9462 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9463 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9464 (expand_vector_condition, vector_element): Likewise.
9465 (subparts_gt): New function.
9466 (get_compute_type): Use subparts_gt.
9467 (count_type_subparts): Delete.
9468 (expand_vector_operations_1): Use subparts_gt instead of
9469 count_type_subparts.
9471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9472 Alan Hayward <alan.hayward@arm.com>
9473 David Sherwood <david.sherwood@arm.com>
9475 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9476 (vect_compile_time_alias): ...this new function. Do the calculation
9477 on poly_ints rather than trees.
9478 (vect_prune_runtime_alias_test_list): Update call accordingly.
9480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9481 Alan Hayward <alan.hayward@arm.com>
9482 David Sherwood <david.sherwood@arm.com>
9484 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9486 (vect_schedule_slp_instance): Likewise.
9488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9489 Alan Hayward <alan.hayward@arm.com>
9490 David Sherwood <david.sherwood@arm.com>
9492 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9493 constant and extern definitions for variable-length vectors.
9494 (vect_get_constant_vectors): Note that the number of units
9495 is known to be constant.
9497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9498 Alan Hayward <alan.hayward@arm.com>
9499 David Sherwood <david.sherwood@arm.com>
9501 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9502 of units as polynomial. Choose between WIDE and NARROW based
9505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9506 Alan Hayward <alan.hayward@arm.com>
9507 David Sherwood <david.sherwood@arm.com>
9509 * tree-vect-stmts.c (simd_clone_subparts): New function.
9510 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9513 Alan Hayward <alan.hayward@arm.com>
9514 David Sherwood <david.sherwood@arm.com>
9516 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9517 vectors as polynomial. Use build_index_vector for
9520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9521 Alan Hayward <alan.hayward@arm.com>
9522 David Sherwood <david.sherwood@arm.com>
9524 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9525 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9526 for variable-length vectors.
9527 (vectorizable_mask_load_store): Treat the number of units as
9528 polynomial, asserting that it is constant if the condition has
9529 already been enforced.
9530 (vectorizable_store, vectorizable_load): Likewise.
9532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9533 Alan Hayward <alan.hayward@arm.com>
9534 David Sherwood <david.sherwood@arm.com>
9536 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9537 of units as polynomial. Punt if we can't tell at compile time
9538 which vector contains the final result.
9540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9541 Alan Hayward <alan.hayward@arm.com>
9542 David Sherwood <david.sherwood@arm.com>
9544 * tree-vect-loop.c (vectorizable_induction): Treat the number
9545 of units as polynomial. Punt on SLP inductions. Use an integer
9546 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9547 cast of such a series for variable-length floating-point
9550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9551 Alan Hayward <alan.hayward@arm.com>
9552 David Sherwood <david.sherwood@arm.com>
9554 * tree.h (build_index_vector): Declare.
9555 * tree.c (build_index_vector): New function.
9556 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9557 of units as polynomial, forcibly converting it to a constant if
9558 vectorizable_reduction has already enforced the condition.
9559 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9560 to create a {1,2,3,...} vector.
9561 (vectorizable_reduction): Treat the number of units as polynomial.
9562 Choose vectype_in based on the largest scalar element size rather
9563 than the smallest number of units. Enforce the restrictions
9566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9567 Alan Hayward <alan.hayward@arm.com>
9568 David Sherwood <david.sherwood@arm.com>
9570 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9571 number of units as polynomial.
9573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9574 Alan Hayward <alan.hayward@arm.com>
9575 David Sherwood <david.sherwood@arm.com>
9577 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9578 * target.def (autovectorize_vector_sizes): Return the vector sizes
9579 by pointer, using vector_sizes rather than a bitmask.
9580 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9581 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9582 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9584 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9585 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9586 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9587 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9588 * omp-general.c (omp_max_vf): Likewise.
9589 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9590 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9591 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9592 * tree-vect-slp.c (vect_slp_bb): Likewise.
9593 * doc/tm.texi: Regenerate.
9594 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9596 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9597 the vector size as a poly_uint64 rather than an unsigned int.
9598 (current_vector_size): Change from an unsigned int to a poly_uint64.
9599 (get_vectype_for_scalar_type): Update accordingly.
9600 * tree.h (build_truth_vector_type): Take the size and number of
9601 units as a poly_uint64 rather than an unsigned int.
9602 (build_vector_type): Add a temporary overload that takes
9603 the number of units as a poly_uint64 rather than an unsigned int.
9604 * tree.c (make_vector_type): Likewise.
9605 (build_truth_vector_type): Take the number of units as a poly_uint64
9606 rather than an unsigned int.
9608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9609 Alan Hayward <alan.hayward@arm.com>
9610 David Sherwood <david.sherwood@arm.com>
9612 * target.def (get_mask_mode): Take the number of units and length
9613 as poly_uint64s rather than unsigned ints.
9614 * targhooks.h (default_get_mask_mode): Update accordingly.
9615 * targhooks.c (default_get_mask_mode): Likewise.
9616 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9617 * doc/tm.texi: Regenerate.
9619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9620 Alan Hayward <alan.hayward@arm.com>
9621 David Sherwood <david.sherwood@arm.com>
9623 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9624 * omp-general.c (omp_max_vf): Likewise.
9625 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9626 (expand_omp_simd): Handle polynomial safelen.
9627 * omp-low.c (omplow_simd_context): Add a default constructor.
9628 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9629 (lower_rec_simd_input_clauses): Update accordingly.
9630 (lower_rec_input_clauses): Likewise.
9632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9633 Alan Hayward <alan.hayward@arm.com>
9634 David Sherwood <david.sherwood@arm.com>
9636 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9637 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9638 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9639 (vect_analyze_slp_cost): Likewise.
9640 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9641 (vect_model_load_cost): Likewise.
9643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9644 Alan Hayward <alan.hayward@arm.com>
9645 David Sherwood <david.sherwood@arm.com>
9647 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9648 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9649 from an unsigned int * to a poly_uint64_pod *.
9650 (calculate_unrolling_factor): New function.
9651 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9654 Alan Hayward <alan.hayward@arm.com>
9655 David Sherwood <david.sherwood@arm.com>
9657 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9658 from an unsigned int to a poly_uint64.
9659 (_loop_vec_info::slp_unrolling_factor): Likewise.
9660 (_loop_vec_info::vectorization_factor): Change from an int
9662 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9663 (vect_get_num_vectors): New function.
9664 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9665 (vect_get_num_copies): Use vect_get_num_vectors.
9666 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9667 to an unsigned int *.
9668 (vect_analyze_data_refs): Change min_vf from an int * to a
9670 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9671 than an unsigned HOST_WIDE_INT.
9672 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9673 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9674 to an unsigned int *.
9675 (vect_analyze_data_ref_dependences): Likewise.
9676 (vect_compute_data_ref_alignment): Handle polynomial vf.
9677 (vect_enhance_data_refs_alignment): Likewise.
9678 (vect_prune_runtime_alias_test_list): Likewise.
9679 (vect_shift_permute_load_chain): Likewise.
9680 (vect_supportable_dr_alignment): Likewise.
9681 (dependence_distance_ge_vf): Take the vectorization factor as a
9682 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9683 (vect_analyze_data_refs): Change min_vf from an int * to a
9685 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9686 vfm1 as a poly_uint64 rather than an int. Make the same change
9687 for the returned bound_scalar.
9688 (vect_gen_vector_loop_niters): Handle polynomial vf.
9689 (vect_do_peeling): Likewise. Update call to
9690 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9691 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9693 * tree-vect-loop.c (vect_determine_vectorization_factor)
9694 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9695 (vect_get_known_peeling_cost): Likewise.
9696 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9697 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9698 (vect_transform_loop): Likewise. Use the lowest possible VF when
9699 updating the upper bounds of the loop.
9700 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9702 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9703 polynomial unroll factors.
9704 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9705 (vect_make_slp_decision): Likewise.
9706 (vect_supported_load_permutation_p): Likewise, and polynomial
9708 (vect_analyze_slp_cost): Handle polynomial vf.
9709 (vect_slp_analyze_node_operations): Likewise.
9710 (vect_slp_analyze_bb_1): Likewise.
9711 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9712 than an unsigned HOST_WIDE_INT.
9713 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9714 (vectorizable_load): Handle polynomial vf.
9715 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9717 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9720 Alan Hayward <alan.hayward@arm.com>
9721 David Sherwood <david.sherwood@arm.com>
9723 * match.pd: Handle bit operations involving three constants
9724 and try to fold one pair.
9726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9728 * tree-vect-loop-manip.c: Include gimple-fold.h.
9729 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9730 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9731 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9732 Add a path that uses a step of VF instead of 1, but disable it
9734 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9735 and niters_no_overflow parameters. Update calls to
9736 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9737 Create a new SSA name if the latter choses to use a ste other
9738 than zero, and return it via niters_vector_mult_vf_var.
9739 * tree-vect-loop.c (vect_transform_loop): Update calls to
9740 vect_do_peeling, vect_gen_vector_loop_niters and
9741 slpeel_make_loop_iterate_ntimes.
9742 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9743 (vect_gen_vector_loop_niters): Update declarations after above changes.
9745 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9747 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9748 128-bit round to integer instructions.
9749 (ceil<mode>2): Likewise.
9750 (btrunc<mode>2): Likewise.
9751 (round<mode>2): Likewise.
9753 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9755 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9756 unaligned VSX load/store on P8/P9.
9757 (expand_block_clear): Allow the use of unaligned VSX
9758 load/store on P8/P9.
9760 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9762 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9764 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9765 swap associated with both a load and a store.
9767 2018-01-02 Andrew Waterman <andrew@sifive.com>
9769 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9770 * config/riscv/riscv.md (clear_cache): Use it.
9772 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9774 * web.c: Remove out-of-date comment.
9776 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9778 * expr.c (fixup_args_size_notes): Check that any existing
9779 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9780 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9781 (emit_single_push_insn): ...here.
9783 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9785 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9786 (const_vector_encoded_nelts): New function.
9787 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9788 (const_vector_int_elt, const_vector_elt): Declare.
9789 * emit-rtl.c (const_vector_int_elt_1): New function.
9790 (const_vector_elt): Likewise.
9791 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9792 of CONST_VECTOR_ELT.
9794 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9796 * expr.c: Include rtx-vector-builder.h.
9797 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9798 directly on the tree encoding.
9799 (const_vector_from_tree): Likewise.
9800 * optabs.c: Include rtx-vector-builder.h.
9801 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9802 sequence of "u" values.
9803 * vec-perm-indices.c: Include rtx-vector-builder.h.
9804 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9805 directly on the vec_perm_indices encoding.
9807 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9809 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9810 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9811 * rtx-vector-builder.h: New file.
9812 * rtx-vector-builder.c: Likewise.
9813 * rtl.h (rtx_def::u2): Add a const_vector field.
9814 (CONST_VECTOR_NPATTERNS): New macro.
9815 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9816 (CONST_VECTOR_DUPLICATE_P): Likewise.
9817 (CONST_VECTOR_STEPPED_P): Likewise.
9818 (CONST_VECTOR_ENCODED_ELT): Likewise.
9819 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9820 (unwrap_const_vec_duplicate): Likewise.
9821 (const_vec_series_p): Check for a non-duplicated vector encoding.
9822 Say that the function only returns true for integer vectors.
9823 * emit-rtl.c: Include rtx-vector-builder.h.
9824 (gen_const_vec_duplicate_1): Delete.
9825 (gen_const_vector): Call gen_const_vec_duplicate instead of
9826 gen_const_vec_duplicate_1.
9827 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9828 (gen_const_vec_duplicate): Use rtx_vector_builder.
9829 (gen_const_vec_series): Likewise.
9830 (gen_rtx_CONST_VECTOR): Likewise.
9831 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9832 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9833 Build a new vector rather than modifying a CONST_VECTOR in-place.
9834 (handle_special_swappables): Update call accordingly.
9835 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9836 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9837 Build a new vector rather than modifying a CONST_VECTOR in-place.
9838 (handle_special_swappables): Update call accordingly.
9840 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9842 * simplify-rtx.c (simplify_const_binary_operation): Use
9843 CONST_VECTOR_ELT instead of XVECEXP.
9845 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9847 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9848 the selector elements to be different from the data elements
9849 if the selector is a VECTOR_CST.
9850 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9851 ssizetype for the selector.
9853 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9855 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9856 before testing each element individually.
9857 * tree-vect-generic.c (lower_vec_perm): Likewise.
9859 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9861 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9862 * selftest-run-tests.c (selftest::run_tests): Call it.
9863 * vector-builder.h (vector_builder::operator ==): New function.
9864 (vector_builder::operator !=): Likewise.
9865 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9866 (vec_perm_indices::all_from_input_p): New function.
9867 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9868 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9869 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9870 instead of reading the VECTOR_CST directly. Detect whether both
9871 vector inputs are the same before constructing the vec_perm_indices,
9872 and update the number of inputs argument accordingly. Use the
9873 utility functions added above. Only construct sel2 if we need to.
9875 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9877 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9878 the broadcast of the low byte.
9879 (expand_mult_highpart): Use an explicit encoding for the permutes.
9880 * optabs-query.c (can_mult_highpart_p): Likewise.
9881 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9882 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9883 (vectorizable_bswap): Likewise.
9884 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9885 explicit encoding for the power-of-2 permutes.
9886 (vect_permute_store_chain): Likewise.
9887 (vect_grouped_load_supported): Likewise.
9888 (vect_permute_load_chain): Likewise.
9890 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9892 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9893 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9894 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9895 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9896 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9897 (vect_gen_perm_mask_any): Likewise.
9899 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9901 * int-vector-builder.h: New file.
9902 * vec-perm-indices.h: Include int-vector-builder.h.
9903 (vec_perm_indices): Redefine as an int_vector_builder.
9904 (auto_vec_perm_indices): Delete.
9905 (vec_perm_builder): Redefine as a stand-alone class.
9906 (vec_perm_indices::vec_perm_indices): New function.
9907 (vec_perm_indices::clamp): Likewise.
9908 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9909 (vec_perm_indices::new_vector): New function.
9910 (vec_perm_indices::new_expanded_vector): Update for new
9911 vec_perm_indices class.
9912 (vec_perm_indices::rotate_inputs): New function.
9913 (vec_perm_indices::all_in_range_p): Operate directly on the
9914 encoded form, without computing elided elements.
9915 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9916 encoding. Update for new vec_perm_indices class.
9917 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
9918 the given vec_perm_builder.
9919 (expand_vec_perm_var): Update vec_perm_builder constructor.
9920 (expand_mult_highpart): Use vec_perm_builder instead of
9921 auto_vec_perm_indices.
9922 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
9923 vec_perm_indices instead of auto_vec_perm_indices. Use a single
9924 or double series encoding as appropriate.
9925 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
9926 vec_perm_indices instead of auto_vec_perm_indices.
9927 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9928 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9929 (vect_permute_store_chain): Likewise.
9930 (vect_grouped_load_supported): Likewise.
9931 (vect_permute_load_chain): Likewise.
9932 (vect_shift_permute_load_chain): Likewise.
9933 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
9934 (vect_transform_slp_perm_load): Likewise.
9935 (vect_schedule_slp_instance): Likewise.
9936 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9937 (vectorizable_mask_load_store): Likewise.
9938 (vectorizable_bswap): Likewise.
9939 (vectorizable_store): Likewise.
9940 (vectorizable_load): Likewise.
9941 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
9942 vec_perm_indices instead of auto_vec_perm_indices. Use
9943 tree_to_vec_perm_builder to read the vector from a tree.
9944 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
9945 vec_perm_builder instead of a vec_perm_indices.
9946 (have_whole_vector_shift): Use vec_perm_builder and
9947 vec_perm_indices instead of auto_vec_perm_indices. Leave the
9948 truncation to calc_vec_perm_mask_for_shift.
9949 (vect_create_epilog_for_reduction): Likewise.
9950 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
9951 from auto_vec_perm_indices to vec_perm_indices.
9952 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9953 instead of changing individual elements.
9954 (aarch64_vectorize_vec_perm_const): Use new_vector to install
9955 the vector in d.perm.
9956 * config/arm/arm.c (expand_vec_perm_d::perm): Change
9957 from auto_vec_perm_indices to vec_perm_indices.
9958 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
9959 instead of changing individual elements.
9960 (arm_vectorize_vec_perm_const): Use new_vector to install
9961 the vector in d.perm.
9962 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
9963 Update vec_perm_builder constructor.
9964 (rs6000_expand_interleave): Likewise.
9965 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
9966 (rs6000_expand_interleave): Likewise.
9968 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9970 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
9971 to qimode could truncate the indices.
9972 * optabs.c (expand_vec_perm_var): Likewise.
9974 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9976 * Makefile.in (OBJS): Add vec-perm-indices.o.
9977 * vec-perm-indices.h: New file.
9978 * vec-perm-indices.c: Likewise.
9979 * target.h (vec_perm_indices): Replace with a forward class
9981 (auto_vec_perm_indices): Move to vec-perm-indices.h.
9982 * optabs.h: Include vec-perm-indices.h.
9983 (expand_vec_perm): Delete.
9984 (selector_fits_mode_p, expand_vec_perm_var): Declare.
9985 (expand_vec_perm_const): Declare.
9986 * target.def (vec_perm_const_ok): Replace with...
9987 (vec_perm_const): ...this new hook.
9988 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
9989 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
9990 * doc/tm.texi: Regenerate.
9991 * optabs.def (vec_perm_const): Delete.
9992 * doc/md.texi (vec_perm_const): Likewise.
9993 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
9994 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
9995 expand_vec_perm for constant permutation vectors. Assert that
9996 the mode of variable permutation vectors is the integer equivalent
9997 of the mode that is being permuted.
9998 * optabs-query.h (selector_fits_mode_p): Declare.
9999 * optabs-query.c: Include vec-perm-indices.h.
10000 (selector_fits_mode_p): New function.
10001 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10002 is defined, instead of checking whether the vec_perm_const_optab
10003 exists. Use targetm.vectorize.vec_perm_const instead of
10004 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10005 fit in the vector mode before using a variable permute.
10006 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10007 vec_perm_indices instead of an rtx.
10008 (expand_vec_perm): Replace with...
10009 (expand_vec_perm_const): ...this new function. Take the selector
10010 as a vec_perm_indices rather than an rtx. Also take the mode of
10011 the selector. Update call to shift_amt_for_vec_perm_mask.
10012 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10013 Use vec_perm_indices::new_expanded_vector to expand the original
10014 selector into bytes. Check whether the indices fit in the vector
10015 mode before using a variable permute.
10016 (expand_vec_perm_var): Make global.
10017 (expand_mult_highpart): Use expand_vec_perm_const.
10018 * fold-const.c: Includes vec-perm-indices.h.
10019 * tree-ssa-forwprop.c: Likewise.
10020 * tree-vect-data-refs.c: Likewise.
10021 * tree-vect-generic.c: Likewise.
10022 * tree-vect-loop.c: Likewise.
10023 * tree-vect-slp.c: Likewise.
10024 * tree-vect-stmts.c: Likewise.
10025 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10027 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10028 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10029 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10030 (aarch64_vectorize_vec_perm_const): ...this new function.
10031 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10032 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10033 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10034 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10035 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10036 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10037 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10039 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10040 check for NEON modes.
10041 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10042 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10043 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10044 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10046 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10047 the old VEC_PERM_CONST conditions.
10048 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10049 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10050 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10051 (ia64_vectorize_vec_perm_const_ok): Merge into...
10052 (ia64_vectorize_vec_perm_const): ...this new function.
10053 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10054 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10055 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10056 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10057 * config/mips/mips.c (mips_expand_vec_perm_const)
10058 (mips_vectorize_vec_perm_const_ok): Merge into...
10059 (mips_vectorize_vec_perm_const): ...this new function.
10060 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10061 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10062 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10063 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10064 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10065 (rs6000_expand_vec_perm_const): Delete.
10066 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10068 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10069 (altivec_expand_vec_perm_const_le): Take each operand individually.
10070 Operate on constant selectors rather than rtxes.
10071 (altivec_expand_vec_perm_const): Likewise. Update call to
10072 altivec_expand_vec_perm_const_le.
10073 (rs6000_expand_vec_perm_const): Delete.
10074 (rs6000_vectorize_vec_perm_const_ok): Delete.
10075 (rs6000_vectorize_vec_perm_const): New function.
10076 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10077 an element count and rtx array.
10078 (rs6000_expand_extract_even): Update call accordingly.
10079 (rs6000_expand_interleave): Likewise.
10080 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10081 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10082 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10083 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10084 (rs6000_expand_vec_perm_const): Delete.
10085 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10086 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10087 (altivec_expand_vec_perm_const_le): Take each operand individually.
10088 Operate on constant selectors rather than rtxes.
10089 (altivec_expand_vec_perm_const): Likewise. Update call to
10090 altivec_expand_vec_perm_const_le.
10091 (rs6000_expand_vec_perm_const): Delete.
10092 (rs6000_vectorize_vec_perm_const_ok): Delete.
10093 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10094 reference to the SPE evmerge intructions.
10095 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10096 an element count and rtx array.
10097 (rs6000_expand_extract_even): Update call accordingly.
10098 (rs6000_expand_interleave): Likewise.
10099 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10100 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10102 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10104 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10106 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10107 vector mode and that that mode matches the mode of the data
10109 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10110 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10111 directly using expand_vec_perm_1 when forcing selectors into
10113 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10115 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10117 * optabs-query.h (can_vec_perm_p): Delete.
10118 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10119 * optabs-query.c (can_vec_perm_p): Split into...
10120 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10121 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10122 particular selector is valid.
10123 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10124 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10125 (vect_grouped_load_supported): Likewise.
10126 (vect_shift_permute_load_chain): Likewise.
10127 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10128 (vect_transform_slp_perm_load): Likewise.
10129 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10130 (vectorizable_bswap): Likewise.
10131 (vect_gen_perm_mask_checked): Likewise.
10132 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10133 implementations of variable permutation vectors into account
10134 when deciding which selector to use.
10135 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10136 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10137 with a false third argument.
10138 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10139 to test whether the constant selector is valid and can_vec_perm_var_p
10140 to test whether a variable selector is valid.
10142 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10144 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10145 * optabs-query.c (can_vec_perm_p): Likewise.
10146 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10147 instead of vec_perm_indices.
10148 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10149 (vect_gen_perm_mask_checked): Likewise,
10150 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10151 (vect_gen_perm_mask_checked): Likewise,
10153 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10155 * optabs-query.h (qimode_for_vec_perm): Declare.
10156 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10157 (qimode_for_vec_perm): ...this new function.
10158 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10160 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10162 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10163 does not have a conditional at the top.
10165 2018-01-02 Richard Biener <rguenther@suse.de>
10167 * ipa-inline.c (big_speedup_p): Fix expression.
10169 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10172 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10175 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10179 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10180 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10181 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10182 cond_taken_branch_cost 3->4.
10184 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10186 PR tree-optimization/83581
10187 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10188 TODO_cleanup_cfg if any changes have been made.
10190 PR middle-end/83608
10191 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10192 convert_modes if target mode has the right side, but different mode
10195 PR middle-end/83609
10196 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10197 last argument when extracting from CONCAT. If either from_real or
10198 from_imag is NULL, use expansion through memory. If result is not
10199 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10200 the parts directly to inner mode, if even that fails, use expansion
10203 PR middle-end/83623
10204 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10205 check for bswap in mode rather than HImode and use that in expand_unop
10208 Copyright (C) 2018 Free Software Foundation, Inc.
10210 Copying and distribution of this file, with or without modification,
10211 are permitted in any medium without royalty provided the copyright
10212 notice and this notice are preserved.