1 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3 * config/nds32/nds32.c (nds32_init_machine_status,
4 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5 strict_aligned_p field.
6 (nds32_expand_to_rtl_hook): New function.
7 (TARGET_EXPAND_TO_RTL_HOOK): Define.
8 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
10 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11 Chung-Ju Wu <jasonwucj@gmail.com>
13 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
14 * config/nds32/nds32-n7.md: New file.
15 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
16 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
18 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
19 * config/nds32/nds32.md (pipeline_model): Add n7.
20 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
21 * config/nds32/pipelines.md: Include n7 settings.
23 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
24 Chung-Ju Wu <jasonwucj@gmail.com>
26 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
27 * config/nds32/nds32-e8.md: New file.
28 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
29 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
31 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
32 * config/nds32/nds32.md (pipeline_model): Add e8.
33 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
34 * config/nds32/pipelines.md: Include e8 settings.
36 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
37 Chung-Ju Wu <jasonwucj@gmail.com>
39 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
40 * config/nds32/nds32-n8.md: New file.
41 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
42 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
44 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
45 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
46 * config/nds32/nds32.md (pipeline_model): Add n8.
47 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
48 * config/nds32/pipelines.md: Include n8 settings.
50 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
51 Chung-Ju Wu <jasonwucj@gmail.com>
53 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
54 * config/nds32/nds32-n9-2r1w.md: New file.
55 * config/nds32/nds32-n9-3r2w.md: New file.
56 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
57 nds32_register_ports): New or modify for cpu n9.
58 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
60 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
61 * config/nds32/nds32-utils.c: New file.
62 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
63 TARGET_MUL_SLOW): Define.
64 * config/nds32/nds32.md (pipeline_model): New attribute.
65 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
66 New options that support cpu n9.
67 * config/nds32/pipelines.md: Include n9 settings.
68 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
70 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
72 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
73 information if necessary.
74 (output_cond_branch_compare_zero): Likewise.
75 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
76 (nds32_target_alignment): Refine for alignment.
77 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
78 (FUNCTION_BOUNDARY): Modify.
79 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
81 * config/nds32/nds32.opt (malways-align, malign-functions): New.
83 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
85 * config/nds32/constants.md (unspec_volatile_element): Add values for
86 TLB operation and data prefetch.
87 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
88 functions for TLB operation and data prefetch.
89 * config/nds32/nds32-intrinsic.md: Likewise.
90 * config/nds32/nds32_intrinsic.h: Likewise.
91 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
92 (nds32_print_operand): Likewise.
93 * config/nds32/nds32.h (nds32_builtins): Likewise.
95 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
96 Andrew Pinski <pinsika@gcc.gnu.org>
99 * match.pd: Use constant_boolean_node of correct type instead of
100 boolean_true_node or boolean_false_node for simplifying
101 pointer comparisons to zero.
103 2018-04-07 Jakub Jelinek <jakub@redhat.com>
105 PR tree-optimization/80021
106 * tree.c (verify_type_variant): Make error call in verify_variant_match
107 translatable and remove final full stop.
109 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
111 * config/nds32/constants.md (unspec_volatile_element): Add
112 UNSPEC_VOLATILE_EH_RETURN.
113 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
114 nds32_output_stack_pop): Support dwarf exception handling process.
115 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
116 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
117 exception handling process.
118 (nds32_compute_stack_frame): Likewise.
119 (nds32_return_addr_rtx): Likewise.
120 (nds32_initial_elimination_offset): Likewise.
121 (nds32_expand_prologue): Likewise.
122 (nds32_expand_epilogue): Likewise.
123 (nds32_dynamic_chain_address): New function.
124 * config/nds32/nds32.h (machine_function): Add fields for dwarf
126 (DYNAMIC_CHAIN_ADDRESS): Define.
127 (EH_RETURN_DATA_REGNO): Define.
128 (EH_RETURN_STACKADJ_RTX): Define.
129 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
130 patterns for dwarf exception handling.
132 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
134 * config/nds32/nds32.h: Clean up obsolete macros.
136 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
138 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
139 Add enum values for particular instructions.
140 * config/nds32/nds32-intrinsic.c: Implementation of expanding
141 particular intrinsic functions.
142 * config/nds32/nds32-intrinsic.md: Likewise.
143 * config/nds32/nds32_intrinsic.h: Likewise.
144 * config/nds32/nds32.h (nds32_builtins): Likewise.
145 * config/nds32/nds32.md (type): Add pbsad and pbsada.
146 (btst, ave): New patterns for particular instructions.
148 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
150 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
151 Add enum values for atomic load/store and memory sync.
152 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
154 * config/nds32/nds32-intrinsic.md: Likewise.
155 * config/nds32/nds32_intrinsic.h: Likewise.
156 * config/nds32/nds32.h (nds32_builtins): Likewise.
158 2018-04-07 Jakub Jelinek <jakub@redhat.com>
160 PR tree-optimization/85257
161 * fold-const.c (native_encode_vector): If not all elts could fit
162 and off is -1, return 0 rather than offset.
163 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
164 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
165 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
166 adjust buffer in native_interpret_expr call.
168 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
170 * config/nds32/constants.md (unspec_volatile_element): Add cache
172 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
173 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
174 * config/nds32/nds32.c (nds32_cctl_names): New.
175 (nds32_print_operand): Handle cache control register names.
176 * config/nds32/nds32.h (nds32_builtins): New enum values.
177 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
179 * config/nds32/nds32.md (type): Add mmu.
180 * config/nds32/pipelines.md (simple_insn): Add mmu.
182 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
184 * config/nds32/nds32.md (type): Remove call.
185 * config/nds32/pipelines.md (simple_insn): Likewise.
187 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
189 * config/nds32/constants.md (unspec_volatile_element): Add
190 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
191 UNSPEC_VOLATILE_FMFCFG.
192 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
193 description for fmfcfg and fmfcsr.
194 (bdesc_1arg): Add fmtcsr.
195 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
196 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
197 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
198 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
199 unspec_fmfcfg): New patterns.
200 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
201 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
202 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
203 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
204 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
205 __nds32__fmfcfg): Define.
207 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
209 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
210 intrinsic register names.
211 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
212 intrinsic register enum values and macros.
214 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
216 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
217 for load/store addressing form.
218 (nds32_print_operand_address): Likewise.
220 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
223 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
224 based on LABEL_REF. Remove useless assertion.
225 (pic_address_needs_scratch): Fix formatting.
226 (sparc_legitimize_pic_address): Minor tweaks.
227 (sparc_delegitimize_address): Adjust assertion accordingly.
228 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
229 into symbolic_operand.
230 (movsi_high_pic_label_ref): Likewise.
231 (movsi_lo_sum_pic_label_ref): Likewise.
232 (movdi_pic_label_ref): Likewise.
233 (movdi_high_pic_label_ref): Likewise.
234 (movdi_lo_sum_pic_label_ref): Likewise.
236 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
238 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
239 custom LIB_SPEC setup.
241 2018-04-06 Ruslan Bukin <br@bsdpad.com>
242 Kito Cheng <kito.cheng@gmail.com>
244 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
245 * config/riscv/freebsd.h: New.
247 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
249 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
250 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
253 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
254 Kito Cheng <kito.cheng@gmail.com>
256 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
257 nds32_output_call, nds32_symbol_binds_local_p): New functions.
258 * config/nds32/nds32-protos.h (nds32_output_call,
259 nds32_output_return): Declare.
260 * config/nds32/nds32.md: Refine all the call and return patterns.
262 2018-04-06 Jakub Jelinek <jakub@redhat.com>
265 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
266 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
268 PR rtl-optimization/84872
269 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
270 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
273 2018-04-06 Tamar Christina <tamar.christina@arm.com>
275 * expr.c (copy_blkmode_to_reg): Revert 254862.
276 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
278 2018-04-06 Richard Biener <rguenther@suse.de>
281 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
282 after seeing a component reference with an adjacent field. Treat
283 refs to arrays at struct end of external decls similar to
284 refs to unconstrained commons.
286 2018-04-06 Jakub Jelinek <jakub@redhat.com>
289 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
290 look through SAVE_EXPRs with non-side-effects argument. Adjust
292 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
295 2018-04-06 Richard Biener <rguenther@suse.de>
298 * alias.c (find_base_term): New wrapper around find_base_term
299 unwinding CSELIB_VAL_PTR changes.
300 (find_base_term): Do not restore CSELIB_VAL_PTR during the
303 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
305 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
307 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
308 constant definitions.
309 ("nop"): lr 0,0 -> nopr r0
310 ("nop_lr0", "nop_lr1"): New insn definitions.
312 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
314 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
315 NDS32_V3PUSH_AVAILABLE_P macro.
317 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
318 Chung-Ju Wu <jasonwucj@gmail.com>
320 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
321 (nds32*-*-*): Add float and fpu_config into supported_defaults.
322 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
323 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
324 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
325 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
326 * config/nds32/constraints.md: New constraints and checking for hard
328 * config/nds32/iterators.md: New mode iterator and attribute for hard
330 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
332 * config/nds32/nds32-fpu.md: New file.
333 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
334 deal with hard float code generation.
335 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
337 (abi_type, float_reg_number): New enum type.
338 * config/nds32/nds32-predicates.c: New predicates for hard float.
339 * config/nds32/nds32-protos.h: Declare functions for hard float.
340 * config/nds32/nds32.c: Implementation for hard float configuration.
341 * config/nds32/nds32.h: Definitions for hard float configuration.
342 * config/nds32/nds32.md: Include hard float machine description and
343 modify patterns for hard float configuration.
344 * config/nds32/nds32.opt: New options for hard float configuration.
345 * config/nds32/predicates.md: New predicates for hard float
348 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
350 * common/config/nds32/nds32-common.c
351 (nds32_option_optimization_table): Enable -mreleax-hint by default.
353 2018-04-05 Jakub Jelinek <jakub@redhat.com>
356 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
357 CONSTRUCTOR_ELT (ctor, ...)->value.
359 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
362 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
364 2018-04-05 Tom de Vries <tom@codesourcery.com>
367 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
370 2018-04-05 Shiva Chen <shiva0217@gmail.com>
371 Kito Cheng <kito.cheng@gmail.com>
373 * config/nds32/constraints.md (U33): Fine-tune checking condition.
374 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
375 * config/nds32/nds32.h (nds32_16bit_address_type): Add
376 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
378 2018-04-05 Shiva Chen <shiva0217@gmail.com>
379 Kito Cheng <kito.cheng@gmail.com>
381 * config/nds32/constraints.md (Ufe): New memory constraint.
382 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
383 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
384 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
386 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
387 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
389 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
391 * config/nds32/nds32.md: Use optimize_size in the condition for
392 alu-shift instructions.
394 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
396 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
398 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
400 * config/nds32/nds32.md (negsi2): Refine pattern.
402 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
403 Chung-Ju Wu <jasonwucj@gmail.com>
405 * config/nds32/iterators.md (shift_rotate): New code iterator.
406 (shift): New code attribute.
407 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
408 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
409 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
410 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
412 (andsi3, *andsi3): Ditto.
413 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
414 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
415 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
416 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
417 nds32_ior_operand, nds32_xor_operand): New predicates.
419 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
421 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
422 (addsi3, subsi3): ... this.
424 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
426 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
428 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
430 * config/nds32/nds32.md: Adjust indention.
432 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
434 * config/nds32/nds32.md (feature): New attribute.
436 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
438 * config/nds32/nds32.md (subtype): New attribute.
440 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
443 * config/arm/arm-builtins.c (arm_expand_builtin): Change
444 expansion to perform a bitwise AND of the argument followed by a
445 boolean negation of the result.
447 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
449 PR rtl-optimization/84878
450 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
451 the basic block. Assert the use reference is not artificial and that
452 it has an associated insn.
454 2018-04-04 Michael Matz <matz@suse.de>
456 * builtins.c (compute_objsize): Pass correct operand
457 to array_at_struct_end_p.
459 2018-04-04 Richard Biener <rguenther@suse.de>
462 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
463 from contexts for DINFO_LEVEL_TERSE and below.
465 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
467 * config/nds32/nds32-doubleword.md (move_<mode>): Require
468 resiter_operand condition.
469 * config/nds32/nds32.md (*move<mode>): Ditto.
471 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
472 Monk Chiang <sh.chiang04@gmail.com>
474 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
476 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
478 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
480 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
481 Kito Cheng <kito.cheng@gmail.com>
483 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
484 nds32_cond_code_str, output_cond_branch,
485 output_cond_branch_compare_zero, nds32_expand_cbranch,
486 nds32_expand_cstore, nds32_expand_movcc,
487 nds32_output_cbranchsi4_equality_zero,
488 nds32_output_cbranchsi4_equality_reg,
489 nds32_output_cbranchsi4_equality_reg_or_const_int,
490 nds32_output_cbranchsi4_greater_less_zero: New functions.
491 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
492 nds32_expand_cstore, nds32_expand_movcc,
493 nds32_output_cbranchsi4_equality_zero,
494 nds32_output_cbranchsi4_equality_reg,
495 nds32_output_cbranchsi4_equality_reg_or_const_int,
496 nds32_output_cbranchsi4_greater_less_zero): Declare.
497 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
498 nds32_rimm11s_operand): New predicates.
499 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
500 * config/nds32/nds32.md: Rewrite all the branch and conditional move
503 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
505 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
506 * config/nds32/nds32.md: Ditto.
507 * config/nds32/pipelines.md: Ditto.
509 2018-04-04 Richard Biener <rguenther@suse.de>
511 PR tree-optimization/85168
512 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
513 propagating abnormals.
515 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
517 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
519 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
520 Kito Cheng <kito.cheng@gmail.com>
522 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
523 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
524 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
525 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
526 * config/nds32/nds32.md (sibcall_internal): New.
527 (sibcall_register): Remove.
528 (sibcall_immediate): Remove.
529 (sibcall_value_internal): New.
530 (sibcall_value_register): Remove.
531 (sibcall_value_immediate): Remove.
532 * config/nds32/predicates.md (nds32_general_register_operand): New.
533 (nds32_call_address_operand): New.
535 2018-04-03 Jakub Jelinek <jakub@redhat.com>
537 PR rtl-optimization/85167
538 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
539 bb_defs if *split_p, instead preinitialize it to NULL.
541 PR tree-optimization/85156
542 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
543 evaluating the argument multiple times.
545 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
547 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
549 (_mm_cvtpd_ps): Likewise.
550 (_mm_cvttpd_epi32): Likewise.
551 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
552 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
553 vector, pixel, and bool following altivec.h include.
555 2018-04-03 Martin Sebor <msebor@redhat.com>
557 * doc/extend.texi (Common Function Attributes): Clarify.
558 (const attribute): Likewise.
559 (pure attribute): Likewise.
561 2018-04-03 Jakub Jelinek <jakub@redhat.com>
564 * config/i386/i386.c (ix86_expand_vector_set): Use
565 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
567 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
569 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
570 instructions when changing rounding bits to preserve precision bits
571 in the x87 control word.
573 2018-04-03 Martin Liska <mliska@suse.cz>
575 PR tree-optimization/82491
576 * rtl.h (strip_offset_and_add): Replace += suboffset with
577 poly_uint64 () + suboffset.
579 2018-03-29 Martin Liska <mliska@suse.cz>
580 Martin Jambor <mjambor@suse.cz>
583 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
584 param_type is not an integral or pointer type.
586 2018-04-03 Richard Biener <rguenther@suse.de>
588 * sese.h (recompute_all_dominators): Remove.
590 2018-04-02 Martin Sebor <msebor@redhat.com>
592 * doc/invoke.texi (-Wrestrict): Fix typos.
594 2018-04-02 Jim Wilson <jimw@sifive.com>
596 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
597 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
598 (<optab>di3, <optab>si3_extend): Likewise.
599 (<optab>si3_mask, <optab>si3_mask_1): New.
600 (<optab>di3_mask, <optab>di3_mask_1): New.
601 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
602 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
603 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
605 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
607 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
610 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
612 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
613 (nds32_canonicalize_comparison): New function.
615 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
616 Kito Cheng <kito.cheng@gmail.com>
617 Kuan-Lin Chen <kuanlinchentw@gmail.com>
619 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
620 * config/nds32/constants.md (unspec_volatile_element): Add
621 UNSPEC_VOLATILE_RELAX_GROUP.
622 * config/nds32/nds32-relax-opt.c: New file.
623 * config/nds32/nds32-predicates.c
624 (nds32_symbol_load_store_p): New function.
625 * config/nds32/nds32-protos.h
626 (nds32_symbol_load_store_p): Declare function.
627 (make_pass_nds32_relax_opt): Declare new rtl pass function.
628 * config/nds32/nds32.c
629 (nds32_register_pass): New function to register pass.
630 (nds32_register_passes): New function to register passes.
631 * config/nds32/nds32.md (relax_group): New pattern.
632 * config/nds32/nds32.opt (mrelax-hint): New option.
633 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
635 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
637 * config/nds32/t-nds32: Modify files dependency.
639 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
641 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
642 (PROFILE_HOOK): Define its implementation.
644 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
646 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
647 type and 32-bit size.
649 2018-04-01 Jakub Jelinek <jakub@redhat.com>
652 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
653 (V_128_256): New mode iterator.
654 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
655 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
656 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
658 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
659 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
661 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
664 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
665 NaN inputs correctly.
667 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
670 * config/rs6000/vsx.md (??r): New mode attribute.
671 (*vsx_mov<mode>_64bit): Use it.
672 (*vsx_mov<mode>_32bit): Likewise.
674 2018-03-30 Martin Sebor <msebor@redhat.com>
676 PR tree-optimization/84818
677 * builtins.c (check_access): Use warning_n.
679 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
682 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
684 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
687 2018-03-30 Julia Koval <julia.koval@intel.com>
690 * x86-tune.def (movx, partial_reg_dependency): Enable for
693 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
696 * lra-constraints.c (process_alt_operands): Move setting
697 this_alternative_matches below.
699 2018-03-29 Martin Liska <mliska@suse.cz>
702 * doc/invoke.texi: Document how LTO works with debug info.
703 Describe auto-load support of binutils. Mention 'x86-64'
704 as valid option value of -march option.
706 2018-03-29 Jakub Jelinek <jakub@redhat.com>
708 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
711 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
712 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
713 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
716 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
719 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
720 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
721 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
722 for RS6000_BTM_POWERPC64.
723 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
724 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
725 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
730 2018-03-28 Carl Love <cel@us.ibm.com>
733 2017-09-27 Carl Love <cel@us.ibm.com>
735 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
736 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
737 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
740 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
742 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
743 instead of __vector bool.
744 (_mm_max_pu8): Likewise.
745 (_mm_min_pi16): Likewise.
747 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
750 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
754 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
756 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
757 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
758 (div_extend): Likewise.
759 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
761 (__builtin_divweuo): Likewise.
762 (__builtin_divdeo): Likewise.
763 (__builtin_divdeuo): Likewise.
765 2018-03-28 Jakub Jelinek <jakub@redhat.com>
768 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
769 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
771 PR tree-optimization/82004
772 * gimple-match-head.c (optimize_pow_to_exp): New function.
773 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
774 Don't fold to exp if optimize_pow_to_exp is false.
776 2018-03-28 Martin Liska <mliska@suse.cz>
779 * calls.c (initialize_argument_information): Fix trailing space.
780 * common.opt: Fix typo and provide better explanation for
781 -fsanitize-coverage option.
782 * config/i386/i386.opt: Fix typo.
784 2018-03-28 Jakub Jelinek <jakub@redhat.com>
785 Martin Liska <mliska@suse.cz>
788 * gimplify.c (asan_poison_variable): Don't do the check for
789 gimplify_omp_ctxp here.
790 (gimplify_decl_expr): Do it here.
791 (gimplify_target_expr): Likewise.
793 2018-03-28 Martin Liska <mliska@suse.cz>
796 * config/i386/i386.c (ix86_function_arg_advance): Do not call
797 chkp_type_bounds_count if MPX is not enabled.
799 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
801 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
803 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
806 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
807 function to create the function decl for complex long double
808 multiply and divide for -mabi=ieeelongdouble.
809 (init_float128_ieee): Call it.
811 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
814 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
815 -fcf-protection=branch -mibt.
816 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
818 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
821 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
823 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
826 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
827 extern array declarations.
829 2018-03-27 Richard Biener <rguenther@suse.de>
832 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
833 explicit single_use checks.
835 2018-03-27 Richard Biener <rguenther@suse.de>
837 PR tree-optimization/85082
838 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
841 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
843 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
844 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
845 Turn on fasynchronous-unwind-tables and funwind-tables.
847 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
850 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
851 (*bmi_blsr_<mode>_ccz): Ditto.
853 2018-03-26 Tom de Vries <tom@codesourcery.com>
855 PR tree-optimization/85063
856 * omp-general.c (offloading_function_p): New function. Factor out
858 * omp-offload.c (pass_omp_target_link::gate): ... here.
859 * omp-general.h (offloading_function_p): Declare.
860 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
861 with attribute omp declare target for offloading functions.
863 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
865 PR tree-optimization/84005
866 * tree-data-ref.h (get_base_for_alignment): Declare.
867 * tree-data-ref.c (get_base_for_alignment_1): New function.
868 (get_base_for_alignment): Likewise.
869 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
870 get_base_for_alignment to find a suitable base object, instead
871 of always using drb->base_address.
873 2018-03-23 Jakub Jelinek <jakub@redhat.com>
876 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
877 known size by default.
879 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
882 * lra-constraints.c (process_alt_operands): Don't match BLKmode
883 and non BLKmode operands.
885 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
888 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
891 2018-03-23 Richard Biener <rguenther@suse.de>
894 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
895 we are going to emit early debug for LTO.
897 2018-03-23 Jakub Jelinek <jakub@redhat.com>
900 * function.c (match_asm_constraints_1): Don't optimize if input
901 doesn't satisfy general_operand predicate for output's mode.
904 * alias.c (write_dependence_p): Don't require for x_canonicalized
905 non-VOIDmode if x has VOIDmode.
908 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
909 just don't try to optimize it rather than assert it never happens.
911 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
913 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
914 macro expansions for definition of ST_INTERNAL_<mode> and
915 LD_INTERNAL_<mode> builtins.
916 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
918 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
920 (altivec_expand_st_builtin): Likewise.
921 (altivec_expand_builtin): Remove calls to deleted functions.
922 (rs6000_address_for_altivec): Delete this function.
923 * config/rs6000/vector.md: Remove expands for
924 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
926 2018-03-22 Sudakshina Das <sudi.das@arm.com>
929 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
930 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
931 re-computing once computed.
932 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
933 (arm_init_machine_status): Initialize
934 machine->static_chain_stack_bytes.
936 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
939 * doc/extend.texi: Add four new prototypes for vec_ld.
940 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
941 definitions for more logical presentation.
942 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
943 entries for V1TI variants of __builtin_altivec_ld builtin.
944 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
945 handling of V1TI variant of LVX icode pattern.
946 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
947 (rs6000_gimple_fold_builtin): Likewise.
948 (altivec_init_builtins): Add code to define
949 __builtin_altivec_lvx_v1ti function.
951 2018-03-22 Jakub Jelinek <jakub@redhat.com>
954 * function.c (match_asm_constraints_1): Don't do the optimization
955 if input isn't a REG, SUBREG, MEM or constant.
957 2018-03-22 Tom de Vries <tom@codesourcery.com>
959 PR tree-optimization/84956
960 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
961 bb_has_abnormal_pred.
963 2018-03-22 Jakub Jelinek <jakub@redhat.com>
966 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
967 DECL_INITIAL (decl) to decl at the end.
968 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
971 2018-03-21 Joseph Myers <joseph@codesourcery.com>
973 * doc/extend.texi (__builtin_tgmath): Document when complex
974 integer types are treated as _Complex _Float64.
976 2018-03-21 Tom de Vries <tom@codesourcery.com>
978 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
980 2018-03-21 Jakub Jelinek <jakub@redhat.com>
982 PR tree-optimization/84960
983 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
984 if it is ENTRY block, move them into single succ of ENTRY in that case.
986 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
988 PR tree-optimization/84811
989 * poly-int.h (poly_span_traits): Remove the T3 parameter and
990 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
991 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
992 (known_subrange_p): Update accordingly. Cast each value involved
993 in the size comparison, rather than casting the result of the
996 2018-03-21 Jakub Jelinek <jakub@redhat.com>
998 PR tree-optimization/84982
999 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1000 by flipping the least significant bit rather than all bits from
1001 bitpos to bitpos + bitsize - 1.
1003 2018-03-21 Nathan Sidwell <nathan@acm.org>
1005 * doc/extend.texi (Deprecated Features): Remove mention of
1006 long-deleted deprecations.
1008 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1011 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1012 * configure: Regenerate.
1014 2018-03-21 Tom de Vries <tom@codesourcery.com>
1016 PR tree-optimization/83126
1017 * tree-parloops.c (num_phis): New function.
1018 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1020 2018-03-21 Nathan Sidwell <nathan@acm.org>
1022 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1023 mention anon-struct/union members and trailing attributes.
1025 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1027 PR tree-optimization/84969
1028 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1029 builtin memset partitions if they set different rhs values.
1031 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1033 PR rtl-optimization/84989
1034 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1035 VEC_DUPLICATE with scalar result mode.
1037 2018-03-21 Martin Liska <mliska@suse.cz>
1040 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1041 not intended return statement.
1043 2018-03-21 Martin Liska <mliska@suse.cz>
1046 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1047 (chkp_find_bound_slots_1): Limit number of iterations.
1049 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1052 * Minor grammar fixes for x86 options.
1054 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1057 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1058 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1060 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1063 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1064 (altivec_lvx_<mode>_1op): Likewise.
1065 (altivec_stvx_<mode>_2op): Likewise.
1066 (altivec_stvx_<mode>_1op): Likewise.
1067 (altivec_lvx_<VM2:mode>): New define_expand.
1068 (altivec_stvx_<VM2:mode>): Likewise.
1069 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1070 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1071 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1072 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1073 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1074 (rs6000_gen_lvx): Likewise.
1075 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1076 (altivec_expand_stv_builtin): Likewise.
1077 (altivec_expand_builtin): Likewise.
1078 * config/rs6000/vector.md: Likewise.
1080 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1083 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1086 2018-03-20 Richard Biener <rguenther@suse.de>
1089 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1090 sign-conversions as zero, fall back to standard scalar_stmt
1093 2018-03-20 Martin Liska <mliska@suse.cz>
1096 * predict.c (rebuild_frequencies): Handle case when we have
1097 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1099 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1102 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1103 flag_section_anchors.
1104 * varasm.c (use_blocks_for_decl_p): Remove hack for
1105 dw2_force_const_mem.
1108 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1110 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1111 be created, use lowpart_subreg of operands[0] rather than operands[0]
1113 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1114 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1115 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1116 and n constraint instead of aarch64_shift_imm_di and Usd.
1117 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1118 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1120 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1123 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1124 to favor GPR over NEON registers.
1125 (<shift>di3_neon): Likewise.
1127 2018-03-20 Tom de Vries <tom@codesourcery.com>
1130 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1131 (nvptx_process_pars): Emit bar.sync asap and alap.
1133 2018-03-20 Tom de Vries <tom@codesourcery.com>
1136 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1137 seen_label if seen_label is already set.
1139 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1142 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1143 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1144 Use 1U instead of 1. Formatting fixes.
1147 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1148 instead of TREE_TYPE (s1) for the return value.
1150 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1152 PR tree-optimization/84946
1153 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1154 bitsize + bitsize in poly_uint64 rather than poly_int64.
1157 * dwarf2asm.c: Include fold-const.c.
1158 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1159 of decl rather than decl itself.
1161 PR rtl-optimization/84643
1162 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1164 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1167 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1168 calling assemble_variable.
1170 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1173 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1174 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1176 2018-03-19 Jim Wilson <jimw@sifive.com>
1179 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1180 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1181 (riscv_first_stack_step): Likewise.
1182 (riscv_option_override): Use STACK_BOUNDARY instead of
1184 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1186 (BIGGEST_ALIGNMENT): Set to 128.
1187 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1188 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1191 2018-03-19 Richard Biener <rguenther@suse.de>
1193 PR tree-optimization/84933
1194 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1195 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1197 2018-03-19 Richard Biener <rguenther@suse.de>
1199 PR tree-optimization/84859
1200 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1201 (cond_if_else_store_replacement): Perform sinking operation on
1202 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1203 Generalize what a BB with a single eligible store is.
1205 2018-03-19 Richard Biener <rguenther@suse.de>
1207 PR tree-optimization/84929
1208 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1209 chrec_is_positive against non-chrec arg.
1211 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1214 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1216 2018-03-18 Martin Liska <mliska@suse.cz>
1218 PR rtl-optimization/84635
1219 * regrename.c (build_def_use): Use matches_mode only when
1222 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1224 PR tree-optimization/84913
1225 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1226 vectorize chains of COND_EXPRs.
1228 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1230 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1232 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1234 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1236 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1238 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1240 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1241 Kito Cheng <kito.cheng@gmail.com>
1243 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1244 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1245 (nds32_adjust_reg_alloc_order): New function.
1246 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1248 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1250 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1251 nds32_print_operand, nds32_print_operand_address): Use
1252 HOST_WIDE_INT_PRINT_DEC instead.
1254 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1256 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1258 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1261 * config/i386/i386.c (initial_ix86_tune_features,
1262 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1264 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1265 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1266 rather than 1u << ix86_tune. Formatting fix.
1267 (ix86_option_override_internal): Change ix86_arch_mask from
1268 unsigned int to unsigned HOST_WIDE_INT, initialize to
1269 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1270 (ix86_function_specific_restore): Likewise.
1272 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1275 * postreload.c (reload_combine_recognize_pattern): Perform
1276 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1277 truncate_int_for_mode the result for the destination's mode.
1280 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1281 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1283 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1286 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1289 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1290 regno_allocno_class_array and sorted_pseudos.
1291 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1292 insns where regno is used.
1294 2018-03-16 Martin Liska <mliska@suse.cz>
1297 * multiple_target.c (create_dispatcher_calls): Redirect
1298 reference in the symbol table.
1300 2018-03-16 Martin Liska <mliska@suse.cz>
1303 * multiple_target.c (create_dispatcher_calls): Redirect also
1306 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1310 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1311 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1312 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1314 2018-03-16 Julia Koval <julia.koval@intel.com>
1316 * doc/invoke.texi (Skylake Server): Add CLWB.
1317 Cannonlake): Remove CLWB.
1319 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1321 PR tree-optimization/84841
1322 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1324 (FLOAT_ONE_CONST_TYPE): Define.
1325 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1326 (sort_by_operand_rank): Put entries with higher constant_type last
1327 rather than first to match comments.
1329 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1331 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1334 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1337 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1338 check and use address of target always.
1340 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1343 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1344 (indirect_thunk_bnd_needed): Likewise.
1345 (indirect_thunks_used): Likewise.
1346 (indirect_thunks_bnd_used): Likewise.
1347 (indirect_return_needed): New.
1348 (indirect_return_bnd_needed): Likewise.
1349 (output_indirect_thunk_function): Add a bool argument for
1351 (output_indirect_thunk_function): Don't generate alias for
1352 function return thunk.
1353 (ix86_code_end): Call output_indirect_thunk_function to generate
1354 function return thunks.
1355 (ix86_output_function_return): Set indirect_return_bnd_needed
1356 and indirect_return_needed instead of indirect_thunk_bnd_needed
1357 and indirect_thunk_needed.
1359 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1361 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1362 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1363 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1365 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1366 Paul Hua <paul.hua.gm@gmail.com>
1369 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1371 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1373 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1374 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1377 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1380 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1381 instead of GET_MODE_SIZE when comparing Units.
1383 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1386 * varasm.c (hash_section): Return an unchangeble hash value
1387 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1388 Return !aarch64_can_use_per_function_literal_pools_p ().
1390 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1393 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1394 rather than address of cmode as last argument to prepare_cmp_insn.
1396 2018-03-15 Julia Koval <julia.koval@intel.com>
1398 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1399 F_AVX512VNNI, F_AVX512BITALG): New.
1401 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1404 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1405 insn for floating-point loads and stores.
1407 2018-03-14 Carl Love <cel@us.ibm.com>
1409 * config/rs6000/rs6000-c.c: Add macro definitions for
1410 ALTIVEC_BUILTIN_VEC_PERMXOR.
1411 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1412 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1413 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1414 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1416 * config/doc/extend.texi: Add prototypes for vec_permxor.
1418 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1421 * diagnostic-show-locus.c (class layout_point): Convert m_line
1422 from int to linenum_type.
1423 (line_span::comparator): Use linenum "compare" function when
1424 comparing line numbers.
1425 (test_line_span): New function.
1426 (layout_range::contains_point): Convert param "row" from int to
1428 (layout_range::intersects_line_p): Likewise.
1429 (layout::will_show_line_p): Likewise.
1430 (layout::print_source_line): Likewise.
1431 (layout::should_print_annotation_line_p): Likewise.
1432 (layout::print_annotation_line): Likewise.
1433 (layout::print_leading_fixits): Likewise.
1434 (layout::annotation_line_showed_range_p): Likewise.
1435 (struct line_corrections): Likewise for field m_row.
1436 (line_corrections::line_corrections): Likewise for param "row".
1437 (layout::print_trailing_fixits): Likewise.
1438 (layout::get_state_at_point): Likewise.
1439 (layout::get_x_bound_for_row): Likewise.
1440 (layout::print_line): Likewise.
1441 (diagnostic_show_locus): Likewise for locals "last_line" and
1443 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1444 * input.c (selftest::test_linenum_comparisons): New function.
1445 (selftest::input_c_tests): Call it.
1446 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1447 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1448 * selftest.h (ASSERT_GT): New macro.
1449 (ASSERT_GT_AT): New macro.
1450 (ASSERT_LT): New macro.
1451 (ASSERT_LT_AT): New macro.
1453 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1455 PR rtl-optimization/84780
1456 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1458 2018-03-14 Martin Liska <mliska@suse.cz>
1460 * tree.c (record_node_allocation_statistics): Use
1461 get_stats_node_kind.
1462 (get_stats_node_kind): New function extracted from
1463 record_node_allocation_statistics.
1464 (free_node): Use get_stats_node_kind.
1466 2018-03-14 Richard Biener <rguenther@suse.de>
1468 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1469 that the value-set of ANTIC_IN doesn't grow.
1472 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1474 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1475 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1477 2018-03-14 Julia Koval <julia.koval@intel.com>
1479 * config.gcc (icelake-client, icelake-server): New.
1481 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1482 (initial_ix86_arch_features): Ditto.
1483 (PTA_SKYLAKE): Add SGX.
1484 (PTA_ICELAKE): Remove.
1485 (PTA_ICELAKE_CLIENT): New.
1486 (PTA_ICELAKE_SERVER): New.
1487 (ix86_option_override_internal): Split up icelake on icelake client and
1489 (get_builtin_code_for_version): Ditto.
1490 (fold_builtin_cpu): Ditto.
1491 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1492 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1493 * config/i386/i386.h (processor_type): Ditto.
1494 * doc/invoke.texi: Ditto.
1496 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1499 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1500 INTEGER_CST offset, add it together with bitpos / 8 and
1501 sign extend based on POINTER_SIZE.
1505 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1508 * config/i386/constraints.md (Yc): New register constraint.
1509 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1510 Use Yc constraint for alternative 2 of operand 0. Remove
1511 preferred_for_speed attribute.
1513 2018-03-14 Richard Biener <rguenther@suse.de>
1515 PR tree-optimization/84830
1516 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1517 with the old one to avoid oscillations.
1519 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1522 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1524 (assign_by_spills): Return a flag of reload assignment failure.
1525 Do not process the reload assignment failures. Do not spill other
1526 reload pseudos if they has the same reg class. Update n if
1528 (lra_assign): Add a return arg. Set up from the result of
1529 assign_by_spills call.
1530 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1531 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1532 usage_insns if it is not NULL.
1533 (spill_hard_reg_in_range): New function.
1534 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1535 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1536 function prototypes.
1537 (lra_assign): Change prototype.
1538 * lra.c (lra): Add code to deal with fails by splitting hard reg
1541 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1543 * config/riscv/riscv.opt (mrelax): New option.
1544 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1545 "norelax" when riscv_mrelax is disabled.
1546 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1548 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1551 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1552 reassociation for int modes.
1554 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1556 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1557 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1559 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1560 * config/aarch64/aarch64-sve.md
1561 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1562 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1563 (*extend<mode><Vwide>2): Rename to...
1564 (aarch64_sve_extend<mode><Vwide>2): ...this.
1565 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1566 renaming the old pattern to...
1567 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1569 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1570 define_expand, renaming the old pattern to...
1571 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1572 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1573 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1574 account when deciding which SVE instruction the optab should use.
1575 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1577 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1579 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1580 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1581 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1582 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1583 (tlsdesc_small_<mode>): Turn a define_expand and use
1584 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1585 (tlsdesc_small_advsimd_<mode>): ...this.
1586 (tlsdesc_small_sve_<mode>): New pattern.
1588 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1590 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1591 (UNSPEC_UMUL_HIGHPART): New constants.
1592 (MUL_HIGHPART): New int iteraor.
1593 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1594 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1596 (*<su>mul<mode>3_highpart): New define_insn.
1598 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1601 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1604 2018-03-13 Martin Liska <mliska@suse.cz>
1607 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1609 (sem_item_optimizer::~sem_item_optimizer): Release it.
1610 (sem_item_optimizer::merge_classes): Register variable aliases.
1611 (sem_item_optimizer::fixup_pt_set): New function.
1612 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1613 * ipa-icf.h: Declare new variables and functions.
1615 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1618 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1619 integer_pow2p@2 and test integer_pow2p in condition.
1620 (A < 0 ? C : 0): Similarly for @1.
1623 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1624 characters starting at p contain '\0' character, don't look beyond
1628 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1629 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1632 * reg-stack.c (change_stack): Change update_end var from int to
1633 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1634 also call set_block_for_insn on the newly added insns and rescan.
1637 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1638 on the last operand.
1641 * tree.c (stabilize_reference_1): Return save_expr (e) for
1642 STATEMENT_LIST even if it doesn't have side-effects.
1644 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1646 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1648 2018-03-12 Renlin Li <renlin.li@arm.com>
1650 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1651 aarch64_output_scalar_simd_mov_immediate.
1653 2018-03-12 Martin Sebor <msebor@redhat.com>
1655 PR tree-optimization/83456
1656 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
1657 for perfectly overlapping calls to memcpy.
1658 (gimple_fold_builtin_memory_chk): Same.
1659 (gimple_fold_builtin_strcpy): Handle no-warning.
1660 (gimple_fold_builtin_stxcpy_chk): Same.
1661 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
1663 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1665 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
1666 parameter. Use it for SFmode.
1667 (rs6000_function_arg_advance_1): Adjust.
1668 (rs6000_function_arg): Adjust.
1669 (rs6000_gimplify_va_arg): Pass false for that new parameter.
1671 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
1673 PR rtl-optimization/84169
1674 PR rtl-optimization/84780
1675 * combine.c (can_combine_p): Check for a 2-insn combination whether
1676 the destination register is used between the two insns, too.
1678 2018-03-12 Richard Biener <rguenther@suse.de>
1680 PR tree-optimization/84803
1681 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
1682 for refs DR analysis didn't process.
1684 2018-03-12 Richard Biener <rguenther@suse.de>
1686 PR tree-optimization/84777
1687 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
1688 force-vectorize loops ignore whether we are optimizing for size.
1690 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
1692 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
1693 (TARGET_MD_ASM_ADJUST): Define.
1695 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
1696 Kito Cheng <kito.cheng@gmail.com>
1697 Chung-Ju Wu <jasonwucj@gmail.com>
1699 * config/nds32/nds32.c (nds32_compute_stack_frame,
1700 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
1701 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
1702 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
1703 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
1704 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
1705 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
1706 * config/nds32/nds32.md (prologue, epilogue): Use macro
1707 NDS32_V3PUSH_AVAILABLE_P to do checking.
1709 2018-03-11 Jakub Jelinek <jakub@redhat.com>
1712 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
1713 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
1714 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
1715 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
1716 addition of most attributes on !orig_type_die or the attribute not
1717 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
1719 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1720 Chung-Ju Wu <jasonwucj@gmail.com>
1722 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
1724 * config/nds32/nds32.opt (mvh): New option.
1726 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1727 Chung-Ju Wu <jasonwucj@gmail.com>
1729 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
1731 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
1732 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
1735 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
1736 Chung-Ju Wu <jasonwucj@gmail.com>
1738 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
1740 * config/nds32/nds32-multiple.md (strlensi): New pattern.
1741 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
1743 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
1744 Kito Cheng <kito.cheng@gmail.com>
1745 Chung-Ju Wu <jasonwucj@gmail.com>
1747 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
1748 UNSPEC_FFMISM and UNSPEC_FLMISM.
1749 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
1750 for ffb, ffmism and flmism.
1751 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
1752 (unspec_ffmism): Ditto.
1753 (unspec_flmism): Ditto.
1754 (nds32_expand_builtin_impl): Check if string extension is available.
1755 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
1756 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
1758 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
1761 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1764 * lra-assigns.c (assign_by_spills): Return a flag of reload
1765 assignment failure. Do not process the reload assignment
1766 failures. Do not spill other reload pseudos if they has the same
1768 (lra_assign): Add a return arg. Set up from the result of
1769 assign_by_spills call.
1770 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1771 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1772 usage_insns if it is not NULL.
1773 (spill_hard_reg_in_range): New function.
1774 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1775 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1776 function prototypes.
1777 (lra_assign): Change prototype.
1778 * lra.c (lra): Add code to deal with fails by splitting hard reg
1781 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
1784 * config/i386/i386.opt: Replace Enforcment with Enforcement.
1786 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
1789 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
1790 (dw_val_node): Add val_symbolic_view.
1791 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
1792 (symview_upper_bound): New.
1793 (new_line_info_table): Initialize symviews_since_reset.
1794 (dwarf2out_source_line): Count symviews_since_reset and set
1795 symview_upper_bound.
1796 (dw_val_equal_p): Handle symview.
1797 (add_AT_symview): New.
1798 (print_dw_val): Handle symview.
1799 (attr_checksum, attr_checksum_ordered): Likewise.
1800 (same_dw_val_p, size_of_die): Likewise.
1801 (value_format, output_die): Likewise.
1802 (add_high_low_attributes): Use add_AT_symview for entry_view.
1803 (dwarf2out_finish): Reset symview_upper_bound, clear
1806 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
1809 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
1810 Add strict argument and use it.
1811 (rs6000_split_multireg_move): Update for new strict argument.
1812 (mem_operand_gpr): Disallow all non-offsettable addresses.
1813 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
1815 2018-03-09 Jakub Jelinek <jakub@redhat.com>
1818 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
1819 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
1820 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
1823 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
1824 decl, use remap_type if we want to use the type.
1826 2018-03-09 Martin Sebor <msebor@redhat.com>
1828 PR tree-optimization/84526
1829 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1831 (builtin_access::generic_overlap): Be prepared to handle non-array
1834 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
1836 PR rtl-optimization/84682
1837 * lra-constraints.c (process_address_1): Check is_address flag
1838 for address constraints.
1839 (process_alt_operands): Likewise.
1840 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
1841 preprocess_constraints.
1842 * recog.h (preprocess_constraints): Add oploc parameter.
1844 * recog.c (preprocess_constraints): Test address_operand for
1845 CT_ADDRESS constraints.
1847 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
1850 * lra-assigns.c (assign_by_spills): Return a flag of reload
1851 assignment failure. Do not process the reload assignment
1852 failures. Do not spill other reload pseudos if they has the same
1854 (lra_assign): Add a return arg. Set up from the result of
1855 assign_by_spills call.
1856 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1857 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1858 usage_insns if it is not NULL.
1859 (spill_hard_reg_in_range): New function.
1860 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1861 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1862 function prototypes.
1863 (lra_assign): Change prototype.
1864 * lra.c (lra): Add code to deal with fails by splitting hard reg
1867 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1870 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
1871 Accept complain bool parameter. Only emit errors if it is true.
1872 (arm_parse_cpu_option_name): Likewise.
1873 (arm_target_thumb_only): Adjust callers of the above.
1874 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
1875 prototype to take a default true bool parameter.
1876 (arm_parse_arch_option_name): Likewise.
1878 2018-03-09 David Malcolm <dmalcolm@redhat.com>
1879 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1883 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
1884 * configure: Regenerate.
1885 * configure.ac ("linker --version-script option"): New.
1886 ("linker soname option"): New.
1888 2018-03-09 Richard Biener <rguenther@suse.de>
1890 PR tree-optimization/84775
1891 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
1892 immediate uses of predicate stmts and mark them modified.
1895 PR tree-optimization/84178
1896 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1898 (version_loop_for_if_conversion): Delay update_ssa call.
1899 (tree_if_conversion): Delay update_ssa until after predicate
1902 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
1905 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
1906 when the function accesses prior frames.
1908 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1911 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
1912 gen_llsym, otherwise call maybe_gen_llsym.
1915 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
1916 has ',' character inside of it.
1918 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1921 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
1922 as clobbering CC_REGNUM.
1924 2018-03-08 Richard Biener <rguenther@suse.de>
1927 * tree-scalar-evolution.c: Include tree-into-ssa.h.
1928 (follow_copies_to_constant): Do not follow SSA names registered
1931 2018-03-08 Richard Biener <rguenther@suse.de>
1933 PR tree-optimization/84178
1934 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
1936 (version_loop_for_if_conversion): Delay update_ssa call.
1937 (tree_if_conversion): Delay update_ssa until after predicate
1940 2018-03-08 David Malcolm <dmalcolm@redhat.com>
1942 PR tree-optimization/84178
1943 * tree-if-conv.c (release_bb_predicate): Remove the
1944 the assertion that the stmts have NULL use_ops.
1945 Discard the statements, asserting that they haven't
1946 yet been added to a BB.
1948 2018-03-08 Richard Biener <rguenther@suse.de>
1950 PR tree-optimization/84746
1951 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
1952 (phi_translate): Pass in destination ANTIC_OUT set.
1953 (phi_translate_1): Likewise. For a simplified result lookup
1954 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
1955 (phi_translate_set): Adjust.
1956 (do_pre_regular_insertion): Likewise.
1957 (do_pre_partial_partial_insertion): Likewise.
1959 2018-03-08 Martin Liska <mliska@suse.cz>
1961 PR gcov-profile/84735
1962 * doc/gcov.texi: Document usage of profile files.
1963 * gcov-io.h: Document changes in the format.
1965 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
1969 * dwarf2out.c (struct dw_line_info_table): Update comments for
1971 (FORCE_RESET_NEXT_VIEW): New.
1972 (FORCE_RESETTING_VIEW_P): New.
1973 (RESETTING_VIEW_P): Check for -1 too.
1974 (ZERO_VIEW_P): Likewise.
1975 (new_line_info_table): Force-reset next view.
1976 (dwarf2out_begin_function): Likewise.
1977 (dwarf2out_source_line): Simplify zero_view_p initialization.
1978 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
1979 view directly. Omit view when omitting .loc at line 0.
1981 2018-03-08 Jakub Jelinek <jakub@redhat.com>
1983 PR tree-optimization/84740
1984 * tree-switch-conversion.c (process_switch): Call build_constructors
1985 only if info.phi_count is non-zero.
1987 PR tree-optimization/84739
1988 * tree-tailcall.c (find_tail_calls): Check call arguments against
1989 DECL_ARGUMENTS (current_function_decl) rather than
1990 DECL_ARGUMENTS (func) when checking for tail recursion.
1992 2018-03-07 Jakub Jelinek <jakub@redhat.com>
1994 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
1995 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
1996 Volker Reichelt's entry and add entries for people that perform
1997 GCC fuzzy testing and report numerous bugs.
1999 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2002 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2003 readonly data in sdata, if that is disabled.
2004 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2005 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2006 -mreadonly-in-sdata option.
2008 2018-03-07 Martin Sebor <msebor@redhat.com>
2010 PR tree-optimization/84468
2011 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2012 basic block when looking for nul assignment.
2014 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2017 * except.h (output_function_exception_table): Adjust prototype.
2018 * except.c (output_function_exception_table): Remove FNNAME parameter
2019 and add SECTION parameter. Ouput one part of the table at a time.
2020 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2021 the first part of the exception table and emit unwind directives.
2022 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2023 (i386_pe_seh_cold_init): Likewise.
2024 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2025 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2026 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2027 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2028 * config/i386/winnt.c: Include except.h.
2029 (struct seh_frame_state): Add reg_offset, after_prologue and
2030 in_cold_section fields.
2031 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2032 (i386_pe_seh_cold_init): New function.
2033 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2034 to seh->in_cold_section.
2035 (seh_emit_push): Record the offset of the push.
2036 (seh_emit_save): Record the offet of the save.
2037 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2038 Test seh->after_prologue to disregard the epilogue.
2039 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2040 (i386_pe_end_cold_function): New function.
2042 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2045 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2046 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2049 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2051 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2052 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2055 * multiple_target.c: Include tree-inline.h and intl.h.
2056 (expand_target_clones): Diagnose and fail if node->definition and
2057 !tree_versionable_function_p (node->decl).
2059 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2061 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2063 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2064 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2065 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2067 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2070 * combine.c (try_combine): Use reg_or_subregno instead of handling
2071 just paradoxical SUBREGs and REGs.
2073 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2075 * config/arc/arc.c (arc_finalize_pic): Remove function.
2076 (arc_must_save_register): We use single base PIC register, remove
2077 checks to save/restore the PIC register.
2078 (arc_expand_prologue): Likewise.
2079 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2081 (arc_verify_short): Likewise.
2082 (arc_attr_type): Likewise.
2083 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2084 (walk_stores): Likewise.
2085 (arc_address_cost): Make it static.
2086 (arc_verify_short): Likewise.
2087 (branch_dest): Likewise.
2088 (arc_attr_type): Likewise.
2089 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2090 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2091 (arc_final_prescan_insn): Remove inserting the nops due to
2092 hardware hazards. It is done in reorg step.
2093 (insn_length_variant_t): Remove.
2094 (insn_length_parameters_t): Likewise.
2095 (arc_insn_length_parameters): Likewise.
2096 (arc_get_insn_variants): Likewise.
2097 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2099 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2102 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2105 PR tree-optimization/84687
2106 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2108 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2110 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2112 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2113 Rename to ppc_speculation_barrier.
2114 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2115 __builtin_ppc_speculation_barrier.
2117 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2120 * combine.c (combine_simplify_rtx): Don't try to simplify if
2121 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2124 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2126 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2127 to 32 bytes when compiling for POWER9.
2129 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2132 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2133 regparm >= 3 with no arg reg available also for calls with
2134 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2137 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2139 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2141 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2144 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2146 2018-03-05 Richard Biener <rguenther@suse.de>
2148 PR tree-optimization/84486
2149 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2150 When inserting a __builtin_assume_aligned call set the LHS
2151 SSA name alignment info accordingly.
2153 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2155 PR tree-optimization/84114
2156 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2157 Avoid reassociation of FLOAT_MODE addition.
2159 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2161 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2162 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2163 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2164 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2165 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2166 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2167 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2169 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2170 __builtin_ia32_wbinvd): New builtins.
2171 (SPECIAL_ARGS2): New.
2172 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2173 (SPECIAL_ARGS2): New.
2174 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2175 (ix86_valid_target_attribute_inner_p): Ditto.
2176 (ix86_init_mmx_sse_builtins): Add special_args2.
2177 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2178 TARGET_WBNOINVD_P): New.
2179 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2180 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2181 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2182 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2183 * config/i386/pconfigintrin.h: New file.
2184 * config/i386/wbnoinvdintrin.h: Ditto.
2185 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2186 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2188 2018-03-05 Richard Biener <rguenther@suse.de>
2190 PR tree-optimization/84670
2191 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2193 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2194 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2195 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2196 doesn't grow if all successors (recursively) were visited at least
2199 2018-03-05 Richard Biener <rguenther@suse.de>
2201 PR tree-optimization/84650
2202 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2203 if executed in the loop pipeline.
2205 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2207 * doc/configfiles.texi (Configuration Files): Move info about
2208 conditionalizing $target-protos.h to...
2209 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2210 differs from $target-protos.h.
2212 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2213 Chung-Ju Wu <jasonwucj@gmail.com>
2215 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2216 * config/nds32/nds32-multiple.md (setmemsi): Define.
2217 * config/nds32/nds32-memory-manipulation.c
2218 (nds32_gen_dup_4_byte_to_word_value): New.
2219 (emit_setmem_word_loop): New.
2220 (emit_setmem_byte_loop): New.
2221 (nds32_expand_setmem_loop): New.
2222 (nds32_expand_setmem_loop_v3m): New.
2223 (nds32_expand_setmem_unroll): New.
2224 (nds32_expand_setmem): New.
2226 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2227 Chung-Ju Wu <jasonwucj@gmail.com>
2229 * config/nds32/nds32-memory-manipulation.c
2230 (nds32_emit_load_store): New.
2231 (nds32_emit_post_inc_load_store): New.
2232 (nds32_emit_mem_move): New.
2233 (nds32_emit_mem_move_block): New.
2234 (nds32_expand_movmemsi_loop_unknown_size): New.
2235 (nds32_expand_movmemsi_loop_known_size): New.
2236 (nds32_expand_movmemsi_loop): New.
2237 (nds32_expand_movmemsi_unroll): New.
2238 (nds32_expand_movmemqi): Rename ...
2239 (nds32_expand_movmemsi): ... to this.
2240 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2241 (movmemsi): ... to this.
2242 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2243 (nds32_expand_movmemsi): ... to this.
2245 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2246 Monk Chiang <sh.chiang04@gmail.com>
2247 Chung-Ju Wu <jasonwucj@gmail.com>
2249 * config/nds32/nds32-protos.h
2250 (nds32_expand_load_multiple): New arguments.
2251 (nds32_expand_store_multiple): Ditto.
2252 (nds32_valid_multiple_load_store): Rename ...
2253 (nds32_valid_multiple_load_store_p): ... to this.
2254 * config/nds32/nds32-memory-manipulation.c
2255 (nds32_expand_load_multiple): Refine implementation.
2256 (nds32_expand_store_multiple): Ditto.
2257 * config/nds32/nds32-multiple.md
2258 (load_multiple): Update nds32_expand_load_multiple interface.
2259 (store_multiple): Update nds32_expand_store_multiple interface.
2260 * config/nds32/nds32-predicates.c
2261 (nds32_valid_multiple_load_store): Rename ...
2262 (nds32_valid_multiple_load_store_p): ... to this and refine
2264 * config/nds32/predicates.md
2265 (nds32_load_multiple_and_update_address_operation): New predicate.
2266 (nds32_store_multiple_and_update_address_operation): New predicate.
2268 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2269 Chung-Ju Wu <jasonwucj@gmail.com>
2271 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2272 (combo): New attribute.
2273 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2275 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2277 * config/nds32/nds32.opt: Change -mcmodel= default value.
2279 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2280 Monk Chiang <sh.chiang04@gmail.com>
2281 Chung-Ju Wu <jasonwucj@gmail.com>
2283 * config/nds32/constants.md (unspec_element): New enum.
2284 * config/nds32/constraints.md (Umw): New constraint.
2285 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2286 * config/nds32/nds32-intrinsic.md: Likewise.
2287 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2288 (nds32_valid_smw_lwm_base_p): New.
2289 (nds32_output_smw_single_word): New.
2290 (nds32_output_lmw_single_word): New.
2291 (nds32_expand_unaligned_load): New.
2292 (nds32_expand_unaligned_store): New.
2293 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2294 (nds32_output_smw_single_word): Declare.
2295 (nds32_output_lmw_single_word): Declare.
2296 (nds32_expand_unaligned_load): Declare.
2297 (nds32_expand_unaligned_store): Declare.
2298 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2299 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2300 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2301 NDS32_BUILTIN_UASTORE_DW.
2302 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2305 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2306 Kito Cheng <kito.cheng@gmail.com>
2307 Chung-Ju Wu <jasonwucj@gmail.com>
2309 * config/nds32/nds32-intrinsic.c
2310 (nds32_expand_builtin_null_ftype_reg): Delete.
2311 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2312 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2313 (nds32_read_argument): New.
2314 (nds32_legitimize_target): Ditto.
2315 (nds32_legitimize_argument): Ditto.
2316 (nds32_check_constant_argument): Ditto.
2317 (nds32_expand_unop_builtin): Ditto.
2318 (nds32_expand_unopimm_builtin): Ditto.
2319 (nds32_expand_binop_builtin): Ditto.
2320 (nds32_builtin_decl_impl): Ditto.
2321 (builtin_description): Ditto.
2322 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2323 (nds32_init_builtins_impl): Ditto.
2324 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2325 (nds32_builtin_decl): New.
2326 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2327 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2329 2018-03-02 Jeff Law <law@redhat.com>
2331 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2332 (redundant_insn, fill_simple_delay_slots): Likewise.
2333 (fill_slots_from_thread): Likewise.
2334 * resource.c (mark_referenced_resources): Likewise.
2335 (mark_set_resources, find_dead_or_set_registers): Likewise.
2337 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2339 * substring-locations.h (format_warning_va): Formatting fix for
2341 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2343 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2344 * substring-locations.c: Include intl.h.
2345 (format_warning_va): Turned into small wrapper around
2346 format_warning_n_va, renamed to ...
2347 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2348 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2350 (format_warning_at_substring_n): New function.
2351 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2352 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2353 format_warning_at_substring with just a shorter name instead of
2354 const function pointer.
2355 (fmtwarn_n): New function.
2356 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2357 appropriate, get rid of all the fmtstr temporaries, move conditionals
2358 with G_() wrapped string literals directly into fmtwarn arguments,
2359 cast dir.len to (int), formatting fixes.
2361 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2363 * doc/invoke.texi: Remove "Cilk Plus" references.
2365 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2366 Richard Biener <rguenther@suse.de>
2369 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2370 for error or warning attributes if CALL_FROM_THUNK_P is set.
2373 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2376 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2377 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2380 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2381 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2383 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2384 instead of -1U in last predictors element's probability member.
2386 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2389 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2390 arguments if they are comparable.
2392 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2394 PR tree-optimization/84634
2395 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2396 masks and masked_loop_p with a single loop_masks, making sure it's
2397 null for bb vectorization.
2399 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2401 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2402 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2403 loop->force_vectorize to check whether there is no alias.
2405 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2408 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2410 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2412 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2413 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2414 instead of a loop around prev_real_insn.
2415 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2419 * config/i386/i386.c (ix86_print_operand): Use conditional
2420 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2423 2018-03-02 Richard Biener <rguenther@suse.de>
2425 PR tree-optimization/84427
2426 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2427 (bitmap_set_subtract_values): Rewrite to handle multiple
2430 (prune_clobbered_mems): Likewise.
2431 (phi_translate): Take edge instead of pred/phiblock.
2432 (phi_translate_1): Likewise.
2433 (phi_translate_set): Likewise. Insert all translated
2434 exprs for a value into the set, keeping possibly multiple
2435 expressions per value.
2436 (compute_antic_aux): Adjust for phi_translate changes.
2437 When intersecting union the expressions and prune those
2438 not in the final value set, keeping possibly multiple
2439 expressions per value. Do not use value-insertion
2440 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2441 all expressions. Add verification that the value-sets
2442 only shrink during iteration.
2443 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2444 (do_pre_regular_insertion): Likewise.
2445 (do_pre_partial_partial_insertion): Likewise.
2447 2018-03-02 Richard Biener <rguenther@suse.de>
2450 * config/darwin.c (saved_debug_info_level): New static global.
2451 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2452 (darwin_asm_lto_end): Restore debug info generation settings.
2454 2018-03-01 Martin Liska <mliska@suse.cz>
2457 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2460 2018-03-01 Richard Biener <rguenther@suse.de>
2463 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2464 type annotation with LTO when debug was disabled at compile-time.
2466 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2468 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2470 (mips_final_postscan_insn): Likewise.
2472 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2474 PR rtl-optimization/84528
2475 * alias.c (init_alias_target): Add commentary.
2476 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2477 a unique base value if the frame pointer is not eliminated
2478 to the stack pointer.
2480 2018-03-01 Tom de Vries <tom@codesourcery.com>
2482 PR rtl-optimization/83327
2483 * lra-int.h (hard_regs_spilled_into): Declare.
2484 * lra.c (hard_regs_spilled_into): Define.
2485 (init_reg_info): Init hard_regs_spilled_into.
2486 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2487 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2488 (process_bb_lives): Handle hard_regs_spilled_into.
2489 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2490 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2492 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2494 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2495 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2496 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2497 * config/rs6000/aix72.h: New file.
2499 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2501 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2502 instead of warning_at with conditional singular and plural messages
2506 * stor-layout.c (update_alignment_for_field): For
2507 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2508 && !DECL_PACKED (field), do the alignment update, just use
2509 only desired_align instead of MAX (type_align, desired_align)
2511 (place_field): Don't do known_align < desired_align handling
2512 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2513 is non-NULL, instead do it after rli->prev_field handling and
2514 only if not within a bitfield word. For DECL_PACKED (field)
2515 use type_align of BITS_PER_UNIT.
2517 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2519 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2520 superfluous parentheses and trailing spaces.
2522 2018-02-28 Richard Biener <rguenther@suse.de>
2524 PR tree-optimization/84584
2525 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2526 SCoPs with fake exit edge.
2528 2018-02-28 Martin Liska <mliska@suse.cz>
2531 * timevar.c (timer::print): Fix format to properly print 100%
2534 2018-02-28 Richard Biener <rguenther@suse.de>
2537 * genmatch.c (capture_info::walk_match): Do not mark
2538 captured expressions without operands as expr_p given
2539 they act more like predicates and should be subject to
2540 "lost tail" side-effect preserving.
2542 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2544 PR rtl-optimization/81611
2545 * auto-inc-dec.c (attempt_change): Move dead note from
2546 mem_insn if it's the next use of regno
2547 (find_address): Take address use of reg holding
2548 non-incremented value. Add parm to limit search to the named
2550 (merge_in_block): Attempt to use a mem insn that is the next
2551 use of the original regno.
2553 2018-02-27 Martin Sebor <msebor@redhat.com>
2556 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2557 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2559 2018-02-27 Martin Sebor <msebor@redhat.com>
2561 PR translation/84207
2562 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2563 n argument to unsigned HOST_WIDE_INT.
2564 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2565 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2566 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2567 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2569 2018-02-27 Richard Biener <rguenther@suse.de>
2571 PR tree-optimization/84512
2572 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2573 Do not use the estimate returned from record_stmt_cost for
2574 the scalar iteration cost but sum properly using add_stmt_cost.
2576 2018-02-27 Richard Biener <rguenther@suse.de>
2578 PR tree-optimization/84466
2579 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2580 Adjust last change to less strictly validate use operands.
2582 2018-02-27 Martin Liska <mliska@suse.cz>
2584 PR gcov-profile/84548
2585 * gcov.c (process_file): Allow partial overlap and consider it
2586 also as group functions.
2587 (output_lines): Properly calculate range of lines for a group.
2589 2018-02-27 Martin Liska <mliska@suse.cz>
2591 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2592 'ggc' suffixes. Change first column width.
2593 (timer::print): Fix formatting of the column.
2595 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2597 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2598 preserve inline entry blocks for the sake of debug inline
2599 entry point markers alone.
2600 (remove_unused_locals): Suggest in comments a better place to
2601 force the preservation of inline entry blocks that are
2602 otherwise unused, but do not preserve them.
2604 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2606 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2608 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2611 * config/i386/constraints.md (Bs): Replace
2612 ix86_indirect_branch_register with
2613 TARGET_INDIRECT_BRANCH_REGISTER.
2615 * config/i386/i386.md (indirect_jump): Likewise.
2616 (tablejump): Likewise.
2617 (*sibcall_memory): Likewise.
2618 (*sibcall_value_memory): Likewise.
2619 Peepholes of indirect call and jump via memory: Likewise.
2620 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2621 (*sibcall_value_GOT_32): Likewise.
2622 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2623 (GOT_memory_operand): Likewise.
2624 (call_insn_operand): Likewise.
2625 (sibcall_insn_operand): Likewise.
2626 (GOT32_symbol_operand): Likewise.
2627 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2629 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2631 PR rtl-optimization/83496
2632 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2633 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2634 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2635 redundant insn, if any.
2636 (relax_delay_slots): Likewise.
2637 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2639 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2641 PR tree-optimization/83965
2642 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2643 that grouped statements are part of a reduction chain. Return
2644 true if the statement is not marked as a reduction itself but
2646 (vect_recog_dot_prod_pattern): Don't check whether the statement
2647 is part of a group here.
2648 (vect_recog_sad_pattern): Likewise.
2649 (vect_recog_widen_sum_pattern): Likewise.
2651 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2654 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2656 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2659 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
2661 (ix86_output_indirect_function_return): New prototype.
2662 (ix86_split_simple_return_pop_internal): Likewise.
2663 * config/i386/i386.c (indirect_return_via_cx): New.
2664 (indirect_return_via_cx_bnd): Likewise.
2665 (indirect_thunk_name): Handle return va CX_REG.
2666 (output_indirect_thunk_function): Create alias for
2667 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
2668 (ix86_output_indirect_jmp): Remove the bool argument.
2669 (ix86_output_indirect_function_return): New function.
2670 (ix86_split_simple_return_pop_internal): Likewise.
2671 * config/i386/i386.md (*indirect_jump): Don't pass false
2672 to ix86_output_indirect_jmp.
2673 (*tablejump_1): Likewise.
2674 (simple_return_pop_internal): Change it to define_insn_and_split.
2675 Call ix86_split_simple_return_pop_internal to split it for
2677 (simple_return_indirect_internal): Call
2678 ix86_output_indirect_function_return instead of
2679 ix86_output_indirect_jmp.
2681 2018-02-26 Jakub Jelinek <jakub@redhat.com>
2684 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
2685 memset and value initialization afterwards.
2687 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
2689 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
2691 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2694 * common/config/aarch64/aarch64-common.c
2695 (aarch_option_optimization_table[]): Switch
2696 off fomit-frame-pointer
2698 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2699 Chung-Ju Wu <jasonwucj@gmail.com>
2701 * config/nds32/nds32-multiple.md (load_multiple): Disallow
2703 (store_multiple): Ditto.
2705 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
2707 * config.gcc: Add --with-cpu support for nds32 target.
2708 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
2709 * config/nds32/nds32.opt: Add -mcpu= option.
2711 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
2713 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
2714 isel=yes): Warn for these deprecated options.
2716 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
2718 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
2719 ISA_2_5_MASKS_EMBEDDED.
2721 2018-02-23 Jakub Jelinek <jakub@redhat.com>
2723 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
2724 p->max as pointers rather than using iterative_hash_expr.
2726 2018-02-23 Carl Love <cel@us.ibm.com>
2728 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
2729 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
2731 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
2732 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
2733 P8V_BUILTIN_VEC_VUNSIGNED2.
2735 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
2738 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
2739 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
2741 * lra-constraints.c (curr_insn_transform): Set up
2742 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
2744 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
2745 * lra-eliminations.c (spill_pseudos): Ditto.
2746 (process_insn_for_elimination): Ditto.
2747 * lra-lives.c (reg_early_clobber_p): Use the new macros.
2748 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
2749 LRA_NON_CLOBBERED_ALT.
2751 2018-02-22 Martin Sebor <msebor@redhat.com>
2753 PR tree-optimization/84480
2754 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
2755 to maybe_diag_stxncpy_trunc. Call it.
2756 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
2757 from gimple_fold_builtin_strcpy. Print inlining stack.
2758 (handle_builtin_stxncpy): Print inlining stack.
2759 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
2761 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
2764 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
2765 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
2766 and -fcheck-pointer-bounds are used together.
2767 (indirect_thunk_prefix): New enum.
2768 (indirect_thunk_need_prefix): New function.
2769 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
2770 "_nt" instead of "_bnd" for NOTRACK prefix.
2771 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
2772 (output_indirect_thunk_function): Likewise.
2774 (ix86_code_end): Update output_indirect_thunk_function calls.
2775 (ix86_output_indirect_branch_via_reg): Replace
2776 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
2777 (ix86_output_indirect_branch_via_push): Likewise.
2778 (ix86_output_function_return): Likewise.
2779 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
2780 incompatible with -fcf-protection=branch and
2781 -fcheck-pointer-bounds.
2783 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2786 * config/aarch64/aarch64.c (aarch64_print_address_internal):
2787 Change gcc_assert call to output_operand_lossage.
2789 2018-02-22 Steve Ellcey <sellcey@cavium.com>
2791 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
2793 2018-02-22 DJ Delorie <dj@redhat.com>
2794 Sebastian Perta <sebastian.perta@renesas.com>
2795 Oleg Endo <olegendo@gcc.gnu.org>
2797 * config/rx/rx.c (rx_rtx_costs): New function.
2798 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
2800 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
2802 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
2804 2018-02-22 Martin Liska <mliska@suse.cz>
2807 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2808 Add "native" as a possible value.
2810 2018-02-22 Martin Liska <mliska@suse.cz>
2813 * config/i386/i386.c (ix86_option_override_internal):
2814 Add "native" as a possible value for -march and -mtune.
2816 2018-02-22 Jakub Jelinek <jakub@redhat.com>
2819 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
2820 to all type variants.
2822 PR tree-optimization/84503
2823 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
2824 width as info->bitpos + info->bitsize - start.
2825 (merged_store_group::merge_overlapping): Simplify width computation.
2826 (check_no_overlap): New function.
2827 (imm_store_chain_info::try_coalesce_bswap): Compute expected
2828 start + width and last_order of the group, fail if check_no_overlap
2830 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
2831 to group if check_no_overlap fails.
2833 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2835 * config/rs6000/altivec.md: Delete contraint arguments to
2836 define_expand, define_split, and define_peephole2, and in
2837 define_insn_and_split if always unused.
2838 * config/rs6000/darwin.md: Ditto.
2839 * config/rs6000/dfp.md: Ditto.
2840 * config/rs6000/rs6000.md: Ditto.
2841 * config/rs6000/sync.md: Ditto.
2842 * config/rs6000/vector.md: Ditto.
2843 * config/rs6000/vsx.md: Ditto.
2845 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
2847 * config/rs6000/altivec.md: Write output control strings as braced
2848 blocks instead of double-quoted strings.
2849 * config/rs6000/darwin.md: Ditto.
2850 * config/rs6000/rs6000.md: Ditto.
2851 * config/rs6000/vector.md: Ditto.
2852 * config/rs6000/vsx.md: Ditto.
2854 2018-02-21 Jason Merrill <jason@redhat.com>
2856 PR c++/84314 - ICE with templates and fastcall attribute.
2857 * attribs.c (build_type_attribute_qual_variant): Remove assert.
2859 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2861 * ipa-cp.c (determine_versionability): Fix comment typos.
2863 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
2866 * ipa-cp.c (determine_versionability): Do not version functions caling
2869 2018-02-21 Martin Liska <mliska@suse.cz>
2872 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
2873 Add "native" as a possible value.
2874 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
2875 the macro when native cpu detection is available.
2877 2018-02-21 Martin Liska <mliska@suse.cz>
2880 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
2881 Add "native" as a possible value.
2882 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
2883 when native cpu detection is available.
2885 2018-02-21 Jakub Jelinek <jakub@redhat.com>
2886 Martin Sebor <msebor@redhat.com>
2888 PR tree-optimization/84478
2889 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
2891 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
2892 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
2893 support which is conservatively correct, for 2 only stay conservative
2894 for maxlen. Formatting and comment capitalization fixes. Add STRICT
2895 argument to the 2 argument get_range_strlen, adjust 6 arg
2896 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
2898 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
2899 (gimple_fold_builtin_strlen): Pass true as last argument to
2902 2018-02-20 Martin Sebor <msebor@redhat.com>
2905 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
2906 (builtin_memref::set_base_and_offset): Same. Handle inner references.
2907 (builtin_memref::builtin_memref): Factor out parts into
2908 set_base_and_offset and call it.
2910 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2913 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
2914 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
2915 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
2916 search at the associated MODE_INT.
2918 2018-02-20 Jeff Law <law@redhat.com>
2921 PR tree-optimization/81592
2923 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
2924 for range data rather than using global data.
2925 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
2926 range data rather than using global data.
2927 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
2928 pass it to children as needed.
2929 (struct directive::fmtresult): Similarly.
2930 (struct directive::set_width): Similarly.
2931 (struct directive::set_precision): Similarly.
2932 (format_integer, format_directive, parse_directive): Similarly.
2933 (format_none): Accept unnamed vr_values parameter.
2934 (format_percent, format_floating, format_character): Similarly.
2935 (format_string, format_plain): Similarly.
2936 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
2937 the EVRP range analyzer for range data rather than using global data.
2938 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
2939 gimple-ssa-evrp-analyze.h
2940 (class sprintf_dom_walker): Add after_dom_children member function.
2941 Add evrp_range_analyzer member.
2942 (sprintf_dom_walker::before_dom_children): Call into the EVRP
2943 range analyzer as needed.
2944 (sprintf_dom_walker::after_dom_children): New member function.
2945 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
2947 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
2948 (evrp_range_analyzer::pop_to_marker): Likewise.
2950 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
2952 PR tree-optimization/84419
2953 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
2954 with the required type if its current type is compatible but
2957 2018-02-20 Jakub Jelinek <jakub@redhat.com>
2960 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
2961 after vectorization.
2963 2018-02-20 Martin Liska <mliska@suse.cz>
2966 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
2967 possible values if we don't have a hint.
2969 2018-02-20 Martin Liska <mliska@suse.cz>
2973 * final.c (shorten_branches): Build align_tab array with one
2975 * opts.c (finish_options): Add alignment option limit check.
2976 (MAX_CODE_ALIGN): Likewise.
2977 (MAX_CODE_ALIGN_VALUE): Likewise.
2978 * doc/invoke.texi: Document maximum allowed option value for
2979 all -falign-* options.
2981 2018-02-19 Jakub Jelinek <jakub@redhat.com>
2984 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
2985 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
2986 * var-tracking.c (emit_note_insn_var_location): Remove all references
2987 to NOTE_INSN_CALL_ARG_LOCATION.
2988 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
2989 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
2990 Use copy_rtx_if_shared.
2991 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
2992 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
2993 (dwarf2out_var_location): Remove handling of
2994 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
2996 * final.c (final_scan_insn): Remove all references to
2997 NOTE_INSN_CALL_ARG_LOCATION.
2998 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
2999 before dumping final insns.
3000 * except.c (emit_note_eh_region_end): Remove all references to
3001 NOTE_INSN_CALL_ARG_LOCATION.
3002 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3003 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3004 * config/arc/arc.c (hwloop_optimize): Likewise.
3005 * config/arm/arm.c (create_fix_barrier): Likewise.
3006 * config/s390/s390.c (s390_chunkify_start): Likewise.
3007 * config/sh/sh.c (find_barrier): Likewise.
3008 * config/i386/i386.c (rest_of_insert_endbranch,
3009 ix86_seh_fixup_eh_fallthru): Likewise.
3010 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3011 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3012 * config/frv/frv.c (frv_function_prologue): Likewise.
3013 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3015 (note_outside_basic_block_p): Remove all references to
3016 NOTE_INSN_CALL_ARG_LOCATION.
3017 * gengtype.c (adjust_field_rtx_def): Likewise.
3018 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3020 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3021 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3024 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3027 PR tree-optimization/84452
3028 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3029 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3032 2018-02-19 Martin Liska <mliska@suse.cz>
3035 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3037 2018-02-19 Martin Liska <mliska@suse.cz>
3038 Richard Sandiford <richard.sandiford@linaro.org>
3040 PR tree-optimization/82491
3041 * gimple-fold.c (get_base_constructor): Make earlier bail out
3044 2018-02-19 Carl Love <cel@us.ibm.com>
3046 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3047 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3049 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3050 P8V_BUILTIN_VEC_NEG.
3052 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3054 * config/rl78/rl78.md (movdf): New define expand.
3056 2018-02-19 Martin Liska <mliska@suse.cz>
3059 * doc/invoke.texi: Fix typo.
3060 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3062 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3064 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3065 handle rs6000_single_float and rs6000_double_float specially for
3068 2018-02-16 Jeff Law <law@redhat.com>
3070 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3073 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3076 * ipa-inline.c (inline_small_functions): Fix a typo.
3078 2018-02-16 Nathan Sidwell <nathan@acm.org>
3080 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3082 2018-02-16 Carl Love <cel@us.ibm.com>
3084 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3085 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3086 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3087 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3088 expansion to P8V_BUILTIN_VEC_FLOAT2.
3090 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3092 PR rtl-optimization/70023
3093 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3094 src_regno into account.
3096 2018-02-16 Carl Love <cel@us.ibm.com>
3098 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3099 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3100 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3101 * config/rs6000/rs6000.c: Remove case statements for
3102 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3103 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3104 and P9V_BUILTIN_VEC_VINSERT4B.
3105 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3106 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3107 * config/rs6000/vsx.md:
3108 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3111 2018-02-16 Carl Love <cel@us.ibm.com>
3113 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3115 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3117 * config/rs6000/rs6000-c.c: Add the definitions for
3118 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3119 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3120 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3121 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3122 definition for insert4b and define insn *insert3b_internal.
3123 * doc/extend.texi: Add documentation for vec_extract4b.
3125 2018-02-16 Nathan Sidwell <nathan@acm.org>
3127 * doc/extend.texi (Backwards Compatibility): Mention friend
3128 injection. Note for-scope is deprecated.
3129 * doc/invoke.texi (-ffriend-injection): Deprecate.
3131 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3133 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3134 that moved to I2, also allow destinations that are a paradoxical
3135 subreg (instead of a normal reg).
3137 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3140 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3143 2018-02-16 Richard Biener <rguenther@suse.de>
3145 PR tree-optimization/84037
3146 PR tree-optimization/84016
3148 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3149 Adjust vec_construct for the fact we need additional higher latency
3150 128bit inserts for AVX256 and AVX512 vector builds.
3151 (ix86_add_stmt_cost): Scale vector construction cost for
3154 2018-02-16 Richard Biener <rguenther@suse.de>
3156 PR tree-optimization/84417
3157 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3158 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3159 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3161 2018-02-16 Martin Liska <mliska@suse.cz>
3164 * internal-fn.def (ASAN_CHECK): Set proper flags.
3165 (ASAN_MARK): Likewise.
3167 2018-02-16 Julia Koval <julia.koval@intel.com>
3169 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3170 from PTA_CANNONLAKE.
3172 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3175 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3176 Use ++iter rather than iter++ for std::list iterators.
3177 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3178 defer deleting them until all nodes in the forest are processed. Do
3179 free even leaf nodes. Change to_process into auto_vec.
3182 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3183 * vec.h (vec_default_construct): Use memset instead of placement new
3184 if BROKEN_VALUE_INITIALIZATION is defined.
3185 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3186 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3189 PR rtl-optimization/83723
3190 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3191 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3192 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3194 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3196 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3198 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3200 PR rtl-optimization/81443
3201 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3202 from inner REGs to paradoxical SUBREGs.
3204 2018-02-16 Richard Biener <rguenther@suse.de>
3206 PR tree-optimization/84399
3207 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3208 For operands we can analyze at their definition make sure we can
3209 analyze them at each use as well.
3211 2018-02-16 Richard Biener <rguenther@suse.de>
3213 PR tree-optimization/84190
3214 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3215 volatile accesses if the decl isn't volatile.
3217 2018-02-15 Jason Merrill <jason@redhat.com>
3219 PR c++/84314 - ICE with templates and fastcall attribute.
3220 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3221 TYPE_CANONICAL on an existing type.
3223 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3225 PR tree-optimization/84383
3226 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3227 dstoff nor call operand_equal_p if dstbase is NULL.
3229 PR tree-optimization/84334
3230 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3231 also a CONSTANT_CLASS_P, punt.
3233 2018-02-14 Jim Wilson <jimw@sifive.com>
3235 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3236 first SMALL_OPERAND check. New local min_second_step. Move assert
3237 to where locals are set. Add TARGET_RVC support.
3238 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3240 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3242 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3244 2018-02-14 Martin Sebor <msebor@redhat.com>
3246 PR tree-optimization/83698
3247 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3248 arrays constrain the offset range to their bounds.
3249 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3250 (builtin_access::overlap): Avoid setting the size of overlap if it's
3252 (maybe_diag_overlap): Also consider arrays when deciding what values
3253 of offsets to include in diagnostics.
3255 2018-02-14 Martin Sebor <msebor@redhat.com>
3258 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3259 that correspond to the kind of a declaration.
3261 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3264 * config/pa/pa.md: Load address of PIC label using the linkage table
3265 if the label is nonlocal.
3267 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3269 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3270 warning message if user requests -maltivec=be.
3271 * doc/invoke.texi: Document deprecation of -maltivec=be.
3273 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3276 * config/rs6000/rs6000-c.c: Update definitions for
3277 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3278 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3280 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3283 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3284 add _get_ssp intrinsics. Remove argument from
3285 __builtin_ia32_rdssp[d|q].
3286 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3287 * config/i386/i386-builtin.def: Remove argument from
3288 __builtin_ia32_rdssp[d|q].
3289 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3290 ix86_expand_special_args_builtin for _rdssp[d|q].
3291 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3292 Clear register before usage.
3293 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3294 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3296 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3298 PR tree-optimization/84357
3299 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3300 operand 1 of an ARRAY_REF too.
3302 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3305 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3306 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3308 (set_of_reg): New struct.
3309 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3310 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3311 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3313 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3314 Split into bitclr, bitset, bitinvert patterns if appropriate.
3315 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3316 use rx_fuse_in_memory_bitop.
3317 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3318 to named insn, correct maximum insn length.
3320 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3323 * machmode.def: Define a complex mode for PARTIAL_INT.
3324 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3326 * doc/rtl.texi: Document CSPImode.
3327 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3329 (msp430_hard_regno_nregs_with_padding): Likewise.
3331 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3334 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3336 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3338 PR rtl-optimization/84169
3339 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3340 we generated a parallel as new i3 and we split that to new i2 and i3
3341 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3342 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3343 those to i2, not i1. Partially rewrite this scan code.
3345 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3348 * stor-layout.c (place_field): For variable length fields, adjust
3349 offset_align afterwards not just based on the field's alignment,
3350 but also on the size.
3353 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3354 of exps and logs in the use_exp2 case.
3356 2018-02-13 Jeff Law <law@redhat.com>
3358 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3361 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3364 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3368 * common.opt (gas-loc-support, gas-locview-support): New.
3369 (ginline-points, ginternal-reset-location-views): New.
3370 * doc/invoke.texi: Document them. Use @itemx where intended.
3371 (gvariable-location-views): Adjust.
3372 * target.def (reset_location_view): New.
3373 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3374 (TARGET_RESET_LOCATION_VIEW): New.
3375 * doc/tm.texi: Rebuilt.
3376 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3377 (dwarf2out_default_as_locview_support): New.
3378 (output_asm_line_debug_info): Use option variables.
3379 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3380 (output_loc_list): Likewise.
3381 (add_high_low_attributes): Check option variables.
3382 Don't output entry view attribute in strict mode.
3383 (gen_inlined_subroutine_die): Check option variables.
3384 (dwarf2out_inline_entry): Likewise.
3385 (init_sections_and_labels): Likewise.
3386 (dwarf2out_early_finish): Likewise.
3387 (maybe_reset_location_view): New, from...
3388 (dwarf2out_var_location): ... here. Call it.
3389 * debug.h (dwarf2out_default_as_loc_support): Declare.
3390 (dwarf2out_default_as_locview_support): Declare.
3391 * hooks.c (hook_int_rtx_insn_0): New.
3392 * hooks.h (hook_int_rtx_insn_0): Declare.
3393 * toplev.c (process_options): Take -gas-loc-support and
3394 -gas-locview-support from dwarf2out. Enable
3395 -gvariable-location-views by default only with locview
3396 assembler support. Enable -ginternal-reset-location-views by
3397 default only if the target defines the corresponding hook.
3398 Enable -ginline-points by default if location views are
3399 enabled; force it disabled if statement frontiers are
3401 * tree-inline.c (expand_call_inline): Check option variables.
3402 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3404 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3406 PR tree-optimization/84321
3407 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3408 handling. Also check whether the anti-range contains any values
3409 that satisfy the mask; switch to a VR_RANGE if not.
3411 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3414 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3416 2018-02-13 Martin Jambor <mjambor@suse.cz>
3419 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3420 of call statements, also set location of a load to a temporary.
3422 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3424 * config/rl78/rl78.c (add_vector_labels): New function.
3425 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3426 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3427 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3428 which checks that no arguments are passed.
3429 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3430 * doc/extend.texi: Documentation for the new attribute.
3432 2018-02-13 Andreas Schwab <schwab@suse.de>
3434 * config/riscv/linux.h (CPP_SPEC): Define.
3436 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3439 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3440 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3441 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3442 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3443 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3444 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3445 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3446 temporarily for AES and PCLMUL builtins.
3448 PR tree-optimization/84339
3449 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3450 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3454 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3455 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3456 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3458 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3460 * omp-simd-clone.h: New file.
3461 * omp-simd-clone.c: Include omp-simd-clone.h.
3462 (expand_simd_clones): No longer static.
3463 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3464 cgraph.h and omp-simd-clone.h.
3465 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3466 (vect_recog_widen_shift_pattern): Formatting fix.
3467 (vect_pattern_recog_1): Don't check optab for calls.
3470 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3471 operands[2] into a REG before using gen_lowpart on it.
3473 2018-02-12 Jeff Law <law@redhat.com>
3476 * config/sh/sh.c (find_barrier): Consider a sibling call
3479 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3480 successfully back substituting a reg.
3482 2018-02-12 Richard Biener <rguenther@suse.de>
3484 PR tree-optimization/84037
3485 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3486 parameter, move visited init to caller.
3487 (vect_slp_analyze_operations): Separate cost from validity
3488 check, initialize visited once for all instances.
3489 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3491 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3493 (vect_model_promotion_demotion_cost): Likewise.
3494 (vectorizable_bswap): Guard cost modeling with !slp_node
3495 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3497 (vectorizable_call): Likewise.
3498 (vectorizable_conversion): Likewise.
3499 (vectorizable_assignment): Likewise.
3500 (vectorizable_shift): Likewise.
3501 (vectorizable_operation): Likewise.
3502 (vectorizable_store): Likewise.
3503 (vectorizable_load): Likewise.
3504 (vectorizable_condition): Likewise.
3505 (vectorizable_comparison): Likewise.
3507 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3510 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3511 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3512 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3514 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3517 * params.def (inline-min-speedup): Increase from 8 to 15.
3518 (max-inline-insns-auto): Decrease from 40 to 30.
3519 * ipa-split.c (consider_split): Add some buffer for function to
3520 be considered inlining candidate.
3521 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3524 2018-02-12 Richard Biener <rguenther@suse.de>
3526 PR tree-optimization/84037
3527 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3528 matched stmts if we cannot swap the non-matched ones.
3530 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3532 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3533 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3534 _mm_maskz_scalef_round_ss): New intrinsics.
3535 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3536 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3537 __builtin_ia32_scalefss_round): Remove.
3538 (__builtin_ia32_scalefsd_mask_round,
3539 __builtin_ia32_scalefss_mask_round): New intrinsics.
3540 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3541 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3542 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3543 "<round_constraint>")): Changed to ...
3544 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3545 "<round_scalar_constraint>")): ... this.
3546 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3547 %0, %1, %2<round_op3>}"): Changed to ...
3548 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3549 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3550 %2<round_scalar_mask_op3>}"): ... this.
3551 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3553 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3555 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3556 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3557 (_mm_maskz_sqrt_round_ss): New intrinsics.
3558 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3559 (__builtin_ia32_sqrtsd_mask_round)
3560 (__builtin_ia32_sqrtss_mask_round): New builtins.
3561 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3562 (__builtin_ia32_sqrtss_round): Remove.
3563 (__builtin_ia32_sqrtsd_mask_round)
3564 (__builtin_ia32_sqrtss_mask_round): New builtins.
3565 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3566 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3567 ((match_operand:VF_128 1 "vector_operand"
3568 "xBm,<round_constraint>")): Changed to ...
3569 ((match_operand:VF_128 1 "vector_operand"
3570 "xBm,<round_scalar_constraint>")): ... this.
3571 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3572 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3573 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3574 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3575 %<iptr>1<round_scalar_mask_op3>}): ... this.
3576 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3577 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3579 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3582 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3583 Cast vec_cmpeq result to correct type.
3584 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3585 Cast vec_cmpgt result to correct type.
3587 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3589 * final.c (final_scan_insn_1): Renamed from...
3590 (final_scan_insn): ... this. New wrapper, to recover
3591 seen from the outermost call in recursive ones.
3592 * config/sparc/sparc.c (output_return): Drop seen from call.
3593 (output_sibcall): Likewise.
3594 * config/visium/visium.c (output_branch): Likewise.
3596 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3598 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3601 2018-02-10 Alan Modra <amodra@gmail.com>
3604 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3605 Specify LR as an input.
3607 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3610 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3611 remove_member_access_dummy_vars): New functions.
3612 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3613 lower_omp_1, execute_lower_omp): Use them.
3615 PR rtl-optimization/84308
3616 * shrink-wrap.c (spread_components): Release todo vector.
3618 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3620 PR rtl-optimization/57193
3621 * ira-color.c (struct allocno_color_data): Add member
3622 conflict_allocno_hard_prefs.
3623 (update_conflict_allocno_hard_prefs): New.
3624 (bucket_allocno_compare_func): Add a preference based on
3625 conflict_allocno_hard_prefs.
3626 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3627 (color_allocnos): Remove a dead code. Initiate
3628 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3630 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3633 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3634 constraint from =wa to wa. Avoid a subreg on the output operand,
3635 instead use a pseudo and subreg it in a move.
3636 (p9_xxbrd_<mode>): Changed to ...
3637 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3638 (p9_xxbrd_v2df): New expander.
3639 (p9_xxbrw_<mode>): Changed to ...
3640 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3641 (p9_xxbrw_v4sf): New expander.
3643 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3645 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3647 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3650 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3651 multiply in 32-bit mode.
3652 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3653 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3656 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3658 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
3659 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
3660 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
3661 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
3663 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
3666 * dwarf2out.c (is_trivial_indirect_ref): New function.
3667 (dwarf2out_late_global_decl): Do not generate a location
3668 attribute for variables that have a non-trivial DECL_VALUE_EXPR
3669 and that are not defined in the current unit.
3671 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3673 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
3674 instead of a libcall for UNORDERED.
3676 2018-02-09 Tamar Christina <tamar.christina@arm.com>
3679 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
3680 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
3682 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3685 * config/s390/s390.c (s390_set_current_function): Invoke
3686 s390_indirect_branch_settings also if fndecl didn't change.
3688 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3690 * config/rs6000/rs6000.md (blockage): Set length to zero.
3692 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
3694 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
3696 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3699 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
3700 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
3704 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
3705 PARALLEL incoming that failed vt_get_decl_and_offset check.
3708 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
3709 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
3711 (get_variable_section): For decls in named .bss* sections pass true as
3712 second argument to bss_initializer_p.
3714 2018-02-09 Marek Polacek <polacek@redhat.com>
3715 Jakub Jelinek <jakub@redhat.com>
3718 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
3719 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
3720 Sync some changes from cxx_fold_indirect_ref.
3722 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
3724 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
3726 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
3727 (BLOCK_INLINE_ENTRY_LABEL): New.
3728 (dwarf2out_var_location): Disregard inline entry markers.
3729 (inline_entry_data): New struct.
3730 (inline_entry_data_hasher): New hashtable type.
3731 (inline_entry_data_hasher::hash): New.
3732 (inline_entry_data_hasher::equal): New.
3733 (inline_entry_data_table): New variable.
3734 (add_high_low_attributes): Add DW_AT_entry_pc and
3735 DW_AT_GNU_entry_view attributes if a pending entry is found
3736 in inline_entry_data_table. Add old entry_pc attribute only
3737 if debug nonbinding markers are disabled.
3738 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
3739 markers are enabled.
3740 (block_within_block_p, dwarf2out_inline_entry): New.
3741 (dwarf2out_finish): Check that no entries remained in
3742 inline_entry_data_table.
3743 * final.c (reemit_insn_block_notes): Handle inline entry notes.
3744 (final_scan_insn, notice_source_line): Likewise.
3745 (rest_of_clean_state): Skip inline entry markers.
3746 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
3748 * gimple.c (gimple_build_debug_inline_entry): New.
3749 * gimple.h (enum gimple_debug_subcode): Add
3750 GIMPLE_DEBUG_INLINE_ENTRY.
3751 (gimple_build_debug_inline_entry): Declare.
3752 (gimple_debug_inline_entry_p): New.
3753 (gimple_debug_nonbind_marker_p): Adjust.
3754 * insn-notes.def (INLINE_ENTRY): New.
3755 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3756 inline entry marker notes.
3757 (print_insn): Likewise.
3758 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
3759 (INSN_DEBUG_MARKER_KIND): Likewise.
3760 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
3761 * tree-inline.c (expand_call_inline): Build and insert
3762 debug_inline_entry stmt.
3763 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
3764 inline entry blocks early, if nonbind markers are enabled.
3765 (dump_scope_block): Dump fragment info.
3766 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
3767 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
3768 (gimple_build_debug_inline_entry): New.
3769 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
3770 Enable/disable inline entry points too.
3771 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
3772 (DEBUG_INSN): Describe inline entry markers.
3774 * common.opt (gvariable-location-views): New.
3775 (gvariable-location-views=incompat5): New.
3776 * config.in: Rebuilt.
3777 * configure: Rebuilt.
3778 * configure.ac: Test assembler for view support.
3779 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
3780 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
3781 * dwarf2out.c (var_loc_view): New typedef.
3782 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
3783 (dwarf2out_locviews_in_attribute): New.
3784 (dwarf2out_locviews_in_loclist): New.
3785 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
3786 (enum dw_line_info_opcode): Add LI_adv_address.
3787 (struct dw_line_info_table): Add view.
3788 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
3789 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
3790 (zero_view_p): New variable.
3791 (ZERO_VIEW_P): New macro.
3792 (output_asm_line_debug_info): New.
3793 (struct var_loc_node): Add view.
3794 (add_AT_view_list, AT_loc_list): New.
3795 (add_var_loc_to_decl): Add view param. Test it against last.
3796 (new_loc_list): Add view params. Record them.
3797 (AT_loc_list_ptr): Handle loc and view lists.
3798 (view_list_to_loc_list_val_node): New.
3799 (print_dw_val): Handle dw_val_class_view_list.
3800 (size_of_die): Likewise.
3801 (value_format): Likewise.
3802 (loc_list_has_views): New.
3803 (gen_llsym): Set vl_symbol too.
3804 (maybe_gen_llsym, skip_loc_list_entry): New.
3805 (dwarf2out_maybe_output_loclist_view_pair): New.
3806 (output_loc_list): Output view list or entries too.
3807 (output_view_list_offset): New.
3808 (output_die): Handle dw_val_class_view_list.
3809 (output_dwarf_version): New.
3810 (output_compilation_unit_header): Use it.
3811 (output_skeleton_debug_sections): Likewise.
3812 (output_rnglists, output_line_info): Likewise.
3813 (output_pubnames, output_aranges): Update version comments.
3814 (output_one_line_info_table): Output view numbers in asm comments.
3815 (dw_loc_list): Determine current endview, pass it to new_loc_list.
3816 Call maybe_gen_llsym.
3817 (loc_list_from_tree_1): Adjust.
3818 (add_AT_location_description): Create view list attribute if
3819 needed, check it's absent otherwise.
3820 (convert_cfa_to_fb_loc_list): Adjust.
3821 (maybe_emit_file): Call output_asm_line_debug_info for test.
3822 (dwarf2out_var_location): Reset views as needed. Precompute
3823 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
3824 attribute. Set view.
3825 (new_line_info_table): Reset next view.
3826 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
3827 (dwarf2out_source_line): Likewise. Output view resets and labels to
3828 the assembler, or select appropriate line info opcodes.
3829 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
3830 (optimize_string_length): Catch it. Adjust.
3831 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
3832 dw_val_class_view_list, and remove it if no longer needed.
3833 (hash_loc_list): Hash view numbers.
3834 (loc_list_hasher::equal): Compare them.
3835 (optimize_location_lists): Check whether a view list symbol is
3836 needed, and whether the locview attribute is present, and
3837 whether they match. Remove the locview attribute if no longer
3839 (index_location_lists): Call skip_loc_list_entry for test.
3840 (dwarf2out_finish): Call output_asm_line_debug_info for test.
3841 Use output_dwarf_version.
3842 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
3843 (struct dw_val_node): Add val_view_list.
3844 * final.c (SEEN_NEXT_VIEW): New.
3845 (set_next_view_needed): New.
3846 (clear_next_view_needed): New.
3847 (maybe_output_next_view): New.
3848 (final_start_function): Rename to...
3849 (final_start_function_1): ... this. Take pointer to FIRST,
3850 add SEEN parameter. Emit param bindings in the initial view.
3851 (final_start_function): Reintroduce SEEN-less interface.
3852 (final): Rename to...
3853 (final_1): ... this. Take SEEN parameter. Output final pending
3854 next view at the end.
3855 (final): Reintroduce seen-less interface.
3856 (final_scan_insn): Output pending next view before switching
3857 sections or ending a block. Mark the next view as needed when
3858 outputting variable locations. Notify debug backend of section
3859 changes, and of location view changes.
3860 (rest_of_handle_final): Adjust.
3861 * toplev.c (process_options): Autodetect value for debug variable
3862 location views option. Warn on incompat5 without -gdwarf-5.
3863 * doc/invoke.texi (gvariable-location-views): New.
3864 (gvariable-location-views=incompat5): New.
3865 (gno-variable-location-views): New.
3867 2018-02-08 David Malcolm <dmalcolm@redhat.com>
3869 PR tree-optimization/84136
3870 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
3871 that the result of find_edge is non-NULL.
3873 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3876 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
3877 storing integer register in SImode. Fix cost of 256 and 512
3878 byte aligned SSE register store.
3880 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
3882 * config/i386/i386.c (ix86_multiplication_cost): Fix
3883 multiplication cost for TARGET_AVX512DQ.
3885 2018-02-08 Marek Polacek <polacek@redhat.com>
3887 PR tree-optimization/84238
3888 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
3891 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3893 PR tree-optimization/84265
3894 * tree-vect-stmts.c (vectorizable_store): Don't treat
3895 VMAT_CONTIGUOUS accesses as grouped.
3896 (vectorizable_load): Likewise.
3898 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3900 PR tree-optimization/81635
3901 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
3902 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
3903 (test_round_for_mask): New functions.
3904 (wide_int_cc_tests): Call test_round_for_mask.
3905 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
3906 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
3907 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
3908 range returned by get_range_info.
3910 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3913 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
3914 * symtab.c: Include builtins.h
3915 (symtab_node::output_to_lto_symbol_table_p): Move here
3916 from lto-streamer-out.c:output_symbol_p.
3917 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
3918 (output_symbol_p): Move all logic to symtab.c
3919 (produce_symtab): Update.
3921 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3923 * config/s390/s390-opts.h (enum indirect_branch): Define.
3924 * config/s390/s390-protos.h (s390_return_addr_from_memory)
3925 (s390_indirect_branch_via_thunk)
3926 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
3927 (enum s390_indirect_branch_type): Define.
3928 * config/s390/s390.c (struct s390_frame_layout, struct
3929 machine_function): Remove.
3930 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
3931 (indirect_branch_table_label_no, indirect_branch_table_name):
3933 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
3934 (enum s390_indirect_branch_option): Define.
3935 (s390_return_addr_from_memory): New function.
3936 (s390_handle_string_attribute): New function.
3937 (s390_attribute_table): Add new attribute handler.
3938 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
3939 (s390_indirect_branch_via_thunk): New function.
3940 (s390_indirect_branch_via_inline_thunk): New function.
3941 (s390_function_ok_for_sibcall): When jumping via thunk disallow
3942 sibling call optimization for non z10 compiles.
3943 (s390_emit_call): Force indirect branch target to be a single
3944 register. Add r1 clobber for non-z10 compiles.
3945 (s390_emit_epilogue): Emit return jump via return_use expander.
3946 (s390_reorg): Handle JUMP_INSNs as execute targets.
3947 (s390_option_override_internal): Perform validity checks for the
3948 new command line options.
3949 (s390_indirect_branch_attrvalue): New function.
3950 (s390_indirect_branch_settings): New function.
3951 (s390_set_current_function): Invoke s390_indirect_branch_settings.
3952 (s390_output_indirect_thunk_function): New function.
3953 (s390_code_end): Implement target hook.
3954 (s390_case_values_threshold): Implement target hook.
3955 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
3957 * config/s390/s390.h (struct s390_frame_layout)
3958 (struct machine_function): Move here from s390.c.
3959 (TARGET_INDIRECT_BRANCH_NOBP_RET)
3960 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
3961 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
3962 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
3963 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
3964 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
3965 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
3966 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
3967 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
3968 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
3969 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
3970 (mnemonic attribute): Add values which aren't recognized
3972 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
3973 pattern for branch conversion. Fix mnemonic attribute.
3974 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
3975 indirect branch via thunk if requested.
3976 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
3977 ("*indirect_jump"): Disable for branch conversion using out of
3979 ("indirect_jump_via_thunk<mode>_z10")
3980 ("indirect_jump_via_thunk<mode>")
3981 ("indirect_jump_via_inlinethunk<mode>_z10")
3982 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
3983 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
3984 ("casesi_jump_via_inlinethunk<mode>_z10")
3985 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
3986 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
3987 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
3988 ("*indirect2_jump"): Disable for branch conversion.
3989 ("casesi_jump"): Turn into expander and expand patterns for branch
3991 ("return_use"): New expander.
3992 ("*return"): Emit return via thunk and rename it to ...
3993 ("*return<mode>"): ... this one.
3994 * config/s390/s390.opt: Add new options and and enum for the
3997 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
3999 * lra-constraints.c (match_reload): Unconditionally use
4000 gen_lowpart_SUBREG, rather than selecting between that
4001 and equivalent gen_rtx_SUBREG code.
4003 2018-02-08 Richard Biener <rguenther@suse.de>
4005 PR tree-optimization/84233
4006 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4007 changed flag instead of boguously re-using phi_inserted.
4009 2018-02-08 Martin Jambor <mjambor@suse.cz>
4011 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4012 static local variables.
4014 2018-02-08 Richard Biener <rguenther@suse.de>
4016 PR tree-optimization/84278
4017 * tree-vect-stmts.c (vectorizable_store): When looking for
4018 smaller vector types to perform grouped strided loads/stores
4019 make sure the mode is supported by the target.
4020 (vectorizable_load): Likewise.
4022 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4024 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4025 Increase LDP/STP opportunities by adding adjacent callee-saves.
4027 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4029 PR rtl-optimization/84068
4030 PR rtl-optimization/83459
4031 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4033 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4035 PR tree-optimization/84224
4036 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4037 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4040 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4043 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4044 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4045 count, remove one USE.
4047 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4049 * doc/install.texi (Configuration): Document the
4050 --with-long-double-format={ibm,ieee} PowerPC configuration
4054 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4055 Convert from define_expand to be define_insn_and_split. Rework
4056 float/double/_Float128 conversions to QI/HI/SImode to work with
4057 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4058 conversions to QI/HImode types did a store and then a load to
4059 truncate the value. For conversions to VSX registers, don't split
4060 the insn, instead emit the code directly. Use the code iterator
4061 any_fix to combine signed and unsigned conversions.
4062 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4063 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4064 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4065 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4066 (fix_<mode>di2_hw): Likewise.
4067 (fixuns_<mode>di2_hw): Likewise.
4068 (fix_<mode>si2_hw): Likewise.
4069 (fixuns_<mode>si2_hw): Likewise.
4070 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4071 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4072 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4073 fix<uns>_trunc<SFDF:mode>si2_p8.
4074 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4076 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4077 (fix<uns>_<mode>_mem): Likewise.
4078 (fctiw<u>z_<mode>_mem): Likewise.
4079 (fix<uns>_<mode>_mem): Likewise.
4080 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4081 the register allocator from doing a direct move to the GPRs to do
4082 a store, and instead use the ISA 3.0 store byte/half-word from
4083 vector register instruction. For IEEE 128-bit floating point,
4084 also optimize stores of 32-bit ints.
4085 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4087 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4089 * genextract.c (push_pathstr_operand): New function to support
4091 (walk_rtx): Call push_pathstr_operand.
4092 (print_path): Support [a-zA-Z].
4094 2018-02-07 Richard Biener <rguenther@suse.de>
4096 PR tree-optimization/84037
4097 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4098 (cse_and_gimplify_to_preheader): Declare.
4099 (vect_get_place_in_interleaving_chain): Likewise.
4100 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4102 (_loop_vec_info::~_loop_vec_info): Delete it.
4103 (cse_and_gimplify_to_preheader): New function.
4104 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4105 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4106 (vectorizable_load): Likewise. For grouped stores always base
4107 the IV on the first element.
4108 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4109 condition before gimplifying.
4111 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4113 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4114 *DIV_EXPR and *MOD_EXPR.
4116 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4119 * config/i386/i386.c (ix86_option_override_internal): Mask out
4120 the CF_SET bit when checking -fcf-protection.
4122 2018-02-07 Tom de Vries <tom@codesourcery.com>
4125 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4128 2018-02-07 Richard Biener <rguenther@suse.de>
4130 PR tree-optimization/84204
4131 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4134 PR tree-optimization/84205
4135 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4136 special-case isl_ast_op_zdiv_r.
4138 PR tree-optimization/84223
4139 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4140 Only add conditions from within the region.
4141 (gather_bbs::after_dom_children): Adjust.
4143 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4146 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4147 * config/avr/avr.md: Only post-reload split REG-REG moves if
4148 either register is GENERAL_REG_P.
4150 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4152 PR tree-optimization/84235
4153 * tree-ssa-scopedtables.c
4154 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4155 if the subtraction is performed in floating point type where NaNs are
4156 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4157 build 1. Formatting fix.
4159 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4162 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4163 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4164 and skip it regardless of bb boundaries. Use CALL_P macro,
4165 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4166 unnecessarily, formatting fix.
4168 2018-02-06 Michael Collison <michael.collison@arm.com>
4170 * config/arm/thumb2.md:
4171 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4172 (*thumb_mov_notscc): Ditto.
4174 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4177 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4178 unsigned_fix, not "s".
4180 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4182 * configure.ac (gcc_fn_eh_frame_ro): New function.
4183 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4184 correct .eh_frame permissions.
4185 * configure: Regenerate.
4187 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4189 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4192 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4194 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4195 Display warning message for -mno-speculate-indirect-jumps.
4197 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4199 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4201 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4203 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4205 PR tree-optimization/84225
4206 * tree-eh.c (find_trapping_overflow): Only call
4207 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4209 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4212 * config/i386/i386.c: Reimplement the check of possible options
4213 -mibt/-mshstk conbination. Change error messages.
4214 * doc/invoke.texi: Fix a typo: remove extra '='.
4216 2018-02-06 Marek Polacek <polacek@redhat.com>
4218 PR tree-optimization/84228
4219 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4221 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4224 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4225 emitted arch directives.
4226 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4227 __ARM_FEATURE_COPROC before changing architectures.
4229 2018-02-06 Richard Biener <rguenther@suse.de>
4231 * config/i386/i386.c (print_reg): Fix typo.
4232 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4234 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4236 * configure: Regenerate.
4238 2018-02-05 Martin Sebor <msebor@redhat.com>
4240 PR tree-optimization/83369
4241 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4244 2018-02-05 Martin Liska <mliska@suse.cz>
4246 * doc/invoke.texi: Cherry-pick upstream r323995.
4248 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4250 * ira.c (ira_init_register_move_cost): Adjust comment.
4252 2018-02-05 Martin Liska <mliska@suse.cz>
4254 PR gcov-profile/84137
4255 * doc/gcov.texi: Fix typo in documentation.
4257 2018-02-05 Martin Liska <mliska@suse.cz>
4259 PR gcov-profile/83879
4260 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4261 using dlopen functionality.
4263 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4265 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4266 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4267 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4268 _mm_maskz_range_round_ss): New intrinsics.
4269 (__builtin_ia32_rangesd128_round)
4270 (__builtin_ia32_rangess128_round): Remove.
4271 (__builtin_ia32_rangesd128_mask_round,
4272 __builtin_ia32_rangess128_mask_round): New builtins.
4273 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4274 __builtin_ia32_rangess128_round): Remove.
4275 (__builtin_ia32_rangesd128_mask_round,
4276 __builtin_ia32_rangess128_mask_round): New builtins.
4277 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4278 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4279 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4280 "<round_saeonly_constraint>")): Changed to ...
4281 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4282 "<round_saeonly_scalar_constraint>")): ... this.
4283 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4284 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4285 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4286 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4287 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4289 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4291 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4293 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4294 Remove all values except native, 8540 and 8548.
4296 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4298 * config/i386/i386.c (ix86_output_function_return): Pass
4299 INVALID_REGNUM, instead of -1, as invalid register number to
4300 indirect_thunk_name and output_indirect_thunk.
4302 2018-02-02 Julia Koval <julia.koval@intel.com>
4304 * config.gcc: Add -march=icelake.
4305 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4306 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4307 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4308 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4309 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4310 (processor_target_table): Add icelake.
4311 (ix86_option_override_internal): Handle new PTAs.
4312 (get_builtin_code_for_version): Handle icelake.
4313 (M_INTEL_COREI7_ICELAKE): New.
4314 (fold_builtin_cpu): Handle icelake.
4315 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4316 * doc/invoke.texi: Add -march=icelake.
4318 2018-02-02 Julia Koval <julia.koval@intel.com>
4320 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4321 to wide_int_bitmask.
4322 * wide-int-bitmask.h: New.
4324 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4327 * config/i386/i386.md: Replace Pmode with word_mode in
4328 builtin_setjmp_setup and builtin_longjmp to support x32.
4330 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4334 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4336 (rs6000_supported_cpu_names): New static variable.
4337 (linux_cpu_translation_table): Likewise.
4338 (elf_platform) <cpu>: Define new static variable and use it.
4339 Translate kernel AT_PLATFORM name to canonical name if needed.
4340 Error if platform name is unknown.
4342 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4345 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4347 2018-02-01 Jeff Law <law@redhat.com>
4350 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4351 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4352 the scratch if RELEASE_VIA_POP is false.
4353 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4354 If we have to save a temporary register, decrement SIZE appropriately.
4355 Pass new arguments to release_scratch_register_on_entry.
4356 (ix86_adjust_stack_and_probe): Likewise.
4357 (ix86_emit_probe_stack_range): Pass new arguments to
4358 release_scratch_register_on_entry.
4360 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4362 PR rtl-optimization/84157
4363 * combine.c (change_zero_ext): Use REG_P predicate in
4364 front of HARD_REGISTER_P predicate.
4366 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4368 * config/avr/avr.c (avr_option_override): Move disabling of
4369 -fdelete-null-pointer-checks to...
4370 * common/config/avr/avr-common.c (avr_option_optimization_table):
4373 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4375 PR tree-optimization/81635
4376 * tree-data-ref.c (split_constant_offset_1): For types that
4377 wrap on overflow, try to use range info to prove that wrapping
4380 2018-02-01 Renlin Li <renlin.li@arm.com>
4383 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4385 (aarch64_register_move_cost): Likewise.
4386 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4388 (REG_CLASS_NAMES): Likewise.
4389 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4390 TAILCALL_ADDR_REGS. Remove IP registers.
4391 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4393 2018-02-01 Richard Biener <rguenther@suse.de>
4395 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4396 for specifying RPO order and allow NULL for that.
4397 * domwalk.c (dom_walker::dom_walker): Likewise.
4398 (dom_walker::walk): Handle NULL RPO order.
4399 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4401 (rewrite_update_dom_walker): Likewise.
4402 (mark_def_dom_walker): Likewise.
4404 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4406 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4407 (aarch64_maybe_expand_sve_subreg_move): Declare.
4408 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4409 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4411 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4412 that are semantically a reverse operation.
4413 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4414 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4415 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4417 (aarch64_can_change_mode_class): For big-endian, forbid changes
4418 between two SVE modes if they have different element sizes.
4420 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4422 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4423 the TImode handling for big-endian targets.
4425 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4427 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4428 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4430 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4431 Remove BSWAP handing for big-endian targets and use the form of
4432 LD1RQ appropariate for the mode.
4434 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4436 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4437 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4440 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4443 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4444 check for operands that need to go through aarch64_sve_reload_be.
4446 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4448 PR tree-optimization/81661
4449 PR tree-optimization/84117
4450 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4451 * tree-eh.c: Include gimplify.h.
4452 (find_trapping_overflow, replace_trapping_overflow,
4453 rewrite_to_non_trapping_overflow): New functions.
4454 * tree-vect-loop.c: Include tree-eh.h.
4455 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4456 * tree-data-ref.c: Include tree-eh.h.
4457 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4459 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4461 PR rtl-optimization/84123
4462 * combine.c (change_zero_ext): Check if hard register satisfies
4463 can_change_dest_mode before calling gen_lowpart_SUBREG.
4465 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4468 * ira.c (ira_init_register_move_cost): Remove assert.
4470 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4472 PR rtl-optimization/84071
4473 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4474 * doc/tm.texi: Regenerate.
4476 2018-01-31 Richard Biener <rguenther@suse.de>
4478 PR tree-optimization/84132
4479 * tree-data-ref.c (analyze_miv_subscript): Properly
4480 check whether evolution_function_is_affine_multivariate_p
4481 before calling gcd_of_steps_may_divide_p.
4483 2018-01-31 Julia Koval <julia.koval@intel.com>
4486 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4487 * config/i386/i386.md (rdpid_rex64) New.
4488 (rdpid): Make 32bit only.
4490 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4493 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4494 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4496 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4499 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4501 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4503 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4505 PR rtl-optimization/84071
4506 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4507 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4509 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4511 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4512 (arc_attribute_table): Add 'aux' attribute.
4513 (arc_in_small_data_p): Consider aux like variables.
4514 (arc_is_aux_reg_p): New function.
4515 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4516 (arc_get_aux_arg): New function.
4517 (prepare_move_operands): Handle aux-register access.
4518 (arc_handle_aux_attribute): New function.
4519 * doc/extend.texi (ARC Variable attributes): Add subsection.
4521 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4523 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4524 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4525 (arc_attribute_table): Add 'uncached' attribute.
4526 (arc_print_operand): Print '.di' flag for uncached memory
4528 (arc_in_small_data_p): Do not consider for small data the uncached
4530 (arc_is_uncached_mem_p): New function.
4531 * config/arc/predicates.md (compact_store_memory_operand): Check
4532 for uncached memory accesses.
4533 (nonvol_nonimm_operand): Likewise.
4534 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4536 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4539 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4540 falign-loops=): Add Optimization flag.
4542 2018-01-30 Jeff Law <law@redhat.com>
4545 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4546 INT_REGISTERS_SAVED. Check it prior to calling
4547 get_scratch_register_on_entry.
4548 (ix86_adjust_stack_and_probe): Similarly.
4549 (ix86_emit_probe_stack_range): Similarly.
4550 (ix86_expand_prologue): Corresponding changes.
4552 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4555 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4556 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4558 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4561 * lra-constraints.c (curr_insn_transform): Process AND in the
4564 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4566 PR rtl-optimization/83986
4567 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4568 dependence against last_pending_memory_flush in addition to
4571 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4573 PR tree-optimization/81611
4574 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4577 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4580 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4583 2018-01-30 Richard Biener <rguenther@suse.de>
4584 Jakub Jelinek <jakub@redhat.com>
4586 PR tree-optimization/84111
4587 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4588 inner loops added during recursion, as they don't have up-to-date
4591 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4594 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4595 (can_inline_edge_by_limits_p): ... here.
4596 (can_early_inline_edge_p, check_callers,
4597 update_caller_keys, update_callee_keys, recursive_inlining,
4598 add_new_edges_to_heap, speculation_useful_p,
4599 inline_small_functions,
4600 inline_small_functions, flatten_function,
4601 inline_to_all_callers_1): Update.
4603 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4605 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4608 2018-01-30 Richard Biener <rguenther@suse.de>
4610 PR tree-optimization/83008
4611 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4612 invariant and constant vector uses in stmts when they need
4615 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4618 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4619 * configure: Regenerate.
4621 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4623 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4625 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4626 Use gen_rtx_REG rather than gen_lowpart.
4628 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4630 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4631 rather than 0 when creating partial subregs.
4633 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4635 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4638 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4641 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4642 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4643 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4644 flags. This restores the settings used before the 2017-07-24.
4645 Turning off pre increment/decrement/modify allows IVOPTS to
4646 optimize DF/SF loops where the index is an int.
4648 2018-01-29 Richard Biener <rguenther@suse.de>
4649 Kelvin Nilsen <kelvin@gcc.gnu.org>
4652 * tree-vect-stmts.c (vectorizable_call): Don't call
4653 targetm.vectorize_builtin_md_vectorized_function if callee is
4656 2018-01-22 Carl Love <cel@us.ibm.com>
4658 * doc/extend.tex: Fix typo in second arg in
4659 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
4661 2018-01-29 Richard Biener <rguenther@suse.de>
4663 PR tree-optimization/84086
4664 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
4665 (flush_ssaname_freelist): When SSA names were released reset
4666 the SCEV hash table.
4668 2018-01-29 Richard Biener <rguenther@suse.de>
4670 PR tree-optimization/84057
4671 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
4672 removed paths when removing edges.
4674 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
4676 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
4677 -mfunction-return=@var{choice}.
4679 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
4682 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
4684 (layout::print_source_line): Likewise.
4685 (test_get_line_width_without_trailing_whitespace): Add test cases.
4687 2018-01-27 Jakub Jelinek <jakub@redhat.com>
4690 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
4693 2018-01-26 Jim Wilson <jimw@sifive.com>
4695 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
4697 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
4700 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4702 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
4703 and CMP + SUB-immediate -> SUBS.
4705 2018-01-26 Martin Sebor <msebor@redhat.com>
4707 PR tree-optimization/83896
4708 * tree-ssa-strlen.c (get_string_len): Rename...
4709 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
4710 Avoid assuming length is constant.
4711 (handle_char_store): Use HOST_WIDE_INT for string length.
4713 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
4716 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
4717 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
4719 2018-01-26 Richard Biener <rguenther@suse.de>
4721 PR rtl-optimization/84003
4722 * dse.c (record_store): Only record redundant stores when
4723 the earlier store aliases at least all accesses the later one does.
4725 2018-01-26 Jakub Jelinek <jakub@redhat.com>
4727 PR rtl-optimization/83985
4728 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
4729 REG_CFA_RESTORE insns.
4730 (delete_unmarked_insns): Don't ignore separate shrink wrapping
4731 REG_CFA_RESTORE insns here.
4734 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
4735 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
4737 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4739 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
4740 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
4741 (arc_init): Likewise.
4742 (arc_override_options): Likewise.
4743 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
4745 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
4747 * config/arc/arc.h (TARGET_DBNZ): Define.
4748 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
4749 properly set the tune attribute.
4750 (dbnz): Use TARGET_DBNZ guard.
4751 * config/arc/arc.opt (mtune): Add core3 option.
4753 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4755 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
4756 recognize new pic like addresses.
4757 (arc_delegitimize_address): Clean up.
4759 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4761 * config/arc/arc-arches.def: Option mrf16 valid for all
4763 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
4764 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
4765 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
4766 * config/arc/arc-tables.opt: Regenerate.
4767 * config/arc/arc.c (arc_conditional_register_usage): Handle
4768 reduced register file case.
4769 (arc_file_start): Set must have build attributes.
4770 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
4772 * config/arc/arc.opt (mrf16): Add new option.
4773 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
4774 * config/arc/genmultilib.awk: Handle new mrf16 option.
4775 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
4776 * config/arc/t-multilib: Regenerate.
4777 * doc/invoke.texi (ARC Options): Document mrf16 option.
4779 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4781 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
4782 * config/arc/arc.c (arc_handle_secure_attribute): New function.
4783 (arc_attribute_table): Add 'secure_call' attribute.
4784 (arc_print_operand): Print secure call operand.
4785 (arc_function_ok_for_sibcall): Don't optimize tail calls when
4787 (arc_is_secure_call_p): New function. * config/arc/arc.md
4788 (call_i): Add support for sjli instruction.
4789 (call_value_i): Likewise.
4790 * config/arc/constraints.md (Csc): New constraint.
4792 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
4793 John Eric Martin <John.Martin@emmicro-us.com>
4795 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
4796 * config/arc/arc.c (_arc_jli_section): New struct.
4797 (arc_jli_section): New type.
4798 (rc_jli_sections): New static variable.
4799 (arc_handle_jli_attribute): New function.
4800 (arc_attribute_table): Add jli_always and jli_fixed attribute.
4801 (arc_file_end): New function.
4802 (TARGET_ASM_FILE_END): Define.
4803 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
4804 (arc_add_jli_section): New function.
4805 (jli_call_scan): Likewise.
4806 (arc_reorg): Call jli_call_scan.
4807 (arc_output_addsi): Remove 'S' from printing asm operand.
4808 (arc_is_jli_call_p): New function.
4809 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
4811 (movhi_insn): Likewise.
4812 (movsi_insn): Likewise.
4813 (movsi_set_cc_insn): Likewise.
4814 (loadqi_update): Likewise.
4815 (load_zeroextendqisi_update): Likewise.
4816 (load_signextendqisi_update): Likewise.
4817 (loadhi_update): Likewise.
4818 (load_zeroextendhisi_update): Likewise.
4819 (load_signextendhisi_update): Likewise.
4820 (loadsi_update): Likewise.
4821 (loadsf_update): Likewise.
4822 (movsicc_insn): Likewise.
4823 (bset_insn): Likewise.
4824 (bxor_insn): Likewise.
4825 (bclr_insn): Likewise.
4826 (bmsk_insn): Likewise.
4827 (bicsi3_insn): Likewise.
4828 (cmpsi_cc_c_insn): Likewise.
4829 (movsi_ne): Likewise.
4830 (movsi_cond_exec): Likewise.
4831 (clrsbsi2): Likewise.
4842 (call_i): Remove 'S' asm letter, add jli instruction.
4843 (call_value_i): Likewise.
4844 * config/arc/arc.op (mjli-always): New option.
4845 * config/arc/constraints.md (Cji): New constraint.
4846 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
4848 (subsf3_fpx): Likewise.
4849 (mulsf3_fpx): Likewise.
4850 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
4852 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
4854 * doc/invoke.texi (ARC): Document mjli-always option.
4856 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
4858 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
4859 avoid addition with 0 and use incw and decw where possible.
4861 2018-01-26 Richard Biener <rguenther@suse.de>
4863 PR tree-optimization/81082
4864 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
4865 association if it requires casting to unsigned.
4866 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
4867 from fold_plusminus_mult_expr to catch important cases late when
4868 range info is available.
4870 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4872 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
4873 * configure.ac (hidden_linkonce): New test.
4874 * configure: Regenerate.
4875 * config.in: Regenerate.
4877 2018-01-26 Julia Koval <julia.koval@intel.com>
4879 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
4880 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
4881 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
4882 _mm_mask_bitshuffle_epi64_mask): Fix type.
4883 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
4884 USI_FTYPE_V4DI_V4DI_USI): Remove.
4885 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
4886 __builtin_ia32_vpshufbitqmb256_mask,
4887 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
4888 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
4889 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
4891 2018-01-26 Alan Modra <amodra@gmail.com>
4894 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
4895 UNSPEC_VBPERMQ. Sort other unspecs.
4897 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
4899 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
4901 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
4904 * predict.c (drop_profile): Do not push/pop cfun; update also
4906 (handle_missing_profiles): Fix logic looking for zero profiles.
4908 2018-01-25 Jakub Jelinek <jakub@redhat.com>
4911 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
4912 on functions with #pragma omp declare simd or functions with simd
4914 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
4915 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
4916 Remove trailing \n from warning_at calls.
4918 2018-01-25 Tom de Vries <tom@codesourcery.com>
4921 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4922 for neutered workers.
4924 2018-01-24 Joseph Myers <joseph@codesourcery.com>
4927 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
4928 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
4930 2018-01-24 Jeff Law <law@redhat.com>
4933 * i386.c (get_probe_interval): Move to earlier point.
4934 (ix86_compute_frame_layout): If -fstack-clash-protection and
4935 the frame is larger than the probe interval, then use pushes
4936 to save registers rather than reg->mem moves.
4937 (ix86_expand_prologue): Remove conditional for int_registers_saved
4940 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
4943 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
4944 min/max for never referenced object.
4946 2018-01-24 Jakub Jelinek <jakub@redhat.com>
4949 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
4951 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
4952 attributes from DECL_ATTRIBUTES (decl) without affecting
4953 DECL_ATTRIBUTES (current_function_decl).
4954 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
4955 functions with non-NULL DECL_ABSTRACT_ORIGIN.
4957 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
4959 PR tree-optimization/83979
4960 * fold-const.c (fold_comparison): Use constant_boolean_node
4961 instead of boolean_{true,false}_node.
4963 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
4965 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
4968 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4970 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
4971 Simplify the clause that sets the length attribute.
4972 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
4973 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
4974 clause that sets the length attribute.
4975 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
4977 2018-01-24 Tom de Vries <tom@codesourcery.com>
4980 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
4981 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
4982 Add strict parameter.
4983 (prevent_branch_around_nothing): Insert dummy insn between branch to
4984 label and label with no ptx insn inbetween.
4985 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
4987 2018-01-24 Tom de Vries <tom@codesourcery.com>
4990 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
4991 for neutered threads in warp.
4992 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
4994 2018-01-24 Richard Biener <rguenther@suse.de>
4996 PR tree-optimization/83176
4997 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5000 2018-01-24 Richard Biener <rguenther@suse.de>
5002 PR tree-optimization/82819
5003 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5004 code generating pluses that are no-ops in the target precision.
5006 2018-01-24 Richard Biener <rguenther@suse.de>
5009 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5011 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5013 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5014 to merge probabilities.
5015 * predict.c (probably_never_executed): Also mark as cold functions
5016 with global 0 profile and guessed local profile.
5017 * profile-count.c (profile_probability::combine_with_count): New
5019 * profile-count.h (profile_probability::operator*,
5020 profile_probability::operator*=, profile_probability::operator/,
5021 profile_probability::operator/=): Reduce precision to adjusted
5022 and set value to guessed on contradictory divisions.
5023 (profile_probability::combine_with_freq): Remove.
5024 (profile_probability::combine_wiht_count): Declare.
5025 (profile_count::force_nonzero):: Set to adjusted.
5026 (profile_count::probability_in):: Set quality to adjusted.
5027 * tree-ssa-tail-merge.c (replace_block_by): Use
5030 2018-01-23 Andrew Waterman <andrew@sifive.com>
5031 Jim Wilson <jimw@sifive.com>
5033 * config/riscv/riscv.c (riscv_stack_boundary): New.
5034 (riscv_option_override): Set riscv_stack_boundary. Handle
5035 riscv_preferred_stack_boundary_arg.
5036 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5037 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5038 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5039 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5040 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5041 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5043 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5046 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5047 of struct ix86_frame.
5048 (ix86_expand_epilogue): Likewise. Add a local variable for
5049 the reg_save_offset field in struct ix86_frame.
5051 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5053 PR tree-optimization/82604
5054 * tree-loop-distribution.c (enum partition_kind): New enum item
5055 PKIND_PARTIAL_MEMSET.
5056 (partition_builtin_p): Support above new enum item.
5057 (generate_code_for_partition): Ditto.
5058 (compute_access_range): Differentiate cases that equality can be
5059 proven at all loops, the innermost loops or no loops.
5060 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5061 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5062 (finalize_partitions, distribute_loop): Don't fuse partition of
5063 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5064 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5067 2018-01-23 Martin Liska <mliska@suse.cz>
5069 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5070 order to ignore the predictor.
5071 (PRED_POLYMORPHIC_CALL): Likewise.
5072 (PRED_RECURSIVE_CALL): Likewise.
5074 2018-01-23 Martin Liska <mliska@suse.cz>
5076 * tree-profile.c (tree_profiling): Print function header to
5077 aware reader which function we are working on.
5078 * value-prof.c (gimple_find_values_to_profile): Do not print
5079 not interesting value histograms.
5081 2018-01-23 Martin Liska <mliska@suse.cz>
5083 * profile-count.h (enum profile_quality): Add
5084 profile_uninitialized as the first value. Do not number values
5085 as they are zero based.
5086 (profile_count::verify): Update sanity check.
5087 (profile_probability::verify): Likewise.
5089 2018-01-23 Nathan Sidwell <nathan@acm.org>
5091 * doc/invoke.texi (ffor-scope): Deprecate.
5093 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5095 PR tree-optimization/83510
5096 * domwalk.c (set_all_edges_as_executable): New function.
5097 (dom_walker::dom_walker): Convert bool param
5098 "skip_unreachable_blocks" to enum reachability. Move setup of
5099 edge flags to set_all_edges_as_executable and only do it when
5100 reachability is REACHABLE_BLOCKS.
5101 * domwalk.h (enum dom_walker::reachability): New enum.
5102 (dom_walker::dom_walker): Convert bool param
5103 "skip_unreachable_blocks" to enum reachability.
5104 (set_all_edges_as_executable): New decl.
5105 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5106 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5108 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5109 but converting true to REACHABLE_BLOCKS.
5110 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5112 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5113 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5114 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5116 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5117 if check_all_array_refs will be called.
5119 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5121 * tree.c (selftest::test_location_wrappers): Add more test
5124 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5126 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5127 (selftest::test_bit_in_range): Likewise.
5129 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5132 * doc/sourcebuild.texi (vect_float): Say that the selector
5133 only describes the situation when -funsafe-math-optimizations is on.
5134 (vect_float_strict): Document.
5136 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5138 PR tree-optimization/83965
5139 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5140 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5141 instead of checking only for a reduction.
5142 (vect_recog_widen_sum_pattern): Likewise.
5144 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5146 * predict.c (probably_never_executed): Only use precise profile info.
5147 (compute_function_frequency): Skip after inlining hack since we now
5148 have quality checking.
5150 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5152 * profile-count.h (profile_probability::very_unlikely,
5153 profile_probability::unlikely, profile_probability::even): Set
5154 precision to guessed.
5156 2018-01-23 Richard Biener <rguenther@suse.de>
5158 PR tree-optimization/83963
5159 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5160 Properly terminate dominator walk when crossing the exit edge not
5161 when visiting its source block.
5163 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5166 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5167 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5169 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5171 PR tree-optimization/83957
5172 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5173 semicolon after for body surrounded by braces.
5175 PR tree-optimization/83081
5176 * profile-count.h (profile_probability::split): New method.
5177 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5178 Use profile_probability::split.
5179 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5180 when splitting a single conditional jump into 2.
5182 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5184 PR tree-optimization/69452
5185 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5188 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5190 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5191 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5192 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5194 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5196 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5197 * config/rl78/rl78.md (movdi): New define_expand.
5198 * config/rl78/rl78.c (rl78_split_movdi): New function.
5200 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5203 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5205 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5206 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5207 128-bit to produce an UNSPEC move to get the double word with the
5208 signbit and then a shift directly to do signbit.
5209 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5210 implementation with a new version that just does either a direct
5211 move or a regular move. Move memory interface to separate insns.
5212 Move insns so they are next to the expander.
5213 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5214 with signbit move. Split big and little endian case.
5215 (signbit<mode>2_dm_mem_le): Likewise.
5216 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5217 (signbit<mode>2_dm2): Likewise.
5219 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5221 * config/rl78/rl78.md (anddi3): New define_expand.
5223 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5225 * config/rl78/rl78.md (umindi3): New define_expand.
5227 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5229 * config/rl78/rl78.md (smindi3): New define_expand.
5231 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5233 * config/rl78/rl78.md (smaxdi3): New define_expand.
5235 2018-01-22 Carl Love <cel@us.ibm.com>
5237 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5238 LVX_V1TI): Add macro expansion.
5239 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5240 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5241 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5242 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5243 Change check to determine if the instruction is a byte reversing
5244 entry. Fix typo in comment.
5245 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5246 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5247 Add def_builtin calls for new builtins.
5248 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5249 Add define_insn expansion.
5251 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5253 * config/rl78/rl78.md (umaxdi3): New define_expand.
5255 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5257 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5258 for non-QImode registers.
5260 2018-01-22 Richard Biener <rguenther@suse.de>
5262 PR tree-optimization/83963
5263 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5264 including the loop exit block.
5265 (scop_detection::merge_sese): Likewise.
5266 (scop_detection::add_scop): Do it here instead.
5268 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5270 * doc/sourcebuild.texi (arm_softfloat): Document.
5272 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5275 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5276 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5277 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5279 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5280 David Edelsohn <dje.gcc@gmail.com>
5283 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5284 Change "crset eq" to "crset 2".
5285 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5286 (*call_indirect_aix<mode>_nospec): Likewise.
5287 (*call_value_indirect_aix<mode>_nospec): Likewise.
5288 (*call_indirect_elfv2<mode>_nospec): Likewise.
5289 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5290 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5291 change assembly output from . to $.
5292 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5293 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5294 (*tablejump<mode>_internal1_nospec): Likewise.
5296 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5299 * config/sh/sh_optimize_sett_clrt.cc:
5300 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5302 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5304 PR tree-optimization/83940
5305 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5306 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5307 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5308 use it to pass back the definition type.
5309 (vect_check_store_rhs): Likewise rhs_dt_out.
5310 (vect_build_gather_load_calls): Add a mask_dt argument and use
5311 it instead of a call to vect_is_simple_use.
5312 (vectorizable_store): Update calls to vect_check_load_store_mask
5313 and vect_check_store_rhs. Use the dt returned by the latter instead
5314 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5315 instead of calls to vect_is_simple_use. Pass the scalar rather
5316 than the vector operand to vect_is_simple_use when handling
5317 second and subsequent copies of an rhs value.
5318 (vectorizable_load): Update calls to vect_check_load_store_mask
5319 and vect_build_gather_load_calls. Use the cached mask_dt and
5320 gs_info.offset_dt instead of calls to vect_is_simple_use.
5322 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5325 * tree-emutls.c: Include gimplify.h.
5326 (lower_emutls_2): New function.
5327 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5328 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5329 it before further processing.
5332 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5333 UINTVAL (trueop1) instead of INTVAL (op1).
5335 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5339 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5340 INCOMING_FRAME_SP_OFFSET if not defined.
5341 (scan_trace): Add ENTRY argument. If true and
5342 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5343 emit a note to adjust the CFA offset.
5344 (create_cfi_notes): Adjust scan_trace callers.
5345 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5346 INCOMING_FRAME_SP_OFFSET in the CIE.
5347 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5348 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5350 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5351 * doc/tm.texi: Regenerated.
5353 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5355 PR rtl-optimization/83147
5356 * lra-constraints.c (remove_inheritance_pseudos): Use
5357 lra_substitute_pseudo_within_insn.
5359 2018-01-19 Tom de Vries <tom@codesourcery.com>
5360 Cesar Philippidis <cesar@codesourcery.com>
5363 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5365 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5368 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5369 spaces for function labels.
5371 2018-01-19 Martin Liska <mliska@suse.cz>
5373 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5374 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5375 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5376 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5377 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5378 (PRED_CONST_RETURN): Change from 69 to 65.
5379 (PRED_NULL_RETURN): Change from 91 to 71.
5380 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5381 (PRED_LOOP_GUARD): Change from 66 to 73.
5383 2018-01-19 Martin Liska <mliska@suse.cz>
5385 * predict.c (predict_insn_def): Add new assert.
5386 (struct branch_predictor): Change type to signed integer.
5387 (test_prediction_value_range): Amend test to cover
5389 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5390 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5391 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5392 (PRED_LOOP_IV_COMPARE): Likewise.
5393 * predict.h (PROB_UNINITIALIZED): Define new constant.
5395 2018-01-19 Martin Liska <mliska@suse.cz>
5397 * predict.c (dump_prediction): Add new format for
5398 analyze_brprob.py script which is enabled with -details
5400 * profile-count.h (precise_p): New function.
5402 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5404 PR tree-optimization/83922
5405 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5406 there are no statements that need masking.
5407 (vect_active_double_reduction_p): New function.
5408 (vect_analyze_loop_operations): Use it when handling phis that
5409 are not in the loop header.
5411 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5413 PR tree-optimization/83914
5414 * tree-vect-loop.c (vectorizable_induction): Don't convert
5415 init_expr or apply the peeling adjustment for inductions
5416 that are nested within the vectorized loop.
5418 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5420 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5423 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5427 * function.h (gimplify_parameters): Add gimple_seq * argument.
5428 * function.c: Include gimple.h and options.h.
5429 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5430 for the added local temporaries if needed.
5431 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5432 if there are any parameter cleanups, wrap whole body into a
5433 try/finally with the cleanups.
5435 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5438 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5439 Use GET_MODE_CLASS for scalar floating point.
5441 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5445 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5446 Fix call of call_cgraph_insertion_hooks.
5448 2018-01-18 Martin Sebor <msebor@redhat.com>
5450 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5452 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5455 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5458 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5461 * common.opt: (-ffile-prefix-map): New option.
5462 * opts.c (common_handle_option): Defer it.
5463 * opts-global.c (handle_common_deferred_options): Handle it.
5464 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5465 * file-prefix-map.h: New file.
5466 (remap_debug_filename, add_debug_prefix_map): ...here.
5467 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5468 * final.c (debug_prefix_map, add_debug_prefix_map
5469 remap_debug_filename): Move to...
5470 * file-prefix-map.c: New file.
5471 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5472 generalize, get rid of alloca(), use strrchr() instead of strchr().
5473 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5474 Implement in terms of add_prefix_map().
5475 (remap_macro_filename, remap_debug_filename): Implement in term of
5477 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5478 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5479 * dbxout.c: Include file-prefix-map.h.
5480 * varasm.c: Likewise.
5481 * vmsdbgout.c: Likewise.
5482 * xcoffout.c: Likewise.
5483 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5484 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5485 * doc/invoke.texi (-ffile-prefix-map): Document.
5486 (-fdebug-prefix-map): Update description.
5488 2018-01-18 Martin Liska <mliska@suse.cz>
5490 * config/i386/i386.c (indirect_thunk_name): Document that also
5492 (output_indirect_thunk): Document why both instructions
5493 (pause and lfence) are generated.
5495 2018-01-18 Richard Biener <rguenther@suse.de>
5497 PR tree-optimization/83887
5498 * graphite-scop-detection.c
5499 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5500 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5501 (scop_detection::merge_sese): Re-implement with a flood-fill
5502 algorithm that properly finds a SESE region if it exists.
5504 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5507 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5508 pointer_diff optimizations use view_convert instead of convert.
5510 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5512 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5513 Generate different code for -mno-speculate-indirect-jumps.
5514 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5515 (*call_indirect_aix<mode>): Disable for
5516 -mno-speculate-indirect-jumps.
5517 (*call_indirect_aix<mode>_nospec): New define_insn.
5518 (*call_value_indirect_aix<mode>): Disable for
5519 -mno-speculate-indirect-jumps.
5520 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5521 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5522 -mno-speculate-indirect-jumps.
5523 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5525 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5527 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5528 long double type, set the flags for noting the default long double
5529 type, even if we don't pass or return a long double type.
5531 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5534 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5537 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5539 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5540 support for merge[hl].
5541 (fold_mergehl_helper): New helper function.
5542 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5543 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5544 (altivec_vmrglw_direct): Add xxmrglw insn.
5546 2018-01-17 Andrew Waterman <andrew@sifive.com>
5548 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5549 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5551 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5554 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5555 call the lto_location_cache before reading the
5556 DECL_SOURCE_LOCATION of the types.
5558 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5559 Richard Sandiford <richard.sandiford@linaro.org>
5561 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5562 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5563 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5564 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5565 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5567 * config/aarch64/constraints.md (aarch64_movti_operand):
5569 * config/aarch64/predicates.md (Uti): Add new constraint.
5571 2018-01-17 Carl Love <cel@us.ibm.com>
5573 * config/rs6000/vsx.md (define_expand xl_len_r,
5574 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5575 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5577 (define_expand, define_insn): Move the shift left from the
5578 define_insn to the define_expand for lxvl and stxvl instructions.
5579 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5580 and XL_LEN_R definitions to PURE.
5582 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5584 * config/i386/i386.c (indirect_thunk_name): Declare regno
5585 as unsigned int. Compare regno with INVALID_REGNUM.
5586 (output_indirect_thunk): Ditto.
5587 (output_indirect_thunk_function): Ditto.
5588 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5589 in the call to output_indirect_thunk_function.
5591 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5594 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5595 rather than the size of inner_type to determine the stack slot size
5596 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5598 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5601 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5604 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5606 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5607 endian Linux systems to optionally enable multilibs for selecting
5608 the long double type if the user configured an explicit type.
5609 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5610 have no long double multilibs if not defined.
5611 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5612 warn if the user used -mabi={ieee,ibm}longdouble and we built
5613 multilibs for long double.
5614 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5615 appropriate multilib option.
5616 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5618 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5619 for building long double multilibs.
5620 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5622 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5624 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5627 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5629 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5632 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5635 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5638 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5640 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5641 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5643 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5645 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5646 different rtl trees depending on TARGET_64BIT.
5647 (rs6000_gen_lvx): Likewise.
5649 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5651 * config/visium/visium.md (nop): Tweak comment.
5652 (hazard_nop): Likewise.
5654 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5656 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
5657 -mspeculate-indirect-jumps.
5658 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
5659 for -mno-speculate-indirect-jumps.
5660 (*call_indirect_elfv2<mode>_nospec): New define_insn.
5661 (*call_value_indirect_elfv2<mode>): Disable for
5662 -mno-speculate-indirect-jumps.
5663 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
5664 (indirect_jump): Emit different RTL for
5665 -mno-speculate-indirect-jumps.
5666 (*indirect_jump<mode>): Disable for
5667 -mno-speculate-indirect-jumps.
5668 (*indirect_jump<mode>_nospec): New define_insn.
5669 (tablejump): Emit different RTL for
5670 -mno-speculate-indirect-jumps.
5671 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
5672 (tablejumpsi_nospec): New define_expand.
5673 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
5674 (tablejumpdi_nospec): New define_expand.
5675 (*tablejump<mode>_internal1): Disable for
5676 -mno-speculate-indirect-jumps.
5677 (*tablejump<mode>_internal1_nospec): New define_insn.
5678 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
5681 2018-01-16 Artyom Skrobov tyomitch@gmail.com
5683 * caller-save.c (insert_save): Drop unnecessary parameter. All
5686 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5687 Richard Biener <rguenth@suse.de>
5690 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
5691 return early, inline manually is_gimple_sizepos. Make sure if we
5692 call gimplify_expr we don't end up with a gimple constant.
5693 * tree.c (variably_modified_type_p): Don't return true for
5694 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
5695 * gimplify.h (is_gimple_sizepos): Remove.
5697 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5699 PR tree-optimization/83857
5700 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
5701 vectorizable_live_operation for pure SLP statements.
5702 (vectorizable_live_operation): Handle PHIs.
5704 2018-01-16 Richard Biener <rguenther@suse.de>
5706 PR tree-optimization/83867
5707 * tree-vect-stmts.c (vect_transform_stmt): Precompute
5708 nested_in_vect_loop_p since the scalar stmt may get invalidated.
5710 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5713 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
5714 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
5715 If off is not INTEGER_CST, issue a may not be aligned warning
5716 rather than isn't aligned. Use isn%'t rather than isn't.
5717 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
5719 <case MULT_EXPR>: Improve the case when bottom and one of the
5720 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
5721 operand, in that case check if the other operand is multiple of
5722 bottom divided by the INTEGER_CST operand.
5724 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5727 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
5728 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
5729 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
5730 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
5731 * config/pa/pa.c (pa_function_arg_advance): Likewise.
5732 (pa_function_arg, pa_arg_partial_bytes): Likewise.
5733 (pa_function_arg_size): New function.
5735 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5737 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
5738 in a separate statement.
5740 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
5742 PR tree-optimization/83847
5743 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
5744 group gathers and scatters.
5746 2018-01-16 Jakub Jelinek <jakub@redhat.com>
5748 PR rtl-optimization/86620
5749 * params.def (max-sched-ready-insns): Bump minimum value to 1.
5751 PR rtl-optimization/83213
5752 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
5753 to last if both are JUMP_INSNs.
5755 PR tree-optimization/83843
5756 * gimple-ssa-store-merging.c
5757 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
5758 store_immediate_info for bswap/nop orig_stores.
5760 2018-01-15 Andrew Waterman <andrew@sifive.com>
5762 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
5764 <UDIV>: Increase cost if !TARGET_DIV.
5766 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
5768 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
5769 (define_attr "cr_logical_3op"): New.
5770 (cceq_ior_compare): Adjust.
5771 (cceq_ior_compare_complement): Adjust.
5772 (*cceq_rev_compare): Adjust.
5773 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
5774 (is_cracked_insn): Adjust.
5775 (insn_must_be_first_in_group): Adjust.
5776 * config/rs6000/40x.md: Adjust.
5777 * config/rs6000/440.md: Adjust.
5778 * config/rs6000/476.md: Adjust.
5779 * config/rs6000/601.md: Adjust.
5780 * config/rs6000/603.md: Adjust.
5781 * config/rs6000/6xx.md: Adjust.
5782 * config/rs6000/7450.md: Adjust.
5783 * config/rs6000/7xx.md: Adjust.
5784 * config/rs6000/8540.md: Adjust.
5785 * config/rs6000/cell.md: Adjust.
5786 * config/rs6000/e300c2c3.md: Adjust.
5787 * config/rs6000/e500mc.md: Adjust.
5788 * config/rs6000/e500mc64.md: Adjust.
5789 * config/rs6000/e5500.md: Adjust.
5790 * config/rs6000/e6500.md: Adjust.
5791 * config/rs6000/mpc.md: Adjust.
5792 * config/rs6000/power4.md: Adjust.
5793 * config/rs6000/power5.md: Adjust.
5794 * config/rs6000/power6.md: Adjust.
5795 * config/rs6000/power7.md: Adjust.
5796 * config/rs6000/power8.md: Adjust.
5797 * config/rs6000/power9.md: Adjust.
5798 * config/rs6000/rs64.md: Adjust.
5799 * config/rs6000/titan.md: Adjust.
5801 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5803 * config/i386/predicates.md (indirect_branch_operand): Rewrite
5804 ix86_indirect_branch_register logic.
5806 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5808 * config/i386/constraints.md (Bs): Update
5809 ix86_indirect_branch_register check. Don't check
5810 ix86_indirect_branch_register with GOT_memory_operand.
5812 * config/i386/predicates.md (GOT_memory_operand): Don't check
5813 ix86_indirect_branch_register here.
5814 (GOT32_symbol_operand): Likewise.
5816 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5818 * config/i386/predicates.md (constant_call_address_operand):
5819 Rewrite ix86_indirect_branch_register logic.
5820 (sibcall_insn_operand): Likewise.
5822 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5824 * config/i386/constraints.md (Bs): Replace
5825 ix86_indirect_branch_thunk_register with
5826 ix86_indirect_branch_register.
5828 * config/i386/i386.md (indirect_jump): Likewise.
5829 (tablejump): Likewise.
5830 (*sibcall_memory): Likewise.
5831 (*sibcall_value_memory): Likewise.
5832 Peepholes of indirect call and jump via memory: Likewise.
5833 * config/i386/i386.opt: Likewise.
5834 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5835 (GOT_memory_operand): Likewise.
5836 (call_insn_operand): Likewise.
5837 (sibcall_insn_operand): Likewise.
5838 (GOT32_symbol_operand): Likewise.
5840 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5843 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
5844 type rather than type addr's type points to.
5845 (expand_omp_atomic_mutex): Likewise.
5846 (expand_omp_atomic): Likewise.
5848 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
5851 * config/i386/i386.c (output_indirect_thunk_function): Use
5852 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
5853 for __x86_return_thunk.
5855 2018-01-15 Richard Biener <rguenther@suse.de>
5858 * expmed.c (extract_bit_field_1): Fix typo.
5860 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5863 * config/arm/iterators.md (VF): New mode iterator.
5864 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
5865 Remove integer-related logic from pattern.
5866 (neon_vabd<mode>_3): Likewise.
5868 2018-01-15 Jakub Jelinek <jakub@redhat.com>
5871 * common.opt (fstrict-overflow): No longer an alias.
5872 (fwrapv-pointer): New option.
5873 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
5874 also for pointer types based on flag_wrapv_pointer.
5875 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
5876 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
5877 opts->x_flag_wrapv got set.
5878 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
5879 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
5880 POINTER_TYPE_OVERFLOW_UNDEFINED.
5881 * match.pd: Likewise in address comparison pattern.
5882 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
5884 2018-01-15 Richard Biener <rguenther@suse.de>
5887 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
5888 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
5889 Reset type names to their identifier if their TYPE_DECL doesn't
5890 have linkage (and thus is used for ODR and devirt).
5891 (save_debug_info_for_decl): Remove.
5892 (save_debug_info_for_type): Likewise.
5893 (add_tree_to_fld_list): Adjust.
5894 * tree-pretty-print.c (dump_generic_node): Make dumping of
5895 type names more robust.
5897 2018-01-15 Richard Biener <rguenther@suse.de>
5899 * BASE-VER: Bump to 8.0.1.
5901 2018-01-14 Martin Sebor <msebor@redhat.com>
5904 * builtins.c (check_access): Avoid warning when the no-warning bit
5907 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
5909 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
5910 * ira-color (allocno_hard_regs_compare): Likewise.
5912 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
5915 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
5916 Use .pushsection/.popsection.
5918 2018-01-14 Martin Sebor <msebor@redhat.com>
5921 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
5923 2018-01-14 Jakub Jelinek <jakub@redhat.com>
5925 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
5926 entry from extra_headers.
5927 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
5928 extra_headers, make the list bitwise identical to the i?86-*-* one.
5930 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5932 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
5933 -mcmodel=large with -mindirect-branch=thunk,
5934 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
5935 -mfunction-return=thunk-extern.
5936 * doc/invoke.texi: Document -mcmodel=large is incompatible with
5937 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
5938 -mfunction-return=thunk and -mfunction-return=thunk-extern.
5940 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5942 * config/i386/i386.c (print_reg): Print the name of the full
5943 integer register without '%'.
5944 (ix86_print_operand): Handle 'V'.
5945 * doc/extend.texi: Document 'V' modifier.
5947 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5949 * config/i386/constraints.md (Bs): Disallow memory operand for
5950 -mindirect-branch-register.
5952 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5953 (GOT_memory_operand): Likewise.
5954 (call_insn_operand): Likewise.
5955 (sibcall_insn_operand): Likewise.
5956 (GOT32_symbol_operand): Likewise.
5957 * config/i386/i386.md (indirect_jump): Call convert_memory_address
5958 for -mindirect-branch-register.
5959 (tablejump): Likewise.
5960 (*sibcall_memory): Likewise.
5961 (*sibcall_value_memory): Likewise.
5962 Disallow peepholes of indirect call and jump via memory for
5963 -mindirect-branch-register.
5964 (*call_pop): Replace m with Bw.
5965 (*call_value_pop): Likewise.
5966 (*sibcall_pop_memory): Replace m with Bs.
5967 * config/i386/i386.opt (mindirect-branch-register): New option.
5968 * doc/invoke.texi: Document -mindirect-branch-register option.
5970 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5972 * config/i386/i386-protos.h (ix86_output_function_return): New.
5973 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
5974 set function_return_type.
5975 (indirect_thunk_name): Add ret_p to indicate thunk for function
5977 (output_indirect_thunk_function): Pass false to
5978 indirect_thunk_name.
5979 (ix86_output_indirect_branch_via_reg): Likewise.
5980 (ix86_output_indirect_branch_via_push): Likewise.
5981 (output_indirect_thunk_function): Create alias for function
5982 return thunk if regno < 0.
5983 (ix86_output_function_return): New function.
5984 (ix86_handle_fndecl_attribute): Handle function_return.
5985 (ix86_attribute_table): Add function_return.
5986 * config/i386/i386.h (machine_function): Add
5987 function_return_type.
5988 * config/i386/i386.md (simple_return_internal): Use
5989 ix86_output_function_return.
5990 (simple_return_internal_long): Likewise.
5991 * config/i386/i386.opt (mfunction-return=): New option.
5992 (indirect_branch): Mention -mfunction-return=.
5993 * doc/extend.texi: Document function_return function attribute.
5994 * doc/invoke.texi: Document -mfunction-return= option.
5996 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
5998 * config/i386/i386-opts.h (indirect_branch): New.
5999 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6000 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6001 with local indirect jump when converting indirect call and jump.
6002 (ix86_set_indirect_branch_type): New.
6003 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6004 (indirectlabelno): New.
6005 (indirect_thunk_needed): Likewise.
6006 (indirect_thunk_bnd_needed): Likewise.
6007 (indirect_thunks_used): Likewise.
6008 (indirect_thunks_bnd_used): Likewise.
6009 (INDIRECT_LABEL): Likewise.
6010 (indirect_thunk_name): Likewise.
6011 (output_indirect_thunk): Likewise.
6012 (output_indirect_thunk_function): Likewise.
6013 (ix86_output_indirect_branch_via_reg): Likewise.
6014 (ix86_output_indirect_branch_via_push): Likewise.
6015 (ix86_output_indirect_branch): Likewise.
6016 (ix86_output_indirect_jmp): Likewise.
6017 (ix86_code_end): Call output_indirect_thunk_function if needed.
6018 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6020 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6021 (ix86_attribute_table): Add indirect_branch.
6022 * config/i386/i386.h (machine_function): Add indirect_branch_type
6023 and has_local_indirect_jump.
6024 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6026 (tablejump): Likewise.
6027 (*indirect_jump): Use ix86_output_indirect_jmp.
6028 (*tablejump_1): Likewise.
6029 (simple_return_indirect_internal): Likewise.
6030 * config/i386/i386.opt (mindirect-branch=): New option.
6031 (indirect_branch): New.
6034 (thunk-inline): Likewise.
6035 (thunk-extern): Likewise.
6036 * doc/extend.texi: Document indirect_branch function attribute.
6037 * doc/invoke.texi: Document -mindirect-branch= option.
6039 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6042 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6044 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6046 * ipa-inline.c (want_inline_small_function_p): Return false if
6047 inlining has already failed with CIF_FINAL_ERROR.
6048 (update_caller_keys): Call want_inline_small_function_p before
6050 (update_callee_keys): Likewise.
6052 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6054 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6056 (rs6000_quadword_masked_address_p): Likewise.
6057 (quad_aligned_load_p): Likewise.
6058 (quad_aligned_store_p): Likewise.
6059 (const_load_sequence_p): Add comment to describe the outer-most loop.
6060 (mimic_memory_attributes_and_flags): New function.
6061 (rs6000_gen_stvx): Likewise.
6062 (replace_swapped_aligned_store): Likewise.
6063 (rs6000_gen_lvx): Likewise.
6064 (replace_swapped_aligned_load): Likewise.
6065 (replace_swapped_load_constant): Capitalize argument name in
6066 comment describing this function.
6067 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6068 and stores that access quad-word aligned addresses and replace
6069 with stvx or lvx instructions when appropriate.
6070 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6071 New function prototype.
6072 (rs6000_quadword_masked_address_p): Likewise.
6073 (rs6000_gen_lvx): Likewise.
6074 (rs6000_gen_stvx): Likewise.
6075 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6076 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6077 when memory address is aligned.
6078 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6079 this split to select lvx instruction when memory address is aligned.
6080 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6081 instruction when memory address is aligned.
6082 (*vsx_le_perm_load_v16qi): Likewise.
6083 (four unnamed splitters): Modify to select the stvx instruction
6084 when memory is aligned.
6086 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6088 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6089 which appears in the queue multiple times.
6091 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6092 Alan Hayward <alan.hayward@arm.com>
6093 David Sherwood <david.sherwood@arm.com>
6095 * tree-vectorizer.h (vec_lower_bound): New structure.
6096 (_loop_vec_info): Add check_nonzero and lower_bounds.
6097 (LOOP_VINFO_CHECK_NONZERO): New macro.
6098 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6099 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6100 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6101 fields. Make seg_len the distance travelled, not including the
6103 (dr_direction_indicator): Declare.
6104 (dr_zero_step_indicator): Likewise.
6105 (dr_known_forward_stride_p): Likewise.
6106 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6108 (runtime_alias_check_p): Allow runtime alias checks with
6110 (operator ==): Compare access_size and align.
6111 (prune_runtime_alias_test_list): Rework for new distinction between
6112 the access_size and seg_len.
6113 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6115 (get_segment_min_max): New function.
6116 (create_intersect_range_checks): Use it.
6117 (dr_step_indicator): New function.
6118 (dr_direction_indicator): Likewise.
6119 (dr_zero_step_indicator): Likewise.
6120 (dr_known_forward_stride_p): Likewise.
6121 * tree-loop-distribution.c (data_ref_segment_size): Return
6122 DR_STEP * (niters - 1).
6123 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6125 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6126 (vect_preserves_scalar_order_p): New function, split out from...
6127 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6128 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6129 (vect_vfa_access_size): New function.
6130 (vect_vfa_align): Likewise.
6131 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6132 (dump_lower_bound): New function.
6133 (vect_check_lower_bound): Likewise.
6134 (vect_small_gap_p): Likewise.
6135 (vectorizable_with_step_bound_p): Likewise.
6136 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6137 depencies if the vectorization factor is 1. Convert the checks
6138 for nonzero steps into checks on the bounds of DR_STEP. Try using
6139 a bunds check for variable steps if the minimum required step is
6140 relatively small. Update calls to the dr_with_seg_len
6141 constructor and to vect_compile_time_alias.
6142 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6144 (vect_loop_versioning): Call it.
6145 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6147 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6150 Alan Hayward <alan.hayward@arm.com>
6151 David Sherwood <david.sherwood@arm.com>
6153 * doc/sourcebuild.texi (vect_scatter_store): Document.
6154 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6156 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6158 * genopinit.c (main): Add supports_vec_scatter_store and
6159 supports_vec_scatter_store_cached to target_optabs.
6160 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6161 IFN_MASK_SCATTER_STORE.
6162 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6164 * internal-fn.h (internal_store_fn_p): Declare.
6165 (internal_fn_stored_value_index): Likewise.
6166 * internal-fn.c (scatter_store_direct): New macro.
6167 (expand_scatter_store_optab_fn): New function.
6168 (direct_scatter_store_optab_supported_p): New macro.
6169 (internal_store_fn_p): New function.
6170 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6171 IFN_MASK_SCATTER_STORE.
6172 (internal_fn_mask_index): Likewise.
6173 (internal_fn_stored_value_index): New function.
6174 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6176 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6177 * optabs-query.c (supports_vec_scatter_store_p): New function.
6178 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6179 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6180 true for scatter stores.
6181 (vect_gather_scatter_fn_p): Handle scatter stores too.
6182 (vect_check_gather_scatter): Consider using scatter stores if
6183 supports_vec_scatter_store_p.
6184 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6186 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6187 internal_fn_stored_value_index.
6188 (check_load_store_masking): Handle scatter stores too.
6189 (vect_get_store_rhs): Make public.
6190 (vectorizable_call): Use internal_store_fn_p.
6191 (vectorizable_store): Handle scatter store internal functions.
6192 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6193 when deciding whether the end of the group has been reached.
6194 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6195 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6196 (mask_scatter_store<mode>): New insns.
6198 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6199 Alan Hayward <alan.hayward@arm.com>
6200 David Sherwood <david.sherwood@arm.com>
6202 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6203 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6204 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6206 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6207 Use vect_truncate_gather_scatter_offset if we can't treat the
6208 operation as a normal gather load or scatter store.
6209 (get_group_load_store_type): Take the gather_scatter_info
6210 as argument. Try using a gather load or scatter store for
6211 single-element groups.
6212 (get_load_store_type): Update calls to get_group_load_store_type
6213 and vect_use_strided_gather_scatters_p.
6215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6216 Alan Hayward <alan.hayward@arm.com>
6217 David Sherwood <david.sherwood@arm.com>
6219 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6220 optional tree argument.
6221 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6223 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6224 but continue to use the current value as a fallback.
6225 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6226 to compare the updates.
6227 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6228 (get_load_store_type): Use it when handling a strided access.
6229 (vect_get_strided_load_store_ops): New function.
6230 (vect_get_data_ptr_increment): Likewise.
6231 (vectorizable_load): Handle strided gather loads. Always pass
6232 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6234 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6235 Alan Hayward <alan.hayward@arm.com>
6236 David Sherwood <david.sherwood@arm.com>
6238 * doc/md.texi (gather_load@var{m}): Document.
6239 (mask_gather_load@var{m}): Likewise.
6240 * genopinit.c (main): Add supports_vec_gather_load and
6241 supports_vec_gather_load_cached to target_optabs.
6242 * optabs-tree.c (init_tree_optimization_optabs): Use
6243 ggc_cleared_alloc to allocate target_optabs.
6244 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6245 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6247 * internal-fn.h (internal_load_fn_p): Declare.
6248 (internal_gather_scatter_fn_p): Likewise.
6249 (internal_fn_mask_index): Likewise.
6250 (internal_gather_scatter_fn_supported_p): Likewise.
6251 * internal-fn.c (gather_load_direct): New macro.
6252 (expand_gather_load_optab_fn): New function.
6253 (direct_gather_load_optab_supported_p): New macro.
6254 (direct_internal_fn_optab): New function.
6255 (internal_load_fn_p): Likewise.
6256 (internal_gather_scatter_fn_p): Likewise.
6257 (internal_fn_mask_index): Likewise.
6258 (internal_gather_scatter_fn_supported_p): Likewise.
6259 * optabs-query.c (supports_at_least_one_mode_p): New function.
6260 (supports_vec_gather_load_p): Likewise.
6261 * optabs-query.h (supports_vec_gather_load_p): Declare.
6262 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6263 and memory_type field.
6264 (NUM_PATTERNS): Bump to 15.
6265 * tree-vect-data-refs.c: Include internal-fn.h.
6266 (vect_gather_scatter_fn_p): New function.
6267 (vect_describe_gather_scatter_call): Likewise.
6268 (vect_check_gather_scatter): Try using internal functions for
6269 gather loads. Recognize existing calls to a gather load function.
6270 (vect_analyze_data_refs): Consider using gather loads if
6271 supports_vec_gather_load_p.
6272 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6273 (vect_get_gather_scatter_offset_type): Likewise.
6274 (vect_convert_mask_for_vectype): Likewise.
6275 (vect_add_conversion_to_patterm): Likewise.
6276 (vect_try_gather_scatter_pattern): Likewise.
6277 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6278 (vect_vect_recog_func_ptrs): Add it.
6279 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6280 internal_fn_mask_index and internal_gather_scatter_fn_p.
6281 (check_load_store_masking): Take the gather_scatter_info as an
6282 argument and handle gather loads.
6283 (vect_get_gather_scatter_ops): New function.
6284 (vectorizable_call): Check internal_load_fn_p.
6285 (vectorizable_load): Likewise. Handle gather load internal
6287 (vectorizable_store): Update call to check_load_store_masking.
6288 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6289 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6290 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6291 (aarch64_gather_scale_operand_d): New predicates.
6292 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6293 (mask_gather_load<mode>): New insns.
6295 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6296 Alan Hayward <alan.hayward@arm.com>
6297 David Sherwood <david.sherwood@arm.com>
6299 * optabs.def (fold_left_plus_optab): New optab.
6300 * doc/md.texi (fold_left_plus_@var{m}): Document.
6301 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6302 * internal-fn.c (fold_left_direct): Define.
6303 (expand_fold_left_optab_fn): Likewise.
6304 (direct_fold_left_optab_supported_p): Likewise.
6305 * fold-const-call.c (fold_const_fold_left): New function.
6306 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6307 * tree-parloops.c (valid_reduction_p): New function.
6308 (gather_scalar_reductions): Use it.
6309 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6310 (vect_finish_replace_stmt): Declare.
6311 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6312 (needs_fold_left_reduction_p): New function, split out from...
6313 (vect_is_simple_reduction): ...here. Accept reductions that
6314 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6315 (vect_force_simple_reduction): Also store the reduction type in
6316 the assignment's STMT_VINFO_REDUC_TYPE.
6317 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6318 (merge_with_identity): New function.
6319 (vect_expand_fold_left): Likewise.
6320 (vectorize_fold_left_reduction): Likewise.
6321 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6322 scalar phi in place for it. Check for target support and reject
6323 cases that would reassociate the operation. Defer the transform
6324 phase to vectorize_fold_left_reduction.
6325 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6326 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6327 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6329 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6331 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6332 call to ifc_temp_var.
6334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6335 Alan Hayward <alan.hayward@arm.com>
6336 David Sherwood <david.sherwood@arm.com>
6338 * target.def (legitimize_address_displacement): Take the original
6339 offset as a poly_int.
6340 * targhooks.h (default_legitimize_address_displacement): Update
6342 * targhooks.c (default_legitimize_address_displacement): Likewise.
6343 * doc/tm.texi: Regenerate.
6344 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6345 as an argument, moving assert of ad->disp == ad->disp_term to...
6346 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6347 Try calling targetm.legitimize_address_displacement before expanding
6348 the address rather than afterwards, and adjust for the new interface.
6349 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6350 Match the new hook interface. Handle SVE addresses.
6351 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6356 * Makefile.in (OBJS): Add early-remat.o.
6357 * target.def (select_early_remat_modes): New hook.
6358 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6359 * doc/tm.texi: Regenerate.
6360 * targhooks.h (default_select_early_remat_modes): Declare.
6361 * targhooks.c (default_select_early_remat_modes): New function.
6362 * timevar.def (TV_EARLY_REMAT): New timevar.
6363 * passes.def (pass_early_remat): New pass.
6364 * tree-pass.h (make_pass_early_remat): Declare.
6365 * early-remat.c: New file.
6366 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6368 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6370 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6371 Alan Hayward <alan.hayward@arm.com>
6372 David Sherwood <david.sherwood@arm.com>
6374 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6375 vfm1 with a bound_epilog parameter.
6376 (vect_do_peeling): Update calls accordingly, and move the prologue
6377 call earlier in the function. Treat the base bound_epilog as 0 for
6378 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6379 this base when peeling for gaps.
6380 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6381 with fully-masked loops.
6382 (vect_estimate_min_profitable_iters): Handle the single peeled
6383 iteration in that case.
6385 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6386 Alan Hayward <alan.hayward@arm.com>
6387 David Sherwood <david.sherwood@arm.com>
6389 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6390 single-element interleaving even if the size is not a power of 2.
6391 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6392 accesses for single-element interleaving if the group size is
6395 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6396 Alan Hayward <alan.hayward@arm.com>
6397 David Sherwood <david.sherwood@arm.com>
6399 * doc/md.texi (fold_extract_last_@var{m}): Document.
6400 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6401 * optabs.def (fold_extract_last_optab): New optab.
6402 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6403 * internal-fn.c (fold_extract_direct): New macro.
6404 (expand_fold_extract_optab_fn): Likewise.
6405 (direct_fold_extract_optab_supported_p): Likewise.
6406 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6407 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6408 EXTRACT_LAST_REDUCTION.
6409 (get_initial_def_for_reduction): Do not create an initial vector
6410 for EXTRACT_LAST_REDUCTION reductions.
6411 (vectorizable_reduction): Leave the scalar phi in place for
6412 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6413 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6414 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6415 transform phase to vectorizable_condition.
6416 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6418 (vect_finish_stmt_generation): ...here.
6419 (vect_finish_replace_stmt): New function.
6420 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6421 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6423 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6425 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6426 Alan Hayward <alan.hayward@arm.com>
6427 David Sherwood <david.sherwood@arm.com>
6429 * doc/md.texi (extract_last_@var{m}): Document.
6430 * optabs.def (extract_last_optab): New optab.
6431 * internal-fn.def (EXTRACT_LAST): New internal function.
6432 * internal-fn.c (cond_unary_direct): New macro.
6433 (expand_cond_unary_optab_fn): Likewise.
6434 (direct_cond_unary_optab_supported_p): Likewise.
6435 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6436 loops using EXTRACT_LAST.
6437 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6438 (extract_last_<mode>): ...this optab.
6439 (vec_extract<mode><Vel>): Update accordingly.
6441 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6442 Alan Hayward <alan.hayward@arm.com>
6443 David Sherwood <david.sherwood@arm.com>
6445 * target.def (empty_mask_is_expensive): New hook.
6446 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6447 * doc/tm.texi: Regenerate.
6448 * targhooks.h (default_empty_mask_is_expensive): Declare.
6449 * targhooks.c (default_empty_mask_is_expensive): New function.
6450 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6451 if the target says that empty masks are expensive.
6452 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6454 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6456 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6457 Alan Hayward <alan.hayward@arm.com>
6458 David Sherwood <david.sherwood@arm.com>
6460 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6461 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6462 (vect_use_loop_mask_for_alignment_p): New function.
6463 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6464 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6465 niters_skip argument. Make sure that the first niters_skip elements
6466 of the first iteration are inactive.
6467 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6468 Update call to vect_set_loop_masks_directly.
6469 (get_misalign_in_elems): New function, split out from...
6470 (vect_gen_prolog_loop_niters): ...here.
6471 (vect_update_init_of_dr): Take a code argument that specifies whether
6472 the adjustment should be added or subtracted.
6473 (vect_update_init_of_drs): Likewise.
6474 (vect_prepare_for_masked_peels): New function.
6475 (vect_do_peeling): Skip prologue peeling if we're using a mask
6476 instead. Update call to vect_update_inits_of_drs.
6477 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6479 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6480 alignment. Do not include the number of peeled iterations in
6481 the minimum threshold in that case.
6482 (vectorizable_induction): Adjust the start value down by
6483 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6484 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6485 Take the number of skipped iterations into account when calculating
6487 * tree-vect-stmts.c (vect_gen_while_not): New function.
6489 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6490 Alan Hayward <alan.hayward@arm.com>
6491 David Sherwood <david.sherwood@arm.com>
6493 * doc/sourcebuild.texi (vect_fully_masked): Document.
6494 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6496 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6498 (vect_analyze_loop_2): ...here. Don't check the vectorization
6499 factor against the number of loop iterations if the loop is
6502 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6503 Alan Hayward <alan.hayward@arm.com>
6504 David Sherwood <david.sherwood@arm.com>
6506 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6507 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6508 (dump_groups): Update accordingly.
6509 (iv_use::mem_type): New member variable.
6510 (address_p): New function.
6511 (record_use): Add a mem_type argument and initialize the new
6513 (record_group_use): Add a mem_type argument. Use address_p.
6514 Remove obsolete null checks of base_object. Update call to record_use.
6515 (find_interesting_uses_op): Update call to record_group_use.
6516 (find_interesting_uses_cond): Likewise.
6517 (find_interesting_uses_address): Likewise.
6518 (get_mem_type_for_internal_fn): New function.
6519 (find_address_like_use): Likewise.
6520 (find_interesting_uses_stmt): Try find_address_like_use before
6521 calling find_interesting_uses_op.
6522 (addr_offset_valid_p): Use the iv mem_type field as the type
6523 of the addressed memory.
6524 (add_autoinc_candidates): Likewise.
6525 (get_address_cost): Likewise.
6526 (split_small_address_groups_p): Use address_p.
6527 (split_address_groups): Likewise.
6528 (add_iv_candidate_for_use): Likewise.
6529 (autoinc_possible_for_pair): Likewise.
6530 (rewrite_groups): Likewise.
6531 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6532 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6533 (get_alias_ptr_type_for_ptr_address): New function.
6534 (rewrite_use_address): Rewrite address uses in calls that were
6535 identified by find_address_like_use.
6537 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6538 Alan Hayward <alan.hayward@arm.com>
6539 David Sherwood <david.sherwood@arm.com>
6541 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6543 * gimple-expr.h (is_gimple_addressable: Likewise.
6544 * gimple-expr.c (is_gimple_address): Likewise.
6545 * internal-fn.c (expand_call_mem_ref): New function.
6546 (expand_mask_load_optab_fn): Use it.
6547 (expand_mask_store_optab_fn): Likewise.
6549 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6550 Alan Hayward <alan.hayward@arm.com>
6551 David Sherwood <david.sherwood@arm.com>
6553 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6554 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6555 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6556 (cond_umax@var{mode}): Document.
6557 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6558 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6559 (cond_umin_optab, cond_umax_optab): New optabs.
6560 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6561 (COND_IOR, COND_XOR): New internal functions.
6562 * internal-fn.h (get_conditional_internal_fn): Declare.
6563 * internal-fn.c (cond_binary_direct): New macro.
6564 (expand_cond_binary_optab_fn): Likewise.
6565 (direct_cond_binary_optab_supported_p): Likewise.
6566 (get_conditional_internal_fn): New function.
6567 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6568 Cope with reduction statements that are vectorized as calls rather
6570 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6571 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6572 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6573 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6574 (UNSPEC_COND_EOR): New unspecs.
6575 (optab): Add mappings for them.
6576 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6577 (sve_int_op, sve_fp_op): New int attributes.
6579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6580 Alan Hayward <alan.hayward@arm.com>
6581 David Sherwood <david.sherwood@arm.com>
6583 * optabs.def (while_ult_optab): New optab.
6584 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6585 * internal-fn.def (WHILE_ULT): New internal function.
6586 * internal-fn.h (direct_internal_fn_supported_p): New override
6587 that takes two types as argument.
6588 * internal-fn.c (while_direct): New macro.
6589 (expand_while_optab_fn): New function.
6590 (convert_optab_supported_p): Likewise.
6591 (direct_while_optab_supported_p): New macro.
6592 * wide-int.h (wi::udiv_ceil): New function.
6593 * tree-vectorizer.h (rgroup_masks): New structure.
6594 (vec_loop_masks): New typedef.
6595 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6597 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6598 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6599 (vect_max_vf): New function.
6600 (slpeel_make_loop_iterate_ntimes): Delete.
6601 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6602 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6603 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6604 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6605 internal-fn.h, stor-layout.h and optabs-query.h.
6606 (vect_set_loop_mask): New function.
6607 (add_preheader_seq): Likewise.
6608 (add_header_seq): Likewise.
6609 (interleave_supported_p): Likewise.
6610 (vect_maybe_permute_loop_masks): Likewise.
6611 (vect_set_loop_masks_directly): Likewise.
6612 (vect_set_loop_condition_masked): Likewise.
6613 (vect_set_loop_condition_unmasked): New function, split out from
6614 slpeel_make_loop_iterate_ntimes.
6615 (slpeel_make_loop_iterate_ntimes): Rename to..
6616 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6617 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6618 (vect_do_peeling): Update call accordingly.
6619 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6621 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6622 mask_compare_type, can_fully_mask_p and fully_masked_p.
6623 (release_vec_loop_masks): New function.
6624 (_loop_vec_info): Use it to free the loop masks.
6625 (can_produce_all_loop_masks_p): New function.
6626 (vect_get_max_nscalars_per_iter): Likewise.
6627 (vect_verify_full_masking): Likewise.
6628 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6629 retries, and free the mask rgroups before retrying. Check loop-wide
6630 reasons for disallowing fully-masked loops. Make the final decision
6631 about whether use a fully-masked loop or not.
6632 (vect_estimate_min_profitable_iters): Do not assume that peeling
6633 for the number of iterations will be needed for fully-masked loops.
6634 (vectorizable_reduction): Disable fully-masked loops.
6635 (vectorizable_live_operation): Likewise.
6636 (vect_halve_mask_nunits): New function.
6637 (vect_double_mask_nunits): Likewise.
6638 (vect_record_loop_mask): Likewise.
6639 (vect_get_loop_mask): Likewise.
6640 (vect_transform_loop): Handle the case in which the final loop
6641 iteration might handle a partial vector. Call vect_set_loop_condition
6642 instead of slpeel_make_loop_iterate_ntimes.
6643 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6644 (check_load_store_masking): New function.
6645 (prepare_load_store_mask): Likewise.
6646 (vectorizable_store): Handle fully-masked loops.
6647 (vectorizable_load): Likewise.
6648 (supportable_widening_operation): Use vect_halve_mask_nunits for
6650 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6651 (vect_gen_while): New function.
6652 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6653 (aarch64_uqdec<mode>): New insn.
6655 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6656 Alan Hayward <alan.hayward@arm.com>
6657 David Sherwood <david.sherwood@arm.com>
6659 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
6660 (reduc_xor_scal_optab): New optabs.
6661 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
6662 (reduc_xor_scal_@var{m}): Document.
6663 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
6664 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
6666 * fold-const-call.c (fold_const_call): Handle them.
6667 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
6668 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
6669 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
6670 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
6671 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
6672 (UNSPEC_XORV): New unspecs.
6673 (optab): Add entries for them.
6674 (BITWISEV): New int iterator.
6675 (bit_reduc_op): New int attributes.
6677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6678 Alan Hayward <alan.hayward@arm.com>
6679 David Sherwood <david.sherwood@arm.com>
6681 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
6682 * internal-fn.def (VEC_SHL_INSERT): New internal function.
6683 * optabs.def (vec_shl_insert_optab): New optab.
6684 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
6685 (duplicate_and_interleave): Likewise.
6686 * tree-vect-loop.c: Include internal-fn.h.
6687 (neutral_op_for_slp_reduction): New function, split out from
6688 get_initial_defs_for_reduction.
6689 (get_initial_def_for_reduction): Handle option 2 for variable-length
6690 vectors by loading the neutral value into a vector and then shifting
6691 the initial value into element 0.
6692 (get_initial_defs_for_reduction): Replace the code argument with
6693 the neutral value calculated by neutral_op_for_slp_reduction.
6694 Use gimple_build_vector for constant-length vectors.
6695 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
6696 but the first group_size elements have a neutral value.
6697 Use duplicate_and_interleave otherwise.
6698 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
6699 Update call to get_initial_defs_for_reduction. Handle SLP
6700 reductions for variable-length vectors by creating one vector
6701 result for each scalar result, with the elements associated
6702 with other scalar results stubbed out with the neutral value.
6703 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
6704 Require IFN_VEC_SHL_INSERT for double reductions on
6705 variable-length vectors, or SLP reductions that have
6706 a neutral value. Require can_duplicate_and_interleave_p
6707 support for variable-length unchained SLP reductions if there
6708 is no neutral value, such as for MIN/MAX reductions. Also require
6709 the number of vector elements to be a multiple of the number of
6710 SLP statements when doing variable-length unchained SLP reductions.
6711 Update call to vect_create_epilog_for_reduction.
6712 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
6713 and remove initial values.
6714 (duplicate_and_interleave): Make public.
6715 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
6716 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
6718 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6719 Alan Hayward <alan.hayward@arm.com>
6720 David Sherwood <david.sherwood@arm.com>
6722 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
6723 (can_duplicate_and_interleave_p): New function.
6724 (vect_get_and_check_slp_defs): Take the vector of statements
6725 rather than just the current one. Remove excess parentheses.
6726 Restriction rejectinon of vect_constant_def and vect_external_def
6727 for variable-length vectors to boolean types, or types for which
6728 can_duplicate_and_interleave_p is false.
6729 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
6730 (duplicate_and_interleave): New function.
6731 (vect_get_constant_vectors): Use gimple_build_vector for
6732 constant-length vectors and suitable variable-length constant
6733 vectors. Use duplicate_and_interleave for other variable-length
6734 vectors. Don't defer the update when inserting new statements.
6736 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6737 Alan Hayward <alan.hayward@arm.com>
6738 David Sherwood <david.sherwood@arm.com>
6740 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
6741 min_profitable_iters doesn't go negative.
6743 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6744 Alan Hayward <alan.hayward@arm.com>
6745 David Sherwood <david.sherwood@arm.com>
6747 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
6748 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
6749 * optabs.def (vec_mask_load_lanes_optab): New optab.
6750 (vec_mask_store_lanes_optab): Likewise.
6751 * internal-fn.def (MASK_LOAD_LANES): New internal function.
6752 (MASK_STORE_LANES): Likewise.
6753 * internal-fn.c (mask_load_lanes_direct): New macro.
6754 (mask_store_lanes_direct): Likewise.
6755 (expand_mask_load_optab_fn): Handle masked operations.
6756 (expand_mask_load_lanes_optab_fn): New macro.
6757 (expand_mask_store_optab_fn): Handle masked operations.
6758 (expand_mask_store_lanes_optab_fn): New macro.
6759 (direct_mask_load_lanes_optab_supported_p): Likewise.
6760 (direct_mask_store_lanes_optab_supported_p): Likewise.
6761 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
6763 (vect_load_lanes_supported): Likewise.
6764 * tree-vect-data-refs.c (strip_conversion): New function.
6765 (can_group_stmts_p): Likewise.
6766 (vect_analyze_data_ref_accesses): Use it instead of checking
6767 for a pair of assignments.
6768 (vect_store_lanes_supported): Take a masked_p parameter.
6769 (vect_load_lanes_supported): Likewise.
6770 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
6771 vect_store_lanes_supported and vect_load_lanes_supported.
6772 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6773 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
6774 parameter. Don't allow gaps for masked accesses.
6775 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
6776 and vect_load_lanes_supported.
6777 (get_load_store_type): Take a masked_p parameter and update
6778 call to get_group_load_store_type.
6779 (vectorizable_store): Update call to get_load_store_type.
6780 Handle IFN_MASK_STORE_LANES.
6781 (vectorizable_load): Update call to get_load_store_type.
6782 Handle IFN_MASK_LOAD_LANES.
6784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6785 Alan Hayward <alan.hayward@arm.com>
6786 David Sherwood <david.sherwood@arm.com>
6788 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
6790 * config/aarch64/aarch64-protos.h
6791 (aarch64_sve_struct_memory_operand_p): Declare.
6792 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
6793 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
6794 (VPRED, vpred): Handle SVE structure modes.
6795 * config/aarch64/constraints.md (Utx): New constraint.
6796 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
6797 (aarch64_sve_struct_nonimmediate_operand): New predicates.
6798 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
6799 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
6800 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
6801 structure modes. Split into pieces after RA.
6802 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
6803 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
6805 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
6806 SVE structure modes.
6807 (aarch64_classify_address): Likewise.
6808 (sizetochar): Move earlier in file.
6809 (aarch64_print_operand): Handle SVE register lists.
6810 (aarch64_array_mode): New function.
6811 (aarch64_sve_struct_memory_operand_p): Likewise.
6812 (TARGET_ARRAY_MODE): Redefine.
6814 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6815 Alan Hayward <alan.hayward@arm.com>
6816 David Sherwood <david.sherwood@arm.com>
6818 * target.def (array_mode): New target hook.
6819 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
6820 * doc/tm.texi: Regenerate.
6821 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
6822 * hooks.c (hook_optmode_mode_uhwi_none): New function.
6823 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
6825 * stor-layout.c (mode_for_array): Likewise. Support polynomial
6828 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6829 Alan Hayward <alan.hayward@arm.com>
6830 David Sherwood <david.sherwood@arm.com>
6832 * fold-const.c (fold_binary_loc): Check the argument types
6833 rather than the result type when testing for a vector operation.
6835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6837 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
6838 * doc/tm.texi: Regenerate.
6840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6841 Alan Hayward <alan.hayward@arm.com>
6842 David Sherwood <david.sherwood@arm.com>
6844 * doc/invoke.texi (-msve-vector-bits=): Document new option.
6845 (sve): Document new AArch64 extension.
6846 * doc/md.texi (w): Extend the description of the AArch64
6847 constraint to include SVE vectors.
6848 (Upl, Upa): Document new AArch64 predicate constraints.
6849 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
6851 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
6852 (msve-vector-bits=): New option.
6853 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
6854 SVE when these are disabled.
6855 (sve): New extension.
6856 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
6857 modes. Adjust their number of units based on aarch64_sve_vg.
6858 (MAX_BITSIZE_MODE_ANY_MODE): Define.
6859 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
6860 aarch64_addr_query_type.
6861 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
6862 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
6863 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
6864 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
6865 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
6866 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
6867 (aarch64_simd_imm_zero_p): Delete.
6868 (aarch64_check_zero_based_sve_index_immediate): Declare.
6869 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6870 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6871 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
6872 (aarch64_sve_float_mul_immediate_p): Likewise.
6873 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6875 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
6876 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
6877 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
6878 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
6879 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
6880 (aarch64_regmode_natural_size): Likewise.
6881 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
6882 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
6884 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
6885 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
6886 for VG and the SVE predicate registers.
6887 (V_ALIASES): Add a "z"-prefixed alias.
6888 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
6889 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
6890 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
6891 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
6892 (REG_CLASS_NAMES): Add entries for them.
6893 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
6894 and the predicate registers.
6895 (aarch64_sve_vg): Declare.
6896 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
6897 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
6898 (REGMODE_NATURAL_SIZE): Define.
6899 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
6901 * config/aarch64/aarch64.c: Include cfgrtl.h.
6902 (simd_immediate_info): Add a constructor for series vectors,
6903 and an associated step field.
6904 (aarch64_sve_vg): New variable.
6905 (aarch64_dbx_register_number): Handle VG and the predicate registers.
6906 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
6907 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
6908 (VEC_ANY_DATA, VEC_STRUCT): New constants.
6909 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
6910 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
6911 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
6912 (aarch64_get_mask_mode): New functions.
6913 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
6914 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
6915 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
6916 predicate modes and predicate registers. Explicitly restrict
6917 GPRs to modes of 16 bytes or smaller. Only allow FP registers
6918 to store a vector mode if it is recognized by
6919 aarch64_classify_vector_mode.
6920 (aarch64_regmode_natural_size): New function.
6921 (aarch64_hard_regno_caller_save_mode): Return the original mode
6923 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
6924 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
6925 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
6926 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
6928 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
6929 does not overlap dest if the function is frame-related. Handle
6931 (aarch64_split_add_offset): New function.
6932 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
6933 them aarch64_add_offset.
6934 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
6935 and update call to aarch64_sub_sp.
6936 (aarch64_add_cfa_expression): New function.
6937 (aarch64_expand_prologue): Pass extra temporary registers to the
6938 functions above. Handle the case in which we need to emit new
6939 DW_CFA_expressions for registers that were originally saved
6940 relative to the stack pointer, but now have to be expressed
6941 relative to the frame pointer.
6942 (aarch64_output_mi_thunk): Pass extra temporary registers to the
6944 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
6945 IP0 and IP1 values for SVE frames.
6946 (aarch64_expand_vec_series): New function.
6947 (aarch64_expand_sve_widened_duplicate): Likewise.
6948 (aarch64_expand_sve_const_vector): Likewise.
6949 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
6950 Handle SVE constants. Use emit_move_insn to move a force_const_mem
6951 into the register, rather than emitting a SET directly.
6952 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
6953 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
6954 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
6955 (offset_9bit_signed_scaled_p): New functions.
6956 (aarch64_replicate_bitmask_imm): New function.
6957 (aarch64_bitmask_imm): Use it.
6958 (aarch64_cannot_force_const_mem): Reject expressions involving
6959 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
6960 (aarch64_classify_index): Handle SVE indices, by requiring
6961 a plain register index with a scale that matches the element size.
6962 (aarch64_classify_address): Handle SVE addresses. Assert that
6963 the mode of the address is VOIDmode or an integer mode.
6964 Update call to aarch64_classify_symbol.
6965 (aarch64_classify_symbolic_expression): Update call to
6966 aarch64_classify_symbol.
6967 (aarch64_const_vec_all_in_range_p): New function.
6968 (aarch64_print_vector_float_operand): Likewise.
6969 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
6970 "vN" for FP registers with SVE modes. Handle (const ...) vectors
6971 and the FP immediates 1.0 and 0.5.
6972 (aarch64_print_address_internal): Handle SVE addresses.
6973 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
6974 (aarch64_regno_regclass): Handle predicate registers.
6975 (aarch64_secondary_reload): Handle big-endian reloads of SVE
6977 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
6978 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
6979 (aarch64_convert_sve_vector_bits): New function.
6980 (aarch64_override_options): Use it to handle -msve-vector-bits=.
6981 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
6983 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
6984 Handle SVE vector and predicate modes. Accept VL-based constants
6985 that need only one temporary register, and VL offsets that require
6986 no temporary registers.
6987 (aarch64_conditional_register_usage): Mark the predicate registers
6988 as fixed if SVE isn't available.
6989 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
6990 Return true for SVE vector and predicate modes.
6991 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
6992 rather than an unsigned int. Handle SVE modes.
6993 (aarch64_preferred_simd_mode): Update call accordingly. Handle
6995 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
6997 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
6998 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
6999 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7000 (aarch64_sve_float_mul_immediate_p): New functions.
7001 (aarch64_sve_valid_immediate): New function.
7002 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7003 Explicitly reject structure modes. Check for INDEX constants.
7004 Handle PTRUE and PFALSE constants.
7005 (aarch64_check_zero_based_sve_index_immediate): New function.
7006 (aarch64_simd_imm_zero_p): Delete.
7007 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7008 vector modes. Accept constants in the range of CNT[BHWD].
7009 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7010 ask for an Advanced SIMD mode.
7011 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7012 (aarch64_simd_vector_alignment): Handle SVE predicates.
7013 (aarch64_vectorize_preferred_vector_alignment): New function.
7014 (aarch64_simd_vector_alignment_reachable): Use it instead of
7016 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7017 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7019 (MAX_VECT_LEN): Delete.
7020 (expand_vec_perm_d): Add a vec_flags field.
7021 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7022 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7023 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7025 (aarch64_evpc_rev): Rename to...
7026 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7027 (aarch64_evpc_rev_global): New function.
7028 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7029 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7031 (aarch64_evpc_sve_tbl): New function.
7032 (aarch64_expand_vec_perm_const_1): Update after rename of
7033 aarch64_evpc_rev. Handle SVE permutes too, trying
7034 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7035 than aarch64_evpc_tbl.
7036 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7037 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7038 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7039 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7040 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7041 (aarch64_expand_sve_vcond): New functions.
7042 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7043 of aarch64_vector_mode_p.
7044 (aarch64_dwarf_poly_indeterminate_value): New function.
7045 (aarch64_compute_pressure_classes): Likewise.
7046 (aarch64_can_change_mode_class): Likewise.
7047 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7048 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7049 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7050 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7051 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7052 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7053 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7054 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7056 (Dn, Dl, Dr): Accept const as well as const_vector.
7057 (Dz): Likewise. Compare against CONST0_RTX.
7058 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7059 of "vector" where appropriate.
7060 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7061 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7062 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7063 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7064 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7065 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7066 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7067 (v_int_equiv): Extend to SVE modes.
7068 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7070 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7071 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7072 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7073 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7074 (SVE_COND_FP_CMP): New int iterators.
7075 (perm_hilo): Handle the new unpack unspecs.
7076 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7078 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7079 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7080 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7081 (aarch64_equality_operator, aarch64_constant_vector_operand)
7082 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7083 (aarch64_sve_nonimmediate_operand): Likewise.
7084 (aarch64_sve_general_operand): Likewise.
7085 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7086 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7087 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7088 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7089 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7090 (aarch64_sve_float_arith_immediate): Likewise.
7091 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7092 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7093 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7094 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7095 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7096 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7097 (aarch64_sve_float_arith_operand): Likewise.
7098 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7099 (aarch64_sve_float_mul_operand): Likewise.
7100 (aarch64_sve_vec_perm_operand): Likewise.
7101 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7102 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7103 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7104 as well as const_vector.
7105 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7106 in file. Use CONST0_RTX and CONSTM1_RTX.
7107 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7108 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7109 Use aarch64_simd_imm_zero.
7110 * config/aarch64/aarch64-sve.md: New file.
7111 * config/aarch64/aarch64.md: Include it.
7112 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7113 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7114 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7115 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7116 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7117 (sve): New attribute.
7118 (enabled): Disable instructions with the sve attribute unless
7120 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7121 aarch64_expand_mov_immediate.
7122 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7123 CNT[BHSD] immediates.
7124 (movti): Split CONST_POLY_INT moves into two halves.
7125 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7126 Split additions that need a temporary here if the destination
7127 is the stack pointer.
7128 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7129 (*add<mode>3_poly_1): New instruction.
7130 (set_clobber_cc): New expander.
7132 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7134 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7135 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7136 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7137 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7138 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7139 Change innermode from fixed_mode_size to machine_mode.
7140 (simplify_subreg): Update call accordingly. Handle a constant-sized
7141 subreg of a variable-length CONST_VECTOR.
7143 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7144 Alan Hayward <alan.hayward@arm.com>
7145 David Sherwood <david.sherwood@arm.com>
7147 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7148 (add_offset_to_base): New function, split out from...
7149 (create_mem_ref): ...here. When handling a scale other than 1,
7150 check first whether the address is valid without the offset.
7151 Add it into the base if so, leaving the index and scale as-is.
7153 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7156 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7157 fold_for_warn before checking if arg2 is INTEGER_CST.
7159 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7161 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7162 (store_multiple_operation): Delete.
7163 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7164 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7165 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7166 guarded by TARGET_STRING.
7167 (rs6000_output_load_multiple): Delete.
7168 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7169 OPTION_MASK_STRING / TARGET_STRING handling.
7170 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7171 (const rs6000_opt_masks) <"string">: Change mask to 0.
7172 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7173 (MASK_STRING): Delete.
7174 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7176 (load_multiple): Delete.
7183 (store_multiple): Delete.
7190 (movmemsi_8reg): Delete.
7191 (corresponding unnamed define_insn): Delete.
7192 (movmemsi_6reg): Delete.
7193 (corresponding unnamed define_insn): Delete.
7194 (movmemsi_4reg): Delete.
7195 (corresponding unnamed define_insn): Delete.
7196 (movmemsi_2reg): Delete.
7197 (corresponding unnamed define_insn): Delete.
7198 (movmemsi_1reg): Delete.
7199 (corresponding unnamed define_insn): Delete.
7200 * config/rs6000/rs6000.opt (mno-string): New.
7201 (mstring): Replace by deprecation warning stub.
7202 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7204 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7206 * regrename.c (regrename_do_replace): If replacing the same
7207 reg multiple times, try to reuse last created gen_raw_REG.
7210 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7211 main to workaround a bug in GDB.
7213 2018-01-12 Tom de Vries <tom@codesourcery.com>
7216 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7218 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7220 PR rtl-optimization/80481
7221 * ira-color.c (get_cap_member): New function.
7222 (allocnos_conflict_by_live_ranges_p): Use it.
7223 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7224 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7226 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7229 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7230 (*saddl_se_1): Ditto.
7232 (*ssubl_se_1): Ditto.
7234 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7236 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7237 rather than wi::to_widest for DR_INITs.
7238 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7239 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7240 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7242 (vect_analyze_group_access_1): Note that here.
7244 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7246 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7247 polynomial type sizes.
7249 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7251 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7252 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7253 (gimple_add_tmp_var): Likewise.
7255 2018-01-12 Martin Liska <mliska@suse.cz>
7257 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7258 (gimple_alloc_sizes): Likewise.
7259 (dump_gimple_statistics): Use PRIu64 in printf format.
7260 * gimple.h: Change uint64_t to int.
7262 2018-01-12 Martin Liska <mliska@suse.cz>
7264 * tree-core.h: Use uint64_t instead of int.
7265 * tree.c (tree_node_counts): Likewise.
7266 (tree_node_sizes): Likewise.
7267 (dump_tree_statistics): Use PRIu64 in printf format.
7269 2018-01-12 Martin Liska <mliska@suse.cz>
7271 * Makefile.in: As qsort_chk is implemented in vec.c, add
7272 vec.o to linkage of gencfn-macros.
7273 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7274 passing the info to record_node_allocation_statistics.
7275 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7277 * ggc-common.c (struct ggc_usage): Add operator== and use
7278 it in operator< and compare function.
7279 * mem-stats.h (struct mem_usage): Likewise.
7280 * vec.c (struct vec_usage): Remove operator< and compare
7281 function. Can be simply inherited.
7283 2018-01-12 Martin Jambor <mjambor@suse.cz>
7286 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7287 * tree-ssa-math-opts.c: Include domwalk.h.
7288 (convert_mult_to_fma_1): New function.
7289 (fma_transformation_info): New type.
7290 (fma_deferring_state): Likewise.
7291 (cancel_fma_deferring): New function.
7292 (result_of_phi): Likewise.
7293 (last_fma_candidate_feeds_initial_phi): Likewise.
7294 (convert_mult_to_fma): Added deferring logic, split actual
7295 transformation to convert_mult_to_fma_1.
7296 (math_opts_dom_walker): New type.
7297 (math_opts_dom_walker::after_dom_children): New method, body moved
7298 here from pass_optimize_widening_mul::execute, added deferring logic
7300 (pass_optimize_widening_mul::execute): Moved most of code to
7301 math_opts_dom_walker::after_dom_children.
7302 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7303 * config/i386/i386.c (ix86_option_override_internal): Added
7304 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7306 2018-01-12 Richard Biener <rguenther@suse.de>
7309 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7310 inline instance vars.
7312 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7315 * config/rx/rx.c (rx_is_restricted_memory_address):
7318 2018-01-12 Richard Biener <rguenther@suse.de>
7320 PR tree-optimization/80846
7321 * target.def (split_reduction): New target hook.
7322 * targhooks.c (default_split_reduction): New function.
7323 * targhooks.h (default_split_reduction): Declare.
7324 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7325 target requests first reduce vectors by combining low and high
7327 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7328 (get_vectype_for_scalar_type_and_size): Export.
7329 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7330 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7331 * doc/tm.texi: Regenerate.
7332 * config/i386/i386.c (ix86_split_reduction): Implement
7333 TARGET_VECTORIZE_SPLIT_REDUCTION.
7335 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7338 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7339 in PIC mode except for TARGET_VXWORKS_RTP.
7340 * config/sparc/sparc.c: Include cfgrtl.h.
7341 (TARGET_INIT_PIC_REG): Define.
7342 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7343 (sparc_pic_register_p): New predicate.
7344 (sparc_legitimate_address_p): Use it.
7345 (sparc_legitimize_pic_address): Likewise.
7346 (sparc_delegitimize_address): Likewise.
7347 (sparc_mode_dependent_address_p): Likewise.
7348 (gen_load_pcrel_sym): Remove 4th parameter.
7349 (load_got_register): Adjust call to above. Remove obsolete stuff.
7350 (sparc_expand_prologue): Do not call load_got_register here.
7351 (sparc_flat_expand_prologue): Likewise.
7352 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7353 (sparc_use_pseudo_pic_reg): New function.
7354 (sparc_init_pic_reg): Likewise.
7355 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7356 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7358 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7360 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7361 Add item for branch_cost.
7363 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7365 PR rtl-optimization/83565
7366 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7367 not extend the result to a larger mode for rotate operations.
7368 (num_sign_bit_copies1): Likewise.
7370 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7373 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7375 Use values-Xc.o for -pedantic.
7376 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7378 2018-01-12 Martin Liska <mliska@suse.cz>
7381 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7383 (possible_polymorphic_call_targets): Use it.
7384 (ipa_devirt): Likewise.
7386 2018-01-12 Martin Liska <mliska@suse.cz>
7388 * profile-count.h (enum profile_quality): Use 0 as invalid
7389 enum value of profile_quality.
7391 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7393 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7394 -mext-string options.
7396 2018-01-12 Richard Biener <rguenther@suse.de>
7398 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7399 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7400 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7402 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7404 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7406 * configure.ac (--with-long-double-format): Add support for the
7407 configuration option to change the default long double format on
7409 * config.gcc (powerpc*-linux*-*): Likewise.
7410 * configure: Regenerate.
7411 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7412 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7413 used without modification.
7415 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7417 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7418 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7419 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7420 MISC_BUILTIN_SPEC_BARRIER.
7421 (rs6000_init_builtins): Likewise.
7422 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7424 (speculation_barrier): New define_insn.
7425 * doc/extend.texi: Document __builtin_speculation_barrier.
7427 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7430 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7431 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7432 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7434 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7435 integral modes instead of "ss" and "sd".
7436 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7437 vectors with 32-bit and 64-bit elements.
7438 (vecdupssescalarmodesuffix): New mode attribute.
7439 (vec_dup<mode>): Use it.
7441 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7444 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7445 frame if argument is passed on stack.
7447 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7450 * ree.c (combine_reaching_defs): Optimize also
7451 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7452 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7454 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7457 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7459 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7462 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7463 after they are computed.
7465 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7467 PR tree-optimization/83695
7468 * gimple-loop-linterchange.cc
7469 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7470 reset cached scev information after interchange.
7471 (pass_linterchange::execute): Remove call to scev_reset_htab.
7473 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7475 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7476 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7477 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7478 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7479 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7480 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7481 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7482 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7483 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7484 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7485 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7486 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7487 (V_lane_reg): Likewise.
7488 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7490 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7491 (vfmal_lane_low<mode>_intrinsic,
7492 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7493 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7494 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7495 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7496 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7497 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7499 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7501 * config/arm/arm-cpus.in (fp16fml): New feature.
7502 (ALL_SIMD): Add fp16fml.
7503 (armv8.2-a): Add fp16fml as an option.
7504 (armv8.3-a): Likewise.
7505 (armv8.4-a): Add fp16fml as part of fp16.
7506 * config/arm/arm.h (TARGET_FP16FML): Define.
7507 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7509 * config/arm/arm-modes.def (V2HF): Define.
7510 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7511 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7512 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7513 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7514 vfmsl_low, vfmsl_high): New set of builtins.
7515 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7516 (vfml_op): New code attribute.
7517 (VFMLHALVES): New int iterator.
7518 (VFML, VFMLSEL): New mode attributes.
7519 (V_reg): Define mapping for V2HF.
7520 (V_hi, V_lo): New mode attributes.
7521 (VF_constraint): Likewise.
7522 (vfml_half, vfml_half_selector): New int attributes.
7523 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7525 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7526 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7528 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7529 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7530 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7531 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7533 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7534 Document new effective target and option set.
7536 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7538 * config/arm/arm-cpus.in (armv8_4): New feature.
7539 (ARMv8_4a): New fgroup.
7540 (armv8.4-a): New arch.
7541 * config/arm/arm-tables.opt: Regenerate.
7542 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7543 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7544 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7545 Add matching rules for -march=armv8.4-a and extensions.
7546 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7548 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7551 * config/rx/rx.md (BW): New mode attribute.
7552 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7554 2018-01-11 Richard Biener <rguenther@suse.de>
7556 PR tree-optimization/83435
7557 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7558 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7559 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7561 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7562 Alan Hayward <alan.hayward@arm.com>
7563 David Sherwood <david.sherwood@arm.com>
7565 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7567 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7568 (aarch64_print_address_internal): Use it to check for a zero offset.
7570 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7571 Alan Hayward <alan.hayward@arm.com>
7572 David Sherwood <david.sherwood@arm.com>
7574 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7575 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7576 Return a poly_int64 rather than a HOST_WIDE_INT.
7577 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7578 rather than a HOST_WIDE_INT.
7579 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7580 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7581 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7582 final_offset from HOST_WIDE_INT to poly_int64.
7583 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7584 to_constant when getting the number of units in an Advanced SIMD
7586 (aarch64_builtin_vectorized_function): Check for a constant number
7588 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7590 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7591 attribute instead of GET_MODE_NUNITS.
7592 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7593 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7594 GET_MODE_SIZE for fixed-size registers.
7595 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7596 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7597 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7598 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7599 (aarch64_print_operand, aarch64_print_address_internal)
7600 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7601 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7602 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7603 Handle polynomial GET_MODE_SIZE.
7604 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7605 wider than SImode without modification.
7606 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7607 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7608 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7609 passing and returning SVE modes.
7610 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7611 rather than GEN_INT.
7612 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7613 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7614 (aarch64_allocate_and_probe_stack_space): Likewise.
7615 (aarch64_layout_frame): Cope with polynomial offsets.
7616 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7617 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7619 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7620 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7621 poly_int64 rather than a HOST_WIDE_INT.
7622 (aarch64_get_separate_components, aarch64_process_components)
7623 (aarch64_expand_prologue, aarch64_expand_epilogue)
7624 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7625 (aarch64_anchor_offset): New function, split out from...
7626 (aarch64_legitimize_address): ...here.
7627 (aarch64_builtin_vectorization_cost): Handle polynomial
7628 TYPE_VECTOR_SUBPARTS.
7629 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7631 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7632 number of elements from the PARALLEL rather than the mode.
7633 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7634 rather than GET_MODE_BITSIZE.
7635 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7636 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7637 (aarch64_expand_vec_perm_const_1): Handle polynomial
7638 d->perm.length () and d->perm elements.
7639 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7640 Apply to_constant to d->perm elements.
7641 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7642 polynomial CONST_VECTOR_NUNITS.
7643 (aarch64_move_pointer): Take amount as a poly_int64 rather
7645 (aarch64_progress_pointer): Avoid temporary variable.
7646 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7647 the mode attribute instead of GET_MODE.
7649 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7650 Alan Hayward <alan.hayward@arm.com>
7651 David Sherwood <david.sherwood@arm.com>
7653 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7654 x exists before using it.
7655 (aarch64_add_constant_internal): Rename to...
7656 (aarch64_add_offset_1): ...this. Replace regnum with separate
7657 src and dest rtxes. Handle the case in which they're different,
7658 including when the offset is zero. Replace scratchreg with an rtx.
7659 Use 2 additions if there is no spare register into which we can
7660 move a 16-bit constant.
7661 (aarch64_add_constant): Delete.
7662 (aarch64_add_offset): Replace reg with separate src and dest
7663 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
7664 Use aarch64_add_offset_1.
7665 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
7666 an rtx rather than an int. Take the delta as a poly_int64
7667 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
7668 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
7669 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
7670 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
7671 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
7673 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
7674 aarch64_add_constant.
7676 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7678 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
7679 Use scalar_float_mode.
7681 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7683 * config/aarch64/aarch64-simd.md
7684 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
7685 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
7686 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
7687 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
7688 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
7689 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
7690 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
7691 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
7692 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
7693 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
7695 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7698 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
7699 targ_options->x_arm_arch_string is non NULL.
7701 2018-01-11 Tamar Christina <tamar.christina@arm.com>
7703 * config/aarch64/aarch64.h
7704 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
7706 2018-01-11 Sudakshina Das <sudi.das@arm.com>
7709 * expmed.c (emit_store_flag_force): Swap if const op0
7710 and change VOIDmode to mode of op0.
7712 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7714 PR rtl-optimization/83761
7715 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
7716 than bytes to mode_for_size.
7718 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7721 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
7722 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
7725 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
7728 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
7729 when in layout mode.
7730 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
7731 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
7734 2018-01-10 Michael Collison <michael.collison@arm.com>
7736 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
7737 * config/aarch64/aarch64-option-extension.def: Add
7738 AARCH64_OPT_EXTENSION of 'fp16fml'.
7739 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7740 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
7741 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
7742 * config/aarch64/constraints.md (Ui7): New constraint.
7743 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
7744 (VFMLA_SEL_W): Ditto.
7747 (VFMLA16_LOW): New int iterator.
7748 (VFMLA16_HIGH): Ditto.
7749 (UNSPEC_FMLAL): New unspec.
7750 (UNSPEC_FMLSL): Ditto.
7751 (UNSPEC_FMLAL2): Ditto.
7752 (UNSPEC_FMLSL2): Ditto.
7753 (f16mac): New code attribute.
7754 * config/aarch64/aarch64-simd-builtins.def
7755 (aarch64_fmlal_lowv2sf): Ditto.
7756 (aarch64_fmlsl_lowv2sf): Ditto.
7757 (aarch64_fmlalq_lowv4sf): Ditto.
7758 (aarch64_fmlslq_lowv4sf): Ditto.
7759 (aarch64_fmlal_highv2sf): Ditto.
7760 (aarch64_fmlsl_highv2sf): Ditto.
7761 (aarch64_fmlalq_highv4sf): Ditto.
7762 (aarch64_fmlslq_highv4sf): Ditto.
7763 (aarch64_fmlal_lane_lowv2sf): Ditto.
7764 (aarch64_fmlsl_lane_lowv2sf): Ditto.
7765 (aarch64_fmlal_laneq_lowv2sf): Ditto.
7766 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
7767 (aarch64_fmlalq_lane_lowv4sf): Ditto.
7768 (aarch64_fmlsl_lane_lowv4sf): Ditto.
7769 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
7770 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
7771 (aarch64_fmlal_lane_highv2sf): Ditto.
7772 (aarch64_fmlsl_lane_highv2sf): Ditto.
7773 (aarch64_fmlal_laneq_highv2sf): Ditto.
7774 (aarch64_fmlsl_laneq_highv2sf): Ditto.
7775 (aarch64_fmlalq_lane_highv4sf): Ditto.
7776 (aarch64_fmlsl_lane_highv4sf): Ditto.
7777 (aarch64_fmlalq_laneq_highv4sf): Ditto.
7778 (aarch64_fmlsl_laneq_highv4sf): Ditto.
7779 * config/aarch64/aarch64-simd.md:
7780 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
7781 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7782 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
7783 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
7784 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
7785 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
7786 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
7787 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
7788 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
7789 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
7790 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
7791 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
7792 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
7793 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
7794 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
7795 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
7796 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
7797 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
7798 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
7799 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
7800 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
7801 (vfmlsl_low_u32): Ditto.
7802 (vfmlalq_low_u32): Ditto.
7803 (vfmlslq_low_u32): Ditto.
7804 (vfmlal_high_u32): Ditto.
7805 (vfmlsl_high_u32): Ditto.
7806 (vfmlalq_high_u32): Ditto.
7807 (vfmlslq_high_u32): Ditto.
7808 (vfmlal_lane_low_u32): Ditto.
7809 (vfmlsl_lane_low_u32): Ditto.
7810 (vfmlal_laneq_low_u32): Ditto.
7811 (vfmlsl_laneq_low_u32): Ditto.
7812 (vfmlalq_lane_low_u32): Ditto.
7813 (vfmlslq_lane_low_u32): Ditto.
7814 (vfmlalq_laneq_low_u32): Ditto.
7815 (vfmlslq_laneq_low_u32): Ditto.
7816 (vfmlal_lane_high_u32): Ditto.
7817 (vfmlsl_lane_high_u32): Ditto.
7818 (vfmlal_laneq_high_u32): Ditto.
7819 (vfmlsl_laneq_high_u32): Ditto.
7820 (vfmlalq_lane_high_u32): Ditto.
7821 (vfmlslq_lane_high_u32): Ditto.
7822 (vfmlalq_laneq_high_u32): Ditto.
7823 (vfmlslq_laneq_high_u32): Ditto.
7824 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
7825 (AARCH64_FL_FOR_ARCH8_4): New.
7826 (AARCH64_ISA_F16FML): New ISA flag.
7827 (TARGET_F16FML): New feature flag for fp16fml.
7828 (doc/invoke.texi): Document new fp16fml option.
7830 2018-01-10 Michael Collison <michael.collison@arm.com>
7832 * config/aarch64/aarch64-builtins.c:
7833 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
7834 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7835 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
7836 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
7837 (AARCH64_ISA_SHA3): New ISA flag.
7838 (TARGET_SHA3): New feature flag for sha3.
7839 * config/aarch64/iterators.md (sha512_op): New int attribute.
7840 (CRYPTO_SHA512): New int iterator.
7841 (UNSPEC_SHA512H): New unspec.
7842 (UNSPEC_SHA512H2): Ditto.
7843 (UNSPEC_SHA512SU0): Ditto.
7844 (UNSPEC_SHA512SU1): Ditto.
7845 * config/aarch64/aarch64-simd-builtins.def
7846 (aarch64_crypto_sha512hqv2di): New builtin.
7847 (aarch64_crypto_sha512h2qv2di): Ditto.
7848 (aarch64_crypto_sha512su0qv2di): Ditto.
7849 (aarch64_crypto_sha512su1qv2di): Ditto.
7850 (aarch64_eor3qv8hi): Ditto.
7851 (aarch64_rax1qv2di): Ditto.
7852 (aarch64_xarqv2di): Ditto.
7853 (aarch64_bcaxqv8hi): Ditto.
7854 * config/aarch64/aarch64-simd.md:
7855 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
7856 (aarch64_crypto_sha512su0qv2di): Ditto.
7857 (aarch64_crypto_sha512su1qv2di): Ditto.
7858 (aarch64_eor3qv8hi): Ditto.
7859 (aarch64_rax1qv2di): Ditto.
7860 (aarch64_xarqv2di): Ditto.
7861 (aarch64_bcaxqv8hi): Ditto.
7862 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
7863 (vsha512h2q_u64): Ditto.
7864 (vsha512su0q_u64): Ditto.
7865 (vsha512su1q_u64): Ditto.
7866 (veor3q_u16): Ditto.
7867 (vrax1q_u64): Ditto.
7869 (vbcaxq_u16): Ditto.
7870 * config/arm/types.md (crypto_sha512): New type attribute.
7871 (crypto_sha3): Ditto.
7872 (doc/invoke.texi): Document new sha3 option.
7874 2018-01-10 Michael Collison <michael.collison@arm.com>
7876 * config/aarch64/aarch64-builtins.c:
7877 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
7878 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7879 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
7880 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
7881 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
7882 (AARCH64_ISA_SM4): New ISA flag.
7883 (TARGET_SM4): New feature flag for sm4.
7884 * config/aarch64/aarch64-simd-builtins.def
7885 (aarch64_sm3ss1qv4si): Ditto.
7886 (aarch64_sm3tt1aq4si): Ditto.
7887 (aarch64_sm3tt1bq4si): Ditto.
7888 (aarch64_sm3tt2aq4si): Ditto.
7889 (aarch64_sm3tt2bq4si): Ditto.
7890 (aarch64_sm3partw1qv4si): Ditto.
7891 (aarch64_sm3partw2qv4si): Ditto.
7892 (aarch64_sm4eqv4si): Ditto.
7893 (aarch64_sm4ekeyqv4si): Ditto.
7894 * config/aarch64/aarch64-simd.md:
7895 (aarch64_sm3ss1qv4si): Ditto.
7896 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
7897 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
7898 (aarch64_sm4eqv4si): Ditto.
7899 (aarch64_sm4ekeyqv4si): Ditto.
7900 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
7901 (sm3part_op): Ditto.
7902 (CRYPTO_SM3TT): Ditto.
7903 (CRYPTO_SM3PART): Ditto.
7904 (UNSPEC_SM3SS1): New unspec.
7905 (UNSPEC_SM3TT1A): Ditto.
7906 (UNSPEC_SM3TT1B): Ditto.
7907 (UNSPEC_SM3TT2A): Ditto.
7908 (UNSPEC_SM3TT2B): Ditto.
7909 (UNSPEC_SM3PARTW1): Ditto.
7910 (UNSPEC_SM3PARTW2): Ditto.
7911 (UNSPEC_SM4E): Ditto.
7912 (UNSPEC_SM4EKEY): Ditto.
7913 * config/aarch64/constraints.md (Ui2): New constraint.
7914 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
7915 * config/arm/types.md (crypto_sm3): New type attribute.
7916 (crypto_sm4): Ditto.
7917 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
7918 (vsm3tt1aq_u32): Ditto.
7919 (vsm3tt1bq_u32): Ditto.
7920 (vsm3tt2aq_u32): Ditto.
7921 (vsm3tt2bq_u32): Ditto.
7922 (vsm3partw1q_u32): Ditto.
7923 (vsm3partw2q_u32): Ditto.
7924 (vsm4eq_u32): Ditto.
7925 (vsm4ekeyq_u32): Ditto.
7926 (doc/invoke.texi): Document new sm4 option.
7928 2018-01-10 Michael Collison <michael.collison@arm.com>
7930 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
7931 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
7932 (AARCH64_FL_FOR_ARCH8_4): New.
7933 (AARCH64_FL_V8_4): New flag.
7934 (doc/invoke.texi): Document new armv8.4-a option.
7936 2018-01-10 Michael Collison <michael.collison@arm.com>
7938 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
7939 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
7940 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
7941 * config/aarch64/aarch64-option-extension.def: Add
7942 AARCH64_OPT_EXTENSION of 'sha2'.
7943 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
7944 (crypto): Disable sha2 and aes if crypto disabled.
7945 (crypto): Enable aes and sha2 if enabled.
7946 (simd): Disable sha2 and aes if simd disabled.
7947 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
7949 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
7950 (TARGET_SHA2): New feature flag for sha2.
7951 (TARGET_AES): New feature flag for aes.
7952 * config/aarch64/aarch64-simd.md:
7953 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
7954 conditional on TARGET_AES.
7955 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
7956 (aarch64_crypto_sha1hsi): Make pattern conditional
7958 (aarch64_crypto_sha1hv4si): Ditto.
7959 (aarch64_be_crypto_sha1hv4si): Ditto.
7960 (aarch64_crypto_sha1su1v4si): Ditto.
7961 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
7962 (aarch64_crypto_sha1su0v4si): Ditto.
7963 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
7964 (aarch64_crypto_sha256su0v4si): Ditto.
7965 (aarch64_crypto_sha256su1v4si): Ditto.
7966 (doc/invoke.texi): Document new aes and sha2 options.
7968 2018-01-10 Martin Sebor <msebor@redhat.com>
7970 PR tree-optimization/83781
7971 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
7974 2018-01-11 Martin Sebor <msebor@gmail.com>
7975 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7977 PR tree-optimization/83501
7978 PR tree-optimization/81703
7980 * tree-ssa-strlen.c (get_string_cst): Rename...
7981 (get_string_len): ...to this. Handle global constants.
7982 (handle_char_store): Adjust.
7984 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
7985 Jim Wilson <jimw@sifive.com>
7987 * config/riscv/riscv-protos.h (riscv_output_return): New.
7988 * config/riscv/riscv.c (struct machine_function): New naked_p field.
7989 (riscv_attribute_table, riscv_output_return),
7990 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
7991 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
7992 (riscv_compute_frame_info): Only compute frame->mask if not a naked
7994 (riscv_expand_prologue): Add early return for naked function.
7995 (riscv_expand_epilogue): Likewise.
7996 (riscv_function_ok_for_sibcall): Return false for naked function.
7997 (riscv_set_current_function): New.
7998 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
7999 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8000 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8001 * doc/extend.texi (RISC-V Function Attributes): New.
8003 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8005 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8006 check for 128-bit long double before checking TCmode.
8007 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8008 128-bit long doubles before checking TFmode or TCmode.
8009 (FLOAT128_IBM_P): Likewise.
8011 2018-01-10 Martin Sebor <msebor@redhat.com>
8013 PR tree-optimization/83671
8014 * builtins.c (c_strlen): Unconditionally return zero for the empty
8016 Use -Warray-bounds for warnings.
8017 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8018 for non-constant array indices with COMPONENT_REF, arrays of
8019 arrays, and pointers to arrays.
8020 (gimple_fold_builtin_strlen): Determine and set length range for
8021 non-constant character arrays.
8023 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8026 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8029 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8031 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8033 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8036 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8037 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8038 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8039 indexed_or_indirect_operand predicate.
8040 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8041 (*vsx_le_perm_load_v8hi): Likewise.
8042 (*vsx_le_perm_load_v16qi): Likewise.
8043 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8044 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8045 (*vsx_le_perm_store_v8hi): Likewise.
8046 (*vsx_le_perm_store_v16qi): Likewise.
8047 (eight unnamed splitters): Likewise.
8049 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8051 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8052 * config/rs6000/emmintrin.h: Likewise.
8053 * config/rs6000/mmintrin.h: Likewise.
8054 * config/rs6000/xmmintrin.h: Likewise.
8056 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8059 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8061 * tree.c (tree_nop_conversion): Return true for location wrapper
8063 (maybe_wrap_with_location): New function.
8064 (selftest::check_strip_nops): New function.
8065 (selftest::test_location_wrappers): New function.
8066 (selftest::tree_c_tests): Call it.
8067 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8068 (maybe_wrap_with_location): New decl.
8069 (EXPR_LOCATION_WRAPPER_P): New macro.
8070 (location_wrapper_p): New inline function.
8071 (tree_strip_any_location_wrapper): New inline function.
8073 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8076 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8077 stack_realign_offset for the largest alignment of stack slot
8079 (ix86_find_max_used_stack_alignment): New function.
8080 (ix86_finalize_stack_frame_flags): Use it. Set
8081 max_used_stack_alignment if we don't realign stack.
8082 * config/i386/i386.h (machine_function): Add
8083 max_used_stack_alignment.
8085 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8087 * config/arm/arm.opt (-mbranch-cost): New option.
8088 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8091 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8094 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8095 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8097 2018-01-10 Richard Biener <rguenther@suse.de>
8100 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8101 early out so it also covers the case where we have a non-NULL
8104 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8106 PR tree-optimization/83753
8107 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8108 for non-strided grouped accesses if the number of elements is 1.
8110 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8113 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8114 * i386.h (TARGET_USE_GATHER): Define.
8115 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8117 2018-01-10 Martin Liska <mliska@suse.cz>
8120 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8121 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8123 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8124 CLEANUP_NO_PARTITIONING is not set.
8126 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8128 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8129 for vectors, as a partial revert of r254296.
8130 * rtl.h (const_vec_p): Delete.
8131 (const_vec_duplicate_p): Don't test for vector CONSTs.
8132 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8133 * expmed.c (make_tree): Likewise.
8136 * common.md (E, F): Use CONSTANT_P instead of checking for
8138 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8139 checking for CONST_VECTOR.
8141 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8144 * predict.c (force_edge_cold): Handle in more sane way edges
8147 2018-01-09 Carl Love <cel@us.ibm.com>
8149 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8151 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8152 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8153 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8154 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8155 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8156 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8157 * config/rs6000/rs6000-protos.h: Add extern defition for
8158 rs6000_generate_float2_double_code.
8159 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8161 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8162 (float2_v2df): Add define_expand.
8164 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8167 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8168 op_mode in the force_to_mode call.
8170 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8172 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8173 instead of checking each element individually.
8174 (aarch64_evpc_uzp): Likewise.
8175 (aarch64_evpc_zip): Likewise.
8176 (aarch64_evpc_ext): Likewise.
8177 (aarch64_evpc_rev): Likewise.
8178 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8179 instead of checking each element individually. Return true without
8181 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8182 whether all selected elements come from the same input, instead of
8183 checking each element individually. Remove calls to gen_rtx_REG,
8184 start_sequence and end_sequence and instead assert that no rtl is
8187 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8189 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8190 order of HIGH and CONST checks.
8192 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8194 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8195 if the destination isn't an SSA_NAME.
8197 2018-01-09 Richard Biener <rguenther@suse.de>
8199 PR tree-optimization/83668
8200 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8202 (canonicalize_loop_form): ... here, renamed from ...
8203 (canonicalize_loop_closed_ssa_form): ... this and amended to
8204 swap successor edges for loop exit blocks to make us use
8205 the RPO order we need for initial schedule generation.
8207 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8209 PR tree-optimization/64811
8210 * match.pd: When optimizing comparisons with Inf, avoid
8211 introducing or losing exceptions from comparisons with NaN.
8213 2018-01-09 Martin Liska <mliska@suse.cz>
8216 * asan.c (shadow_mem_size): Add gcc_assert.
8218 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8220 Don't save registers in main().
8223 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8224 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8225 * config/avr/avr.c (avr_set_current_function): Don't error if
8226 naked, OS_task or OS_main are specified at the same time.
8227 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8229 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8231 * common/config/avr/avr-common.c (avr_option_optimization_table):
8232 Switch on -mmain-is-OS_task for optimizing compilations.
8234 2018-01-09 Richard Biener <rguenther@suse.de>
8236 PR tree-optimization/83572
8237 * graphite.c: Include cfganal.h.
8238 (graphite_transform_loops): Connect infinite loops to exit
8239 and remove fake edges at the end.
8241 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8243 * ipa-inline.c (edge_badness): Revert accidental checkin.
8245 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8248 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8249 symbols; not inline clones.
8251 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8254 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8255 hard registers. Formatting fixes.
8257 PR preprocessor/83722
8258 * gcc.c (try_generate_repro): Pass
8259 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8260 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8263 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8264 Kito Cheng <kito.cheng@gmail.com>
8266 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8267 (riscv_leaf_function_p): Delete.
8268 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8270 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8272 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8274 (do_ifelse): New function.
8275 (do_isel): New function.
8276 (do_sub3): New function.
8277 (do_add3): New function.
8278 (do_load_mask_compare): New function.
8279 (do_overlap_load_compare): New function.
8280 (expand_compare_loop): New function.
8281 (expand_block_compare): Call expand_compare_loop() when appropriate.
8282 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8284 (-mblock-compare-inline-loop-limit): New option.
8286 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8289 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8290 Reverse order of second and third operands in first alternative.
8291 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8292 of first and second elements in UNSPEC_VPERMR vector.
8293 (altivec_expand_vec_perm_le): Likewise.
8295 2018-01-08 Jeff Law <law@redhat.com>
8297 PR rtl-optimizatin/81308
8298 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8299 (process_switch): If group_case_labels makes a change, then set
8301 (pass_convert_switch::execute): If a switch is converted, then
8302 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8304 PR rtl-optimization/81308
8305 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8308 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8310 PR target/83663 - Revert r255946
8311 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8312 generation for cases where splatting a value is not useful.
8313 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8314 across a vec_duplicate and a paradoxical subreg forming a vector
8315 mode to a vec_concat.
8317 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8319 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8320 -march=armv8.3-a variants.
8321 * config/arm/t-multilib: Likewise.
8322 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8324 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8326 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8328 (cceq_ior_compare_complement): Give it a name so I can use it, and
8329 change boolean_or_operator predicate to boolean_operator so it can
8330 be used to generate a crand.
8331 (eqne): New code iterator.
8332 (bd/bd_neg): New code_attrs.
8333 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8334 a single define_insn.
8335 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8336 decrement (bdnzt/bdnzf/bdzt/bdzf).
8337 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8338 with the new names of the branch decrement patterns, and added the
8339 names of the branch decrement conditional patterns.
8341 2018-01-08 Richard Biener <rguenther@suse.de>
8343 PR tree-optimization/83563
8344 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8347 2018-01-08 Richard Biener <rguenther@suse.de>
8350 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8352 2018-01-08 Richard Biener <rguenther@suse.de>
8354 PR tree-optimization/83685
8355 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8356 references to abnormals.
8358 2018-01-08 Richard Biener <rguenther@suse.de>
8361 * dwarf2out.c (output_indirect_strings): Handle empty
8362 skeleton_debug_str_hash.
8363 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8365 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8367 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8368 (emit_store_direct): Likewise.
8369 (arc_trampoline_adjust_address): Likewise.
8370 (arc_asm_trampoline_template): New function.
8371 (arc_initialize_trampoline): Use asm_trampoline_template.
8372 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8373 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8374 * config/arc/arc.md (flush_icache): Delete pattern.
8376 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8378 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8379 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8382 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8385 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8386 by not USED_FOR_TARGET.
8387 (make_pass_resolve_sw_modes): Likewise.
8389 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8391 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8394 2018-01-08 Richard Biener <rguenther@suse.de>
8397 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8399 2018-01-08 Richard Biener <rguenther@suse.de>
8402 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8404 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8407 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8408 basic blocks with a small number of successors.
8409 (convert_control_dep_chain_into_preds): Improve handling of
8411 (dump_predicates): Split apart into...
8412 (dump_pred_chain): ...here...
8413 (dump_pred_info): ...and here.
8414 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8415 (can_chain_union_be_invalidated_p): Improve check for invalidation
8417 (uninit_uses_cannot_happen): Avoid unnecessary if
8418 convert_control_dep_chain_into_preds yielded nothing.
8420 2018-01-06 Martin Sebor <msebor@redhat.com>
8422 PR tree-optimization/83640
8423 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8424 subtracting negative offset from size.
8425 (builtin_access::overlap): Adjust offset bounds of the access to fall
8426 within the size of the object if possible.
8428 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8430 PR rtl-optimization/83699
8431 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8432 extract_bit_field_as_subreg to cases in which the extracted
8433 value is also a vector.
8435 * lra-constraints.c (process_alt_operands): Test for the equivalence
8436 substitutions when detecting a possible reload cycle.
8438 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8441 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8442 by default if flag_selective_schedling{,2}. Formatting fixes.
8444 PR rtl-optimization/83682
8445 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8446 if it has non-VECTOR_MODE element mode.
8447 (vec_duplicate_p): Likewise.
8450 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8451 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8453 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8456 * config/i386/i386-builtin.def
8457 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8458 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8459 Require also OPTION_MASK_ISA_AVX512F in addition to
8460 OPTION_MASK_ISA_GFNI.
8461 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8462 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8463 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8464 to OPTION_MASK_ISA_GFNI.
8465 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8466 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8467 OPTION_MASK_ISA_AVX512BW.
8468 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8469 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8470 addition to OPTION_MASK_ISA_GFNI.
8471 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8472 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8473 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8474 to OPTION_MASK_ISA_GFNI.
8475 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8476 a requirement for all ISAs rather than any of them with a few
8478 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8480 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8481 bitmasks to be enabled with 3 exceptions, instead of requiring any
8482 enabled ISA with lots of exceptions.
8483 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8484 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8485 Change avx512bw in isa attribute to avx512f.
8486 * config/i386/sgxintrin.h: Add license boilerplate.
8487 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8488 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8489 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8490 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8492 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8493 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8494 temporarily sse2 rather than sse if not enabled already.
8497 * config/i386/sse.md (VI248_VLBW): Rename to ...
8498 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8499 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8500 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8501 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8502 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8503 mode iterator instead of VI248_VLBW.
8505 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8507 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8508 (record_modified): Skip clobbers; add debug output.
8509 (param_change_prob): Use sreal frequencies.
8511 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8513 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8514 punt for user-aligned variables.
8516 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8518 * tree-chrec.c (chrec_contains_symbols): Return true for
8521 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8524 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8525 of (x|y) == x for BICS pattern.
8527 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8529 PR tree-optimization/83605
8530 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8531 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8534 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8536 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8537 * config/epiphany/rtems.h: New file.
8539 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8540 Uros Bizjak <ubizjak@gmail.com>
8543 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8544 QIreg_operand instead of register_operand predicate.
8545 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8546 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8547 comments instead of -fmitigate[-_]rop.
8549 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8552 * cgraphunit.c (symbol_table::compile): Switch to text_section
8553 before calling assembly_start debug hook.
8554 * run-rtl-passes.c (run_rtl_passes): Likewise.
8557 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8559 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8560 range_int_cst_p rather than !symbolic_range_p before calling
8561 extract_range_from_multiplicative_op_1.
8563 2018-01-04 Jeff Law <law@redhat.com>
8565 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8566 redundant test in assertion.
8568 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8570 * doc/rtl.texi: Document machine_mode wrapper classes.
8572 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8574 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8577 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8579 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8580 the VEC_PERM_EXPR fold to fail.
8582 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8585 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8586 to switched_sections.
8588 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8591 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8594 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8597 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8598 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8600 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8603 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8604 is BLKmode and bitpos not zero or mode change is needed.
8606 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8609 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8612 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8615 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8616 instead of MULT rtx. Update all corresponding splitters.
8618 (*ssub<modesuffix>): Ditto.
8620 (*cmp_sadd_di): Update split patterns.
8621 (*cmp_sadd_si): Ditto.
8622 (*cmp_sadd_sidi): Ditto.
8623 (*cmp_ssub_di): Ditto.
8624 (*cmp_ssub_si): Ditto.
8625 (*cmp_ssub_sidi): Ditto.
8626 * config/alpha/predicates.md (const23_operand): New predicate.
8627 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8628 Look for ASHIFT, not MULT inner operand.
8629 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8631 2018-01-04 Martin Liska <mliska@suse.cz>
8633 PR gcov-profile/83669
8634 * gcov.c (output_intermediate_file): Add version to intermediate
8636 * doc/gcov.texi: Document new field 'version' in intermediate
8637 file format. Fix location of '-k' option of gcov command.
8639 2018-01-04 Martin Liska <mliska@suse.cz>
8642 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8644 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8646 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8648 2018-01-03 Martin Sebor <msebor@redhat.com>
8650 PR tree-optimization/83655
8651 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8652 checking calls with invalid arguments.
8654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8656 * tree-vect-stmts.c (vect_get_store_rhs): New function.
8657 (vectorizable_mask_load_store): Delete.
8658 (vectorizable_call): Return false for masked loads and stores.
8659 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
8660 instead of gimple_assign_rhs1.
8661 (vectorizable_load): Handle IFN_MASK_LOAD.
8662 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
8664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8666 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
8668 (vectorizable_mask_load_store): ...here.
8669 (vectorizable_load): ...and here.
8671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8673 * tree-vect-stmts.c (vect_build_all_ones_mask)
8674 (vect_build_zero_merge_argument): New functions, split out from...
8675 (vectorizable_load): ...here.
8677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8679 * tree-vect-stmts.c (vect_check_store_rhs): New function,
8681 (vectorizable_mask_load_store): ...here.
8682 (vectorizable_store): ...and here.
8684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8686 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
8688 (vectorizable_mask_load_store): ...here.
8690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8692 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
8693 (vect_model_store_cost): Take a vec_load_store_type instead of a
8695 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
8696 (vect_model_store_cost): Take a vec_load_store_type instead of a
8698 (vectorizable_mask_load_store): Update accordingly.
8699 (vectorizable_store): Likewise.
8700 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
8702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8704 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
8705 IFN_MASK_LOAD calls here rather than...
8706 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
8708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8709 Alan Hayward <alan.hayward@arm.com>
8710 David Sherwood <david.sherwood@arm.com>
8712 * expmed.c (extract_bit_field_1): For vector extracts,
8713 fall back to extract_bit_field_as_subreg if vec_extract
8716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8717 Alan Hayward <alan.hayward@arm.com>
8718 David Sherwood <david.sherwood@arm.com>
8720 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
8721 they are variable or constant sized.
8722 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
8723 slots for constant-sized data.
8725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8726 Alan Hayward <alan.hayward@arm.com>
8727 David Sherwood <david.sherwood@arm.com>
8729 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
8730 handling COND_EXPRs with boolean comparisons, try to find a better
8731 basis for the mask type than the boolean itself.
8733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8735 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
8736 is calculated and how it can be overridden.
8737 * genmodes.c (max_bitsize_mode_any_mode): New variable.
8738 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
8740 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
8743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8744 Alan Hayward <alan.hayward@arm.com>
8745 David Sherwood <david.sherwood@arm.com>
8747 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
8748 Remove the mode argument.
8749 (aarch64_simd_valid_immediate): Remove the mode and inverse
8751 * config/aarch64/iterators.md (bitsize): New iterator.
8752 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
8753 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
8754 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
8755 aarch64_simd_valid_immediate.
8756 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
8757 (aarch64_reg_or_bic_imm): Likewise.
8758 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
8759 with an insn_type enum and msl with a modifier_type enum.
8760 Replace element_width with a scalar_mode. Change the shift
8761 to unsigned int. Add constructors for scalar_float_mode and
8762 scalar_int_mode elements.
8763 (aarch64_vect_float_const_representable_p): Delete.
8764 (aarch64_can_const_movi_rtx_p)
8765 (aarch64_simd_scalar_immediate_valid_for_move)
8766 (aarch64_simd_make_constant): Update call to
8767 aarch64_simd_valid_immediate.
8768 (aarch64_advsimd_valid_immediate_hs): New function.
8769 (aarch64_advsimd_valid_immediate): Likewise.
8770 (aarch64_simd_valid_immediate): Remove mode and inverse
8771 arguments. Rewrite to use the above. Use const_vec_duplicate_p
8772 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
8773 and aarch64_float_const_representable_p on the result.
8774 (aarch64_output_simd_mov_immediate): Remove mode argument.
8775 Update call to aarch64_simd_valid_immediate and use of
8776 simd_immediate_info.
8777 (aarch64_output_scalar_simd_mov_immediate): Update call
8780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8781 Alan Hayward <alan.hayward@arm.com>
8782 David Sherwood <david.sherwood@arm.com>
8784 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
8785 (mode_nunits): Likewise CONST_MODE_NUNITS.
8786 * machmode.def (ADJUST_NUNITS): Document.
8787 * genmodes.c (mode_data::need_nunits_adj): New field.
8788 (blank_mode): Update accordingly.
8789 (adj_nunits): New variable.
8790 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
8792 (emit_mode_size_inline): Set need_bytesize_adj for all modes
8793 listed in adj_nunits.
8794 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
8795 listed in adj_nunits. Don't emit case statements for such modes.
8796 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
8797 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
8798 nothing if adj_nunits is nonnull.
8799 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
8800 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
8801 (emit_mode_fbit): Update use of print_maybe_const_decl.
8802 (emit_move_size): Likewise. Treat the array as non-const
8804 (emit_mode_adjustments): Handle adj_nunits.
8806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8808 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
8809 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
8810 (VECTOR_MODES): Use it.
8811 (make_vector_modes): Take the prefix as an argument.
8813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8814 Alan Hayward <alan.hayward@arm.com>
8815 David Sherwood <david.sherwood@arm.com>
8817 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
8818 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
8819 for MODE_VECTOR_BOOL.
8820 * machmode.def (VECTOR_BOOL_MODE): Document.
8821 * genmodes.c (VECTOR_BOOL_MODE): New macro.
8822 (make_vector_bool_mode): New function.
8823 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
8825 * lto-streamer-in.c (lto_input_mode_table): Likewise.
8826 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
8828 * stor-layout.c (int_mode_for_mode): Likewise.
8829 * tree.c (build_vector_type_for_mode): Likewise.
8830 * varasm.c (output_constant_pool_2): Likewise.
8831 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
8832 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
8833 for MODE_VECTOR_BOOL.
8834 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
8835 of mode class checks.
8836 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
8837 instead of a list of mode class checks.
8838 (expand_vector_scalar_condition): Likewise.
8839 (type_for_widest_vector_mode): Handle BImode as an inner mode.
8841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8842 Alan Hayward <alan.hayward@arm.com>
8843 David Sherwood <david.sherwood@arm.com>
8845 * machmode.h (mode_size): Change from unsigned short to
8847 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
8848 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8849 or if measurement_type is not polynomial.
8850 (fixed_size_mode::includes_p): Check for constant-sized modes.
8851 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
8852 return a poly_uint16 rather than an unsigned short.
8853 (emit_mode_size): Change the type of mode_size from unsigned short
8854 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
8855 (emit_mode_adjustments): Cope with polynomial vector sizes.
8856 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8858 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8860 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
8861 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
8862 * caller-save.c (setup_save_areas): Likewise.
8863 (replace_reg_with_saved_mem): Likewise.
8864 * calls.c (emit_library_call_value_1): Likewise.
8865 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
8866 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
8867 (gen_lowpart_for_combine): Likewise.
8868 * convert.c (convert_to_integer_1): Likewise.
8869 * cse.c (equiv_constant, cse_insn): Likewise.
8870 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
8871 (cselib_subst_to_values): Likewise.
8872 * dce.c (word_dce_process_block): Likewise.
8873 * df-problems.c (df_word_lr_mark_ref): Likewise.
8874 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
8875 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
8876 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
8877 (rtl_for_decl_location): Likewise.
8878 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
8879 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
8880 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
8881 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
8882 (expand_expr_real_1): Likewise.
8883 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
8884 (pad_below): Likewise.
8885 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8886 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
8887 * ira.c (get_subreg_tracking_sizes): Likewise.
8888 * ira-build.c (ira_create_allocno_objects): Likewise.
8889 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
8890 (ira_sort_regnos_for_alter_reg): Likewise.
8891 * ira-costs.c (record_operand_costs): Likewise.
8892 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
8893 (resolve_simple_move): Likewise.
8894 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
8895 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
8896 (lra_constraints): Likewise.
8897 (CONST_POOL_OK_P): Reject variable-sized modes.
8898 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
8899 (add_pseudo_to_slot, lra_spill): Likewise.
8900 * omp-low.c (omp_clause_aligned_alignment): Likewise.
8901 * optabs-query.c (get_best_extraction_insn): Likewise.
8902 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
8903 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
8904 (expand_mult_highpart, valid_multiword_target_p): Likewise.
8905 * recog.c (offsettable_address_addr_space_p): Likewise.
8906 * regcprop.c (maybe_mode_change): Likewise.
8907 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
8908 * regrename.c (build_def_use): Likewise.
8909 * regstat.c (dump_reg_info): Likewise.
8910 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
8911 (find_reloads, find_reloads_subreg_address): Likewise.
8912 * reload1.c (eliminate_regs_1): Likewise.
8913 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
8914 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
8915 (simplify_binary_operation_1, simplify_subreg): Likewise.
8916 * targhooks.c (default_function_arg_padding): Likewise.
8917 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
8918 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
8919 (verify_gimple_assign_ternary): Likewise.
8920 * tree-inline.c (estimate_move_cost): Likewise.
8921 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8922 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
8923 (get_address_cost_ainc): Likewise.
8924 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
8925 (vect_supportable_dr_alignment): Likewise.
8926 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8927 (vectorizable_reduction): Likewise.
8928 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
8929 (vectorizable_operation, vectorizable_load): Likewise.
8930 * tree.c (build_same_sized_truth_vector_type): Likewise.
8931 * valtrack.c (cleanup_auto_inc_dec): Likewise.
8932 * var-tracking.c (emit_note_insn_var_location): Likewise.
8933 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
8934 (ADDR_VEC_ALIGN): Likewise.
8936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8937 Alan Hayward <alan.hayward@arm.com>
8938 David Sherwood <david.sherwood@arm.com>
8940 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
8942 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
8943 or if measurement_type is polynomial.
8944 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
8945 * combine.c (make_extraction): Likewise.
8946 * dse.c (find_shift_sequence): Likewise.
8947 * dwarf2out.c (mem_loc_descriptor): Likewise.
8948 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
8949 (extract_bit_field, extract_low_bits): Likewise.
8950 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
8951 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
8952 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
8953 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
8954 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
8955 * reload.c (find_reloads): Likewise.
8956 * reload1.c (alter_reg): Likewise.
8957 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
8958 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
8959 * tree-if-conv.c (predicate_mem_writes): Likewise.
8960 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
8961 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
8962 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
8963 * valtrack.c (dead_debug_insert_temp): Likewise.
8964 * varasm.c (mergeable_constant_section): Likewise.
8965 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
8967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8968 Alan Hayward <alan.hayward@arm.com>
8969 David Sherwood <david.sherwood@arm.com>
8971 * expr.c (expand_assignment): Cope with polynomial mode sizes
8972 when assigning to a CONCAT.
8974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8975 Alan Hayward <alan.hayward@arm.com>
8976 David Sherwood <david.sherwood@arm.com>
8978 * machmode.h (mode_precision): Change from unsigned short to
8980 (mode_to_precision): Return a poly_uint16 rather than an unsigned
8982 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
8983 or if measurement_type is not polynomial.
8984 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
8985 in which the mode is already known to be a scalar_int_mode.
8986 * genmodes.c (emit_mode_precision): Change the type of mode_precision
8987 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
8989 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
8990 for GET_MODE_PRECISION.
8991 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
8992 for GET_MODE_PRECISION.
8993 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
8995 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
8996 (expand_field_assignment, make_extraction): Likewise.
8997 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
8998 (get_last_value): Likewise.
8999 * convert.c (convert_to_integer_1): Likewise.
9000 * cse.c (cse_insn): Likewise.
9001 * expr.c (expand_expr_real_1): Likewise.
9002 * lra-constraints.c (simplify_operand_subreg): Likewise.
9003 * optabs-query.c (can_atomic_load_p): Likewise.
9004 * optabs.c (expand_atomic_load): Likewise.
9005 (expand_atomic_store): Likewise.
9006 * ree.c (combine_reaching_defs): Likewise.
9007 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9008 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9009 * tree.h (type_has_mode_precision_p): Likewise.
9010 * ubsan.c (instrument_si_overflow): Likewise.
9012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9013 Alan Hayward <alan.hayward@arm.com>
9014 David Sherwood <david.sherwood@arm.com>
9016 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9017 polynomial numbers of units.
9018 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9019 (valid_vector_subparts_p): New function.
9020 (build_vector_type): Remove temporary shim and take the number
9021 of units as a poly_uint64 rather than an int.
9022 (build_opaque_vector_type): Take the number of units as a
9023 poly_uint64 rather than an int.
9024 * tree.c (build_vector_from_ctor): Handle polynomial
9025 TYPE_VECTOR_SUBPARTS.
9026 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9027 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9028 (build_vector_from_val): If the number of units is variable,
9029 use build_vec_duplicate_cst for constant operands and
9030 VEC_DUPLICATE_EXPR otherwise.
9031 (make_vector_type): Remove temporary is_constant ().
9032 (build_vector_type, build_opaque_vector_type): Take the number of
9033 units as a poly_uint64 rather than an int.
9034 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9036 * cfgexpand.c (expand_debug_expr): Likewise.
9037 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9038 (store_constructor, expand_expr_real_1): Likewise.
9039 (const_scalar_mask_from_tree): Likewise.
9040 * fold-const-call.c (fold_const_reduction): Likewise.
9041 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9042 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9043 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9044 (fold_relational_const): Likewise.
9045 (native_interpret_vector): Likewise. Change the size from an
9046 int to an unsigned int.
9047 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9048 TYPE_VECTOR_SUBPARTS.
9049 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9050 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9051 duplicating a non-constant operand into a variable-length vector.
9052 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9053 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9054 * ipa-icf.c (sem_variable::equals): Likewise.
9055 * match.pd: Likewise.
9056 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9057 * print-tree.c (print_node): Likewise.
9058 * stor-layout.c (layout_type): Likewise.
9059 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9060 * tree-cfg.c (verify_gimple_comparison): Likewise.
9061 (verify_gimple_assign_binary): Likewise.
9062 (verify_gimple_assign_ternary): Likewise.
9063 (verify_gimple_assign_single): Likewise.
9064 * tree-pretty-print.c (dump_generic_node): Likewise.
9065 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9066 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9067 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9068 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9069 (vect_shift_permute_load_chain): Likewise.
9070 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9071 (expand_vector_condition, optimize_vector_constructor): Likewise.
9072 (lower_vec_perm, get_compute_type): Likewise.
9073 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9074 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9075 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9076 (vect_recog_mask_conversion_pattern): Likewise.
9077 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9078 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9079 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9080 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9081 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9082 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9083 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9084 (supportable_widening_operation): Likewise.
9085 (supportable_narrowing_operation): Likewise.
9086 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9088 * varasm.c (output_constant): Likewise.
9090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9091 Alan Hayward <alan.hayward@arm.com>
9092 David Sherwood <david.sherwood@arm.com>
9094 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9095 so that both the length == 3 and length != 3 cases set up their
9096 own permute vectors. Add comments explaining why we know the
9097 number of elements is constant.
9098 (vect_permute_load_chain): Likewise.
9100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9101 Alan Hayward <alan.hayward@arm.com>
9102 David Sherwood <david.sherwood@arm.com>
9104 * machmode.h (mode_nunits): Change from unsigned char to
9106 (ONLY_FIXED_SIZE_MODES): New macro.
9107 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9108 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9109 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9111 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9112 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9113 or if measurement_type is not polynomial.
9114 * genmodes.c (ZERO_COEFFS): New macro.
9115 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9117 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9118 Use ZERO_COEFFS when emitting initializers.
9119 * data-streamer.h (bp_pack_poly_value): New function.
9120 (bp_unpack_poly_value): Likewise.
9121 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9122 for GET_MODE_NUNITS.
9123 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9124 for GET_MODE_NUNITS.
9125 * tree.c (make_vector_type): Remove temporary shim and make
9126 the real function take the number of units as a poly_uint64
9128 (build_vector_type_for_mode): Handle polynomial nunits.
9129 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9130 * emit-rtl.c (const_vec_series_p_1): Likewise.
9131 (gen_rtx_CONST_VECTOR): Likewise.
9132 * fold-const.c (test_vec_duplicate_folding): Likewise.
9133 * genrecog.c (validate_pattern): Likewise.
9134 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9135 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9136 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9137 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9138 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9139 * rtlanal.c (subreg_get_info): Likewise.
9140 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9141 (vect_grouped_load_supported): Likewise.
9142 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9143 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9144 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9145 (simplify_const_unary_operation, simplify_binary_operation_1)
9146 (simplify_const_binary_operation, simplify_ternary_operation)
9147 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9148 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9149 instead of CONST_VECTOR_NUNITS.
9150 * varasm.c (output_constant_pool_2): Likewise.
9151 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9152 explicit-encoded elements in the XVEC for variable-length vectors.
9154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9156 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9159 Alan Hayward <alan.hayward@arm.com>
9160 David Sherwood <david.sherwood@arm.com>
9162 * coretypes.h (fixed_size_mode): Declare.
9163 (fixed_size_mode_pod): New typedef.
9164 * builtins.h (target_builtins::x_apply_args_mode)
9165 (target_builtins::x_apply_result_mode): Change type to
9166 fixed_size_mode_pod.
9167 * builtins.c (apply_args_size, apply_result_size, result_vector)
9168 (expand_builtin_apply_args_1, expand_builtin_apply)
9169 (expand_builtin_return): Update accordingly.
9171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9173 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9174 * cselib.c (cselib_hash_rtx): Likewise.
9175 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9176 CONST_VECTOR encoding.
9178 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9179 Jeff Law <law@redhat.com>
9182 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9183 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9184 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9185 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9188 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9189 explicitly probe *sp in a noreturn function if there were any callee
9190 register saves or frame pointer is needed.
9192 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9195 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9196 BLKmode for ternary, binary or unary expressions.
9199 * var-tracking.c (delete_vta_debug_insn): New inline function.
9200 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9201 insns from get_insns () to NULL instead of each bb separately.
9202 Use delete_vta_debug_insn. No longer static.
9203 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9204 delete_vta_debug_insns callers.
9205 * rtl.h (delete_vta_debug_insns): Declare.
9206 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9207 instead of variable_tracking_main.
9209 2018-01-03 Martin Sebor <msebor@redhat.com>
9211 PR tree-optimization/83603
9212 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9213 arguments past the endof the argument list in functions declared
9214 without a prototype.
9215 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9216 Avoid checking when arguments are null.
9218 2018-01-03 Martin Sebor <msebor@redhat.com>
9221 * doc/extend.texi (attribute const): Fix a typo.
9222 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9223 issuing -Wsuggest-attribute for void functions.
9225 2018-01-03 Martin Sebor <msebor@redhat.com>
9227 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9228 offset_int::from instead of wide_int::to_shwi.
9229 (maybe_diag_overlap): Remove assertion.
9230 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9231 * gimple-ssa-sprintf.c (format_directive): Same.
9232 (parse_directive): Same.
9233 (sprintf_dom_walker::compute_format_length): Same.
9234 (try_substitute_return_value): Same.
9236 2018-01-03 Jeff Law <law@redhat.com>
9239 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9240 non-constant residual for zero at runtime and avoid probing in
9241 that case. Reorganize code for trailing problem to mirror handling
9244 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9246 PR tree-optimization/83501
9247 * tree-ssa-strlen.c (get_string_cst): New.
9248 (handle_char_store): Call get_string_cst.
9250 2018-01-03 Martin Liska <mliska@suse.cz>
9252 PR tree-optimization/83593
9253 * tree-ssa-strlen.c: Include tree-cfg.h.
9254 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9255 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9256 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9258 (strlen_dom_walker::before_dom_children): Call
9259 gimple_purge_dead_eh_edges. Dump tranformation with details
9261 (strlen_dom_walker::before_dom_children): Update call by adding
9262 new argument cleanup_eh.
9263 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9265 2018-01-03 Martin Liska <mliska@suse.cz>
9268 * cif-code.def (VARIADIC_THUNK): New enum value.
9269 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9272 2018-01-03 Jan Beulich <jbeulich@suse.com>
9274 * sse.md (mov<mode>_internal): Tighten condition for when to use
9275 vmovdqu<ssescalarsize> for TI and OI modes.
9277 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9279 Update copyright years.
9281 2018-01-03 Martin Liska <mliska@suse.cz>
9284 * ipa-visibility.c (function_and_variable_visibility): Skip
9285 functions with noipa attribure.
9287 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9289 * gcc.c (process_command): Update copyright notice dates.
9290 * gcov-dump.c (print_version): Ditto.
9291 * gcov.c (print_version): Ditto.
9292 * gcov-tool.c (print_version): Ditto.
9293 * gengtype.c (create_file): Ditto.
9294 * doc/cpp.texi: Bump @copying's copyright year.
9295 * doc/cppinternals.texi: Ditto.
9296 * doc/gcc.texi: Ditto.
9297 * doc/gccint.texi: Ditto.
9298 * doc/gcov.texi: Ditto.
9299 * doc/install.texi: Ditto.
9300 * doc/invoke.texi: Ditto.
9302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9304 * vector-builder.h (vector_builder::m_full_nelts): Change from
9305 unsigned int to poly_uint64.
9306 (vector_builder::full_nelts): Update prototype accordingly.
9307 (vector_builder::new_vector): Likewise.
9308 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9309 (vector_builder::operator ==): Likewise.
9310 (vector_builder::finalize): Likewise.
9311 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9312 Take the number of elements as a poly_uint64 rather than an
9314 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9315 from unsigned int to poly_uint64.
9316 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9317 (vec_perm_indices::new_vector): Likewise.
9318 (vec_perm_indices::length): Likewise.
9319 (vec_perm_indices::nelts_per_input): Likewise.
9320 (vec_perm_indices::input_nelts): Likewise.
9321 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9322 number of elements per input as a poly_uint64 rather than an
9323 unsigned int. Use the original encoding for variable-length
9324 vectors, rather than clamping each individual element.
9325 For the second and subsequent elements in each pattern,
9326 clamp the step and base before clamping their sum.
9327 (vec_perm_indices::series_p): Handle polynomial element counts.
9328 (vec_perm_indices::all_in_range_p): Likewise.
9329 (vec_perm_indices_to_tree): Likewise.
9330 (vec_perm_indices_to_rtx): Likewise.
9331 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9332 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9333 (tree_vector_builder::new_binary_operation): Handle polynomial
9334 element counts. Return false if we need to know the number
9335 of elements at compile time.
9336 * fold-const.c (fold_vec_perm): Punt if the number of elements
9337 isn't known at compile time.
9339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9341 * vec-perm-indices.h (vec_perm_builder): Change element type
9342 from HOST_WIDE_INT to poly_int64.
9343 (vec_perm_indices::element_type): Update accordingly.
9344 (vec_perm_indices::clamp): Handle polynomial element_types.
9345 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9346 (vec_perm_indices::all_in_range_p): Likewise.
9347 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9349 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9350 polynomial vec_perm_indices element types.
9351 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9352 * fold-const.c (fold_vec_perm): Likewise.
9353 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9354 * tree-vect-generic.c (lower_vec_perm): Likewise.
9355 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9356 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9357 element type to HOST_WIDE_INT.
9359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9360 Alan Hayward <alan.hayward@arm.com>
9361 David Sherwood <david.sherwood@arm.com>
9363 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9364 rather than an int. Use plus_constant.
9365 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9366 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9369 Alan Hayward <alan.hayward@arm.com>
9370 David Sherwood <david.sherwood@arm.com>
9372 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9373 a HOST_WIDE_INT to a poly_int64.
9375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9376 Alan Hayward <alan.hayward@arm.com>
9377 David Sherwood <david.sherwood@arm.com>
9379 * calls.c (load_register_parameters): Cope with polynomial
9380 mode sizes. Require a constant size for BLKmode parameters
9381 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9382 forces a parameter to be padded at the lsb end in order to
9383 fill a complete number of words, require the parameter size
9384 to be ordered wrt UNITS_PER_WORD.
9386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9387 Alan Hayward <alan.hayward@arm.com>
9388 David Sherwood <david.sherwood@arm.com>
9390 * reload1.c (spill_stack_slot_width): Change element type
9391 from unsigned int to poly_uint64_pod.
9392 (alter_reg): Treat mode sizes as polynomial.
9394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9395 Alan Hayward <alan.hayward@arm.com>
9396 David Sherwood <david.sherwood@arm.com>
9398 * reload.c (complex_word_subreg_p): New function.
9399 (reload_inner_reg_of_subreg, push_reload): Use it.
9401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9402 Alan Hayward <alan.hayward@arm.com>
9403 David Sherwood <david.sherwood@arm.com>
9405 * lra-constraints.c (process_alt_operands): Reject matched
9406 operands whose sizes aren't ordered.
9407 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9414 that the mode size is in the set {1, 2, 4, 8, 16}.
9416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9417 Alan Hayward <alan.hayward@arm.com>
9418 David Sherwood <david.sherwood@arm.com>
9420 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9421 Use plus_constant instead of gen_rtx_PLUS.
9423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9424 Alan Hayward <alan.hayward@arm.com>
9425 David Sherwood <david.sherwood@arm.com>
9427 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9428 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9429 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9430 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9431 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9432 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9433 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9434 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9435 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9436 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9438 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9439 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9440 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9441 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9442 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9443 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9444 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9445 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9446 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9447 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9449 * expr.c (emit_move_resolve_push): Treat the input and result
9450 of PUSH_ROUNDING as a poly_int64.
9451 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9452 (emit_push_insn): Likewise.
9453 * lra-eliminations.c (mark_not_eliminable): Likewise.
9454 * recog.c (push_operand): Likewise.
9455 * reload1.c (elimination_effects): Likewise.
9456 * rtlanal.c (nonzero_bits1): Likewise.
9457 * calls.c (store_one_arg): Likewise. Require the padding to be
9458 known at compile time.
9460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9461 Alan Hayward <alan.hayward@arm.com>
9462 David Sherwood <david.sherwood@arm.com>
9464 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9465 Use plus_constant instead of gen_rtx_PLUS.
9467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9468 Alan Hayward <alan.hayward@arm.com>
9469 David Sherwood <david.sherwood@arm.com>
9471 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9475 Alan Hayward <alan.hayward@arm.com>
9476 David Sherwood <david.sherwood@arm.com>
9478 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9479 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9480 via stack temporaries. Treat the mode size as polynomial too.
9482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9483 Alan Hayward <alan.hayward@arm.com>
9484 David Sherwood <david.sherwood@arm.com>
9486 * expr.c (expand_expr_real_2): When handling conversions involving
9487 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9488 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9489 as a poly_uint64 too.
9491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9492 Alan Hayward <alan.hayward@arm.com>
9493 David Sherwood <david.sherwood@arm.com>
9495 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
9502 REGMODE_NATURAL_SIZE.
9503 * expmed.c (store_bit_field_1): Likewise.
9504 * expr.c (store_constructor): Likewise.
9505 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9506 and polynomial REGMODE_NATURAL_SIZE.
9507 (gen_lowpart_common): Likewise.
9508 * reginfo.c (record_subregs_of_mode): Likewise.
9509 * rtlanal.c (read_modify_subreg_p): Likewise.
9511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9512 Alan Hayward <alan.hayward@arm.com>
9513 David Sherwood <david.sherwood@arm.com>
9515 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9516 numbers of elements.
9518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9519 Alan Hayward <alan.hayward@arm.com>
9520 David Sherwood <david.sherwood@arm.com>
9522 * match.pd: Cope with polynomial numbers of vector elements.
9524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9525 Alan Hayward <alan.hayward@arm.com>
9526 David Sherwood <david.sherwood@arm.com>
9528 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9529 in a POINTER_PLUS_EXPR.
9531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9532 Alan Hayward <alan.hayward@arm.com>
9533 David Sherwood <david.sherwood@arm.com>
9535 * omp-simd-clone.c (simd_clone_subparts): New function.
9536 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9537 (ipa_simd_modify_function_body): Likewise.
9539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9540 Alan Hayward <alan.hayward@arm.com>
9541 David Sherwood <david.sherwood@arm.com>
9543 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9544 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9545 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9546 (expand_vector_condition, vector_element): Likewise.
9547 (subparts_gt): New function.
9548 (get_compute_type): Use subparts_gt.
9549 (count_type_subparts): Delete.
9550 (expand_vector_operations_1): Use subparts_gt instead of
9551 count_type_subparts.
9553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9554 Alan Hayward <alan.hayward@arm.com>
9555 David Sherwood <david.sherwood@arm.com>
9557 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9558 (vect_compile_time_alias): ...this new function. Do the calculation
9559 on poly_ints rather than trees.
9560 (vect_prune_runtime_alias_test_list): Update call accordingly.
9562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9563 Alan Hayward <alan.hayward@arm.com>
9564 David Sherwood <david.sherwood@arm.com>
9566 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9568 (vect_schedule_slp_instance): Likewise.
9570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9571 Alan Hayward <alan.hayward@arm.com>
9572 David Sherwood <david.sherwood@arm.com>
9574 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9575 constant and extern definitions for variable-length vectors.
9576 (vect_get_constant_vectors): Note that the number of units
9577 is known to be constant.
9579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9580 Alan Hayward <alan.hayward@arm.com>
9581 David Sherwood <david.sherwood@arm.com>
9583 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9584 of units as polynomial. Choose between WIDE and NARROW based
9587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9588 Alan Hayward <alan.hayward@arm.com>
9589 David Sherwood <david.sherwood@arm.com>
9591 * tree-vect-stmts.c (simd_clone_subparts): New function.
9592 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9595 Alan Hayward <alan.hayward@arm.com>
9596 David Sherwood <david.sherwood@arm.com>
9598 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9599 vectors as polynomial. Use build_index_vector for
9602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9603 Alan Hayward <alan.hayward@arm.com>
9604 David Sherwood <david.sherwood@arm.com>
9606 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9607 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9608 for variable-length vectors.
9609 (vectorizable_mask_load_store): Treat the number of units as
9610 polynomial, asserting that it is constant if the condition has
9611 already been enforced.
9612 (vectorizable_store, vectorizable_load): Likewise.
9614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9615 Alan Hayward <alan.hayward@arm.com>
9616 David Sherwood <david.sherwood@arm.com>
9618 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9619 of units as polynomial. Punt if we can't tell at compile time
9620 which vector contains the final result.
9622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9623 Alan Hayward <alan.hayward@arm.com>
9624 David Sherwood <david.sherwood@arm.com>
9626 * tree-vect-loop.c (vectorizable_induction): Treat the number
9627 of units as polynomial. Punt on SLP inductions. Use an integer
9628 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9629 cast of such a series for variable-length floating-point
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.h (build_index_vector): Declare.
9637 * tree.c (build_index_vector): New function.
9638 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9639 of units as polynomial, forcibly converting it to a constant if
9640 vectorizable_reduction has already enforced the condition.
9641 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9642 to create a {1,2,3,...} vector.
9643 (vectorizable_reduction): Treat the number of units as polynomial.
9644 Choose vectype_in based on the largest scalar element size rather
9645 than the smallest number of units. Enforce the restrictions
9648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9649 Alan Hayward <alan.hayward@arm.com>
9650 David Sherwood <david.sherwood@arm.com>
9652 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9653 number of units as polynomial.
9655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9656 Alan Hayward <alan.hayward@arm.com>
9657 David Sherwood <david.sherwood@arm.com>
9659 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
9660 * target.def (autovectorize_vector_sizes): Return the vector sizes
9661 by pointer, using vector_sizes rather than a bitmask.
9662 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
9663 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
9664 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
9666 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
9667 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
9668 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
9669 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
9670 * omp-general.c (omp_max_vf): Likewise.
9671 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9672 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
9673 * tree-vect-loop.c (vect_analyze_loop): Likewise.
9674 * tree-vect-slp.c (vect_slp_bb): Likewise.
9675 * doc/tm.texi: Regenerate.
9676 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
9678 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
9679 the vector size as a poly_uint64 rather than an unsigned int.
9680 (current_vector_size): Change from an unsigned int to a poly_uint64.
9681 (get_vectype_for_scalar_type): Update accordingly.
9682 * tree.h (build_truth_vector_type): Take the size and number of
9683 units as a poly_uint64 rather than an unsigned int.
9684 (build_vector_type): Add a temporary overload that takes
9685 the number of units as a poly_uint64 rather than an unsigned int.
9686 * tree.c (make_vector_type): Likewise.
9687 (build_truth_vector_type): Take the number of units as a poly_uint64
9688 rather than an unsigned int.
9690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9691 Alan Hayward <alan.hayward@arm.com>
9692 David Sherwood <david.sherwood@arm.com>
9694 * target.def (get_mask_mode): Take the number of units and length
9695 as poly_uint64s rather than unsigned ints.
9696 * targhooks.h (default_get_mask_mode): Update accordingly.
9697 * targhooks.c (default_get_mask_mode): Likewise.
9698 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
9699 * doc/tm.texi: Regenerate.
9701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9702 Alan Hayward <alan.hayward@arm.com>
9703 David Sherwood <david.sherwood@arm.com>
9705 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
9706 * omp-general.c (omp_max_vf): Likewise.
9707 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
9708 (expand_omp_simd): Handle polynomial safelen.
9709 * omp-low.c (omplow_simd_context): Add a default constructor.
9710 (omplow_simd_context::max_vf): Change from int to poly_uint64.
9711 (lower_rec_simd_input_clauses): Update accordingly.
9712 (lower_rec_input_clauses): Likewise.
9714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9715 Alan Hayward <alan.hayward@arm.com>
9716 David Sherwood <david.sherwood@arm.com>
9718 * tree-vectorizer.h (vect_nunits_for_cost): New function.
9719 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
9720 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
9721 (vect_analyze_slp_cost): Likewise.
9722 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
9723 (vect_model_load_cost): Likewise.
9725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9726 Alan Hayward <alan.hayward@arm.com>
9727 David Sherwood <david.sherwood@arm.com>
9729 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
9730 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
9731 from an unsigned int * to a poly_uint64_pod *.
9732 (calculate_unrolling_factor): New function.
9733 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
9735 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9736 Alan Hayward <alan.hayward@arm.com>
9737 David Sherwood <david.sherwood@arm.com>
9739 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
9740 from an unsigned int to a poly_uint64.
9741 (_loop_vec_info::slp_unrolling_factor): Likewise.
9742 (_loop_vec_info::vectorization_factor): Change from an int
9744 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
9745 (vect_get_num_vectors): New function.
9746 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
9747 (vect_get_num_copies): Use vect_get_num_vectors.
9748 (vect_analyze_data_ref_dependences): Change max_vf from an int *
9749 to an unsigned int *.
9750 (vect_analyze_data_refs): Change min_vf from an int * to a
9752 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9753 than an unsigned HOST_WIDE_INT.
9754 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
9755 (vect_analyze_data_ref_dependence): Change max_vf from an int *
9756 to an unsigned int *.
9757 (vect_analyze_data_ref_dependences): Likewise.
9758 (vect_compute_data_ref_alignment): Handle polynomial vf.
9759 (vect_enhance_data_refs_alignment): Likewise.
9760 (vect_prune_runtime_alias_test_list): Likewise.
9761 (vect_shift_permute_load_chain): Likewise.
9762 (vect_supportable_dr_alignment): Likewise.
9763 (dependence_distance_ge_vf): Take the vectorization factor as a
9764 poly_uint64 rather than an unsigned HOST_WIDE_INT.
9765 (vect_analyze_data_refs): Change min_vf from an int * to a
9767 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
9768 vfm1 as a poly_uint64 rather than an int. Make the same change
9769 for the returned bound_scalar.
9770 (vect_gen_vector_loop_niters): Handle polynomial vf.
9771 (vect_do_peeling): Likewise. Update call to
9772 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
9773 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
9775 * tree-vect-loop.c (vect_determine_vectorization_factor)
9776 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
9777 (vect_get_known_peeling_cost): Likewise.
9778 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
9779 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
9780 (vect_transform_loop): Likewise. Use the lowest possible VF when
9781 updating the upper bounds of the loop.
9782 (vect_min_worthwhile_factor): Make static. Return an unsigned int
9784 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
9785 polynomial unroll factors.
9786 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
9787 (vect_make_slp_decision): Likewise.
9788 (vect_supported_load_permutation_p): Likewise, and polynomial
9790 (vect_analyze_slp_cost): Handle polynomial vf.
9791 (vect_slp_analyze_node_operations): Likewise.
9792 (vect_slp_analyze_bb_1): Likewise.
9793 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
9794 than an unsigned HOST_WIDE_INT.
9795 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
9796 (vectorizable_load): Handle polynomial vf.
9797 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
9799 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
9801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9802 Alan Hayward <alan.hayward@arm.com>
9803 David Sherwood <david.sherwood@arm.com>
9805 * match.pd: Handle bit operations involving three constants
9806 and try to fold one pair.
9808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9810 * tree-vect-loop-manip.c: Include gimple-fold.h.
9811 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
9812 niters_maybe_zero parameters. Handle other cases besides a step of 1.
9813 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
9814 Add a path that uses a step of VF instead of 1, but disable it
9816 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
9817 and niters_no_overflow parameters. Update calls to
9818 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
9819 Create a new SSA name if the latter choses to use a ste other
9820 than zero, and return it via niters_vector_mult_vf_var.
9821 * tree-vect-loop.c (vect_transform_loop): Update calls to
9822 vect_do_peeling, vect_gen_vector_loop_niters and
9823 slpeel_make_loop_iterate_ntimes.
9824 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
9825 (vect_gen_vector_loop_niters): Update declarations after above changes.
9827 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
9829 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
9830 128-bit round to integer instructions.
9831 (ceil<mode>2): Likewise.
9832 (btrunc<mode>2): Likewise.
9833 (round<mode>2): Likewise.
9835 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9837 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
9838 unaligned VSX load/store on P8/P9.
9839 (expand_block_clear): Allow the use of unaligned VSX
9840 load/store on P8/P9.
9842 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9844 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
9846 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
9847 swap associated with both a load and a store.
9849 2018-01-02 Andrew Waterman <andrew@sifive.com>
9851 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
9852 * config/riscv/riscv.md (clear_cache): Use it.
9854 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
9856 * web.c: Remove out-of-date comment.
9858 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9860 * expr.c (fixup_args_size_notes): Check that any existing
9861 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
9862 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
9863 (emit_single_push_insn): ...here.
9865 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9867 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
9868 (const_vector_encoded_nelts): New function.
9869 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
9870 (const_vector_int_elt, const_vector_elt): Declare.
9871 * emit-rtl.c (const_vector_int_elt_1): New function.
9872 (const_vector_elt): Likewise.
9873 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
9874 of CONST_VECTOR_ELT.
9876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9878 * expr.c: Include rtx-vector-builder.h.
9879 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
9880 directly on the tree encoding.
9881 (const_vector_from_tree): Likewise.
9882 * optabs.c: Include rtx-vector-builder.h.
9883 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
9884 sequence of "u" values.
9885 * vec-perm-indices.c: Include rtx-vector-builder.h.
9886 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
9887 directly on the vec_perm_indices encoding.
9889 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9891 * doc/rtl.texi (const_vector): Describe new encoding scheme.
9892 * Makefile.in (OBJS): Add rtx-vector-builder.o.
9893 * rtx-vector-builder.h: New file.
9894 * rtx-vector-builder.c: Likewise.
9895 * rtl.h (rtx_def::u2): Add a const_vector field.
9896 (CONST_VECTOR_NPATTERNS): New macro.
9897 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
9898 (CONST_VECTOR_DUPLICATE_P): Likewise.
9899 (CONST_VECTOR_STEPPED_P): Likewise.
9900 (CONST_VECTOR_ENCODED_ELT): Likewise.
9901 (const_vec_duplicate_p): Check for a duplicated vector encoding.
9902 (unwrap_const_vec_duplicate): Likewise.
9903 (const_vec_series_p): Check for a non-duplicated vector encoding.
9904 Say that the function only returns true for integer vectors.
9905 * emit-rtl.c: Include rtx-vector-builder.h.
9906 (gen_const_vec_duplicate_1): Delete.
9907 (gen_const_vector): Call gen_const_vec_duplicate instead of
9908 gen_const_vec_duplicate_1.
9909 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
9910 (gen_const_vec_duplicate): Use rtx_vector_builder.
9911 (gen_const_vec_series): Likewise.
9912 (gen_rtx_CONST_VECTOR): Likewise.
9913 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
9914 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9915 Build a new vector rather than modifying a CONST_VECTOR in-place.
9916 (handle_special_swappables): Update call accordingly.
9917 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
9918 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
9919 Build a new vector rather than modifying a CONST_VECTOR in-place.
9920 (handle_special_swappables): Update call accordingly.
9922 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9924 * simplify-rtx.c (simplify_const_binary_operation): Use
9925 CONST_VECTOR_ELT instead of XVECEXP.
9927 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9929 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
9930 the selector elements to be different from the data elements
9931 if the selector is a VECTOR_CST.
9932 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
9933 ssizetype for the selector.
9935 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9937 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
9938 before testing each element individually.
9939 * tree-vect-generic.c (lower_vec_perm): Likewise.
9941 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9943 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
9944 * selftest-run-tests.c (selftest::run_tests): Call it.
9945 * vector-builder.h (vector_builder::operator ==): New function.
9946 (vector_builder::operator !=): Likewise.
9947 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
9948 (vec_perm_indices::all_from_input_p): New function.
9949 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9950 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
9951 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
9952 instead of reading the VECTOR_CST directly. Detect whether both
9953 vector inputs are the same before constructing the vec_perm_indices,
9954 and update the number of inputs argument accordingly. Use the
9955 utility functions added above. Only construct sel2 if we need to.
9957 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9959 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
9960 the broadcast of the low byte.
9961 (expand_mult_highpart): Use an explicit encoding for the permutes.
9962 * optabs-query.c (can_mult_highpart_p): Likewise.
9963 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
9964 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9965 (vectorizable_bswap): Likewise.
9966 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
9967 explicit encoding for the power-of-2 permutes.
9968 (vect_permute_store_chain): Likewise.
9969 (vect_grouped_load_supported): Likewise.
9970 (vect_permute_load_chain): Likewise.
9972 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9974 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
9975 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
9976 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
9977 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9978 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
9979 (vect_gen_perm_mask_any): Likewise.
9981 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
9983 * int-vector-builder.h: New file.
9984 * vec-perm-indices.h: Include int-vector-builder.h.
9985 (vec_perm_indices): Redefine as an int_vector_builder.
9986 (auto_vec_perm_indices): Delete.
9987 (vec_perm_builder): Redefine as a stand-alone class.
9988 (vec_perm_indices::vec_perm_indices): New function.
9989 (vec_perm_indices::clamp): Likewise.
9990 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
9991 (vec_perm_indices::new_vector): New function.
9992 (vec_perm_indices::new_expanded_vector): Update for new
9993 vec_perm_indices class.
9994 (vec_perm_indices::rotate_inputs): New function.
9995 (vec_perm_indices::all_in_range_p): Operate directly on the
9996 encoded form, without computing elided elements.
9997 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
9998 encoding. Update for new vec_perm_indices class.
9999 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10000 the given vec_perm_builder.
10001 (expand_vec_perm_var): Update vec_perm_builder constructor.
10002 (expand_mult_highpart): Use vec_perm_builder instead of
10003 auto_vec_perm_indices.
10004 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10005 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10006 or double series encoding as appropriate.
10007 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10008 vec_perm_indices instead of auto_vec_perm_indices.
10009 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10010 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10011 (vect_permute_store_chain): Likewise.
10012 (vect_grouped_load_supported): Likewise.
10013 (vect_permute_load_chain): Likewise.
10014 (vect_shift_permute_load_chain): Likewise.
10015 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10016 (vect_transform_slp_perm_load): Likewise.
10017 (vect_schedule_slp_instance): Likewise.
10018 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10019 (vectorizable_mask_load_store): Likewise.
10020 (vectorizable_bswap): Likewise.
10021 (vectorizable_store): Likewise.
10022 (vectorizable_load): Likewise.
10023 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10024 vec_perm_indices instead of auto_vec_perm_indices. Use
10025 tree_to_vec_perm_builder to read the vector from a tree.
10026 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10027 vec_perm_builder instead of a vec_perm_indices.
10028 (have_whole_vector_shift): Use vec_perm_builder and
10029 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10030 truncation to calc_vec_perm_mask_for_shift.
10031 (vect_create_epilog_for_reduction): Likewise.
10032 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10033 from auto_vec_perm_indices to vec_perm_indices.
10034 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10035 instead of changing individual elements.
10036 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10037 the vector in d.perm.
10038 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10039 from auto_vec_perm_indices to vec_perm_indices.
10040 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10041 instead of changing individual elements.
10042 (arm_vectorize_vec_perm_const): Use new_vector to install
10043 the vector in d.perm.
10044 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10045 Update vec_perm_builder constructor.
10046 (rs6000_expand_interleave): Likewise.
10047 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10048 (rs6000_expand_interleave): Likewise.
10050 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10052 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10053 to qimode could truncate the indices.
10054 * optabs.c (expand_vec_perm_var): Likewise.
10056 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10058 * Makefile.in (OBJS): Add vec-perm-indices.o.
10059 * vec-perm-indices.h: New file.
10060 * vec-perm-indices.c: Likewise.
10061 * target.h (vec_perm_indices): Replace with a forward class
10063 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10064 * optabs.h: Include vec-perm-indices.h.
10065 (expand_vec_perm): Delete.
10066 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10067 (expand_vec_perm_const): Declare.
10068 * target.def (vec_perm_const_ok): Replace with...
10069 (vec_perm_const): ...this new hook.
10070 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10071 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10072 * doc/tm.texi: Regenerate.
10073 * optabs.def (vec_perm_const): Delete.
10074 * doc/md.texi (vec_perm_const): Likewise.
10075 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10076 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10077 expand_vec_perm for constant permutation vectors. Assert that
10078 the mode of variable permutation vectors is the integer equivalent
10079 of the mode that is being permuted.
10080 * optabs-query.h (selector_fits_mode_p): Declare.
10081 * optabs-query.c: Include vec-perm-indices.h.
10082 (selector_fits_mode_p): New function.
10083 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10084 is defined, instead of checking whether the vec_perm_const_optab
10085 exists. Use targetm.vectorize.vec_perm_const instead of
10086 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10087 fit in the vector mode before using a variable permute.
10088 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10089 vec_perm_indices instead of an rtx.
10090 (expand_vec_perm): Replace with...
10091 (expand_vec_perm_const): ...this new function. Take the selector
10092 as a vec_perm_indices rather than an rtx. Also take the mode of
10093 the selector. Update call to shift_amt_for_vec_perm_mask.
10094 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10095 Use vec_perm_indices::new_expanded_vector to expand the original
10096 selector into bytes. Check whether the indices fit in the vector
10097 mode before using a variable permute.
10098 (expand_vec_perm_var): Make global.
10099 (expand_mult_highpart): Use expand_vec_perm_const.
10100 * fold-const.c: Includes vec-perm-indices.h.
10101 * tree-ssa-forwprop.c: Likewise.
10102 * tree-vect-data-refs.c: Likewise.
10103 * tree-vect-generic.c: Likewise.
10104 * tree-vect-loop.c: Likewise.
10105 * tree-vect-slp.c: Likewise.
10106 * tree-vect-stmts.c: Likewise.
10107 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10109 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10110 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10111 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10112 (aarch64_vectorize_vec_perm_const): ...this new function.
10113 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10114 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10115 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10116 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10117 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10118 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10119 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10121 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10122 check for NEON modes.
10123 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10124 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10125 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10126 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10128 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10129 the old VEC_PERM_CONST conditions.
10130 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10131 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10132 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10133 (ia64_vectorize_vec_perm_const_ok): Merge into...
10134 (ia64_vectorize_vec_perm_const): ...this new function.
10135 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10136 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10137 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10138 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10139 * config/mips/mips.c (mips_expand_vec_perm_const)
10140 (mips_vectorize_vec_perm_const_ok): Merge into...
10141 (mips_vectorize_vec_perm_const): ...this new function.
10142 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10143 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10144 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10145 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10146 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10147 (rs6000_expand_vec_perm_const): Delete.
10148 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10150 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10151 (altivec_expand_vec_perm_const_le): Take each operand individually.
10152 Operate on constant selectors rather than rtxes.
10153 (altivec_expand_vec_perm_const): Likewise. Update call to
10154 altivec_expand_vec_perm_const_le.
10155 (rs6000_expand_vec_perm_const): Delete.
10156 (rs6000_vectorize_vec_perm_const_ok): Delete.
10157 (rs6000_vectorize_vec_perm_const): New function.
10158 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10159 an element count and rtx array.
10160 (rs6000_expand_extract_even): Update call accordingly.
10161 (rs6000_expand_interleave): Likewise.
10162 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10163 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10164 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10165 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10166 (rs6000_expand_vec_perm_const): Delete.
10167 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10168 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10169 (altivec_expand_vec_perm_const_le): Take each operand individually.
10170 Operate on constant selectors rather than rtxes.
10171 (altivec_expand_vec_perm_const): Likewise. Update call to
10172 altivec_expand_vec_perm_const_le.
10173 (rs6000_expand_vec_perm_const): Delete.
10174 (rs6000_vectorize_vec_perm_const_ok): Delete.
10175 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10176 reference to the SPE evmerge intructions.
10177 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10178 an element count and rtx array.
10179 (rs6000_expand_extract_even): Update call accordingly.
10180 (rs6000_expand_interleave): Likewise.
10181 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10182 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10184 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10186 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10188 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10189 vector mode and that that mode matches the mode of the data
10191 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10192 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10193 directly using expand_vec_perm_1 when forcing selectors into
10195 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10197 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10199 * optabs-query.h (can_vec_perm_p): Delete.
10200 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10201 * optabs-query.c (can_vec_perm_p): Split into...
10202 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10203 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10204 particular selector is valid.
10205 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10206 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10207 (vect_grouped_load_supported): Likewise.
10208 (vect_shift_permute_load_chain): Likewise.
10209 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10210 (vect_transform_slp_perm_load): Likewise.
10211 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10212 (vectorizable_bswap): Likewise.
10213 (vect_gen_perm_mask_checked): Likewise.
10214 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10215 implementations of variable permutation vectors into account
10216 when deciding which selector to use.
10217 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10218 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10219 with a false third argument.
10220 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10221 to test whether the constant selector is valid and can_vec_perm_var_p
10222 to test whether a variable selector is valid.
10224 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10226 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10227 * optabs-query.c (can_vec_perm_p): Likewise.
10228 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10229 instead of vec_perm_indices.
10230 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10231 (vect_gen_perm_mask_checked): Likewise,
10232 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10233 (vect_gen_perm_mask_checked): Likewise,
10235 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10237 * optabs-query.h (qimode_for_vec_perm): Declare.
10238 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10239 (qimode_for_vec_perm): ...this new function.
10240 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10242 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10244 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10245 does not have a conditional at the top.
10247 2018-01-02 Richard Biener <rguenther@suse.de>
10249 * ipa-inline.c (big_speedup_p): Fix expression.
10251 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10254 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10257 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10261 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10262 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10263 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10264 cond_taken_branch_cost 3->4.
10266 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10268 PR tree-optimization/83581
10269 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10270 TODO_cleanup_cfg if any changes have been made.
10272 PR middle-end/83608
10273 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10274 convert_modes if target mode has the right side, but different mode
10277 PR middle-end/83609
10278 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10279 last argument when extracting from CONCAT. If either from_real or
10280 from_imag is NULL, use expansion through memory. If result is not
10281 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10282 the parts directly to inner mode, if even that fails, use expansion
10285 PR middle-end/83623
10286 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10287 check for bswap in mode rather than HImode and use that in expand_unop
10290 Copyright (C) 2018 Free Software Foundation, Inc.
10292 Copying and distribution of this file, with or without modification,
10293 are permitted in any medium without royalty provided the copyright
10294 notice and this notice are preserved.