1 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3 * config/arm/arm-cpus.in (mode26): Delete.
4 (armv4): Delete mode26 reference.
5 * config/arm/arm.c (arm_configure_build_target): Delete use of
8 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
10 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
12 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
13 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
15 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
16 (floatunsdidf2): Ditto.
18 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
20 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
21 (fixuns_trunc<mode>si2_avx512f): Ditto.
22 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
23 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
24 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
26 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
28 PR rtl-optimization/79985
29 * df-scan.c (df_insn_refs_collect): Remove special case for
30 global registers and asm statements.
32 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
34 * extend.texi (Global Register Variables): Rewrite the bullet list.
35 Note that the register is available for allocation. Note that access
36 via inline asm must use constraints. Add note about async-signal
37 handlers. Remove paragraph about automagic register selection.
39 2018-05-23 Richard Biener <rguenther@suse.de>
41 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
42 of fixed offset from memset VN.
44 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
46 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
48 (alloc_cand_and_find_basis): Initialize first_interp field.
49 (slsr_process_mul): Modify first_interp field.
50 (slsr_process_add): Likewise.
51 (slsr_process_cast): Modify first_interp field for each new
53 (slsr_process_copy): Likewise.
54 (dump_candidate): Dump first_interp field.
55 (replace_mult_candidate): Process all interpretations, not just
57 (replace_rhs_if_not_dup): Likewise.
58 (replace_one_candidate): Likewise.
60 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
62 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
64 (aarch64_needs_frame_chain): New function.
65 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
67 2018-05-23 Sudakshina Das <sudi.das@arm.com>
70 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
71 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
72 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
73 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
74 as true for strict-align.
75 (aarch64_can_inline_p): Perform checks even when callee has no
76 attributes to check for strict alignment.
77 * doc/extend.texi (AArch64 Function Attributes): Document
79 * doc/invoke.texi: (AArch64 Options): Likewise.
81 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
83 PR tree-optimization/85853
84 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
85 the handling of the root of the node to...
86 (vect_slp_analyze_node_operations_1): ...this new function,
87 and run the whole thing with the child nodes' def types
88 set according to their SLP node's def type.
90 2018-05-23 Richard Biener <rguenther@suse.de>
93 * tree-data-ref.c (create_runtime_alias_checks): Defer
94 and ignore overflow warnings.
96 2018-05-23 Yury Gribov <tetra2005@gmail.com>
98 PR tree-optimization/85822
99 * tree-vrp.c (is_masked_range_test): Fix handling of negative
102 2018-05-23 Richard Biener <rguenther@suse.de>
104 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
105 memset constants via native_interpret_expr.
107 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
110 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
112 (cgraph_node::create_alias): Likewise.
113 (cgraph_node::get_availability): Check ifunc_resolver instead
114 of looking up ifunc attribute.
115 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
116 * varasm.c (do_assemble_alias): Likewise.
117 (assemble_alias): Likewise.
118 (default_binds_local_p_3): Likewise.
119 * cgraph.h (cgraph_node): Add ifunc_resolver.
120 (cgraph_node::only_called_directly_or_aliased_p): Return false
122 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
124 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
125 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
126 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
127 instead of looking up ifunc attribute.
129 2018-05-22 Luis Machado <luis.machado@linaro.org>
131 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
133 2018-05-22 Martin Sebor <msebor@redhat.com>
136 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
137 only when expasion succeeds.
138 (expand_builtin_strcmp): Same.
139 (expand_builtin_strncmp): Same.
141 2018-05-22 Martin Sebor <msebor@redhat.com>
143 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
145 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
146 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
148 * config/aarch64/aarch64-ldpstp.md: Replace uses of
149 aarch64_mem_pair_operand with memory_operand and delete operand swapping
151 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
152 Add check for legitimate_address.
153 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
154 (aarch64_swap_ldrstr_operands): New.
155 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
158 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
159 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
161 * config/aarch64/aarch64.md: New patterns to generate stp
163 (store_pair_sw, store_pair_dw): New patterns to generate stp for
164 single words and double words.
165 (load_pair_sw, load_pair_dw): Likewise.
166 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
168 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
170 * config/aarch64/aarch64-ldpstp.md: Modify peephole
171 for different mode ldpstp and add peephole for merged zero stores.
173 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
175 (aarch64_gen_store_pair): Rename calls to match new patterns.
176 (aarch64_gen_load_pair): Rename calls to match new patterns.
177 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
178 (load_pair<DREG:mode><DREG2:mode>): ... This.
179 (store_pair<mode>): Rename to...
180 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
181 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
183 (V_INT_EQUIV): Handle SImode.
184 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
187 2018-05-22 Martin Sebor <msebor@redhat.com>
190 * calls.c (maybe_warn_nonstring_arg): Use string length to set
191 or ajust the presumed bound on an operation to avoid unnecessary
194 2018-05-22 Martin Sebor <msebor@redhat.com>
196 PR tree-optimization/85826
197 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
198 assuming that a DECL necesarily has a constant size.
200 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
203 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
205 2018-05-22 Richard Biener <rguenther@suse.de>
207 PR tree-optimization/85834
208 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
209 non-constant and non-zero memset arguments.
211 2018-05-22 Martin Liska <mliska@suse.cz>
214 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
216 2018-05-22 Richard Biener <rguenther@suse.de>
218 PR tree-optimization/85863
219 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
220 comparisons when vectype is specified.
221 (vectorizable_condition): Do not specify vectype for
222 vect_is_simple_cond when SLP vectorizing.
224 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
227 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
228 define __ibm128 as long double.
229 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
230 as a distinct type when IEEE 128-bit support is enabled.
231 (init_float128_ieee): Fix up conversions between IFmode and IEEE
232 128-bit types to use the correct functions.
233 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
234 convert between 128-bit floating point types that have different
235 modes but the same representation, instead of using gen_lowpart to
237 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
239 (IFKF_reg): New attributes to give the register constraints for
241 (extend<mode>tf2_internal): New insns to mark an explicit
242 conversion between 128-bit floating point types that have a
243 different mode but share the same representation.
245 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
247 PR tree-optimization/85814
248 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
249 a null return from get_strinfo when unsharing the next
250 strinfo in the chain.
252 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
255 * varasm.c (weak_finish): Clean up weak_decls.
257 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
259 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
260 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
261 UNSPEC_UADALP values.
262 * config/aarch64/iterators.md (ABAL): New int iterator.
265 (sur): Add mappings for the above.
266 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
268 (aarch64_<sur>abal<mode>_4): Likewise.
269 (aarch64_<sur>adalp<mode>_3): Likewise.
270 (<sur>sadv16qi): New define_expand.
272 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
274 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
275 (*movdf_internal): Ditto.
276 (*rcpsf2_sse): Ditto.
277 (*rsqrtsf2_sse): Ditto.
278 (*sqrt<mode>2_sse): Ditto.
280 2018-05-21 Tamar Christina <tamar.christina@arm.com>
282 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
284 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
285 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
286 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
287 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
289 * config/aarch64/arm_neon.h: Likewise.
290 * config/aarch64/iterators.md (VQ_I): New.
292 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
294 * config.gcc: Add arc/t-multilib-linux to tmake_file for
296 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
299 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
301 * config/nds32/constraints.md (S): New constraint.
302 * config/nds32/nds32.md (call_internal): Use constraint S.
303 (call_value_internal): Likewise.
304 (sibcall_internal): Likewise.
305 (sibcall_value_internal): Likewise.
307 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
308 Chung-Ju Wu <jasonwucj@gmail.com>
310 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
313 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
314 Chung-Ju Wu <jasonwucj@gmail.com>
316 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
317 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
318 (nds32_rtx_costs_impl): Simplify.
319 (nds32_address_cost_impl): Simplify.
320 (nds32_init_rtx_costs): New function.
321 (nds32_rtx_costs_speed_prefer): Likewise.
322 (nds32_rtx_costs_size_prefer): Likewise.
323 (nds32_address_cost_speed_prefer): Likewise.
324 (nds32_address_cost_speed_fwprop): Likewise.
325 (nds32_address_cost_size_prefer): Likewise.
326 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
327 * config/nds32/nds32.c (nds32_option_override): Use
328 nds32_init_rtx_costs function.
330 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
332 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
333 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
334 (TARGET_PIPELINE_N8): Likewise.
335 (TARGET_PIPELINE_N10): Likewise.
336 (TARGET_PIPELINE_N13): Likewise.
337 (TARGET_PIPELINE_GRAYWOLF): Likewise.
339 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
341 * config/nds32/nds32-fpu.md: Update copyright year.
343 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
345 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
347 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
349 * config/nds32/nds32.c
350 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
351 * config/nds32/nds32.opt (minline-asm-r15): New option.
353 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
355 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
357 * config/nds32/nds32.md (abssi2): New pattern.
359 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
361 * config/i386/i386.md (rex64namesuffix): New mode attribute.
362 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
363 Merge insn pattern from sse_cvtsi2ss<round_name> and
364 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
365 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
366 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
367 using SWI48 mode iterator.
368 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
369 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
370 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
371 pattern from sse_cvttss2si<round_saeonly_name>
372 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
373 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
374 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
375 using SWI48 mode iterator.
376 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
377 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
378 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
379 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
380 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
381 using SWI48 mode iterator.
382 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
383 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
384 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
385 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
386 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
388 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
389 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
390 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
391 pattern from sse_cvttsd2si<round_saeonly_name>
392 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
394 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
396 * config/nds32/nds32-md-auxiliary.c
397 (nds32_valid_smw_lwm_base_p): Refine.
398 (nds32_output_smw_single_word): Refine.
399 (nds32_output_smw_double_word): New.
400 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
402 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
404 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
405 (nds32_output_stack_pop): Refine.
406 (nds32_expand_unaligned_load): Refine.
407 (nds32_expand_unaligned_store): Refine.
409 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
410 Chung-Ju Wu <jasonwucj@gmail.com>
412 * config/nds32/constants.md: Add TP_REGNUM constant.
413 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
414 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
416 * config/nds32/nds32-doubleword.md: Consider flag_pic.
417 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
418 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
419 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
420 and PIC code generation.
421 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
423 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
425 * config/nds32/nds32.md: Support TLS and PIC.
426 * config/nds32/nds32.c: Support TLS and PIC.
427 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
428 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
431 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
433 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
436 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
437 Chung-Ju Wu <jasonwucj@gmail.com>
439 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
440 * config/nds32/nds32-md-auxiliary.c
441 (symbolic_reference_mentioned_p): New.
442 (nds32_legitimize_ict_address): New.
443 (nds32_expand_ict_move): New.
444 (nds32_indirect_call_referenced_p): New.
445 (nds32_symbol_binds_local_p): Delete.
446 (nds32_long_call_p): Modify.
447 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
448 * config/nds32/nds32-protos.h
449 (symbolic_reference_mentioned_p): Declare.
450 (nds32_legitimize_ict_address): Declare.
451 (nds32_expand_ict_move): Declare.
452 (nds32_indirect_call_referenced_p): Declare.
453 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
454 (nds32_relax_group): Use nds32_ict_const_p as condition.
455 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
456 (nds32_asm_file_start): Output ict_model directive in asm code.
457 (nds32_legitimate_address_p): Consider indirect call.
458 (nds32_print_operand): Consider indirect call.
459 (nds32_print_operand_address): Consider indirect call.
460 (nds32_insert_attributes): Handle "indirect_call" attribute.
461 (TARGET_LEGITIMATE_ADDRESS_P): Define.
462 (TARGET_LEGITIMATE_CONSTANT_P): Define.
463 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
464 (TARGET_DELEGITIMIZE_ADDRESS): Define.
465 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
466 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
467 (TARGET_ICT_MODEL_SMALL): Define.
468 (TARGET_ICT_MODEL_LARGE): Define.
469 * config/nds32/nds32.md (movsi): Consider ict model.
470 (call, call_value): Consider ict model.
471 (sibcall, sibcall_value): Consider ict model.
472 * config/nds32/nds32.opt (mict-model): New option.
473 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
476 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
477 Monk Chiang <sh.chiang04@gmail.com>
478 Jim Wilson <jimw@sifive.com>
480 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
481 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
482 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
483 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
484 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
485 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
486 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
487 compute save_libcall_adjustment properly.
488 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
489 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
490 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
491 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
492 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
493 (ABI_SPEC): Handle mabi=ilp32e.
494 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
496 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
497 <-march>: Add rv32e as an example.
499 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
502 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
503 (intra_create_variable_infos): Handle C++ constructors.
505 2018-05-18 Martin Liska <mliska@suse.cz>
507 * passes.def: Remove a redundant pass.
509 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
512 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
514 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
516 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
518 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
520 (armv2, armv2a, armv3, armv3m): Delete architectures.
521 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
522 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
523 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
525 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
526 (*mulsidi3adddi): Likewise.
527 (mulsidi3): Likewise.
528 (*mulsidi3_nov6): Likewise.
529 (umulsidi3): Likewise.
530 (umulsidi3_nov6): Likewise.
531 (umaddsidi4): Likewise.
532 (*umulsidi3adddi): Likewise.
533 (smulsi3_highpart): Likewise.
534 (*smulsi3_highpart_nov6): Likewise.
535 (umulsi3_highpart): Likewise.
536 (*umulsi3_highpart_nov6): Likewise.
537 * config/arm/arm.h (arm_arch3m): Delete.
538 * config/arm/arm.c (arm_arch3m): Delete.
539 (arm_option_override_internal): Update armv3-related comment.
540 (arm_configure_build_target): Delete use of isa_bit_mode32.
541 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
542 (arm_rtx_costs_internal): Delete check of arm_arch3m.
543 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
546 * config/arm/arm-protos.h (arm_arch3m): Delete.
547 * config/arm/arm-tables.opt: Regenerate.
548 * config/arm/arm-tune.md: Likewise.
549 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
550 deleted architectures.
552 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
554 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
555 (armv5t, armv5te): New features.
556 (ARMv5, ARMv5e): Delete fgroups.
557 (ARMv5t, ARMv5te): Adjust for above changes.
559 (armv5, armv5e): Delete arches.
560 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
562 (*call_reg_arm): Likewise.
563 (*call_value_reg_armv5): Likewise.
564 (*call_value_reg_arm): Likewise.
565 (*call_symbol): Likewise.
566 (*call_value_symbol): Likewise.
567 (*sibcall_insn): Likewise.
568 (*sibcall_value_insn): Likewise.
570 (prefetch): Likewise.
571 (define_split and define_peephole2 dependent on arm_arch5):
573 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
575 (TARGET_ARM_QBIT): Likewise.
576 (TARGET_DSP_MULTIPLY): Likewise.
577 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
578 (arm_arch5, arm_arch5e): Delete.
579 (arm_arch5t, arm_arch5te): Declare.
580 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
581 (arm_arch5t): Declare.
582 (arm_option_reconfigure_globals): Update for the above.
583 (arm_options_perform_arch_sanity_checks): Update comment, replace
584 use of arm_arch5 with arm_arch5t.
585 (use_return_insn): Likewise.
586 (arm_emit_call_insn): Likewise.
587 (output_return_instruction): Likewise.
588 (arm_final_prescan_insn): Likewise.
589 (arm_coproc_builtin_available): Likewise.
590 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
591 arm_arch5e with arm_arch5t and arm_arch5te.
592 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
593 (arm_arch5t, arm_arch5te): Declare.
594 * config/arm/arm-tables.opt: Regenerate.
595 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
596 * config/arm/t-multilib: Likewise.
597 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
598 instead of arm_arch5.
599 (*call_reg_thumb1): Likewise.
600 (*call_value_reg_thumb1_v5): Likewise.
601 (*call_value_reg_thumb1): Likewise.
602 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
604 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
606 2018-05-18 Martin Liska <mliska@suse.cz>
608 PR gcov-profile/84846
609 * doc/gcov.texi: Document -t option of gcov tool.
611 2018-05-18 Martin Liska <mliska@suse.cz>
613 PR gcov-profile/84846
614 * gcov.c (print_usage): Add new -t option.
615 (process_args): Handle the option.
616 (generate_results): Use stdout as output when requested by
619 2018-05-18 Martin Liska <mliska@suse.cz>
621 PR gcov-profile/84846
622 * coverage.c (coverage_init): Write PWD to .gcno file.
623 * doc/gcov.texi: Document how working directory is printed.
624 * gcov-dump.c (dump_gcov_file): Print PWD.
625 * gcov.c (output_intermediate_file): Likewise.
626 (read_graph_file): Read PWD string.
627 (output_lines): Print PWD.
629 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
632 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
633 for retval and return false if all args to phi are zero.
635 2018-05-18 Richard Biener <rguenther@suse.de>
637 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
639 (evrp_dom_walker::before_dom_children): Call it.
641 2018-05-18 Richard Biener <rguenther@suse.de>
643 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
644 results when processing array refs with variable index.
646 2018-05-18 Toon Moene <toon@moene.org>
648 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
649 directly after that of -floop-interchange. Indicate that both
650 options are enabled by default when specifying -O3.
652 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
654 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
655 iterator. Delete separate integer-mode vec_set<mode> expander.
656 (aarch64_simd_vec_setv2di): Delete.
657 (vec_setv2di): Delete.
658 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
659 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
660 the "w, r" alternative.
662 2018-05-18 Martin Liska <mliska@suse.cz>
664 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
665 * tree-pass.h (make_pass_lower_switch_O0): New function.
666 * tree-switch-conversion.c (node_has_low_bound): Remove.
667 (node_has_high_bound): Likewise.
668 (node_is_bounded): Likewise.
669 (class pass_lower_switch): Make it a template type and create
671 (pass_lower_switch::execute): Add template argument.
672 (make_pass_lower_switch): New function.
673 (make_pass_lower_switch_O0): New function.
674 (do_jump_if_equal): Remove.
675 (emit_case_nodes): Simplify to just handle all 3 cases and leave
676 all the hard work to tree optimization passes.
678 2018-05-18 Martin Liska <mliska@suse.cz>
680 * dbgcnt.c (limit_low): Renamed from limit.
681 (limit_high): New variable.
682 (dbg_cnt_is_enabled): Check for upper limit.
683 (dbg_cnt): Adjust dumping.
684 (dbg_cnt_set_limit_by_index): Add new argument for high
686 (dbg_cnt_set_limit_by_name): Likewise.
687 (dbg_cnt_process_single_pair): Parse new format.
688 (dbg_cnt_process_opt): Use strtok.
689 (dbg_cnt_list_all_counters): Remove 'value' and add
691 * doc/invoke.texi: Document changes.
693 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
695 * doc/sourcebuild.texi (scalar_all_fma): Document.
696 * tree.def (FMA_EXPR): Delete.
697 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
698 * internal-fn.c (ternary_direct): New macro.
699 (expand_ternary_optab_fn): Likewise.
700 (direct_ternary_optab_supported_p): Likewise.
701 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
702 * builtins.c (fold_builtin_fma): Delete.
703 (fold_builtin_3): Don't call it.
704 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
705 * expr.c (expand_expr_real_2): Likewise.
706 * fold-const.c (operand_equal_p): Likewise.
707 (fold_ternary_loc): Likewise.
708 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
709 * gimple.c (DEFTREECODE): Likewise.
710 * gimplify.c (gimplify_expr): Likewise.
711 * optabs-tree.c (optab_for_tree_code): Likewise.
712 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
713 * tree-eh.c (operation_could_trap_p): Likewise.
714 (stmt_could_throw_1_p): Likewise.
715 * tree-inline.c (estimate_operator_cost): Likewise.
716 * tree-pretty-print.c (dump_generic_node): Likewise.
717 (op_code_prio): Likewise.
718 * tree-ssa-loop-im.c (stmt_cost): Likewise.
719 * tree-ssa-operands.c (get_expr_operands): Likewise.
720 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
721 * fold-const-call.h (fold_fma): Delete.
722 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
723 CFN_FNMA and CFN_FNMS.
725 * genmatch.c (combined_fn): New enum.
726 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
727 (commutative_op): New function.
728 (commutate): Use it. Handle more than 2 operands.
729 (dt_operand::gen_gimple_expr): Use commutative_op.
730 (parser::parse_expr): Allow :c to be used with non-binary
731 operators if the commutative operand is known.
732 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
733 CFN_FMS, CFN_FNMA and CFN_FNMS.
734 (backprop::process_assign_use): Remove FMA_EXPR handling.
735 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
736 (gen_hsa_fma): New function.
737 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
738 IFN_FNMA and IFN_FNMS.
739 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
740 * gimple-fold.h (follow_all_ssa_edges): Declare.
741 * gimple-fold.c (follow_all_ssa_edges): New function.
742 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
743 gimple_build interface and use follow_all_ssa_edges to fold the result.
744 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
745 instead of checking for optabs directly.
746 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
747 rather than FMA_EXPRs.
748 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
749 call to IFN_FMA instead of an FMA_EXPR.
751 2018-05-17 Jim Wilson <jimw@sifive.com>
753 * expr.c (do_tablejump): When converting index to Pmode, if we have a
754 sign extended promoted subreg, and the range does not have the sign bit
755 set, then do a sign extend.
757 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
758 test, check for sign extended subreg and/or constant operands, and
759 do a sign extend in that case.
761 2018-05-17 Steve Ellcey <sellcey@cavium.com>
763 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
764 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
766 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
767 Change logics_shift_reg to logics_shift_imm.
768 (thunderx2t99_fp_loadpair_basic): Delete.
769 (thunderx2t99_fp_storepair_basic): Delete.
770 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
771 (thunderx2t99_asimd_polynomial): Delete.
772 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
773 and neon_fp_mul_d_scalar_q.
774 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
775 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
776 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
777 (thunderx2t99_asimd_lut): Add missing tbl types.
778 (thunderx2t99_asimd_ext): Delete.
779 (thunderx2t99_asimd_load1_1_mult): Delete.
780 (thunderx2t99_asimd_load1_2_mult): Delete.
781 (thunderx2t99_asimd_load1_ldp): New.
782 (thunderx2t99_asimd_load1): New.
783 (thunderx2t99_asimd_load2): Add missing *load2* types.
784 (thunderx2t99_asimd_load3): New.
785 (thunderx2t99_asimd_load4): New.
786 (thunderx2t99_asimd_store1_1_mult): Delete.
787 (thunderx2t99_asimd_store1_2_mult): Delete.
788 (thunderx2t99_asimd_store2_mult): Delete.
789 (thunderx2t99_asimd_store2_onelane): Delete.
790 (thunderx2t99_asimd_store_stp): New.
791 (thunderx2t99_asimd_store1): New.
792 (thunderx2t99_asimd_store2): New.
793 (thunderx2t99_asimd_store3): New.
794 (thunderx2t99_asimd_store4): New.
796 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
798 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
799 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
801 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
802 Segher Boessenkool <segher@kernel.crashing.org>
805 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
808 2018-05-17 Richard Biener <rguenther@suse.de>
810 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
811 for pruning loop and prune defs feeding only already visited PHIs.
813 2018-05-17 Richard Biener <rguenther@suse.de>
815 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
817 2018-05-17 Bin Cheng <bin.cheng@arm.com>
818 Richard Biener <rguenther@suse.de>
820 PR tree-optimization/85793
821 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
822 for VMAT_ELEMENTWISE.
824 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
826 * internal-fn.h (lookup_internal_fn): Declare
827 * internal-fn.c (lookup_internal_fn): New function.
828 * gimple.c (gimple_build_call_from_tree): Handle calls to
830 * gimple-pretty-print.c (dump_gimple_call): Print "." before
831 internal function names.
832 * tree-pretty-print.c (dump_generic_node): Likewise.
833 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
835 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
837 * gimple-fold.h (gimple_build): Make the function forms take
838 combined_fn rather than built_in_function.
839 (gimple_simplify): Likewise.
840 * gimple-match-head.c (gimple_simplify): Likewise.
841 * gimple-fold.c (gimple_build): Likewise.
842 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
843 rather than gimple_build_call_internal.
844 (get_initial_defs_for_reduction): Likewise.
845 (vect_create_epilog_for_reduction): Likewise.
846 (vectorizable_live_operation): Likewise.
848 2018-05-17 Martin Liska <mliska@suse.cz>
850 * gimple-ssa-sprintf.c (format_directive): Do not use
851 space in between 'G_' and '('.
853 2018-05-17 Jakub Jelinek <jakub@redhat.com>
856 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
857 even if the mask is not all ones.
860 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
862 (ix86_gimple_fold_builtin): Likewise.
865 * config/i386/i386.c: Include tree-vector-builder.h.
866 (ix86_vector_shift_count): New function.
867 (ix86_fold_builtin): Fold shift builtins by scalar count.
868 (ix86_gimple_fold_builtin): Likewise.
870 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
871 _mm512_setzero): New intrinsics.
873 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
874 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
876 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
877 code generation for cases where splatting a value is not useful.
878 * simplify-rtx.c (simplify_ternary_operation): Simplify
879 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
880 mode to a vec_concat.
882 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
884 * config.gcc: Support "goldmont-plus".
885 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
887 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
888 PROCESSOR_GOLDMONT_PLUS.
889 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
890 (processor_target_table): Add "goldmont-plus".
891 (PTA_GOLDMONT_PLUS): Define.
892 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
893 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
894 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
895 (fold_builtin_cpu): Add "goldmont-plus".
896 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
897 (ix86_option_override_internal): Add "goldmont-plus".
898 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
899 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
900 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
901 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
903 2018-05-17 Richard Biener <rguenther@suse.de>
905 PR tree-optimization/85757
906 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
907 remove defs that only feed that PHI from further processing.
909 2018-05-16 Jim Wilson <jimw@sifive.com>
911 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
913 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
915 2018-05-16 Mark Wielaard <mark@klomp.org>
917 * dwarf2out.c (count_index_strings): New function.
918 (output_indirect_strings): Call count_index_strings and generate
919 header for dwarf_version >= 5.
921 2018-05-16 Mark Wielaard <mark@klomp.org>
923 * dwarf2out.c (dwarf_FORM): New function.
924 (set_indirect_string): Use dwarf_FORM.
925 (reset_indirect_string): Likewise.
926 (size_of_die): Likewise.
927 (value_format): Likewise.
928 (output_die): Likewise.
929 (add_skeleton_AT_string): Likewise.
930 (output_macinfo_op): Likewise.
931 (index_string): Likewise.
932 (output_index_string_offset): Likewise.
933 (output_index_string): Likewise.
934 (count_index_strings): Likewise.
936 2018-05-16 Carl Love <cel@us.ibm.com>
938 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
939 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
941 2018-05-16 Martin Jambor <mjambor@suse.cz>
943 * ipa-prop.c (ipa_free_all_edge_args): Remove.
944 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
946 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
948 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
949 (fnma<mode>4): Likewise.
950 (fms<mode>4): Likewise.
951 (fnms<mode>4): Likewise.
952 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
953 (aarch64_fnma<mode>4): Likewise.
954 (aarch64_fms<mode>4): Likewise.
955 (aarch64_fnms<mode>4): Likewise.
956 (aarch64_fnmadd<mode>4): Likewise.
958 2018-05-16 Jason Merrill <jason@redhat.com>
960 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
962 2018-05-16 Richard Biener <rguenther@suse.de>
964 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
965 (dump_stmt_cost): Declare.
966 (add_stmt_cost): Dump cost we add.
967 (add_stmt_costs): New function.
968 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
970 (vect_analyze_stmt): Adjust prototype.
971 (vectorizable_condition): Likewise.
972 (vectorizable_live_operation): Likewise.
973 (vectorizable_reduction): Likewise.
974 (vectorizable_induction): Likewise.
975 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
976 cost vector to pass to vectorizable_ and record afterwards.
977 (vect_model_reduction_cost): Take cost vector argument and adjust.
978 (vect_model_induction_cost): Likewise.
979 (vectorizable_reduction): Likewise.
980 (vectorizable_induction): Likewise.
981 (vectorizable_live_operation): Likewise.
982 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
983 SLP_TREE_NUMBER_OF_VEC_STMTS.
984 (vect_analyze_slp_cost_1): Remove.
985 (vect_analyze_slp_cost): Likewise.
986 (vect_slp_analyze_node_operations): Take visited args and
987 a target cost vector. Avoid processing already visited stmt sets.
988 (vect_slp_analyze_operations): Use a local cost vector to gather
989 costs and register those of non-discarded instances.
990 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
991 (vect_schedule_slp_instance): Remove copying of
992 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
994 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
995 adding cost. Record cost entry location.
996 (vect_prologue_cost_for_slp_op): Function to compute cost of
997 a constant or invariant generated for SLP vect in the prologue,
998 split out from vect_analyze_slp_cost_1.
999 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1000 (vect_model_promotion_demotion_cost): Likewise.
1001 (vect_model_store_cost): Likewise, make static.
1002 (vect_model_load_cost): Likewise.
1003 (vectorizable_bswap): Add cost vector arg and adjust.
1004 (vectorizable_call): Likewise.
1005 (vectorizable_simd_clone_call): Likewise.
1006 (vectorizable_conversion): Likewise.
1007 (vectorizable_assignment): Likewise.
1008 (vectorizable_shift): Likewise.
1009 (vectorizable_operation): Likewise.
1010 (vectorizable_store): Likewise.
1011 (vectorizable_load): Likewise.
1012 (vectorizable_condition): Likewise.
1013 (vectorizable_comparison): Likewise.
1014 (can_vectorize_live_stmts): Likewise.
1015 (vect_analyze_stmt): Likewise.
1016 (vect_transform_stmt): Adjust calls to vectorizable_*.
1017 * tree-vectorizer.c: Include gimple-pretty-print.h.
1018 (dump_stmt_cost): New function.
1020 2018-05-16 Richard Biener <rguenther@suse.de>
1022 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1023 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1024 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1025 (check_name): New callback.
1026 (dse_classify_store): Track cycles via a visited bitmap of PHI
1027 defs and simplify handling of in-loop and across loop dead stores
1028 and properly fail for loop-variant refs. Handle byte-tracking with
1029 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1032 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1034 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1035 (vect_get_mask_type_for_stmt): Likewise.
1036 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1038 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1039 to determine the statement's vector type and the vector type that
1040 should be used for calculating nunits. Deal with cases in which
1041 the type has to be deferred.
1042 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1043 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1044 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1045 (vect_determine_vf_for_stmt): New functions, split out from...
1046 (vect_determine_vectorization_factor): ...here.
1047 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1048 (vect_get_mask_type_for_stmt): New functions, split out from
1049 vect_determine_vectorization_factor.
1051 2018-05-16 Richard Biener <rguenther@suse.de>
1053 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1054 verify the [VEC_]COND_EXPR embedded comparison.
1056 2018-05-15 Martin Sebor <msebor@redhat.com>
1058 PR tree-optimization/85753
1059 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1060 RECORD_TYPE in addition to ARRAY_TYPE.
1062 2018-05-15 Martin Sebor <msebor@redhat.com>
1065 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1067 2018-05-15 Richard Biener <rguenther@suse.de>
1069 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1070 add by_clobber_p one. Change algorithm to collect all defs
1071 representing uses we need to walk and try reducing them to
1072 a single one before failing.
1073 (dse_dom_walker::dse_optimize_stmt): Adjust.
1075 2018-05-13 Mark Wielaard <mark@klomp.org>
1077 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1078 (size_of_loc_descr): Likewise.
1079 (output_loc_operands): Likewise.
1080 (output_loc_operands_raw): Likewise.
1081 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1082 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1083 (hash_loc_operands): Likewise.
1084 (compare_loc_operands): Likewise.
1086 2018-05-14 Mark Wielaard <mark@klomp.org>
1088 * dwarf2out.c (count_index_addrs): New function.
1089 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1091 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1093 PR tree-optimization/83648
1094 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1095 return value as malloc candidate.
1097 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1100 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1101 as true in call to suggest_attribute.
1103 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1105 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1106 -mreadonly-in-sdata.
1108 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1110 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1112 (aarch64_crypto_aesd_fused): Likewise.
1114 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1116 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1117 (movsi_aarch64): Likewise.
1118 (load_pairsi): Likewise.
1119 (load_pairdi): Likewise.
1120 (store_pairsi): Likewise.
1121 (store_pairdi): Likewise.
1122 (load_pairsf): Likewise.
1123 (load_pairdf): Likewise.
1124 (store_pairsf): Likewise.
1125 (store_pairdf): Likewise.
1126 (zero_extend): Likewise.
1127 (trunc): Swap alternatives.
1128 (fcvt_target): Add '?' to prefer w over r.
1130 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1133 * config/i386/i386.md: Disallow non-commutative arithmetics in
1134 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1135 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1136 in the peephole2 before it.
1138 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1140 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1141 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1142 (ix86_handle_option): Handle -mcldemote.
1143 * config.gcc: New header.
1144 * config/i386/cldemoteintrin.h: New file.
1145 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1146 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1148 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1149 OPTION_MASK_ISA_CLDEMOTE.
1150 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1151 (ix86_valid_target_attribute_inner_p): Ditto.
1152 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1153 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1154 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1155 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1156 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1158 * config/i386/i386.opt: Add -mcldemote.
1159 * config/i386/x86intrin.h: New header.
1160 * doc/invoke.texi: Add -mcldemote.
1162 2018-05-14 Richard Biener <rguenther@suse.de>
1164 * doc/match-and-simplify.texi: Adjust :s documentation.
1166 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1168 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1169 intended memcpy size.
1170 (REORDER_45): Likewise.
1172 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1174 * sort.cc: New file.
1175 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1176 * vec.c (qsort_chk): Use gcc_qsort.
1177 * Makefile.in (OBJS-libcommon): Add sort.o.
1178 (build/sort.o): New target. Use it...
1179 (BUILD_RTL): ... here, and...
1180 (build/gencfn-macros): ... here, and...
1181 (build/genmatch): ... here.
1183 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1184 Chung-Ju Wu <jasonwucj@gmail.com>
1186 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1187 * config/nds32/nds32-graywolf.md: New file.
1188 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1189 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1191 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1192 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1193 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1194 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1195 * config/nds32/pipelines.md: Include n15 settings.
1197 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1198 Chung-Ju Wu <jasonwucj@gmail.com>
1200 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1201 * config/nds32/nds32-n13.md: New file.
1202 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1203 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1205 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1206 * config/nds32/nds32.md (pipeline_model): Add n13.
1207 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1208 * config/nds32/pipelines.md: Include n13 settings.
1210 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1211 Chung-Ju Wu <jasonwucj@gmail.com>
1213 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1214 * config/nds32/nds32-n10.md: New file.
1215 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1216 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1218 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1219 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1220 * config/nds32/nds32.md (pipeline_model): Add n10.
1221 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1222 * config/nds32/pipelines.md: Include n10 settings.
1224 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1225 Kito Cheng <kito.cheng@gmail.com>
1226 Chung-Ju Wu <jasonwucj@gmail.com>
1228 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1229 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1230 Add enum values for DSP extension instructions.
1231 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1233 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1234 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1236 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1237 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1238 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1239 * config/nds32/nds32-intrinsic.md: Likewise.
1240 * config/nds32/nds32_intrinsic.h: Likewise.
1241 * config/nds32/nds32-md-auxiliary.c: Likewise.
1242 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1243 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1244 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1245 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1246 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1247 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1248 TYPE_DMAC in switch statement.
1249 * config/nds32/nds32.c: New checking and implementation for DSP
1250 extension instructions.
1251 * config/nds32/nds32.h: Likewise.
1252 * config/nds32/nds32.md: Likewise.
1253 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1254 * config/nds32/predicates.md: Implement new predicates for DSP
1257 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1259 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1260 Reformat alternatives and attributes so it is easier to identify
1261 which constraints/attributes go with which instruction.
1262 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1263 (mov<mode>_softfloat32, FMOVE64): Likewise.
1264 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1265 (mov<mode>_softfloat64, FMOVE64): Likewise.
1267 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1269 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1271 (Basic PowerPC Built-in Functions): The new name of the
1272 subsection previously known as "PowerPC Built-in Functions".
1273 (Basic PowerPC Built-in Functions Available on all Configurations):
1275 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1276 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1277 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1278 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1280 2018-05-11 Martin Jambor <mjambor@suse.cz>
1283 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1286 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1289 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1291 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1293 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1294 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1295 (ix86_handle_option): Handle -mwaitpkg.
1296 * config.gcc: New header.
1297 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1298 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1299 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1301 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1302 OPTION_MASK_ISA_WAITPKG.
1303 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1304 (ix86_option_override_internal): Add PTA_WAITPKG.
1305 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1306 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1307 IX86_BUILTIN_TPAUSE.
1308 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1309 __builtin_ia32_umwait and __builtin_ia32_tpause.
1310 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1311 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1312 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1313 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1314 UNSPECV_TPAUSE): New.
1315 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1316 * config/i386/i386.opt: Add -mwaitpkg.
1317 * config/i386/waitpkgintrin.h: New file.
1318 * config/i386/x86intrin.h: New header.
1319 * doc/invoke.texi: Add -mwaitpkg.
1321 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1324 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1326 (cortex-m0): Use armv6s-m isa.
1327 (cortex-m0plus): Likewise.
1328 (cortex-m1): Likewise.
1329 (cortex-m0.small-multiply): Likewise.
1330 (cortex-m0plus.small-multiply): Likewise.
1331 (cortex-m1.small-multiply): Likewise.
1333 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1334 Jakub Jelinek <jakub@redhat.com>
1336 PR tree-optimization/85692
1337 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1338 source permute as well.
1340 2018-05-11 Martin Liska <mliska@suse.cz>
1343 * doc/extend.texi: Document LLVM style format for no_sanitize
1346 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1348 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1349 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1350 (mode_supports_vsx_dform_quad): Likewise.
1351 (mode_supports_vmx_dform): Move these functions to be next to the
1352 other mode_supports functions.
1353 (mode_supports_dq_form): Likewise.
1354 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1355 mode_supports_dq_form.
1356 (reg_offset_addressing_ok_p): Likewise.
1357 (offsettable_ok_by_alignment): Likewise.
1358 (rs6000_legitimate_offset_address_p): Likewise.
1359 (legitimate_lo_sum_address_p): Likewise.
1360 (rs6000_legitimize_address): Likewise.
1361 (rs6000_legitimize_reload_address): Likewise.
1362 (rs6000_secondary_reload_inner): Likewise.
1363 (rs6000_preferred_reload_class): Likewise.
1364 (rs6000_output_move_128bit): Likewise.
1366 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1368 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1369 Generate SImode target register for null target.
1370 <case IX86_BUILTIN_XGETBV>: Ditto.
1371 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1372 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1374 2018-05-10 Carl Love <cel@us.ibm.com>
1376 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1377 dcbtt and dcbtstt if operands[2] is 0.
1379 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1382 * config/i386/sse.md (usadv64qi): New expander.
1384 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1386 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1387 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1388 -maltivec=be support.
1389 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1390 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1391 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1392 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1393 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1394 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1395 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1396 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1397 altivec_vsumsws): Adjust.
1398 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1399 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1400 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1402 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1403 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1404 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1405 (altivec_lve<VI_char>x): Delete expand.
1406 (*altivec_lve<VI_char>x_internal): Rename to...
1407 (altivec_lve<VI_char>x): ... this.
1408 (altivec_lvxl_<mode>): Delete expand.
1409 (*altivec_lvxl_<mode>_internal): Rename to ...
1410 (altivec_lvxl_<mode>): ... this.
1411 (altivec_stvxl_<mode>): Delete expand.
1412 (*altivec_stvxl_<mode>_internal): Rename to ...
1413 (altivec_stvxl_<mode>): ... this.
1414 (altivec_stve<VI_char>x): Delete expand.
1415 (*altivec_stve<VI_char>x_internal): Rename to ...
1416 (altivec_stve<VI_char>x): ... this.
1417 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1418 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1419 reduc_plus_scal_<mode>): Adjust.
1420 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1422 (rs6000_cpu_cpp_builtins): Adjust.
1423 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1424 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1425 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1426 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1427 -maltivec=be support.
1428 (rs6000_split_vec_extract_var): Adjust.
1429 (rs6000_split_v4si_init): Adjust.
1430 (swap_selector_for_mode): Delete.
1431 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1432 altivec_expand_stvex_be): Delete.
1433 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1434 -maltivec=be support.
1435 (rs6000_gimple_fold_builtin): Ditto.
1436 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1438 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1439 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1440 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1441 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1442 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1443 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1444 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1445 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1446 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1447 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1448 anonymous split): Adjust.
1449 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1450 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1452 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1454 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1455 when --with-gxx-include-dir is also specified.
1456 * configure: Regenerate.
1458 2018-05-09 Jim Wilson <jimw@sifive.com>
1461 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1462 * config/riscv/t-withmultilib: New.
1463 * config/riscv/withmultilib.h: New.
1464 * doc/install.texi: Document RISC-V --with-multilib-list support.
1466 2018-05-09 Richard Biener <rguenther@suse.de>
1468 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1470 (vect_bb_vectorization_profitable_p): Adjust. Compute
1471 actual scalar cost using the cost vector and the add_stmt_cost
1474 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1476 PR rtl-optimization/85645
1477 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1478 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1480 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1482 PR rtl-optimization/85645
1483 * shrink-wrap.c (spread_components): Return a boolean saying if
1484 anything was changed.
1485 (try_shrink_wrapping_separate): Iterate spread_components until
1486 nothing changes anymore.
1488 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1490 PR rtl-optimization/85645
1491 * regrename.c (build_def_use): Also kill the chains that include the
1492 destination of a REG_CFA_REGISTER note.
1494 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1496 PR rtl-optimization/85645
1497 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1498 insn that has a REG_CFA_REGISTER note.
1500 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1502 * cfgexpand.c (expand_clobber): New function.
1503 (expand_gimple_stmt_1): Use it.
1504 * tree-vect-stmts.c (vect_clobber_variable): New function,
1506 (vectorizable_simd_clone_call): ...here.
1507 (vectorizable_store): Emit a clobber either side of an
1508 IFN_STORE_LANES sequence.
1509 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1511 2018-05-09 Tom de Vries <tom@codesourcery.com>
1514 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1515 (define_insn "trap_if_false"): Add exit after trap.
1517 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1519 PR rtl-optimization/85638
1520 * bb-reorder.c: Include common/common-target.h.
1521 (create_forwarder_block): New function extracted from...
1522 (fix_up_crossing_landing_pad): ...here. Rename into...
1523 (dw2_fix_up_crossing_landing_pad): ...this.
1524 (sjlj_fix_up_crossing_landing_pad): New function.
1525 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1526 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1527 from both partitions and exit the loop after one iteration.
1529 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1532 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1534 (Basic PowerPC Built-in Functions): The new name of the
1535 subsection previously known as "PowerPC Built-in Functions".
1536 (Basic PowerPC Built-in Functions Available on all Configurations):
1538 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1540 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1542 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1544 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1547 2018-05-08 Jim Wilson <jimw@sifive.com>
1549 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1550 (LD_EMUL_SUFFIX): New.
1551 (LINK_SPEC): Use it.
1553 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1555 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1557 (Basic PowerPC Built-in Functions): The new name of the
1558 subsection previously known as "PowerPC Built-in Functions".
1559 (Basic PowerPC Built-in Functions Available on all Configurations):
1561 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1563 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1565 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1567 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1570 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1573 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1574 after cmpelim optimization.
1576 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1578 * config.gcc: Support "goldmont".
1579 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1580 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1582 * config/i386/i386.c (m_GOLDMONT): Define.
1583 (processor_target_table): Add "goldmont".
1584 (PTA_GOLDMONT): Define.
1585 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1586 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1587 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1588 (fold_builtin_cpu): Add "goldmont".
1589 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1590 (ix86_option_override_internal): Add "goldmont".
1591 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1592 (processor_type): Add PROCESSOR_GOLDMONT.
1593 * config/i386/i386.md: Add CPU "glm".
1594 * config/i386/glm.md: New file.
1595 * config/i386/x86-tune.def: Add m_GOLDMONT.
1596 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1598 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1601 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1603 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1604 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1605 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1608 * config/i386/i386.c (ix86_fold_builtin): Handle
1609 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1612 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1613 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1615 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1618 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1619 (check_arch): Likewise.
1620 (check_fpu): Return the result rather than printing it.
1621 (end arch): Fix operator precedence.
1622 (end cpu): Likewise.
1623 (END): Print the result from check_fpu.
1625 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1626 Alan Hayward <alan.hayward@arm.com>
1627 David Sherwood <david.sherwood@arm.com>
1629 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1630 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1631 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1632 (*fcmuo<mode>_and): New patterns.
1634 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1636 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1637 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1638 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1639 (cmp_op, sve_imm_con): New code attributes.
1640 (SVE_COND_INT_CMP, imm_con): Delete.
1641 (cmp_op): Remove above unspecs from int attribute.
1642 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1644 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1645 comparison-specific unspecs.
1646 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1647 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1648 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1649 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1650 (*vec_fcm<cmp_op><mode>): Rename to...
1651 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1652 (*vec_fcmuo<mode>): Rename to...
1653 (*fcmuo<mode>): ...this and adjust likewise.
1654 (*pred_fcm<cmp_op><mode>): New pattern.
1655 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1656 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1658 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1660 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1661 (aarch64_emit_sve_predicated_cond): New function.
1662 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1663 (aarch64_emit_unspec_cond_or): Replace with...
1664 (aarch64_emit_sve_or_conds): ...this new function. Use
1665 aarch64_emit_sve_ptrue_op for the individual comparisons and
1666 aarch64_emit_binop to OR them together.
1667 (aarch64_emit_inverted_unspec_cond): Replace with...
1668 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1669 aarch64_emit_sve_ptrue_op for the comparison and
1670 aarch64_emit_unop to invert the result.
1671 (aarch64_expand_sve_vec_cmp_float): Update after the above
1672 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1674 2018-05-07 Nathan Sidwell <nathan@acm.org>
1676 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1677 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1678 (Backwards Compatibility): Likewise.
1680 2018-05-07 Luis Machado <luis.machado@linaro.org>
1684 2018-05-07 Luis Machado <luis.machado@linaro.org>
1686 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1687 <prefetch_dynamic_strides>: New const bool field.
1688 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1689 prefetch_dynamic_strides.
1690 (exynosm1_prefetch_tune): Likewise.
1691 (thunderxt88_prefetch_tune): Likewise.
1692 (thunderx_prefetch_tune): Likewise.
1693 (thunderx2t99_prefetch_tune): Likewise.
1694 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1695 (aarch64_override_options_internal): Update to set
1696 PARAM_PREFETCH_DYNAMIC_STRIDES.
1697 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1698 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1699 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1700 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1701 prefetch-dynamic-strides setting.
1703 2018-05-07 Luis Machado <luis.machado@linaro.org>
1705 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1706 <minimum_stride>: New const int field.
1707 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1708 minimum_stride field.
1709 (exynosm1_prefetch_tune): Likewise.
1710 (thunderxt88_prefetch_tune): Likewise.
1711 (thunderx_prefetch_tune): Likewise.
1712 (thunderx2t99_prefetch_tune): Likewise.
1713 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1714 (aarch64_override_options_internal): Update to set
1715 PARAM_PREFETCH_MINIMUM_STRIDE.
1716 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1717 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1718 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1719 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1720 stride is constant and is below the minimum stride threshold.
1722 2018-05-07 Luis Machado <luis.machado@linaro.org>
1724 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1727 2018-05-07 Luis Machado <luis.machado@linaro.org>
1729 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1730 <prefetch_dynamic_strides>: New const bool field.
1731 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1732 prefetch_dynamic_strides.
1733 (exynosm1_prefetch_tune): Likewise.
1734 (thunderxt88_prefetch_tune): Likewise.
1735 (thunderx_prefetch_tune): Likewise.
1736 (thunderx2t99_prefetch_tune): Likewise.
1737 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1738 (aarch64_override_options_internal): Update to set
1739 PARAM_PREFETCH_DYNAMIC_STRIDES.
1740 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1741 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1742 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1743 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1744 prefetch-dynamic-strides setting.
1746 2018-05-07 Luis Machado <luis.machado@linaro.org>
1748 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1749 <minimum_stride>: New const int field.
1750 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1751 minimum_stride field.
1752 (exynosm1_prefetch_tune): Likewise.
1753 (thunderxt88_prefetch_tune): Likewise.
1754 (thunderx_prefetch_tune): Likewise.
1755 (thunderx2t99_prefetch_tune): Likewise.
1756 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1757 (aarch64_override_options_internal): Update to set
1758 PARAM_PREFETCH_MINIMUM_STRIDE.
1759 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1760 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1761 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1762 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1763 stride is constant and is below the minimum stride threshold.
1765 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1768 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1769 the type is addressable. Don't force op into register if it has
1772 2018-05-05 Roland McGrath <mcgrathr@google.com>
1775 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1776 any section for which we don't know a specific type it should have,
1777 regardless of name. Previously this was done only for the exact
1778 names ".init_array", ".fini_array", and ".preinit_array".
1779 (default_elf_asm_named_section): Add comment about
1780 relationship with default_section_type_flags and SECTION_NOTYPE.
1781 (get_section): Don't consider it a type conflict if one side has
1782 SECTION_NOTYPE and the other doesn't, as long as neither has the
1783 SECTION_BSS et al used in the default_section_type_flags logic.
1785 2018-05-05 Tom de Vries <tom@codesourcery.com>
1788 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1789 (workaround_barsyncs): New function.
1790 (nvptx_reorg): Use workaround_barsyncs.
1791 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1792 (define_expand "nvptx_membar_cta"): New define_expand.
1793 (define_insn "*nvptx_membar_cta"): New insn.
1795 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1797 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1798 To improve optimization opportunities.
1799 * builtin-types.def: The new needed builtin types for the above.
1801 2018-05-04 Richard Biener <rguenther@suse.de>
1803 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1804 * gimple-ssa-store-merging.c
1805 (imm_store_chain_info::output_merged_store): Remove redundant create,
1806 release split_store vector contents on failure.
1807 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1808 scalar stmt vector on cache hit.
1810 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1812 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1814 * config.gcc (powerpc-xilinx-eabi*): Remove.
1815 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1817 (fusion_addis_mem_combo_load): Ditto.
1818 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1820 (rs6000_cpu_cpp_builtins): Ditto.
1821 * config/rs6000/rs6000-linux.c
1822 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1823 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1824 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1826 (rs6000_setup_reg_addr_masks): Ditto.
1827 (rs6000_init_hard_regno_mode_ok): Ditto.
1828 (rs6000_option_override_internal): Ditto.
1829 (legitimate_lo_sum_address_p): Ditto.
1830 (rs6000_legitimize_address): Ditto.
1831 (rs6000_legitimize_reload_address): Ditto.
1832 (rs6000_legitimate_address_p): Ditto.
1833 (abi_v4_pass_in_fpr): Ditto.
1834 (setup_incoming_varargs): Ditto.
1835 (rs6000_gimplify_va_arg): Ditto.
1836 (rs6000_split_multireg_move): Ditto.
1837 (rs6000_savres_strategy): Ditto.
1838 (rs6000_emit_prologue_components): Ditto.
1839 (rs6000_emit_epilogue_components): Ditto.
1840 (rs6000_emit_prologue): Ditto.
1841 (rs6000_emit_epilogue): Ditto.
1842 (rs6000_elf_file_end): Ditto.
1843 (rs6000_function_value): Ditto.
1844 (rs6000_libcall_value): Ditto.
1845 * config/rs6000/rs6000.h: Ditto.
1846 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1847 (TARGET_MINMAX): ... this. New.
1848 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1849 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1850 (*movsi_internal1_single): Delete.
1851 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1852 mfpu=, mxilinx-fpu): Delete.
1853 * config/rs6000/singlefp.h: Delete.
1854 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1855 * config/rs6000/t-rs6000: Ditto.
1856 * config/rs6000/t-xilinx: Delete.
1857 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1858 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1859 (VStype_simple): Delete.
1860 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1861 * config/rs6000/xfpu.h: Delete.
1862 * config/rs6000/xfpu.md: Delete.
1863 * config/rs6000/xilinx.h: Delete.
1864 * config/rs6000/xilinx.opt: Delete.
1865 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1866 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1868 2018-05-04 Tom de Vries <tom@codesourcery.com>
1871 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1874 2018-05-04 Richard Biener <rguenther@suse.de>
1877 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1878 (expand_complex_div_wide): Likewise.
1879 (expand_complex_operations_1): Likewise.
1880 (expand_complex_libcall): Preserve EH info of the original stmt.
1881 (tree_lower_complex): Handle removed blocks.
1882 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1883 on complex multiplication and division libcall builtins.
1885 2018-05-04 Richard Biener <rguenther@suse.de>
1888 * fold-const.c (negate_expr_p): Restrict negation of operand
1889 zero of a division to when we know that can happen without
1891 (fold_negate_expr_1): Likewise.
1893 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1896 * real.h (real_nextafter): Declare.
1897 * real.c (real_nextafter): New function.
1898 * fold-const-call.c (fold_const_nextafter): New function.
1899 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1900 CASE_CFN_NEXTTOWARD.
1901 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1902 even when arg1_mode is different from arg0_mode.
1904 2018-05-03 Nathan Sidwell <nathan@acm.org>
1906 * doc/extend.texi (Deprecated Features): Remove
1908 (Backwards Compatibility): Likewise.
1909 * doc/invoke.texi (C++ Language Options): Likewise.
1910 (C++ Dialect Options): Likewise.
1912 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1915 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1916 _mm512_mask_mullox_epi64): New intrinsics.
1918 2018-05-03 Tom de Vries <tom@codesourcery.com>
1921 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1922 dump files): Add offload-tree.
1924 2018-05-03 Richard Biener <rguenther@suse.de>
1926 PR tree-optimization/85615
1927 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1928 to loops not nested in BBs loop father to avoid creating multi-entry
1931 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1933 PR tree-optimization/70291
1934 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1935 arguments. Change return type to tree. Emit libcall as a new
1936 statement rather than replacing existing one when inplace_p is true.
1937 (expand_complex_multiplication_components): New function.
1938 (expand_complex_multiplication): Expand floating-point complex
1939 multiplication using the above.
1940 (expand_complex_division): Rename inner_type parameter to type.
1941 Update expand_complex_libcall call-site.
1942 (expand_complex_operations_1): Update expand_complex_multiplication
1943 and expand_complex_division call-sites.
1945 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1948 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1949 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1950 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1951 the highest significant bit of the shift count mask is clear. In
1952 check whether and[sq]i3 is needed verify that all significant bits
1953 of the shift count other than the highest are set.
1955 2018-05-02 Tom de Vries <tom@codesourcery.com>
1958 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1959 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1960 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1961 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1962 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1963 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1964 __builtin_goacc_parlevel_size.
1966 2018-05-02 Richard Biener <rguenther@suse.de>
1968 PR tree-optimization/85597
1969 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1970 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1973 2018-05-02 Tom de Vries <tom@codesourcery.com>
1976 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1977 dump files): Add ltrans-tree.
1979 2018-05-02 Tom de Vries <tom@codesourcery.com>
1982 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1983 dump files): Add wpa-ipa.
1985 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1987 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1988 powerpc*-*-linux*paired* target.
1989 * config/rs6000/750cl.h: Delete.
1990 * config/rs6000/paired.h: Delete.
1991 * config/rs6000/paired.md: Delete.
1992 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1994 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1995 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1996 comment. Remove paired float support.
1997 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1998 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2000 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2001 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2003 * config/rs6000/rs6000.c: Remove paired float support.
2004 (paired_expand_vector_init, paired_expand_vector_move,
2005 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2006 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2007 paired_expand_builtin, paired_expand_predicate_builtin,
2008 paired_init_builtins): Delete.
2009 * config/rs6000/rs6000.h: Remove paired float support.
2010 * config/rs6000/rs6000.md: Remove paired float support.
2011 (move_from_CR_ov_bit): Delete.
2012 * config/rs6000/rs6000.opt (mpaired): Delete.
2013 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2014 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2016 2018-05-02 Richard Biener <rguenther@suse.de>
2019 * gimplify.c (gimplify_save_expr): When in SSA form allow
2020 SAVE_EXPRs to compute to SSA vars.
2022 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2025 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2026 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2027 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2028 clobber operands[2], instead use a new pseudo. Formatting fixes.
2030 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2032 PR tree-optimization/85586
2033 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2034 exit early for statements in the same group if the accesses are
2037 2018-05-02 Tom de Vries <tom@codesourcery.com>
2040 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2043 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2045 PR tree-optimization/85143
2046 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2048 2018-05-01 Tom de Vries <tom@codesourcery.com>
2051 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2052 not found" error message.
2054 2018-05-01 Tom de Vries <tom@codesourcery.com>
2057 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2058 * vec.c (test_ordered_remove_if): New function.
2059 (vec_c_tests): Call test_ordered_remove_if.
2060 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2061 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2062 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2063 VEC_ORDERED_REMOVE_IF.
2065 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2067 PR tree-optimization/82665
2068 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2069 pointer subtraction where arguments come from a memchr call.
2071 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2073 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2074 --push-state --as-needed and --pop-state instead of --as-needed and
2075 --no-as-needed if ld supports it.
2076 * configure: Regenerated.
2079 * doc/install.texi2html: Replace _002d with - and _002a with * in
2080 generated html files using sed.
2082 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2085 * gcc-rich-location.c (blank_line_before_p): New function.
2086 (use_new_line): New function.
2087 (gcc_rich_location::add_fixit_insert_formatted): New function.
2088 * gcc-rich-location.h
2089 (gcc_rich_location::add_fixit_insert_formatted): New function.
2091 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2093 * selftest.c (assert_streq): Rename "expected" and "actual" to
2094 "val1" and "val2". Extend NULL-handling to cover both inputs
2095 symmetrically, while still requiring both to be non-NULL for a pass.
2096 * selftest.h (assert_streq): Rename "expected" and "actual" to
2098 (ASSERT_EQ): Likewise.
2099 (ASSERT_EQ_AT): Likewise.
2100 (ASSERT_KNOWN_EQ): Likewise.
2101 (ASSERT_KNOWN_EQ_AT): Likewise.
2102 (ASSERT_NE): Likewise.
2103 (ASSERT_MAYBE_NE): Likewise.
2104 (ASSERT_MAYBE_NE_AT): Likewise.
2105 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2106 the assertion to pass.
2107 (ASSERT_STREQ_AT): Likewise.
2109 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2111 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2112 interaction with -pie.
2114 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2116 * selftest.h: Fix alphabetization of per-source-file selftest
2119 2018-04-30 Jason Merrill <jason@redhat.com>
2121 PR c++/61982 - dead stores to destroyed objects.
2122 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2125 2018-04-30 Jason Merrill <jason@redhat.com>
2127 * tree.c (build_clobber): New.
2128 * tree.h: Declare it.
2129 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2131 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2133 * diagnostic-show-locus.c (layout::layout): Update for
2134 location_get_source_line returning a char_span.
2135 (struct char_span): Move to input.h.
2136 (struct correction): Update for fields in char_span becoming
2138 (struct source_line): Update for location_get_source_line
2139 returning a char_span.
2140 (layout::print_line): Likewise.
2141 * edit-context.c (edited_file::print_content): Likewise.
2142 (edited_file::print_diff_hunk): Likewise.
2143 (edited_file::print_run_of_changed_lines): Likewise.
2144 (edited_file::get_num_lines): Likewise.
2145 (edited_line::edited_line): Likewise.
2146 * final.c (asm_show_source): Likewise.
2147 * input.c (location_get_source_line): Convert return type
2148 from const char * to char_span, losing the final "line_len"
2150 (dump_location_info): Update for the above.
2151 (get_substring_ranges_for_loc): Likewise. Use a char_span
2152 when handling the literal within the line.
2153 (test_reading_source_line): Update for location_get_source_line
2154 returning a char_span.
2155 * input.h (class char_span): Move here from
2156 diagnostic-show-locus.c, converting from a struct to a class.
2157 Make data members private.
2158 (char_span::operator bool): New.
2159 (char_span::length): New.
2160 (char_span::get_buffer): New.
2161 (char_span::operator[]): New.
2162 (char_span::subspan): Make const.
2163 (char_span::xstrdup): New.
2164 (location_get_source_line): Convert return type from const char *
2165 to char_span, losing the final "line_size" param.
2167 2018-04-30 Jan Hubicka <jh@suse.cz>
2169 * lto-wrapper.c (ltrans_priorities): New static var.
2170 (cmp_priority): New.
2171 (run_gcc): Read priorities and if doing parallel build order
2172 the Makefile by them.
2174 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2176 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2178 2018-04-30 Richard Biener <rguenther@suse.de>
2180 * tree-cfg.c (verify_address): Remove base argument, add
2181 flag whether to check TREE_ADDRESSABLE and do that.
2182 (verify_expr): Remove.
2183 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2184 (verify_gimple_assign_single): Likewise.
2185 (verify_gimple_switch): Likewise.
2186 (verify_expr_location_1): Dereference tp once. Add (disabled)
2187 piece from verify_expr.
2188 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2190 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2192 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2194 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2196 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2197 (small_data_pattern): Likewise.
2198 (arc_rewrite_small_data): Likewise.
2199 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2200 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2201 (get_symbol_alignment): New function.
2202 (legitimate_small_data_address_p): Likewise.
2203 (legitimate_scaled_address): Update, call
2204 legitimate_small_data_address_p.
2205 (output_sdata): New static variable.
2206 (arc_print_operand): Update how we handle small data operands.
2207 (arc_print_operand_address): Likewise.
2208 (arc_legitimate_address_p): Update, use
2209 legitimate_small_data_address_p.
2210 (arc_rewrite_small_data_p): Remove.
2211 (arc_rewrite_small_data_1): Likewise.
2212 (arc_rewrite_small_data): Likewise.
2213 (small_data_pattern): Likewise.
2214 (compact_sda_memory_operand): Update to use
2215 legitimate_small_data_address_p and get_symbol_alignment.
2216 (prepare_move_operands): Don't rewite sdata pattern.
2217 (prepare_extend_operands): Remove.
2218 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2220 (zero_extendqisi2): Likewise.
2221 (zero_extendhisi2): Likewise.
2222 (extendqihi2): Likewise.
2223 (extendqisi2): Likewise.
2224 (extendhisi2): Likewise.
2228 * config/arc/constraints.md (Usd): Change it to memory constraint.
2230 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2232 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2233 as source of std instructions.
2234 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2235 allow 6-bit constants as source for store instructions.
2236 (movdi_insn): Update instruction pattern to allow 6-bit constants
2237 as source for store instructions.
2239 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2241 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2243 2018-04-30 Nathan Sidwell <nathan@acm.org>
2244 Sandra Loosemore <sandra@codesourcery.com>
2246 * dumpfile.c (dump_open): Allow '-' for stdout.
2247 * doc/invoke.texi (Developer Options): Document dump filename
2248 determination early. Document stdin/stdout selection.
2250 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2252 Microblaze Target: PIC data text relative
2254 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2255 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2257 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2258 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2259 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2260 New addressing mode for data-text relative position indepenedent code.
2261 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2262 'ADDRESS_SYMBOLIC_TXT_REL'.
2263 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2264 (microblaze_legitimate_pic_operand): Exclude function calls from
2265 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2266 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2268 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2269 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2270 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2271 for 'address + offset'.
2272 (microblaze_expand_prologue): Add new function prologue call for
2274 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2275 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2276 table in case of TARGET_PIC_DATA_TEXT_REL.
2277 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2278 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2279 Add new macros 'UNSPEC_TEXT',
2280 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2281 + exclude function calls from 'UNSPEC_PLT' in case of data text
2283 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2284 new target hook for generating address diff vector tables in case of
2286 * doc/tm.texi : Regenerate.
2287 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2288 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2289 of addr diff vector generation.
2290 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2291 target hook definition.
2292 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2293 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2294 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2295 option for data text relative.
2297 2018-04-30 Richard Biener <rguenther@suse.de>
2299 * tree-chrec.h (evolution_function_is_constant_p): Remove
2301 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2303 2018-04-30 Richard Biener <rguenther@suse.de>
2306 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2308 2018-04-30 Richard Biener <rguenther@suse.de>
2310 PR tree-optimization/28364
2311 PR tree-optimization/85275
2312 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2313 copying first exit test.
2315 2018-04-28 Mark Wielaard <mark@klomp.org>
2317 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2319 (dwarf_AT): Handle DW_AT_addr_base.
2320 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2322 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2325 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2326 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2327 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2328 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2330 2018-04-28 Richard Biener <rguenther@suse.de>
2332 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2333 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2334 to reflect use. Only add interesting stmts.
2336 2018-04-27 Martin Jambor <mjambor@suse.cz>
2339 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2340 the jump function allows for passing through aggregate values.
2342 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2344 * input.h (in_system_header_at): Convert from macro to inline
2346 (from_macro_expansion_at): Likewise.
2347 (from_macro_definition_at): Likewise.
2349 2018-04-27 Jeff Law <law@redhat.com>
2351 * config.gcc: Mark tile* targets as deprecated/obsolete.
2353 2018-04-27 Richard Biener <rguenther@suse.de>
2355 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2358 2018-04-27 Richard Biener <rguenther@suse.de>
2360 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2362 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2364 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2365 with Yd constraint. Set "preferred_for_speed" attribute from
2366 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2368 (*movdi_internal): Ditto.
2369 (movti_interunit splitters): Remove
2370 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2371 (movdi_interunit splitters): Ditto.
2372 * config/i386/constraints.md (Ye): Remove.
2373 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2375 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2378 * config/aarch64/constraints.md (Usg): Limit to 31.
2381 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2383 PR tree-optimization/85529
2384 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2385 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2386 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2387 zero extension or masking of the MSB bit.
2388 (optimize_range_tests): Add FIRST_BB argument, pass it through
2389 to optimize_range_tests_var_bound.
2390 (maybe_optimize_range_tests, reassociate_bb): Adjust
2391 optimize_range_tests callers.
2393 2018-04-26 Richard Biener <rguenther@suse.de>
2394 Jakub Jelinek <jakub@redhat.com>
2396 * cgraph.h (symbol_table): Just declare debug method here.
2397 * symtab.c (symbol_table::debug): Define.
2399 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2401 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2403 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2405 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2406 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2407 (*movdi_internal): Substitute Yi and Yj constraint with x
2408 and Ym and Yn constraint with y constraint. Update "isa"
2409 attribute and set "preferred_for_speed" attribute from
2410 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2411 (*movsi_internal): Ditto.
2412 (*movdf_internal): Ditto.
2413 (*movsf_internal): Ditto.
2414 (*zero_extendsidi2): Ditto.
2415 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2416 (sse2_loadld): Ditto.
2417 (*vec_extract<ssevecmodelower>_0): Ditto.
2418 (*vec_extractv4si_0_zext_sse4): Ditto.
2419 (vec_concatv2di): Ditto.
2420 (*vec_dup<mode>): Ditto.
2421 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2422 * config/i386/constraints.md (Yi): Remove.
2427 2018-04-26 Nathan Sidwell <nathan@acm.org>
2429 * dumpfile.c (dump_open): New.
2430 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2431 (dump_finish): Detect stdio/stderr by value not name.
2433 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2435 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2437 2018-04-26 Tom de Vries <tom@codesourcery.com>
2440 * config/nvptx/nvptx.c (verify_neutering_jumps)
2441 (verify_neutering_labels): New function
2442 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2444 2018-04-26 Tom de Vries <tom@codesourcery.com>
2447 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2448 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2451 2018-04-26 Richard Biener <rguenther@suse.de>
2452 Tom de Vries <tom@codesourcery.com>
2455 * lto-streamer-out.c (output_function): Fixup loops if required to match
2456 discovery done in the reader.
2458 2018-04-26 Richard Biener <rguenther@suse.de>
2460 PR tree-optimization/85116
2461 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2462 have a loop exit from the single latch predecessor. Remove
2463 case of header with just condition.
2464 (ch_base::copy_headers): Exclude infinite loops from any
2466 (pass_ch::execute): Record exits.
2468 2018-04-26 Richard Biener <rguenther@suse.de>
2470 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2471 prologue cost vector and pass it to vect_get_load_cost.
2472 (vect_get_peeling_costs_all_drs): Likewise.
2473 (vect_peeling_hash_get_lowest_cost): Likewise.
2474 (vect_enhance_data_refs_alignment): Likewise.
2476 2018-04-26 Richard Biener <rguenther@suse.de>
2479 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2480 checking of integer<->pointer conversions.
2481 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2482 sign-/zero-extending pointer types.
2483 (expand_omp_for_static_chunk): Likewise.
2485 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2486 Jean Lee <xiaoyur347@gmail.com>
2488 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2489 (TARGET_ASAN_SHADOW_OFFSET): Define.
2490 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2491 true for -fsanitize=address.
2493 2018-04-25 Mark Wielaard <mark@klomp.org>
2495 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2498 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2500 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2501 than "alu", remove explicit "memory" and "imm_disp" attributes.
2502 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2505 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2506 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2507 gen_lowpart_no_emit.
2509 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2512 * config/i386/i386.c (ix86_expand_builtin): Change memory
2513 operand to XI, extend p0 to Pmode.
2514 * config/i386/i386.md: Change unspec volatile and operand
2515 1 mode to XI, change operand 0 mode to P.
2517 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2519 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2520 GET_MODE_MASK before any checking.
2521 (nds32_can_use_bset_p): Likewise.
2522 (nds32_can_use_btgl_p): Likewise.
2524 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2526 * config/nds32/nds32-doubleword.md: New define_split pattern for
2527 illegal register number.
2529 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2531 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2533 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2535 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2537 2018-04-25 Richard Biener <rguenther@suse.de>
2539 * lto-streamer.h (LTO_major_version): Bump to 8.
2541 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2543 * BASE-VER: Set to 9.0.0.
2545 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2547 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2548 in __abskf2 and __powikf2.
2550 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2553 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2554 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2555 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2556 Use the above on operand 2. Reindent.
2557 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2559 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2562 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2564 * config/i386/i386.opt (mcet): Removed.
2565 * doc/install.texi: Remove -mcet documentation.
2566 * doc/invoke.texi: Likewise.
2568 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2571 * doc/install.texi: Remove -mcet from bootstrap-cet.
2573 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2576 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2577 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2581 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2582 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2583 containing a CONST_VECTOR.
2585 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2587 * doc/install.texi: Update newlib dependency for nvptx.
2589 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2592 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2593 instead of INTVAL when shifting x left.
2595 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2597 PR tree-optimization/85478
2598 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2599 vect_grouped_store_supported for single element vectors.
2601 2018-04-24 Richard Biener <rguenther@suse.de>
2604 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2605 load cost increase to the case of non-constant step.
2607 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2610 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2611 destination if any_malformed_asm.
2613 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2616 * expr.c (store_field): In the bitfield case, if the value comes from
2617 a function call and is returned in registers by means of a PARALLEL,
2618 do not change the mode of the temporary unless BLKmode and VOIDmode.
2620 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2622 PR rtl-optimization/85423
2623 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2624 dependencies to debug insns when the previous insn is non-debug.
2626 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2628 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2629 enums into a single definition.
2630 (fls): Fix predicates and printing.
2633 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2635 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2636 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2637 and short u6 immediate.
2638 (check_if_valid_sleep_operand): Remove.
2639 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2641 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2643 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2644 flag_always_save_lp condition.
2645 * config/nds32/nds32.opt (malways-save-lp): New option.
2647 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2649 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2650 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2651 * config/nds32/nds32.h
2652 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2653 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2655 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2657 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2658 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2660 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2661 Chung-Ju Wu <jasonwucj@gmail.com>
2663 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2665 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2666 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2668 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2670 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2672 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2674 * config/nds32/nds32-protos.h (nds32_data_alignment,
2675 nds32_local_alignment): Declare.
2676 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2677 nds32_local_alignment): New functions.
2678 (TARGET_CONSTANT_ALIGNMENT): Define.
2679 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2681 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2683 * config/nds32/nds32.c
2684 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2685 (TARGET_MODES_TIEABLE_P): Likewise.
2687 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2689 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2692 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2693 Chung-Ju Wu <jasonwucj@gmail.com>
2695 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2696 for unaligned access.
2697 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2699 * config/nds32/nds32-intrinsic.md: Likewise.
2700 * config/nds32/nds32_intrinsic.h: Likewise.
2701 * config/nds32/nds32.h (nds32_builtins): Likewise.
2702 * config/nds32/nds32.opt (munaligned-access): New option.
2703 * config/nds32/nds32.c (nds32_asm_file_start): Display
2704 flag_unaligned_access status.
2706 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2708 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2709 -mno-relax is present.
2710 * config/riscv/linux.h (LINK_SPEC): Ditto.
2712 2018-04-20 Martin Sebor <msebor@redhat.com>
2715 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2717 (gimple_fold_builtin_stxcpy_chk): Same.
2718 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2720 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2723 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2724 __powikf2 when long double is IEEE 128-bit.
2726 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2728 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2729 step to make sure stack always aligned.
2731 2018-04-20 Carl Love <cel@us.ibm.com>
2734 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2735 size check for arg0.
2737 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2738 Tom de Vries <tom@codesourcery.com>
2741 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2742 Emit insns for calls too.
2743 (nvptx_find_par): Always look for worker-level predecessor insn.
2744 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2746 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2747 (nvptx_process_pars): Propagate frames for calls.
2749 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2752 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2754 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2755 (ix86_handle_option): Don't handle OPT_mibt.
2756 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2758 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2760 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2761 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2762 (ix86_target_macros): Define __CET__ with flag_cf_protection
2763 for -fcf-protection.
2764 * config/i386/i386.c (isa2_opts): Remove -mibt.
2765 * config/i386/i386.h (TARGET_IBT): Removed.
2766 (TARGET_IBT_P): Likewise.
2767 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2768 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2769 * config/i386/i386.opt (mcet): Update help message.
2772 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2773 -mcet as an alias for -mshstk.
2775 2018-04-20 Richard Biener <rguenther@suse.de>
2778 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2779 complexity by forcing a single use of the multiply operand.
2781 2018-04-20 Martin Jambor <mjambor@suse.cz>
2784 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2785 recursion dependency to only apply to non-clones.
2787 2018-04-20 Martin Jambor <mjambor@suse.cz>
2790 * ipa-cp.c (create_specialized_node): Check that clones of
2791 self-recursive edges exist during IPA-CP.
2793 2018-04-19 Toon Moene <toon@moene.org>
2795 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2798 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2800 PR tree-optimization/85467
2801 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2802 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2803 VECTOR_CST element to type.
2805 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2808 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2809 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2810 (builtin_longjmp): Likewise.
2811 (save_stack_nonlocal): New pattern.
2812 (restore_stack_nonlocal): Likewise.
2814 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2817 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2818 Replace ASM_OUTPUT_LABEL with fprintf.
2820 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2823 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2824 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2825 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2826 define __IBT__ and __SHSTK__ for -fcf-protection.
2827 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2829 (ix86_trampoline_init): Likewise.
2830 (x86_output_mi_thunk): Likewise.
2831 (ix86_notrack_prefixed_insn_p): Likewise.
2832 (ix86_option_override_internal): Don't disallow -fcf-protection.
2833 * config/i386/i386.md (rdssp<mode>): Also enable for
2835 (incssp<mode>): Likewise.
2836 (nop_endbr): Likewise.
2837 * config/i386/i386.opt (mcet): Change help message to built-in
2841 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2842 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2843 enable CET built-in functions.
2845 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2847 * common/config/i386/i386-common.c
2848 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2849 OPTION_MASK_ISA_MOVDIRI_UNSET,
2850 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2851 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2852 * config.gcc (movdirintrin.h): New header.
2853 * config/i386/cpuid.h (bit_MOVDIRI,
2854 bit_MOVDIR64B): New bits.
2855 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2857 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2858 (VOID, PVOID, PCVOID)): New function types.
2859 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2860 __builtin_ia32_directstoreu_u64,
2861 __builtin_ia32_movdir64b): New builtins.
2862 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2863 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2865 (ix86_valid_target_attribute_inner_p): Ditto.
2866 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2867 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2868 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2869 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2870 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2871 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2872 (movdiri<mode>, movdir64b_<mode>): New.
2873 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2874 * config/i386/immintrin.h: Include movdirintrin.h.
2875 * config/i386/movdirintrin.h: New file.
2876 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2878 2018-04-19 Richard Biener <rguenther@suse.de>
2881 * cfg.c (clear_bb_flags): When loop state says we have
2882 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2884 2018-04-19 Richard Biener <rguenther@suse.de>
2886 PR tree-optimization/84737
2887 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2888 copying restrict info.
2889 (vect_setup_realignment): Use it.
2890 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2891 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2892 the first DR to all generated stores.
2893 (vectorizable_load): Likewise for loads.
2895 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2897 PR tree-optimization/85446
2898 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2899 the integral and pointer types to have the same precision.
2901 * doc/install.texi: Document --disable-cet being the default and
2904 2018-04-18 Martin Liska <mliska@suse.cz>
2906 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2909 2018-04-18 Martin Liska <mliska@suse.cz>
2912 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2915 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2916 arguments if they are comparable.
2918 2018-04-18 Martin Liska <mliska@suse.cz>
2921 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2924 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2927 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2930 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2931 ENDBR after calling __morestack.
2933 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2936 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2937 by using gcc_base_ver to generate a gcc_driver_version, and use
2938 it when generating GCC_DRIVER_NAME.
2939 * configure: Regenerate.
2941 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2944 * config.gcc: Obsolete powerpc*-*-*spe*.
2946 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2949 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2950 (stabstr_D): Change type of unum from unsigned int to
2951 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2954 2018-04-17 Jim Wilson <jimw@sifive.com>
2957 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2958 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2959 Set arg_pointer_offset after using pretend_args_size.
2961 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2963 PR rtl-optimization/85431
2964 * dse.c (record_store): Ignore zero width stores.
2967 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2968 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2969 __builtin_stack_restore rather than after it.
2970 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2971 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2972 argument instead of virtual_dynamic_stack_rtx.
2974 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2976 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2978 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2979 Add note to error message to explain internal mapping of overloaded
2980 built-in function name to non-overloaded built-in function name.
2981 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2984 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2987 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2988 where the inputs overlap with the output.
2990 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2992 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2993 (=v, v) alternative and explicit "memory" attribute.
2994 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2995 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2997 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2998 "sselog1" type instead of "sselog".
2999 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3000 "sselog". Remove explicit "memory" attribute.
3001 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3002 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3004 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3005 "sselog1" type instead of "sselog". Remove explicit "memory"
3007 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3008 use "sselog1" type instead of "sselog". Remove explicit "memory"
3010 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3011 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3013 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3014 "sselog1" type instead of "sselog". Remove explicit "memory"
3016 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3017 use "sselog1" type instead of "sselog". Remove explicit "memory"
3021 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3024 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3027 2018-04-17 Martin Jambor <mjambor@suse.cz>
3030 * ipa-cp.c (create_specialized_node): Call
3031 expand_all_artificial_thunks if necessary.
3033 2018-04-17 Martin Liska <mliska@suse.cz>
3036 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3037 in message, remote space in between '_G' and '('.
3039 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3042 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3043 avx512f_vmcmp<mode>3<round_saeonly_name>,
3044 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3045 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3046 avx512f_rndscale<mode><round_saeonly_name>,
3047 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3048 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3049 Use %<iptr>2 instead of %2 for -masm=intel.
3050 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3051 avx512f_vcvttss2usi<round_saeonly_name>,
3052 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3054 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3055 avx512f_vcvttsd2usi<round_saeonly_name>,
3056 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3057 Use %q1 instead of %1 for -masm=intel.
3058 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3059 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3060 of %3 for -masm=intel.
3061 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3063 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3065 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3067 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3068 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3070 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3072 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3073 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3074 %g1 and one with %0 and %1.
3075 (avx512er_vmrcp28<mode><round_saeonly_name>,
3076 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3078 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3079 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3080 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3081 of %0 and %{%4%} for -masm=intel.
3082 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3083 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3084 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3085 order of %0 and %{%5%}%{z%} for -masm=intel.
3087 2018-04-17 Jan Hubicka <jh@suse.cz>
3090 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3092 2018-04-17 Martin Liska <mliska@suse.cz>
3095 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3096 for target_clone error message. Make default implementation
3097 clone to be a local declaration.
3098 (separate_attrs): Add new argument and check for an empty
3100 (expand_target_clones): Handle it.
3101 (ipa_target_clone): Make redirection just for target_clones
3104 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3105 Tom de Vries <tom@codesourcery.com>
3108 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3109 tiled basic blocks without omp continue statements.
3111 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3114 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3115 vec_extract expression as having side effects to make sure it gets
3118 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3121 * config/i386/i386.c (get_builtin_code_for_version): Check
3124 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3127 * gcc/config.gcc: Support "skylake".
3128 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3130 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3131 (processor_target_table): Add "skylake".
3132 (ix86_option_override_internal): Add "skylake".
3133 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3134 PROCESSOR_CANNONLAKE.
3135 (get_builtin_code_for_version): Fix priority for
3136 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3137 PROCESSOR_SKYLAKE-AVX512.
3138 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3139 (processor_type): Add PROCESSOR_SKYLAKE.
3141 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3142 Jason Merrill <jason@redhat.com>
3145 * convert.c (convert_to_integer_1): Use direct recursion for
3146 enumeral types and types with a precision less than the number
3147 of bits in their mode.
3149 2018-04-16 Julia Koval <julia.koval@intel.com>
3152 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3153 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3155 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3158 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3159 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3160 and -mno-direct-move.
3162 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3165 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3166 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3167 Ensure negative shifts result in {0}.
3169 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3171 PR rtl-optimization/79916
3172 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3173 regs (if any) to define how to gnerate SD moves when LRA is in
3176 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3178 PR rtl-optimization/85393
3179 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3180 * except.c (expand_dw2_landing_pad_for_region): Make static.
3181 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3182 a label and unconditional jump to old_bb, rather than
3183 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3186 PR rtl-optimization/85376
3187 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3188 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3189 instead of a specific value.
3191 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3192 Bin Cheng <bin.cheng@arm.com>
3194 PR tree-optimization/82965
3195 PR tree-optimization/83991
3196 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3197 by_profile_only parameter.
3198 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3199 information if the loop was predicted to iterate too many times.
3200 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3202 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3205 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3208 2018-04-13 Martin Liska <mliska@suse.cz>
3209 Jakub Jelinek <jakub@redhat.com>
3212 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3213 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3214 * builtins.c (expand_builtin_memory_copy_args): Use
3215 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3216 handle dest_addr == pc_rtx.
3218 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3221 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3222 asked to not generate direct moves.
3223 (fix_trunc<mode>si2_stfiwx): Similar.
3224 (fix_trunc<mode>si2_internal): Similar.
3226 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3229 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3230 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3231 lookup if dest in some wider mode is known to be const0_rtx and
3232 if so, record permanent equivalence for it to be ZERO_EXTEND of
3233 the narrower mode destination.
3235 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3237 * lto-streamer-out.c (output_function): Revert 259346.
3238 * omp-expand.c (expand_oacc_for): Likewise.
3240 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3242 PR rtl-optimization/85354
3243 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3244 * sel-sched.c (sel_global_init): ... here.
3246 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3249 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3250 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3251 mode for PE-COFF targets.
3252 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3253 (i386_pe_asm_lto_end): Likewise.
3254 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3255 (TARGET_ASM_LTO_END): Likewise.
3256 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3257 (i386_pe_asm_lto_start): New function.
3258 (i386_pe_asm_lto_end): Likewise.
3260 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3261 Richard Biener <rguenther@suse.de>
3264 * lto-streamer-out.c (output_function): Fix CFG loop state before
3266 * omp-expand.c (expand_oacc_for): Handle calls to internal
3267 functions like regular functions.
3269 2018-04-12 Richard Biener <rguenther@suse.de>
3272 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3273 for the early LTO debug to properly generate references to it
3274 during DIE emission. Do not re-use that for the skeleton for
3276 (dwarf2out_early_finish): Likewise.
3278 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3281 * config/i386/sse.md
3282 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3283 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3284 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3285 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3286 and output is a reg, avoid creating invalid lowpart subreg, but
3287 instead split into a 512-bit move. Don't split if not AVX512VL,
3288 input is xmm16+ reg and output is a mem.
3289 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3290 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3291 xmm16+ reg and output is a mem.
3293 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3295 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3296 also for flag_dwarf2_cfi_asm.
3298 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3300 PR rtl-optimization/85342
3301 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3302 a bool scalar var inside of the loop instead. Don't try to update
3303 recog_data.operand after failed apply_change_group.
3305 2018-04-12 Tom de Vries <tom@codesourcery.com>
3308 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3309 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3310 array with flexible array member as array without given dimension.
3311 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3312 argument for undefined param to true.
3314 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3317 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3318 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3319 from PowerPC section.
3320 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3321 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3322 help text that is too long.
3323 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3324 help text that is too long.
3325 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3326 help text that is too long.
3328 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3330 * config/alpha/alpha.md (stack_probe_internal): Rename
3331 from "probe_stack". Update all callers.
3333 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3335 PR rtl-optimization/84566
3336 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3337 sched_macro_fuse_insns.
3339 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3342 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3343 (compute_block_dependences): ... from here.
3345 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3347 PR tree-optimization/85331
3348 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3349 from int to HOST_WIDE_INT.
3351 2018-04-11 Martin Jambor <mjambor@suse.cz>
3354 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3355 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3356 not the same as the source val.
3357 (cgraph_edge_brings_value_p): New parameter.
3358 (gather_edges_for_value): Pass destination value to
3359 cgraph_edge_brings_value_p.
3360 (perhaps_add_new_callers): Likewise.
3361 (get_info_about_necessary_edges): Likewise and exclude values brought
3362 only by self-recursive edges.
3363 (create_specialized_node): Redirect only clones of self-calling edges.
3364 (+self_recursive_pass_through_p): New function.
3365 (find_more_scalar_values_for_callers_subset): Use it.
3366 (find_aggregate_values_for_callers_subset): Likewise.
3367 (known_aggs_to_agg_replacement_list): Removed.
3368 (decide_whether_version_node): Re-calculate known constants for all
3369 remaining context clones.
3371 2018-04-11 Richard Biener <rguenther@suse.de>
3374 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3375 from early DWARF output.
3376 (dwarf2out_early_finish): Output line info unconditionally into
3377 early DWARF and add reference to it.
3379 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3382 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3383 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3384 other than V2DFmode using iptr mode attribute.
3385 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3387 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3389 PR rtl-optimization/84659
3390 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3392 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3395 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3397 (output_loc_list): Pass address of a dummy size variable even in the
3398 locview handling loop.
3399 (index_location_lists): Add comment on why skip_loc_list_entry can't
3402 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3405 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3408 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3411 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3412 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3413 and -mstring-compare-inline-limit.
3415 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3418 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3419 for stack clash protection in a register whenever we need it to be in
3422 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3424 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3425 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3427 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3430 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3432 (mlong-double-): Ditto.
3433 * config/rs6000/sysv4.opt (msdata=): Ditto.
3434 (mtls-size=): Ditto.
3436 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3438 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3439 erroneous entries for
3440 "vector int vec_ldl (int, long int *)", and
3441 "vector unsigned int vec_ldl (int, unsigned long int *)".
3442 Add comments and entries for
3443 "vector bool char vec_ldl (int, bool char *)",
3444 "vector bool short vec_ldl (int, bool short *)",
3445 "vector bool int vec_ldl (int, bool int *)",
3446 "vector bool long long vec_ldl (int, bool long long *)",
3447 "vector pixel vec_ldl (int, pixel *)",
3448 "vector long long vec_ldl (int, long long *)",
3449 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3450 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3451 type tree bool_long_long_type_node and correct definition of
3452 bool_V2DI_type_node to make reference to this new type tree.
3453 (rs6000_mangle_type): Replace erroneous reference to
3454 bool_long_type_node with bool_long_long_type_node.
3455 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3456 comments to emphasize sign distinctions for char and int types and
3457 replace RS6000_BTI_bool_long constant with
3458 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3459 use of RS6000_BTI_pixel.
3460 (bool_long_type_node): Remove this macro definition.
3461 (bool_long_long_type_node): New macro definition
3463 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3465 PR rtl-optimization/85300
3466 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3467 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3468 simplify_unary_operation fails.
3470 2018-04-10 Martin Liska <mliska@suse.cz>
3472 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3473 cgraph_edge and ipa_ref.
3475 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3479 * config/i386/sse.md
3480 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3481 computation of the VEC_MERGE selector from mask.
3482 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3483 Fix decoding of the VEC_MERGE selector into mask.
3485 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3487 PR tree-optimization/85286
3488 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3490 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3492 * final.c (final_1): Set insn_last_address as well as
3493 insn_current_address.
3495 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3498 * explow.c (emit_stack_probe): Call validize_mem on memory location
3499 before passing it to gen_probe_stack. Create address operand and
3500 legitimize it for the probe_stack_address case.
3502 2018-04-09 Jan Hubicka <jh@suse.cz>
3505 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3506 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3507 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3509 (free_lang_data): Rebuild type inheritance graph.
3511 2018-04-09 Martin Sebor <msebor@redhat.com>
3513 * invoke.texi (-finline-small-functions): Mention other optimization
3515 (-findirect-inlining, -fpartial-inlining): Same.
3516 (-finline-functions-called-once): Same.
3517 (-freorder-blocks-and-partition): Same.
3519 2018-04-09 Jan Hubicka <jh@suse.cz>
3522 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3523 jumps; choose last target that matches the criteria (i.e.
3524 no partition changes for non-crossing jumps).
3525 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3526 support for redirecting crossing jumps to non-crossing.
3528 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3530 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3531 also for naked functions.
3533 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3535 * config/arc/arc.md (add_shift): New pattern.
3536 (add_shift2): Likewise.
3537 (sub_shift): Likewise.
3538 (sub_shift_cmp0_noout): Likewise.
3539 (compare_si_ashiftsi): Likewise.
3540 (xbfu_cmp0_noout): New combine pattern.
3541 (xbfu_cmp0"): Likewise.
3542 (movsi_set_cc_insn): Place the predicable variant first.
3543 (commutative_binary_cmp0_noout): Remove clobber.
3544 (commutative_binary_cmp0): New pattern.
3545 (noncommutative_binary_cmp0): Likewise.
3546 (noncommutative_binary_cmp0_noout): Likewise.
3547 (noncommutative_binary_comparison_result_used): Removed.
3548 (rsub_cmp0): New pattern.
3549 (rsub_cmp0_noout): Likewise.
3550 (extzvsi): Changed, keep only meaningful variants.
3551 (SQH, SEZ): New iterators.
3552 (SQH_postfix): New mode attribute.
3553 (SEZ_prefix): New code attribute.
3554 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3555 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3556 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3558 (noncommutative_operator): Check the availability of barrel
3561 2018-04-09 Richard Biener <rguenther@suse.de>
3563 PR tree-optimization/85284
3564 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3565 Only use the niter constraining form of simple_iv when the exit
3568 2018-04-09 Tom de Vries <tom@codesourcery.com>
3571 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3572 (define_expand "*memory_barrier"): New define_expand.
3573 (define_insn "memory_barrier"): New insn.
3575 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3577 PR rtl-optimization/80463
3578 PR rtl-optimization/83972
3579 PR rtl-optimization/83480
3581 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3582 correct producer for the insn.
3583 (tidy_control_flow): Fixup seqnos in case of debug insns.
3585 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3587 PR rtl-optimization/83913
3589 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3590 different sched-times when merging exprs.
3592 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3594 PR rtl-optimization/83962
3596 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3597 tidy_fallthru_edge and tidy_control_flow.
3599 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3601 PR rtl-optimization/83530
3603 * sel-sched.c (force_next_insn): New global variable.
3604 (remove_insn_for_debug): When force_next_insn is true, also leave only
3605 next insn in the ready list.
3606 (sel_sched_region): When the region wasn't scheduled, make another pass
3607 over it with force_next_insn set to 1.
3609 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3611 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3613 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3614 for interrupt control.
3615 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3616 functions for interrupt control.
3617 * config/nds32/nds32-intrinsic.md: Likewise.
3618 * config/nds32/nds32_intrinsic.h: Likewise.
3619 * config/nds32/nds32.h (nds32_builtins): Likewise.
3621 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3623 * config/nds32/nds32.c (nds32_init_machine_status,
3624 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3625 strict_aligned_p field.
3626 (nds32_expand_to_rtl_hook): New function.
3627 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3628 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3630 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3631 Chung-Ju Wu <jasonwucj@gmail.com>
3633 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3634 * config/nds32/nds32-n7.md: New file.
3635 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3636 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3638 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3639 * config/nds32/nds32.md (pipeline_model): Add n7.
3640 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3641 * config/nds32/pipelines.md: Include n7 settings.
3643 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3644 Chung-Ju Wu <jasonwucj@gmail.com>
3646 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3647 * config/nds32/nds32-e8.md: New file.
3648 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3649 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3651 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3652 * config/nds32/nds32.md (pipeline_model): Add e8.
3653 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3654 * config/nds32/pipelines.md: Include e8 settings.
3656 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3657 Chung-Ju Wu <jasonwucj@gmail.com>
3659 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3660 * config/nds32/nds32-n8.md: New file.
3661 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3662 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3664 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3665 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3666 * config/nds32/nds32.md (pipeline_model): Add n8.
3667 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3668 * config/nds32/pipelines.md: Include n8 settings.
3670 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3671 Chung-Ju Wu <jasonwucj@gmail.com>
3673 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3674 * config/nds32/nds32-n9-2r1w.md: New file.
3675 * config/nds32/nds32-n9-3r2w.md: New file.
3676 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3677 nds32_register_ports): New or modify for cpu n9.
3678 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3680 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3681 * config/nds32/nds32-utils.c: New file.
3682 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3683 TARGET_MUL_SLOW): Define.
3684 * config/nds32/nds32.md (pipeline_model): New attribute.
3685 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3686 New options that support cpu n9.
3687 * config/nds32/pipelines.md: Include n9 settings.
3688 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3690 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3692 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3693 information if necessary.
3694 (output_cond_branch_compare_zero): Likewise.
3695 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3696 (nds32_target_alignment): Refine for alignment.
3697 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3698 (FUNCTION_BOUNDARY): Modify.
3699 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3701 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3703 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3705 * config/nds32/constants.md (unspec_volatile_element): Add values for
3706 TLB operation and data prefetch.
3707 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3708 functions for TLB operation and data prefetch.
3709 * config/nds32/nds32-intrinsic.md: Likewise.
3710 * config/nds32/nds32_intrinsic.h: Likewise.
3711 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3712 (nds32_print_operand): Likewise.
3713 * config/nds32/nds32.h (nds32_builtins): Likewise.
3715 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3716 Andrew Pinski <pinsika@gcc.gnu.org>
3719 * match.pd: Use constant_boolean_node of correct type instead of
3720 boolean_true_node or boolean_false_node for simplifying
3721 pointer comparisons to zero.
3723 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3725 PR tree-optimization/80021
3726 * tree.c (verify_type_variant): Make error call in verify_variant_match
3727 translatable and remove final full stop.
3729 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3731 * config/nds32/constants.md (unspec_volatile_element): Add
3732 UNSPEC_VOLATILE_EH_RETURN.
3733 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3734 nds32_output_stack_pop): Support dwarf exception handling process.
3735 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3736 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3737 exception handling process.
3738 (nds32_compute_stack_frame): Likewise.
3739 (nds32_return_addr_rtx): Likewise.
3740 (nds32_initial_elimination_offset): Likewise.
3741 (nds32_expand_prologue): Likewise.
3742 (nds32_expand_epilogue): Likewise.
3743 (nds32_dynamic_chain_address): New function.
3744 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3746 (DYNAMIC_CHAIN_ADDRESS): Define.
3747 (EH_RETURN_DATA_REGNO): Define.
3748 (EH_RETURN_STACKADJ_RTX): Define.
3749 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3750 patterns for dwarf exception handling.
3752 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3754 * config/nds32/nds32.h: Clean up obsolete macros.
3756 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3758 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3759 Add enum values for particular instructions.
3760 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3761 particular intrinsic functions.
3762 * config/nds32/nds32-intrinsic.md: Likewise.
3763 * config/nds32/nds32_intrinsic.h: Likewise.
3764 * config/nds32/nds32.h (nds32_builtins): Likewise.
3765 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3766 (btst, ave): New patterns for particular instructions.
3768 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3770 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3771 Add enum values for atomic load/store and memory sync.
3772 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3774 * config/nds32/nds32-intrinsic.md: Likewise.
3775 * config/nds32/nds32_intrinsic.h: Likewise.
3776 * config/nds32/nds32.h (nds32_builtins): Likewise.
3778 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3780 PR tree-optimization/85257
3781 * fold-const.c (native_encode_vector): If not all elts could fit
3782 and off is -1, return 0 rather than offset.
3783 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3784 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3785 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3786 adjust buffer in native_interpret_expr call.
3788 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3790 * config/nds32/constants.md (unspec_volatile_element): Add cache
3791 control enum values.
3792 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3793 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3794 * config/nds32/nds32.c (nds32_cctl_names): New.
3795 (nds32_print_operand): Handle cache control register names.
3796 * config/nds32/nds32.h (nds32_builtins): New enum values.
3797 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3799 * config/nds32/nds32.md (type): Add mmu.
3800 * config/nds32/pipelines.md (simple_insn): Add mmu.
3802 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3804 * config/nds32/nds32.md (type): Remove call.
3805 * config/nds32/pipelines.md (simple_insn): Likewise.
3807 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3809 * config/nds32/constants.md (unspec_volatile_element): Add
3810 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3811 UNSPEC_VOLATILE_FMFCFG.
3812 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3813 description for fmfcfg and fmfcsr.
3814 (bdesc_1arg): Add fmtcsr.
3815 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3816 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3817 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3818 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3819 unspec_fmfcfg): New patterns.
3820 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3821 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3822 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3823 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3824 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3825 __nds32__fmfcfg): Define.
3827 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3829 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3830 intrinsic register names.
3831 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3832 intrinsic register enum values and macros.
3834 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3836 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3837 for load/store addressing form.
3838 (nds32_print_operand_address): Likewise.
3840 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3843 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3844 based on LABEL_REF. Remove useless assertion.
3845 (pic_address_needs_scratch): Fix formatting.
3846 (sparc_legitimize_pic_address): Minor tweaks.
3847 (sparc_delegitimize_address): Adjust assertion accordingly.
3848 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3849 into symbolic_operand.
3850 (movsi_high_pic_label_ref): Likewise.
3851 (movsi_lo_sum_pic_label_ref): Likewise.
3852 (movdi_pic_label_ref): Likewise.
3853 (movdi_high_pic_label_ref): Likewise.
3854 (movdi_lo_sum_pic_label_ref): Likewise.
3856 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3858 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3859 custom LIB_SPEC setup.
3861 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3862 Kito Cheng <kito.cheng@gmail.com>
3864 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3865 * config/riscv/freebsd.h: New.
3867 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3869 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3870 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3873 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3874 Kito Cheng <kito.cheng@gmail.com>
3876 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3877 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3878 * config/nds32/nds32-protos.h (nds32_output_call,
3879 nds32_output_return): Declare.
3880 * config/nds32/nds32.md: Refine all the call and return patterns.
3882 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3885 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3886 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3888 PR rtl-optimization/84872
3889 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3890 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3893 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3895 * expr.c (copy_blkmode_to_reg): Revert 254862.
3896 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3898 2018-04-06 Richard Biener <rguenther@suse.de>
3901 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3902 after seeing a component reference with an adjacent field. Treat
3903 refs to arrays at struct end of external decls similar to
3904 refs to unconstrained commons.
3906 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3909 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3910 look through SAVE_EXPRs with non-side-effects argument. Adjust
3912 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3915 2018-04-06 Richard Biener <rguenther@suse.de>
3918 * alias.c (find_base_term): New wrapper around find_base_term
3919 unwinding CSELIB_VAL_PTR changes.
3920 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3923 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3925 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3927 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3928 constant definitions.
3929 ("nop"): lr 0,0 -> nopr r0
3930 ("nop_lr0", "nop_lr1"): New insn definitions.
3932 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3934 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3935 NDS32_V3PUSH_AVAILABLE_P macro.
3937 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3938 Chung-Ju Wu <jasonwucj@gmail.com>
3940 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3941 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3942 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3943 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3944 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3945 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3946 * config/nds32/constraints.md: New constraints and checking for hard
3947 float configuration.
3948 * config/nds32/iterators.md: New mode iterator and attribute for hard
3949 float configuration.
3950 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3952 * config/nds32/nds32-fpu.md: New file.
3953 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3954 deal with hard float code generation.
3955 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3957 (abi_type, float_reg_number): New enum type.
3958 * config/nds32/nds32-predicates.c: New predicates for hard float.
3959 * config/nds32/nds32-protos.h: Declare functions for hard float.
3960 * config/nds32/nds32.c: Implementation for hard float configuration.
3961 * config/nds32/nds32.h: Definitions for hard float configuration.
3962 * config/nds32/nds32.md: Include hard float machine description and
3963 modify patterns for hard float configuration.
3964 * config/nds32/nds32.opt: New options for hard float configuration.
3965 * config/nds32/predicates.md: New predicates for hard float
3968 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3970 * common/config/nds32/nds32-common.c
3971 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3973 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3976 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3977 CONSTRUCTOR_ELT (ctor, ...)->value.
3979 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3982 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3984 2018-04-05 Tom de Vries <tom@codesourcery.com>
3987 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3990 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3991 Kito Cheng <kito.cheng@gmail.com>
3993 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3994 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3995 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3996 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3998 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3999 Kito Cheng <kito.cheng@gmail.com>
4001 * config/nds32/constraints.md (Ufe): New memory constraint.
4002 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4003 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4004 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4006 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4007 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4009 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4011 * config/nds32/nds32.md: Use optimize_size in the condition for
4012 alu-shift instructions.
4014 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4016 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4018 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4020 * config/nds32/nds32.md (negsi2): Refine pattern.
4022 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4023 Chung-Ju Wu <jasonwucj@gmail.com>
4025 * config/nds32/iterators.md (shift_rotate): New code iterator.
4026 (shift): New code attribute.
4027 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4028 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4029 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4030 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4031 bit-wise operations.
4032 (andsi3, *andsi3): Ditto.
4033 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4034 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4035 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4036 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4037 nds32_ior_operand, nds32_xor_operand): New predicates.
4039 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4041 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4042 (addsi3, subsi3): ... this.
4044 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4046 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4048 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4050 * config/nds32/nds32.md: Adjust indention.
4052 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4054 * config/nds32/nds32.md (feature): New attribute.
4056 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4058 * config/nds32/nds32.md (subtype): New attribute.
4060 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4063 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4064 expansion to perform a bitwise AND of the argument followed by a
4065 boolean negation of the result.
4067 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4069 PR rtl-optimization/84878
4070 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4071 the basic block. Assert the use reference is not artificial and that
4072 it has an associated insn.
4074 2018-04-04 Michael Matz <matz@suse.de>
4076 * builtins.c (compute_objsize): Pass correct operand
4077 to array_at_struct_end_p.
4079 2018-04-04 Richard Biener <rguenther@suse.de>
4082 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4083 from contexts for DINFO_LEVEL_TERSE and below.
4085 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4087 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4088 resiter_operand condition.
4089 * config/nds32/nds32.md (*move<mode>): Ditto.
4091 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4092 Monk Chiang <sh.chiang04@gmail.com>
4094 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4096 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4098 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4100 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4101 Kito Cheng <kito.cheng@gmail.com>
4103 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4104 nds32_cond_code_str, output_cond_branch,
4105 output_cond_branch_compare_zero, nds32_expand_cbranch,
4106 nds32_expand_cstore, nds32_expand_movcc,
4107 nds32_output_cbranchsi4_equality_zero,
4108 nds32_output_cbranchsi4_equality_reg,
4109 nds32_output_cbranchsi4_equality_reg_or_const_int,
4110 nds32_output_cbranchsi4_greater_less_zero: New functions.
4111 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4112 nds32_expand_cstore, nds32_expand_movcc,
4113 nds32_output_cbranchsi4_equality_zero,
4114 nds32_output_cbranchsi4_equality_reg,
4115 nds32_output_cbranchsi4_equality_reg_or_const_int,
4116 nds32_output_cbranchsi4_greater_less_zero): Declare.
4117 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4118 nds32_rimm11s_operand): New predicates.
4119 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4120 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4123 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4125 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4126 * config/nds32/nds32.md: Ditto.
4127 * config/nds32/pipelines.md: Ditto.
4129 2018-04-04 Richard Biener <rguenther@suse.de>
4131 PR tree-optimization/85168
4132 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4133 propagating abnormals.
4135 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4137 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4139 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4140 Kito Cheng <kito.cheng@gmail.com>
4142 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4143 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4144 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4145 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4146 * config/nds32/nds32.md (sibcall_internal): New.
4147 (sibcall_register): Remove.
4148 (sibcall_immediate): Remove.
4149 (sibcall_value_internal): New.
4150 (sibcall_value_register): Remove.
4151 (sibcall_value_immediate): Remove.
4152 * config/nds32/predicates.md (nds32_general_register_operand): New.
4153 (nds32_call_address_operand): New.
4155 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4157 PR rtl-optimization/85167
4158 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4159 bb_defs if *split_p, instead preinitialize it to NULL.
4161 PR tree-optimization/85156
4162 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4163 evaluating the argument multiple times.
4165 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4167 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4169 (_mm_cvtpd_ps): Likewise.
4170 (_mm_cvttpd_epi32): Likewise.
4171 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4172 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4173 vector, pixel, and bool following altivec.h include.
4175 2018-04-03 Martin Sebor <msebor@redhat.com>
4177 * doc/extend.texi (Common Function Attributes): Clarify.
4178 (const attribute): Likewise.
4179 (pure attribute): Likewise.
4181 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4184 * config/i386/i386.c (ix86_expand_vector_set): Use
4185 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4187 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4189 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4190 instructions when changing rounding bits to preserve precision bits
4191 in the x87 control word.
4193 2018-04-03 Martin Liska <mliska@suse.cz>
4195 PR tree-optimization/82491
4196 * rtl.h (strip_offset_and_add): Replace += suboffset with
4197 poly_uint64 () + suboffset.
4199 2018-03-29 Martin Liska <mliska@suse.cz>
4200 Martin Jambor <mjambor@suse.cz>
4203 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4204 param_type is not an integral or pointer type.
4206 2018-04-03 Richard Biener <rguenther@suse.de>
4208 * sese.h (recompute_all_dominators): Remove.
4210 2018-04-02 Martin Sebor <msebor@redhat.com>
4212 * doc/invoke.texi (-Wrestrict): Fix typos.
4214 2018-04-02 Jim Wilson <jimw@sifive.com>
4216 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4217 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4218 (<optab>di3, <optab>si3_extend): Likewise.
4219 (<optab>si3_mask, <optab>si3_mask_1): New.
4220 (<optab>di3_mask, <optab>di3_mask_1): New.
4221 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4222 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4223 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4225 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4227 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4230 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4232 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4233 (nds32_canonicalize_comparison): New function.
4235 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4236 Kito Cheng <kito.cheng@gmail.com>
4237 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4239 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4240 * config/nds32/constants.md (unspec_volatile_element): Add
4241 UNSPEC_VOLATILE_RELAX_GROUP.
4242 * config/nds32/nds32-relax-opt.c: New file.
4243 * config/nds32/nds32-predicates.c
4244 (nds32_symbol_load_store_p): New function.
4245 * config/nds32/nds32-protos.h
4246 (nds32_symbol_load_store_p): Declare function.
4247 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4248 * config/nds32/nds32.c
4249 (nds32_register_pass): New function to register pass.
4250 (nds32_register_passes): New function to register passes.
4251 * config/nds32/nds32.md (relax_group): New pattern.
4252 * config/nds32/nds32.opt (mrelax-hint): New option.
4253 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4255 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4257 * config/nds32/t-nds32: Modify files dependency.
4259 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4261 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4262 (PROFILE_HOOK): Define its implementation.
4264 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4266 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4267 type and 32-bit size.
4269 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4272 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4273 (V_128_256): New mode iterator.
4274 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4275 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4276 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4278 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4279 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4281 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4284 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4285 NaN inputs correctly.
4287 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4290 * config/rs6000/vsx.md (??r): New mode attribute.
4291 (*vsx_mov<mode>_64bit): Use it.
4292 (*vsx_mov<mode>_32bit): Likewise.
4294 2018-03-30 Martin Sebor <msebor@redhat.com>
4296 PR tree-optimization/84818
4297 * builtins.c (check_access): Use warning_n.
4299 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4302 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4304 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4307 2018-03-30 Julia Koval <julia.koval@intel.com>
4310 * x86-tune.def (movx, partial_reg_dependency): Enable for
4313 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4316 * lra-constraints.c (process_alt_operands): Move setting
4317 this_alternative_matches below.
4319 2018-03-29 Martin Liska <mliska@suse.cz>
4322 * doc/invoke.texi: Document how LTO works with debug info.
4323 Describe auto-load support of binutils. Mention 'x86-64'
4324 as valid option value of -march option.
4326 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4328 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4331 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4332 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4333 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4336 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4339 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4340 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4341 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4342 for RS6000_BTM_POWERPC64.
4343 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4344 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4345 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4350 2018-03-28 Carl Love <cel@us.ibm.com>
4353 2017-09-27 Carl Love <cel@us.ibm.com>
4355 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4356 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4357 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4360 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4362 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4363 instead of __vector bool.
4364 (_mm_max_pu8): Likewise.
4365 (_mm_min_pi16): Likewise.
4367 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4370 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4371 (DIVWEUO): Likewise.
4373 (DIVDEUO): Likewise.
4374 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4375 DIVWEUO and DIVDEUO.
4376 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4377 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4378 (div_extend): Likewise.
4379 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4381 (__builtin_divweuo): Likewise.
4382 (__builtin_divdeo): Likewise.
4383 (__builtin_divdeuo): Likewise.
4385 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4388 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4389 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4391 PR tree-optimization/82004
4392 * gimple-match-head.c (optimize_pow_to_exp): New function.
4393 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4394 Don't fold to exp if optimize_pow_to_exp is false.
4396 2018-03-28 Martin Liska <mliska@suse.cz>
4399 * calls.c (initialize_argument_information): Fix trailing space.
4400 * common.opt: Fix typo and provide better explanation for
4401 -fsanitize-coverage option.
4402 * config/i386/i386.opt: Fix typo.
4404 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4405 Martin Liska <mliska@suse.cz>
4408 * gimplify.c (asan_poison_variable): Don't do the check for
4409 gimplify_omp_ctxp here.
4410 (gimplify_decl_expr): Do it here.
4411 (gimplify_target_expr): Likewise.
4413 2018-03-28 Martin Liska <mliska@suse.cz>
4416 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4417 chkp_type_bounds_count if MPX is not enabled.
4419 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4421 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4423 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4426 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4427 function to create the function decl for complex long double
4428 multiply and divide for -mabi=ieeelongdouble.
4429 (init_float128_ieee): Call it.
4431 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4434 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4435 -fcf-protection=branch -mibt.
4436 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4438 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4441 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4443 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4446 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4447 extern array declarations.
4449 2018-03-27 Richard Biener <rguenther@suse.de>
4452 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4453 explicit single_use checks.
4455 2018-03-27 Richard Biener <rguenther@suse.de>
4457 PR tree-optimization/85082
4458 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4461 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4463 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4464 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4465 Turn on fasynchronous-unwind-tables and funwind-tables.
4467 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4470 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4471 (*bmi_blsr_<mode>_ccz): Ditto.
4473 2018-03-26 Tom de Vries <tom@codesourcery.com>
4475 PR tree-optimization/85063
4476 * omp-general.c (offloading_function_p): New function. Factor out
4478 * omp-offload.c (pass_omp_target_link::gate): ... here.
4479 * omp-general.h (offloading_function_p): Declare.
4480 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4481 with attribute omp declare target for offloading functions.
4483 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4485 PR tree-optimization/84005
4486 * tree-data-ref.h (get_base_for_alignment): Declare.
4487 * tree-data-ref.c (get_base_for_alignment_1): New function.
4488 (get_base_for_alignment): Likewise.
4489 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4490 get_base_for_alignment to find a suitable base object, instead
4491 of always using drb->base_address.
4493 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4496 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4497 known size by default.
4499 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4502 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4503 and non BLKmode operands.
4505 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4508 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4509 Clean up attributes.
4511 2018-03-23 Richard Biener <rguenther@suse.de>
4514 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4515 we are going to emit early debug for LTO.
4517 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4520 * function.c (match_asm_constraints_1): Don't optimize if input
4521 doesn't satisfy general_operand predicate for output's mode.
4524 * alias.c (write_dependence_p): Don't require for x_canonicalized
4525 non-VOIDmode if x has VOIDmode.
4528 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4529 just don't try to optimize it rather than assert it never happens.
4531 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4533 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4534 macro expansions for definition of ST_INTERNAL_<mode> and
4535 LD_INTERNAL_<mode> builtins.
4536 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4538 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4540 (altivec_expand_st_builtin): Likewise.
4541 (altivec_expand_builtin): Remove calls to deleted functions.
4542 (rs6000_address_for_altivec): Delete this function.
4543 * config/rs6000/vector.md: Remove expands for
4544 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4546 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4549 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4550 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4551 re-computing once computed.
4552 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4553 (arm_init_machine_status): Initialize
4554 machine->static_chain_stack_bytes.
4556 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4559 * doc/extend.texi: Add four new prototypes for vec_ld.
4560 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4561 definitions for more logical presentation.
4562 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4563 entries for V1TI variants of __builtin_altivec_ld builtin.
4564 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4565 handling of V1TI variant of LVX icode pattern.
4566 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4567 (rs6000_gimple_fold_builtin): Likewise.
4568 (altivec_init_builtins): Add code to define
4569 __builtin_altivec_lvx_v1ti function.
4571 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4574 * function.c (match_asm_constraints_1): Don't do the optimization
4575 if input isn't a REG, SUBREG, MEM or constant.
4577 2018-03-22 Tom de Vries <tom@codesourcery.com>
4579 PR tree-optimization/84956
4580 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4581 bb_has_abnormal_pred.
4583 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4586 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4587 DECL_INITIAL (decl) to decl at the end.
4588 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4591 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4593 * doc/extend.texi (__builtin_tgmath): Document when complex
4594 integer types are treated as _Complex _Float64.
4596 2018-03-21 Tom de Vries <tom@codesourcery.com>
4598 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4600 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4602 PR tree-optimization/84960
4603 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4604 if it is ENTRY block, move them into single succ of ENTRY in that case.
4606 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4608 PR tree-optimization/84811
4609 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4610 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4611 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4612 (known_subrange_p): Update accordingly. Cast each value involved
4613 in the size comparison, rather than casting the result of the
4616 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4618 PR tree-optimization/84982
4619 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4620 by flipping the least significant bit rather than all bits from
4621 bitpos to bitpos + bitsize - 1.
4623 2018-03-21 Nathan Sidwell <nathan@acm.org>
4625 * doc/extend.texi (Deprecated Features): Remove mention of
4626 long-deleted deprecations.
4628 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4631 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4632 * configure: Regenerate.
4634 2018-03-21 Tom de Vries <tom@codesourcery.com>
4636 PR tree-optimization/83126
4637 * tree-parloops.c (num_phis): New function.
4638 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4640 2018-03-21 Nathan Sidwell <nathan@acm.org>
4642 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4643 mention anon-struct/union members and trailing attributes.
4645 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4647 PR tree-optimization/84969
4648 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4649 builtin memset partitions if they set different rhs values.
4651 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4653 PR rtl-optimization/84989
4654 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4655 VEC_DUPLICATE with scalar result mode.
4657 2018-03-21 Martin Liska <mliska@suse.cz>
4660 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4661 not intended return statement.
4663 2018-03-21 Martin Liska <mliska@suse.cz>
4666 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4667 (chkp_find_bound_slots_1): Limit number of iterations.
4669 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4672 * Minor grammar fixes for x86 options.
4674 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4677 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4678 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4680 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4683 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4684 (altivec_lvx_<mode>_1op): Likewise.
4685 (altivec_stvx_<mode>_2op): Likewise.
4686 (altivec_stvx_<mode>_1op): Likewise.
4687 (altivec_lvx_<VM2:mode>): New define_expand.
4688 (altivec_stvx_<VM2:mode>): Likewise.
4689 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4690 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4691 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4692 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4693 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4694 (rs6000_gen_lvx): Likewise.
4695 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4696 (altivec_expand_stv_builtin): Likewise.
4697 (altivec_expand_builtin): Likewise.
4698 * config/rs6000/vector.md: Likewise.
4700 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4703 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4706 2018-03-20 Richard Biener <rguenther@suse.de>
4709 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4710 sign-conversions as zero, fall back to standard scalar_stmt
4713 2018-03-20 Martin Liska <mliska@suse.cz>
4716 * predict.c (rebuild_frequencies): Handle case when we have
4717 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4719 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4722 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4723 flag_section_anchors.
4724 * varasm.c (use_blocks_for_decl_p): Remove hack for
4725 dw2_force_const_mem.
4728 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4730 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4731 be created, use lowpart_subreg of operands[0] rather than operands[0]
4733 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4734 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4735 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4736 and n constraint instead of aarch64_shift_imm_di and Usd.
4737 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4738 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4740 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4743 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4744 to favor GPR over NEON registers.
4745 (<shift>di3_neon): Likewise.
4747 2018-03-20 Tom de Vries <tom@codesourcery.com>
4750 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4751 (nvptx_process_pars): Emit bar.sync asap and alap.
4753 2018-03-20 Tom de Vries <tom@codesourcery.com>
4756 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4757 seen_label if seen_label is already set.
4759 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4762 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4763 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4764 Use 1U instead of 1. Formatting fixes.
4767 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4768 instead of TREE_TYPE (s1) for the return value.
4770 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4772 PR tree-optimization/84946
4773 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4774 bitsize + bitsize in poly_uint64 rather than poly_int64.
4777 * dwarf2asm.c: Include fold-const.c.
4778 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4779 of decl rather than decl itself.
4781 PR rtl-optimization/84643
4782 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4784 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4787 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4788 calling assemble_variable.
4790 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4793 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4794 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4796 2018-03-19 Jim Wilson <jimw@sifive.com>
4799 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4800 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4801 (riscv_first_stack_step): Likewise.
4802 (riscv_option_override): Use STACK_BOUNDARY instead of
4804 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4806 (BIGGEST_ALIGNMENT): Set to 128.
4807 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4808 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4811 2018-03-19 Richard Biener <rguenther@suse.de>
4813 PR tree-optimization/84933
4814 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4815 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4817 2018-03-19 Richard Biener <rguenther@suse.de>
4819 PR tree-optimization/84859
4820 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4821 (cond_if_else_store_replacement): Perform sinking operation on
4822 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4823 Generalize what a BB with a single eligible store is.
4825 2018-03-19 Richard Biener <rguenther@suse.de>
4827 PR tree-optimization/84929
4828 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4829 chrec_is_positive against non-chrec arg.
4831 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4834 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4836 2018-03-18 Martin Liska <mliska@suse.cz>
4838 PR rtl-optimization/84635
4839 * regrename.c (build_def_use): Use matches_mode only when
4842 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4844 PR tree-optimization/84913
4845 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4846 vectorize chains of COND_EXPRs.
4848 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4850 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4852 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4854 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4856 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4858 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4860 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4861 Kito Cheng <kito.cheng@gmail.com>
4863 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4864 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4865 (nds32_adjust_reg_alloc_order): New function.
4866 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4868 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4870 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4871 nds32_print_operand, nds32_print_operand_address): Use
4872 HOST_WIDE_INT_PRINT_DEC instead.
4874 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4876 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4878 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4881 * config/i386/i386.c (initial_ix86_tune_features,
4882 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4884 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4885 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4886 rather than 1u << ix86_tune. Formatting fix.
4887 (ix86_option_override_internal): Change ix86_arch_mask from
4888 unsigned int to unsigned HOST_WIDE_INT, initialize to
4889 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4890 (ix86_function_specific_restore): Likewise.
4892 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4895 * postreload.c (reload_combine_recognize_pattern): Perform
4896 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4897 truncate_int_for_mode the result for the destination's mode.
4900 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4901 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4903 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4906 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4909 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4910 regno_allocno_class_array and sorted_pseudos.
4911 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4912 insns where regno is used.
4914 2018-03-16 Martin Liska <mliska@suse.cz>
4917 * multiple_target.c (create_dispatcher_calls): Redirect
4918 reference in the symbol table.
4920 2018-03-16 Martin Liska <mliska@suse.cz>
4923 * multiple_target.c (create_dispatcher_calls): Redirect also
4926 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4930 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4931 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4932 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4934 2018-03-16 Julia Koval <julia.koval@intel.com>
4936 * doc/invoke.texi (Skylake Server): Add CLWB.
4937 Cannonlake): Remove CLWB.
4939 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4941 PR tree-optimization/84841
4942 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4944 (FLOAT_ONE_CONST_TYPE): Define.
4945 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4946 (sort_by_operand_rank): Put entries with higher constant_type last
4947 rather than first to match comments.
4949 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4951 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4954 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4957 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4958 check and use address of target always.
4960 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4963 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4964 (indirect_thunk_bnd_needed): Likewise.
4965 (indirect_thunks_used): Likewise.
4966 (indirect_thunks_bnd_used): Likewise.
4967 (indirect_return_needed): New.
4968 (indirect_return_bnd_needed): Likewise.
4969 (output_indirect_thunk_function): Add a bool argument for
4971 (output_indirect_thunk_function): Don't generate alias for
4972 function return thunk.
4973 (ix86_code_end): Call output_indirect_thunk_function to generate
4974 function return thunks.
4975 (ix86_output_function_return): Set indirect_return_bnd_needed
4976 and indirect_return_needed instead of indirect_thunk_bnd_needed
4977 and indirect_thunk_needed.
4979 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4981 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4982 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4983 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4985 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4986 Paul Hua <paul.hua.gm@gmail.com>
4989 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4991 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4993 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4994 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4997 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5000 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5001 instead of GET_MODE_SIZE when comparing Units.
5003 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5006 * varasm.c (hash_section): Return an unchangeble hash value
5007 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5008 Return !aarch64_can_use_per_function_literal_pools_p ().
5010 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5013 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5014 rather than address of cmode as last argument to prepare_cmp_insn.
5016 2018-03-15 Julia Koval <julia.koval@intel.com>
5018 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5019 F_AVX512VNNI, F_AVX512BITALG): New.
5021 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5024 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5025 insn for floating-point loads and stores.
5027 2018-03-14 Carl Love <cel@us.ibm.com>
5029 * config/rs6000/rs6000-c.c: Add macro definitions for
5030 ALTIVEC_BUILTIN_VEC_PERMXOR.
5031 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5032 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5033 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5034 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5036 * config/doc/extend.texi: Add prototypes for vec_permxor.
5038 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5041 * diagnostic-show-locus.c (class layout_point): Convert m_line
5042 from int to linenum_type.
5043 (line_span::comparator): Use linenum "compare" function when
5044 comparing line numbers.
5045 (test_line_span): New function.
5046 (layout_range::contains_point): Convert param "row" from int to
5048 (layout_range::intersects_line_p): Likewise.
5049 (layout::will_show_line_p): Likewise.
5050 (layout::print_source_line): Likewise.
5051 (layout::should_print_annotation_line_p): Likewise.
5052 (layout::print_annotation_line): Likewise.
5053 (layout::print_leading_fixits): Likewise.
5054 (layout::annotation_line_showed_range_p): Likewise.
5055 (struct line_corrections): Likewise for field m_row.
5056 (line_corrections::line_corrections): Likewise for param "row".
5057 (layout::print_trailing_fixits): Likewise.
5058 (layout::get_state_at_point): Likewise.
5059 (layout::get_x_bound_for_row): Likewise.
5060 (layout::print_line): Likewise.
5061 (diagnostic_show_locus): Likewise for locals "last_line" and
5063 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5064 * input.c (selftest::test_linenum_comparisons): New function.
5065 (selftest::input_c_tests): Call it.
5066 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5067 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5068 * selftest.h (ASSERT_GT): New macro.
5069 (ASSERT_GT_AT): New macro.
5070 (ASSERT_LT): New macro.
5071 (ASSERT_LT_AT): New macro.
5073 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5075 PR rtl-optimization/84780
5076 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5078 2018-03-14 Martin Liska <mliska@suse.cz>
5080 * tree.c (record_node_allocation_statistics): Use
5081 get_stats_node_kind.
5082 (get_stats_node_kind): New function extracted from
5083 record_node_allocation_statistics.
5084 (free_node): Use get_stats_node_kind.
5086 2018-03-14 Richard Biener <rguenther@suse.de>
5088 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5089 that the value-set of ANTIC_IN doesn't grow.
5092 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5094 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5095 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5097 2018-03-14 Julia Koval <julia.koval@intel.com>
5099 * config.gcc (icelake-client, icelake-server): New.
5101 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5102 (initial_ix86_arch_features): Ditto.
5103 (PTA_SKYLAKE): Add SGX.
5104 (PTA_ICELAKE): Remove.
5105 (PTA_ICELAKE_CLIENT): New.
5106 (PTA_ICELAKE_SERVER): New.
5107 (ix86_option_override_internal): Split up icelake on icelake client and
5109 (get_builtin_code_for_version): Ditto.
5110 (fold_builtin_cpu): Ditto.
5111 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5112 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5113 * config/i386/i386.h (processor_type): Ditto.
5114 * doc/invoke.texi: Ditto.
5116 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5119 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5120 INTEGER_CST offset, add it together with bitpos / 8 and
5121 sign extend based on POINTER_SIZE.
5125 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5128 * config/i386/constraints.md (Yc): New register constraint.
5129 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5130 Use Yc constraint for alternative 2 of operand 0. Remove
5131 preferred_for_speed attribute.
5133 2018-03-14 Richard Biener <rguenther@suse.de>
5135 PR tree-optimization/84830
5136 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5137 with the old one to avoid oscillations.
5139 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5142 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5144 (assign_by_spills): Return a flag of reload assignment failure.
5145 Do not process the reload assignment failures. Do not spill other
5146 reload pseudos if they has the same reg class. Update n if
5148 (lra_assign): Add a return arg. Set up from the result of
5149 assign_by_spills call.
5150 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5151 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5152 usage_insns if it is not NULL.
5153 (spill_hard_reg_in_range): New function.
5154 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5155 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5156 function prototypes.
5157 (lra_assign): Change prototype.
5158 * lra.c (lra): Add code to deal with fails by splitting hard reg
5161 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5163 * config/riscv/riscv.opt (mrelax): New option.
5164 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5165 "norelax" when riscv_mrelax is disabled.
5166 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5168 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5171 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5172 reassociation for int modes.
5174 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5176 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5177 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5179 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5180 * config/aarch64/aarch64-sve.md
5181 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5182 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5183 (*extend<mode><Vwide>2): Rename to...
5184 (aarch64_sve_extend<mode><Vwide>2): ...this.
5185 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5186 renaming the old pattern to...
5187 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5189 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5190 define_expand, renaming the old pattern to...
5191 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5192 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5193 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5194 account when deciding which SVE instruction the optab should use.
5195 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5197 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5199 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5200 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5201 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5202 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5203 (tlsdesc_small_<mode>): Turn a define_expand and use
5204 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5205 (tlsdesc_small_advsimd_<mode>): ...this.
5206 (tlsdesc_small_sve_<mode>): New pattern.
5208 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5210 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5211 (UNSPEC_UMUL_HIGHPART): New constants.
5212 (MUL_HIGHPART): New int iteraor.
5213 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5214 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5216 (*<su>mul<mode>3_highpart): New define_insn.
5218 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5221 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5224 2018-03-13 Martin Liska <mliska@suse.cz>
5227 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5229 (sem_item_optimizer::~sem_item_optimizer): Release it.
5230 (sem_item_optimizer::merge_classes): Register variable aliases.
5231 (sem_item_optimizer::fixup_pt_set): New function.
5232 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5233 * ipa-icf.h: Declare new variables and functions.
5235 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5238 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5239 integer_pow2p@2 and test integer_pow2p in condition.
5240 (A < 0 ? C : 0): Similarly for @1.
5243 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5244 characters starting at p contain '\0' character, don't look beyond
5248 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5249 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5252 * reg-stack.c (change_stack): Change update_end var from int to
5253 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5254 also call set_block_for_insn on the newly added insns and rescan.
5257 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5258 on the last operand.
5261 * tree.c (stabilize_reference_1): Return save_expr (e) for
5262 STATEMENT_LIST even if it doesn't have side-effects.
5264 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5266 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5268 2018-03-12 Renlin Li <renlin.li@arm.com>
5270 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5271 aarch64_output_scalar_simd_mov_immediate.
5273 2018-03-12 Martin Sebor <msebor@redhat.com>
5275 PR tree-optimization/83456
5276 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5277 for perfectly overlapping calls to memcpy.
5278 (gimple_fold_builtin_memory_chk): Same.
5279 (gimple_fold_builtin_strcpy): Handle no-warning.
5280 (gimple_fold_builtin_stxcpy_chk): Same.
5281 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5283 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5285 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5286 parameter. Use it for SFmode.
5287 (rs6000_function_arg_advance_1): Adjust.
5288 (rs6000_function_arg): Adjust.
5289 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5291 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5293 PR rtl-optimization/84169
5294 PR rtl-optimization/84780
5295 * combine.c (can_combine_p): Check for a 2-insn combination whether
5296 the destination register is used between the two insns, too.
5298 2018-03-12 Richard Biener <rguenther@suse.de>
5300 PR tree-optimization/84803
5301 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5302 for refs DR analysis didn't process.
5304 2018-03-12 Richard Biener <rguenther@suse.de>
5306 PR tree-optimization/84777
5307 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5308 force-vectorize loops ignore whether we are optimizing for size.
5310 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5312 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5313 (TARGET_MD_ASM_ADJUST): Define.
5315 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5316 Kito Cheng <kito.cheng@gmail.com>
5317 Chung-Ju Wu <jasonwucj@gmail.com>
5319 * config/nds32/nds32.c (nds32_compute_stack_frame,
5320 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5321 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5322 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5323 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5324 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5325 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5326 * config/nds32/nds32.md (prologue, epilogue): Use macro
5327 NDS32_V3PUSH_AVAILABLE_P to do checking.
5329 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5332 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5333 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5334 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5335 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5336 addition of most attributes on !orig_type_die or the attribute not
5337 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5339 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5340 Chung-Ju Wu <jasonwucj@gmail.com>
5342 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5344 * config/nds32/nds32.opt (mvh): New option.
5346 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5347 Chung-Ju Wu <jasonwucj@gmail.com>
5349 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5351 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5352 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5355 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5356 Chung-Ju Wu <jasonwucj@gmail.com>
5358 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5360 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5361 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5363 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5364 Kito Cheng <kito.cheng@gmail.com>
5365 Chung-Ju Wu <jasonwucj@gmail.com>
5367 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5368 UNSPEC_FFMISM and UNSPEC_FLMISM.
5369 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5370 for ffb, ffmism and flmism.
5371 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5372 (unspec_ffmism): Ditto.
5373 (unspec_flmism): Ditto.
5374 (nds32_expand_builtin_impl): Check if string extension is available.
5375 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5376 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5378 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5381 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5384 * lra-assigns.c (assign_by_spills): Return a flag of reload
5385 assignment failure. Do not process the reload assignment
5386 failures. Do not spill other reload pseudos if they has the same
5388 (lra_assign): Add a return arg. Set up from the result of
5389 assign_by_spills call.
5390 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5391 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5392 usage_insns if it is not NULL.
5393 (spill_hard_reg_in_range): New function.
5394 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5395 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5396 function prototypes.
5397 (lra_assign): Change prototype.
5398 * lra.c (lra): Add code to deal with fails by splitting hard reg
5401 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5404 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5406 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5409 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5410 (dw_val_node): Add val_symbolic_view.
5411 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5412 (symview_upper_bound): New.
5413 (new_line_info_table): Initialize symviews_since_reset.
5414 (dwarf2out_source_line): Count symviews_since_reset and set
5415 symview_upper_bound.
5416 (dw_val_equal_p): Handle symview.
5417 (add_AT_symview): New.
5418 (print_dw_val): Handle symview.
5419 (attr_checksum, attr_checksum_ordered): Likewise.
5420 (same_dw_val_p, size_of_die): Likewise.
5421 (value_format, output_die): Likewise.
5422 (add_high_low_attributes): Use add_AT_symview for entry_view.
5423 (dwarf2out_finish): Reset symview_upper_bound, clear
5426 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5429 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5430 Add strict argument and use it.
5431 (rs6000_split_multireg_move): Update for new strict argument.
5432 (mem_operand_gpr): Disallow all non-offsettable addresses.
5433 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5435 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5438 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5439 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5440 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5443 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5444 decl, use remap_type if we want to use the type.
5446 2018-03-09 Martin Sebor <msebor@redhat.com>
5448 PR tree-optimization/84526
5449 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5451 (builtin_access::generic_overlap): Be prepared to handle non-array
5454 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5456 PR rtl-optimization/84682
5457 * lra-constraints.c (process_address_1): Check is_address flag
5458 for address constraints.
5459 (process_alt_operands): Likewise.
5460 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5461 preprocess_constraints.
5462 * recog.h (preprocess_constraints): Add oploc parameter.
5464 * recog.c (preprocess_constraints): Test address_operand for
5465 CT_ADDRESS constraints.
5467 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5470 * lra-assigns.c (assign_by_spills): Return a flag of reload
5471 assignment failure. Do not process the reload assignment
5472 failures. Do not spill other reload pseudos if they has the same
5474 (lra_assign): Add a return arg. Set up from the result of
5475 assign_by_spills call.
5476 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5477 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5478 usage_insns if it is not NULL.
5479 (spill_hard_reg_in_range): New function.
5480 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5481 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5482 function prototypes.
5483 (lra_assign): Change prototype.
5484 * lra.c (lra): Add code to deal with fails by splitting hard reg
5487 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5490 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5491 Accept complain bool parameter. Only emit errors if it is true.
5492 (arm_parse_cpu_option_name): Likewise.
5493 (arm_target_thumb_only): Adjust callers of the above.
5494 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5495 prototype to take a default true bool parameter.
5496 (arm_parse_arch_option_name): Likewise.
5498 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5499 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5503 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5504 * configure: Regenerate.
5505 * configure.ac ("linker --version-script option"): New.
5506 ("linker soname option"): New.
5508 2018-03-09 Richard Biener <rguenther@suse.de>
5510 PR tree-optimization/84775
5511 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5512 immediate uses of predicate stmts and mark them modified.
5515 PR tree-optimization/84178
5516 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5518 (version_loop_for_if_conversion): Delay update_ssa call.
5519 (tree_if_conversion): Delay update_ssa until after predicate
5522 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5525 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5526 when the function accesses prior frames.
5528 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5531 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5532 gen_llsym, otherwise call maybe_gen_llsym.
5535 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5536 has ',' character inside of it.
5538 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5541 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5542 as clobbering CC_REGNUM.
5544 2018-03-08 Richard Biener <rguenther@suse.de>
5547 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5548 (follow_copies_to_constant): Do not follow SSA names registered
5551 2018-03-08 Richard Biener <rguenther@suse.de>
5553 PR tree-optimization/84178
5554 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5556 (version_loop_for_if_conversion): Delay update_ssa call.
5557 (tree_if_conversion): Delay update_ssa until after predicate
5560 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5562 PR tree-optimization/84178
5563 * tree-if-conv.c (release_bb_predicate): Remove the
5564 the assertion that the stmts have NULL use_ops.
5565 Discard the statements, asserting that they haven't
5566 yet been added to a BB.
5568 2018-03-08 Richard Biener <rguenther@suse.de>
5570 PR tree-optimization/84746
5571 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5572 (phi_translate): Pass in destination ANTIC_OUT set.
5573 (phi_translate_1): Likewise. For a simplified result lookup
5574 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5575 (phi_translate_set): Adjust.
5576 (do_pre_regular_insertion): Likewise.
5577 (do_pre_partial_partial_insertion): Likewise.
5579 2018-03-08 Martin Liska <mliska@suse.cz>
5581 PR gcov-profile/84735
5582 * doc/gcov.texi: Document usage of profile files.
5583 * gcov-io.h: Document changes in the format.
5585 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5589 * dwarf2out.c (struct dw_line_info_table): Update comments for
5591 (FORCE_RESET_NEXT_VIEW): New.
5592 (FORCE_RESETTING_VIEW_P): New.
5593 (RESETTING_VIEW_P): Check for -1 too.
5594 (ZERO_VIEW_P): Likewise.
5595 (new_line_info_table): Force-reset next view.
5596 (dwarf2out_begin_function): Likewise.
5597 (dwarf2out_source_line): Simplify zero_view_p initialization.
5598 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5599 view directly. Omit view when omitting .loc at line 0.
5601 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5603 PR tree-optimization/84740
5604 * tree-switch-conversion.c (process_switch): Call build_constructors
5605 only if info.phi_count is non-zero.
5607 PR tree-optimization/84739
5608 * tree-tailcall.c (find_tail_calls): Check call arguments against
5609 DECL_ARGUMENTS (current_function_decl) rather than
5610 DECL_ARGUMENTS (func) when checking for tail recursion.
5612 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5614 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5615 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5616 Volker Reichelt's entry and add entries for people that perform
5617 GCC fuzzy testing and report numerous bugs.
5619 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5622 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5623 readonly data in sdata, if that is disabled.
5624 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5625 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5626 -mreadonly-in-sdata option.
5628 2018-03-07 Martin Sebor <msebor@redhat.com>
5630 PR tree-optimization/84468
5631 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5632 basic block when looking for nul assignment.
5634 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5637 * except.h (output_function_exception_table): Adjust prototype.
5638 * except.c (output_function_exception_table): Remove FNNAME parameter
5639 and add SECTION parameter. Ouput one part of the table at a time.
5640 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5641 the first part of the exception table and emit unwind directives.
5642 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5643 (i386_pe_seh_cold_init): Likewise.
5644 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5645 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5646 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5647 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5648 * config/i386/winnt.c: Include except.h.
5649 (struct seh_frame_state): Add reg_offset, after_prologue and
5650 in_cold_section fields.
5651 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5652 (i386_pe_seh_cold_init): New function.
5653 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5654 to seh->in_cold_section.
5655 (seh_emit_push): Record the offset of the push.
5656 (seh_emit_save): Record the offet of the save.
5657 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5658 Test seh->after_prologue to disregard the epilogue.
5659 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5660 (i386_pe_end_cold_function): New function.
5662 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5665 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5666 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5669 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5671 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5672 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5675 * multiple_target.c: Include tree-inline.h and intl.h.
5676 (expand_target_clones): Diagnose and fail if node->definition and
5677 !tree_versionable_function_p (node->decl).
5679 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5681 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5683 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5684 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5685 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5687 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5690 * combine.c (try_combine): Use reg_or_subregno instead of handling
5691 just paradoxical SUBREGs and REGs.
5693 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5695 * config/arc/arc.c (arc_finalize_pic): Remove function.
5696 (arc_must_save_register): We use single base PIC register, remove
5697 checks to save/restore the PIC register.
5698 (arc_expand_prologue): Likewise.
5699 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5701 (arc_verify_short): Likewise.
5702 (arc_attr_type): Likewise.
5703 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5704 (walk_stores): Likewise.
5705 (arc_address_cost): Make it static.
5706 (arc_verify_short): Likewise.
5707 (branch_dest): Likewise.
5708 (arc_attr_type): Likewise.
5709 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5710 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5711 (arc_final_prescan_insn): Remove inserting the nops due to
5712 hardware hazards. It is done in reorg step.
5713 (insn_length_variant_t): Remove.
5714 (insn_length_parameters_t): Likewise.
5715 (arc_insn_length_parameters): Likewise.
5716 (arc_get_insn_variants): Likewise.
5717 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5719 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5722 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5725 PR tree-optimization/84687
5726 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5728 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5730 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5732 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5733 Rename to ppc_speculation_barrier.
5734 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5735 __builtin_ppc_speculation_barrier.
5737 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5740 * combine.c (combine_simplify_rtx): Don't try to simplify if
5741 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5744 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5746 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5747 to 32 bytes when compiling for POWER9.
5749 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5752 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5753 regparm >= 3 with no arg reg available also for calls with
5754 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5757 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5759 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5761 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5764 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5766 2018-03-05 Richard Biener <rguenther@suse.de>
5768 PR tree-optimization/84486
5769 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5770 When inserting a __builtin_assume_aligned call set the LHS
5771 SSA name alignment info accordingly.
5773 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5775 PR tree-optimization/84114
5776 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5777 Avoid reassociation of FLOAT_MODE addition.
5779 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5781 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5782 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5783 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5784 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5785 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5786 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5787 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5789 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5790 __builtin_ia32_wbinvd): New builtins.
5791 (SPECIAL_ARGS2): New.
5792 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5793 (SPECIAL_ARGS2): New.
5794 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5795 (ix86_valid_target_attribute_inner_p): Ditto.
5796 (ix86_init_mmx_sse_builtins): Add special_args2.
5797 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5798 TARGET_WBNOINVD_P): New.
5799 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5800 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5801 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5802 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5803 * config/i386/pconfigintrin.h: New file.
5804 * config/i386/wbnoinvdintrin.h: Ditto.
5805 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5806 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5808 2018-03-05 Richard Biener <rguenther@suse.de>
5810 PR tree-optimization/84670
5811 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5813 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5814 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5815 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5816 doesn't grow if all successors (recursively) were visited at least
5819 2018-03-05 Richard Biener <rguenther@suse.de>
5821 PR tree-optimization/84650
5822 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5823 if executed in the loop pipeline.
5825 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5827 * doc/configfiles.texi (Configuration Files): Move info about
5828 conditionalizing $target-protos.h to...
5829 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5830 differs from $target-protos.h.
5832 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5833 Chung-Ju Wu <jasonwucj@gmail.com>
5835 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5836 * config/nds32/nds32-multiple.md (setmemsi): Define.
5837 * config/nds32/nds32-memory-manipulation.c
5838 (nds32_gen_dup_4_byte_to_word_value): New.
5839 (emit_setmem_word_loop): New.
5840 (emit_setmem_byte_loop): New.
5841 (nds32_expand_setmem_loop): New.
5842 (nds32_expand_setmem_loop_v3m): New.
5843 (nds32_expand_setmem_unroll): New.
5844 (nds32_expand_setmem): New.
5846 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5847 Chung-Ju Wu <jasonwucj@gmail.com>
5849 * config/nds32/nds32-memory-manipulation.c
5850 (nds32_emit_load_store): New.
5851 (nds32_emit_post_inc_load_store): New.
5852 (nds32_emit_mem_move): New.
5853 (nds32_emit_mem_move_block): New.
5854 (nds32_expand_movmemsi_loop_unknown_size): New.
5855 (nds32_expand_movmemsi_loop_known_size): New.
5856 (nds32_expand_movmemsi_loop): New.
5857 (nds32_expand_movmemsi_unroll): New.
5858 (nds32_expand_movmemqi): Rename ...
5859 (nds32_expand_movmemsi): ... to this.
5860 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5861 (movmemsi): ... to this.
5862 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5863 (nds32_expand_movmemsi): ... to this.
5865 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5866 Monk Chiang <sh.chiang04@gmail.com>
5867 Chung-Ju Wu <jasonwucj@gmail.com>
5869 * config/nds32/nds32-protos.h
5870 (nds32_expand_load_multiple): New arguments.
5871 (nds32_expand_store_multiple): Ditto.
5872 (nds32_valid_multiple_load_store): Rename ...
5873 (nds32_valid_multiple_load_store_p): ... to this.
5874 * config/nds32/nds32-memory-manipulation.c
5875 (nds32_expand_load_multiple): Refine implementation.
5876 (nds32_expand_store_multiple): Ditto.
5877 * config/nds32/nds32-multiple.md
5878 (load_multiple): Update nds32_expand_load_multiple interface.
5879 (store_multiple): Update nds32_expand_store_multiple interface.
5880 * config/nds32/nds32-predicates.c
5881 (nds32_valid_multiple_load_store): Rename ...
5882 (nds32_valid_multiple_load_store_p): ... to this and refine
5884 * config/nds32/predicates.md
5885 (nds32_load_multiple_and_update_address_operation): New predicate.
5886 (nds32_store_multiple_and_update_address_operation): New predicate.
5888 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5889 Chung-Ju Wu <jasonwucj@gmail.com>
5891 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5892 (combo): New attribute.
5893 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5895 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5897 * config/nds32/nds32.opt: Change -mcmodel= default value.
5899 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5900 Monk Chiang <sh.chiang04@gmail.com>
5901 Chung-Ju Wu <jasonwucj@gmail.com>
5903 * config/nds32/constants.md (unspec_element): New enum.
5904 * config/nds32/constraints.md (Umw): New constraint.
5905 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5906 * config/nds32/nds32-intrinsic.md: Likewise.
5907 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5908 (nds32_valid_smw_lwm_base_p): New.
5909 (nds32_output_smw_single_word): New.
5910 (nds32_output_lmw_single_word): New.
5911 (nds32_expand_unaligned_load): New.
5912 (nds32_expand_unaligned_store): New.
5913 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5914 (nds32_output_smw_single_word): Declare.
5915 (nds32_output_lmw_single_word): Declare.
5916 (nds32_expand_unaligned_load): Declare.
5917 (nds32_expand_unaligned_store): Declare.
5918 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5919 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5920 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5921 NDS32_BUILTIN_UASTORE_DW.
5922 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5925 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5926 Kito Cheng <kito.cheng@gmail.com>
5927 Chung-Ju Wu <jasonwucj@gmail.com>
5929 * config/nds32/nds32-intrinsic.c
5930 (nds32_expand_builtin_null_ftype_reg): Delete.
5931 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5932 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5933 (nds32_read_argument): New.
5934 (nds32_legitimize_target): Ditto.
5935 (nds32_legitimize_argument): Ditto.
5936 (nds32_check_constant_argument): Ditto.
5937 (nds32_expand_unop_builtin): Ditto.
5938 (nds32_expand_unopimm_builtin): Ditto.
5939 (nds32_expand_binop_builtin): Ditto.
5940 (nds32_builtin_decl_impl): Ditto.
5941 (builtin_description): Ditto.
5942 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5943 (nds32_init_builtins_impl): Ditto.
5944 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5945 (nds32_builtin_decl): New.
5946 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5947 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5949 2018-03-02 Jeff Law <law@redhat.com>
5951 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5952 (redundant_insn, fill_simple_delay_slots): Likewise.
5953 (fill_slots_from_thread): Likewise.
5954 * resource.c (mark_referenced_resources): Likewise.
5955 (mark_set_resources, find_dead_or_set_registers): Likewise.
5957 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5959 * substring-locations.h (format_warning_va): Formatting fix for
5961 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5963 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5964 * substring-locations.c: Include intl.h.
5965 (format_warning_va): Turned into small wrapper around
5966 format_warning_n_va, renamed to ...
5967 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5968 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5970 (format_warning_at_substring_n): New function.
5971 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5972 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5973 format_warning_at_substring with just a shorter name instead of
5974 const function pointer.
5975 (fmtwarn_n): New function.
5976 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5977 appropriate, get rid of all the fmtstr temporaries, move conditionals
5978 with G_() wrapped string literals directly into fmtwarn arguments,
5979 cast dir.len to (int), formatting fixes.
5981 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5983 * doc/invoke.texi: Remove "Cilk Plus" references.
5985 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5986 Richard Biener <rguenther@suse.de>
5989 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5990 for error or warning attributes if CALL_FROM_THUNK_P is set.
5993 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5996 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5997 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6000 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6001 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6003 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6004 instead of -1U in last predictors element's probability member.
6006 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6009 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6010 arguments if they are comparable.
6012 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6014 PR tree-optimization/84634
6015 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6016 masks and masked_loop_p with a single loop_masks, making sure it's
6017 null for bb vectorization.
6019 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6021 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6022 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6023 loop->force_vectorize to check whether there is no alias.
6025 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6028 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6030 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6032 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6033 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6034 instead of a loop around prev_real_insn.
6035 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6039 * config/i386/i386.c (ix86_print_operand): Use conditional
6040 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6043 2018-03-02 Richard Biener <rguenther@suse.de>
6045 PR tree-optimization/84427
6046 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6047 (bitmap_set_subtract_values): Rewrite to handle multiple
6050 (prune_clobbered_mems): Likewise.
6051 (phi_translate): Take edge instead of pred/phiblock.
6052 (phi_translate_1): Likewise.
6053 (phi_translate_set): Likewise. Insert all translated
6054 exprs for a value into the set, keeping possibly multiple
6055 expressions per value.
6056 (compute_antic_aux): Adjust for phi_translate changes.
6057 When intersecting union the expressions and prune those
6058 not in the final value set, keeping possibly multiple
6059 expressions per value. Do not use value-insertion
6060 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6061 all expressions. Add verification that the value-sets
6062 only shrink during iteration.
6063 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6064 (do_pre_regular_insertion): Likewise.
6065 (do_pre_partial_partial_insertion): Likewise.
6067 2018-03-02 Richard Biener <rguenther@suse.de>
6070 * config/darwin.c (saved_debug_info_level): New static global.
6071 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6072 (darwin_asm_lto_end): Restore debug info generation settings.
6074 2018-03-01 Martin Liska <mliska@suse.cz>
6077 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6080 2018-03-01 Richard Biener <rguenther@suse.de>
6083 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6084 type annotation with LTO when debug was disabled at compile-time.
6086 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6088 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6090 (mips_final_postscan_insn): Likewise.
6092 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6094 PR rtl-optimization/84528
6095 * alias.c (init_alias_target): Add commentary.
6096 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6097 a unique base value if the frame pointer is not eliminated
6098 to the stack pointer.
6100 2018-03-01 Tom de Vries <tom@codesourcery.com>
6102 PR rtl-optimization/83327
6103 * lra-int.h (hard_regs_spilled_into): Declare.
6104 * lra.c (hard_regs_spilled_into): Define.
6105 (init_reg_info): Init hard_regs_spilled_into.
6106 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6107 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6108 (process_bb_lives): Handle hard_regs_spilled_into.
6109 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6110 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6112 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6114 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6115 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6116 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6117 * config/rs6000/aix72.h: New file.
6119 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6121 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6122 instead of warning_at with conditional singular and plural messages
6126 * stor-layout.c (update_alignment_for_field): For
6127 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6128 && !DECL_PACKED (field), do the alignment update, just use
6129 only desired_align instead of MAX (type_align, desired_align)
6131 (place_field): Don't do known_align < desired_align handling
6132 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6133 is non-NULL, instead do it after rli->prev_field handling and
6134 only if not within a bitfield word. For DECL_PACKED (field)
6135 use type_align of BITS_PER_UNIT.
6137 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6139 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6140 superfluous parentheses and trailing spaces.
6142 2018-02-28 Richard Biener <rguenther@suse.de>
6144 PR tree-optimization/84584
6145 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6146 SCoPs with fake exit edge.
6148 2018-02-28 Martin Liska <mliska@suse.cz>
6151 * timevar.c (timer::print): Fix format to properly print 100%
6154 2018-02-28 Richard Biener <rguenther@suse.de>
6157 * genmatch.c (capture_info::walk_match): Do not mark
6158 captured expressions without operands as expr_p given
6159 they act more like predicates and should be subject to
6160 "lost tail" side-effect preserving.
6162 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6164 PR rtl-optimization/81611
6165 * auto-inc-dec.c (attempt_change): Move dead note from
6166 mem_insn if it's the next use of regno
6167 (find_address): Take address use of reg holding
6168 non-incremented value. Add parm to limit search to the named
6170 (merge_in_block): Attempt to use a mem insn that is the next
6171 use of the original regno.
6173 2018-02-27 Martin Sebor <msebor@redhat.com>
6176 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6177 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6179 2018-02-27 Martin Sebor <msebor@redhat.com>
6181 PR translation/84207
6182 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6183 n argument to unsigned HOST_WIDE_INT.
6184 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6185 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6186 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6187 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6189 2018-02-27 Richard Biener <rguenther@suse.de>
6191 PR tree-optimization/84512
6192 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6193 Do not use the estimate returned from record_stmt_cost for
6194 the scalar iteration cost but sum properly using add_stmt_cost.
6196 2018-02-27 Richard Biener <rguenther@suse.de>
6198 PR tree-optimization/84466
6199 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6200 Adjust last change to less strictly validate use operands.
6202 2018-02-27 Martin Liska <mliska@suse.cz>
6204 PR gcov-profile/84548
6205 * gcov.c (process_file): Allow partial overlap and consider it
6206 also as group functions.
6207 (output_lines): Properly calculate range of lines for a group.
6209 2018-02-27 Martin Liska <mliska@suse.cz>
6211 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6212 'ggc' suffixes. Change first column width.
6213 (timer::print): Fix formatting of the column.
6215 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6217 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6218 preserve inline entry blocks for the sake of debug inline
6219 entry point markers alone.
6220 (remove_unused_locals): Suggest in comments a better place to
6221 force the preservation of inline entry blocks that are
6222 otherwise unused, but do not preserve them.
6224 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6226 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6228 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6231 * config/i386/constraints.md (Bs): Replace
6232 ix86_indirect_branch_register with
6233 TARGET_INDIRECT_BRANCH_REGISTER.
6235 * config/i386/i386.md (indirect_jump): Likewise.
6236 (tablejump): Likewise.
6237 (*sibcall_memory): Likewise.
6238 (*sibcall_value_memory): Likewise.
6239 Peepholes of indirect call and jump via memory: Likewise.
6240 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6241 (*sibcall_value_GOT_32): Likewise.
6242 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6243 (GOT_memory_operand): Likewise.
6244 (call_insn_operand): Likewise.
6245 (sibcall_insn_operand): Likewise.
6246 (GOT32_symbol_operand): Likewise.
6247 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6249 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6251 PR rtl-optimization/83496
6252 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6253 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6254 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6255 redundant insn, if any.
6256 (relax_delay_slots): Likewise.
6257 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6259 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6261 PR tree-optimization/83965
6262 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6263 that grouped statements are part of a reduction chain. Return
6264 true if the statement is not marked as a reduction itself but
6266 (vect_recog_dot_prod_pattern): Don't check whether the statement
6267 is part of a group here.
6268 (vect_recog_sad_pattern): Likewise.
6269 (vect_recog_widen_sum_pattern): Likewise.
6271 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6274 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6276 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6279 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6281 (ix86_output_indirect_function_return): New prototype.
6282 (ix86_split_simple_return_pop_internal): Likewise.
6283 * config/i386/i386.c (indirect_return_via_cx): New.
6284 (indirect_return_via_cx_bnd): Likewise.
6285 (indirect_thunk_name): Handle return va CX_REG.
6286 (output_indirect_thunk_function): Create alias for
6287 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6288 (ix86_output_indirect_jmp): Remove the bool argument.
6289 (ix86_output_indirect_function_return): New function.
6290 (ix86_split_simple_return_pop_internal): Likewise.
6291 * config/i386/i386.md (*indirect_jump): Don't pass false
6292 to ix86_output_indirect_jmp.
6293 (*tablejump_1): Likewise.
6294 (simple_return_pop_internal): Change it to define_insn_and_split.
6295 Call ix86_split_simple_return_pop_internal to split it for
6297 (simple_return_indirect_internal): Call
6298 ix86_output_indirect_function_return instead of
6299 ix86_output_indirect_jmp.
6301 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6304 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6305 memset and value initialization afterwards.
6307 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6309 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6311 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6314 * common/config/aarch64/aarch64-common.c
6315 (aarch_option_optimization_table[]): Switch
6316 off fomit-frame-pointer
6318 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6319 Chung-Ju Wu <jasonwucj@gmail.com>
6321 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6323 (store_multiple): Ditto.
6325 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6327 * config.gcc: Add --with-cpu support for nds32 target.
6328 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6329 * config/nds32/nds32.opt: Add -mcpu= option.
6331 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6333 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6334 isel=yes): Warn for these deprecated options.
6336 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6338 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6339 ISA_2_5_MASKS_EMBEDDED.
6341 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6343 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6344 p->max as pointers rather than using iterative_hash_expr.
6346 2018-02-23 Carl Love <cel@us.ibm.com>
6348 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6349 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6351 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6352 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6353 P8V_BUILTIN_VEC_VUNSIGNED2.
6355 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6358 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6359 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6361 * lra-constraints.c (curr_insn_transform): Set up
6362 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6364 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6365 * lra-eliminations.c (spill_pseudos): Ditto.
6366 (process_insn_for_elimination): Ditto.
6367 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6368 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6369 LRA_NON_CLOBBERED_ALT.
6371 2018-02-22 Martin Sebor <msebor@redhat.com>
6373 PR tree-optimization/84480
6374 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6375 to maybe_diag_stxncpy_trunc. Call it.
6376 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6377 from gimple_fold_builtin_strcpy. Print inlining stack.
6378 (handle_builtin_stxncpy): Print inlining stack.
6379 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6381 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6384 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6385 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6386 and -fcheck-pointer-bounds are used together.
6387 (indirect_thunk_prefix): New enum.
6388 (indirect_thunk_need_prefix): New function.
6389 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6390 "_nt" instead of "_bnd" for NOTRACK prefix.
6391 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6392 (output_indirect_thunk_function): Likewise.
6394 (ix86_code_end): Update output_indirect_thunk_function calls.
6395 (ix86_output_indirect_branch_via_reg): Replace
6396 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6397 (ix86_output_indirect_branch_via_push): Likewise.
6398 (ix86_output_function_return): Likewise.
6399 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6400 incompatible with -fcf-protection=branch and
6401 -fcheck-pointer-bounds.
6403 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6406 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6407 Change gcc_assert call to output_operand_lossage.
6409 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6411 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6413 2018-02-22 DJ Delorie <dj@redhat.com>
6414 Sebastian Perta <sebastian.perta@renesas.com>
6415 Oleg Endo <olegendo@gcc.gnu.org>
6417 * config/rx/rx.c (rx_rtx_costs): New function.
6418 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6420 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6422 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6424 2018-02-22 Martin Liska <mliska@suse.cz>
6427 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6428 Add "native" as a possible value.
6430 2018-02-22 Martin Liska <mliska@suse.cz>
6433 * config/i386/i386.c (ix86_option_override_internal):
6434 Add "native" as a possible value for -march and -mtune.
6436 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6439 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6440 to all type variants.
6442 PR tree-optimization/84503
6443 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6444 width as info->bitpos + info->bitsize - start.
6445 (merged_store_group::merge_overlapping): Simplify width computation.
6446 (check_no_overlap): New function.
6447 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6448 start + width and last_order of the group, fail if check_no_overlap
6450 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6451 to group if check_no_overlap fails.
6453 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6455 * config/rs6000/altivec.md: Delete contraint arguments to
6456 define_expand, define_split, and define_peephole2, and in
6457 define_insn_and_split if always unused.
6458 * config/rs6000/darwin.md: Ditto.
6459 * config/rs6000/dfp.md: Ditto.
6460 * config/rs6000/rs6000.md: Ditto.
6461 * config/rs6000/sync.md: Ditto.
6462 * config/rs6000/vector.md: Ditto.
6463 * config/rs6000/vsx.md: Ditto.
6465 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6467 * config/rs6000/altivec.md: Write output control strings as braced
6468 blocks instead of double-quoted strings.
6469 * config/rs6000/darwin.md: Ditto.
6470 * config/rs6000/rs6000.md: Ditto.
6471 * config/rs6000/vector.md: Ditto.
6472 * config/rs6000/vsx.md: Ditto.
6474 2018-02-21 Jason Merrill <jason@redhat.com>
6476 PR c++/84314 - ICE with templates and fastcall attribute.
6477 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6479 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6481 * ipa-cp.c (determine_versionability): Fix comment typos.
6483 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6486 * ipa-cp.c (determine_versionability): Do not version functions caling
6489 2018-02-21 Martin Liska <mliska@suse.cz>
6492 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6493 Add "native" as a possible value.
6494 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6495 the macro when native cpu detection is available.
6497 2018-02-21 Martin Liska <mliska@suse.cz>
6500 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6501 Add "native" as a possible value.
6502 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6503 when native cpu detection is available.
6505 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6506 Martin Sebor <msebor@redhat.com>
6508 PR tree-optimization/84478
6509 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6511 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6512 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6513 support which is conservatively correct, for 2 only stay conservative
6514 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6515 argument to the 2 argument get_range_strlen, adjust 6 arg
6516 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6518 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6519 (gimple_fold_builtin_strlen): Pass true as last argument to
6522 2018-02-20 Martin Sebor <msebor@redhat.com>
6525 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6526 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6527 (builtin_memref::builtin_memref): Factor out parts into
6528 set_base_and_offset and call it.
6530 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6533 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6534 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6535 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6536 search at the associated MODE_INT.
6538 2018-02-20 Jeff Law <law@redhat.com>
6541 PR tree-optimization/81592
6543 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6544 for range data rather than using global data.
6545 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6546 range data rather than using global data.
6547 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6548 pass it to children as needed.
6549 (struct directive::fmtresult): Similarly.
6550 (struct directive::set_width): Similarly.
6551 (struct directive::set_precision): Similarly.
6552 (format_integer, format_directive, parse_directive): Similarly.
6553 (format_none): Accept unnamed vr_values parameter.
6554 (format_percent, format_floating, format_character): Similarly.
6555 (format_string, format_plain): Similarly.
6556 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6557 the EVRP range analyzer for range data rather than using global data.
6558 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6559 gimple-ssa-evrp-analyze.h
6560 (class sprintf_dom_walker): Add after_dom_children member function.
6561 Add evrp_range_analyzer member.
6562 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6563 range analyzer as needed.
6564 (sprintf_dom_walker::after_dom_children): New member function.
6565 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6567 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6568 (evrp_range_analyzer::pop_to_marker): Likewise.
6570 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6572 PR tree-optimization/84419
6573 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6574 with the required type if its current type is compatible but
6577 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6580 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6581 after vectorization.
6583 2018-02-20 Martin Liska <mliska@suse.cz>
6586 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6587 possible values if we don't have a hint.
6589 2018-02-20 Martin Liska <mliska@suse.cz>
6593 * final.c (shorten_branches): Build align_tab array with one
6595 * opts.c (finish_options): Add alignment option limit check.
6596 (MAX_CODE_ALIGN): Likewise.
6597 (MAX_CODE_ALIGN_VALUE): Likewise.
6598 * doc/invoke.texi: Document maximum allowed option value for
6599 all -falign-* options.
6601 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6604 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6605 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6606 * var-tracking.c (emit_note_insn_var_location): Remove all references
6607 to NOTE_INSN_CALL_ARG_LOCATION.
6608 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6609 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6610 Use copy_rtx_if_shared.
6611 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6612 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6613 (dwarf2out_var_location): Remove handling of
6614 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6616 * final.c (final_scan_insn): Remove all references to
6617 NOTE_INSN_CALL_ARG_LOCATION.
6618 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6619 before dumping final insns.
6620 * except.c (emit_note_eh_region_end): Remove all references to
6621 NOTE_INSN_CALL_ARG_LOCATION.
6622 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6623 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6624 * config/arc/arc.c (hwloop_optimize): Likewise.
6625 * config/arm/arm.c (create_fix_barrier): Likewise.
6626 * config/s390/s390.c (s390_chunkify_start): Likewise.
6627 * config/sh/sh.c (find_barrier): Likewise.
6628 * config/i386/i386.c (rest_of_insert_endbranch,
6629 ix86_seh_fixup_eh_fallthru): Likewise.
6630 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6631 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6632 * config/frv/frv.c (frv_function_prologue): Likewise.
6633 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6635 (note_outside_basic_block_p): Remove all references to
6636 NOTE_INSN_CALL_ARG_LOCATION.
6637 * gengtype.c (adjust_field_rtx_def): Likewise.
6638 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6640 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6641 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6644 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6647 PR tree-optimization/84452
6648 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6649 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6652 2018-02-19 Martin Liska <mliska@suse.cz>
6655 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6657 2018-02-19 Martin Liska <mliska@suse.cz>
6658 Richard Sandiford <richard.sandiford@linaro.org>
6660 PR tree-optimization/82491
6661 * gimple-fold.c (get_base_constructor): Make earlier bail out
6664 2018-02-19 Carl Love <cel@us.ibm.com>
6666 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6667 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6669 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6670 P8V_BUILTIN_VEC_NEG.
6672 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6674 * config/rl78/rl78.md (movdf): New define expand.
6676 2018-02-19 Martin Liska <mliska@suse.cz>
6679 * doc/invoke.texi: Fix typo.
6680 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6682 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6684 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6685 handle rs6000_single_float and rs6000_double_float specially for
6688 2018-02-16 Jeff Law <law@redhat.com>
6690 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6693 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6696 * ipa-inline.c (inline_small_functions): Fix a typo.
6698 2018-02-16 Nathan Sidwell <nathan@acm.org>
6700 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6702 2018-02-16 Carl Love <cel@us.ibm.com>
6704 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6705 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6706 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6707 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6708 expansion to P8V_BUILTIN_VEC_FLOAT2.
6710 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6712 PR rtl-optimization/70023
6713 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6714 src_regno into account.
6716 2018-02-16 Carl Love <cel@us.ibm.com>
6718 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6719 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6720 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6721 * config/rs6000/rs6000.c: Remove case statements for
6722 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6723 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6724 and P9V_BUILTIN_VEC_VINSERT4B.
6725 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6726 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6727 * config/rs6000/vsx.md:
6728 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6731 2018-02-16 Carl Love <cel@us.ibm.com>
6733 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6735 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6737 * config/rs6000/rs6000-c.c: Add the definitions for
6738 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6739 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6740 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6741 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6742 definition for insert4b and define insn *insert3b_internal.
6743 * doc/extend.texi: Add documentation for vec_extract4b.
6745 2018-02-16 Nathan Sidwell <nathan@acm.org>
6747 * doc/extend.texi (Backwards Compatibility): Mention friend
6748 injection. Note for-scope is deprecated.
6749 * doc/invoke.texi (-ffriend-injection): Deprecate.
6751 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6753 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6754 that moved to I2, also allow destinations that are a paradoxical
6755 subreg (instead of a normal reg).
6757 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6760 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6763 2018-02-16 Richard Biener <rguenther@suse.de>
6765 PR tree-optimization/84037
6766 PR tree-optimization/84016
6768 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6769 Adjust vec_construct for the fact we need additional higher latency
6770 128bit inserts for AVX256 and AVX512 vector builds.
6771 (ix86_add_stmt_cost): Scale vector construction cost for
6774 2018-02-16 Richard Biener <rguenther@suse.de>
6776 PR tree-optimization/84417
6777 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6778 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6779 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6781 2018-02-16 Martin Liska <mliska@suse.cz>
6784 * internal-fn.def (ASAN_CHECK): Set proper flags.
6785 (ASAN_MARK): Likewise.
6787 2018-02-16 Julia Koval <julia.koval@intel.com>
6789 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6790 from PTA_CANNONLAKE.
6792 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6795 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6796 Use ++iter rather than iter++ for std::list iterators.
6797 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6798 defer deleting them until all nodes in the forest are processed. Do
6799 free even leaf nodes. Change to_process into auto_vec.
6802 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6803 * vec.h (vec_default_construct): Use memset instead of placement new
6804 if BROKEN_VALUE_INITIALIZATION is defined.
6805 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6806 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6809 PR rtl-optimization/83723
6810 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6811 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6812 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6814 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6816 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6818 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6820 PR rtl-optimization/81443
6821 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6822 from inner REGs to paradoxical SUBREGs.
6824 2018-02-16 Richard Biener <rguenther@suse.de>
6826 PR tree-optimization/84399
6827 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6828 For operands we can analyze at their definition make sure we can
6829 analyze them at each use as well.
6831 2018-02-16 Richard Biener <rguenther@suse.de>
6833 PR tree-optimization/84190
6834 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6835 volatile accesses if the decl isn't volatile.
6837 2018-02-15 Jason Merrill <jason@redhat.com>
6839 PR c++/84314 - ICE with templates and fastcall attribute.
6840 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6841 TYPE_CANONICAL on an existing type.
6843 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6845 PR tree-optimization/84383
6846 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6847 dstoff nor call operand_equal_p if dstbase is NULL.
6849 PR tree-optimization/84334
6850 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6851 also a CONSTANT_CLASS_P, punt.
6853 2018-02-14 Jim Wilson <jimw@sifive.com>
6855 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6856 first SMALL_OPERAND check. New local min_second_step. Move assert
6857 to where locals are set. Add TARGET_RVC support.
6858 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6860 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6862 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6864 2018-02-14 Martin Sebor <msebor@redhat.com>
6866 PR tree-optimization/83698
6867 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6868 arrays constrain the offset range to their bounds.
6869 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6870 (builtin_access::overlap): Avoid setting the size of overlap if it's
6872 (maybe_diag_overlap): Also consider arrays when deciding what values
6873 of offsets to include in diagnostics.
6875 2018-02-14 Martin Sebor <msebor@redhat.com>
6878 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6879 that correspond to the kind of a declaration.
6881 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6884 * config/pa/pa.md: Load address of PIC label using the linkage table
6885 if the label is nonlocal.
6887 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6889 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6890 warning message if user requests -maltivec=be.
6891 * doc/invoke.texi: Document deprecation of -maltivec=be.
6893 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6896 * config/rs6000/rs6000-c.c: Update definitions for
6897 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6898 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6900 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6903 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6904 add _get_ssp intrinsics. Remove argument from
6905 __builtin_ia32_rdssp[d|q].
6906 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6907 * config/i386/i386-builtin.def: Remove argument from
6908 __builtin_ia32_rdssp[d|q].
6909 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6910 ix86_expand_special_args_builtin for _rdssp[d|q].
6911 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6912 Clear register before usage.
6913 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6914 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6916 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6918 PR tree-optimization/84357
6919 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6920 operand 1 of an ARRAY_REF too.
6922 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6925 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6926 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6928 (set_of_reg): New struct.
6929 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6930 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6931 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6933 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6934 Split into bitclr, bitset, bitinvert patterns if appropriate.
6935 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6936 use rx_fuse_in_memory_bitop.
6937 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6938 to named insn, correct maximum insn length.
6940 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6943 * machmode.def: Define a complex mode for PARTIAL_INT.
6944 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6946 * doc/rtl.texi: Document CSPImode.
6947 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6949 (msp430_hard_regno_nregs_with_padding): Likewise.
6951 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6954 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6956 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6958 PR rtl-optimization/84169
6959 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6960 we generated a parallel as new i3 and we split that to new i2 and i3
6961 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6962 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6963 those to i2, not i1. Partially rewrite this scan code.
6965 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6968 * stor-layout.c (place_field): For variable length fields, adjust
6969 offset_align afterwards not just based on the field's alignment,
6970 but also on the size.
6973 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6974 of exps and logs in the use_exp2 case.
6976 2018-02-13 Jeff Law <law@redhat.com>
6978 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6981 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6984 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6988 * common.opt (gas-loc-support, gas-locview-support): New.
6989 (ginline-points, ginternal-reset-location-views): New.
6990 * doc/invoke.texi: Document them. Use @itemx where intended.
6991 (gvariable-location-views): Adjust.
6992 * target.def (reset_location_view): New.
6993 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6994 (TARGET_RESET_LOCATION_VIEW): New.
6995 * doc/tm.texi: Rebuilt.
6996 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6997 (dwarf2out_default_as_locview_support): New.
6998 (output_asm_line_debug_info): Use option variables.
6999 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7000 (output_loc_list): Likewise.
7001 (add_high_low_attributes): Check option variables.
7002 Don't output entry view attribute in strict mode.
7003 (gen_inlined_subroutine_die): Check option variables.
7004 (dwarf2out_inline_entry): Likewise.
7005 (init_sections_and_labels): Likewise.
7006 (dwarf2out_early_finish): Likewise.
7007 (maybe_reset_location_view): New, from...
7008 (dwarf2out_var_location): ... here. Call it.
7009 * debug.h (dwarf2out_default_as_loc_support): Declare.
7010 (dwarf2out_default_as_locview_support): Declare.
7011 * hooks.c (hook_int_rtx_insn_0): New.
7012 * hooks.h (hook_int_rtx_insn_0): Declare.
7013 * toplev.c (process_options): Take -gas-loc-support and
7014 -gas-locview-support from dwarf2out. Enable
7015 -gvariable-location-views by default only with locview
7016 assembler support. Enable -ginternal-reset-location-views by
7017 default only if the target defines the corresponding hook.
7018 Enable -ginline-points by default if location views are
7019 enabled; force it disabled if statement frontiers are
7021 * tree-inline.c (expand_call_inline): Check option variables.
7022 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7024 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7026 PR tree-optimization/84321
7027 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7028 handling. Also check whether the anti-range contains any values
7029 that satisfy the mask; switch to a VR_RANGE if not.
7031 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7034 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7036 2018-02-13 Martin Jambor <mjambor@suse.cz>
7039 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7040 of call statements, also set location of a load to a temporary.
7042 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7044 * config/rl78/rl78.c (add_vector_labels): New function.
7045 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7046 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7047 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7048 which checks that no arguments are passed.
7049 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7050 * doc/extend.texi: Documentation for the new attribute.
7052 2018-02-13 Andreas Schwab <schwab@suse.de>
7054 * config/riscv/linux.h (CPP_SPEC): Define.
7056 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7059 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7060 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7061 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7062 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7063 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7064 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7065 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7066 temporarily for AES and PCLMUL builtins.
7068 PR tree-optimization/84339
7069 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7070 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7074 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7075 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7076 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7078 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7080 * omp-simd-clone.h: New file.
7081 * omp-simd-clone.c: Include omp-simd-clone.h.
7082 (expand_simd_clones): No longer static.
7083 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7084 cgraph.h and omp-simd-clone.h.
7085 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7086 (vect_recog_widen_shift_pattern): Formatting fix.
7087 (vect_pattern_recog_1): Don't check optab for calls.
7090 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7091 operands[2] into a REG before using gen_lowpart on it.
7093 2018-02-12 Jeff Law <law@redhat.com>
7096 * config/sh/sh.c (find_barrier): Consider a sibling call
7099 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7100 successfully back substituting a reg.
7102 2018-02-12 Richard Biener <rguenther@suse.de>
7104 PR tree-optimization/84037
7105 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7106 parameter, move visited init to caller.
7107 (vect_slp_analyze_operations): Separate cost from validity
7108 check, initialize visited once for all instances.
7109 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7111 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7113 (vect_model_promotion_demotion_cost): Likewise.
7114 (vectorizable_bswap): Guard cost modeling with !slp_node
7115 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7117 (vectorizable_call): Likewise.
7118 (vectorizable_conversion): Likewise.
7119 (vectorizable_assignment): Likewise.
7120 (vectorizable_shift): Likewise.
7121 (vectorizable_operation): Likewise.
7122 (vectorizable_store): Likewise.
7123 (vectorizable_load): Likewise.
7124 (vectorizable_condition): Likewise.
7125 (vectorizable_comparison): Likewise.
7127 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7130 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7131 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7132 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7134 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7137 * params.def (inline-min-speedup): Increase from 8 to 15.
7138 (max-inline-insns-auto): Decrease from 40 to 30.
7139 * ipa-split.c (consider_split): Add some buffer for function to
7140 be considered inlining candidate.
7141 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7144 2018-02-12 Richard Biener <rguenther@suse.de>
7146 PR tree-optimization/84037
7147 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7148 matched stmts if we cannot swap the non-matched ones.
7150 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7152 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7153 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7154 _mm_maskz_scalef_round_ss): New intrinsics.
7155 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7156 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7157 __builtin_ia32_scalefss_round): Remove.
7158 (__builtin_ia32_scalefsd_mask_round,
7159 __builtin_ia32_scalefss_mask_round): New intrinsics.
7160 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7161 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7162 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7163 "<round_constraint>")): Changed to ...
7164 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7165 "<round_scalar_constraint>")): ... this.
7166 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7167 %0, %1, %2<round_op3>}"): Changed to ...
7168 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7169 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7170 %2<round_scalar_mask_op3>}"): ... this.
7171 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7173 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7175 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7176 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7177 (_mm_maskz_sqrt_round_ss): New intrinsics.
7178 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7179 (__builtin_ia32_sqrtsd_mask_round)
7180 (__builtin_ia32_sqrtss_mask_round): New builtins.
7181 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7182 (__builtin_ia32_sqrtss_round): Remove.
7183 (__builtin_ia32_sqrtsd_mask_round)
7184 (__builtin_ia32_sqrtss_mask_round): New builtins.
7185 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7186 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7187 ((match_operand:VF_128 1 "vector_operand"
7188 "xBm,<round_constraint>")): Changed to ...
7189 ((match_operand:VF_128 1 "vector_operand"
7190 "xBm,<round_scalar_constraint>")): ... this.
7191 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7192 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7193 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7194 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7195 %<iptr>1<round_scalar_mask_op3>}): ... this.
7196 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7197 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7199 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7202 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7203 Cast vec_cmpeq result to correct type.
7204 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7205 Cast vec_cmpgt result to correct type.
7207 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7209 * final.c (final_scan_insn_1): Renamed from...
7210 (final_scan_insn): ... this. New wrapper, to recover
7211 seen from the outermost call in recursive ones.
7212 * config/sparc/sparc.c (output_return): Drop seen from call.
7213 (output_sibcall): Likewise.
7214 * config/visium/visium.c (output_branch): Likewise.
7216 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7218 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7221 2018-02-10 Alan Modra <amodra@gmail.com>
7224 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7225 Specify LR as an input.
7227 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7230 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7231 remove_member_access_dummy_vars): New functions.
7232 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7233 lower_omp_1, execute_lower_omp): Use them.
7235 PR rtl-optimization/84308
7236 * shrink-wrap.c (spread_components): Release todo vector.
7238 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7240 PR rtl-optimization/57193
7241 * ira-color.c (struct allocno_color_data): Add member
7242 conflict_allocno_hard_prefs.
7243 (update_conflict_allocno_hard_prefs): New.
7244 (bucket_allocno_compare_func): Add a preference based on
7245 conflict_allocno_hard_prefs.
7246 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7247 (color_allocnos): Remove a dead code. Initiate
7248 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7250 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7253 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7254 constraint from =wa to wa. Avoid a subreg on the output operand,
7255 instead use a pseudo and subreg it in a move.
7256 (p9_xxbrd_<mode>): Changed to ...
7257 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7258 (p9_xxbrd_v2df): New expander.
7259 (p9_xxbrw_<mode>): Changed to ...
7260 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7261 (p9_xxbrw_v4sf): New expander.
7263 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7265 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7267 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7270 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7271 multiply in 32-bit mode.
7272 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7273 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7276 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7278 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7279 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7280 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7281 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7283 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7286 * dwarf2out.c (is_trivial_indirect_ref): New function.
7287 (dwarf2out_late_global_decl): Do not generate a location
7288 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7289 and that are not defined in the current unit.
7291 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7293 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7294 instead of a libcall for UNORDERED.
7296 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7299 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7300 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7302 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7305 * config/s390/s390.c (s390_set_current_function): Invoke
7306 s390_indirect_branch_settings also if fndecl didn't change.
7308 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7310 * config/rs6000/rs6000.md (blockage): Set length to zero.
7312 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7314 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7316 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7319 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7320 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7324 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7325 PARALLEL incoming that failed vt_get_decl_and_offset check.
7328 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7329 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7331 (get_variable_section): For decls in named .bss* sections pass true as
7332 second argument to bss_initializer_p.
7334 2018-02-09 Marek Polacek <polacek@redhat.com>
7335 Jakub Jelinek <jakub@redhat.com>
7338 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7339 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7340 Sync some changes from cxx_fold_indirect_ref.
7342 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7344 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7346 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7347 (BLOCK_INLINE_ENTRY_LABEL): New.
7348 (dwarf2out_var_location): Disregard inline entry markers.
7349 (inline_entry_data): New struct.
7350 (inline_entry_data_hasher): New hashtable type.
7351 (inline_entry_data_hasher::hash): New.
7352 (inline_entry_data_hasher::equal): New.
7353 (inline_entry_data_table): New variable.
7354 (add_high_low_attributes): Add DW_AT_entry_pc and
7355 DW_AT_GNU_entry_view attributes if a pending entry is found
7356 in inline_entry_data_table. Add old entry_pc attribute only
7357 if debug nonbinding markers are disabled.
7358 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7359 markers are enabled.
7360 (block_within_block_p, dwarf2out_inline_entry): New.
7361 (dwarf2out_finish): Check that no entries remained in
7362 inline_entry_data_table.
7363 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7364 (final_scan_insn, notice_source_line): Likewise.
7365 (rest_of_clean_state): Skip inline entry markers.
7366 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7368 * gimple.c (gimple_build_debug_inline_entry): New.
7369 * gimple.h (enum gimple_debug_subcode): Add
7370 GIMPLE_DEBUG_INLINE_ENTRY.
7371 (gimple_build_debug_inline_entry): Declare.
7372 (gimple_debug_inline_entry_p): New.
7373 (gimple_debug_nonbind_marker_p): Adjust.
7374 * insn-notes.def (INLINE_ENTRY): New.
7375 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7376 inline entry marker notes.
7377 (print_insn): Likewise.
7378 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7379 (INSN_DEBUG_MARKER_KIND): Likewise.
7380 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7381 * tree-inline.c (expand_call_inline): Build and insert
7382 debug_inline_entry stmt.
7383 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7384 inline entry blocks early, if nonbind markers are enabled.
7385 (dump_scope_block): Dump fragment info.
7386 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7387 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7388 (gimple_build_debug_inline_entry): New.
7389 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7390 Enable/disable inline entry points too.
7391 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7392 (DEBUG_INSN): Describe inline entry markers.
7394 * common.opt (gvariable-location-views): New.
7395 (gvariable-location-views=incompat5): New.
7396 * config.in: Rebuilt.
7397 * configure: Rebuilt.
7398 * configure.ac: Test assembler for view support.
7399 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7400 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7401 * dwarf2out.c (var_loc_view): New typedef.
7402 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7403 (dwarf2out_locviews_in_attribute): New.
7404 (dwarf2out_locviews_in_loclist): New.
7405 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7406 (enum dw_line_info_opcode): Add LI_adv_address.
7407 (struct dw_line_info_table): Add view.
7408 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7409 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7410 (zero_view_p): New variable.
7411 (ZERO_VIEW_P): New macro.
7412 (output_asm_line_debug_info): New.
7413 (struct var_loc_node): Add view.
7414 (add_AT_view_list, AT_loc_list): New.
7415 (add_var_loc_to_decl): Add view param. Test it against last.
7416 (new_loc_list): Add view params. Record them.
7417 (AT_loc_list_ptr): Handle loc and view lists.
7418 (view_list_to_loc_list_val_node): New.
7419 (print_dw_val): Handle dw_val_class_view_list.
7420 (size_of_die): Likewise.
7421 (value_format): Likewise.
7422 (loc_list_has_views): New.
7423 (gen_llsym): Set vl_symbol too.
7424 (maybe_gen_llsym, skip_loc_list_entry): New.
7425 (dwarf2out_maybe_output_loclist_view_pair): New.
7426 (output_loc_list): Output view list or entries too.
7427 (output_view_list_offset): New.
7428 (output_die): Handle dw_val_class_view_list.
7429 (output_dwarf_version): New.
7430 (output_compilation_unit_header): Use it.
7431 (output_skeleton_debug_sections): Likewise.
7432 (output_rnglists, output_line_info): Likewise.
7433 (output_pubnames, output_aranges): Update version comments.
7434 (output_one_line_info_table): Output view numbers in asm comments.
7435 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7436 Call maybe_gen_llsym.
7437 (loc_list_from_tree_1): Adjust.
7438 (add_AT_location_description): Create view list attribute if
7439 needed, check it's absent otherwise.
7440 (convert_cfa_to_fb_loc_list): Adjust.
7441 (maybe_emit_file): Call output_asm_line_debug_info for test.
7442 (dwarf2out_var_location): Reset views as needed. Precompute
7443 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7444 attribute. Set view.
7445 (new_line_info_table): Reset next view.
7446 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7447 (dwarf2out_source_line): Likewise. Output view resets and labels to
7448 the assembler, or select appropriate line info opcodes.
7449 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7450 (optimize_string_length): Catch it. Adjust.
7451 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7452 dw_val_class_view_list, and remove it if no longer needed.
7453 (hash_loc_list): Hash view numbers.
7454 (loc_list_hasher::equal): Compare them.
7455 (optimize_location_lists): Check whether a view list symbol is
7456 needed, and whether the locview attribute is present, and
7457 whether they match. Remove the locview attribute if no longer
7459 (index_location_lists): Call skip_loc_list_entry for test.
7460 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7461 Use output_dwarf_version.
7462 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7463 (struct dw_val_node): Add val_view_list.
7464 * final.c (SEEN_NEXT_VIEW): New.
7465 (set_next_view_needed): New.
7466 (clear_next_view_needed): New.
7467 (maybe_output_next_view): New.
7468 (final_start_function): Rename to...
7469 (final_start_function_1): ... this. Take pointer to FIRST,
7470 add SEEN parameter. Emit param bindings in the initial view.
7471 (final_start_function): Reintroduce SEEN-less interface.
7472 (final): Rename to...
7473 (final_1): ... this. Take SEEN parameter. Output final pending
7474 next view at the end.
7475 (final): Reintroduce seen-less interface.
7476 (final_scan_insn): Output pending next view before switching
7477 sections or ending a block. Mark the next view as needed when
7478 outputting variable locations. Notify debug backend of section
7479 changes, and of location view changes.
7480 (rest_of_handle_final): Adjust.
7481 * toplev.c (process_options): Autodetect value for debug variable
7482 location views option. Warn on incompat5 without -gdwarf-5.
7483 * doc/invoke.texi (gvariable-location-views): New.
7484 (gvariable-location-views=incompat5): New.
7485 (gno-variable-location-views): New.
7487 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7489 PR tree-optimization/84136
7490 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7491 that the result of find_edge is non-NULL.
7493 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7496 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7497 storing integer register in SImode. Fix cost of 256 and 512
7498 byte aligned SSE register store.
7500 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7502 * config/i386/i386.c (ix86_multiplication_cost): Fix
7503 multiplication cost for TARGET_AVX512DQ.
7505 2018-02-08 Marek Polacek <polacek@redhat.com>
7507 PR tree-optimization/84238
7508 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7511 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7513 PR tree-optimization/84265
7514 * tree-vect-stmts.c (vectorizable_store): Don't treat
7515 VMAT_CONTIGUOUS accesses as grouped.
7516 (vectorizable_load): Likewise.
7518 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7520 PR tree-optimization/81635
7521 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7522 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7523 (test_round_for_mask): New functions.
7524 (wide_int_cc_tests): Call test_round_for_mask.
7525 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7526 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7527 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7528 range returned by get_range_info.
7530 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7533 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7534 * symtab.c: Include builtins.h
7535 (symtab_node::output_to_lto_symbol_table_p): Move here
7536 from lto-streamer-out.c:output_symbol_p.
7537 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7538 (output_symbol_p): Move all logic to symtab.c
7539 (produce_symtab): Update.
7541 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7543 * config/s390/s390-opts.h (enum indirect_branch): Define.
7544 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7545 (s390_indirect_branch_via_thunk)
7546 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7547 (enum s390_indirect_branch_type): Define.
7548 * config/s390/s390.c (struct s390_frame_layout, struct
7549 machine_function): Remove.
7550 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7551 (indirect_branch_table_label_no, indirect_branch_table_name):
7553 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7554 (enum s390_indirect_branch_option): Define.
7555 (s390_return_addr_from_memory): New function.
7556 (s390_handle_string_attribute): New function.
7557 (s390_attribute_table): Add new attribute handler.
7558 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7559 (s390_indirect_branch_via_thunk): New function.
7560 (s390_indirect_branch_via_inline_thunk): New function.
7561 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7562 sibling call optimization for non z10 compiles.
7563 (s390_emit_call): Force indirect branch target to be a single
7564 register. Add r1 clobber for non-z10 compiles.
7565 (s390_emit_epilogue): Emit return jump via return_use expander.
7566 (s390_reorg): Handle JUMP_INSNs as execute targets.
7567 (s390_option_override_internal): Perform validity checks for the
7568 new command line options.
7569 (s390_indirect_branch_attrvalue): New function.
7570 (s390_indirect_branch_settings): New function.
7571 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7572 (s390_output_indirect_thunk_function): New function.
7573 (s390_code_end): Implement target hook.
7574 (s390_case_values_threshold): Implement target hook.
7575 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7577 * config/s390/s390.h (struct s390_frame_layout)
7578 (struct machine_function): Move here from s390.c.
7579 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7580 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7581 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7582 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7583 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7584 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7585 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7586 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7587 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7588 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7589 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7590 (mnemonic attribute): Add values which aren't recognized
7592 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7593 pattern for branch conversion. Fix mnemonic attribute.
7594 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7595 indirect branch via thunk if requested.
7596 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7597 ("*indirect_jump"): Disable for branch conversion using out of
7599 ("indirect_jump_via_thunk<mode>_z10")
7600 ("indirect_jump_via_thunk<mode>")
7601 ("indirect_jump_via_inlinethunk<mode>_z10")
7602 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7603 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7604 ("casesi_jump_via_inlinethunk<mode>_z10")
7605 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7606 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7607 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7608 ("*indirect2_jump"): Disable for branch conversion.
7609 ("casesi_jump"): Turn into expander and expand patterns for branch
7611 ("return_use"): New expander.
7612 ("*return"): Emit return via thunk and rename it to ...
7613 ("*return<mode>"): ... this one.
7614 * config/s390/s390.opt: Add new options and and enum for the
7617 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7619 * lra-constraints.c (match_reload): Unconditionally use
7620 gen_lowpart_SUBREG, rather than selecting between that
7621 and equivalent gen_rtx_SUBREG code.
7623 2018-02-08 Richard Biener <rguenther@suse.de>
7625 PR tree-optimization/84233
7626 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7627 changed flag instead of boguously re-using phi_inserted.
7629 2018-02-08 Martin Jambor <mjambor@suse.cz>
7631 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7632 static local variables.
7634 2018-02-08 Richard Biener <rguenther@suse.de>
7636 PR tree-optimization/84278
7637 * tree-vect-stmts.c (vectorizable_store): When looking for
7638 smaller vector types to perform grouped strided loads/stores
7639 make sure the mode is supported by the target.
7640 (vectorizable_load): Likewise.
7642 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7644 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7645 Increase LDP/STP opportunities by adding adjacent callee-saves.
7647 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7649 PR rtl-optimization/84068
7650 PR rtl-optimization/83459
7651 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7653 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7655 PR tree-optimization/84224
7656 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7657 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7660 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7663 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7664 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7665 count, remove one USE.
7667 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7669 * doc/install.texi (Configuration): Document the
7670 --with-long-double-format={ibm,ieee} PowerPC configuration
7674 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7675 Convert from define_expand to be define_insn_and_split. Rework
7676 float/double/_Float128 conversions to QI/HI/SImode to work with
7677 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7678 conversions to QI/HImode types did a store and then a load to
7679 truncate the value. For conversions to VSX registers, don't split
7680 the insn, instead emit the code directly. Use the code iterator
7681 any_fix to combine signed and unsigned conversions.
7682 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7683 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7684 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7685 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7686 (fix_<mode>di2_hw): Likewise.
7687 (fixuns_<mode>di2_hw): Likewise.
7688 (fix_<mode>si2_hw): Likewise.
7689 (fixuns_<mode>si2_hw): Likewise.
7690 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7691 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7692 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7693 fix<uns>_trunc<SFDF:mode>si2_p8.
7694 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7696 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7697 (fix<uns>_<mode>_mem): Likewise.
7698 (fctiw<u>z_<mode>_mem): Likewise.
7699 (fix<uns>_<mode>_mem): Likewise.
7700 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7701 the register allocator from doing a direct move to the GPRs to do
7702 a store, and instead use the ISA 3.0 store byte/half-word from
7703 vector register instruction. For IEEE 128-bit floating point,
7704 also optimize stores of 32-bit ints.
7705 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7707 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7709 * genextract.c (push_pathstr_operand): New function to support
7711 (walk_rtx): Call push_pathstr_operand.
7712 (print_path): Support [a-zA-Z].
7714 2018-02-07 Richard Biener <rguenther@suse.de>
7716 PR tree-optimization/84037
7717 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7718 (cse_and_gimplify_to_preheader): Declare.
7719 (vect_get_place_in_interleaving_chain): Likewise.
7720 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7722 (_loop_vec_info::~_loop_vec_info): Delete it.
7723 (cse_and_gimplify_to_preheader): New function.
7724 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7725 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7726 (vectorizable_load): Likewise. For grouped stores always base
7727 the IV on the first element.
7728 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7729 condition before gimplifying.
7731 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7733 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7734 *DIV_EXPR and *MOD_EXPR.
7736 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7739 * config/i386/i386.c (ix86_option_override_internal): Mask out
7740 the CF_SET bit when checking -fcf-protection.
7742 2018-02-07 Tom de Vries <tom@codesourcery.com>
7745 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7748 2018-02-07 Richard Biener <rguenther@suse.de>
7750 PR tree-optimization/84204
7751 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7754 PR tree-optimization/84205
7755 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7756 special-case isl_ast_op_zdiv_r.
7758 PR tree-optimization/84223
7759 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7760 Only add conditions from within the region.
7761 (gather_bbs::after_dom_children): Adjust.
7763 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7766 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7767 * config/avr/avr.md: Only post-reload split REG-REG moves if
7768 either register is GENERAL_REG_P.
7770 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7772 PR tree-optimization/84235
7773 * tree-ssa-scopedtables.c
7774 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7775 if the subtraction is performed in floating point type where NaNs are
7776 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7777 build 1. Formatting fix.
7779 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7782 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7783 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7784 and skip it regardless of bb boundaries. Use CALL_P macro,
7785 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7786 unnecessarily, formatting fix.
7788 2018-02-06 Michael Collison <michael.collison@arm.com>
7790 * config/arm/thumb2.md:
7791 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7792 (*thumb_mov_notscc): Ditto.
7794 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7797 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7798 unsigned_fix, not "s".
7800 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7802 * configure.ac (gcc_fn_eh_frame_ro): New function.
7803 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7804 correct .eh_frame permissions.
7805 * configure: Regenerate.
7807 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7809 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7812 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7814 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7815 Display warning message for -mno-speculate-indirect-jumps.
7817 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7819 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7821 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7823 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7825 PR tree-optimization/84225
7826 * tree-eh.c (find_trapping_overflow): Only call
7827 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7829 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7832 * config/i386/i386.c: Reimplement the check of possible options
7833 -mibt/-mshstk conbination. Change error messages.
7834 * doc/invoke.texi: Fix a typo: remove extra '='.
7836 2018-02-06 Marek Polacek <polacek@redhat.com>
7838 PR tree-optimization/84228
7839 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7841 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7844 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7845 emitted arch directives.
7846 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7847 __ARM_FEATURE_COPROC before changing architectures.
7849 2018-02-06 Richard Biener <rguenther@suse.de>
7851 * config/i386/i386.c (print_reg): Fix typo.
7852 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7854 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7856 * configure: Regenerate.
7858 2018-02-05 Martin Sebor <msebor@redhat.com>
7860 PR tree-optimization/83369
7861 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7864 2018-02-05 Martin Liska <mliska@suse.cz>
7866 * doc/invoke.texi: Cherry-pick upstream r323995.
7868 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7870 * ira.c (ira_init_register_move_cost): Adjust comment.
7872 2018-02-05 Martin Liska <mliska@suse.cz>
7874 PR gcov-profile/84137
7875 * doc/gcov.texi: Fix typo in documentation.
7877 2018-02-05 Martin Liska <mliska@suse.cz>
7879 PR gcov-profile/83879
7880 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7881 using dlopen functionality.
7883 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7885 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7886 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7887 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7888 _mm_maskz_range_round_ss): New intrinsics.
7889 (__builtin_ia32_rangesd128_round)
7890 (__builtin_ia32_rangess128_round): Remove.
7891 (__builtin_ia32_rangesd128_mask_round,
7892 __builtin_ia32_rangess128_mask_round): New builtins.
7893 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7894 __builtin_ia32_rangess128_round): Remove.
7895 (__builtin_ia32_rangesd128_mask_round,
7896 __builtin_ia32_rangess128_mask_round): New builtins.
7897 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7898 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7899 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7900 "<round_saeonly_constraint>")): Changed to ...
7901 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7902 "<round_saeonly_scalar_constraint>")): ... this.
7903 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7904 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7905 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7906 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7907 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7909 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7911 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7913 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7914 Remove all values except native, 8540 and 8548.
7916 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7918 * config/i386/i386.c (ix86_output_function_return): Pass
7919 INVALID_REGNUM, instead of -1, as invalid register number to
7920 indirect_thunk_name and output_indirect_thunk.
7922 2018-02-02 Julia Koval <julia.koval@intel.com>
7924 * config.gcc: Add -march=icelake.
7925 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7926 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7927 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7928 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7929 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7930 (processor_target_table): Add icelake.
7931 (ix86_option_override_internal): Handle new PTAs.
7932 (get_builtin_code_for_version): Handle icelake.
7933 (M_INTEL_COREI7_ICELAKE): New.
7934 (fold_builtin_cpu): Handle icelake.
7935 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7936 * doc/invoke.texi: Add -march=icelake.
7938 2018-02-02 Julia Koval <julia.koval@intel.com>
7940 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7941 to wide_int_bitmask.
7942 * wide-int-bitmask.h: New.
7944 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7947 * config/i386/i386.md: Replace Pmode with word_mode in
7948 builtin_setjmp_setup and builtin_longjmp to support x32.
7950 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7954 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7956 (rs6000_supported_cpu_names): New static variable.
7957 (linux_cpu_translation_table): Likewise.
7958 (elf_platform) <cpu>: Define new static variable and use it.
7959 Translate kernel AT_PLATFORM name to canonical name if needed.
7960 Error if platform name is unknown.
7962 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7965 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7967 2018-02-01 Jeff Law <law@redhat.com>
7970 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7971 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7972 the scratch if RELEASE_VIA_POP is false.
7973 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7974 If we have to save a temporary register, decrement SIZE appropriately.
7975 Pass new arguments to release_scratch_register_on_entry.
7976 (ix86_adjust_stack_and_probe): Likewise.
7977 (ix86_emit_probe_stack_range): Pass new arguments to
7978 release_scratch_register_on_entry.
7980 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7982 PR rtl-optimization/84157
7983 * combine.c (change_zero_ext): Use REG_P predicate in
7984 front of HARD_REGISTER_P predicate.
7986 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7988 * config/avr/avr.c (avr_option_override): Move disabling of
7989 -fdelete-null-pointer-checks to...
7990 * common/config/avr/avr-common.c (avr_option_optimization_table):
7993 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7995 PR tree-optimization/81635
7996 * tree-data-ref.c (split_constant_offset_1): For types that
7997 wrap on overflow, try to use range info to prove that wrapping
8000 2018-02-01 Renlin Li <renlin.li@arm.com>
8003 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8005 (aarch64_register_move_cost): Likewise.
8006 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8008 (REG_CLASS_NAMES): Likewise.
8009 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8010 TAILCALL_ADDR_REGS. Remove IP registers.
8011 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8013 2018-02-01 Richard Biener <rguenther@suse.de>
8015 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8016 for specifying RPO order and allow NULL for that.
8017 * domwalk.c (dom_walker::dom_walker): Likewise.
8018 (dom_walker::walk): Handle NULL RPO order.
8019 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8021 (rewrite_update_dom_walker): Likewise.
8022 (mark_def_dom_walker): Likewise.
8024 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8026 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8027 (aarch64_maybe_expand_sve_subreg_move): Declare.
8028 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8029 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8031 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8032 that are semantically a reverse operation.
8033 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8034 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8035 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8037 (aarch64_can_change_mode_class): For big-endian, forbid changes
8038 between two SVE modes if they have different element sizes.
8040 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8042 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8043 the TImode handling for big-endian targets.
8045 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8047 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8048 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8050 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8051 Remove BSWAP handing for big-endian targets and use the form of
8052 LD1RQ appropariate for the mode.
8054 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8056 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8057 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8060 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8063 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8064 check for operands that need to go through aarch64_sve_reload_be.
8066 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8068 PR tree-optimization/81661
8069 PR tree-optimization/84117
8070 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8071 * tree-eh.c: Include gimplify.h.
8072 (find_trapping_overflow, replace_trapping_overflow,
8073 rewrite_to_non_trapping_overflow): New functions.
8074 * tree-vect-loop.c: Include tree-eh.h.
8075 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8076 * tree-data-ref.c: Include tree-eh.h.
8077 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8079 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8081 PR rtl-optimization/84123
8082 * combine.c (change_zero_ext): Check if hard register satisfies
8083 can_change_dest_mode before calling gen_lowpart_SUBREG.
8085 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8088 * ira.c (ira_init_register_move_cost): Remove assert.
8090 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8092 PR rtl-optimization/84071
8093 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8094 * doc/tm.texi: Regenerate.
8096 2018-01-31 Richard Biener <rguenther@suse.de>
8098 PR tree-optimization/84132
8099 * tree-data-ref.c (analyze_miv_subscript): Properly
8100 check whether evolution_function_is_affine_multivariate_p
8101 before calling gcd_of_steps_may_divide_p.
8103 2018-01-31 Julia Koval <julia.koval@intel.com>
8106 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8107 * config/i386/i386.md (rdpid_rex64) New.
8108 (rdpid): Make 32bit only.
8110 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8113 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8114 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8116 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8119 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8121 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8123 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8125 PR rtl-optimization/84071
8126 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8127 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8129 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8131 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8132 (arc_attribute_table): Add 'aux' attribute.
8133 (arc_in_small_data_p): Consider aux like variables.
8134 (arc_is_aux_reg_p): New function.
8135 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8136 (arc_get_aux_arg): New function.
8137 (prepare_move_operands): Handle aux-register access.
8138 (arc_handle_aux_attribute): New function.
8139 * doc/extend.texi (ARC Variable attributes): Add subsection.
8141 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8143 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8144 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8145 (arc_attribute_table): Add 'uncached' attribute.
8146 (arc_print_operand): Print '.di' flag for uncached memory
8148 (arc_in_small_data_p): Do not consider for small data the uncached
8150 (arc_is_uncached_mem_p): New function.
8151 * config/arc/predicates.md (compact_store_memory_operand): Check
8152 for uncached memory accesses.
8153 (nonvol_nonimm_operand): Likewise.
8154 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8156 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8159 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8160 falign-loops=): Add Optimization flag.
8162 2018-01-30 Jeff Law <law@redhat.com>
8165 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8166 INT_REGISTERS_SAVED. Check it prior to calling
8167 get_scratch_register_on_entry.
8168 (ix86_adjust_stack_and_probe): Similarly.
8169 (ix86_emit_probe_stack_range): Similarly.
8170 (ix86_expand_prologue): Corresponding changes.
8172 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8175 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8176 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8178 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8181 * lra-constraints.c (curr_insn_transform): Process AND in the
8184 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8186 PR rtl-optimization/83986
8187 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8188 dependence against last_pending_memory_flush in addition to
8191 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8193 PR tree-optimization/81611
8194 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8197 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8200 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8203 2018-01-30 Richard Biener <rguenther@suse.de>
8204 Jakub Jelinek <jakub@redhat.com>
8206 PR tree-optimization/84111
8207 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8208 inner loops added during recursion, as they don't have up-to-date
8211 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8214 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8215 (can_inline_edge_by_limits_p): ... here.
8216 (can_early_inline_edge_p, check_callers,
8217 update_caller_keys, update_callee_keys, recursive_inlining,
8218 add_new_edges_to_heap, speculation_useful_p,
8219 inline_small_functions,
8220 inline_small_functions, flatten_function,
8221 inline_to_all_callers_1): Update.
8223 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8225 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8228 2018-01-30 Richard Biener <rguenther@suse.de>
8230 PR tree-optimization/83008
8231 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8232 invariant and constant vector uses in stmts when they need
8235 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8238 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8239 * configure: Regenerate.
8241 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8243 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8245 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8246 Use gen_rtx_REG rather than gen_lowpart.
8248 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8250 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8251 rather than 0 when creating partial subregs.
8253 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8255 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8258 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8261 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8262 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8263 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8264 flags. This restores the settings used before the 2017-07-24.
8265 Turning off pre increment/decrement/modify allows IVOPTS to
8266 optimize DF/SF loops where the index is an int.
8268 2018-01-29 Richard Biener <rguenther@suse.de>
8269 Kelvin Nilsen <kelvin@gcc.gnu.org>
8272 * tree-vect-stmts.c (vectorizable_call): Don't call
8273 targetm.vectorize_builtin_md_vectorized_function if callee is
8276 2018-01-22 Carl Love <cel@us.ibm.com>
8278 * doc/extend.tex: Fix typo in second arg in
8279 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8281 2018-01-29 Richard Biener <rguenther@suse.de>
8283 PR tree-optimization/84086
8284 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8285 (flush_ssaname_freelist): When SSA names were released reset
8286 the SCEV hash table.
8288 2018-01-29 Richard Biener <rguenther@suse.de>
8290 PR tree-optimization/84057
8291 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8292 removed paths when removing edges.
8294 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8296 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8297 -mfunction-return=@var{choice}.
8299 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8302 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8304 (layout::print_source_line): Likewise.
8305 (test_get_line_width_without_trailing_whitespace): Add test cases.
8307 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8310 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8313 2018-01-26 Jim Wilson <jimw@sifive.com>
8315 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8317 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8320 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8322 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8323 and CMP + SUB-immediate -> SUBS.
8325 2018-01-26 Martin Sebor <msebor@redhat.com>
8327 PR tree-optimization/83896
8328 * tree-ssa-strlen.c (get_string_len): Rename...
8329 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8330 Avoid assuming length is constant.
8331 (handle_char_store): Use HOST_WIDE_INT for string length.
8333 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8336 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8337 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8339 2018-01-26 Richard Biener <rguenther@suse.de>
8341 PR rtl-optimization/84003
8342 * dse.c (record_store): Only record redundant stores when
8343 the earlier store aliases at least all accesses the later one does.
8345 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8347 PR rtl-optimization/83985
8348 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8349 REG_CFA_RESTORE insns.
8350 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8351 REG_CFA_RESTORE insns here.
8354 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8355 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8357 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8359 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8360 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8361 (arc_init): Likewise.
8362 (arc_override_options): Likewise.
8363 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8365 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8367 * config/arc/arc.h (TARGET_DBNZ): Define.
8368 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8369 properly set the tune attribute.
8370 (dbnz): Use TARGET_DBNZ guard.
8371 * config/arc/arc.opt (mtune): Add core3 option.
8373 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8375 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8376 recognize new pic like addresses.
8377 (arc_delegitimize_address): Clean up.
8379 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8381 * config/arc/arc-arches.def: Option mrf16 valid for all
8383 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8384 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8385 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8386 * config/arc/arc-tables.opt: Regenerate.
8387 * config/arc/arc.c (arc_conditional_register_usage): Handle
8388 reduced register file case.
8389 (arc_file_start): Set must have build attributes.
8390 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8392 * config/arc/arc.opt (mrf16): Add new option.
8393 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8394 * config/arc/genmultilib.awk: Handle new mrf16 option.
8395 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8396 * config/arc/t-multilib: Regenerate.
8397 * doc/invoke.texi (ARC Options): Document mrf16 option.
8399 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8401 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8402 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8403 (arc_attribute_table): Add 'secure_call' attribute.
8404 (arc_print_operand): Print secure call operand.
8405 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8407 (arc_is_secure_call_p): New function. * config/arc/arc.md
8408 (call_i): Add support for sjli instruction.
8409 (call_value_i): Likewise.
8410 * config/arc/constraints.md (Csc): New constraint.
8412 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8413 John Eric Martin <John.Martin@emmicro-us.com>
8415 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8416 * config/arc/arc.c (_arc_jli_section): New struct.
8417 (arc_jli_section): New type.
8418 (rc_jli_sections): New static variable.
8419 (arc_handle_jli_attribute): New function.
8420 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8421 (arc_file_end): New function.
8422 (TARGET_ASM_FILE_END): Define.
8423 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8424 (arc_add_jli_section): New function.
8425 (jli_call_scan): Likewise.
8426 (arc_reorg): Call jli_call_scan.
8427 (arc_output_addsi): Remove 'S' from printing asm operand.
8428 (arc_is_jli_call_p): New function.
8429 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8431 (movhi_insn): Likewise.
8432 (movsi_insn): Likewise.
8433 (movsi_set_cc_insn): Likewise.
8434 (loadqi_update): Likewise.
8435 (load_zeroextendqisi_update): Likewise.
8436 (load_signextendqisi_update): Likewise.
8437 (loadhi_update): Likewise.
8438 (load_zeroextendhisi_update): Likewise.
8439 (load_signextendhisi_update): Likewise.
8440 (loadsi_update): Likewise.
8441 (loadsf_update): Likewise.
8442 (movsicc_insn): Likewise.
8443 (bset_insn): Likewise.
8444 (bxor_insn): Likewise.
8445 (bclr_insn): Likewise.
8446 (bmsk_insn): Likewise.
8447 (bicsi3_insn): Likewise.
8448 (cmpsi_cc_c_insn): Likewise.
8449 (movsi_ne): Likewise.
8450 (movsi_cond_exec): Likewise.
8451 (clrsbsi2): Likewise.
8462 (call_i): Remove 'S' asm letter, add jli instruction.
8463 (call_value_i): Likewise.
8464 * config/arc/arc.op (mjli-always): New option.
8465 * config/arc/constraints.md (Cji): New constraint.
8466 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8468 (subsf3_fpx): Likewise.
8469 (mulsf3_fpx): Likewise.
8470 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8472 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8474 * doc/invoke.texi (ARC): Document mjli-always option.
8476 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8478 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8479 avoid addition with 0 and use incw and decw where possible.
8481 2018-01-26 Richard Biener <rguenther@suse.de>
8483 PR tree-optimization/81082
8484 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8485 association if it requires casting to unsigned.
8486 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8487 from fold_plusminus_mult_expr to catch important cases late when
8488 range info is available.
8490 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8492 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8493 * configure.ac (hidden_linkonce): New test.
8494 * configure: Regenerate.
8495 * config.in: Regenerate.
8497 2018-01-26 Julia Koval <julia.koval@intel.com>
8499 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8500 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8501 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8502 _mm_mask_bitshuffle_epi64_mask): Fix type.
8503 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8504 USI_FTYPE_V4DI_V4DI_USI): Remove.
8505 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8506 __builtin_ia32_vpshufbitqmb256_mask,
8507 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8508 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8509 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8511 2018-01-26 Alan Modra <amodra@gmail.com>
8514 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8515 UNSPEC_VBPERMQ. Sort other unspecs.
8517 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8519 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8521 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8524 * predict.c (drop_profile): Do not push/pop cfun; update also
8526 (handle_missing_profiles): Fix logic looking for zero profiles.
8528 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8531 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8532 on functions with #pragma omp declare simd or functions with simd
8534 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8535 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8536 Remove trailing \n from warning_at calls.
8538 2018-01-25 Tom de Vries <tom@codesourcery.com>
8541 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8542 for neutered workers.
8544 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8547 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8548 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8550 2018-01-24 Jeff Law <law@redhat.com>
8553 * i386.c (get_probe_interval): Move to earlier point.
8554 (ix86_compute_frame_layout): If -fstack-clash-protection and
8555 the frame is larger than the probe interval, then use pushes
8556 to save registers rather than reg->mem moves.
8557 (ix86_expand_prologue): Remove conditional for int_registers_saved
8560 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8563 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8564 min/max for never referenced object.
8566 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8569 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8571 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8572 attributes from DECL_ATTRIBUTES (decl) without affecting
8573 DECL_ATTRIBUTES (current_function_decl).
8574 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8575 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8577 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8579 PR tree-optimization/83979
8580 * fold-const.c (fold_comparison): Use constant_boolean_node
8581 instead of boolean_{true,false}_node.
8583 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8585 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8588 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8590 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8591 Simplify the clause that sets the length attribute.
8592 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8593 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8594 clause that sets the length attribute.
8595 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8597 2018-01-24 Tom de Vries <tom@codesourcery.com>
8600 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8601 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8602 Add strict parameter.
8603 (prevent_branch_around_nothing): Insert dummy insn between branch to
8604 label and label with no ptx insn inbetween.
8605 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8607 2018-01-24 Tom de Vries <tom@codesourcery.com>
8610 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8611 for neutered threads in warp.
8612 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8614 2018-01-24 Richard Biener <rguenther@suse.de>
8616 PR tree-optimization/83176
8617 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8620 2018-01-24 Richard Biener <rguenther@suse.de>
8622 PR tree-optimization/82819
8623 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8624 code generating pluses that are no-ops in the target precision.
8626 2018-01-24 Richard Biener <rguenther@suse.de>
8629 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8631 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8633 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8634 to merge probabilities.
8635 * predict.c (probably_never_executed): Also mark as cold functions
8636 with global 0 profile and guessed local profile.
8637 * profile-count.c (profile_probability::combine_with_count): New
8639 * profile-count.h (profile_probability::operator*,
8640 profile_probability::operator*=, profile_probability::operator/,
8641 profile_probability::operator/=): Reduce precision to adjusted
8642 and set value to guessed on contradictory divisions.
8643 (profile_probability::combine_with_freq): Remove.
8644 (profile_probability::combine_wiht_count): Declare.
8645 (profile_count::force_nonzero):: Set to adjusted.
8646 (profile_count::probability_in):: Set quality to adjusted.
8647 * tree-ssa-tail-merge.c (replace_block_by): Use
8650 2018-01-23 Andrew Waterman <andrew@sifive.com>
8651 Jim Wilson <jimw@sifive.com>
8653 * config/riscv/riscv.c (riscv_stack_boundary): New.
8654 (riscv_option_override): Set riscv_stack_boundary. Handle
8655 riscv_preferred_stack_boundary_arg.
8656 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8657 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8658 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8659 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8660 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8661 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8663 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8666 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8667 of struct ix86_frame.
8668 (ix86_expand_epilogue): Likewise. Add a local variable for
8669 the reg_save_offset field in struct ix86_frame.
8671 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8673 PR tree-optimization/82604
8674 * tree-loop-distribution.c (enum partition_kind): New enum item
8675 PKIND_PARTIAL_MEMSET.
8676 (partition_builtin_p): Support above new enum item.
8677 (generate_code_for_partition): Ditto.
8678 (compute_access_range): Differentiate cases that equality can be
8679 proven at all loops, the innermost loops or no loops.
8680 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8681 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8682 (finalize_partitions, distribute_loop): Don't fuse partition of
8683 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8684 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8687 2018-01-23 Martin Liska <mliska@suse.cz>
8689 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8690 order to ignore the predictor.
8691 (PRED_POLYMORPHIC_CALL): Likewise.
8692 (PRED_RECURSIVE_CALL): Likewise.
8694 2018-01-23 Martin Liska <mliska@suse.cz>
8696 * tree-profile.c (tree_profiling): Print function header to
8697 aware reader which function we are working on.
8698 * value-prof.c (gimple_find_values_to_profile): Do not print
8699 not interesting value histograms.
8701 2018-01-23 Martin Liska <mliska@suse.cz>
8703 * profile-count.h (enum profile_quality): Add
8704 profile_uninitialized as the first value. Do not number values
8705 as they are zero based.
8706 (profile_count::verify): Update sanity check.
8707 (profile_probability::verify): Likewise.
8709 2018-01-23 Nathan Sidwell <nathan@acm.org>
8711 * doc/invoke.texi (ffor-scope): Deprecate.
8713 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8715 PR tree-optimization/83510
8716 * domwalk.c (set_all_edges_as_executable): New function.
8717 (dom_walker::dom_walker): Convert bool param
8718 "skip_unreachable_blocks" to enum reachability. Move setup of
8719 edge flags to set_all_edges_as_executable and only do it when
8720 reachability is REACHABLE_BLOCKS.
8721 * domwalk.h (enum dom_walker::reachability): New enum.
8722 (dom_walker::dom_walker): Convert bool param
8723 "skip_unreachable_blocks" to enum reachability.
8724 (set_all_edges_as_executable): New decl.
8725 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8726 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8728 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8729 but converting true to REACHABLE_BLOCKS.
8730 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8732 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8733 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8734 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8736 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8737 if check_all_array_refs will be called.
8739 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8741 * tree.c (selftest::test_location_wrappers): Add more test
8744 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8746 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8747 (selftest::test_bit_in_range): Likewise.
8749 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8752 * doc/sourcebuild.texi (vect_float): Say that the selector
8753 only describes the situation when -funsafe-math-optimizations is on.
8754 (vect_float_strict): Document.
8756 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8758 PR tree-optimization/83965
8759 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8760 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8761 instead of checking only for a reduction.
8762 (vect_recog_widen_sum_pattern): Likewise.
8764 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8766 * predict.c (probably_never_executed): Only use precise profile info.
8767 (compute_function_frequency): Skip after inlining hack since we now
8768 have quality checking.
8770 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8772 * profile-count.h (profile_probability::very_unlikely,
8773 profile_probability::unlikely, profile_probability::even): Set
8774 precision to guessed.
8776 2018-01-23 Richard Biener <rguenther@suse.de>
8778 PR tree-optimization/83963
8779 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8780 Properly terminate dominator walk when crossing the exit edge not
8781 when visiting its source block.
8783 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8786 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8787 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8789 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8791 PR tree-optimization/83957
8792 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8793 semicolon after for body surrounded by braces.
8795 PR tree-optimization/83081
8796 * profile-count.h (profile_probability::split): New method.
8797 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8798 Use profile_probability::split.
8799 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8800 when splitting a single conditional jump into 2.
8802 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8804 PR tree-optimization/69452
8805 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8808 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8810 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8811 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8812 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8814 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8816 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8817 * config/rl78/rl78.md (movdi): New define_expand.
8818 * config/rl78/rl78.c (rl78_split_movdi): New function.
8820 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8823 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8825 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8826 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8827 128-bit to produce an UNSPEC move to get the double word with the
8828 signbit and then a shift directly to do signbit.
8829 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8830 implementation with a new version that just does either a direct
8831 move or a regular move. Move memory interface to separate insns.
8832 Move insns so they are next to the expander.
8833 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8834 with signbit move. Split big and little endian case.
8835 (signbit<mode>2_dm_mem_le): Likewise.
8836 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8837 (signbit<mode>2_dm2): Likewise.
8839 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8841 * config/rl78/rl78.md (anddi3): New define_expand.
8843 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8845 * config/rl78/rl78.md (umindi3): New define_expand.
8847 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8849 * config/rl78/rl78.md (smindi3): New define_expand.
8851 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8853 * config/rl78/rl78.md (smaxdi3): New define_expand.
8855 2018-01-22 Carl Love <cel@us.ibm.com>
8857 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8858 LVX_V1TI): Add macro expansion.
8859 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8860 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8861 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8862 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8863 Change check to determine if the instruction is a byte reversing
8864 entry. Fix typo in comment.
8865 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8866 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8867 Add def_builtin calls for new builtins.
8868 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8869 Add define_insn expansion.
8871 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8873 * config/rl78/rl78.md (umaxdi3): New define_expand.
8875 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8877 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8878 for non-QImode registers.
8880 2018-01-22 Richard Biener <rguenther@suse.de>
8882 PR tree-optimization/83963
8883 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8884 including the loop exit block.
8885 (scop_detection::merge_sese): Likewise.
8886 (scop_detection::add_scop): Do it here instead.
8888 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8890 * doc/sourcebuild.texi (arm_softfloat): Document.
8892 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8895 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8896 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8897 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8899 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8900 David Edelsohn <dje.gcc@gmail.com>
8903 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8904 Change "crset eq" to "crset 2".
8905 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8906 (*call_indirect_aix<mode>_nospec): Likewise.
8907 (*call_value_indirect_aix<mode>_nospec): Likewise.
8908 (*call_indirect_elfv2<mode>_nospec): Likewise.
8909 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8910 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8911 change assembly output from . to $.
8912 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8913 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8914 (*tablejump<mode>_internal1_nospec): Likewise.
8916 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8919 * config/sh/sh_optimize_sett_clrt.cc:
8920 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8922 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8924 PR tree-optimization/83940
8925 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8926 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8927 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8928 use it to pass back the definition type.
8929 (vect_check_store_rhs): Likewise rhs_dt_out.
8930 (vect_build_gather_load_calls): Add a mask_dt argument and use
8931 it instead of a call to vect_is_simple_use.
8932 (vectorizable_store): Update calls to vect_check_load_store_mask
8933 and vect_check_store_rhs. Use the dt returned by the latter instead
8934 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8935 instead of calls to vect_is_simple_use. Pass the scalar rather
8936 than the vector operand to vect_is_simple_use when handling
8937 second and subsequent copies of an rhs value.
8938 (vectorizable_load): Update calls to vect_check_load_store_mask
8939 and vect_build_gather_load_calls. Use the cached mask_dt and
8940 gs_info.offset_dt instead of calls to vect_is_simple_use.
8942 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8945 * tree-emutls.c: Include gimplify.h.
8946 (lower_emutls_2): New function.
8947 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8948 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8949 it before further processing.
8952 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8953 UINTVAL (trueop1) instead of INTVAL (op1).
8955 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8959 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8960 INCOMING_FRAME_SP_OFFSET if not defined.
8961 (scan_trace): Add ENTRY argument. If true and
8962 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8963 emit a note to adjust the CFA offset.
8964 (create_cfi_notes): Adjust scan_trace callers.
8965 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8966 INCOMING_FRAME_SP_OFFSET in the CIE.
8967 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8968 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8970 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8971 * doc/tm.texi: Regenerated.
8973 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8975 PR rtl-optimization/83147
8976 * lra-constraints.c (remove_inheritance_pseudos): Use
8977 lra_substitute_pseudo_within_insn.
8979 2018-01-19 Tom de Vries <tom@codesourcery.com>
8980 Cesar Philippidis <cesar@codesourcery.com>
8983 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8985 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8988 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8989 spaces for function labels.
8991 2018-01-19 Martin Liska <mliska@suse.cz>
8993 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8994 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8995 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8996 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8997 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8998 (PRED_CONST_RETURN): Change from 69 to 65.
8999 (PRED_NULL_RETURN): Change from 91 to 71.
9000 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9001 (PRED_LOOP_GUARD): Change from 66 to 73.
9003 2018-01-19 Martin Liska <mliska@suse.cz>
9005 * predict.c (predict_insn_def): Add new assert.
9006 (struct branch_predictor): Change type to signed integer.
9007 (test_prediction_value_range): Amend test to cover
9009 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9010 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9011 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9012 (PRED_LOOP_IV_COMPARE): Likewise.
9013 * predict.h (PROB_UNINITIALIZED): Define new constant.
9015 2018-01-19 Martin Liska <mliska@suse.cz>
9017 * predict.c (dump_prediction): Add new format for
9018 analyze_brprob.py script which is enabled with -details
9020 * profile-count.h (precise_p): New function.
9022 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9024 PR tree-optimization/83922
9025 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9026 there are no statements that need masking.
9027 (vect_active_double_reduction_p): New function.
9028 (vect_analyze_loop_operations): Use it when handling phis that
9029 are not in the loop header.
9031 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9033 PR tree-optimization/83914
9034 * tree-vect-loop.c (vectorizable_induction): Don't convert
9035 init_expr or apply the peeling adjustment for inductions
9036 that are nested within the vectorized loop.
9038 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9040 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9043 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9047 * function.h (gimplify_parameters): Add gimple_seq * argument.
9048 * function.c: Include gimple.h and options.h.
9049 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9050 for the added local temporaries if needed.
9051 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9052 if there are any parameter cleanups, wrap whole body into a
9053 try/finally with the cleanups.
9055 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9058 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9059 Use GET_MODE_CLASS for scalar floating point.
9061 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9065 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9066 Fix call of call_cgraph_insertion_hooks.
9068 2018-01-18 Martin Sebor <msebor@redhat.com>
9070 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9072 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9075 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9078 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9081 * common.opt: (-ffile-prefix-map): New option.
9082 * opts.c (common_handle_option): Defer it.
9083 * opts-global.c (handle_common_deferred_options): Handle it.
9084 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9085 * file-prefix-map.h: New file.
9086 (remap_debug_filename, add_debug_prefix_map): ...here.
9087 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9088 * final.c (debug_prefix_map, add_debug_prefix_map
9089 remap_debug_filename): Move to...
9090 * file-prefix-map.c: New file.
9091 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9092 generalize, get rid of alloca(), use strrchr() instead of strchr().
9093 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9094 Implement in terms of add_prefix_map().
9095 (remap_macro_filename, remap_debug_filename): Implement in term of
9097 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9098 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9099 * dbxout.c: Include file-prefix-map.h.
9100 * varasm.c: Likewise.
9101 * vmsdbgout.c: Likewise.
9102 * xcoffout.c: Likewise.
9103 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9104 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9105 * doc/invoke.texi (-ffile-prefix-map): Document.
9106 (-fdebug-prefix-map): Update description.
9108 2018-01-18 Martin Liska <mliska@suse.cz>
9110 * config/i386/i386.c (indirect_thunk_name): Document that also
9112 (output_indirect_thunk): Document why both instructions
9113 (pause and lfence) are generated.
9115 2018-01-18 Richard Biener <rguenther@suse.de>
9117 PR tree-optimization/83887
9118 * graphite-scop-detection.c
9119 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9120 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9121 (scop_detection::merge_sese): Re-implement with a flood-fill
9122 algorithm that properly finds a SESE region if it exists.
9124 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9127 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9128 pointer_diff optimizations use view_convert instead of convert.
9130 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9132 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9133 Generate different code for -mno-speculate-indirect-jumps.
9134 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9135 (*call_indirect_aix<mode>): Disable for
9136 -mno-speculate-indirect-jumps.
9137 (*call_indirect_aix<mode>_nospec): New define_insn.
9138 (*call_value_indirect_aix<mode>): Disable for
9139 -mno-speculate-indirect-jumps.
9140 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9141 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9142 -mno-speculate-indirect-jumps.
9143 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9145 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9147 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9148 long double type, set the flags for noting the default long double
9149 type, even if we don't pass or return a long double type.
9151 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9154 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9157 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9159 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9160 support for merge[hl].
9161 (fold_mergehl_helper): New helper function.
9162 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9163 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9164 (altivec_vmrglw_direct): Add xxmrglw insn.
9166 2018-01-17 Andrew Waterman <andrew@sifive.com>
9168 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9169 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9171 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9174 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9175 call the lto_location_cache before reading the
9176 DECL_SOURCE_LOCATION of the types.
9178 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9179 Richard Sandiford <richard.sandiford@linaro.org>
9181 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9182 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9183 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9184 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9185 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9187 * config/aarch64/constraints.md (aarch64_movti_operand):
9189 * config/aarch64/predicates.md (Uti): Add new constraint.
9191 2018-01-17 Carl Love <cel@us.ibm.com>
9193 * config/rs6000/vsx.md (define_expand xl_len_r,
9194 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9195 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9197 (define_expand, define_insn): Move the shift left from the
9198 define_insn to the define_expand for lxvl and stxvl instructions.
9199 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9200 and XL_LEN_R definitions to PURE.
9202 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9204 * config/i386/i386.c (indirect_thunk_name): Declare regno
9205 as unsigned int. Compare regno with INVALID_REGNUM.
9206 (output_indirect_thunk): Ditto.
9207 (output_indirect_thunk_function): Ditto.
9208 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9209 in the call to output_indirect_thunk_function.
9211 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9214 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9215 rather than the size of inner_type to determine the stack slot size
9216 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9218 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9221 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9224 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9226 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9227 endian Linux systems to optionally enable multilibs for selecting
9228 the long double type if the user configured an explicit type.
9229 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9230 have no long double multilibs if not defined.
9231 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9232 warn if the user used -mabi={ieee,ibm}longdouble and we built
9233 multilibs for long double.
9234 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9235 appropriate multilib option.
9236 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9238 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9239 for building long double multilibs.
9240 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9242 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9244 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9247 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9249 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9252 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9255 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9258 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9260 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9261 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9263 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9265 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9266 different rtl trees depending on TARGET_64BIT.
9267 (rs6000_gen_lvx): Likewise.
9269 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9271 * config/visium/visium.md (nop): Tweak comment.
9272 (hazard_nop): Likewise.
9274 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9276 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9277 -mspeculate-indirect-jumps.
9278 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9279 for -mno-speculate-indirect-jumps.
9280 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9281 (*call_value_indirect_elfv2<mode>): Disable for
9282 -mno-speculate-indirect-jumps.
9283 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9284 (indirect_jump): Emit different RTL for
9285 -mno-speculate-indirect-jumps.
9286 (*indirect_jump<mode>): Disable for
9287 -mno-speculate-indirect-jumps.
9288 (*indirect_jump<mode>_nospec): New define_insn.
9289 (tablejump): Emit different RTL for
9290 -mno-speculate-indirect-jumps.
9291 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9292 (tablejumpsi_nospec): New define_expand.
9293 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9294 (tablejumpdi_nospec): New define_expand.
9295 (*tablejump<mode>_internal1): Disable for
9296 -mno-speculate-indirect-jumps.
9297 (*tablejump<mode>_internal1_nospec): New define_insn.
9298 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9301 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9303 * caller-save.c (insert_save): Drop unnecessary parameter. All
9306 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9307 Richard Biener <rguenth@suse.de>
9310 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9311 return early, inline manually is_gimple_sizepos. Make sure if we
9312 call gimplify_expr we don't end up with a gimple constant.
9313 * tree.c (variably_modified_type_p): Don't return true for
9314 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9315 * gimplify.h (is_gimple_sizepos): Remove.
9317 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9319 PR tree-optimization/83857
9320 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9321 vectorizable_live_operation for pure SLP statements.
9322 (vectorizable_live_operation): Handle PHIs.
9324 2018-01-16 Richard Biener <rguenther@suse.de>
9326 PR tree-optimization/83867
9327 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9328 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9330 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9333 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9334 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9335 If off is not INTEGER_CST, issue a may not be aligned warning
9336 rather than isn't aligned. Use isn%'t rather than isn't.
9337 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9339 <case MULT_EXPR>: Improve the case when bottom and one of the
9340 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9341 operand, in that case check if the other operand is multiple of
9342 bottom divided by the INTEGER_CST operand.
9344 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9347 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9348 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9349 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9350 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9351 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9352 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9353 (pa_function_arg_size): New function.
9355 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9357 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9358 in a separate statement.
9360 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9362 PR tree-optimization/83847
9363 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9364 group gathers and scatters.
9366 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9368 PR rtl-optimization/86620
9369 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9371 PR rtl-optimization/83213
9372 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9373 to last if both are JUMP_INSNs.
9375 PR tree-optimization/83843
9376 * gimple-ssa-store-merging.c
9377 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9378 store_immediate_info for bswap/nop orig_stores.
9380 2018-01-15 Andrew Waterman <andrew@sifive.com>
9382 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9384 <UDIV>: Increase cost if !TARGET_DIV.
9386 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9388 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9389 (define_attr "cr_logical_3op"): New.
9390 (cceq_ior_compare): Adjust.
9391 (cceq_ior_compare_complement): Adjust.
9392 (*cceq_rev_compare): Adjust.
9393 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9394 (is_cracked_insn): Adjust.
9395 (insn_must_be_first_in_group): Adjust.
9396 * config/rs6000/40x.md: Adjust.
9397 * config/rs6000/440.md: Adjust.
9398 * config/rs6000/476.md: Adjust.
9399 * config/rs6000/601.md: Adjust.
9400 * config/rs6000/603.md: Adjust.
9401 * config/rs6000/6xx.md: Adjust.
9402 * config/rs6000/7450.md: Adjust.
9403 * config/rs6000/7xx.md: Adjust.
9404 * config/rs6000/8540.md: Adjust.
9405 * config/rs6000/cell.md: Adjust.
9406 * config/rs6000/e300c2c3.md: Adjust.
9407 * config/rs6000/e500mc.md: Adjust.
9408 * config/rs6000/e500mc64.md: Adjust.
9409 * config/rs6000/e5500.md: Adjust.
9410 * config/rs6000/e6500.md: Adjust.
9411 * config/rs6000/mpc.md: Adjust.
9412 * config/rs6000/power4.md: Adjust.
9413 * config/rs6000/power5.md: Adjust.
9414 * config/rs6000/power6.md: Adjust.
9415 * config/rs6000/power7.md: Adjust.
9416 * config/rs6000/power8.md: Adjust.
9417 * config/rs6000/power9.md: Adjust.
9418 * config/rs6000/rs64.md: Adjust.
9419 * config/rs6000/titan.md: Adjust.
9421 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9423 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9424 ix86_indirect_branch_register logic.
9426 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9428 * config/i386/constraints.md (Bs): Update
9429 ix86_indirect_branch_register check. Don't check
9430 ix86_indirect_branch_register with GOT_memory_operand.
9432 * config/i386/predicates.md (GOT_memory_operand): Don't check
9433 ix86_indirect_branch_register here.
9434 (GOT32_symbol_operand): Likewise.
9436 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9438 * config/i386/predicates.md (constant_call_address_operand):
9439 Rewrite ix86_indirect_branch_register logic.
9440 (sibcall_insn_operand): Likewise.
9442 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9444 * config/i386/constraints.md (Bs): Replace
9445 ix86_indirect_branch_thunk_register with
9446 ix86_indirect_branch_register.
9448 * config/i386/i386.md (indirect_jump): Likewise.
9449 (tablejump): Likewise.
9450 (*sibcall_memory): Likewise.
9451 (*sibcall_value_memory): Likewise.
9452 Peepholes of indirect call and jump via memory: Likewise.
9453 * config/i386/i386.opt: Likewise.
9454 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9455 (GOT_memory_operand): Likewise.
9456 (call_insn_operand): Likewise.
9457 (sibcall_insn_operand): Likewise.
9458 (GOT32_symbol_operand): Likewise.
9460 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9463 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9464 type rather than type addr's type points to.
9465 (expand_omp_atomic_mutex): Likewise.
9466 (expand_omp_atomic): Likewise.
9468 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9471 * config/i386/i386.c (output_indirect_thunk_function): Use
9472 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9473 for __x86_return_thunk.
9475 2018-01-15 Richard Biener <rguenther@suse.de>
9478 * expmed.c (extract_bit_field_1): Fix typo.
9480 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9483 * config/arm/iterators.md (VF): New mode iterator.
9484 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9485 Remove integer-related logic from pattern.
9486 (neon_vabd<mode>_3): Likewise.
9488 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9491 * common.opt (fstrict-overflow): No longer an alias.
9492 (fwrapv-pointer): New option.
9493 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9494 also for pointer types based on flag_wrapv_pointer.
9495 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9496 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9497 opts->x_flag_wrapv got set.
9498 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9499 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9500 POINTER_TYPE_OVERFLOW_UNDEFINED.
9501 * match.pd: Likewise in address comparison pattern.
9502 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9504 2018-01-15 Richard Biener <rguenther@suse.de>
9507 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9508 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9509 Reset type names to their identifier if their TYPE_DECL doesn't
9510 have linkage (and thus is used for ODR and devirt).
9511 (save_debug_info_for_decl): Remove.
9512 (save_debug_info_for_type): Likewise.
9513 (add_tree_to_fld_list): Adjust.
9514 * tree-pretty-print.c (dump_generic_node): Make dumping of
9515 type names more robust.
9517 2018-01-15 Richard Biener <rguenther@suse.de>
9519 * BASE-VER: Bump to 8.0.1.
9521 2018-01-14 Martin Sebor <msebor@redhat.com>
9524 * builtins.c (check_access): Avoid warning when the no-warning bit
9527 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9529 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9530 * ira-color (allocno_hard_regs_compare): Likewise.
9532 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9535 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9536 Use .pushsection/.popsection.
9538 2018-01-14 Martin Sebor <msebor@redhat.com>
9541 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9543 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9545 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9546 entry from extra_headers.
9547 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9548 extra_headers, make the list bitwise identical to the i?86-*-* one.
9550 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9552 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9553 -mcmodel=large with -mindirect-branch=thunk,
9554 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9555 -mfunction-return=thunk-extern.
9556 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9557 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9558 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9560 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9562 * config/i386/i386.c (print_reg): Print the name of the full
9563 integer register without '%'.
9564 (ix86_print_operand): Handle 'V'.
9565 * doc/extend.texi: Document 'V' modifier.
9567 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9569 * config/i386/constraints.md (Bs): Disallow memory operand for
9570 -mindirect-branch-register.
9572 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9573 (GOT_memory_operand): Likewise.
9574 (call_insn_operand): Likewise.
9575 (sibcall_insn_operand): Likewise.
9576 (GOT32_symbol_operand): Likewise.
9577 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9578 for -mindirect-branch-register.
9579 (tablejump): Likewise.
9580 (*sibcall_memory): Likewise.
9581 (*sibcall_value_memory): Likewise.
9582 Disallow peepholes of indirect call and jump via memory for
9583 -mindirect-branch-register.
9584 (*call_pop): Replace m with Bw.
9585 (*call_value_pop): Likewise.
9586 (*sibcall_pop_memory): Replace m with Bs.
9587 * config/i386/i386.opt (mindirect-branch-register): New option.
9588 * doc/invoke.texi: Document -mindirect-branch-register option.
9590 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9592 * config/i386/i386-protos.h (ix86_output_function_return): New.
9593 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9594 set function_return_type.
9595 (indirect_thunk_name): Add ret_p to indicate thunk for function
9597 (output_indirect_thunk_function): Pass false to
9598 indirect_thunk_name.
9599 (ix86_output_indirect_branch_via_reg): Likewise.
9600 (ix86_output_indirect_branch_via_push): Likewise.
9601 (output_indirect_thunk_function): Create alias for function
9602 return thunk if regno < 0.
9603 (ix86_output_function_return): New function.
9604 (ix86_handle_fndecl_attribute): Handle function_return.
9605 (ix86_attribute_table): Add function_return.
9606 * config/i386/i386.h (machine_function): Add
9607 function_return_type.
9608 * config/i386/i386.md (simple_return_internal): Use
9609 ix86_output_function_return.
9610 (simple_return_internal_long): Likewise.
9611 * config/i386/i386.opt (mfunction-return=): New option.
9612 (indirect_branch): Mention -mfunction-return=.
9613 * doc/extend.texi: Document function_return function attribute.
9614 * doc/invoke.texi: Document -mfunction-return= option.
9616 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9618 * config/i386/i386-opts.h (indirect_branch): New.
9619 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9620 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9621 with local indirect jump when converting indirect call and jump.
9622 (ix86_set_indirect_branch_type): New.
9623 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9624 (indirectlabelno): New.
9625 (indirect_thunk_needed): Likewise.
9626 (indirect_thunk_bnd_needed): Likewise.
9627 (indirect_thunks_used): Likewise.
9628 (indirect_thunks_bnd_used): Likewise.
9629 (INDIRECT_LABEL): Likewise.
9630 (indirect_thunk_name): Likewise.
9631 (output_indirect_thunk): Likewise.
9632 (output_indirect_thunk_function): Likewise.
9633 (ix86_output_indirect_branch_via_reg): Likewise.
9634 (ix86_output_indirect_branch_via_push): Likewise.
9635 (ix86_output_indirect_branch): Likewise.
9636 (ix86_output_indirect_jmp): Likewise.
9637 (ix86_code_end): Call output_indirect_thunk_function if needed.
9638 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9640 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9641 (ix86_attribute_table): Add indirect_branch.
9642 * config/i386/i386.h (machine_function): Add indirect_branch_type
9643 and has_local_indirect_jump.
9644 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9646 (tablejump): Likewise.
9647 (*indirect_jump): Use ix86_output_indirect_jmp.
9648 (*tablejump_1): Likewise.
9649 (simple_return_indirect_internal): Likewise.
9650 * config/i386/i386.opt (mindirect-branch=): New option.
9651 (indirect_branch): New.
9654 (thunk-inline): Likewise.
9655 (thunk-extern): Likewise.
9656 * doc/extend.texi: Document indirect_branch function attribute.
9657 * doc/invoke.texi: Document -mindirect-branch= option.
9659 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9662 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9664 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9666 * ipa-inline.c (want_inline_small_function_p): Return false if
9667 inlining has already failed with CIF_FINAL_ERROR.
9668 (update_caller_keys): Call want_inline_small_function_p before
9670 (update_callee_keys): Likewise.
9672 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9674 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9676 (rs6000_quadword_masked_address_p): Likewise.
9677 (quad_aligned_load_p): Likewise.
9678 (quad_aligned_store_p): Likewise.
9679 (const_load_sequence_p): Add comment to describe the outer-most loop.
9680 (mimic_memory_attributes_and_flags): New function.
9681 (rs6000_gen_stvx): Likewise.
9682 (replace_swapped_aligned_store): Likewise.
9683 (rs6000_gen_lvx): Likewise.
9684 (replace_swapped_aligned_load): Likewise.
9685 (replace_swapped_load_constant): Capitalize argument name in
9686 comment describing this function.
9687 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9688 and stores that access quad-word aligned addresses and replace
9689 with stvx or lvx instructions when appropriate.
9690 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9691 New function prototype.
9692 (rs6000_quadword_masked_address_p): Likewise.
9693 (rs6000_gen_lvx): Likewise.
9694 (rs6000_gen_stvx): Likewise.
9695 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9696 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9697 when memory address is aligned.
9698 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9699 this split to select lvx instruction when memory address is aligned.
9700 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9701 instruction when memory address is aligned.
9702 (*vsx_le_perm_load_v16qi): Likewise.
9703 (four unnamed splitters): Modify to select the stvx instruction
9704 when memory is aligned.
9706 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9708 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9709 which appears in the queue multiple times.
9711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9712 Alan Hayward <alan.hayward@arm.com>
9713 David Sherwood <david.sherwood@arm.com>
9715 * tree-vectorizer.h (vec_lower_bound): New structure.
9716 (_loop_vec_info): Add check_nonzero and lower_bounds.
9717 (LOOP_VINFO_CHECK_NONZERO): New macro.
9718 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9719 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9720 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9721 fields. Make seg_len the distance travelled, not including the
9723 (dr_direction_indicator): Declare.
9724 (dr_zero_step_indicator): Likewise.
9725 (dr_known_forward_stride_p): Likewise.
9726 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9728 (runtime_alias_check_p): Allow runtime alias checks with
9730 (operator ==): Compare access_size and align.
9731 (prune_runtime_alias_test_list): Rework for new distinction between
9732 the access_size and seg_len.
9733 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9735 (get_segment_min_max): New function.
9736 (create_intersect_range_checks): Use it.
9737 (dr_step_indicator): New function.
9738 (dr_direction_indicator): Likewise.
9739 (dr_zero_step_indicator): Likewise.
9740 (dr_known_forward_stride_p): Likewise.
9741 * tree-loop-distribution.c (data_ref_segment_size): Return
9742 DR_STEP * (niters - 1).
9743 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9745 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9746 (vect_preserves_scalar_order_p): New function, split out from...
9747 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9748 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9749 (vect_vfa_access_size): New function.
9750 (vect_vfa_align): Likewise.
9751 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9752 (dump_lower_bound): New function.
9753 (vect_check_lower_bound): Likewise.
9754 (vect_small_gap_p): Likewise.
9755 (vectorizable_with_step_bound_p): Likewise.
9756 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9757 depencies if the vectorization factor is 1. Convert the checks
9758 for nonzero steps into checks on the bounds of DR_STEP. Try using
9759 a bunds check for variable steps if the minimum required step is
9760 relatively small. Update calls to the dr_with_seg_len
9761 constructor and to vect_compile_time_alias.
9762 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9764 (vect_loop_versioning): Call it.
9765 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9767 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9770 Alan Hayward <alan.hayward@arm.com>
9771 David Sherwood <david.sherwood@arm.com>
9773 * doc/sourcebuild.texi (vect_scatter_store): Document.
9774 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9776 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9778 * genopinit.c (main): Add supports_vec_scatter_store and
9779 supports_vec_scatter_store_cached to target_optabs.
9780 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9781 IFN_MASK_SCATTER_STORE.
9782 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9784 * internal-fn.h (internal_store_fn_p): Declare.
9785 (internal_fn_stored_value_index): Likewise.
9786 * internal-fn.c (scatter_store_direct): New macro.
9787 (expand_scatter_store_optab_fn): New function.
9788 (direct_scatter_store_optab_supported_p): New macro.
9789 (internal_store_fn_p): New function.
9790 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9791 IFN_MASK_SCATTER_STORE.
9792 (internal_fn_mask_index): Likewise.
9793 (internal_fn_stored_value_index): New function.
9794 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9796 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9797 * optabs-query.c (supports_vec_scatter_store_p): New function.
9798 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9799 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9800 true for scatter stores.
9801 (vect_gather_scatter_fn_p): Handle scatter stores too.
9802 (vect_check_gather_scatter): Consider using scatter stores if
9803 supports_vec_scatter_store_p.
9804 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9806 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9807 internal_fn_stored_value_index.
9808 (check_load_store_masking): Handle scatter stores too.
9809 (vect_get_store_rhs): Make public.
9810 (vectorizable_call): Use internal_store_fn_p.
9811 (vectorizable_store): Handle scatter store internal functions.
9812 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9813 when deciding whether the end of the group has been reached.
9814 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9815 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9816 (mask_scatter_store<mode>): New insns.
9818 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9819 Alan Hayward <alan.hayward@arm.com>
9820 David Sherwood <david.sherwood@arm.com>
9822 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9823 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9824 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9826 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9827 Use vect_truncate_gather_scatter_offset if we can't treat the
9828 operation as a normal gather load or scatter store.
9829 (get_group_load_store_type): Take the gather_scatter_info
9830 as argument. Try using a gather load or scatter store for
9831 single-element groups.
9832 (get_load_store_type): Update calls to get_group_load_store_type
9833 and vect_use_strided_gather_scatters_p.
9835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9836 Alan Hayward <alan.hayward@arm.com>
9837 David Sherwood <david.sherwood@arm.com>
9839 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9840 optional tree argument.
9841 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9843 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9844 but continue to use the current value as a fallback.
9845 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9846 to compare the updates.
9847 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9848 (get_load_store_type): Use it when handling a strided access.
9849 (vect_get_strided_load_store_ops): New function.
9850 (vect_get_data_ptr_increment): Likewise.
9851 (vectorizable_load): Handle strided gather loads. Always pass
9852 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9854 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9855 Alan Hayward <alan.hayward@arm.com>
9856 David Sherwood <david.sherwood@arm.com>
9858 * doc/md.texi (gather_load@var{m}): Document.
9859 (mask_gather_load@var{m}): Likewise.
9860 * genopinit.c (main): Add supports_vec_gather_load and
9861 supports_vec_gather_load_cached to target_optabs.
9862 * optabs-tree.c (init_tree_optimization_optabs): Use
9863 ggc_cleared_alloc to allocate target_optabs.
9864 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9865 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9867 * internal-fn.h (internal_load_fn_p): Declare.
9868 (internal_gather_scatter_fn_p): Likewise.
9869 (internal_fn_mask_index): Likewise.
9870 (internal_gather_scatter_fn_supported_p): Likewise.
9871 * internal-fn.c (gather_load_direct): New macro.
9872 (expand_gather_load_optab_fn): New function.
9873 (direct_gather_load_optab_supported_p): New macro.
9874 (direct_internal_fn_optab): New function.
9875 (internal_load_fn_p): Likewise.
9876 (internal_gather_scatter_fn_p): Likewise.
9877 (internal_fn_mask_index): Likewise.
9878 (internal_gather_scatter_fn_supported_p): Likewise.
9879 * optabs-query.c (supports_at_least_one_mode_p): New function.
9880 (supports_vec_gather_load_p): Likewise.
9881 * optabs-query.h (supports_vec_gather_load_p): Declare.
9882 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9883 and memory_type field.
9884 (NUM_PATTERNS): Bump to 15.
9885 * tree-vect-data-refs.c: Include internal-fn.h.
9886 (vect_gather_scatter_fn_p): New function.
9887 (vect_describe_gather_scatter_call): Likewise.
9888 (vect_check_gather_scatter): Try using internal functions for
9889 gather loads. Recognize existing calls to a gather load function.
9890 (vect_analyze_data_refs): Consider using gather loads if
9891 supports_vec_gather_load_p.
9892 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9893 (vect_get_gather_scatter_offset_type): Likewise.
9894 (vect_convert_mask_for_vectype): Likewise.
9895 (vect_add_conversion_to_patterm): Likewise.
9896 (vect_try_gather_scatter_pattern): Likewise.
9897 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9898 (vect_vect_recog_func_ptrs): Add it.
9899 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9900 internal_fn_mask_index and internal_gather_scatter_fn_p.
9901 (check_load_store_masking): Take the gather_scatter_info as an
9902 argument and handle gather loads.
9903 (vect_get_gather_scatter_ops): New function.
9904 (vectorizable_call): Check internal_load_fn_p.
9905 (vectorizable_load): Likewise. Handle gather load internal
9907 (vectorizable_store): Update call to check_load_store_masking.
9908 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9909 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9910 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9911 (aarch64_gather_scale_operand_d): New predicates.
9912 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9913 (mask_gather_load<mode>): New insns.
9915 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9916 Alan Hayward <alan.hayward@arm.com>
9917 David Sherwood <david.sherwood@arm.com>
9919 * optabs.def (fold_left_plus_optab): New optab.
9920 * doc/md.texi (fold_left_plus_@var{m}): Document.
9921 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9922 * internal-fn.c (fold_left_direct): Define.
9923 (expand_fold_left_optab_fn): Likewise.
9924 (direct_fold_left_optab_supported_p): Likewise.
9925 * fold-const-call.c (fold_const_fold_left): New function.
9926 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9927 * tree-parloops.c (valid_reduction_p): New function.
9928 (gather_scalar_reductions): Use it.
9929 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9930 (vect_finish_replace_stmt): Declare.
9931 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9932 (needs_fold_left_reduction_p): New function, split out from...
9933 (vect_is_simple_reduction): ...here. Accept reductions that
9934 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9935 (vect_force_simple_reduction): Also store the reduction type in
9936 the assignment's STMT_VINFO_REDUC_TYPE.
9937 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9938 (merge_with_identity): New function.
9939 (vect_expand_fold_left): Likewise.
9940 (vectorize_fold_left_reduction): Likewise.
9941 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9942 scalar phi in place for it. Check for target support and reject
9943 cases that would reassociate the operation. Defer the transform
9944 phase to vectorize_fold_left_reduction.
9945 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9946 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9947 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9949 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9951 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9952 call to ifc_temp_var.
9954 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9955 Alan Hayward <alan.hayward@arm.com>
9956 David Sherwood <david.sherwood@arm.com>
9958 * target.def (legitimize_address_displacement): Take the original
9959 offset as a poly_int.
9960 * targhooks.h (default_legitimize_address_displacement): Update
9962 * targhooks.c (default_legitimize_address_displacement): Likewise.
9963 * doc/tm.texi: Regenerate.
9964 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9965 as an argument, moving assert of ad->disp == ad->disp_term to...
9966 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9967 Try calling targetm.legitimize_address_displacement before expanding
9968 the address rather than afterwards, and adjust for the new interface.
9969 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9970 Match the new hook interface. Handle SVE addresses.
9971 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9974 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9976 * Makefile.in (OBJS): Add early-remat.o.
9977 * target.def (select_early_remat_modes): New hook.
9978 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9979 * doc/tm.texi: Regenerate.
9980 * targhooks.h (default_select_early_remat_modes): Declare.
9981 * targhooks.c (default_select_early_remat_modes): New function.
9982 * timevar.def (TV_EARLY_REMAT): New timevar.
9983 * passes.def (pass_early_remat): New pass.
9984 * tree-pass.h (make_pass_early_remat): Declare.
9985 * early-remat.c: New file.
9986 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9988 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9990 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9991 Alan Hayward <alan.hayward@arm.com>
9992 David Sherwood <david.sherwood@arm.com>
9994 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9995 vfm1 with a bound_epilog parameter.
9996 (vect_do_peeling): Update calls accordingly, and move the prologue
9997 call earlier in the function. Treat the base bound_epilog as 0 for
9998 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9999 this base when peeling for gaps.
10000 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10001 with fully-masked loops.
10002 (vect_estimate_min_profitable_iters): Handle the single peeled
10003 iteration in that case.
10005 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10006 Alan Hayward <alan.hayward@arm.com>
10007 David Sherwood <david.sherwood@arm.com>
10009 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10010 single-element interleaving even if the size is not a power of 2.
10011 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10012 accesses for single-element interleaving if the group size is
10015 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10016 Alan Hayward <alan.hayward@arm.com>
10017 David Sherwood <david.sherwood@arm.com>
10019 * doc/md.texi (fold_extract_last_@var{m}): Document.
10020 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10021 * optabs.def (fold_extract_last_optab): New optab.
10022 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10023 * internal-fn.c (fold_extract_direct): New macro.
10024 (expand_fold_extract_optab_fn): Likewise.
10025 (direct_fold_extract_optab_supported_p): Likewise.
10026 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10027 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10028 EXTRACT_LAST_REDUCTION.
10029 (get_initial_def_for_reduction): Do not create an initial vector
10030 for EXTRACT_LAST_REDUCTION reductions.
10031 (vectorizable_reduction): Leave the scalar phi in place for
10032 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10033 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10034 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10035 transform phase to vectorizable_condition.
10036 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10038 (vect_finish_stmt_generation): ...here.
10039 (vect_finish_replace_stmt): New function.
10040 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10041 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10043 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10046 Alan Hayward <alan.hayward@arm.com>
10047 David Sherwood <david.sherwood@arm.com>
10049 * doc/md.texi (extract_last_@var{m}): Document.
10050 * optabs.def (extract_last_optab): New optab.
10051 * internal-fn.def (EXTRACT_LAST): New internal function.
10052 * internal-fn.c (cond_unary_direct): New macro.
10053 (expand_cond_unary_optab_fn): Likewise.
10054 (direct_cond_unary_optab_supported_p): Likewise.
10055 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10056 loops using EXTRACT_LAST.
10057 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10058 (extract_last_<mode>): ...this optab.
10059 (vec_extract<mode><Vel>): Update accordingly.
10061 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10062 Alan Hayward <alan.hayward@arm.com>
10063 David Sherwood <david.sherwood@arm.com>
10065 * target.def (empty_mask_is_expensive): New hook.
10066 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10067 * doc/tm.texi: Regenerate.
10068 * targhooks.h (default_empty_mask_is_expensive): Declare.
10069 * targhooks.c (default_empty_mask_is_expensive): New function.
10070 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10071 if the target says that empty masks are expensive.
10072 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10074 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10076 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10077 Alan Hayward <alan.hayward@arm.com>
10078 David Sherwood <david.sherwood@arm.com>
10080 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10081 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10082 (vect_use_loop_mask_for_alignment_p): New function.
10083 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10084 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10085 niters_skip argument. Make sure that the first niters_skip elements
10086 of the first iteration are inactive.
10087 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10088 Update call to vect_set_loop_masks_directly.
10089 (get_misalign_in_elems): New function, split out from...
10090 (vect_gen_prolog_loop_niters): ...here.
10091 (vect_update_init_of_dr): Take a code argument that specifies whether
10092 the adjustment should be added or subtracted.
10093 (vect_update_init_of_drs): Likewise.
10094 (vect_prepare_for_masked_peels): New function.
10095 (vect_do_peeling): Skip prologue peeling if we're using a mask
10096 instead. Update call to vect_update_inits_of_drs.
10097 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10099 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10100 alignment. Do not include the number of peeled iterations in
10101 the minimum threshold in that case.
10102 (vectorizable_induction): Adjust the start value down by
10103 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10104 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10105 Take the number of skipped iterations into account when calculating
10107 * tree-vect-stmts.c (vect_gen_while_not): New function.
10109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10110 Alan Hayward <alan.hayward@arm.com>
10111 David Sherwood <david.sherwood@arm.com>
10113 * doc/sourcebuild.texi (vect_fully_masked): Document.
10114 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10115 default value to 0.
10116 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10118 (vect_analyze_loop_2): ...here. Don't check the vectorization
10119 factor against the number of loop iterations if the loop is
10122 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10123 Alan Hayward <alan.hayward@arm.com>
10124 David Sherwood <david.sherwood@arm.com>
10126 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10127 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10128 (dump_groups): Update accordingly.
10129 (iv_use::mem_type): New member variable.
10130 (address_p): New function.
10131 (record_use): Add a mem_type argument and initialize the new
10133 (record_group_use): Add a mem_type argument. Use address_p.
10134 Remove obsolete null checks of base_object. Update call to record_use.
10135 (find_interesting_uses_op): Update call to record_group_use.
10136 (find_interesting_uses_cond): Likewise.
10137 (find_interesting_uses_address): Likewise.
10138 (get_mem_type_for_internal_fn): New function.
10139 (find_address_like_use): Likewise.
10140 (find_interesting_uses_stmt): Try find_address_like_use before
10141 calling find_interesting_uses_op.
10142 (addr_offset_valid_p): Use the iv mem_type field as the type
10143 of the addressed memory.
10144 (add_autoinc_candidates): Likewise.
10145 (get_address_cost): Likewise.
10146 (split_small_address_groups_p): Use address_p.
10147 (split_address_groups): Likewise.
10148 (add_iv_candidate_for_use): Likewise.
10149 (autoinc_possible_for_pair): Likewise.
10150 (rewrite_groups): Likewise.
10151 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10152 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10153 (get_alias_ptr_type_for_ptr_address): New function.
10154 (rewrite_use_address): Rewrite address uses in calls that were
10155 identified by find_address_like_use.
10157 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10158 Alan Hayward <alan.hayward@arm.com>
10159 David Sherwood <david.sherwood@arm.com>
10161 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10163 * gimple-expr.h (is_gimple_addressable: Likewise.
10164 * gimple-expr.c (is_gimple_address): Likewise.
10165 * internal-fn.c (expand_call_mem_ref): New function.
10166 (expand_mask_load_optab_fn): Use it.
10167 (expand_mask_store_optab_fn): Likewise.
10169 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10170 Alan Hayward <alan.hayward@arm.com>
10171 David Sherwood <david.sherwood@arm.com>
10173 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10174 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10175 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10176 (cond_umax@var{mode}): Document.
10177 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10178 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10179 (cond_umin_optab, cond_umax_optab): New optabs.
10180 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10181 (COND_IOR, COND_XOR): New internal functions.
10182 * internal-fn.h (get_conditional_internal_fn): Declare.
10183 * internal-fn.c (cond_binary_direct): New macro.
10184 (expand_cond_binary_optab_fn): Likewise.
10185 (direct_cond_binary_optab_supported_p): Likewise.
10186 (get_conditional_internal_fn): New function.
10187 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10188 Cope with reduction statements that are vectorized as calls rather
10190 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10191 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10192 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10193 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10194 (UNSPEC_COND_EOR): New unspecs.
10195 (optab): Add mappings for them.
10196 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10197 (sve_int_op, sve_fp_op): New int attributes.
10199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10200 Alan Hayward <alan.hayward@arm.com>
10201 David Sherwood <david.sherwood@arm.com>
10203 * optabs.def (while_ult_optab): New optab.
10204 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10205 * internal-fn.def (WHILE_ULT): New internal function.
10206 * internal-fn.h (direct_internal_fn_supported_p): New override
10207 that takes two types as argument.
10208 * internal-fn.c (while_direct): New macro.
10209 (expand_while_optab_fn): New function.
10210 (convert_optab_supported_p): Likewise.
10211 (direct_while_optab_supported_p): New macro.
10212 * wide-int.h (wi::udiv_ceil): New function.
10213 * tree-vectorizer.h (rgroup_masks): New structure.
10214 (vec_loop_masks): New typedef.
10215 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10216 and fully_masked_p.
10217 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10218 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10219 (vect_max_vf): New function.
10220 (slpeel_make_loop_iterate_ntimes): Delete.
10221 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10222 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10223 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10224 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10225 internal-fn.h, stor-layout.h and optabs-query.h.
10226 (vect_set_loop_mask): New function.
10227 (add_preheader_seq): Likewise.
10228 (add_header_seq): Likewise.
10229 (interleave_supported_p): Likewise.
10230 (vect_maybe_permute_loop_masks): Likewise.
10231 (vect_set_loop_masks_directly): Likewise.
10232 (vect_set_loop_condition_masked): Likewise.
10233 (vect_set_loop_condition_unmasked): New function, split out from
10234 slpeel_make_loop_iterate_ntimes.
10235 (slpeel_make_loop_iterate_ntimes): Rename to..
10236 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10237 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10238 (vect_do_peeling): Update call accordingly.
10239 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10241 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10242 mask_compare_type, can_fully_mask_p and fully_masked_p.
10243 (release_vec_loop_masks): New function.
10244 (_loop_vec_info): Use it to free the loop masks.
10245 (can_produce_all_loop_masks_p): New function.
10246 (vect_get_max_nscalars_per_iter): Likewise.
10247 (vect_verify_full_masking): Likewise.
10248 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10249 retries, and free the mask rgroups before retrying. Check loop-wide
10250 reasons for disallowing fully-masked loops. Make the final decision
10251 about whether use a fully-masked loop or not.
10252 (vect_estimate_min_profitable_iters): Do not assume that peeling
10253 for the number of iterations will be needed for fully-masked loops.
10254 (vectorizable_reduction): Disable fully-masked loops.
10255 (vectorizable_live_operation): Likewise.
10256 (vect_halve_mask_nunits): New function.
10257 (vect_double_mask_nunits): Likewise.
10258 (vect_record_loop_mask): Likewise.
10259 (vect_get_loop_mask): Likewise.
10260 (vect_transform_loop): Handle the case in which the final loop
10261 iteration might handle a partial vector. Call vect_set_loop_condition
10262 instead of slpeel_make_loop_iterate_ntimes.
10263 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10264 (check_load_store_masking): New function.
10265 (prepare_load_store_mask): Likewise.
10266 (vectorizable_store): Handle fully-masked loops.
10267 (vectorizable_load): Likewise.
10268 (supportable_widening_operation): Use vect_halve_mask_nunits for
10270 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10271 (vect_gen_while): New function.
10272 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10273 (aarch64_uqdec<mode>): New insn.
10275 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10276 Alan Hayward <alan.hayward@arm.com>
10277 David Sherwood <david.sherwood@arm.com>
10279 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10280 (reduc_xor_scal_optab): New optabs.
10281 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10282 (reduc_xor_scal_@var{m}): Document.
10283 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10284 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10285 internal functions.
10286 * fold-const-call.c (fold_const_call): Handle them.
10287 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10288 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10289 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10290 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10291 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10292 (UNSPEC_XORV): New unspecs.
10293 (optab): Add entries for them.
10294 (BITWISEV): New int iterator.
10295 (bit_reduc_op): New int attributes.
10297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10298 Alan Hayward <alan.hayward@arm.com>
10299 David Sherwood <david.sherwood@arm.com>
10301 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10302 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10303 * optabs.def (vec_shl_insert_optab): New optab.
10304 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10305 (duplicate_and_interleave): Likewise.
10306 * tree-vect-loop.c: Include internal-fn.h.
10307 (neutral_op_for_slp_reduction): New function, split out from
10308 get_initial_defs_for_reduction.
10309 (get_initial_def_for_reduction): Handle option 2 for variable-length
10310 vectors by loading the neutral value into a vector and then shifting
10311 the initial value into element 0.
10312 (get_initial_defs_for_reduction): Replace the code argument with
10313 the neutral value calculated by neutral_op_for_slp_reduction.
10314 Use gimple_build_vector for constant-length vectors.
10315 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10316 but the first group_size elements have a neutral value.
10317 Use duplicate_and_interleave otherwise.
10318 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10319 Update call to get_initial_defs_for_reduction. Handle SLP
10320 reductions for variable-length vectors by creating one vector
10321 result for each scalar result, with the elements associated
10322 with other scalar results stubbed out with the neutral value.
10323 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10324 Require IFN_VEC_SHL_INSERT for double reductions on
10325 variable-length vectors, or SLP reductions that have
10326 a neutral value. Require can_duplicate_and_interleave_p
10327 support for variable-length unchained SLP reductions if there
10328 is no neutral value, such as for MIN/MAX reductions. Also require
10329 the number of vector elements to be a multiple of the number of
10330 SLP statements when doing variable-length unchained SLP reductions.
10331 Update call to vect_create_epilog_for_reduction.
10332 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10333 and remove initial values.
10334 (duplicate_and_interleave): Make public.
10335 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10336 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10338 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10339 Alan Hayward <alan.hayward@arm.com>
10340 David Sherwood <david.sherwood@arm.com>
10342 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10343 (can_duplicate_and_interleave_p): New function.
10344 (vect_get_and_check_slp_defs): Take the vector of statements
10345 rather than just the current one. Remove excess parentheses.
10346 Restriction rejectinon of vect_constant_def and vect_external_def
10347 for variable-length vectors to boolean types, or types for which
10348 can_duplicate_and_interleave_p is false.
10349 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10350 (duplicate_and_interleave): New function.
10351 (vect_get_constant_vectors): Use gimple_build_vector for
10352 constant-length vectors and suitable variable-length constant
10353 vectors. Use duplicate_and_interleave for other variable-length
10354 vectors. Don't defer the update when inserting new statements.
10356 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10357 Alan Hayward <alan.hayward@arm.com>
10358 David Sherwood <david.sherwood@arm.com>
10360 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10361 min_profitable_iters doesn't go negative.
10363 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10364 Alan Hayward <alan.hayward@arm.com>
10365 David Sherwood <david.sherwood@arm.com>
10367 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10368 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10369 * optabs.def (vec_mask_load_lanes_optab): New optab.
10370 (vec_mask_store_lanes_optab): Likewise.
10371 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10372 (MASK_STORE_LANES): Likewise.
10373 * internal-fn.c (mask_load_lanes_direct): New macro.
10374 (mask_store_lanes_direct): Likewise.
10375 (expand_mask_load_optab_fn): Handle masked operations.
10376 (expand_mask_load_lanes_optab_fn): New macro.
10377 (expand_mask_store_optab_fn): Handle masked operations.
10378 (expand_mask_store_lanes_optab_fn): New macro.
10379 (direct_mask_load_lanes_optab_supported_p): Likewise.
10380 (direct_mask_store_lanes_optab_supported_p): Likewise.
10381 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10383 (vect_load_lanes_supported): Likewise.
10384 * tree-vect-data-refs.c (strip_conversion): New function.
10385 (can_group_stmts_p): Likewise.
10386 (vect_analyze_data_ref_accesses): Use it instead of checking
10387 for a pair of assignments.
10388 (vect_store_lanes_supported): Take a masked_p parameter.
10389 (vect_load_lanes_supported): Likewise.
10390 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10391 vect_store_lanes_supported and vect_load_lanes_supported.
10392 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10393 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10394 parameter. Don't allow gaps for masked accesses.
10395 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10396 and vect_load_lanes_supported.
10397 (get_load_store_type): Take a masked_p parameter and update
10398 call to get_group_load_store_type.
10399 (vectorizable_store): Update call to get_load_store_type.
10400 Handle IFN_MASK_STORE_LANES.
10401 (vectorizable_load): Update call to get_load_store_type.
10402 Handle IFN_MASK_LOAD_LANES.
10404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10405 Alan Hayward <alan.hayward@arm.com>
10406 David Sherwood <david.sherwood@arm.com>
10408 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10410 * config/aarch64/aarch64-protos.h
10411 (aarch64_sve_struct_memory_operand_p): Declare.
10412 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10413 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10414 (VPRED, vpred): Handle SVE structure modes.
10415 * config/aarch64/constraints.md (Utx): New constraint.
10416 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10417 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10418 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10419 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10420 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10421 structure modes. Split into pieces after RA.
10422 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10423 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10425 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10426 SVE structure modes.
10427 (aarch64_classify_address): Likewise.
10428 (sizetochar): Move earlier in file.
10429 (aarch64_print_operand): Handle SVE register lists.
10430 (aarch64_array_mode): New function.
10431 (aarch64_sve_struct_memory_operand_p): Likewise.
10432 (TARGET_ARRAY_MODE): Redefine.
10434 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10435 Alan Hayward <alan.hayward@arm.com>
10436 David Sherwood <david.sherwood@arm.com>
10438 * target.def (array_mode): New target hook.
10439 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10440 * doc/tm.texi: Regenerate.
10441 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10442 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10443 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10444 targetm.array_mode.
10445 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10448 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10449 Alan Hayward <alan.hayward@arm.com>
10450 David Sherwood <david.sherwood@arm.com>
10452 * fold-const.c (fold_binary_loc): Check the argument types
10453 rather than the result type when testing for a vector operation.
10455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10457 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10458 * doc/tm.texi: Regenerate.
10460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10461 Alan Hayward <alan.hayward@arm.com>
10462 David Sherwood <david.sherwood@arm.com>
10464 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10465 (sve): Document new AArch64 extension.
10466 * doc/md.texi (w): Extend the description of the AArch64
10467 constraint to include SVE vectors.
10468 (Upl, Upa): Document new AArch64 predicate constraints.
10469 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10471 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10472 (msve-vector-bits=): New option.
10473 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10474 SVE when these are disabled.
10475 (sve): New extension.
10476 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10477 modes. Adjust their number of units based on aarch64_sve_vg.
10478 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10479 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10480 aarch64_addr_query_type.
10481 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10482 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10483 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10484 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10485 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10486 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10487 (aarch64_simd_imm_zero_p): Delete.
10488 (aarch64_check_zero_based_sve_index_immediate): Declare.
10489 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10490 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10491 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10492 (aarch64_sve_float_mul_immediate_p): Likewise.
10493 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10494 rather than an rtx.
10495 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10496 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10497 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10498 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10499 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10500 (aarch64_regmode_natural_size): Likewise.
10501 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10502 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10504 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10505 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10506 for VG and the SVE predicate registers.
10507 (V_ALIASES): Add a "z"-prefixed alias.
10508 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10509 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10510 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10511 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10512 (REG_CLASS_NAMES): Add entries for them.
10513 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10514 and the predicate registers.
10515 (aarch64_sve_vg): Declare.
10516 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10517 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10518 (REGMODE_NATURAL_SIZE): Define.
10519 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10521 * config/aarch64/aarch64.c: Include cfgrtl.h.
10522 (simd_immediate_info): Add a constructor for series vectors,
10523 and an associated step field.
10524 (aarch64_sve_vg): New variable.
10525 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10526 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10527 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10528 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10529 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10530 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10531 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10532 (aarch64_get_mask_mode): New functions.
10533 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10534 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10535 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10536 predicate modes and predicate registers. Explicitly restrict
10537 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10538 to store a vector mode if it is recognized by
10539 aarch64_classify_vector_mode.
10540 (aarch64_regmode_natural_size): New function.
10541 (aarch64_hard_regno_caller_save_mode): Return the original mode
10543 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10544 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10545 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10546 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10548 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10549 does not overlap dest if the function is frame-related. Handle
10551 (aarch64_split_add_offset): New function.
10552 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10553 them aarch64_add_offset.
10554 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10555 and update call to aarch64_sub_sp.
10556 (aarch64_add_cfa_expression): New function.
10557 (aarch64_expand_prologue): Pass extra temporary registers to the
10558 functions above. Handle the case in which we need to emit new
10559 DW_CFA_expressions for registers that were originally saved
10560 relative to the stack pointer, but now have to be expressed
10561 relative to the frame pointer.
10562 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10564 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10565 IP0 and IP1 values for SVE frames.
10566 (aarch64_expand_vec_series): New function.
10567 (aarch64_expand_sve_widened_duplicate): Likewise.
10568 (aarch64_expand_sve_const_vector): Likewise.
10569 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10570 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10571 into the register, rather than emitting a SET directly.
10572 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10573 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10574 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10575 (offset_9bit_signed_scaled_p): New functions.
10576 (aarch64_replicate_bitmask_imm): New function.
10577 (aarch64_bitmask_imm): Use it.
10578 (aarch64_cannot_force_const_mem): Reject expressions involving
10579 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10580 (aarch64_classify_index): Handle SVE indices, by requiring
10581 a plain register index with a scale that matches the element size.
10582 (aarch64_classify_address): Handle SVE addresses. Assert that
10583 the mode of the address is VOIDmode or an integer mode.
10584 Update call to aarch64_classify_symbol.
10585 (aarch64_classify_symbolic_expression): Update call to
10586 aarch64_classify_symbol.
10587 (aarch64_const_vec_all_in_range_p): New function.
10588 (aarch64_print_vector_float_operand): Likewise.
10589 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10590 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10591 and the FP immediates 1.0 and 0.5.
10592 (aarch64_print_address_internal): Handle SVE addresses.
10593 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10594 (aarch64_regno_regclass): Handle predicate registers.
10595 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10597 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10598 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10599 (aarch64_convert_sve_vector_bits): New function.
10600 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10601 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10602 rather than an rtx.
10603 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10604 Handle SVE vector and predicate modes. Accept VL-based constants
10605 that need only one temporary register, and VL offsets that require
10606 no temporary registers.
10607 (aarch64_conditional_register_usage): Mark the predicate registers
10608 as fixed if SVE isn't available.
10609 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10610 Return true for SVE vector and predicate modes.
10611 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10612 rather than an unsigned int. Handle SVE modes.
10613 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10615 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10617 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10618 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10619 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10620 (aarch64_sve_float_mul_immediate_p): New functions.
10621 (aarch64_sve_valid_immediate): New function.
10622 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10623 Explicitly reject structure modes. Check for INDEX constants.
10624 Handle PTRUE and PFALSE constants.
10625 (aarch64_check_zero_based_sve_index_immediate): New function.
10626 (aarch64_simd_imm_zero_p): Delete.
10627 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10628 vector modes. Accept constants in the range of CNT[BHWD].
10629 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10630 ask for an Advanced SIMD mode.
10631 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10632 (aarch64_simd_vector_alignment): Handle SVE predicates.
10633 (aarch64_vectorize_preferred_vector_alignment): New function.
10634 (aarch64_simd_vector_alignment_reachable): Use it instead of
10636 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10637 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10639 (MAX_VECT_LEN): Delete.
10640 (expand_vec_perm_d): Add a vec_flags field.
10641 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10642 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10643 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10645 (aarch64_evpc_rev): Rename to...
10646 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10647 (aarch64_evpc_rev_global): New function.
10648 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10649 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10651 (aarch64_evpc_sve_tbl): New function.
10652 (aarch64_expand_vec_perm_const_1): Update after rename of
10653 aarch64_evpc_rev. Handle SVE permutes too, trying
10654 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10655 than aarch64_evpc_tbl.
10656 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10657 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10658 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10659 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10660 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10661 (aarch64_expand_sve_vcond): New functions.
10662 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10663 of aarch64_vector_mode_p.
10664 (aarch64_dwarf_poly_indeterminate_value): New function.
10665 (aarch64_compute_pressure_classes): Likewise.
10666 (aarch64_can_change_mode_class): Likewise.
10667 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10668 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10669 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10670 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10671 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10672 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10673 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10674 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10676 (Dn, Dl, Dr): Accept const as well as const_vector.
10677 (Dz): Likewise. Compare against CONST0_RTX.
10678 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10679 of "vector" where appropriate.
10680 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10681 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10682 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10683 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10684 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10685 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10686 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10687 (v_int_equiv): Extend to SVE modes.
10688 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10690 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10691 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10692 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10693 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10694 (SVE_COND_FP_CMP): New int iterators.
10695 (perm_hilo): Handle the new unpack unspecs.
10696 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10698 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10699 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10700 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10701 (aarch64_equality_operator, aarch64_constant_vector_operand)
10702 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10703 (aarch64_sve_nonimmediate_operand): Likewise.
10704 (aarch64_sve_general_operand): Likewise.
10705 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10706 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10707 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10708 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10709 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10710 (aarch64_sve_float_arith_immediate): Likewise.
10711 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10712 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10713 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10714 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10715 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10716 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10717 (aarch64_sve_float_arith_operand): Likewise.
10718 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10719 (aarch64_sve_float_mul_operand): Likewise.
10720 (aarch64_sve_vec_perm_operand): Likewise.
10721 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10722 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10723 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10724 as well as const_vector.
10725 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10726 in file. Use CONST0_RTX and CONSTM1_RTX.
10727 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10728 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10729 Use aarch64_simd_imm_zero.
10730 * config/aarch64/aarch64-sve.md: New file.
10731 * config/aarch64/aarch64.md: Include it.
10732 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10733 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10734 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10735 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10736 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10737 (sve): New attribute.
10738 (enabled): Disable instructions with the sve attribute unless
10740 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10741 aarch64_expand_mov_immediate.
10742 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10743 CNT[BHSD] immediates.
10744 (movti): Split CONST_POLY_INT moves into two halves.
10745 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10746 Split additions that need a temporary here if the destination
10747 is the stack pointer.
10748 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10749 (*add<mode>3_poly_1): New instruction.
10750 (set_clobber_cc): New expander.
10752 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10754 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10755 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10756 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10757 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10758 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10759 Change innermode from fixed_mode_size to machine_mode.
10760 (simplify_subreg): Update call accordingly. Handle a constant-sized
10761 subreg of a variable-length CONST_VECTOR.
10763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10764 Alan Hayward <alan.hayward@arm.com>
10765 David Sherwood <david.sherwood@arm.com>
10767 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10768 (add_offset_to_base): New function, split out from...
10769 (create_mem_ref): ...here. When handling a scale other than 1,
10770 check first whether the address is valid without the offset.
10771 Add it into the base if so, leaving the index and scale as-is.
10773 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10776 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10777 fold_for_warn before checking if arg2 is INTEGER_CST.
10779 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10781 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10782 (store_multiple_operation): Delete.
10783 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10784 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10785 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10786 guarded by TARGET_STRING.
10787 (rs6000_output_load_multiple): Delete.
10788 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10789 OPTION_MASK_STRING / TARGET_STRING handling.
10790 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10791 (const rs6000_opt_masks) <"string">: Change mask to 0.
10792 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10793 (MASK_STRING): Delete.
10794 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10796 (load_multiple): Delete.
10803 (store_multiple): Delete.
10810 (movmemsi_8reg): Delete.
10811 (corresponding unnamed define_insn): Delete.
10812 (movmemsi_6reg): Delete.
10813 (corresponding unnamed define_insn): Delete.
10814 (movmemsi_4reg): Delete.
10815 (corresponding unnamed define_insn): Delete.
10816 (movmemsi_2reg): Delete.
10817 (corresponding unnamed define_insn): Delete.
10818 (movmemsi_1reg): Delete.
10819 (corresponding unnamed define_insn): Delete.
10820 * config/rs6000/rs6000.opt (mno-string): New.
10821 (mstring): Replace by deprecation warning stub.
10822 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10824 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10826 * regrename.c (regrename_do_replace): If replacing the same
10827 reg multiple times, try to reuse last created gen_raw_REG.
10830 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10831 main to workaround a bug in GDB.
10833 2018-01-12 Tom de Vries <tom@codesourcery.com>
10836 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10838 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10840 PR rtl-optimization/80481
10841 * ira-color.c (get_cap_member): New function.
10842 (allocnos_conflict_by_live_ranges_p): Use it.
10843 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10844 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10846 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10849 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10850 (*saddl_se_1): Ditto.
10851 (*ssubsi_1): Ditto.
10852 (*ssubl_se_1): Ditto.
10854 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10856 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10857 rather than wi::to_widest for DR_INITs.
10858 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10859 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10860 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10862 (vect_analyze_group_access_1): Note that here.
10864 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10866 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10867 polynomial type sizes.
10869 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10871 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10872 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10873 (gimple_add_tmp_var): Likewise.
10875 2018-01-12 Martin Liska <mliska@suse.cz>
10877 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10878 (gimple_alloc_sizes): Likewise.
10879 (dump_gimple_statistics): Use PRIu64 in printf format.
10880 * gimple.h: Change uint64_t to int.
10882 2018-01-12 Martin Liska <mliska@suse.cz>
10884 * tree-core.h: Use uint64_t instead of int.
10885 * tree.c (tree_node_counts): Likewise.
10886 (tree_node_sizes): Likewise.
10887 (dump_tree_statistics): Use PRIu64 in printf format.
10889 2018-01-12 Martin Liska <mliska@suse.cz>
10891 * Makefile.in: As qsort_chk is implemented in vec.c, add
10892 vec.o to linkage of gencfn-macros.
10893 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10894 passing the info to record_node_allocation_statistics.
10895 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10897 * ggc-common.c (struct ggc_usage): Add operator== and use
10898 it in operator< and compare function.
10899 * mem-stats.h (struct mem_usage): Likewise.
10900 * vec.c (struct vec_usage): Remove operator< and compare
10901 function. Can be simply inherited.
10903 2018-01-12 Martin Jambor <mjambor@suse.cz>
10906 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10907 * tree-ssa-math-opts.c: Include domwalk.h.
10908 (convert_mult_to_fma_1): New function.
10909 (fma_transformation_info): New type.
10910 (fma_deferring_state): Likewise.
10911 (cancel_fma_deferring): New function.
10912 (result_of_phi): Likewise.
10913 (last_fma_candidate_feeds_initial_phi): Likewise.
10914 (convert_mult_to_fma): Added deferring logic, split actual
10915 transformation to convert_mult_to_fma_1.
10916 (math_opts_dom_walker): New type.
10917 (math_opts_dom_walker::after_dom_children): New method, body moved
10918 here from pass_optimize_widening_mul::execute, added deferring logic
10920 (pass_optimize_widening_mul::execute): Moved most of code to
10921 math_opts_dom_walker::after_dom_children.
10922 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10923 * config/i386/i386.c (ix86_option_override_internal): Added
10924 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10926 2018-01-12 Richard Biener <rguenther@suse.de>
10929 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10930 inline instance vars.
10932 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10935 * config/rx/rx.c (rx_is_restricted_memory_address):
10936 Handle SUBREG case.
10938 2018-01-12 Richard Biener <rguenther@suse.de>
10940 PR tree-optimization/80846
10941 * target.def (split_reduction): New target hook.
10942 * targhooks.c (default_split_reduction): New function.
10943 * targhooks.h (default_split_reduction): Declare.
10944 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10945 target requests first reduce vectors by combining low and high
10947 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10948 (get_vectype_for_scalar_type_and_size): Export.
10949 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10950 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10951 * doc/tm.texi: Regenerate.
10952 * config/i386/i386.c (ix86_split_reduction): Implement
10953 TARGET_VECTORIZE_SPLIT_REDUCTION.
10955 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10958 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10959 in PIC mode except for TARGET_VXWORKS_RTP.
10960 * config/sparc/sparc.c: Include cfgrtl.h.
10961 (TARGET_INIT_PIC_REG): Define.
10962 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10963 (sparc_pic_register_p): New predicate.
10964 (sparc_legitimate_address_p): Use it.
10965 (sparc_legitimize_pic_address): Likewise.
10966 (sparc_delegitimize_address): Likewise.
10967 (sparc_mode_dependent_address_p): Likewise.
10968 (gen_load_pcrel_sym): Remove 4th parameter.
10969 (load_got_register): Adjust call to above. Remove obsolete stuff.
10970 (sparc_expand_prologue): Do not call load_got_register here.
10971 (sparc_flat_expand_prologue): Likewise.
10972 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10973 (sparc_use_pseudo_pic_reg): New function.
10974 (sparc_init_pic_reg): Likewise.
10975 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10976 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10978 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10980 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10981 Add item for branch_cost.
10983 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10985 PR rtl-optimization/83565
10986 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10987 not extend the result to a larger mode for rotate operations.
10988 (num_sign_bit_copies1): Likewise.
10990 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10993 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10995 Use values-Xc.o for -pedantic.
10996 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10998 2018-01-12 Martin Liska <mliska@suse.cz>
11001 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11003 (possible_polymorphic_call_targets): Use it.
11004 (ipa_devirt): Likewise.
11006 2018-01-12 Martin Liska <mliska@suse.cz>
11008 * profile-count.h (enum profile_quality): Use 0 as invalid
11009 enum value of profile_quality.
11011 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11013 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11014 -mext-string options.
11016 2018-01-12 Richard Biener <rguenther@suse.de>
11018 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11019 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11020 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11022 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11024 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11026 * configure.ac (--with-long-double-format): Add support for the
11027 configuration option to change the default long double format on
11029 * config.gcc (powerpc*-linux*-*): Likewise.
11030 * configure: Regenerate.
11031 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11032 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11033 used without modification.
11035 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11037 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11038 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11039 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11040 MISC_BUILTIN_SPEC_BARRIER.
11041 (rs6000_init_builtins): Likewise.
11042 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11044 (speculation_barrier): New define_insn.
11045 * doc/extend.texi: Document __builtin_speculation_barrier.
11047 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11050 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11051 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11052 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11054 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11055 integral modes instead of "ss" and "sd".
11056 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11057 vectors with 32-bit and 64-bit elements.
11058 (vecdupssescalarmodesuffix): New mode attribute.
11059 (vec_dup<mode>): Use it.
11061 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11064 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11065 frame if argument is passed on stack.
11067 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11070 * ree.c (combine_reaching_defs): Optimize also
11071 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11072 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11074 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11076 PR middle-end/83189
11077 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11079 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11081 PR middle-end/83718
11082 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11083 after they are computed.
11085 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11087 PR tree-optimization/83695
11088 * gimple-loop-linterchange.cc
11089 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11090 reset cached scev information after interchange.
11091 (pass_linterchange::execute): Remove call to scev_reset_htab.
11093 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11095 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11096 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11097 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11098 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11099 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11100 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11101 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11102 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11103 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11104 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11105 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11106 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11107 (V_lane_reg): Likewise.
11108 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11110 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11111 (vfmal_lane_low<mode>_intrinsic,
11112 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11113 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11114 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11115 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11116 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11117 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11119 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11121 * config/arm/arm-cpus.in (fp16fml): New feature.
11122 (ALL_SIMD): Add fp16fml.
11123 (armv8.2-a): Add fp16fml as an option.
11124 (armv8.3-a): Likewise.
11125 (armv8.4-a): Add fp16fml as part of fp16.
11126 * config/arm/arm.h (TARGET_FP16FML): Define.
11127 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11129 * config/arm/arm-modes.def (V2HF): Define.
11130 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11131 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11132 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11133 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11134 vfmsl_low, vfmsl_high): New set of builtins.
11135 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11136 (vfml_op): New code attribute.
11137 (VFMLHALVES): New int iterator.
11138 (VFML, VFMLSEL): New mode attributes.
11139 (V_reg): Define mapping for V2HF.
11140 (V_hi, V_lo): New mode attributes.
11141 (VF_constraint): Likewise.
11142 (vfml_half, vfml_half_selector): New int attributes.
11143 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11145 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11146 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11148 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11149 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11150 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11151 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11153 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11154 Document new effective target and option set.
11156 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11158 * config/arm/arm-cpus.in (armv8_4): New feature.
11159 (ARMv8_4a): New fgroup.
11160 (armv8.4-a): New arch.
11161 * config/arm/arm-tables.opt: Regenerate.
11162 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11163 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11164 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11165 Add matching rules for -march=armv8.4-a and extensions.
11166 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11168 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11171 * config/rx/rx.md (BW): New mode attribute.
11172 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11174 2018-01-11 Richard Biener <rguenther@suse.de>
11176 PR tree-optimization/83435
11177 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11178 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11179 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11181 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11182 Alan Hayward <alan.hayward@arm.com>
11183 David Sherwood <david.sherwood@arm.com>
11185 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11187 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11188 (aarch64_print_address_internal): Use it to check for a zero offset.
11190 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11191 Alan Hayward <alan.hayward@arm.com>
11192 David Sherwood <david.sherwood@arm.com>
11194 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11195 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11196 Return a poly_int64 rather than a HOST_WIDE_INT.
11197 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11198 rather than a HOST_WIDE_INT.
11199 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11200 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11201 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11202 final_offset from HOST_WIDE_INT to poly_int64.
11203 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11204 to_constant when getting the number of units in an Advanced SIMD
11206 (aarch64_builtin_vectorized_function): Check for a constant number
11208 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11210 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11211 attribute instead of GET_MODE_NUNITS.
11212 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11213 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11214 GET_MODE_SIZE for fixed-size registers.
11215 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11216 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11217 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11218 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11219 (aarch64_print_operand, aarch64_print_address_internal)
11220 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11221 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11222 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11223 Handle polynomial GET_MODE_SIZE.
11224 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11225 wider than SImode without modification.
11226 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11227 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11228 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11229 passing and returning SVE modes.
11230 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11231 rather than GEN_INT.
11232 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11233 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11234 (aarch64_allocate_and_probe_stack_space): Likewise.
11235 (aarch64_layout_frame): Cope with polynomial offsets.
11236 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11237 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11238 polynomial offsets.
11239 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11240 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11241 poly_int64 rather than a HOST_WIDE_INT.
11242 (aarch64_get_separate_components, aarch64_process_components)
11243 (aarch64_expand_prologue, aarch64_expand_epilogue)
11244 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11245 (aarch64_anchor_offset): New function, split out from...
11246 (aarch64_legitimize_address): ...here.
11247 (aarch64_builtin_vectorization_cost): Handle polynomial
11248 TYPE_VECTOR_SUBPARTS.
11249 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11251 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11252 number of elements from the PARALLEL rather than the mode.
11253 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11254 rather than GET_MODE_BITSIZE.
11255 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11256 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11257 (aarch64_expand_vec_perm_const_1): Handle polynomial
11258 d->perm.length () and d->perm elements.
11259 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11260 Apply to_constant to d->perm elements.
11261 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11262 polynomial CONST_VECTOR_NUNITS.
11263 (aarch64_move_pointer): Take amount as a poly_int64 rather
11265 (aarch64_progress_pointer): Avoid temporary variable.
11266 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11267 the mode attribute instead of GET_MODE.
11269 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11270 Alan Hayward <alan.hayward@arm.com>
11271 David Sherwood <david.sherwood@arm.com>
11273 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11274 x exists before using it.
11275 (aarch64_add_constant_internal): Rename to...
11276 (aarch64_add_offset_1): ...this. Replace regnum with separate
11277 src and dest rtxes. Handle the case in which they're different,
11278 including when the offset is zero. Replace scratchreg with an rtx.
11279 Use 2 additions if there is no spare register into which we can
11280 move a 16-bit constant.
11281 (aarch64_add_constant): Delete.
11282 (aarch64_add_offset): Replace reg with separate src and dest
11283 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11284 Use aarch64_add_offset_1.
11285 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11286 an rtx rather than an int. Take the delta as a poly_int64
11287 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11288 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11289 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11290 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11291 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11292 and aarch64_add_sp.
11293 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11294 aarch64_add_constant.
11296 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11298 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11299 Use scalar_float_mode.
11301 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11303 * config/aarch64/aarch64-simd.md
11304 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11305 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11306 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11307 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11308 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11309 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11310 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11311 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11312 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11313 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11315 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11318 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11319 targ_options->x_arm_arch_string is non NULL.
11321 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11323 * config/aarch64/aarch64.h
11324 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11326 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11329 * expmed.c (emit_store_flag_force): Swap if const op0
11330 and change VOIDmode to mode of op0.
11332 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11334 PR rtl-optimization/83761
11335 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11336 than bytes to mode_for_size.
11338 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11340 PR middle-end/83189
11341 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11342 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11345 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11347 PR middle-end/83575
11348 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11349 when in layout mode.
11350 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11351 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11354 2018-01-10 Michael Collison <michael.collison@arm.com>
11356 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11357 * config/aarch64/aarch64-option-extension.def: Add
11358 AARCH64_OPT_EXTENSION of 'fp16fml'.
11359 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11360 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11361 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11362 * config/aarch64/constraints.md (Ui7): New constraint.
11363 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11364 (VFMLA_SEL_W): Ditto.
11367 (VFMLA16_LOW): New int iterator.
11368 (VFMLA16_HIGH): Ditto.
11369 (UNSPEC_FMLAL): New unspec.
11370 (UNSPEC_FMLSL): Ditto.
11371 (UNSPEC_FMLAL2): Ditto.
11372 (UNSPEC_FMLSL2): Ditto.
11373 (f16mac): New code attribute.
11374 * config/aarch64/aarch64-simd-builtins.def
11375 (aarch64_fmlal_lowv2sf): Ditto.
11376 (aarch64_fmlsl_lowv2sf): Ditto.
11377 (aarch64_fmlalq_lowv4sf): Ditto.
11378 (aarch64_fmlslq_lowv4sf): Ditto.
11379 (aarch64_fmlal_highv2sf): Ditto.
11380 (aarch64_fmlsl_highv2sf): Ditto.
11381 (aarch64_fmlalq_highv4sf): Ditto.
11382 (aarch64_fmlslq_highv4sf): Ditto.
11383 (aarch64_fmlal_lane_lowv2sf): Ditto.
11384 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11385 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11386 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11387 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11388 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11389 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11390 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11391 (aarch64_fmlal_lane_highv2sf): Ditto.
11392 (aarch64_fmlsl_lane_highv2sf): Ditto.
11393 (aarch64_fmlal_laneq_highv2sf): Ditto.
11394 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11395 (aarch64_fmlalq_lane_highv4sf): Ditto.
11396 (aarch64_fmlsl_lane_highv4sf): Ditto.
11397 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11398 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11399 * config/aarch64/aarch64-simd.md:
11400 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11401 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11402 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11403 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11404 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11405 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11406 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11407 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11408 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11409 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11410 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11411 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11412 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11413 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11414 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11415 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11416 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11417 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11418 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11419 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11420 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11421 (vfmlsl_low_u32): Ditto.
11422 (vfmlalq_low_u32): Ditto.
11423 (vfmlslq_low_u32): Ditto.
11424 (vfmlal_high_u32): Ditto.
11425 (vfmlsl_high_u32): Ditto.
11426 (vfmlalq_high_u32): Ditto.
11427 (vfmlslq_high_u32): Ditto.
11428 (vfmlal_lane_low_u32): Ditto.
11429 (vfmlsl_lane_low_u32): Ditto.
11430 (vfmlal_laneq_low_u32): Ditto.
11431 (vfmlsl_laneq_low_u32): Ditto.
11432 (vfmlalq_lane_low_u32): Ditto.
11433 (vfmlslq_lane_low_u32): Ditto.
11434 (vfmlalq_laneq_low_u32): Ditto.
11435 (vfmlslq_laneq_low_u32): Ditto.
11436 (vfmlal_lane_high_u32): Ditto.
11437 (vfmlsl_lane_high_u32): Ditto.
11438 (vfmlal_laneq_high_u32): Ditto.
11439 (vfmlsl_laneq_high_u32): Ditto.
11440 (vfmlalq_lane_high_u32): Ditto.
11441 (vfmlslq_lane_high_u32): Ditto.
11442 (vfmlalq_laneq_high_u32): Ditto.
11443 (vfmlslq_laneq_high_u32): Ditto.
11444 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11445 (AARCH64_FL_FOR_ARCH8_4): New.
11446 (AARCH64_ISA_F16FML): New ISA flag.
11447 (TARGET_F16FML): New feature flag for fp16fml.
11448 (doc/invoke.texi): Document new fp16fml option.
11450 2018-01-10 Michael Collison <michael.collison@arm.com>
11452 * config/aarch64/aarch64-builtins.c:
11453 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11454 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11455 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11456 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11457 (AARCH64_ISA_SHA3): New ISA flag.
11458 (TARGET_SHA3): New feature flag for sha3.
11459 * config/aarch64/iterators.md (sha512_op): New int attribute.
11460 (CRYPTO_SHA512): New int iterator.
11461 (UNSPEC_SHA512H): New unspec.
11462 (UNSPEC_SHA512H2): Ditto.
11463 (UNSPEC_SHA512SU0): Ditto.
11464 (UNSPEC_SHA512SU1): Ditto.
11465 * config/aarch64/aarch64-simd-builtins.def
11466 (aarch64_crypto_sha512hqv2di): New builtin.
11467 (aarch64_crypto_sha512h2qv2di): Ditto.
11468 (aarch64_crypto_sha512su0qv2di): Ditto.
11469 (aarch64_crypto_sha512su1qv2di): Ditto.
11470 (aarch64_eor3qv8hi): Ditto.
11471 (aarch64_rax1qv2di): Ditto.
11472 (aarch64_xarqv2di): Ditto.
11473 (aarch64_bcaxqv8hi): Ditto.
11474 * config/aarch64/aarch64-simd.md:
11475 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11476 (aarch64_crypto_sha512su0qv2di): Ditto.
11477 (aarch64_crypto_sha512su1qv2di): Ditto.
11478 (aarch64_eor3qv8hi): Ditto.
11479 (aarch64_rax1qv2di): Ditto.
11480 (aarch64_xarqv2di): Ditto.
11481 (aarch64_bcaxqv8hi): Ditto.
11482 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11483 (vsha512h2q_u64): Ditto.
11484 (vsha512su0q_u64): Ditto.
11485 (vsha512su1q_u64): Ditto.
11486 (veor3q_u16): Ditto.
11487 (vrax1q_u64): Ditto.
11488 (vxarq_u64): Ditto.
11489 (vbcaxq_u16): Ditto.
11490 * config/arm/types.md (crypto_sha512): New type attribute.
11491 (crypto_sha3): Ditto.
11492 (doc/invoke.texi): Document new sha3 option.
11494 2018-01-10 Michael Collison <michael.collison@arm.com>
11496 * config/aarch64/aarch64-builtins.c:
11497 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11498 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11499 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11500 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11501 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11502 (AARCH64_ISA_SM4): New ISA flag.
11503 (TARGET_SM4): New feature flag for sm4.
11504 * config/aarch64/aarch64-simd-builtins.def
11505 (aarch64_sm3ss1qv4si): Ditto.
11506 (aarch64_sm3tt1aq4si): Ditto.
11507 (aarch64_sm3tt1bq4si): Ditto.
11508 (aarch64_sm3tt2aq4si): Ditto.
11509 (aarch64_sm3tt2bq4si): Ditto.
11510 (aarch64_sm3partw1qv4si): Ditto.
11511 (aarch64_sm3partw2qv4si): Ditto.
11512 (aarch64_sm4eqv4si): Ditto.
11513 (aarch64_sm4ekeyqv4si): Ditto.
11514 * config/aarch64/aarch64-simd.md:
11515 (aarch64_sm3ss1qv4si): Ditto.
11516 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11517 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11518 (aarch64_sm4eqv4si): Ditto.
11519 (aarch64_sm4ekeyqv4si): Ditto.
11520 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11521 (sm3part_op): Ditto.
11522 (CRYPTO_SM3TT): Ditto.
11523 (CRYPTO_SM3PART): Ditto.
11524 (UNSPEC_SM3SS1): New unspec.
11525 (UNSPEC_SM3TT1A): Ditto.
11526 (UNSPEC_SM3TT1B): Ditto.
11527 (UNSPEC_SM3TT2A): Ditto.
11528 (UNSPEC_SM3TT2B): Ditto.
11529 (UNSPEC_SM3PARTW1): Ditto.
11530 (UNSPEC_SM3PARTW2): Ditto.
11531 (UNSPEC_SM4E): Ditto.
11532 (UNSPEC_SM4EKEY): Ditto.
11533 * config/aarch64/constraints.md (Ui2): New constraint.
11534 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11535 * config/arm/types.md (crypto_sm3): New type attribute.
11536 (crypto_sm4): Ditto.
11537 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11538 (vsm3tt1aq_u32): Ditto.
11539 (vsm3tt1bq_u32): Ditto.
11540 (vsm3tt2aq_u32): Ditto.
11541 (vsm3tt2bq_u32): Ditto.
11542 (vsm3partw1q_u32): Ditto.
11543 (vsm3partw2q_u32): Ditto.
11544 (vsm4eq_u32): Ditto.
11545 (vsm4ekeyq_u32): Ditto.
11546 (doc/invoke.texi): Document new sm4 option.
11548 2018-01-10 Michael Collison <michael.collison@arm.com>
11550 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11551 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11552 (AARCH64_FL_FOR_ARCH8_4): New.
11553 (AARCH64_FL_V8_4): New flag.
11554 (doc/invoke.texi): Document new armv8.4-a option.
11556 2018-01-10 Michael Collison <michael.collison@arm.com>
11558 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11559 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11560 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11561 * config/aarch64/aarch64-option-extension.def: Add
11562 AARCH64_OPT_EXTENSION of 'sha2'.
11563 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11564 (crypto): Disable sha2 and aes if crypto disabled.
11565 (crypto): Enable aes and sha2 if enabled.
11566 (simd): Disable sha2 and aes if simd disabled.
11567 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11569 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11570 (TARGET_SHA2): New feature flag for sha2.
11571 (TARGET_AES): New feature flag for aes.
11572 * config/aarch64/aarch64-simd.md:
11573 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11574 conditional on TARGET_AES.
11575 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11576 (aarch64_crypto_sha1hsi): Make pattern conditional
11578 (aarch64_crypto_sha1hv4si): Ditto.
11579 (aarch64_be_crypto_sha1hv4si): Ditto.
11580 (aarch64_crypto_sha1su1v4si): Ditto.
11581 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11582 (aarch64_crypto_sha1su0v4si): Ditto.
11583 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11584 (aarch64_crypto_sha256su0v4si): Ditto.
11585 (aarch64_crypto_sha256su1v4si): Ditto.
11586 (doc/invoke.texi): Document new aes and sha2 options.
11588 2018-01-10 Martin Sebor <msebor@redhat.com>
11590 PR tree-optimization/83781
11591 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11594 2018-01-11 Martin Sebor <msebor@gmail.com>
11595 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11597 PR tree-optimization/83501
11598 PR tree-optimization/81703
11600 * tree-ssa-strlen.c (get_string_cst): Rename...
11601 (get_string_len): ...to this. Handle global constants.
11602 (handle_char_store): Adjust.
11604 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11605 Jim Wilson <jimw@sifive.com>
11607 * config/riscv/riscv-protos.h (riscv_output_return): New.
11608 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11609 (riscv_attribute_table, riscv_output_return),
11610 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11611 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11612 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11614 (riscv_expand_prologue): Add early return for naked function.
11615 (riscv_expand_epilogue): Likewise.
11616 (riscv_function_ok_for_sibcall): Return false for naked function.
11617 (riscv_set_current_function): New.
11618 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11619 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11620 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11621 * doc/extend.texi (RISC-V Function Attributes): New.
11623 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11625 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11626 check for 128-bit long double before checking TCmode.
11627 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11628 128-bit long doubles before checking TFmode or TCmode.
11629 (FLOAT128_IBM_P): Likewise.
11631 2018-01-10 Martin Sebor <msebor@redhat.com>
11633 PR tree-optimization/83671
11634 * builtins.c (c_strlen): Unconditionally return zero for the empty
11636 Use -Warray-bounds for warnings.
11637 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11638 for non-constant array indices with COMPONENT_REF, arrays of
11639 arrays, and pointers to arrays.
11640 (gimple_fold_builtin_strlen): Determine and set length range for
11641 non-constant character arrays.
11643 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11645 PR middle-end/81897
11646 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11649 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11651 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11653 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11656 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11657 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11658 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11659 indexed_or_indirect_operand predicate.
11660 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11661 (*vsx_le_perm_load_v8hi): Likewise.
11662 (*vsx_le_perm_load_v16qi): Likewise.
11663 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11664 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11665 (*vsx_le_perm_store_v8hi): Likewise.
11666 (*vsx_le_perm_store_v16qi): Likewise.
11667 (eight unnamed splitters): Likewise.
11669 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11671 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11672 * config/rs6000/emmintrin.h: Likewise.
11673 * config/rs6000/mmintrin.h: Likewise.
11674 * config/rs6000/xmmintrin.h: Likewise.
11676 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11679 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11681 * tree.c (tree_nop_conversion): Return true for location wrapper
11683 (maybe_wrap_with_location): New function.
11684 (selftest::check_strip_nops): New function.
11685 (selftest::test_location_wrappers): New function.
11686 (selftest::tree_c_tests): Call it.
11687 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11688 (maybe_wrap_with_location): New decl.
11689 (EXPR_LOCATION_WRAPPER_P): New macro.
11690 (location_wrapper_p): New inline function.
11691 (tree_strip_any_location_wrapper): New inline function.
11693 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11696 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11697 stack_realign_offset for the largest alignment of stack slot
11699 (ix86_find_max_used_stack_alignment): New function.
11700 (ix86_finalize_stack_frame_flags): Use it. Set
11701 max_used_stack_alignment if we don't realign stack.
11702 * config/i386/i386.h (machine_function): Add
11703 max_used_stack_alignment.
11705 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11707 * config/arm/arm.opt (-mbranch-cost): New option.
11708 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11711 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11714 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11715 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11717 2018-01-10 Richard Biener <rguenther@suse.de>
11720 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11721 early out so it also covers the case where we have a non-NULL
11724 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11726 PR tree-optimization/83753
11727 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11728 for non-strided grouped accesses if the number of elements is 1.
11730 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11733 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11734 * i386.h (TARGET_USE_GATHER): Define.
11735 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11737 2018-01-10 Martin Liska <mliska@suse.cz>
11740 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11741 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11743 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11744 CLEANUP_NO_PARTITIONING is not set.
11746 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11748 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11749 for vectors, as a partial revert of r254296.
11750 * rtl.h (const_vec_p): Delete.
11751 (const_vec_duplicate_p): Don't test for vector CONSTs.
11752 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11753 * expmed.c (make_tree): Likewise.
11756 * common.md (E, F): Use CONSTANT_P instead of checking for
11758 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11759 checking for CONST_VECTOR.
11761 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11763 PR middle-end/83575
11764 * predict.c (force_edge_cold): Handle in more sane way edges
11765 with no prediction.
11767 2018-01-09 Carl Love <cel@us.ibm.com>
11769 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11771 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11772 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11773 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11774 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11775 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11776 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11777 * config/rs6000/rs6000-protos.h: Add extern defition for
11778 rs6000_generate_float2_double_code.
11779 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11781 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11782 (float2_v2df): Add define_expand.
11784 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11787 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11788 op_mode in the force_to_mode call.
11790 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11792 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11793 instead of checking each element individually.
11794 (aarch64_evpc_uzp): Likewise.
11795 (aarch64_evpc_zip): Likewise.
11796 (aarch64_evpc_ext): Likewise.
11797 (aarch64_evpc_rev): Likewise.
11798 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11799 instead of checking each element individually. Return true without
11801 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11802 whether all selected elements come from the same input, instead of
11803 checking each element individually. Remove calls to gen_rtx_REG,
11804 start_sequence and end_sequence and instead assert that no rtl is
11807 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11809 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11810 order of HIGH and CONST checks.
11812 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11814 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11815 if the destination isn't an SSA_NAME.
11817 2018-01-09 Richard Biener <rguenther@suse.de>
11819 PR tree-optimization/83668
11820 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11822 (canonicalize_loop_form): ... here, renamed from ...
11823 (canonicalize_loop_closed_ssa_form): ... this and amended to
11824 swap successor edges for loop exit blocks to make us use
11825 the RPO order we need for initial schedule generation.
11827 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11829 PR tree-optimization/64811
11830 * match.pd: When optimizing comparisons with Inf, avoid
11831 introducing or losing exceptions from comparisons with NaN.
11833 2018-01-09 Martin Liska <mliska@suse.cz>
11836 * asan.c (shadow_mem_size): Add gcc_assert.
11838 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11840 Don't save registers in main().
11843 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11844 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11845 * config/avr/avr.c (avr_set_current_function): Don't error if
11846 naked, OS_task or OS_main are specified at the same time.
11847 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11849 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11851 * common/config/avr/avr-common.c (avr_option_optimization_table):
11852 Switch on -mmain-is-OS_task for optimizing compilations.
11854 2018-01-09 Richard Biener <rguenther@suse.de>
11856 PR tree-optimization/83572
11857 * graphite.c: Include cfganal.h.
11858 (graphite_transform_loops): Connect infinite loops to exit
11859 and remove fake edges at the end.
11861 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11863 * ipa-inline.c (edge_badness): Revert accidental checkin.
11865 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11868 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11869 symbols; not inline clones.
11871 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11874 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11875 hard registers. Formatting fixes.
11877 PR preprocessor/83722
11878 * gcc.c (try_generate_repro): Pass
11879 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11880 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11883 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11884 Kito Cheng <kito.cheng@gmail.com>
11886 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11887 (riscv_leaf_function_p): Delete.
11888 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11890 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11892 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11894 (do_ifelse): New function.
11895 (do_isel): New function.
11896 (do_sub3): New function.
11897 (do_add3): New function.
11898 (do_load_mask_compare): New function.
11899 (do_overlap_load_compare): New function.
11900 (expand_compare_loop): New function.
11901 (expand_block_compare): Call expand_compare_loop() when appropriate.
11902 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11903 option description.
11904 (-mblock-compare-inline-loop-limit): New option.
11906 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11909 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11910 Reverse order of second and third operands in first alternative.
11911 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11912 of first and second elements in UNSPEC_VPERMR vector.
11913 (altivec_expand_vec_perm_le): Likewise.
11915 2018-01-08 Jeff Law <law@redhat.com>
11917 PR rtl-optimizatin/81308
11918 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11919 (process_switch): If group_case_labels makes a change, then set
11921 (pass_convert_switch::execute): If a switch is converted, then
11922 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11924 PR rtl-optimization/81308
11925 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11928 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11930 PR target/83663 - Revert r255946
11931 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11932 generation for cases where splatting a value is not useful.
11933 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11934 across a vec_duplicate and a paradoxical subreg forming a vector
11935 mode to a vec_concat.
11937 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11939 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11940 -march=armv8.3-a variants.
11941 * config/arm/t-multilib: Likewise.
11942 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11944 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11946 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11948 (cceq_ior_compare_complement): Give it a name so I can use it, and
11949 change boolean_or_operator predicate to boolean_operator so it can
11950 be used to generate a crand.
11951 (eqne): New code iterator.
11952 (bd/bd_neg): New code_attrs.
11953 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11954 a single define_insn.
11955 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11956 decrement (bdnzt/bdnzf/bdzt/bdzf).
11957 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11958 with the new names of the branch decrement patterns, and added the
11959 names of the branch decrement conditional patterns.
11961 2018-01-08 Richard Biener <rguenther@suse.de>
11963 PR tree-optimization/83563
11964 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11967 2018-01-08 Richard Biener <rguenther@suse.de>
11969 PR middle-end/83713
11970 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11972 2018-01-08 Richard Biener <rguenther@suse.de>
11974 PR tree-optimization/83685
11975 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11976 references to abnormals.
11978 2018-01-08 Richard Biener <rguenther@suse.de>
11981 * dwarf2out.c (output_indirect_strings): Handle empty
11982 skeleton_debug_str_hash.
11983 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11985 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11987 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11988 (emit_store_direct): Likewise.
11989 (arc_trampoline_adjust_address): Likewise.
11990 (arc_asm_trampoline_template): New function.
11991 (arc_initialize_trampoline): Use asm_trampoline_template.
11992 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11993 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11994 * config/arc/arc.md (flush_icache): Delete pattern.
11996 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11998 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11999 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12002 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12005 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12006 by not USED_FOR_TARGET.
12007 (make_pass_resolve_sw_modes): Likewise.
12009 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12011 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12014 2018-01-08 Richard Biener <rguenther@suse.de>
12016 PR middle-end/83580
12017 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12019 2018-01-08 Richard Biener <rguenther@suse.de>
12021 PR middle-end/83517
12022 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12024 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12026 PR middle-end/81897
12027 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12028 basic blocks with a small number of successors.
12029 (convert_control_dep_chain_into_preds): Improve handling of
12031 (dump_predicates): Split apart into...
12032 (dump_pred_chain): ...here...
12033 (dump_pred_info): ...and here.
12034 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12035 (can_chain_union_be_invalidated_p): Improve check for invalidation
12037 (uninit_uses_cannot_happen): Avoid unnecessary if
12038 convert_control_dep_chain_into_preds yielded nothing.
12040 2018-01-06 Martin Sebor <msebor@redhat.com>
12042 PR tree-optimization/83640
12043 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12044 subtracting negative offset from size.
12045 (builtin_access::overlap): Adjust offset bounds of the access to fall
12046 within the size of the object if possible.
12048 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12050 PR rtl-optimization/83699
12051 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12052 extract_bit_field_as_subreg to cases in which the extracted
12053 value is also a vector.
12055 * lra-constraints.c (process_alt_operands): Test for the equivalence
12056 substitutions when detecting a possible reload cycle.
12058 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12061 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12062 by default if flag_selective_schedling{,2}. Formatting fixes.
12064 PR rtl-optimization/83682
12065 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12066 if it has non-VECTOR_MODE element mode.
12067 (vec_duplicate_p): Likewise.
12069 PR middle-end/83694
12070 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12071 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12073 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12076 * config/i386/i386-builtin.def
12077 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12078 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12079 Require also OPTION_MASK_ISA_AVX512F in addition to
12080 OPTION_MASK_ISA_GFNI.
12081 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12082 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12083 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12084 to OPTION_MASK_ISA_GFNI.
12085 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12086 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12087 OPTION_MASK_ISA_AVX512BW.
12088 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12089 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12090 addition to OPTION_MASK_ISA_GFNI.
12091 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12092 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12093 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12094 to OPTION_MASK_ISA_GFNI.
12095 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12096 a requirement for all ISAs rather than any of them with a few
12098 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12100 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12101 bitmasks to be enabled with 3 exceptions, instead of requiring any
12102 enabled ISA with lots of exceptions.
12103 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12104 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12105 Change avx512bw in isa attribute to avx512f.
12106 * config/i386/sgxintrin.h: Add license boilerplate.
12107 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12108 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12109 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12110 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12112 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12113 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12114 temporarily sse2 rather than sse if not enabled already.
12117 * config/i386/sse.md (VI248_VLBW): Rename to ...
12118 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12119 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12120 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12121 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12122 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12123 mode iterator instead of VI248_VLBW.
12125 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12127 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12128 (record_modified): Skip clobbers; add debug output.
12129 (param_change_prob): Use sreal frequencies.
12131 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12133 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12134 punt for user-aligned variables.
12136 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12138 * tree-chrec.c (chrec_contains_symbols): Return true for
12141 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12144 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12145 of (x|y) == x for BICS pattern.
12147 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12149 PR tree-optimization/83605
12150 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12151 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12154 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12156 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12157 * config/epiphany/rtems.h: New file.
12159 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12160 Uros Bizjak <ubizjak@gmail.com>
12163 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12164 QIreg_operand instead of register_operand predicate.
12165 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12166 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12167 comments instead of -fmitigate[-_]rop.
12169 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12172 * cgraphunit.c (symbol_table::compile): Switch to text_section
12173 before calling assembly_start debug hook.
12174 * run-rtl-passes.c (run_rtl_passes): Likewise.
12177 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12179 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12180 range_int_cst_p rather than !symbolic_range_p before calling
12181 extract_range_from_multiplicative_op_1.
12183 2018-01-04 Jeff Law <law@redhat.com>
12185 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12186 redundant test in assertion.
12188 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12190 * doc/rtl.texi: Document machine_mode wrapper classes.
12192 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12194 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12195 using tree_to_uhwi.
12197 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12199 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12200 the VEC_PERM_EXPR fold to fail.
12202 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12205 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12206 to switched_sections.
12208 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12211 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12212 test for d.testing.
12214 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12217 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12218 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12220 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12223 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12224 is BLKmode and bitpos not zero or mode change is needed.
12226 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12229 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12232 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12235 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12236 instead of MULT rtx. Update all corresponding splitters.
12237 (*saddl_se): Ditto.
12238 (*ssub<modesuffix>): Ditto.
12239 (*ssubl_se): Ditto.
12240 (*cmp_sadd_di): Update split patterns.
12241 (*cmp_sadd_si): Ditto.
12242 (*cmp_sadd_sidi): Ditto.
12243 (*cmp_ssub_di): Ditto.
12244 (*cmp_ssub_si): Ditto.
12245 (*cmp_ssub_sidi): Ditto.
12246 * config/alpha/predicates.md (const23_operand): New predicate.
12247 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12248 Look for ASHIFT, not MULT inner operand.
12249 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12251 2018-01-04 Martin Liska <mliska@suse.cz>
12253 PR gcov-profile/83669
12254 * gcov.c (output_intermediate_file): Add version to intermediate
12256 * doc/gcov.texi: Document new field 'version' in intermediate
12257 file format. Fix location of '-k' option of gcov command.
12259 2018-01-04 Martin Liska <mliska@suse.cz>
12262 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12264 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12266 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12268 2018-01-03 Martin Sebor <msebor@redhat.com>
12270 PR tree-optimization/83655
12271 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12272 checking calls with invalid arguments.
12274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12276 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12277 (vectorizable_mask_load_store): Delete.
12278 (vectorizable_call): Return false for masked loads and stores.
12279 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12280 instead of gimple_assign_rhs1.
12281 (vectorizable_load): Handle IFN_MASK_LOAD.
12282 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12286 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12288 (vectorizable_mask_load_store): ...here.
12289 (vectorizable_load): ...and here.
12291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12293 * tree-vect-stmts.c (vect_build_all_ones_mask)
12294 (vect_build_zero_merge_argument): New functions, split out from...
12295 (vectorizable_load): ...here.
12297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12299 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12301 (vectorizable_mask_load_store): ...here.
12302 (vectorizable_store): ...and here.
12304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12306 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12308 (vectorizable_mask_load_store): ...here.
12310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12312 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12313 (vect_model_store_cost): Take a vec_load_store_type instead of a
12315 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12316 (vect_model_store_cost): Take a vec_load_store_type instead of a
12318 (vectorizable_mask_load_store): Update accordingly.
12319 (vectorizable_store): Likewise.
12320 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12324 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12325 IFN_MASK_LOAD calls here rather than...
12326 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12329 Alan Hayward <alan.hayward@arm.com>
12330 David Sherwood <david.sherwood@arm.com>
12332 * expmed.c (extract_bit_field_1): For vector extracts,
12333 fall back to extract_bit_field_as_subreg if vec_extract
12336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12337 Alan Hayward <alan.hayward@arm.com>
12338 David Sherwood <david.sherwood@arm.com>
12340 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12341 they are variable or constant sized.
12342 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12343 slots for constant-sized data.
12345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12346 Alan Hayward <alan.hayward@arm.com>
12347 David Sherwood <david.sherwood@arm.com>
12349 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12350 handling COND_EXPRs with boolean comparisons, try to find a better
12351 basis for the mask type than the boolean itself.
12353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12355 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12356 is calculated and how it can be overridden.
12357 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12358 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12360 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12364 Alan Hayward <alan.hayward@arm.com>
12365 David Sherwood <david.sherwood@arm.com>
12367 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12368 Remove the mode argument.
12369 (aarch64_simd_valid_immediate): Remove the mode and inverse
12371 * config/aarch64/iterators.md (bitsize): New iterator.
12372 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12373 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12374 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12375 aarch64_simd_valid_immediate.
12376 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12377 (aarch64_reg_or_bic_imm): Likewise.
12378 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12379 with an insn_type enum and msl with a modifier_type enum.
12380 Replace element_width with a scalar_mode. Change the shift
12381 to unsigned int. Add constructors for scalar_float_mode and
12382 scalar_int_mode elements.
12383 (aarch64_vect_float_const_representable_p): Delete.
12384 (aarch64_can_const_movi_rtx_p)
12385 (aarch64_simd_scalar_immediate_valid_for_move)
12386 (aarch64_simd_make_constant): Update call to
12387 aarch64_simd_valid_immediate.
12388 (aarch64_advsimd_valid_immediate_hs): New function.
12389 (aarch64_advsimd_valid_immediate): Likewise.
12390 (aarch64_simd_valid_immediate): Remove mode and inverse
12391 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12392 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12393 and aarch64_float_const_representable_p on the result.
12394 (aarch64_output_simd_mov_immediate): Remove mode argument.
12395 Update call to aarch64_simd_valid_immediate and use of
12396 simd_immediate_info.
12397 (aarch64_output_scalar_simd_mov_immediate): Update call
12400 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12401 Alan Hayward <alan.hayward@arm.com>
12402 David Sherwood <david.sherwood@arm.com>
12404 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12405 (mode_nunits): Likewise CONST_MODE_NUNITS.
12406 * machmode.def (ADJUST_NUNITS): Document.
12407 * genmodes.c (mode_data::need_nunits_adj): New field.
12408 (blank_mode): Update accordingly.
12409 (adj_nunits): New variable.
12410 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12412 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12413 listed in adj_nunits.
12414 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12415 listed in adj_nunits. Don't emit case statements for such modes.
12416 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12417 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12418 nothing if adj_nunits is nonnull.
12419 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12420 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12421 (emit_mode_fbit): Update use of print_maybe_const_decl.
12422 (emit_move_size): Likewise. Treat the array as non-const
12424 (emit_mode_adjustments): Handle adj_nunits.
12426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12428 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12429 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12430 (VECTOR_MODES): Use it.
12431 (make_vector_modes): Take the prefix as an argument.
12433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12434 Alan Hayward <alan.hayward@arm.com>
12435 David Sherwood <david.sherwood@arm.com>
12437 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12438 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12439 for MODE_VECTOR_BOOL.
12440 * machmode.def (VECTOR_BOOL_MODE): Document.
12441 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12442 (make_vector_bool_mode): New function.
12443 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12445 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12446 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12448 * stor-layout.c (int_mode_for_mode): Likewise.
12449 * tree.c (build_vector_type_for_mode): Likewise.
12450 * varasm.c (output_constant_pool_2): Likewise.
12451 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12452 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12453 for MODE_VECTOR_BOOL.
12454 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12455 of mode class checks.
12456 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12457 instead of a list of mode class checks.
12458 (expand_vector_scalar_condition): Likewise.
12459 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12462 Alan Hayward <alan.hayward@arm.com>
12463 David Sherwood <david.sherwood@arm.com>
12465 * machmode.h (mode_size): Change from unsigned short to
12467 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12468 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12469 or if measurement_type is not polynomial.
12470 (fixed_size_mode::includes_p): Check for constant-sized modes.
12471 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12472 return a poly_uint16 rather than an unsigned short.
12473 (emit_mode_size): Change the type of mode_size from unsigned short
12474 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12475 (emit_mode_adjustments): Cope with polynomial vector sizes.
12476 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12478 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12480 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12481 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12482 * caller-save.c (setup_save_areas): Likewise.
12483 (replace_reg_with_saved_mem): Likewise.
12484 * calls.c (emit_library_call_value_1): Likewise.
12485 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12486 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12487 (gen_lowpart_for_combine): Likewise.
12488 * convert.c (convert_to_integer_1): Likewise.
12489 * cse.c (equiv_constant, cse_insn): Likewise.
12490 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12491 (cselib_subst_to_values): Likewise.
12492 * dce.c (word_dce_process_block): Likewise.
12493 * df-problems.c (df_word_lr_mark_ref): Likewise.
12494 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12495 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12496 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12497 (rtl_for_decl_location): Likewise.
12498 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12499 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12500 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12501 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12502 (expand_expr_real_1): Likewise.
12503 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12504 (pad_below): Likewise.
12505 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12506 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12507 * ira.c (get_subreg_tracking_sizes): Likewise.
12508 * ira-build.c (ira_create_allocno_objects): Likewise.
12509 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12510 (ira_sort_regnos_for_alter_reg): Likewise.
12511 * ira-costs.c (record_operand_costs): Likewise.
12512 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12513 (resolve_simple_move): Likewise.
12514 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12515 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12516 (lra_constraints): Likewise.
12517 (CONST_POOL_OK_P): Reject variable-sized modes.
12518 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12519 (add_pseudo_to_slot, lra_spill): Likewise.
12520 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12521 * optabs-query.c (get_best_extraction_insn): Likewise.
12522 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12523 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12524 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12525 * recog.c (offsettable_address_addr_space_p): Likewise.
12526 * regcprop.c (maybe_mode_change): Likewise.
12527 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12528 * regrename.c (build_def_use): Likewise.
12529 * regstat.c (dump_reg_info): Likewise.
12530 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12531 (find_reloads, find_reloads_subreg_address): Likewise.
12532 * reload1.c (eliminate_regs_1): Likewise.
12533 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12534 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12535 (simplify_binary_operation_1, simplify_subreg): Likewise.
12536 * targhooks.c (default_function_arg_padding): Likewise.
12537 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12538 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12539 (verify_gimple_assign_ternary): Likewise.
12540 * tree-inline.c (estimate_move_cost): Likewise.
12541 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12542 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12543 (get_address_cost_ainc): Likewise.
12544 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12545 (vect_supportable_dr_alignment): Likewise.
12546 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12547 (vectorizable_reduction): Likewise.
12548 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12549 (vectorizable_operation, vectorizable_load): Likewise.
12550 * tree.c (build_same_sized_truth_vector_type): Likewise.
12551 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12552 * var-tracking.c (emit_note_insn_var_location): Likewise.
12553 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12554 (ADDR_VEC_ALIGN): Likewise.
12556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12557 Alan Hayward <alan.hayward@arm.com>
12558 David Sherwood <david.sherwood@arm.com>
12560 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12562 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12563 or if measurement_type is polynomial.
12564 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12565 * combine.c (make_extraction): Likewise.
12566 * dse.c (find_shift_sequence): Likewise.
12567 * dwarf2out.c (mem_loc_descriptor): Likewise.
12568 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12569 (extract_bit_field, extract_low_bits): Likewise.
12570 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12571 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12572 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12573 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12574 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12575 * reload.c (find_reloads): Likewise.
12576 * reload1.c (alter_reg): Likewise.
12577 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12578 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12579 * tree-if-conv.c (predicate_mem_writes): Likewise.
12580 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12581 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12582 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12583 * valtrack.c (dead_debug_insert_temp): Likewise.
12584 * varasm.c (mergeable_constant_section): Likewise.
12585 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12588 Alan Hayward <alan.hayward@arm.com>
12589 David Sherwood <david.sherwood@arm.com>
12591 * expr.c (expand_assignment): Cope with polynomial mode sizes
12592 when assigning to a CONCAT.
12594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12595 Alan Hayward <alan.hayward@arm.com>
12596 David Sherwood <david.sherwood@arm.com>
12598 * machmode.h (mode_precision): Change from unsigned short to
12600 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12602 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12603 or if measurement_type is not polynomial.
12604 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12605 in which the mode is already known to be a scalar_int_mode.
12606 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12607 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12609 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12610 for GET_MODE_PRECISION.
12611 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12612 for GET_MODE_PRECISION.
12613 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12615 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12616 (expand_field_assignment, make_extraction): Likewise.
12617 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12618 (get_last_value): Likewise.
12619 * convert.c (convert_to_integer_1): Likewise.
12620 * cse.c (cse_insn): Likewise.
12621 * expr.c (expand_expr_real_1): Likewise.
12622 * lra-constraints.c (simplify_operand_subreg): Likewise.
12623 * optabs-query.c (can_atomic_load_p): Likewise.
12624 * optabs.c (expand_atomic_load): Likewise.
12625 (expand_atomic_store): Likewise.
12626 * ree.c (combine_reaching_defs): Likewise.
12627 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12628 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12629 * tree.h (type_has_mode_precision_p): Likewise.
12630 * ubsan.c (instrument_si_overflow): Likewise.
12632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12633 Alan Hayward <alan.hayward@arm.com>
12634 David Sherwood <david.sherwood@arm.com>
12636 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12637 polynomial numbers of units.
12638 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12639 (valid_vector_subparts_p): New function.
12640 (build_vector_type): Remove temporary shim and take the number
12641 of units as a poly_uint64 rather than an int.
12642 (build_opaque_vector_type): Take the number of units as a
12643 poly_uint64 rather than an int.
12644 * tree.c (build_vector_from_ctor): Handle polynomial
12645 TYPE_VECTOR_SUBPARTS.
12646 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12647 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12648 (build_vector_from_val): If the number of units is variable,
12649 use build_vec_duplicate_cst for constant operands and
12650 VEC_DUPLICATE_EXPR otherwise.
12651 (make_vector_type): Remove temporary is_constant ().
12652 (build_vector_type, build_opaque_vector_type): Take the number of
12653 units as a poly_uint64 rather than an int.
12654 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12656 * cfgexpand.c (expand_debug_expr): Likewise.
12657 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12658 (store_constructor, expand_expr_real_1): Likewise.
12659 (const_scalar_mask_from_tree): Likewise.
12660 * fold-const-call.c (fold_const_reduction): Likewise.
12661 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12662 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12663 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12664 (fold_relational_const): Likewise.
12665 (native_interpret_vector): Likewise. Change the size from an
12666 int to an unsigned int.
12667 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12668 TYPE_VECTOR_SUBPARTS.
12669 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12670 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12671 duplicating a non-constant operand into a variable-length vector.
12672 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12673 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12674 * ipa-icf.c (sem_variable::equals): Likewise.
12675 * match.pd: Likewise.
12676 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12677 * print-tree.c (print_node): Likewise.
12678 * stor-layout.c (layout_type): Likewise.
12679 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12680 * tree-cfg.c (verify_gimple_comparison): Likewise.
12681 (verify_gimple_assign_binary): Likewise.
12682 (verify_gimple_assign_ternary): Likewise.
12683 (verify_gimple_assign_single): Likewise.
12684 * tree-pretty-print.c (dump_generic_node): Likewise.
12685 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12686 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12687 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12688 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12689 (vect_shift_permute_load_chain): Likewise.
12690 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12691 (expand_vector_condition, optimize_vector_constructor): Likewise.
12692 (lower_vec_perm, get_compute_type): Likewise.
12693 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12694 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12695 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12696 (vect_recog_mask_conversion_pattern): Likewise.
12697 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12698 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12699 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12700 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12701 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12702 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12703 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12704 (supportable_widening_operation): Likewise.
12705 (supportable_narrowing_operation): Likewise.
12706 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12708 * varasm.c (output_constant): Likewise.
12710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12711 Alan Hayward <alan.hayward@arm.com>
12712 David Sherwood <david.sherwood@arm.com>
12714 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12715 so that both the length == 3 and length != 3 cases set up their
12716 own permute vectors. Add comments explaining why we know the
12717 number of elements is constant.
12718 (vect_permute_load_chain): Likewise.
12720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12721 Alan Hayward <alan.hayward@arm.com>
12722 David Sherwood <david.sherwood@arm.com>
12724 * machmode.h (mode_nunits): Change from unsigned char to
12726 (ONLY_FIXED_SIZE_MODES): New macro.
12727 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12728 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12729 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12731 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12732 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12733 or if measurement_type is not polynomial.
12734 * genmodes.c (ZERO_COEFFS): New macro.
12735 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12737 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12738 Use ZERO_COEFFS when emitting initializers.
12739 * data-streamer.h (bp_pack_poly_value): New function.
12740 (bp_unpack_poly_value): Likewise.
12741 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12742 for GET_MODE_NUNITS.
12743 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12744 for GET_MODE_NUNITS.
12745 * tree.c (make_vector_type): Remove temporary shim and make
12746 the real function take the number of units as a poly_uint64
12747 rather than an int.
12748 (build_vector_type_for_mode): Handle polynomial nunits.
12749 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12750 * emit-rtl.c (const_vec_series_p_1): Likewise.
12751 (gen_rtx_CONST_VECTOR): Likewise.
12752 * fold-const.c (test_vec_duplicate_folding): Likewise.
12753 * genrecog.c (validate_pattern): Likewise.
12754 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12755 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12756 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12757 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12758 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12759 * rtlanal.c (subreg_get_info): Likewise.
12760 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12761 (vect_grouped_load_supported): Likewise.
12762 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12763 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12764 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12765 (simplify_const_unary_operation, simplify_binary_operation_1)
12766 (simplify_const_binary_operation, simplify_ternary_operation)
12767 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12768 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12769 instead of CONST_VECTOR_NUNITS.
12770 * varasm.c (output_constant_pool_2): Likewise.
12771 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12772 explicit-encoded elements in the XVEC for variable-length vectors.
12774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12776 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12779 Alan Hayward <alan.hayward@arm.com>
12780 David Sherwood <david.sherwood@arm.com>
12782 * coretypes.h (fixed_size_mode): Declare.
12783 (fixed_size_mode_pod): New typedef.
12784 * builtins.h (target_builtins::x_apply_args_mode)
12785 (target_builtins::x_apply_result_mode): Change type to
12786 fixed_size_mode_pod.
12787 * builtins.c (apply_args_size, apply_result_size, result_vector)
12788 (expand_builtin_apply_args_1, expand_builtin_apply)
12789 (expand_builtin_return): Update accordingly.
12791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12793 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12794 * cselib.c (cselib_hash_rtx): Likewise.
12795 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12796 CONST_VECTOR encoding.
12798 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12799 Jeff Law <law@redhat.com>
12802 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12803 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12804 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12805 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12808 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12809 explicitly probe *sp in a noreturn function if there were any callee
12810 register saves or frame pointer is needed.
12812 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12815 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12816 BLKmode for ternary, binary or unary expressions.
12819 * var-tracking.c (delete_vta_debug_insn): New inline function.
12820 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12821 insns from get_insns () to NULL instead of each bb separately.
12822 Use delete_vta_debug_insn. No longer static.
12823 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12824 delete_vta_debug_insns callers.
12825 * rtl.h (delete_vta_debug_insns): Declare.
12826 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12827 instead of variable_tracking_main.
12829 2018-01-03 Martin Sebor <msebor@redhat.com>
12831 PR tree-optimization/83603
12832 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12833 arguments past the endof the argument list in functions declared
12834 without a prototype.
12835 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12836 Avoid checking when arguments are null.
12838 2018-01-03 Martin Sebor <msebor@redhat.com>
12841 * doc/extend.texi (attribute const): Fix a typo.
12842 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12843 issuing -Wsuggest-attribute for void functions.
12845 2018-01-03 Martin Sebor <msebor@redhat.com>
12847 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12848 offset_int::from instead of wide_int::to_shwi.
12849 (maybe_diag_overlap): Remove assertion.
12850 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12851 * gimple-ssa-sprintf.c (format_directive): Same.
12852 (parse_directive): Same.
12853 (sprintf_dom_walker::compute_format_length): Same.
12854 (try_substitute_return_value): Same.
12856 2018-01-03 Jeff Law <law@redhat.com>
12858 PR middle-end/83654
12859 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12860 non-constant residual for zero at runtime and avoid probing in
12861 that case. Reorganize code for trailing problem to mirror handling
12864 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12866 PR tree-optimization/83501
12867 * tree-ssa-strlen.c (get_string_cst): New.
12868 (handle_char_store): Call get_string_cst.
12870 2018-01-03 Martin Liska <mliska@suse.cz>
12872 PR tree-optimization/83593
12873 * tree-ssa-strlen.c: Include tree-cfg.h.
12874 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12875 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12876 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12878 (strlen_dom_walker::before_dom_children): Call
12879 gimple_purge_dead_eh_edges. Dump tranformation with details
12881 (strlen_dom_walker::before_dom_children): Update call by adding
12882 new argument cleanup_eh.
12883 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12885 2018-01-03 Martin Liska <mliska@suse.cz>
12888 * cif-code.def (VARIADIC_THUNK): New enum value.
12889 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12892 2018-01-03 Jan Beulich <jbeulich@suse.com>
12894 * sse.md (mov<mode>_internal): Tighten condition for when to use
12895 vmovdqu<ssescalarsize> for TI and OI modes.
12897 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12899 Update copyright years.
12901 2018-01-03 Martin Liska <mliska@suse.cz>
12904 * ipa-visibility.c (function_and_variable_visibility): Skip
12905 functions with noipa attribure.
12907 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12909 * gcc.c (process_command): Update copyright notice dates.
12910 * gcov-dump.c (print_version): Ditto.
12911 * gcov.c (print_version): Ditto.
12912 * gcov-tool.c (print_version): Ditto.
12913 * gengtype.c (create_file): Ditto.
12914 * doc/cpp.texi: Bump @copying's copyright year.
12915 * doc/cppinternals.texi: Ditto.
12916 * doc/gcc.texi: Ditto.
12917 * doc/gccint.texi: Ditto.
12918 * doc/gcov.texi: Ditto.
12919 * doc/install.texi: Ditto.
12920 * doc/invoke.texi: Ditto.
12922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12924 * vector-builder.h (vector_builder::m_full_nelts): Change from
12925 unsigned int to poly_uint64.
12926 (vector_builder::full_nelts): Update prototype accordingly.
12927 (vector_builder::new_vector): Likewise.
12928 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12929 (vector_builder::operator ==): Likewise.
12930 (vector_builder::finalize): Likewise.
12931 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12932 Take the number of elements as a poly_uint64 rather than an
12934 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12935 from unsigned int to poly_uint64.
12936 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12937 (vec_perm_indices::new_vector): Likewise.
12938 (vec_perm_indices::length): Likewise.
12939 (vec_perm_indices::nelts_per_input): Likewise.
12940 (vec_perm_indices::input_nelts): Likewise.
12941 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12942 number of elements per input as a poly_uint64 rather than an
12943 unsigned int. Use the original encoding for variable-length
12944 vectors, rather than clamping each individual element.
12945 For the second and subsequent elements in each pattern,
12946 clamp the step and base before clamping their sum.
12947 (vec_perm_indices::series_p): Handle polynomial element counts.
12948 (vec_perm_indices::all_in_range_p): Likewise.
12949 (vec_perm_indices_to_tree): Likewise.
12950 (vec_perm_indices_to_rtx): Likewise.
12951 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12952 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12953 (tree_vector_builder::new_binary_operation): Handle polynomial
12954 element counts. Return false if we need to know the number
12955 of elements at compile time.
12956 * fold-const.c (fold_vec_perm): Punt if the number of elements
12957 isn't known at compile time.
12959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12961 * vec-perm-indices.h (vec_perm_builder): Change element type
12962 from HOST_WIDE_INT to poly_int64.
12963 (vec_perm_indices::element_type): Update accordingly.
12964 (vec_perm_indices::clamp): Handle polynomial element_types.
12965 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12966 (vec_perm_indices::all_in_range_p): Likewise.
12967 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12969 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12970 polynomial vec_perm_indices element types.
12971 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12972 * fold-const.c (fold_vec_perm): Likewise.
12973 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12974 * tree-vect-generic.c (lower_vec_perm): Likewise.
12975 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12976 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12977 element type to HOST_WIDE_INT.
12979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12980 Alan Hayward <alan.hayward@arm.com>
12981 David Sherwood <david.sherwood@arm.com>
12983 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12984 rather than an int. Use plus_constant.
12985 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12986 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12989 Alan Hayward <alan.hayward@arm.com>
12990 David Sherwood <david.sherwood@arm.com>
12992 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12993 a HOST_WIDE_INT to a poly_int64.
12995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12996 Alan Hayward <alan.hayward@arm.com>
12997 David Sherwood <david.sherwood@arm.com>
12999 * calls.c (load_register_parameters): Cope with polynomial
13000 mode sizes. Require a constant size for BLKmode parameters
13001 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13002 forces a parameter to be padded at the lsb end in order to
13003 fill a complete number of words, require the parameter size
13004 to be ordered wrt UNITS_PER_WORD.
13006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13007 Alan Hayward <alan.hayward@arm.com>
13008 David Sherwood <david.sherwood@arm.com>
13010 * reload1.c (spill_stack_slot_width): Change element type
13011 from unsigned int to poly_uint64_pod.
13012 (alter_reg): Treat mode sizes as polynomial.
13014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13015 Alan Hayward <alan.hayward@arm.com>
13016 David Sherwood <david.sherwood@arm.com>
13018 * reload.c (complex_word_subreg_p): New function.
13019 (reload_inner_reg_of_subreg, push_reload): Use it.
13021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13022 Alan Hayward <alan.hayward@arm.com>
13023 David Sherwood <david.sherwood@arm.com>
13025 * lra-constraints.c (process_alt_operands): Reject matched
13026 operands whose sizes aren't ordered.
13027 (match_reload): Refer to this check here.
13029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13030 Alan Hayward <alan.hayward@arm.com>
13031 David Sherwood <david.sherwood@arm.com>
13033 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13034 that the mode size is in the set {1, 2, 4, 8, 16}.
13036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13037 Alan Hayward <alan.hayward@arm.com>
13038 David Sherwood <david.sherwood@arm.com>
13040 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13041 Use plus_constant instead of gen_rtx_PLUS.
13043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13044 Alan Hayward <alan.hayward@arm.com>
13045 David Sherwood <david.sherwood@arm.com>
13047 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13048 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13049 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13050 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13051 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13052 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13053 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13054 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13055 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13056 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13058 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13059 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13060 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13061 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13062 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13063 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13064 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13065 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13066 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13067 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13069 * expr.c (emit_move_resolve_push): Treat the input and result
13070 of PUSH_ROUNDING as a poly_int64.
13071 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13072 (emit_push_insn): Likewise.
13073 * lra-eliminations.c (mark_not_eliminable): Likewise.
13074 * recog.c (push_operand): Likewise.
13075 * reload1.c (elimination_effects): Likewise.
13076 * rtlanal.c (nonzero_bits1): Likewise.
13077 * calls.c (store_one_arg): Likewise. Require the padding to be
13078 known at compile time.
13080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13081 Alan Hayward <alan.hayward@arm.com>
13082 David Sherwood <david.sherwood@arm.com>
13084 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13085 Use plus_constant instead of gen_rtx_PLUS.
13087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13088 Alan Hayward <alan.hayward@arm.com>
13089 David Sherwood <david.sherwood@arm.com>
13091 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13092 rather than an int.
13094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13095 Alan Hayward <alan.hayward@arm.com>
13096 David Sherwood <david.sherwood@arm.com>
13098 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13099 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13100 via stack temporaries. Treat the mode size as polynomial too.
13102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13103 Alan Hayward <alan.hayward@arm.com>
13104 David Sherwood <david.sherwood@arm.com>
13106 * expr.c (expand_expr_real_2): When handling conversions involving
13107 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13108 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13109 as a poly_uint64 too.
13111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13112 Alan Hayward <alan.hayward@arm.com>
13113 David Sherwood <david.sherwood@arm.com>
13115 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13118 Alan Hayward <alan.hayward@arm.com>
13119 David Sherwood <david.sherwood@arm.com>
13121 * combine.c (can_change_dest_mode): Handle polynomial
13122 REGMODE_NATURAL_SIZE.
13123 * expmed.c (store_bit_field_1): Likewise.
13124 * expr.c (store_constructor): Likewise.
13125 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13126 and polynomial REGMODE_NATURAL_SIZE.
13127 (gen_lowpart_common): Likewise.
13128 * reginfo.c (record_subregs_of_mode): Likewise.
13129 * rtlanal.c (read_modify_subreg_p): Likewise.
13131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13132 Alan Hayward <alan.hayward@arm.com>
13133 David Sherwood <david.sherwood@arm.com>
13135 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13136 numbers of elements.
13138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13139 Alan Hayward <alan.hayward@arm.com>
13140 David Sherwood <david.sherwood@arm.com>
13142 * match.pd: Cope with polynomial numbers of vector elements.
13144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13145 Alan Hayward <alan.hayward@arm.com>
13146 David Sherwood <david.sherwood@arm.com>
13148 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13149 in a POINTER_PLUS_EXPR.
13151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13152 Alan Hayward <alan.hayward@arm.com>
13153 David Sherwood <david.sherwood@arm.com>
13155 * omp-simd-clone.c (simd_clone_subparts): New function.
13156 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13157 (ipa_simd_modify_function_body): Likewise.
13159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13160 Alan Hayward <alan.hayward@arm.com>
13161 David Sherwood <david.sherwood@arm.com>
13163 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13164 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13165 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13166 (expand_vector_condition, vector_element): Likewise.
13167 (subparts_gt): New function.
13168 (get_compute_type): Use subparts_gt.
13169 (count_type_subparts): Delete.
13170 (expand_vector_operations_1): Use subparts_gt instead of
13171 count_type_subparts.
13173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13174 Alan Hayward <alan.hayward@arm.com>
13175 David Sherwood <david.sherwood@arm.com>
13177 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13178 (vect_compile_time_alias): ...this new function. Do the calculation
13179 on poly_ints rather than trees.
13180 (vect_prune_runtime_alias_test_list): Update call accordingly.
13182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13183 Alan Hayward <alan.hayward@arm.com>
13184 David Sherwood <david.sherwood@arm.com>
13186 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13188 (vect_schedule_slp_instance): Likewise.
13190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13191 Alan Hayward <alan.hayward@arm.com>
13192 David Sherwood <david.sherwood@arm.com>
13194 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13195 constant and extern definitions for variable-length vectors.
13196 (vect_get_constant_vectors): Note that the number of units
13197 is known to be constant.
13199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13200 Alan Hayward <alan.hayward@arm.com>
13201 David Sherwood <david.sherwood@arm.com>
13203 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13204 of units as polynomial. Choose between WIDE and NARROW based
13207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13208 Alan Hayward <alan.hayward@arm.com>
13209 David Sherwood <david.sherwood@arm.com>
13211 * tree-vect-stmts.c (simd_clone_subparts): New function.
13212 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13215 Alan Hayward <alan.hayward@arm.com>
13216 David Sherwood <david.sherwood@arm.com>
13218 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13219 vectors as polynomial. Use build_index_vector for
13220 IFN_GOMP_SIMD_LANE.
13222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13223 Alan Hayward <alan.hayward@arm.com>
13224 David Sherwood <david.sherwood@arm.com>
13226 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13227 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13228 for variable-length vectors.
13229 (vectorizable_mask_load_store): Treat the number of units as
13230 polynomial, asserting that it is constant if the condition has
13231 already been enforced.
13232 (vectorizable_store, vectorizable_load): Likewise.
13234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13235 Alan Hayward <alan.hayward@arm.com>
13236 David Sherwood <david.sherwood@arm.com>
13238 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13239 of units as polynomial. Punt if we can't tell at compile time
13240 which vector contains the final result.
13242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13243 Alan Hayward <alan.hayward@arm.com>
13244 David Sherwood <david.sherwood@arm.com>
13246 * tree-vect-loop.c (vectorizable_induction): Treat the number
13247 of units as polynomial. Punt on SLP inductions. Use an integer
13248 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13249 cast of such a series for variable-length floating-point
13252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13253 Alan Hayward <alan.hayward@arm.com>
13254 David Sherwood <david.sherwood@arm.com>
13256 * tree.h (build_index_vector): Declare.
13257 * tree.c (build_index_vector): New function.
13258 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13259 of units as polynomial, forcibly converting it to a constant if
13260 vectorizable_reduction has already enforced the condition.
13261 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13262 to create a {1,2,3,...} vector.
13263 (vectorizable_reduction): Treat the number of units as polynomial.
13264 Choose vectype_in based on the largest scalar element size rather
13265 than the smallest number of units. Enforce the restrictions
13268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13269 Alan Hayward <alan.hayward@arm.com>
13270 David Sherwood <david.sherwood@arm.com>
13272 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13273 number of units as polynomial.
13275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13276 Alan Hayward <alan.hayward@arm.com>
13277 David Sherwood <david.sherwood@arm.com>
13279 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13280 * target.def (autovectorize_vector_sizes): Return the vector sizes
13281 by pointer, using vector_sizes rather than a bitmask.
13282 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13283 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13284 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13286 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13287 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13288 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13289 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13290 * omp-general.c (omp_max_vf): Likewise.
13291 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13292 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13293 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13294 * tree-vect-slp.c (vect_slp_bb): Likewise.
13295 * doc/tm.texi: Regenerate.
13296 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13298 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13299 the vector size as a poly_uint64 rather than an unsigned int.
13300 (current_vector_size): Change from an unsigned int to a poly_uint64.
13301 (get_vectype_for_scalar_type): Update accordingly.
13302 * tree.h (build_truth_vector_type): Take the size and number of
13303 units as a poly_uint64 rather than an unsigned int.
13304 (build_vector_type): Add a temporary overload that takes
13305 the number of units as a poly_uint64 rather than an unsigned int.
13306 * tree.c (make_vector_type): Likewise.
13307 (build_truth_vector_type): Take the number of units as a poly_uint64
13308 rather than an unsigned int.
13310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13311 Alan Hayward <alan.hayward@arm.com>
13312 David Sherwood <david.sherwood@arm.com>
13314 * target.def (get_mask_mode): Take the number of units and length
13315 as poly_uint64s rather than unsigned ints.
13316 * targhooks.h (default_get_mask_mode): Update accordingly.
13317 * targhooks.c (default_get_mask_mode): Likewise.
13318 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13319 * doc/tm.texi: Regenerate.
13321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13322 Alan Hayward <alan.hayward@arm.com>
13323 David Sherwood <david.sherwood@arm.com>
13325 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13326 * omp-general.c (omp_max_vf): Likewise.
13327 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13328 (expand_omp_simd): Handle polynomial safelen.
13329 * omp-low.c (omplow_simd_context): Add a default constructor.
13330 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13331 (lower_rec_simd_input_clauses): Update accordingly.
13332 (lower_rec_input_clauses): Likewise.
13334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13335 Alan Hayward <alan.hayward@arm.com>
13336 David Sherwood <david.sherwood@arm.com>
13338 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13339 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13340 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13341 (vect_analyze_slp_cost): Likewise.
13342 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13343 (vect_model_load_cost): Likewise.
13345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13346 Alan Hayward <alan.hayward@arm.com>
13347 David Sherwood <david.sherwood@arm.com>
13349 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13350 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13351 from an unsigned int * to a poly_uint64_pod *.
13352 (calculate_unrolling_factor): New function.
13353 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13356 Alan Hayward <alan.hayward@arm.com>
13357 David Sherwood <david.sherwood@arm.com>
13359 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13360 from an unsigned int to a poly_uint64.
13361 (_loop_vec_info::slp_unrolling_factor): Likewise.
13362 (_loop_vec_info::vectorization_factor): Change from an int
13364 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13365 (vect_get_num_vectors): New function.
13366 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13367 (vect_get_num_copies): Use vect_get_num_vectors.
13368 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13369 to an unsigned int *.
13370 (vect_analyze_data_refs): Change min_vf from an int * to a
13372 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13373 than an unsigned HOST_WIDE_INT.
13374 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13375 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13376 to an unsigned int *.
13377 (vect_analyze_data_ref_dependences): Likewise.
13378 (vect_compute_data_ref_alignment): Handle polynomial vf.
13379 (vect_enhance_data_refs_alignment): Likewise.
13380 (vect_prune_runtime_alias_test_list): Likewise.
13381 (vect_shift_permute_load_chain): Likewise.
13382 (vect_supportable_dr_alignment): Likewise.
13383 (dependence_distance_ge_vf): Take the vectorization factor as a
13384 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13385 (vect_analyze_data_refs): Change min_vf from an int * to a
13387 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13388 vfm1 as a poly_uint64 rather than an int. Make the same change
13389 for the returned bound_scalar.
13390 (vect_gen_vector_loop_niters): Handle polynomial vf.
13391 (vect_do_peeling): Likewise. Update call to
13392 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13393 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13395 * tree-vect-loop.c (vect_determine_vectorization_factor)
13396 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13397 (vect_get_known_peeling_cost): Likewise.
13398 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13399 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13400 (vect_transform_loop): Likewise. Use the lowest possible VF when
13401 updating the upper bounds of the loop.
13402 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13403 rather than an int.
13404 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13405 polynomial unroll factors.
13406 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13407 (vect_make_slp_decision): Likewise.
13408 (vect_supported_load_permutation_p): Likewise, and polynomial
13410 (vect_analyze_slp_cost): Handle polynomial vf.
13411 (vect_slp_analyze_node_operations): Likewise.
13412 (vect_slp_analyze_bb_1): Likewise.
13413 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13414 than an unsigned HOST_WIDE_INT.
13415 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13416 (vectorizable_load): Handle polynomial vf.
13417 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13419 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13422 Alan Hayward <alan.hayward@arm.com>
13423 David Sherwood <david.sherwood@arm.com>
13425 * match.pd: Handle bit operations involving three constants
13426 and try to fold one pair.
13428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13430 * tree-vect-loop-manip.c: Include gimple-fold.h.
13431 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13432 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13433 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13434 Add a path that uses a step of VF instead of 1, but disable it
13436 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13437 and niters_no_overflow parameters. Update calls to
13438 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13439 Create a new SSA name if the latter choses to use a ste other
13440 than zero, and return it via niters_vector_mult_vf_var.
13441 * tree-vect-loop.c (vect_transform_loop): Update calls to
13442 vect_do_peeling, vect_gen_vector_loop_niters and
13443 slpeel_make_loop_iterate_ntimes.
13444 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13445 (vect_gen_vector_loop_niters): Update declarations after above changes.
13447 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13449 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13450 128-bit round to integer instructions.
13451 (ceil<mode>2): Likewise.
13452 (btrunc<mode>2): Likewise.
13453 (round<mode>2): Likewise.
13455 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13457 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13458 unaligned VSX load/store on P8/P9.
13459 (expand_block_clear): Allow the use of unaligned VSX
13460 load/store on P8/P9.
13462 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13464 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13466 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13467 swap associated with both a load and a store.
13469 2018-01-02 Andrew Waterman <andrew@sifive.com>
13471 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13472 * config/riscv/riscv.md (clear_cache): Use it.
13474 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13476 * web.c: Remove out-of-date comment.
13478 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13480 * expr.c (fixup_args_size_notes): Check that any existing
13481 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13482 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13483 (emit_single_push_insn): ...here.
13485 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13487 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13488 (const_vector_encoded_nelts): New function.
13489 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13490 (const_vector_int_elt, const_vector_elt): Declare.
13491 * emit-rtl.c (const_vector_int_elt_1): New function.
13492 (const_vector_elt): Likewise.
13493 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13494 of CONST_VECTOR_ELT.
13496 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13498 * expr.c: Include rtx-vector-builder.h.
13499 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13500 directly on the tree encoding.
13501 (const_vector_from_tree): Likewise.
13502 * optabs.c: Include rtx-vector-builder.h.
13503 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13504 sequence of "u" values.
13505 * vec-perm-indices.c: Include rtx-vector-builder.h.
13506 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13507 directly on the vec_perm_indices encoding.
13509 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13511 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13512 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13513 * rtx-vector-builder.h: New file.
13514 * rtx-vector-builder.c: Likewise.
13515 * rtl.h (rtx_def::u2): Add a const_vector field.
13516 (CONST_VECTOR_NPATTERNS): New macro.
13517 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13518 (CONST_VECTOR_DUPLICATE_P): Likewise.
13519 (CONST_VECTOR_STEPPED_P): Likewise.
13520 (CONST_VECTOR_ENCODED_ELT): Likewise.
13521 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13522 (unwrap_const_vec_duplicate): Likewise.
13523 (const_vec_series_p): Check for a non-duplicated vector encoding.
13524 Say that the function only returns true for integer vectors.
13525 * emit-rtl.c: Include rtx-vector-builder.h.
13526 (gen_const_vec_duplicate_1): Delete.
13527 (gen_const_vector): Call gen_const_vec_duplicate instead of
13528 gen_const_vec_duplicate_1.
13529 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13530 (gen_const_vec_duplicate): Use rtx_vector_builder.
13531 (gen_const_vec_series): Likewise.
13532 (gen_rtx_CONST_VECTOR): Likewise.
13533 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13534 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13535 Build a new vector rather than modifying a CONST_VECTOR in-place.
13536 (handle_special_swappables): Update call accordingly.
13537 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13538 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13539 Build a new vector rather than modifying a CONST_VECTOR in-place.
13540 (handle_special_swappables): Update call accordingly.
13542 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13544 * simplify-rtx.c (simplify_const_binary_operation): Use
13545 CONST_VECTOR_ELT instead of XVECEXP.
13547 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13549 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13550 the selector elements to be different from the data elements
13551 if the selector is a VECTOR_CST.
13552 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13553 ssizetype for the selector.
13555 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13557 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13558 before testing each element individually.
13559 * tree-vect-generic.c (lower_vec_perm): Likewise.
13561 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13563 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13564 * selftest-run-tests.c (selftest::run_tests): Call it.
13565 * vector-builder.h (vector_builder::operator ==): New function.
13566 (vector_builder::operator !=): Likewise.
13567 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13568 (vec_perm_indices::all_from_input_p): New function.
13569 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13570 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13571 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13572 instead of reading the VECTOR_CST directly. Detect whether both
13573 vector inputs are the same before constructing the vec_perm_indices,
13574 and update the number of inputs argument accordingly. Use the
13575 utility functions added above. Only construct sel2 if we need to.
13577 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13579 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13580 the broadcast of the low byte.
13581 (expand_mult_highpart): Use an explicit encoding for the permutes.
13582 * optabs-query.c (can_mult_highpart_p): Likewise.
13583 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13584 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13585 (vectorizable_bswap): Likewise.
13586 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13587 explicit encoding for the power-of-2 permutes.
13588 (vect_permute_store_chain): Likewise.
13589 (vect_grouped_load_supported): Likewise.
13590 (vect_permute_load_chain): Likewise.
13592 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13594 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13595 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13596 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13597 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13598 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13599 (vect_gen_perm_mask_any): Likewise.
13601 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13603 * int-vector-builder.h: New file.
13604 * vec-perm-indices.h: Include int-vector-builder.h.
13605 (vec_perm_indices): Redefine as an int_vector_builder.
13606 (auto_vec_perm_indices): Delete.
13607 (vec_perm_builder): Redefine as a stand-alone class.
13608 (vec_perm_indices::vec_perm_indices): New function.
13609 (vec_perm_indices::clamp): Likewise.
13610 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13611 (vec_perm_indices::new_vector): New function.
13612 (vec_perm_indices::new_expanded_vector): Update for new
13613 vec_perm_indices class.
13614 (vec_perm_indices::rotate_inputs): New function.
13615 (vec_perm_indices::all_in_range_p): Operate directly on the
13616 encoded form, without computing elided elements.
13617 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13618 encoding. Update for new vec_perm_indices class.
13619 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13620 the given vec_perm_builder.
13621 (expand_vec_perm_var): Update vec_perm_builder constructor.
13622 (expand_mult_highpart): Use vec_perm_builder instead of
13623 auto_vec_perm_indices.
13624 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13625 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13626 or double series encoding as appropriate.
13627 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13628 vec_perm_indices instead of auto_vec_perm_indices.
13629 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13630 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13631 (vect_permute_store_chain): Likewise.
13632 (vect_grouped_load_supported): Likewise.
13633 (vect_permute_load_chain): Likewise.
13634 (vect_shift_permute_load_chain): Likewise.
13635 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13636 (vect_transform_slp_perm_load): Likewise.
13637 (vect_schedule_slp_instance): Likewise.
13638 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13639 (vectorizable_mask_load_store): Likewise.
13640 (vectorizable_bswap): Likewise.
13641 (vectorizable_store): Likewise.
13642 (vectorizable_load): Likewise.
13643 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13644 vec_perm_indices instead of auto_vec_perm_indices. Use
13645 tree_to_vec_perm_builder to read the vector from a tree.
13646 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13647 vec_perm_builder instead of a vec_perm_indices.
13648 (have_whole_vector_shift): Use vec_perm_builder and
13649 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13650 truncation to calc_vec_perm_mask_for_shift.
13651 (vect_create_epilog_for_reduction): Likewise.
13652 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13653 from auto_vec_perm_indices to vec_perm_indices.
13654 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13655 instead of changing individual elements.
13656 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13657 the vector in d.perm.
13658 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13659 from auto_vec_perm_indices to vec_perm_indices.
13660 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13661 instead of changing individual elements.
13662 (arm_vectorize_vec_perm_const): Use new_vector to install
13663 the vector in d.perm.
13664 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13665 Update vec_perm_builder constructor.
13666 (rs6000_expand_interleave): Likewise.
13667 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13668 (rs6000_expand_interleave): Likewise.
13670 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13672 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13673 to qimode could truncate the indices.
13674 * optabs.c (expand_vec_perm_var): Likewise.
13676 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13678 * Makefile.in (OBJS): Add vec-perm-indices.o.
13679 * vec-perm-indices.h: New file.
13680 * vec-perm-indices.c: Likewise.
13681 * target.h (vec_perm_indices): Replace with a forward class
13683 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13684 * optabs.h: Include vec-perm-indices.h.
13685 (expand_vec_perm): Delete.
13686 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13687 (expand_vec_perm_const): Declare.
13688 * target.def (vec_perm_const_ok): Replace with...
13689 (vec_perm_const): ...this new hook.
13690 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13691 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13692 * doc/tm.texi: Regenerate.
13693 * optabs.def (vec_perm_const): Delete.
13694 * doc/md.texi (vec_perm_const): Likewise.
13695 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13696 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13697 expand_vec_perm for constant permutation vectors. Assert that
13698 the mode of variable permutation vectors is the integer equivalent
13699 of the mode that is being permuted.
13700 * optabs-query.h (selector_fits_mode_p): Declare.
13701 * optabs-query.c: Include vec-perm-indices.h.
13702 (selector_fits_mode_p): New function.
13703 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13704 is defined, instead of checking whether the vec_perm_const_optab
13705 exists. Use targetm.vectorize.vec_perm_const instead of
13706 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13707 fit in the vector mode before using a variable permute.
13708 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13709 vec_perm_indices instead of an rtx.
13710 (expand_vec_perm): Replace with...
13711 (expand_vec_perm_const): ...this new function. Take the selector
13712 as a vec_perm_indices rather than an rtx. Also take the mode of
13713 the selector. Update call to shift_amt_for_vec_perm_mask.
13714 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13715 Use vec_perm_indices::new_expanded_vector to expand the original
13716 selector into bytes. Check whether the indices fit in the vector
13717 mode before using a variable permute.
13718 (expand_vec_perm_var): Make global.
13719 (expand_mult_highpart): Use expand_vec_perm_const.
13720 * fold-const.c: Includes vec-perm-indices.h.
13721 * tree-ssa-forwprop.c: Likewise.
13722 * tree-vect-data-refs.c: Likewise.
13723 * tree-vect-generic.c: Likewise.
13724 * tree-vect-loop.c: Likewise.
13725 * tree-vect-slp.c: Likewise.
13726 * tree-vect-stmts.c: Likewise.
13727 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13729 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13730 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13731 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13732 (aarch64_vectorize_vec_perm_const): ...this new function.
13733 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13734 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13735 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13736 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13737 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13738 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13739 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13741 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13742 check for NEON modes.
13743 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13744 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13745 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13746 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13748 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13749 the old VEC_PERM_CONST conditions.
13750 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13751 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13752 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13753 (ia64_vectorize_vec_perm_const_ok): Merge into...
13754 (ia64_vectorize_vec_perm_const): ...this new function.
13755 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13756 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13757 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13758 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13759 * config/mips/mips.c (mips_expand_vec_perm_const)
13760 (mips_vectorize_vec_perm_const_ok): Merge into...
13761 (mips_vectorize_vec_perm_const): ...this new function.
13762 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13763 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13764 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13765 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13766 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13767 (rs6000_expand_vec_perm_const): Delete.
13768 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13770 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13771 (altivec_expand_vec_perm_const_le): Take each operand individually.
13772 Operate on constant selectors rather than rtxes.
13773 (altivec_expand_vec_perm_const): Likewise. Update call to
13774 altivec_expand_vec_perm_const_le.
13775 (rs6000_expand_vec_perm_const): Delete.
13776 (rs6000_vectorize_vec_perm_const_ok): Delete.
13777 (rs6000_vectorize_vec_perm_const): New function.
13778 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13779 an element count and rtx array.
13780 (rs6000_expand_extract_even): Update call accordingly.
13781 (rs6000_expand_interleave): Likewise.
13782 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13783 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13784 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13785 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13786 (rs6000_expand_vec_perm_const): Delete.
13787 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13788 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13789 (altivec_expand_vec_perm_const_le): Take each operand individually.
13790 Operate on constant selectors rather than rtxes.
13791 (altivec_expand_vec_perm_const): Likewise. Update call to
13792 altivec_expand_vec_perm_const_le.
13793 (rs6000_expand_vec_perm_const): Delete.
13794 (rs6000_vectorize_vec_perm_const_ok): Delete.
13795 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13796 reference to the SPE evmerge intructions.
13797 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13798 an element count and rtx array.
13799 (rs6000_expand_extract_even): Update call accordingly.
13800 (rs6000_expand_interleave): Likewise.
13801 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13802 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13804 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13806 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13808 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13809 vector mode and that that mode matches the mode of the data
13811 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13812 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13813 directly using expand_vec_perm_1 when forcing selectors into
13815 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13819 * optabs-query.h (can_vec_perm_p): Delete.
13820 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13821 * optabs-query.c (can_vec_perm_p): Split into...
13822 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13823 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13824 particular selector is valid.
13825 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13826 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13827 (vect_grouped_load_supported): Likewise.
13828 (vect_shift_permute_load_chain): Likewise.
13829 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13830 (vect_transform_slp_perm_load): Likewise.
13831 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13832 (vectorizable_bswap): Likewise.
13833 (vect_gen_perm_mask_checked): Likewise.
13834 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13835 implementations of variable permutation vectors into account
13836 when deciding which selector to use.
13837 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13838 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13839 with a false third argument.
13840 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13841 to test whether the constant selector is valid and can_vec_perm_var_p
13842 to test whether a variable selector is valid.
13844 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13846 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13847 * optabs-query.c (can_vec_perm_p): Likewise.
13848 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13849 instead of vec_perm_indices.
13850 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13851 (vect_gen_perm_mask_checked): Likewise,
13852 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13853 (vect_gen_perm_mask_checked): Likewise,
13855 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13857 * optabs-query.h (qimode_for_vec_perm): Declare.
13858 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13859 (qimode_for_vec_perm): ...this new function.
13860 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13862 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13864 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13865 does not have a conditional at the top.
13867 2018-01-02 Richard Biener <rguenther@suse.de>
13869 * ipa-inline.c (big_speedup_p): Fix expression.
13871 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13874 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13877 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13881 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13882 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13883 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13884 cond_taken_branch_cost 3->4.
13886 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13888 PR tree-optimization/83581
13889 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13890 TODO_cleanup_cfg if any changes have been made.
13892 PR middle-end/83608
13893 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13894 convert_modes if target mode has the right side, but different mode
13897 PR middle-end/83609
13898 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13899 last argument when extracting from CONCAT. If either from_real or
13900 from_imag is NULL, use expansion through memory. If result is not
13901 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13902 the parts directly to inner mode, if even that fails, use expansion
13905 PR middle-end/83623
13906 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13907 check for bswap in mode rather than HImode and use that in expand_unop
13910 Copyright (C) 2018 Free Software Foundation, Inc.
13912 Copying and distribution of this file, with or without modification,
13913 are permitted in any medium without royalty provided the copyright
13914 notice and this notice are preserved.