1 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
5 (aarch64_needs_frame_chain): New function.
6 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
8 2018-05-23 Sudakshina Das <sudi.das@arm.com>
11 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
12 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
13 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
14 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
15 as true for strict-align.
16 (aarch64_can_inline_p): Perform checks even when callee has no
17 attributes to check for strict alignment.
18 * doc/extend.texi (AArch64 Function Attributes): Document
20 * doc/invoke.texi: (AArch64 Options): Likewise.
22 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
24 PR tree-optimization/85853
25 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
26 the handling of the root of the node to...
27 (vect_slp_analyze_node_operations_1): ...this new function,
28 and run the whole thing with the child nodes' def types
29 set according to their SLP node's def type.
31 2018-05-23 Richard Biener <rguenther@suse.de>
34 * tree-data-ref.c (create_runtime_alias_checks): Defer
35 and ignore overflow warnings.
37 2018-05-23 Yury Gribov <tetra2005@gmail.com>
39 PR tree-optimization/85822
40 * tree-vrp.c (is_masked_range_test): Fix handling of negative
43 2018-05-23 Richard Biener <rguenther@suse.de>
45 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
46 memset constants via native_interpret_expr.
48 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
51 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
53 (cgraph_node::create_alias): Likewise.
54 (cgraph_node::get_availability): Check ifunc_resolver instead
55 of looking up ifunc attribute.
56 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
57 * varasm.c (do_assemble_alias): Likewise.
58 (assemble_alias): Likewise.
59 (default_binds_local_p_3): Likewise.
60 * cgraph.h (cgraph_node): Add ifunc_resolver.
61 (cgraph_node::only_called_directly_or_aliased_p): Return false
63 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
65 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
66 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
67 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
68 instead of looking up ifunc attribute.
70 2018-05-22 Luis Machado <luis.machado@linaro.org>
72 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
74 2018-05-22 Martin Sebor <msebor@redhat.com>
77 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
78 only when expasion succeeds.
79 (expand_builtin_strcmp): Same.
80 (expand_builtin_strncmp): Same.
82 2018-05-22 Martin Sebor <msebor@redhat.com>
84 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
86 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
87 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
89 * config/aarch64/aarch64-ldpstp.md: Replace uses of
90 aarch64_mem_pair_operand with memory_operand and delete operand swapping
92 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
93 Add check for legitimate_address.
94 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
95 (aarch64_swap_ldrstr_operands): New.
96 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
99 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
100 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
102 * config/aarch64/aarch64.md: New patterns to generate stp
104 (store_pair_sw, store_pair_dw): New patterns to generate stp for
105 single words and double words.
106 (load_pair_sw, load_pair_dw): Likewise.
107 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
109 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
111 * config/aarch64/aarch64-ldpstp.md: Modify peephole
112 for different mode ldpstp and add peephole for merged zero stores.
114 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
116 (aarch64_gen_store_pair): Rename calls to match new patterns.
117 (aarch64_gen_load_pair): Rename calls to match new patterns.
118 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
119 (load_pair<DREG:mode><DREG2:mode>): ... This.
120 (store_pair<mode>): Rename to...
121 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
122 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
124 (V_INT_EQUIV): Handle SImode.
125 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
128 2018-05-22 Martin Sebor <msebor@redhat.com>
131 * calls.c (maybe_warn_nonstring_arg): Use string length to set
132 or ajust the presumed bound on an operation to avoid unnecessary
135 2018-05-22 Martin Sebor <msebor@redhat.com>
137 PR tree-optimization/85826
138 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
139 assuming that a DECL necesarily has a constant size.
141 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
144 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
146 2018-05-22 Richard Biener <rguenther@suse.de>
148 PR tree-optimization/85834
149 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
150 non-constant and non-zero memset arguments.
152 2018-05-22 Martin Liska <mliska@suse.cz>
155 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
157 2018-05-22 Richard Biener <rguenther@suse.de>
159 PR tree-optimization/85863
160 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
161 comparisons when vectype is specified.
162 (vectorizable_condition): Do not specify vectype for
163 vect_is_simple_cond when SLP vectorizing.
165 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
168 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
169 define __ibm128 as long double.
170 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
171 as a distinct type when IEEE 128-bit support is enabled.
172 (init_float128_ieee): Fix up conversions between IFmode and IEEE
173 128-bit types to use the correct functions.
174 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
175 convert between 128-bit floating point types that have different
176 modes but the same representation, instead of using gen_lowpart to
178 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
180 (IFKF_reg): New attributes to give the register constraints for
182 (extend<mode>tf2_internal): New insns to mark an explicit
183 conversion between 128-bit floating point types that have a
184 different mode but share the same representation.
186 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
188 PR tree-optimization/85814
189 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
190 a null return from get_strinfo when unsharing the next
191 strinfo in the chain.
193 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
196 * varasm.c (weak_finish): Clean up weak_decls.
198 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
200 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
201 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
202 UNSPEC_UADALP values.
203 * config/aarch64/iterators.md (ABAL): New int iterator.
206 (sur): Add mappings for the above.
207 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
209 (aarch64_<sur>abal<mode>_4): Likewise.
210 (aarch64_<sur>adalp<mode>_3): Likewise.
211 (<sur>sadv16qi): New define_expand.
213 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
215 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
216 (*movdf_internal): Ditto.
217 (*rcpsf2_sse): Ditto.
218 (*rsqrtsf2_sse): Ditto.
219 (*sqrt<mode>2_sse): Ditto.
221 2018-05-21 Tamar Christina <tamar.christina@arm.com>
223 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
225 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
226 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
227 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
228 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
230 * config/aarch64/arm_neon.h: Likewise.
231 * config/aarch64/iterators.md (VQ_I): New.
233 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
235 * config.gcc: Add arc/t-multilib-linux to tmake_file for
237 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
240 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
242 * config/nds32/constraints.md (S): New constraint.
243 * config/nds32/nds32.md (call_internal): Use constraint S.
244 (call_value_internal): Likewise.
245 (sibcall_internal): Likewise.
246 (sibcall_value_internal): Likewise.
248 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
249 Chung-Ju Wu <jasonwucj@gmail.com>
251 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
254 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
255 Chung-Ju Wu <jasonwucj@gmail.com>
257 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
258 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
259 (nds32_rtx_costs_impl): Simplify.
260 (nds32_address_cost_impl): Simplify.
261 (nds32_init_rtx_costs): New function.
262 (nds32_rtx_costs_speed_prefer): Likewise.
263 (nds32_rtx_costs_size_prefer): Likewise.
264 (nds32_address_cost_speed_prefer): Likewise.
265 (nds32_address_cost_speed_fwprop): Likewise.
266 (nds32_address_cost_size_prefer): Likewise.
267 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
268 * config/nds32/nds32.c (nds32_option_override): Use
269 nds32_init_rtx_costs function.
271 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
273 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
274 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
275 (TARGET_PIPELINE_N8): Likewise.
276 (TARGET_PIPELINE_N10): Likewise.
277 (TARGET_PIPELINE_N13): Likewise.
278 (TARGET_PIPELINE_GRAYWOLF): Likewise.
280 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
282 * config/nds32/nds32-fpu.md: Update copyright year.
284 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
286 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
288 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
290 * config/nds32/nds32.c
291 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
292 * config/nds32/nds32.opt (minline-asm-r15): New option.
294 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
296 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
298 * config/nds32/nds32.md (abssi2): New pattern.
300 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
302 * config/i386/i386.md (rex64namesuffix): New mode attribute.
303 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
304 Merge insn pattern from sse_cvtsi2ss<round_name> and
305 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
306 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
307 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
308 using SWI48 mode iterator.
309 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
310 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
311 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
312 pattern from sse_cvttss2si<round_saeonly_name>
313 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
314 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
315 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
316 using SWI48 mode iterator.
317 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
318 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
319 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
320 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
321 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
322 using SWI48 mode iterator.
323 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
324 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
325 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
326 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
327 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
329 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
330 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
331 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
332 pattern from sse_cvttsd2si<round_saeonly_name>
333 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
335 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
337 * config/nds32/nds32-md-auxiliary.c
338 (nds32_valid_smw_lwm_base_p): Refine.
339 (nds32_output_smw_single_word): Refine.
340 (nds32_output_smw_double_word): New.
341 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
343 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
345 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
346 (nds32_output_stack_pop): Refine.
347 (nds32_expand_unaligned_load): Refine.
348 (nds32_expand_unaligned_store): Refine.
350 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
351 Chung-Ju Wu <jasonwucj@gmail.com>
353 * config/nds32/constants.md: Add TP_REGNUM constant.
354 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
355 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
357 * config/nds32/nds32-doubleword.md: Consider flag_pic.
358 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
359 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
360 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
361 and PIC code generation.
362 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
364 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
366 * config/nds32/nds32.md: Support TLS and PIC.
367 * config/nds32/nds32.c: Support TLS and PIC.
368 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
369 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
372 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
374 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
377 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
378 Chung-Ju Wu <jasonwucj@gmail.com>
380 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
381 * config/nds32/nds32-md-auxiliary.c
382 (symbolic_reference_mentioned_p): New.
383 (nds32_legitimize_ict_address): New.
384 (nds32_expand_ict_move): New.
385 (nds32_indirect_call_referenced_p): New.
386 (nds32_symbol_binds_local_p): Delete.
387 (nds32_long_call_p): Modify.
388 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
389 * config/nds32/nds32-protos.h
390 (symbolic_reference_mentioned_p): Declare.
391 (nds32_legitimize_ict_address): Declare.
392 (nds32_expand_ict_move): Declare.
393 (nds32_indirect_call_referenced_p): Declare.
394 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
395 (nds32_relax_group): Use nds32_ict_const_p as condition.
396 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
397 (nds32_asm_file_start): Output ict_model directive in asm code.
398 (nds32_legitimate_address_p): Consider indirect call.
399 (nds32_print_operand): Consider indirect call.
400 (nds32_print_operand_address): Consider indirect call.
401 (nds32_insert_attributes): Handle "indirect_call" attribute.
402 (TARGET_LEGITIMATE_ADDRESS_P): Define.
403 (TARGET_LEGITIMATE_CONSTANT_P): Define.
404 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
405 (TARGET_DELEGITIMIZE_ADDRESS): Define.
406 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
407 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
408 (TARGET_ICT_MODEL_SMALL): Define.
409 (TARGET_ICT_MODEL_LARGE): Define.
410 * config/nds32/nds32.md (movsi): Consider ict model.
411 (call, call_value): Consider ict model.
412 (sibcall, sibcall_value): Consider ict model.
413 * config/nds32/nds32.opt (mict-model): New option.
414 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
417 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
418 Monk Chiang <sh.chiang04@gmail.com>
419 Jim Wilson <jimw@sifive.com>
421 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
422 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
423 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
424 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
425 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
426 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
427 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
428 compute save_libcall_adjustment properly.
429 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
430 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
431 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
432 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
433 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
434 (ABI_SPEC): Handle mabi=ilp32e.
435 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
437 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
438 <-march>: Add rv32e as an example.
440 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
443 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
444 (intra_create_variable_infos): Handle C++ constructors.
446 2018-05-18 Martin Liska <mliska@suse.cz>
448 * passes.def: Remove a redundant pass.
450 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
453 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
455 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
457 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
459 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
461 (armv2, armv2a, armv3, armv3m): Delete architectures.
462 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
463 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
464 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
466 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
467 (*mulsidi3adddi): Likewise.
468 (mulsidi3): Likewise.
469 (*mulsidi3_nov6): Likewise.
470 (umulsidi3): Likewise.
471 (umulsidi3_nov6): Likewise.
472 (umaddsidi4): Likewise.
473 (*umulsidi3adddi): Likewise.
474 (smulsi3_highpart): Likewise.
475 (*smulsi3_highpart_nov6): Likewise.
476 (umulsi3_highpart): Likewise.
477 (*umulsi3_highpart_nov6): Likewise.
478 * config/arm/arm.h (arm_arch3m): Delete.
479 * config/arm/arm.c (arm_arch3m): Delete.
480 (arm_option_override_internal): Update armv3-related comment.
481 (arm_configure_build_target): Delete use of isa_bit_mode32.
482 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
483 (arm_rtx_costs_internal): Delete check of arm_arch3m.
484 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
487 * config/arm/arm-protos.h (arm_arch3m): Delete.
488 * config/arm/arm-tables.opt: Regenerate.
489 * config/arm/arm-tune.md: Likewise.
490 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
491 deleted architectures.
493 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
495 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
496 (armv5t, armv5te): New features.
497 (ARMv5, ARMv5e): Delete fgroups.
498 (ARMv5t, ARMv5te): Adjust for above changes.
500 (armv5, armv5e): Delete arches.
501 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
503 (*call_reg_arm): Likewise.
504 (*call_value_reg_armv5): Likewise.
505 (*call_value_reg_arm): Likewise.
506 (*call_symbol): Likewise.
507 (*call_value_symbol): Likewise.
508 (*sibcall_insn): Likewise.
509 (*sibcall_value_insn): Likewise.
511 (prefetch): Likewise.
512 (define_split and define_peephole2 dependent on arm_arch5):
514 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
516 (TARGET_ARM_QBIT): Likewise.
517 (TARGET_DSP_MULTIPLY): Likewise.
518 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
519 (arm_arch5, arm_arch5e): Delete.
520 (arm_arch5t, arm_arch5te): Declare.
521 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
522 (arm_arch5t): Declare.
523 (arm_option_reconfigure_globals): Update for the above.
524 (arm_options_perform_arch_sanity_checks): Update comment, replace
525 use of arm_arch5 with arm_arch5t.
526 (use_return_insn): Likewise.
527 (arm_emit_call_insn): Likewise.
528 (output_return_instruction): Likewise.
529 (arm_final_prescan_insn): Likewise.
530 (arm_coproc_builtin_available): Likewise.
531 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
532 arm_arch5e with arm_arch5t and arm_arch5te.
533 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
534 (arm_arch5t, arm_arch5te): Declare.
535 * config/arm/arm-tables.opt: Regenerate.
536 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
537 * config/arm/t-multilib: Likewise.
538 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
539 instead of arm_arch5.
540 (*call_reg_thumb1): Likewise.
541 (*call_value_reg_thumb1_v5): Likewise.
542 (*call_value_reg_thumb1): Likewise.
543 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
545 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
547 2018-05-18 Martin Liska <mliska@suse.cz>
549 PR gcov-profile/84846
550 * doc/gcov.texi: Document -t option of gcov tool.
552 2018-05-18 Martin Liska <mliska@suse.cz>
554 PR gcov-profile/84846
555 * gcov.c (print_usage): Add new -t option.
556 (process_args): Handle the option.
557 (generate_results): Use stdout as output when requested by
560 2018-05-18 Martin Liska <mliska@suse.cz>
562 PR gcov-profile/84846
563 * coverage.c (coverage_init): Write PWD to .gcno file.
564 * doc/gcov.texi: Document how working directory is printed.
565 * gcov-dump.c (dump_gcov_file): Print PWD.
566 * gcov.c (output_intermediate_file): Likewise.
567 (read_graph_file): Read PWD string.
568 (output_lines): Print PWD.
570 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
573 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
574 for retval and return false if all args to phi are zero.
576 2018-05-18 Richard Biener <rguenther@suse.de>
578 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
580 (evrp_dom_walker::before_dom_children): Call it.
582 2018-05-18 Richard Biener <rguenther@suse.de>
584 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
585 results when processing array refs with variable index.
587 2018-05-18 Toon Moene <toon@moene.org>
589 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
590 directly after that of -floop-interchange. Indicate that both
591 options are enabled by default when specifying -O3.
593 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
595 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
596 iterator. Delete separate integer-mode vec_set<mode> expander.
597 (aarch64_simd_vec_setv2di): Delete.
598 (vec_setv2di): Delete.
599 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
600 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
601 the "w, r" alternative.
603 2018-05-18 Martin Liska <mliska@suse.cz>
605 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
606 * tree-pass.h (make_pass_lower_switch_O0): New function.
607 * tree-switch-conversion.c (node_has_low_bound): Remove.
608 (node_has_high_bound): Likewise.
609 (node_is_bounded): Likewise.
610 (class pass_lower_switch): Make it a template type and create
612 (pass_lower_switch::execute): Add template argument.
613 (make_pass_lower_switch): New function.
614 (make_pass_lower_switch_O0): New function.
615 (do_jump_if_equal): Remove.
616 (emit_case_nodes): Simplify to just handle all 3 cases and leave
617 all the hard work to tree optimization passes.
619 2018-05-18 Martin Liska <mliska@suse.cz>
621 * dbgcnt.c (limit_low): Renamed from limit.
622 (limit_high): New variable.
623 (dbg_cnt_is_enabled): Check for upper limit.
624 (dbg_cnt): Adjust dumping.
625 (dbg_cnt_set_limit_by_index): Add new argument for high
627 (dbg_cnt_set_limit_by_name): Likewise.
628 (dbg_cnt_process_single_pair): Parse new format.
629 (dbg_cnt_process_opt): Use strtok.
630 (dbg_cnt_list_all_counters): Remove 'value' and add
632 * doc/invoke.texi: Document changes.
634 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
636 * doc/sourcebuild.texi (scalar_all_fma): Document.
637 * tree.def (FMA_EXPR): Delete.
638 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
639 * internal-fn.c (ternary_direct): New macro.
640 (expand_ternary_optab_fn): Likewise.
641 (direct_ternary_optab_supported_p): Likewise.
642 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
643 * builtins.c (fold_builtin_fma): Delete.
644 (fold_builtin_3): Don't call it.
645 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
646 * expr.c (expand_expr_real_2): Likewise.
647 * fold-const.c (operand_equal_p): Likewise.
648 (fold_ternary_loc): Likewise.
649 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
650 * gimple.c (DEFTREECODE): Likewise.
651 * gimplify.c (gimplify_expr): Likewise.
652 * optabs-tree.c (optab_for_tree_code): Likewise.
653 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
654 * tree-eh.c (operation_could_trap_p): Likewise.
655 (stmt_could_throw_1_p): Likewise.
656 * tree-inline.c (estimate_operator_cost): Likewise.
657 * tree-pretty-print.c (dump_generic_node): Likewise.
658 (op_code_prio): Likewise.
659 * tree-ssa-loop-im.c (stmt_cost): Likewise.
660 * tree-ssa-operands.c (get_expr_operands): Likewise.
661 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
662 * fold-const-call.h (fold_fma): Delete.
663 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
664 CFN_FNMA and CFN_FNMS.
666 * genmatch.c (combined_fn): New enum.
667 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
668 (commutative_op): New function.
669 (commutate): Use it. Handle more than 2 operands.
670 (dt_operand::gen_gimple_expr): Use commutative_op.
671 (parser::parse_expr): Allow :c to be used with non-binary
672 operators if the commutative operand is known.
673 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
674 CFN_FMS, CFN_FNMA and CFN_FNMS.
675 (backprop::process_assign_use): Remove FMA_EXPR handling.
676 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
677 (gen_hsa_fma): New function.
678 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
679 IFN_FNMA and IFN_FNMS.
680 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
681 * gimple-fold.h (follow_all_ssa_edges): Declare.
682 * gimple-fold.c (follow_all_ssa_edges): New function.
683 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
684 gimple_build interface and use follow_all_ssa_edges to fold the result.
685 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
686 instead of checking for optabs directly.
687 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
688 rather than FMA_EXPRs.
689 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
690 call to IFN_FMA instead of an FMA_EXPR.
692 2018-05-17 Jim Wilson <jimw@sifive.com>
694 * expr.c (do_tablejump): When converting index to Pmode, if we have a
695 sign extended promoted subreg, and the range does not have the sign bit
696 set, then do a sign extend.
698 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
699 test, check for sign extended subreg and/or constant operands, and
700 do a sign extend in that case.
702 2018-05-17 Steve Ellcey <sellcey@cavium.com>
704 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
705 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
707 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
708 Change logics_shift_reg to logics_shift_imm.
709 (thunderx2t99_fp_loadpair_basic): Delete.
710 (thunderx2t99_fp_storepair_basic): Delete.
711 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
712 (thunderx2t99_asimd_polynomial): Delete.
713 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
714 and neon_fp_mul_d_scalar_q.
715 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
716 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
717 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
718 (thunderx2t99_asimd_lut): Add missing tbl types.
719 (thunderx2t99_asimd_ext): Delete.
720 (thunderx2t99_asimd_load1_1_mult): Delete.
721 (thunderx2t99_asimd_load1_2_mult): Delete.
722 (thunderx2t99_asimd_load1_ldp): New.
723 (thunderx2t99_asimd_load1): New.
724 (thunderx2t99_asimd_load2): Add missing *load2* types.
725 (thunderx2t99_asimd_load3): New.
726 (thunderx2t99_asimd_load4): New.
727 (thunderx2t99_asimd_store1_1_mult): Delete.
728 (thunderx2t99_asimd_store1_2_mult): Delete.
729 (thunderx2t99_asimd_store2_mult): Delete.
730 (thunderx2t99_asimd_store2_onelane): Delete.
731 (thunderx2t99_asimd_store_stp): New.
732 (thunderx2t99_asimd_store1): New.
733 (thunderx2t99_asimd_store2): New.
734 (thunderx2t99_asimd_store3): New.
735 (thunderx2t99_asimd_store4): New.
737 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
739 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
740 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
742 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
743 Segher Boessenkool <segher@kernel.crashing.org>
746 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
749 2018-05-17 Richard Biener <rguenther@suse.de>
751 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
752 for pruning loop and prune defs feeding only already visited PHIs.
754 2018-05-17 Richard Biener <rguenther@suse.de>
756 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
758 2018-05-17 Bin Cheng <bin.cheng@arm.com>
759 Richard Biener <rguenther@suse.de>
761 PR tree-optimization/85793
762 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
763 for VMAT_ELEMENTWISE.
765 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
767 * internal-fn.h (lookup_internal_fn): Declare
768 * internal-fn.c (lookup_internal_fn): New function.
769 * gimple.c (gimple_build_call_from_tree): Handle calls to
771 * gimple-pretty-print.c (dump_gimple_call): Print "." before
772 internal function names.
773 * tree-pretty-print.c (dump_generic_node): Likewise.
774 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
776 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
778 * gimple-fold.h (gimple_build): Make the function forms take
779 combined_fn rather than built_in_function.
780 (gimple_simplify): Likewise.
781 * gimple-match-head.c (gimple_simplify): Likewise.
782 * gimple-fold.c (gimple_build): Likewise.
783 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
784 rather than gimple_build_call_internal.
785 (get_initial_defs_for_reduction): Likewise.
786 (vect_create_epilog_for_reduction): Likewise.
787 (vectorizable_live_operation): Likewise.
789 2018-05-17 Martin Liska <mliska@suse.cz>
791 * gimple-ssa-sprintf.c (format_directive): Do not use
792 space in between 'G_' and '('.
794 2018-05-17 Jakub Jelinek <jakub@redhat.com>
797 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
798 even if the mask is not all ones.
801 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
803 (ix86_gimple_fold_builtin): Likewise.
806 * config/i386/i386.c: Include tree-vector-builder.h.
807 (ix86_vector_shift_count): New function.
808 (ix86_fold_builtin): Fold shift builtins by scalar count.
809 (ix86_gimple_fold_builtin): Likewise.
811 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
812 _mm512_setzero): New intrinsics.
814 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
815 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
817 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
818 code generation for cases where splatting a value is not useful.
819 * simplify-rtx.c (simplify_ternary_operation): Simplify
820 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
821 mode to a vec_concat.
823 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
825 * config.gcc: Support "goldmont-plus".
826 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
828 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
829 PROCESSOR_GOLDMONT_PLUS.
830 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
831 (processor_target_table): Add "goldmont-plus".
832 (PTA_GOLDMONT_PLUS): Define.
833 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
834 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
835 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
836 (fold_builtin_cpu): Add "goldmont-plus".
837 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
838 (ix86_option_override_internal): Add "goldmont-plus".
839 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
840 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
841 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
842 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
844 2018-05-17 Richard Biener <rguenther@suse.de>
846 PR tree-optimization/85757
847 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
848 remove defs that only feed that PHI from further processing.
850 2018-05-16 Jim Wilson <jimw@sifive.com>
852 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
854 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
856 2018-05-16 Mark Wielaard <mark@klomp.org>
858 * dwarf2out.c (count_index_strings): New function.
859 (output_indirect_strings): Call count_index_strings and generate
860 header for dwarf_version >= 5.
862 2018-05-16 Mark Wielaard <mark@klomp.org>
864 * dwarf2out.c (dwarf_FORM): New function.
865 (set_indirect_string): Use dwarf_FORM.
866 (reset_indirect_string): Likewise.
867 (size_of_die): Likewise.
868 (value_format): Likewise.
869 (output_die): Likewise.
870 (add_skeleton_AT_string): Likewise.
871 (output_macinfo_op): Likewise.
872 (index_string): Likewise.
873 (output_index_string_offset): Likewise.
874 (output_index_string): Likewise.
875 (count_index_strings): Likewise.
877 2018-05-16 Carl Love <cel@us.ibm.com>
879 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
880 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
882 2018-05-16 Martin Jambor <mjambor@suse.cz>
884 * ipa-prop.c (ipa_free_all_edge_args): Remove.
885 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
887 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
889 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
890 (fnma<mode>4): Likewise.
891 (fms<mode>4): Likewise.
892 (fnms<mode>4): Likewise.
893 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
894 (aarch64_fnma<mode>4): Likewise.
895 (aarch64_fms<mode>4): Likewise.
896 (aarch64_fnms<mode>4): Likewise.
897 (aarch64_fnmadd<mode>4): Likewise.
899 2018-05-16 Jason Merrill <jason@redhat.com>
901 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
903 2018-05-16 Richard Biener <rguenther@suse.de>
905 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
906 (dump_stmt_cost): Declare.
907 (add_stmt_cost): Dump cost we add.
908 (add_stmt_costs): New function.
909 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
911 (vect_analyze_stmt): Adjust prototype.
912 (vectorizable_condition): Likewise.
913 (vectorizable_live_operation): Likewise.
914 (vectorizable_reduction): Likewise.
915 (vectorizable_induction): Likewise.
916 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
917 cost vector to pass to vectorizable_ and record afterwards.
918 (vect_model_reduction_cost): Take cost vector argument and adjust.
919 (vect_model_induction_cost): Likewise.
920 (vectorizable_reduction): Likewise.
921 (vectorizable_induction): Likewise.
922 (vectorizable_live_operation): Likewise.
923 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
924 SLP_TREE_NUMBER_OF_VEC_STMTS.
925 (vect_analyze_slp_cost_1): Remove.
926 (vect_analyze_slp_cost): Likewise.
927 (vect_slp_analyze_node_operations): Take visited args and
928 a target cost vector. Avoid processing already visited stmt sets.
929 (vect_slp_analyze_operations): Use a local cost vector to gather
930 costs and register those of non-discarded instances.
931 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
932 (vect_schedule_slp_instance): Remove copying of
933 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
935 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
936 adding cost. Record cost entry location.
937 (vect_prologue_cost_for_slp_op): Function to compute cost of
938 a constant or invariant generated for SLP vect in the prologue,
939 split out from vect_analyze_slp_cost_1.
940 (vect_model_simple_cost): Make static. Adjust for SLP costing.
941 (vect_model_promotion_demotion_cost): Likewise.
942 (vect_model_store_cost): Likewise, make static.
943 (vect_model_load_cost): Likewise.
944 (vectorizable_bswap): Add cost vector arg and adjust.
945 (vectorizable_call): Likewise.
946 (vectorizable_simd_clone_call): Likewise.
947 (vectorizable_conversion): Likewise.
948 (vectorizable_assignment): Likewise.
949 (vectorizable_shift): Likewise.
950 (vectorizable_operation): Likewise.
951 (vectorizable_store): Likewise.
952 (vectorizable_load): Likewise.
953 (vectorizable_condition): Likewise.
954 (vectorizable_comparison): Likewise.
955 (can_vectorize_live_stmts): Likewise.
956 (vect_analyze_stmt): Likewise.
957 (vect_transform_stmt): Adjust calls to vectorizable_*.
958 * tree-vectorizer.c: Include gimple-pretty-print.h.
959 (dump_stmt_cost): New function.
961 2018-05-16 Richard Biener <rguenther@suse.de>
963 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
964 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
965 * tree-ssa-dse.c: Include tree-ssa-loop.h.
966 (check_name): New callback.
967 (dse_classify_store): Track cycles via a visited bitmap of PHI
968 defs and simplify handling of in-loop and across loop dead stores
969 and properly fail for loop-variant refs. Handle byte-tracking with
970 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
973 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
975 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
976 (vect_get_mask_type_for_stmt): Likewise.
977 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
979 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
980 to determine the statement's vector type and the vector type that
981 should be used for calculating nunits. Deal with cases in which
982 the type has to be deferred.
983 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
984 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
985 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
986 (vect_determine_vf_for_stmt): New functions, split out from...
987 (vect_determine_vectorization_factor): ...here.
988 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
989 (vect_get_mask_type_for_stmt): New functions, split out from
990 vect_determine_vectorization_factor.
992 2018-05-16 Richard Biener <rguenther@suse.de>
994 * tree-cfg.c (verify_gimple_assign_ternary): Properly
995 verify the [VEC_]COND_EXPR embedded comparison.
997 2018-05-15 Martin Sebor <msebor@redhat.com>
999 PR tree-optimization/85753
1000 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1001 RECORD_TYPE in addition to ARRAY_TYPE.
1003 2018-05-15 Martin Sebor <msebor@redhat.com>
1006 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1008 2018-05-15 Richard Biener <rguenther@suse.de>
1010 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1011 add by_clobber_p one. Change algorithm to collect all defs
1012 representing uses we need to walk and try reducing them to
1013 a single one before failing.
1014 (dse_dom_walker::dse_optimize_stmt): Adjust.
1016 2018-05-13 Mark Wielaard <mark@klomp.org>
1018 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1019 (size_of_loc_descr): Likewise.
1020 (output_loc_operands): Likewise.
1021 (output_loc_operands_raw): Likewise.
1022 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1023 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1024 (hash_loc_operands): Likewise.
1025 (compare_loc_operands): Likewise.
1027 2018-05-14 Mark Wielaard <mark@klomp.org>
1029 * dwarf2out.c (count_index_addrs): New function.
1030 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1032 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1034 PR tree-optimization/83648
1035 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1036 return value as malloc candidate.
1038 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1041 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1042 as true in call to suggest_attribute.
1044 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1046 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1047 -mreadonly-in-sdata.
1049 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1051 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1053 (aarch64_crypto_aesd_fused): Likewise.
1055 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1057 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1058 (movsi_aarch64): Likewise.
1059 (load_pairsi): Likewise.
1060 (load_pairdi): Likewise.
1061 (store_pairsi): Likewise.
1062 (store_pairdi): Likewise.
1063 (load_pairsf): Likewise.
1064 (load_pairdf): Likewise.
1065 (store_pairsf): Likewise.
1066 (store_pairdf): Likewise.
1067 (zero_extend): Likewise.
1068 (trunc): Swap alternatives.
1069 (fcvt_target): Add '?' to prefer w over r.
1071 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1074 * config/i386/i386.md: Disallow non-commutative arithmetics in
1075 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1076 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1077 in the peephole2 before it.
1079 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1081 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1082 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1083 (ix86_handle_option): Handle -mcldemote.
1084 * config.gcc: New header.
1085 * config/i386/cldemoteintrin.h: New file.
1086 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1087 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1089 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1090 OPTION_MASK_ISA_CLDEMOTE.
1091 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1092 (ix86_valid_target_attribute_inner_p): Ditto.
1093 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1094 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1095 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1096 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1097 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1099 * config/i386/i386.opt: Add -mcldemote.
1100 * config/i386/x86intrin.h: New header.
1101 * doc/invoke.texi: Add -mcldemote.
1103 2018-05-14 Richard Biener <rguenther@suse.de>
1105 * doc/match-and-simplify.texi: Adjust :s documentation.
1107 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1109 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1110 intended memcpy size.
1111 (REORDER_45): Likewise.
1113 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1115 * sort.cc: New file.
1116 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1117 * vec.c (qsort_chk): Use gcc_qsort.
1118 * Makefile.in (OBJS-libcommon): Add sort.o.
1119 (build/sort.o): New target. Use it...
1120 (BUILD_RTL): ... here, and...
1121 (build/gencfn-macros): ... here, and...
1122 (build/genmatch): ... here.
1124 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1125 Chung-Ju Wu <jasonwucj@gmail.com>
1127 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1128 * config/nds32/nds32-graywolf.md: New file.
1129 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1130 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1132 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1133 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1134 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1135 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1136 * config/nds32/pipelines.md: Include n15 settings.
1138 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1139 Chung-Ju Wu <jasonwucj@gmail.com>
1141 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1142 * config/nds32/nds32-n13.md: New file.
1143 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1144 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1146 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1147 * config/nds32/nds32.md (pipeline_model): Add n13.
1148 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1149 * config/nds32/pipelines.md: Include n13 settings.
1151 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1152 Chung-Ju Wu <jasonwucj@gmail.com>
1154 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1155 * config/nds32/nds32-n10.md: New file.
1156 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1157 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1159 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1160 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1161 * config/nds32/nds32.md (pipeline_model): Add n10.
1162 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1163 * config/nds32/pipelines.md: Include n10 settings.
1165 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1166 Kito Cheng <kito.cheng@gmail.com>
1167 Chung-Ju Wu <jasonwucj@gmail.com>
1169 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1170 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1171 Add enum values for DSP extension instructions.
1172 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1174 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1175 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1177 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1178 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1179 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1180 * config/nds32/nds32-intrinsic.md: Likewise.
1181 * config/nds32/nds32_intrinsic.h: Likewise.
1182 * config/nds32/nds32-md-auxiliary.c: Likewise.
1183 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1184 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1185 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1186 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1187 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1188 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1189 TYPE_DMAC in switch statement.
1190 * config/nds32/nds32.c: New checking and implementation for DSP
1191 extension instructions.
1192 * config/nds32/nds32.h: Likewise.
1193 * config/nds32/nds32.md: Likewise.
1194 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1195 * config/nds32/predicates.md: Implement new predicates for DSP
1198 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1200 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1201 Reformat alternatives and attributes so it is easier to identify
1202 which constraints/attributes go with which instruction.
1203 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1204 (mov<mode>_softfloat32, FMOVE64): Likewise.
1205 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1206 (mov<mode>_softfloat64, FMOVE64): Likewise.
1208 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1210 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1212 (Basic PowerPC Built-in Functions): The new name of the
1213 subsection previously known as "PowerPC Built-in Functions".
1214 (Basic PowerPC Built-in Functions Available on all Configurations):
1216 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1217 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1218 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1219 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1221 2018-05-11 Martin Jambor <mjambor@suse.cz>
1224 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1227 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1230 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1232 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1234 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1235 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1236 (ix86_handle_option): Handle -mwaitpkg.
1237 * config.gcc: New header.
1238 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1239 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1240 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1242 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1243 OPTION_MASK_ISA_WAITPKG.
1244 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1245 (ix86_option_override_internal): Add PTA_WAITPKG.
1246 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1247 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1248 IX86_BUILTIN_TPAUSE.
1249 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1250 __builtin_ia32_umwait and __builtin_ia32_tpause.
1251 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1252 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1253 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1254 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1255 UNSPECV_TPAUSE): New.
1256 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1257 * config/i386/i386.opt: Add -mwaitpkg.
1258 * config/i386/waitpkgintrin.h: New file.
1259 * config/i386/x86intrin.h: New header.
1260 * doc/invoke.texi: Add -mwaitpkg.
1262 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1265 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1267 (cortex-m0): Use armv6s-m isa.
1268 (cortex-m0plus): Likewise.
1269 (cortex-m1): Likewise.
1270 (cortex-m0.small-multiply): Likewise.
1271 (cortex-m0plus.small-multiply): Likewise.
1272 (cortex-m1.small-multiply): Likewise.
1274 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1275 Jakub Jelinek <jakub@redhat.com>
1277 PR tree-optimization/85692
1278 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1279 source permute as well.
1281 2018-05-11 Martin Liska <mliska@suse.cz>
1284 * doc/extend.texi: Document LLVM style format for no_sanitize
1287 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1289 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1290 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1291 (mode_supports_vsx_dform_quad): Likewise.
1292 (mode_supports_vmx_dform): Move these functions to be next to the
1293 other mode_supports functions.
1294 (mode_supports_dq_form): Likewise.
1295 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1296 mode_supports_dq_form.
1297 (reg_offset_addressing_ok_p): Likewise.
1298 (offsettable_ok_by_alignment): Likewise.
1299 (rs6000_legitimate_offset_address_p): Likewise.
1300 (legitimate_lo_sum_address_p): Likewise.
1301 (rs6000_legitimize_address): Likewise.
1302 (rs6000_legitimize_reload_address): Likewise.
1303 (rs6000_secondary_reload_inner): Likewise.
1304 (rs6000_preferred_reload_class): Likewise.
1305 (rs6000_output_move_128bit): Likewise.
1307 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1309 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1310 Generate SImode target register for null target.
1311 <case IX86_BUILTIN_XGETBV>: Ditto.
1312 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1313 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1315 2018-05-10 Carl Love <cel@us.ibm.com>
1317 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1318 dcbtt and dcbtstt if operands[2] is 0.
1320 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1323 * config/i386/sse.md (usadv64qi): New expander.
1325 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1327 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1328 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1329 -maltivec=be support.
1330 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1331 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1332 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1333 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1334 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1335 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1336 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1337 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1338 altivec_vsumsws): Adjust.
1339 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1340 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1341 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1343 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1344 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1345 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1346 (altivec_lve<VI_char>x): Delete expand.
1347 (*altivec_lve<VI_char>x_internal): Rename to...
1348 (altivec_lve<VI_char>x): ... this.
1349 (altivec_lvxl_<mode>): Delete expand.
1350 (*altivec_lvxl_<mode>_internal): Rename to ...
1351 (altivec_lvxl_<mode>): ... this.
1352 (altivec_stvxl_<mode>): Delete expand.
1353 (*altivec_stvxl_<mode>_internal): Rename to ...
1354 (altivec_stvxl_<mode>): ... this.
1355 (altivec_stve<VI_char>x): Delete expand.
1356 (*altivec_stve<VI_char>x_internal): Rename to ...
1357 (altivec_stve<VI_char>x): ... this.
1358 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1359 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1360 reduc_plus_scal_<mode>): Adjust.
1361 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1363 (rs6000_cpu_cpp_builtins): Adjust.
1364 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1365 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1366 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1367 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1368 -maltivec=be support.
1369 (rs6000_split_vec_extract_var): Adjust.
1370 (rs6000_split_v4si_init): Adjust.
1371 (swap_selector_for_mode): Delete.
1372 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1373 altivec_expand_stvex_be): Delete.
1374 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1375 -maltivec=be support.
1376 (rs6000_gimple_fold_builtin): Ditto.
1377 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1379 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1380 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1381 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1382 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1383 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1384 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1385 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1386 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1387 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1388 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1389 anonymous split): Adjust.
1390 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1391 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1393 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1395 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1396 when --with-gxx-include-dir is also specified.
1397 * configure: Regenerate.
1399 2018-05-09 Jim Wilson <jimw@sifive.com>
1402 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1403 * config/riscv/t-withmultilib: New.
1404 * config/riscv/withmultilib.h: New.
1405 * doc/install.texi: Document RISC-V --with-multilib-list support.
1407 2018-05-09 Richard Biener <rguenther@suse.de>
1409 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1411 (vect_bb_vectorization_profitable_p): Adjust. Compute
1412 actual scalar cost using the cost vector and the add_stmt_cost
1415 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1417 PR rtl-optimization/85645
1418 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1419 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1421 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1423 PR rtl-optimization/85645
1424 * shrink-wrap.c (spread_components): Return a boolean saying if
1425 anything was changed.
1426 (try_shrink_wrapping_separate): Iterate spread_components until
1427 nothing changes anymore.
1429 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1431 PR rtl-optimization/85645
1432 * regrename.c (build_def_use): Also kill the chains that include the
1433 destination of a REG_CFA_REGISTER note.
1435 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1437 PR rtl-optimization/85645
1438 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1439 insn that has a REG_CFA_REGISTER note.
1441 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1443 * cfgexpand.c (expand_clobber): New function.
1444 (expand_gimple_stmt_1): Use it.
1445 * tree-vect-stmts.c (vect_clobber_variable): New function,
1447 (vectorizable_simd_clone_call): ...here.
1448 (vectorizable_store): Emit a clobber either side of an
1449 IFN_STORE_LANES sequence.
1450 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1452 2018-05-09 Tom de Vries <tom@codesourcery.com>
1455 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1456 (define_insn "trap_if_false"): Add exit after trap.
1458 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1460 PR rtl-optimization/85638
1461 * bb-reorder.c: Include common/common-target.h.
1462 (create_forwarder_block): New function extracted from...
1463 (fix_up_crossing_landing_pad): ...here. Rename into...
1464 (dw2_fix_up_crossing_landing_pad): ...this.
1465 (sjlj_fix_up_crossing_landing_pad): New function.
1466 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1467 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1468 from both partitions and exit the loop after one iteration.
1470 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1473 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1475 (Basic PowerPC Built-in Functions): The new name of the
1476 subsection previously known as "PowerPC Built-in Functions".
1477 (Basic PowerPC Built-in Functions Available on all Configurations):
1479 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1481 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1483 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1485 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1488 2018-05-08 Jim Wilson <jimw@sifive.com>
1490 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1491 (LD_EMUL_SUFFIX): New.
1492 (LINK_SPEC): Use it.
1494 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1496 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1498 (Basic PowerPC Built-in Functions): The new name of the
1499 subsection previously known as "PowerPC Built-in Functions".
1500 (Basic PowerPC Built-in Functions Available on all Configurations):
1502 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1504 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1506 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1508 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1511 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1514 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1515 after cmpelim optimization.
1517 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1519 * config.gcc: Support "goldmont".
1520 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1521 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1523 * config/i386/i386.c (m_GOLDMONT): Define.
1524 (processor_target_table): Add "goldmont".
1525 (PTA_GOLDMONT): Define.
1526 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1527 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1528 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1529 (fold_builtin_cpu): Add "goldmont".
1530 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1531 (ix86_option_override_internal): Add "goldmont".
1532 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1533 (processor_type): Add PROCESSOR_GOLDMONT.
1534 * config/i386/i386.md: Add CPU "glm".
1535 * config/i386/glm.md: New file.
1536 * config/i386/x86-tune.def: Add m_GOLDMONT.
1537 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1539 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1542 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1544 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1545 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1546 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1549 * config/i386/i386.c (ix86_fold_builtin): Handle
1550 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1553 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1554 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1556 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1559 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1560 (check_arch): Likewise.
1561 (check_fpu): Return the result rather than printing it.
1562 (end arch): Fix operator precedence.
1563 (end cpu): Likewise.
1564 (END): Print the result from check_fpu.
1566 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1567 Alan Hayward <alan.hayward@arm.com>
1568 David Sherwood <david.sherwood@arm.com>
1570 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1571 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1572 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1573 (*fcmuo<mode>_and): New patterns.
1575 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1577 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1578 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1579 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1580 (cmp_op, sve_imm_con): New code attributes.
1581 (SVE_COND_INT_CMP, imm_con): Delete.
1582 (cmp_op): Remove above unspecs from int attribute.
1583 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1585 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1586 comparison-specific unspecs.
1587 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1588 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1589 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1590 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1591 (*vec_fcm<cmp_op><mode>): Rename to...
1592 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1593 (*vec_fcmuo<mode>): Rename to...
1594 (*fcmuo<mode>): ...this and adjust likewise.
1595 (*pred_fcm<cmp_op><mode>): New pattern.
1596 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1597 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1599 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1601 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1602 (aarch64_emit_sve_predicated_cond): New function.
1603 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1604 (aarch64_emit_unspec_cond_or): Replace with...
1605 (aarch64_emit_sve_or_conds): ...this new function. Use
1606 aarch64_emit_sve_ptrue_op for the individual comparisons and
1607 aarch64_emit_binop to OR them together.
1608 (aarch64_emit_inverted_unspec_cond): Replace with...
1609 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1610 aarch64_emit_sve_ptrue_op for the comparison and
1611 aarch64_emit_unop to invert the result.
1612 (aarch64_expand_sve_vec_cmp_float): Update after the above
1613 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1615 2018-05-07 Nathan Sidwell <nathan@acm.org>
1617 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1618 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1619 (Backwards Compatibility): Likewise.
1621 2018-05-07 Luis Machado <luis.machado@linaro.org>
1625 2018-05-07 Luis Machado <luis.machado@linaro.org>
1627 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1628 <prefetch_dynamic_strides>: New const bool field.
1629 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1630 prefetch_dynamic_strides.
1631 (exynosm1_prefetch_tune): Likewise.
1632 (thunderxt88_prefetch_tune): Likewise.
1633 (thunderx_prefetch_tune): Likewise.
1634 (thunderx2t99_prefetch_tune): Likewise.
1635 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1636 (aarch64_override_options_internal): Update to set
1637 PARAM_PREFETCH_DYNAMIC_STRIDES.
1638 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1639 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1640 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1641 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1642 prefetch-dynamic-strides setting.
1644 2018-05-07 Luis Machado <luis.machado@linaro.org>
1646 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1647 <minimum_stride>: New const int field.
1648 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1649 minimum_stride field.
1650 (exynosm1_prefetch_tune): Likewise.
1651 (thunderxt88_prefetch_tune): Likewise.
1652 (thunderx_prefetch_tune): Likewise.
1653 (thunderx2t99_prefetch_tune): Likewise.
1654 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1655 (aarch64_override_options_internal): Update to set
1656 PARAM_PREFETCH_MINIMUM_STRIDE.
1657 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1658 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1659 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1660 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1661 stride is constant and is below the minimum stride threshold.
1663 2018-05-07 Luis Machado <luis.machado@linaro.org>
1665 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1668 2018-05-07 Luis Machado <luis.machado@linaro.org>
1670 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1671 <prefetch_dynamic_strides>: New const bool field.
1672 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1673 prefetch_dynamic_strides.
1674 (exynosm1_prefetch_tune): Likewise.
1675 (thunderxt88_prefetch_tune): Likewise.
1676 (thunderx_prefetch_tune): Likewise.
1677 (thunderx2t99_prefetch_tune): Likewise.
1678 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1679 (aarch64_override_options_internal): Update to set
1680 PARAM_PREFETCH_DYNAMIC_STRIDES.
1681 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1682 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1683 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1684 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1685 prefetch-dynamic-strides setting.
1687 2018-05-07 Luis Machado <luis.machado@linaro.org>
1689 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1690 <minimum_stride>: New const int field.
1691 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1692 minimum_stride field.
1693 (exynosm1_prefetch_tune): Likewise.
1694 (thunderxt88_prefetch_tune): Likewise.
1695 (thunderx_prefetch_tune): Likewise.
1696 (thunderx2t99_prefetch_tune): Likewise.
1697 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1698 (aarch64_override_options_internal): Update to set
1699 PARAM_PREFETCH_MINIMUM_STRIDE.
1700 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1701 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1702 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1703 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1704 stride is constant and is below the minimum stride threshold.
1706 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1709 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1710 the type is addressable. Don't force op into register if it has
1713 2018-05-05 Roland McGrath <mcgrathr@google.com>
1716 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1717 any section for which we don't know a specific type it should have,
1718 regardless of name. Previously this was done only for the exact
1719 names ".init_array", ".fini_array", and ".preinit_array".
1720 (default_elf_asm_named_section): Add comment about
1721 relationship with default_section_type_flags and SECTION_NOTYPE.
1722 (get_section): Don't consider it a type conflict if one side has
1723 SECTION_NOTYPE and the other doesn't, as long as neither has the
1724 SECTION_BSS et al used in the default_section_type_flags logic.
1726 2018-05-05 Tom de Vries <tom@codesourcery.com>
1729 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1730 (workaround_barsyncs): New function.
1731 (nvptx_reorg): Use workaround_barsyncs.
1732 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1733 (define_expand "nvptx_membar_cta"): New define_expand.
1734 (define_insn "*nvptx_membar_cta"): New insn.
1736 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1738 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1739 To improve optimization opportunities.
1740 * builtin-types.def: The new needed builtin types for the above.
1742 2018-05-04 Richard Biener <rguenther@suse.de>
1744 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1745 * gimple-ssa-store-merging.c
1746 (imm_store_chain_info::output_merged_store): Remove redundant create,
1747 release split_store vector contents on failure.
1748 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1749 scalar stmt vector on cache hit.
1751 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1753 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1755 * config.gcc (powerpc-xilinx-eabi*): Remove.
1756 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1758 (fusion_addis_mem_combo_load): Ditto.
1759 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1761 (rs6000_cpu_cpp_builtins): Ditto.
1762 * config/rs6000/rs6000-linux.c
1763 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1764 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1765 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1767 (rs6000_setup_reg_addr_masks): Ditto.
1768 (rs6000_init_hard_regno_mode_ok): Ditto.
1769 (rs6000_option_override_internal): Ditto.
1770 (legitimate_lo_sum_address_p): Ditto.
1771 (rs6000_legitimize_address): Ditto.
1772 (rs6000_legitimize_reload_address): Ditto.
1773 (rs6000_legitimate_address_p): Ditto.
1774 (abi_v4_pass_in_fpr): Ditto.
1775 (setup_incoming_varargs): Ditto.
1776 (rs6000_gimplify_va_arg): Ditto.
1777 (rs6000_split_multireg_move): Ditto.
1778 (rs6000_savres_strategy): Ditto.
1779 (rs6000_emit_prologue_components): Ditto.
1780 (rs6000_emit_epilogue_components): Ditto.
1781 (rs6000_emit_prologue): Ditto.
1782 (rs6000_emit_epilogue): Ditto.
1783 (rs6000_elf_file_end): Ditto.
1784 (rs6000_function_value): Ditto.
1785 (rs6000_libcall_value): Ditto.
1786 * config/rs6000/rs6000.h: Ditto.
1787 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1788 (TARGET_MINMAX): ... this. New.
1789 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1790 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1791 (*movsi_internal1_single): Delete.
1792 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1793 mfpu=, mxilinx-fpu): Delete.
1794 * config/rs6000/singlefp.h: Delete.
1795 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1796 * config/rs6000/t-rs6000: Ditto.
1797 * config/rs6000/t-xilinx: Delete.
1798 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1799 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1800 (VStype_simple): Delete.
1801 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1802 * config/rs6000/xfpu.h: Delete.
1803 * config/rs6000/xfpu.md: Delete.
1804 * config/rs6000/xilinx.h: Delete.
1805 * config/rs6000/xilinx.opt: Delete.
1806 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1807 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1809 2018-05-04 Tom de Vries <tom@codesourcery.com>
1812 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1815 2018-05-04 Richard Biener <rguenther@suse.de>
1818 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1819 (expand_complex_div_wide): Likewise.
1820 (expand_complex_operations_1): Likewise.
1821 (expand_complex_libcall): Preserve EH info of the original stmt.
1822 (tree_lower_complex): Handle removed blocks.
1823 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1824 on complex multiplication and division libcall builtins.
1826 2018-05-04 Richard Biener <rguenther@suse.de>
1829 * fold-const.c (negate_expr_p): Restrict negation of operand
1830 zero of a division to when we know that can happen without
1832 (fold_negate_expr_1): Likewise.
1834 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1837 * real.h (real_nextafter): Declare.
1838 * real.c (real_nextafter): New function.
1839 * fold-const-call.c (fold_const_nextafter): New function.
1840 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1841 CASE_CFN_NEXTTOWARD.
1842 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1843 even when arg1_mode is different from arg0_mode.
1845 2018-05-03 Nathan Sidwell <nathan@acm.org>
1847 * doc/extend.texi (Deprecated Features): Remove
1849 (Backwards Compatibility): Likewise.
1850 * doc/invoke.texi (C++ Language Options): Likewise.
1851 (C++ Dialect Options): Likewise.
1853 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1856 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1857 _mm512_mask_mullox_epi64): New intrinsics.
1859 2018-05-03 Tom de Vries <tom@codesourcery.com>
1862 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1863 dump files): Add offload-tree.
1865 2018-05-03 Richard Biener <rguenther@suse.de>
1867 PR tree-optimization/85615
1868 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1869 to loops not nested in BBs loop father to avoid creating multi-entry
1872 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1874 PR tree-optimization/70291
1875 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1876 arguments. Change return type to tree. Emit libcall as a new
1877 statement rather than replacing existing one when inplace_p is true.
1878 (expand_complex_multiplication_components): New function.
1879 (expand_complex_multiplication): Expand floating-point complex
1880 multiplication using the above.
1881 (expand_complex_division): Rename inner_type parameter to type.
1882 Update expand_complex_libcall call-site.
1883 (expand_complex_operations_1): Update expand_complex_multiplication
1884 and expand_complex_division call-sites.
1886 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1889 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1890 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1891 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1892 the highest significant bit of the shift count mask is clear. In
1893 check whether and[sq]i3 is needed verify that all significant bits
1894 of the shift count other than the highest are set.
1896 2018-05-02 Tom de Vries <tom@codesourcery.com>
1899 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1900 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1901 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1902 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1903 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1904 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1905 __builtin_goacc_parlevel_size.
1907 2018-05-02 Richard Biener <rguenther@suse.de>
1909 PR tree-optimization/85597
1910 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1911 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1914 2018-05-02 Tom de Vries <tom@codesourcery.com>
1917 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1918 dump files): Add ltrans-tree.
1920 2018-05-02 Tom de Vries <tom@codesourcery.com>
1923 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1924 dump files): Add wpa-ipa.
1926 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1928 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1929 powerpc*-*-linux*paired* target.
1930 * config/rs6000/750cl.h: Delete.
1931 * config/rs6000/paired.h: Delete.
1932 * config/rs6000/paired.md: Delete.
1933 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1935 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1936 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1937 comment. Remove paired float support.
1938 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1939 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1941 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1942 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1944 * config/rs6000/rs6000.c: Remove paired float support.
1945 (paired_expand_vector_init, paired_expand_vector_move,
1946 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1947 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1948 paired_expand_builtin, paired_expand_predicate_builtin,
1949 paired_init_builtins): Delete.
1950 * config/rs6000/rs6000.h: Remove paired float support.
1951 * config/rs6000/rs6000.md: Remove paired float support.
1952 (move_from_CR_ov_bit): Delete.
1953 * config/rs6000/rs6000.opt (mpaired): Delete.
1954 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1955 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1957 2018-05-02 Richard Biener <rguenther@suse.de>
1960 * gimplify.c (gimplify_save_expr): When in SSA form allow
1961 SAVE_EXPRs to compute to SSA vars.
1963 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1966 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1967 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1968 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1969 clobber operands[2], instead use a new pseudo. Formatting fixes.
1971 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1973 PR tree-optimization/85586
1974 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1975 exit early for statements in the same group if the accesses are
1978 2018-05-02 Tom de Vries <tom@codesourcery.com>
1981 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1984 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1986 PR tree-optimization/85143
1987 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1989 2018-05-01 Tom de Vries <tom@codesourcery.com>
1992 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1993 not found" error message.
1995 2018-05-01 Tom de Vries <tom@codesourcery.com>
1998 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1999 * vec.c (test_ordered_remove_if): New function.
2000 (vec_c_tests): Call test_ordered_remove_if.
2001 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2002 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2003 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2004 VEC_ORDERED_REMOVE_IF.
2006 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2008 PR tree-optimization/82665
2009 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2010 pointer subtraction where arguments come from a memchr call.
2012 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2014 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2015 --push-state --as-needed and --pop-state instead of --as-needed and
2016 --no-as-needed if ld supports it.
2017 * configure: Regenerated.
2020 * doc/install.texi2html: Replace _002d with - and _002a with * in
2021 generated html files using sed.
2023 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2026 * gcc-rich-location.c (blank_line_before_p): New function.
2027 (use_new_line): New function.
2028 (gcc_rich_location::add_fixit_insert_formatted): New function.
2029 * gcc-rich-location.h
2030 (gcc_rich_location::add_fixit_insert_formatted): New function.
2032 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2034 * selftest.c (assert_streq): Rename "expected" and "actual" to
2035 "val1" and "val2". Extend NULL-handling to cover both inputs
2036 symmetrically, while still requiring both to be non-NULL for a pass.
2037 * selftest.h (assert_streq): Rename "expected" and "actual" to
2039 (ASSERT_EQ): Likewise.
2040 (ASSERT_EQ_AT): Likewise.
2041 (ASSERT_KNOWN_EQ): Likewise.
2042 (ASSERT_KNOWN_EQ_AT): Likewise.
2043 (ASSERT_NE): Likewise.
2044 (ASSERT_MAYBE_NE): Likewise.
2045 (ASSERT_MAYBE_NE_AT): Likewise.
2046 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2047 the assertion to pass.
2048 (ASSERT_STREQ_AT): Likewise.
2050 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2052 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2053 interaction with -pie.
2055 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2057 * selftest.h: Fix alphabetization of per-source-file selftest
2060 2018-04-30 Jason Merrill <jason@redhat.com>
2062 PR c++/61982 - dead stores to destroyed objects.
2063 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2066 2018-04-30 Jason Merrill <jason@redhat.com>
2068 * tree.c (build_clobber): New.
2069 * tree.h: Declare it.
2070 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2072 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2074 * diagnostic-show-locus.c (layout::layout): Update for
2075 location_get_source_line returning a char_span.
2076 (struct char_span): Move to input.h.
2077 (struct correction): Update for fields in char_span becoming
2079 (struct source_line): Update for location_get_source_line
2080 returning a char_span.
2081 (layout::print_line): Likewise.
2082 * edit-context.c (edited_file::print_content): Likewise.
2083 (edited_file::print_diff_hunk): Likewise.
2084 (edited_file::print_run_of_changed_lines): Likewise.
2085 (edited_file::get_num_lines): Likewise.
2086 (edited_line::edited_line): Likewise.
2087 * final.c (asm_show_source): Likewise.
2088 * input.c (location_get_source_line): Convert return type
2089 from const char * to char_span, losing the final "line_len"
2091 (dump_location_info): Update for the above.
2092 (get_substring_ranges_for_loc): Likewise. Use a char_span
2093 when handling the literal within the line.
2094 (test_reading_source_line): Update for location_get_source_line
2095 returning a char_span.
2096 * input.h (class char_span): Move here from
2097 diagnostic-show-locus.c, converting from a struct to a class.
2098 Make data members private.
2099 (char_span::operator bool): New.
2100 (char_span::length): New.
2101 (char_span::get_buffer): New.
2102 (char_span::operator[]): New.
2103 (char_span::subspan): Make const.
2104 (char_span::xstrdup): New.
2105 (location_get_source_line): Convert return type from const char *
2106 to char_span, losing the final "line_size" param.
2108 2018-04-30 Jan Hubicka <jh@suse.cz>
2110 * lto-wrapper.c (ltrans_priorities): New static var.
2111 (cmp_priority): New.
2112 (run_gcc): Read priorities and if doing parallel build order
2113 the Makefile by them.
2115 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2117 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2119 2018-04-30 Richard Biener <rguenther@suse.de>
2121 * tree-cfg.c (verify_address): Remove base argument, add
2122 flag whether to check TREE_ADDRESSABLE and do that.
2123 (verify_expr): Remove.
2124 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2125 (verify_gimple_assign_single): Likewise.
2126 (verify_gimple_switch): Likewise.
2127 (verify_expr_location_1): Dereference tp once. Add (disabled)
2128 piece from verify_expr.
2129 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2131 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2133 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2135 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2137 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2138 (small_data_pattern): Likewise.
2139 (arc_rewrite_small_data): Likewise.
2140 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2141 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2142 (get_symbol_alignment): New function.
2143 (legitimate_small_data_address_p): Likewise.
2144 (legitimate_scaled_address): Update, call
2145 legitimate_small_data_address_p.
2146 (output_sdata): New static variable.
2147 (arc_print_operand): Update how we handle small data operands.
2148 (arc_print_operand_address): Likewise.
2149 (arc_legitimate_address_p): Update, use
2150 legitimate_small_data_address_p.
2151 (arc_rewrite_small_data_p): Remove.
2152 (arc_rewrite_small_data_1): Likewise.
2153 (arc_rewrite_small_data): Likewise.
2154 (small_data_pattern): Likewise.
2155 (compact_sda_memory_operand): Update to use
2156 legitimate_small_data_address_p and get_symbol_alignment.
2157 (prepare_move_operands): Don't rewite sdata pattern.
2158 (prepare_extend_operands): Remove.
2159 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2161 (zero_extendqisi2): Likewise.
2162 (zero_extendhisi2): Likewise.
2163 (extendqihi2): Likewise.
2164 (extendqisi2): Likewise.
2165 (extendhisi2): Likewise.
2169 * config/arc/constraints.md (Usd): Change it to memory constraint.
2171 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2173 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2174 as source of std instructions.
2175 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2176 allow 6-bit constants as source for store instructions.
2177 (movdi_insn): Update instruction pattern to allow 6-bit constants
2178 as source for store instructions.
2180 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2182 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2184 2018-04-30 Nathan Sidwell <nathan@acm.org>
2185 Sandra Loosemore <sandra@codesourcery.com>
2187 * dumpfile.c (dump_open): Allow '-' for stdout.
2188 * doc/invoke.texi (Developer Options): Document dump filename
2189 determination early. Document stdin/stdout selection.
2191 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2193 Microblaze Target: PIC data text relative
2195 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2196 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2198 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2199 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2200 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2201 New addressing mode for data-text relative position indepenedent code.
2202 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2203 'ADDRESS_SYMBOLIC_TXT_REL'.
2204 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2205 (microblaze_legitimate_pic_operand): Exclude function calls from
2206 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2207 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2209 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2210 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2211 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2212 for 'address + offset'.
2213 (microblaze_expand_prologue): Add new function prologue call for
2215 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2216 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2217 table in case of TARGET_PIC_DATA_TEXT_REL.
2218 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2219 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2220 Add new macros 'UNSPEC_TEXT',
2221 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2222 + exclude function calls from 'UNSPEC_PLT' in case of data text
2224 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2225 new target hook for generating address diff vector tables in case of
2227 * doc/tm.texi : Regenerate.
2228 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2229 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2230 of addr diff vector generation.
2231 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2232 target hook definition.
2233 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2234 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2235 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2236 option for data text relative.
2238 2018-04-30 Richard Biener <rguenther@suse.de>
2240 * tree-chrec.h (evolution_function_is_constant_p): Remove
2242 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2244 2018-04-30 Richard Biener <rguenther@suse.de>
2247 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2249 2018-04-30 Richard Biener <rguenther@suse.de>
2251 PR tree-optimization/28364
2252 PR tree-optimization/85275
2253 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2254 copying first exit test.
2256 2018-04-28 Mark Wielaard <mark@klomp.org>
2258 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2260 (dwarf_AT): Handle DW_AT_addr_base.
2261 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2263 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2266 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2267 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2268 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2269 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2271 2018-04-28 Richard Biener <rguenther@suse.de>
2273 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2274 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2275 to reflect use. Only add interesting stmts.
2277 2018-04-27 Martin Jambor <mjambor@suse.cz>
2280 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2281 the jump function allows for passing through aggregate values.
2283 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2285 * input.h (in_system_header_at): Convert from macro to inline
2287 (from_macro_expansion_at): Likewise.
2288 (from_macro_definition_at): Likewise.
2290 2018-04-27 Jeff Law <law@redhat.com>
2292 * config.gcc: Mark tile* targets as deprecated/obsolete.
2294 2018-04-27 Richard Biener <rguenther@suse.de>
2296 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2299 2018-04-27 Richard Biener <rguenther@suse.de>
2301 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2303 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2305 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2306 with Yd constraint. Set "preferred_for_speed" attribute from
2307 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2309 (*movdi_internal): Ditto.
2310 (movti_interunit splitters): Remove
2311 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2312 (movdi_interunit splitters): Ditto.
2313 * config/i386/constraints.md (Ye): Remove.
2314 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2316 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2319 * config/aarch64/constraints.md (Usg): Limit to 31.
2322 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2324 PR tree-optimization/85529
2325 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2326 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2327 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2328 zero extension or masking of the MSB bit.
2329 (optimize_range_tests): Add FIRST_BB argument, pass it through
2330 to optimize_range_tests_var_bound.
2331 (maybe_optimize_range_tests, reassociate_bb): Adjust
2332 optimize_range_tests callers.
2334 2018-04-26 Richard Biener <rguenther@suse.de>
2335 Jakub Jelinek <jakub@redhat.com>
2337 * cgraph.h (symbol_table): Just declare debug method here.
2338 * symtab.c (symbol_table::debug): Define.
2340 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2342 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2344 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2346 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2347 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2348 (*movdi_internal): Substitute Yi and Yj constraint with x
2349 and Ym and Yn constraint with y constraint. Update "isa"
2350 attribute and set "preferred_for_speed" attribute from
2351 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2352 (*movsi_internal): Ditto.
2353 (*movdf_internal): Ditto.
2354 (*movsf_internal): Ditto.
2355 (*zero_extendsidi2): Ditto.
2356 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2357 (sse2_loadld): Ditto.
2358 (*vec_extract<ssevecmodelower>_0): Ditto.
2359 (*vec_extractv4si_0_zext_sse4): Ditto.
2360 (vec_concatv2di): Ditto.
2361 (*vec_dup<mode>): Ditto.
2362 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2363 * config/i386/constraints.md (Yi): Remove.
2368 2018-04-26 Nathan Sidwell <nathan@acm.org>
2370 * dumpfile.c (dump_open): New.
2371 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2372 (dump_finish): Detect stdio/stderr by value not name.
2374 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2376 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2378 2018-04-26 Tom de Vries <tom@codesourcery.com>
2381 * config/nvptx/nvptx.c (verify_neutering_jumps)
2382 (verify_neutering_labels): New function
2383 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2385 2018-04-26 Tom de Vries <tom@codesourcery.com>
2388 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2389 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2392 2018-04-26 Richard Biener <rguenther@suse.de>
2393 Tom de Vries <tom@codesourcery.com>
2396 * lto-streamer-out.c (output_function): Fixup loops if required to match
2397 discovery done in the reader.
2399 2018-04-26 Richard Biener <rguenther@suse.de>
2401 PR tree-optimization/85116
2402 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2403 have a loop exit from the single latch predecessor. Remove
2404 case of header with just condition.
2405 (ch_base::copy_headers): Exclude infinite loops from any
2407 (pass_ch::execute): Record exits.
2409 2018-04-26 Richard Biener <rguenther@suse.de>
2411 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2412 prologue cost vector and pass it to vect_get_load_cost.
2413 (vect_get_peeling_costs_all_drs): Likewise.
2414 (vect_peeling_hash_get_lowest_cost): Likewise.
2415 (vect_enhance_data_refs_alignment): Likewise.
2417 2018-04-26 Richard Biener <rguenther@suse.de>
2420 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2421 checking of integer<->pointer conversions.
2422 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2423 sign-/zero-extending pointer types.
2424 (expand_omp_for_static_chunk): Likewise.
2426 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2427 Jean Lee <xiaoyur347@gmail.com>
2429 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2430 (TARGET_ASAN_SHADOW_OFFSET): Define.
2431 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2432 true for -fsanitize=address.
2434 2018-04-25 Mark Wielaard <mark@klomp.org>
2436 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2439 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2441 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2442 than "alu", remove explicit "memory" and "imm_disp" attributes.
2443 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2446 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2447 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2448 gen_lowpart_no_emit.
2450 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2453 * config/i386/i386.c (ix86_expand_builtin): Change memory
2454 operand to XI, extend p0 to Pmode.
2455 * config/i386/i386.md: Change unspec volatile and operand
2456 1 mode to XI, change operand 0 mode to P.
2458 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2460 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2461 GET_MODE_MASK before any checking.
2462 (nds32_can_use_bset_p): Likewise.
2463 (nds32_can_use_btgl_p): Likewise.
2465 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2467 * config/nds32/nds32-doubleword.md: New define_split pattern for
2468 illegal register number.
2470 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2472 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2474 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2476 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2478 2018-04-25 Richard Biener <rguenther@suse.de>
2480 * lto-streamer.h (LTO_major_version): Bump to 8.
2482 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2484 * BASE-VER: Set to 9.0.0.
2486 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2488 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2489 in __abskf2 and __powikf2.
2491 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2494 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2495 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2496 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2497 Use the above on operand 2. Reindent.
2498 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2500 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2503 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2505 * config/i386/i386.opt (mcet): Removed.
2506 * doc/install.texi: Remove -mcet documentation.
2507 * doc/invoke.texi: Likewise.
2509 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2512 * doc/install.texi: Remove -mcet from bootstrap-cet.
2514 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2517 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2518 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2522 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2523 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2524 containing a CONST_VECTOR.
2526 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2528 * doc/install.texi: Update newlib dependency for nvptx.
2530 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2533 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2534 instead of INTVAL when shifting x left.
2536 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2538 PR tree-optimization/85478
2539 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2540 vect_grouped_store_supported for single element vectors.
2542 2018-04-24 Richard Biener <rguenther@suse.de>
2545 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2546 load cost increase to the case of non-constant step.
2548 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2551 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2552 destination if any_malformed_asm.
2554 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2557 * expr.c (store_field): In the bitfield case, if the value comes from
2558 a function call and is returned in registers by means of a PARALLEL,
2559 do not change the mode of the temporary unless BLKmode and VOIDmode.
2561 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2563 PR rtl-optimization/85423
2564 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2565 dependencies to debug insns when the previous insn is non-debug.
2567 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2569 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2570 enums into a single definition.
2571 (fls): Fix predicates and printing.
2574 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2576 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2577 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2578 and short u6 immediate.
2579 (check_if_valid_sleep_operand): Remove.
2580 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2582 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2584 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2585 flag_always_save_lp condition.
2586 * config/nds32/nds32.opt (malways-save-lp): New option.
2588 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2590 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2591 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2592 * config/nds32/nds32.h
2593 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2594 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2596 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2598 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2599 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2601 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2602 Chung-Ju Wu <jasonwucj@gmail.com>
2604 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2606 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2607 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2609 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2611 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2613 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2615 * config/nds32/nds32-protos.h (nds32_data_alignment,
2616 nds32_local_alignment): Declare.
2617 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2618 nds32_local_alignment): New functions.
2619 (TARGET_CONSTANT_ALIGNMENT): Define.
2620 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2622 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2624 * config/nds32/nds32.c
2625 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2626 (TARGET_MODES_TIEABLE_P): Likewise.
2628 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2630 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2633 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2634 Chung-Ju Wu <jasonwucj@gmail.com>
2636 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2637 for unaligned access.
2638 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2640 * config/nds32/nds32-intrinsic.md: Likewise.
2641 * config/nds32/nds32_intrinsic.h: Likewise.
2642 * config/nds32/nds32.h (nds32_builtins): Likewise.
2643 * config/nds32/nds32.opt (munaligned-access): New option.
2644 * config/nds32/nds32.c (nds32_asm_file_start): Display
2645 flag_unaligned_access status.
2647 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2649 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2650 -mno-relax is present.
2651 * config/riscv/linux.h (LINK_SPEC): Ditto.
2653 2018-04-20 Martin Sebor <msebor@redhat.com>
2656 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2658 (gimple_fold_builtin_stxcpy_chk): Same.
2659 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2661 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2664 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2665 __powikf2 when long double is IEEE 128-bit.
2667 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2669 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2670 step to make sure stack always aligned.
2672 2018-04-20 Carl Love <cel@us.ibm.com>
2675 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2676 size check for arg0.
2678 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2679 Tom de Vries <tom@codesourcery.com>
2682 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2683 Emit insns for calls too.
2684 (nvptx_find_par): Always look for worker-level predecessor insn.
2685 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2687 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2688 (nvptx_process_pars): Propagate frames for calls.
2690 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2693 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2695 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2696 (ix86_handle_option): Don't handle OPT_mibt.
2697 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2699 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2701 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2702 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2703 (ix86_target_macros): Define __CET__ with flag_cf_protection
2704 for -fcf-protection.
2705 * config/i386/i386.c (isa2_opts): Remove -mibt.
2706 * config/i386/i386.h (TARGET_IBT): Removed.
2707 (TARGET_IBT_P): Likewise.
2708 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2709 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2710 * config/i386/i386.opt (mcet): Update help message.
2713 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2714 -mcet as an alias for -mshstk.
2716 2018-04-20 Richard Biener <rguenther@suse.de>
2719 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2720 complexity by forcing a single use of the multiply operand.
2722 2018-04-20 Martin Jambor <mjambor@suse.cz>
2725 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2726 recursion dependency to only apply to non-clones.
2728 2018-04-20 Martin Jambor <mjambor@suse.cz>
2731 * ipa-cp.c (create_specialized_node): Check that clones of
2732 self-recursive edges exist during IPA-CP.
2734 2018-04-19 Toon Moene <toon@moene.org>
2736 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2739 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2741 PR tree-optimization/85467
2742 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2743 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2744 VECTOR_CST element to type.
2746 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2749 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2750 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2751 (builtin_longjmp): Likewise.
2752 (save_stack_nonlocal): New pattern.
2753 (restore_stack_nonlocal): Likewise.
2755 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2758 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2759 Replace ASM_OUTPUT_LABEL with fprintf.
2761 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2764 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2765 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2766 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2767 define __IBT__ and __SHSTK__ for -fcf-protection.
2768 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2770 (ix86_trampoline_init): Likewise.
2771 (x86_output_mi_thunk): Likewise.
2772 (ix86_notrack_prefixed_insn_p): Likewise.
2773 (ix86_option_override_internal): Don't disallow -fcf-protection.
2774 * config/i386/i386.md (rdssp<mode>): Also enable for
2776 (incssp<mode>): Likewise.
2777 (nop_endbr): Likewise.
2778 * config/i386/i386.opt (mcet): Change help message to built-in
2782 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2783 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2784 enable CET built-in functions.
2786 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2788 * common/config/i386/i386-common.c
2789 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2790 OPTION_MASK_ISA_MOVDIRI_UNSET,
2791 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2792 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2793 * config.gcc (movdirintrin.h): New header.
2794 * config/i386/cpuid.h (bit_MOVDIRI,
2795 bit_MOVDIR64B): New bits.
2796 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2798 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2799 (VOID, PVOID, PCVOID)): New function types.
2800 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2801 __builtin_ia32_directstoreu_u64,
2802 __builtin_ia32_movdir64b): New builtins.
2803 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2804 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2806 (ix86_valid_target_attribute_inner_p): Ditto.
2807 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2808 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2809 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2810 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2811 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2812 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2813 (movdiri<mode>, movdir64b_<mode>): New.
2814 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2815 * config/i386/immintrin.h: Include movdirintrin.h.
2816 * config/i386/movdirintrin.h: New file.
2817 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2819 2018-04-19 Richard Biener <rguenther@suse.de>
2822 * cfg.c (clear_bb_flags): When loop state says we have
2823 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2825 2018-04-19 Richard Biener <rguenther@suse.de>
2827 PR tree-optimization/84737
2828 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2829 copying restrict info.
2830 (vect_setup_realignment): Use it.
2831 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2832 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2833 the first DR to all generated stores.
2834 (vectorizable_load): Likewise for loads.
2836 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2838 PR tree-optimization/85446
2839 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2840 the integral and pointer types to have the same precision.
2842 * doc/install.texi: Document --disable-cet being the default and
2845 2018-04-18 Martin Liska <mliska@suse.cz>
2847 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2850 2018-04-18 Martin Liska <mliska@suse.cz>
2853 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2856 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2857 arguments if they are comparable.
2859 2018-04-18 Martin Liska <mliska@suse.cz>
2862 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2865 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2868 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2871 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2872 ENDBR after calling __morestack.
2874 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2877 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2878 by using gcc_base_ver to generate a gcc_driver_version, and use
2879 it when generating GCC_DRIVER_NAME.
2880 * configure: Regenerate.
2882 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2885 * config.gcc: Obsolete powerpc*-*-*spe*.
2887 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2890 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2891 (stabstr_D): Change type of unum from unsigned int to
2892 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2895 2018-04-17 Jim Wilson <jimw@sifive.com>
2898 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2899 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2900 Set arg_pointer_offset after using pretend_args_size.
2902 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2904 PR rtl-optimization/85431
2905 * dse.c (record_store): Ignore zero width stores.
2908 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2909 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2910 __builtin_stack_restore rather than after it.
2911 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2912 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2913 argument instead of virtual_dynamic_stack_rtx.
2915 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2917 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2919 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2920 Add note to error message to explain internal mapping of overloaded
2921 built-in function name to non-overloaded built-in function name.
2922 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2925 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2928 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2929 where the inputs overlap with the output.
2931 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2933 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2934 (=v, v) alternative and explicit "memory" attribute.
2935 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2936 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2938 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2939 "sselog1" type instead of "sselog".
2940 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2941 "sselog". Remove explicit "memory" attribute.
2942 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2943 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2945 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2946 "sselog1" type instead of "sselog". Remove explicit "memory"
2948 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2949 use "sselog1" type instead of "sselog". Remove explicit "memory"
2951 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2952 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2954 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2955 "sselog1" type instead of "sselog". Remove explicit "memory"
2957 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2958 use "sselog1" type instead of "sselog". Remove explicit "memory"
2962 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2965 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2968 2018-04-17 Martin Jambor <mjambor@suse.cz>
2971 * ipa-cp.c (create_specialized_node): Call
2972 expand_all_artificial_thunks if necessary.
2974 2018-04-17 Martin Liska <mliska@suse.cz>
2977 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2978 in message, remote space in between '_G' and '('.
2980 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2983 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2984 avx512f_vmcmp<mode>3<round_saeonly_name>,
2985 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2986 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2987 avx512f_rndscale<mode><round_saeonly_name>,
2988 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2989 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2990 Use %<iptr>2 instead of %2 for -masm=intel.
2991 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2992 avx512f_vcvttss2usi<round_saeonly_name>,
2993 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2995 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2996 avx512f_vcvttsd2usi<round_saeonly_name>,
2997 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2998 Use %q1 instead of %1 for -masm=intel.
2999 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3000 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3001 of %3 for -masm=intel.
3002 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3004 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3006 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3008 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3009 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3011 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3013 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3014 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3015 %g1 and one with %0 and %1.
3016 (avx512er_vmrcp28<mode><round_saeonly_name>,
3017 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3019 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3020 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3021 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3022 of %0 and %{%4%} for -masm=intel.
3023 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3024 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3025 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3026 order of %0 and %{%5%}%{z%} for -masm=intel.
3028 2018-04-17 Jan Hubicka <jh@suse.cz>
3031 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3033 2018-04-17 Martin Liska <mliska@suse.cz>
3036 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3037 for target_clone error message. Make default implementation
3038 clone to be a local declaration.
3039 (separate_attrs): Add new argument and check for an empty
3041 (expand_target_clones): Handle it.
3042 (ipa_target_clone): Make redirection just for target_clones
3045 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3046 Tom de Vries <tom@codesourcery.com>
3049 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3050 tiled basic blocks without omp continue statements.
3052 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3055 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3056 vec_extract expression as having side effects to make sure it gets
3059 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3062 * config/i386/i386.c (get_builtin_code_for_version): Check
3065 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3068 * gcc/config.gcc: Support "skylake".
3069 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3071 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3072 (processor_target_table): Add "skylake".
3073 (ix86_option_override_internal): Add "skylake".
3074 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3075 PROCESSOR_CANNONLAKE.
3076 (get_builtin_code_for_version): Fix priority for
3077 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3078 PROCESSOR_SKYLAKE-AVX512.
3079 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3080 (processor_type): Add PROCESSOR_SKYLAKE.
3082 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3083 Jason Merrill <jason@redhat.com>
3086 * convert.c (convert_to_integer_1): Use direct recursion for
3087 enumeral types and types with a precision less than the number
3088 of bits in their mode.
3090 2018-04-16 Julia Koval <julia.koval@intel.com>
3093 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3094 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3096 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3099 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3100 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3101 and -mno-direct-move.
3103 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3106 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3107 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3108 Ensure negative shifts result in {0}.
3110 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3112 PR rtl-optimization/79916
3113 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3114 regs (if any) to define how to gnerate SD moves when LRA is in
3117 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3119 PR rtl-optimization/85393
3120 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3121 * except.c (expand_dw2_landing_pad_for_region): Make static.
3122 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3123 a label and unconditional jump to old_bb, rather than
3124 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3127 PR rtl-optimization/85376
3128 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3129 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3130 instead of a specific value.
3132 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3133 Bin Cheng <bin.cheng@arm.com>
3135 PR tree-optimization/82965
3136 PR tree-optimization/83991
3137 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3138 by_profile_only parameter.
3139 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3140 information if the loop was predicted to iterate too many times.
3141 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3143 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3146 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3149 2018-04-13 Martin Liska <mliska@suse.cz>
3150 Jakub Jelinek <jakub@redhat.com>
3153 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3154 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3155 * builtins.c (expand_builtin_memory_copy_args): Use
3156 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3157 handle dest_addr == pc_rtx.
3159 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3162 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3163 asked to not generate direct moves.
3164 (fix_trunc<mode>si2_stfiwx): Similar.
3165 (fix_trunc<mode>si2_internal): Similar.
3167 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3170 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3171 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3172 lookup if dest in some wider mode is known to be const0_rtx and
3173 if so, record permanent equivalence for it to be ZERO_EXTEND of
3174 the narrower mode destination.
3176 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3178 * lto-streamer-out.c (output_function): Revert 259346.
3179 * omp-expand.c (expand_oacc_for): Likewise.
3181 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3183 PR rtl-optimization/85354
3184 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3185 * sel-sched.c (sel_global_init): ... here.
3187 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3190 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3191 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3192 mode for PE-COFF targets.
3193 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3194 (i386_pe_asm_lto_end): Likewise.
3195 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3196 (TARGET_ASM_LTO_END): Likewise.
3197 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3198 (i386_pe_asm_lto_start): New function.
3199 (i386_pe_asm_lto_end): Likewise.
3201 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3202 Richard Biener <rguenther@suse.de>
3205 * lto-streamer-out.c (output_function): Fix CFG loop state before
3207 * omp-expand.c (expand_oacc_for): Handle calls to internal
3208 functions like regular functions.
3210 2018-04-12 Richard Biener <rguenther@suse.de>
3213 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3214 for the early LTO debug to properly generate references to it
3215 during DIE emission. Do not re-use that for the skeleton for
3217 (dwarf2out_early_finish): Likewise.
3219 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3222 * config/i386/sse.md
3223 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3224 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3225 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3226 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3227 and output is a reg, avoid creating invalid lowpart subreg, but
3228 instead split into a 512-bit move. Don't split if not AVX512VL,
3229 input is xmm16+ reg and output is a mem.
3230 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3231 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3232 xmm16+ reg and output is a mem.
3234 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3236 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3237 also for flag_dwarf2_cfi_asm.
3239 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3241 PR rtl-optimization/85342
3242 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3243 a bool scalar var inside of the loop instead. Don't try to update
3244 recog_data.operand after failed apply_change_group.
3246 2018-04-12 Tom de Vries <tom@codesourcery.com>
3249 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3250 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3251 array with flexible array member as array without given dimension.
3252 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3253 argument for undefined param to true.
3255 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3258 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3259 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3260 from PowerPC section.
3261 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3262 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3263 help text that is too long.
3264 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3265 help text that is too long.
3266 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3267 help text that is too long.
3269 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3271 * config/alpha/alpha.md (stack_probe_internal): Rename
3272 from "probe_stack". Update all callers.
3274 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3276 PR rtl-optimization/84566
3277 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3278 sched_macro_fuse_insns.
3280 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3283 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3284 (compute_block_dependences): ... from here.
3286 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3288 PR tree-optimization/85331
3289 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3290 from int to HOST_WIDE_INT.
3292 2018-04-11 Martin Jambor <mjambor@suse.cz>
3295 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3296 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3297 not the same as the source val.
3298 (cgraph_edge_brings_value_p): New parameter.
3299 (gather_edges_for_value): Pass destination value to
3300 cgraph_edge_brings_value_p.
3301 (perhaps_add_new_callers): Likewise.
3302 (get_info_about_necessary_edges): Likewise and exclude values brought
3303 only by self-recursive edges.
3304 (create_specialized_node): Redirect only clones of self-calling edges.
3305 (+self_recursive_pass_through_p): New function.
3306 (find_more_scalar_values_for_callers_subset): Use it.
3307 (find_aggregate_values_for_callers_subset): Likewise.
3308 (known_aggs_to_agg_replacement_list): Removed.
3309 (decide_whether_version_node): Re-calculate known constants for all
3310 remaining context clones.
3312 2018-04-11 Richard Biener <rguenther@suse.de>
3315 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3316 from early DWARF output.
3317 (dwarf2out_early_finish): Output line info unconditionally into
3318 early DWARF and add reference to it.
3320 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3323 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3324 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3325 other than V2DFmode using iptr mode attribute.
3326 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3328 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3330 PR rtl-optimization/84659
3331 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3333 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3336 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3338 (output_loc_list): Pass address of a dummy size variable even in the
3339 locview handling loop.
3340 (index_location_lists): Add comment on why skip_loc_list_entry can't
3343 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3346 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3349 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3352 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3353 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3354 and -mstring-compare-inline-limit.
3356 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3359 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3360 for stack clash protection in a register whenever we need it to be in
3363 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3365 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3366 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3368 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3371 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3373 (mlong-double-): Ditto.
3374 * config/rs6000/sysv4.opt (msdata=): Ditto.
3375 (mtls-size=): Ditto.
3377 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3379 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3380 erroneous entries for
3381 "vector int vec_ldl (int, long int *)", and
3382 "vector unsigned int vec_ldl (int, unsigned long int *)".
3383 Add comments and entries for
3384 "vector bool char vec_ldl (int, bool char *)",
3385 "vector bool short vec_ldl (int, bool short *)",
3386 "vector bool int vec_ldl (int, bool int *)",
3387 "vector bool long long vec_ldl (int, bool long long *)",
3388 "vector pixel vec_ldl (int, pixel *)",
3389 "vector long long vec_ldl (int, long long *)",
3390 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3391 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3392 type tree bool_long_long_type_node and correct definition of
3393 bool_V2DI_type_node to make reference to this new type tree.
3394 (rs6000_mangle_type): Replace erroneous reference to
3395 bool_long_type_node with bool_long_long_type_node.
3396 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3397 comments to emphasize sign distinctions for char and int types and
3398 replace RS6000_BTI_bool_long constant with
3399 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3400 use of RS6000_BTI_pixel.
3401 (bool_long_type_node): Remove this macro definition.
3402 (bool_long_long_type_node): New macro definition
3404 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3406 PR rtl-optimization/85300
3407 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3408 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3409 simplify_unary_operation fails.
3411 2018-04-10 Martin Liska <mliska@suse.cz>
3413 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3414 cgraph_edge and ipa_ref.
3416 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3420 * config/i386/sse.md
3421 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3422 computation of the VEC_MERGE selector from mask.
3423 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3424 Fix decoding of the VEC_MERGE selector into mask.
3426 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3428 PR tree-optimization/85286
3429 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3431 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3433 * final.c (final_1): Set insn_last_address as well as
3434 insn_current_address.
3436 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3439 * explow.c (emit_stack_probe): Call validize_mem on memory location
3440 before passing it to gen_probe_stack. Create address operand and
3441 legitimize it for the probe_stack_address case.
3443 2018-04-09 Jan Hubicka <jh@suse.cz>
3446 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3447 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3448 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3450 (free_lang_data): Rebuild type inheritance graph.
3452 2018-04-09 Martin Sebor <msebor@redhat.com>
3454 * invoke.texi (-finline-small-functions): Mention other optimization
3456 (-findirect-inlining, -fpartial-inlining): Same.
3457 (-finline-functions-called-once): Same.
3458 (-freorder-blocks-and-partition): Same.
3460 2018-04-09 Jan Hubicka <jh@suse.cz>
3463 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3464 jumps; choose last target that matches the criteria (i.e.
3465 no partition changes for non-crossing jumps).
3466 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3467 support for redirecting crossing jumps to non-crossing.
3469 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3471 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3472 also for naked functions.
3474 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3476 * config/arc/arc.md (add_shift): New pattern.
3477 (add_shift2): Likewise.
3478 (sub_shift): Likewise.
3479 (sub_shift_cmp0_noout): Likewise.
3480 (compare_si_ashiftsi): Likewise.
3481 (xbfu_cmp0_noout): New combine pattern.
3482 (xbfu_cmp0"): Likewise.
3483 (movsi_set_cc_insn): Place the predicable variant first.
3484 (commutative_binary_cmp0_noout): Remove clobber.
3485 (commutative_binary_cmp0): New pattern.
3486 (noncommutative_binary_cmp0): Likewise.
3487 (noncommutative_binary_cmp0_noout): Likewise.
3488 (noncommutative_binary_comparison_result_used): Removed.
3489 (rsub_cmp0): New pattern.
3490 (rsub_cmp0_noout): Likewise.
3491 (extzvsi): Changed, keep only meaningful variants.
3492 (SQH, SEZ): New iterators.
3493 (SQH_postfix): New mode attribute.
3494 (SEZ_prefix): New code attribute.
3495 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3496 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3497 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3499 (noncommutative_operator): Check the availability of barrel
3502 2018-04-09 Richard Biener <rguenther@suse.de>
3504 PR tree-optimization/85284
3505 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3506 Only use the niter constraining form of simple_iv when the exit
3509 2018-04-09 Tom de Vries <tom@codesourcery.com>
3512 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3513 (define_expand "*memory_barrier"): New define_expand.
3514 (define_insn "memory_barrier"): New insn.
3516 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3518 PR rtl-optimization/80463
3519 PR rtl-optimization/83972
3520 PR rtl-optimization/83480
3522 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3523 correct producer for the insn.
3524 (tidy_control_flow): Fixup seqnos in case of debug insns.
3526 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3528 PR rtl-optimization/83913
3530 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3531 different sched-times when merging exprs.
3533 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3535 PR rtl-optimization/83962
3537 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3538 tidy_fallthru_edge and tidy_control_flow.
3540 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3542 PR rtl-optimization/83530
3544 * sel-sched.c (force_next_insn): New global variable.
3545 (remove_insn_for_debug): When force_next_insn is true, also leave only
3546 next insn in the ready list.
3547 (sel_sched_region): When the region wasn't scheduled, make another pass
3548 over it with force_next_insn set to 1.
3550 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3552 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3554 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3555 for interrupt control.
3556 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3557 functions for interrupt control.
3558 * config/nds32/nds32-intrinsic.md: Likewise.
3559 * config/nds32/nds32_intrinsic.h: Likewise.
3560 * config/nds32/nds32.h (nds32_builtins): Likewise.
3562 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3564 * config/nds32/nds32.c (nds32_init_machine_status,
3565 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3566 strict_aligned_p field.
3567 (nds32_expand_to_rtl_hook): New function.
3568 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3569 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3571 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3572 Chung-Ju Wu <jasonwucj@gmail.com>
3574 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3575 * config/nds32/nds32-n7.md: New file.
3576 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3577 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3579 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3580 * config/nds32/nds32.md (pipeline_model): Add n7.
3581 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3582 * config/nds32/pipelines.md: Include n7 settings.
3584 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3585 Chung-Ju Wu <jasonwucj@gmail.com>
3587 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3588 * config/nds32/nds32-e8.md: New file.
3589 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3590 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3592 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3593 * config/nds32/nds32.md (pipeline_model): Add e8.
3594 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3595 * config/nds32/pipelines.md: Include e8 settings.
3597 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3598 Chung-Ju Wu <jasonwucj@gmail.com>
3600 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3601 * config/nds32/nds32-n8.md: New file.
3602 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3603 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3605 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3606 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3607 * config/nds32/nds32.md (pipeline_model): Add n8.
3608 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3609 * config/nds32/pipelines.md: Include n8 settings.
3611 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3612 Chung-Ju Wu <jasonwucj@gmail.com>
3614 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3615 * config/nds32/nds32-n9-2r1w.md: New file.
3616 * config/nds32/nds32-n9-3r2w.md: New file.
3617 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3618 nds32_register_ports): New or modify for cpu n9.
3619 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3621 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3622 * config/nds32/nds32-utils.c: New file.
3623 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3624 TARGET_MUL_SLOW): Define.
3625 * config/nds32/nds32.md (pipeline_model): New attribute.
3626 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3627 New options that support cpu n9.
3628 * config/nds32/pipelines.md: Include n9 settings.
3629 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3631 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3633 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3634 information if necessary.
3635 (output_cond_branch_compare_zero): Likewise.
3636 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3637 (nds32_target_alignment): Refine for alignment.
3638 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3639 (FUNCTION_BOUNDARY): Modify.
3640 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3642 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3644 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3646 * config/nds32/constants.md (unspec_volatile_element): Add values for
3647 TLB operation and data prefetch.
3648 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3649 functions for TLB operation and data prefetch.
3650 * config/nds32/nds32-intrinsic.md: Likewise.
3651 * config/nds32/nds32_intrinsic.h: Likewise.
3652 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3653 (nds32_print_operand): Likewise.
3654 * config/nds32/nds32.h (nds32_builtins): Likewise.
3656 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3657 Andrew Pinski <pinsika@gcc.gnu.org>
3660 * match.pd: Use constant_boolean_node of correct type instead of
3661 boolean_true_node or boolean_false_node for simplifying
3662 pointer comparisons to zero.
3664 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3666 PR tree-optimization/80021
3667 * tree.c (verify_type_variant): Make error call in verify_variant_match
3668 translatable and remove final full stop.
3670 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3672 * config/nds32/constants.md (unspec_volatile_element): Add
3673 UNSPEC_VOLATILE_EH_RETURN.
3674 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3675 nds32_output_stack_pop): Support dwarf exception handling process.
3676 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3677 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3678 exception handling process.
3679 (nds32_compute_stack_frame): Likewise.
3680 (nds32_return_addr_rtx): Likewise.
3681 (nds32_initial_elimination_offset): Likewise.
3682 (nds32_expand_prologue): Likewise.
3683 (nds32_expand_epilogue): Likewise.
3684 (nds32_dynamic_chain_address): New function.
3685 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3687 (DYNAMIC_CHAIN_ADDRESS): Define.
3688 (EH_RETURN_DATA_REGNO): Define.
3689 (EH_RETURN_STACKADJ_RTX): Define.
3690 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3691 patterns for dwarf exception handling.
3693 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3695 * config/nds32/nds32.h: Clean up obsolete macros.
3697 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3699 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3700 Add enum values for particular instructions.
3701 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3702 particular intrinsic functions.
3703 * config/nds32/nds32-intrinsic.md: Likewise.
3704 * config/nds32/nds32_intrinsic.h: Likewise.
3705 * config/nds32/nds32.h (nds32_builtins): Likewise.
3706 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3707 (btst, ave): New patterns for particular instructions.
3709 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3711 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3712 Add enum values for atomic load/store and memory sync.
3713 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3715 * config/nds32/nds32-intrinsic.md: Likewise.
3716 * config/nds32/nds32_intrinsic.h: Likewise.
3717 * config/nds32/nds32.h (nds32_builtins): Likewise.
3719 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3721 PR tree-optimization/85257
3722 * fold-const.c (native_encode_vector): If not all elts could fit
3723 and off is -1, return 0 rather than offset.
3724 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3725 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3726 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3727 adjust buffer in native_interpret_expr call.
3729 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3731 * config/nds32/constants.md (unspec_volatile_element): Add cache
3732 control enum values.
3733 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3734 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3735 * config/nds32/nds32.c (nds32_cctl_names): New.
3736 (nds32_print_operand): Handle cache control register names.
3737 * config/nds32/nds32.h (nds32_builtins): New enum values.
3738 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3740 * config/nds32/nds32.md (type): Add mmu.
3741 * config/nds32/pipelines.md (simple_insn): Add mmu.
3743 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3745 * config/nds32/nds32.md (type): Remove call.
3746 * config/nds32/pipelines.md (simple_insn): Likewise.
3748 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3750 * config/nds32/constants.md (unspec_volatile_element): Add
3751 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3752 UNSPEC_VOLATILE_FMFCFG.
3753 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3754 description for fmfcfg and fmfcsr.
3755 (bdesc_1arg): Add fmtcsr.
3756 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3757 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3758 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3759 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3760 unspec_fmfcfg): New patterns.
3761 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3762 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3763 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3764 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3765 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3766 __nds32__fmfcfg): Define.
3768 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3770 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3771 intrinsic register names.
3772 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3773 intrinsic register enum values and macros.
3775 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3777 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3778 for load/store addressing form.
3779 (nds32_print_operand_address): Likewise.
3781 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3784 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3785 based on LABEL_REF. Remove useless assertion.
3786 (pic_address_needs_scratch): Fix formatting.
3787 (sparc_legitimize_pic_address): Minor tweaks.
3788 (sparc_delegitimize_address): Adjust assertion accordingly.
3789 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3790 into symbolic_operand.
3791 (movsi_high_pic_label_ref): Likewise.
3792 (movsi_lo_sum_pic_label_ref): Likewise.
3793 (movdi_pic_label_ref): Likewise.
3794 (movdi_high_pic_label_ref): Likewise.
3795 (movdi_lo_sum_pic_label_ref): Likewise.
3797 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3799 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3800 custom LIB_SPEC setup.
3802 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3803 Kito Cheng <kito.cheng@gmail.com>
3805 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3806 * config/riscv/freebsd.h: New.
3808 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3810 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3811 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3814 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3815 Kito Cheng <kito.cheng@gmail.com>
3817 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3818 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3819 * config/nds32/nds32-protos.h (nds32_output_call,
3820 nds32_output_return): Declare.
3821 * config/nds32/nds32.md: Refine all the call and return patterns.
3823 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3826 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3827 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3829 PR rtl-optimization/84872
3830 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3831 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3834 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3836 * expr.c (copy_blkmode_to_reg): Revert 254862.
3837 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3839 2018-04-06 Richard Biener <rguenther@suse.de>
3842 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3843 after seeing a component reference with an adjacent field. Treat
3844 refs to arrays at struct end of external decls similar to
3845 refs to unconstrained commons.
3847 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3850 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3851 look through SAVE_EXPRs with non-side-effects argument. Adjust
3853 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3856 2018-04-06 Richard Biener <rguenther@suse.de>
3859 * alias.c (find_base_term): New wrapper around find_base_term
3860 unwinding CSELIB_VAL_PTR changes.
3861 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3864 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3866 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3868 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3869 constant definitions.
3870 ("nop"): lr 0,0 -> nopr r0
3871 ("nop_lr0", "nop_lr1"): New insn definitions.
3873 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3875 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3876 NDS32_V3PUSH_AVAILABLE_P macro.
3878 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3879 Chung-Ju Wu <jasonwucj@gmail.com>
3881 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3882 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3883 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3884 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3885 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3886 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3887 * config/nds32/constraints.md: New constraints and checking for hard
3888 float configuration.
3889 * config/nds32/iterators.md: New mode iterator and attribute for hard
3890 float configuration.
3891 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3893 * config/nds32/nds32-fpu.md: New file.
3894 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3895 deal with hard float code generation.
3896 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3898 (abi_type, float_reg_number): New enum type.
3899 * config/nds32/nds32-predicates.c: New predicates for hard float.
3900 * config/nds32/nds32-protos.h: Declare functions for hard float.
3901 * config/nds32/nds32.c: Implementation for hard float configuration.
3902 * config/nds32/nds32.h: Definitions for hard float configuration.
3903 * config/nds32/nds32.md: Include hard float machine description and
3904 modify patterns for hard float configuration.
3905 * config/nds32/nds32.opt: New options for hard float configuration.
3906 * config/nds32/predicates.md: New predicates for hard float
3909 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3911 * common/config/nds32/nds32-common.c
3912 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3914 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3917 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3918 CONSTRUCTOR_ELT (ctor, ...)->value.
3920 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3923 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3925 2018-04-05 Tom de Vries <tom@codesourcery.com>
3928 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3931 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3932 Kito Cheng <kito.cheng@gmail.com>
3934 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3935 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3936 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3937 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3939 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3940 Kito Cheng <kito.cheng@gmail.com>
3942 * config/nds32/constraints.md (Ufe): New memory constraint.
3943 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3944 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3945 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3947 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3948 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3950 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3952 * config/nds32/nds32.md: Use optimize_size in the condition for
3953 alu-shift instructions.
3955 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3957 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3959 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3961 * config/nds32/nds32.md (negsi2): Refine pattern.
3963 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3964 Chung-Ju Wu <jasonwucj@gmail.com>
3966 * config/nds32/iterators.md (shift_rotate): New code iterator.
3967 (shift): New code attribute.
3968 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3969 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3970 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3971 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3972 bit-wise operations.
3973 (andsi3, *andsi3): Ditto.
3974 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3975 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3976 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3977 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3978 nds32_ior_operand, nds32_xor_operand): New predicates.
3980 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3982 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3983 (addsi3, subsi3): ... this.
3985 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3987 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3989 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3991 * config/nds32/nds32.md: Adjust indention.
3993 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3995 * config/nds32/nds32.md (feature): New attribute.
3997 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3999 * config/nds32/nds32.md (subtype): New attribute.
4001 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4004 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4005 expansion to perform a bitwise AND of the argument followed by a
4006 boolean negation of the result.
4008 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4010 PR rtl-optimization/84878
4011 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4012 the basic block. Assert the use reference is not artificial and that
4013 it has an associated insn.
4015 2018-04-04 Michael Matz <matz@suse.de>
4017 * builtins.c (compute_objsize): Pass correct operand
4018 to array_at_struct_end_p.
4020 2018-04-04 Richard Biener <rguenther@suse.de>
4023 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4024 from contexts for DINFO_LEVEL_TERSE and below.
4026 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4028 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4029 resiter_operand condition.
4030 * config/nds32/nds32.md (*move<mode>): Ditto.
4032 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4033 Monk Chiang <sh.chiang04@gmail.com>
4035 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4037 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4039 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4041 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4042 Kito Cheng <kito.cheng@gmail.com>
4044 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4045 nds32_cond_code_str, output_cond_branch,
4046 output_cond_branch_compare_zero, nds32_expand_cbranch,
4047 nds32_expand_cstore, nds32_expand_movcc,
4048 nds32_output_cbranchsi4_equality_zero,
4049 nds32_output_cbranchsi4_equality_reg,
4050 nds32_output_cbranchsi4_equality_reg_or_const_int,
4051 nds32_output_cbranchsi4_greater_less_zero: New functions.
4052 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4053 nds32_expand_cstore, nds32_expand_movcc,
4054 nds32_output_cbranchsi4_equality_zero,
4055 nds32_output_cbranchsi4_equality_reg,
4056 nds32_output_cbranchsi4_equality_reg_or_const_int,
4057 nds32_output_cbranchsi4_greater_less_zero): Declare.
4058 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4059 nds32_rimm11s_operand): New predicates.
4060 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4061 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4064 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4066 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4067 * config/nds32/nds32.md: Ditto.
4068 * config/nds32/pipelines.md: Ditto.
4070 2018-04-04 Richard Biener <rguenther@suse.de>
4072 PR tree-optimization/85168
4073 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4074 propagating abnormals.
4076 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4078 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4080 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4081 Kito Cheng <kito.cheng@gmail.com>
4083 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4084 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4085 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4086 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4087 * config/nds32/nds32.md (sibcall_internal): New.
4088 (sibcall_register): Remove.
4089 (sibcall_immediate): Remove.
4090 (sibcall_value_internal): New.
4091 (sibcall_value_register): Remove.
4092 (sibcall_value_immediate): Remove.
4093 * config/nds32/predicates.md (nds32_general_register_operand): New.
4094 (nds32_call_address_operand): New.
4096 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4098 PR rtl-optimization/85167
4099 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4100 bb_defs if *split_p, instead preinitialize it to NULL.
4102 PR tree-optimization/85156
4103 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4104 evaluating the argument multiple times.
4106 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4108 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4110 (_mm_cvtpd_ps): Likewise.
4111 (_mm_cvttpd_epi32): Likewise.
4112 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4113 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4114 vector, pixel, and bool following altivec.h include.
4116 2018-04-03 Martin Sebor <msebor@redhat.com>
4118 * doc/extend.texi (Common Function Attributes): Clarify.
4119 (const attribute): Likewise.
4120 (pure attribute): Likewise.
4122 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4125 * config/i386/i386.c (ix86_expand_vector_set): Use
4126 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4128 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4130 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4131 instructions when changing rounding bits to preserve precision bits
4132 in the x87 control word.
4134 2018-04-03 Martin Liska <mliska@suse.cz>
4136 PR tree-optimization/82491
4137 * rtl.h (strip_offset_and_add): Replace += suboffset with
4138 poly_uint64 () + suboffset.
4140 2018-03-29 Martin Liska <mliska@suse.cz>
4141 Martin Jambor <mjambor@suse.cz>
4144 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4145 param_type is not an integral or pointer type.
4147 2018-04-03 Richard Biener <rguenther@suse.de>
4149 * sese.h (recompute_all_dominators): Remove.
4151 2018-04-02 Martin Sebor <msebor@redhat.com>
4153 * doc/invoke.texi (-Wrestrict): Fix typos.
4155 2018-04-02 Jim Wilson <jimw@sifive.com>
4157 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4158 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4159 (<optab>di3, <optab>si3_extend): Likewise.
4160 (<optab>si3_mask, <optab>si3_mask_1): New.
4161 (<optab>di3_mask, <optab>di3_mask_1): New.
4162 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4163 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4164 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4166 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4168 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4171 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4173 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4174 (nds32_canonicalize_comparison): New function.
4176 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4177 Kito Cheng <kito.cheng@gmail.com>
4178 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4180 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4181 * config/nds32/constants.md (unspec_volatile_element): Add
4182 UNSPEC_VOLATILE_RELAX_GROUP.
4183 * config/nds32/nds32-relax-opt.c: New file.
4184 * config/nds32/nds32-predicates.c
4185 (nds32_symbol_load_store_p): New function.
4186 * config/nds32/nds32-protos.h
4187 (nds32_symbol_load_store_p): Declare function.
4188 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4189 * config/nds32/nds32.c
4190 (nds32_register_pass): New function to register pass.
4191 (nds32_register_passes): New function to register passes.
4192 * config/nds32/nds32.md (relax_group): New pattern.
4193 * config/nds32/nds32.opt (mrelax-hint): New option.
4194 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4196 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4198 * config/nds32/t-nds32: Modify files dependency.
4200 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4202 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4203 (PROFILE_HOOK): Define its implementation.
4205 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4207 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4208 type and 32-bit size.
4210 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4213 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4214 (V_128_256): New mode iterator.
4215 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4216 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4217 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4219 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4220 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4222 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4225 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4226 NaN inputs correctly.
4228 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4231 * config/rs6000/vsx.md (??r): New mode attribute.
4232 (*vsx_mov<mode>_64bit): Use it.
4233 (*vsx_mov<mode>_32bit): Likewise.
4235 2018-03-30 Martin Sebor <msebor@redhat.com>
4237 PR tree-optimization/84818
4238 * builtins.c (check_access): Use warning_n.
4240 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4243 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4245 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4248 2018-03-30 Julia Koval <julia.koval@intel.com>
4251 * x86-tune.def (movx, partial_reg_dependency): Enable for
4254 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4257 * lra-constraints.c (process_alt_operands): Move setting
4258 this_alternative_matches below.
4260 2018-03-29 Martin Liska <mliska@suse.cz>
4263 * doc/invoke.texi: Document how LTO works with debug info.
4264 Describe auto-load support of binutils. Mention 'x86-64'
4265 as valid option value of -march option.
4267 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4269 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4272 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4273 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4274 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4277 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4280 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4281 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4282 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4283 for RS6000_BTM_POWERPC64.
4284 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4285 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4286 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4291 2018-03-28 Carl Love <cel@us.ibm.com>
4294 2017-09-27 Carl Love <cel@us.ibm.com>
4296 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4297 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4298 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4301 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4303 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4304 instead of __vector bool.
4305 (_mm_max_pu8): Likewise.
4306 (_mm_min_pi16): Likewise.
4308 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4311 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4312 (DIVWEUO): Likewise.
4314 (DIVDEUO): Likewise.
4315 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4316 DIVWEUO and DIVDEUO.
4317 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4318 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4319 (div_extend): Likewise.
4320 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4322 (__builtin_divweuo): Likewise.
4323 (__builtin_divdeo): Likewise.
4324 (__builtin_divdeuo): Likewise.
4326 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4329 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4330 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4332 PR tree-optimization/82004
4333 * gimple-match-head.c (optimize_pow_to_exp): New function.
4334 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4335 Don't fold to exp if optimize_pow_to_exp is false.
4337 2018-03-28 Martin Liska <mliska@suse.cz>
4340 * calls.c (initialize_argument_information): Fix trailing space.
4341 * common.opt: Fix typo and provide better explanation for
4342 -fsanitize-coverage option.
4343 * config/i386/i386.opt: Fix typo.
4345 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4346 Martin Liska <mliska@suse.cz>
4349 * gimplify.c (asan_poison_variable): Don't do the check for
4350 gimplify_omp_ctxp here.
4351 (gimplify_decl_expr): Do it here.
4352 (gimplify_target_expr): Likewise.
4354 2018-03-28 Martin Liska <mliska@suse.cz>
4357 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4358 chkp_type_bounds_count if MPX is not enabled.
4360 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4362 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4364 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4367 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4368 function to create the function decl for complex long double
4369 multiply and divide for -mabi=ieeelongdouble.
4370 (init_float128_ieee): Call it.
4372 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4375 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4376 -fcf-protection=branch -mibt.
4377 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4379 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4382 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4384 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4387 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4388 extern array declarations.
4390 2018-03-27 Richard Biener <rguenther@suse.de>
4393 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4394 explicit single_use checks.
4396 2018-03-27 Richard Biener <rguenther@suse.de>
4398 PR tree-optimization/85082
4399 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4402 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4404 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4405 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4406 Turn on fasynchronous-unwind-tables and funwind-tables.
4408 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4411 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4412 (*bmi_blsr_<mode>_ccz): Ditto.
4414 2018-03-26 Tom de Vries <tom@codesourcery.com>
4416 PR tree-optimization/85063
4417 * omp-general.c (offloading_function_p): New function. Factor out
4419 * omp-offload.c (pass_omp_target_link::gate): ... here.
4420 * omp-general.h (offloading_function_p): Declare.
4421 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4422 with attribute omp declare target for offloading functions.
4424 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4426 PR tree-optimization/84005
4427 * tree-data-ref.h (get_base_for_alignment): Declare.
4428 * tree-data-ref.c (get_base_for_alignment_1): New function.
4429 (get_base_for_alignment): Likewise.
4430 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4431 get_base_for_alignment to find a suitable base object, instead
4432 of always using drb->base_address.
4434 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4437 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4438 known size by default.
4440 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4443 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4444 and non BLKmode operands.
4446 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4449 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4450 Clean up attributes.
4452 2018-03-23 Richard Biener <rguenther@suse.de>
4455 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4456 we are going to emit early debug for LTO.
4458 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4461 * function.c (match_asm_constraints_1): Don't optimize if input
4462 doesn't satisfy general_operand predicate for output's mode.
4465 * alias.c (write_dependence_p): Don't require for x_canonicalized
4466 non-VOIDmode if x has VOIDmode.
4469 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4470 just don't try to optimize it rather than assert it never happens.
4472 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4474 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4475 macro expansions for definition of ST_INTERNAL_<mode> and
4476 LD_INTERNAL_<mode> builtins.
4477 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4479 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4481 (altivec_expand_st_builtin): Likewise.
4482 (altivec_expand_builtin): Remove calls to deleted functions.
4483 (rs6000_address_for_altivec): Delete this function.
4484 * config/rs6000/vector.md: Remove expands for
4485 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4487 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4490 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4491 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4492 re-computing once computed.
4493 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4494 (arm_init_machine_status): Initialize
4495 machine->static_chain_stack_bytes.
4497 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4500 * doc/extend.texi: Add four new prototypes for vec_ld.
4501 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4502 definitions for more logical presentation.
4503 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4504 entries for V1TI variants of __builtin_altivec_ld builtin.
4505 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4506 handling of V1TI variant of LVX icode pattern.
4507 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4508 (rs6000_gimple_fold_builtin): Likewise.
4509 (altivec_init_builtins): Add code to define
4510 __builtin_altivec_lvx_v1ti function.
4512 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4515 * function.c (match_asm_constraints_1): Don't do the optimization
4516 if input isn't a REG, SUBREG, MEM or constant.
4518 2018-03-22 Tom de Vries <tom@codesourcery.com>
4520 PR tree-optimization/84956
4521 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4522 bb_has_abnormal_pred.
4524 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4527 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4528 DECL_INITIAL (decl) to decl at the end.
4529 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4532 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4534 * doc/extend.texi (__builtin_tgmath): Document when complex
4535 integer types are treated as _Complex _Float64.
4537 2018-03-21 Tom de Vries <tom@codesourcery.com>
4539 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4541 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4543 PR tree-optimization/84960
4544 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4545 if it is ENTRY block, move them into single succ of ENTRY in that case.
4547 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4549 PR tree-optimization/84811
4550 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4551 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4552 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4553 (known_subrange_p): Update accordingly. Cast each value involved
4554 in the size comparison, rather than casting the result of the
4557 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4559 PR tree-optimization/84982
4560 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4561 by flipping the least significant bit rather than all bits from
4562 bitpos to bitpos + bitsize - 1.
4564 2018-03-21 Nathan Sidwell <nathan@acm.org>
4566 * doc/extend.texi (Deprecated Features): Remove mention of
4567 long-deleted deprecations.
4569 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4572 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4573 * configure: Regenerate.
4575 2018-03-21 Tom de Vries <tom@codesourcery.com>
4577 PR tree-optimization/83126
4578 * tree-parloops.c (num_phis): New function.
4579 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4581 2018-03-21 Nathan Sidwell <nathan@acm.org>
4583 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4584 mention anon-struct/union members and trailing attributes.
4586 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4588 PR tree-optimization/84969
4589 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4590 builtin memset partitions if they set different rhs values.
4592 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4594 PR rtl-optimization/84989
4595 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4596 VEC_DUPLICATE with scalar result mode.
4598 2018-03-21 Martin Liska <mliska@suse.cz>
4601 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4602 not intended return statement.
4604 2018-03-21 Martin Liska <mliska@suse.cz>
4607 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4608 (chkp_find_bound_slots_1): Limit number of iterations.
4610 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4613 * Minor grammar fixes for x86 options.
4615 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4618 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4619 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4621 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4624 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4625 (altivec_lvx_<mode>_1op): Likewise.
4626 (altivec_stvx_<mode>_2op): Likewise.
4627 (altivec_stvx_<mode>_1op): Likewise.
4628 (altivec_lvx_<VM2:mode>): New define_expand.
4629 (altivec_stvx_<VM2:mode>): Likewise.
4630 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4631 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4632 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4633 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4634 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4635 (rs6000_gen_lvx): Likewise.
4636 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4637 (altivec_expand_stv_builtin): Likewise.
4638 (altivec_expand_builtin): Likewise.
4639 * config/rs6000/vector.md: Likewise.
4641 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4644 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4647 2018-03-20 Richard Biener <rguenther@suse.de>
4650 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4651 sign-conversions as zero, fall back to standard scalar_stmt
4654 2018-03-20 Martin Liska <mliska@suse.cz>
4657 * predict.c (rebuild_frequencies): Handle case when we have
4658 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4660 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4663 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4664 flag_section_anchors.
4665 * varasm.c (use_blocks_for_decl_p): Remove hack for
4666 dw2_force_const_mem.
4669 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4671 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4672 be created, use lowpart_subreg of operands[0] rather than operands[0]
4674 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4675 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4676 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4677 and n constraint instead of aarch64_shift_imm_di and Usd.
4678 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4679 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4681 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4684 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4685 to favor GPR over NEON registers.
4686 (<shift>di3_neon): Likewise.
4688 2018-03-20 Tom de Vries <tom@codesourcery.com>
4691 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4692 (nvptx_process_pars): Emit bar.sync asap and alap.
4694 2018-03-20 Tom de Vries <tom@codesourcery.com>
4697 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4698 seen_label if seen_label is already set.
4700 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4703 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4704 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4705 Use 1U instead of 1. Formatting fixes.
4708 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4709 instead of TREE_TYPE (s1) for the return value.
4711 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4713 PR tree-optimization/84946
4714 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4715 bitsize + bitsize in poly_uint64 rather than poly_int64.
4718 * dwarf2asm.c: Include fold-const.c.
4719 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4720 of decl rather than decl itself.
4722 PR rtl-optimization/84643
4723 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4725 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4728 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4729 calling assemble_variable.
4731 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4734 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4735 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4737 2018-03-19 Jim Wilson <jimw@sifive.com>
4740 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4741 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4742 (riscv_first_stack_step): Likewise.
4743 (riscv_option_override): Use STACK_BOUNDARY instead of
4745 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4747 (BIGGEST_ALIGNMENT): Set to 128.
4748 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4749 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4752 2018-03-19 Richard Biener <rguenther@suse.de>
4754 PR tree-optimization/84933
4755 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4756 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4758 2018-03-19 Richard Biener <rguenther@suse.de>
4760 PR tree-optimization/84859
4761 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4762 (cond_if_else_store_replacement): Perform sinking operation on
4763 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4764 Generalize what a BB with a single eligible store is.
4766 2018-03-19 Richard Biener <rguenther@suse.de>
4768 PR tree-optimization/84929
4769 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4770 chrec_is_positive against non-chrec arg.
4772 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4775 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4777 2018-03-18 Martin Liska <mliska@suse.cz>
4779 PR rtl-optimization/84635
4780 * regrename.c (build_def_use): Use matches_mode only when
4783 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4785 PR tree-optimization/84913
4786 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4787 vectorize chains of COND_EXPRs.
4789 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4791 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4793 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4795 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4797 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4799 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4801 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4802 Kito Cheng <kito.cheng@gmail.com>
4804 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4805 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4806 (nds32_adjust_reg_alloc_order): New function.
4807 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4809 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4811 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4812 nds32_print_operand, nds32_print_operand_address): Use
4813 HOST_WIDE_INT_PRINT_DEC instead.
4815 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4817 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4819 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4822 * config/i386/i386.c (initial_ix86_tune_features,
4823 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4825 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4826 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4827 rather than 1u << ix86_tune. Formatting fix.
4828 (ix86_option_override_internal): Change ix86_arch_mask from
4829 unsigned int to unsigned HOST_WIDE_INT, initialize to
4830 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4831 (ix86_function_specific_restore): Likewise.
4833 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4836 * postreload.c (reload_combine_recognize_pattern): Perform
4837 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4838 truncate_int_for_mode the result for the destination's mode.
4841 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4842 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4844 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4847 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4850 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4851 regno_allocno_class_array and sorted_pseudos.
4852 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4853 insns where regno is used.
4855 2018-03-16 Martin Liska <mliska@suse.cz>
4858 * multiple_target.c (create_dispatcher_calls): Redirect
4859 reference in the symbol table.
4861 2018-03-16 Martin Liska <mliska@suse.cz>
4864 * multiple_target.c (create_dispatcher_calls): Redirect also
4867 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4871 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4872 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4873 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4875 2018-03-16 Julia Koval <julia.koval@intel.com>
4877 * doc/invoke.texi (Skylake Server): Add CLWB.
4878 Cannonlake): Remove CLWB.
4880 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4882 PR tree-optimization/84841
4883 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4885 (FLOAT_ONE_CONST_TYPE): Define.
4886 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4887 (sort_by_operand_rank): Put entries with higher constant_type last
4888 rather than first to match comments.
4890 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4892 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4895 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4898 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4899 check and use address of target always.
4901 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4904 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4905 (indirect_thunk_bnd_needed): Likewise.
4906 (indirect_thunks_used): Likewise.
4907 (indirect_thunks_bnd_used): Likewise.
4908 (indirect_return_needed): New.
4909 (indirect_return_bnd_needed): Likewise.
4910 (output_indirect_thunk_function): Add a bool argument for
4912 (output_indirect_thunk_function): Don't generate alias for
4913 function return thunk.
4914 (ix86_code_end): Call output_indirect_thunk_function to generate
4915 function return thunks.
4916 (ix86_output_function_return): Set indirect_return_bnd_needed
4917 and indirect_return_needed instead of indirect_thunk_bnd_needed
4918 and indirect_thunk_needed.
4920 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4922 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4923 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4924 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4926 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4927 Paul Hua <paul.hua.gm@gmail.com>
4930 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4932 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4934 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4935 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4938 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4941 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4942 instead of GET_MODE_SIZE when comparing Units.
4944 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4947 * varasm.c (hash_section): Return an unchangeble hash value
4948 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4949 Return !aarch64_can_use_per_function_literal_pools_p ().
4951 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4954 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4955 rather than address of cmode as last argument to prepare_cmp_insn.
4957 2018-03-15 Julia Koval <julia.koval@intel.com>
4959 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4960 F_AVX512VNNI, F_AVX512BITALG): New.
4962 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4965 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4966 insn for floating-point loads and stores.
4968 2018-03-14 Carl Love <cel@us.ibm.com>
4970 * config/rs6000/rs6000-c.c: Add macro definitions for
4971 ALTIVEC_BUILTIN_VEC_PERMXOR.
4972 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4973 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4974 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4975 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4977 * config/doc/extend.texi: Add prototypes for vec_permxor.
4979 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4982 * diagnostic-show-locus.c (class layout_point): Convert m_line
4983 from int to linenum_type.
4984 (line_span::comparator): Use linenum "compare" function when
4985 comparing line numbers.
4986 (test_line_span): New function.
4987 (layout_range::contains_point): Convert param "row" from int to
4989 (layout_range::intersects_line_p): Likewise.
4990 (layout::will_show_line_p): Likewise.
4991 (layout::print_source_line): Likewise.
4992 (layout::should_print_annotation_line_p): Likewise.
4993 (layout::print_annotation_line): Likewise.
4994 (layout::print_leading_fixits): Likewise.
4995 (layout::annotation_line_showed_range_p): Likewise.
4996 (struct line_corrections): Likewise for field m_row.
4997 (line_corrections::line_corrections): Likewise for param "row".
4998 (layout::print_trailing_fixits): Likewise.
4999 (layout::get_state_at_point): Likewise.
5000 (layout::get_x_bound_for_row): Likewise.
5001 (layout::print_line): Likewise.
5002 (diagnostic_show_locus): Likewise for locals "last_line" and
5004 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5005 * input.c (selftest::test_linenum_comparisons): New function.
5006 (selftest::input_c_tests): Call it.
5007 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5008 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5009 * selftest.h (ASSERT_GT): New macro.
5010 (ASSERT_GT_AT): New macro.
5011 (ASSERT_LT): New macro.
5012 (ASSERT_LT_AT): New macro.
5014 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5016 PR rtl-optimization/84780
5017 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5019 2018-03-14 Martin Liska <mliska@suse.cz>
5021 * tree.c (record_node_allocation_statistics): Use
5022 get_stats_node_kind.
5023 (get_stats_node_kind): New function extracted from
5024 record_node_allocation_statistics.
5025 (free_node): Use get_stats_node_kind.
5027 2018-03-14 Richard Biener <rguenther@suse.de>
5029 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5030 that the value-set of ANTIC_IN doesn't grow.
5033 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5035 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5036 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5038 2018-03-14 Julia Koval <julia.koval@intel.com>
5040 * config.gcc (icelake-client, icelake-server): New.
5042 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5043 (initial_ix86_arch_features): Ditto.
5044 (PTA_SKYLAKE): Add SGX.
5045 (PTA_ICELAKE): Remove.
5046 (PTA_ICELAKE_CLIENT): New.
5047 (PTA_ICELAKE_SERVER): New.
5048 (ix86_option_override_internal): Split up icelake on icelake client and
5050 (get_builtin_code_for_version): Ditto.
5051 (fold_builtin_cpu): Ditto.
5052 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5053 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5054 * config/i386/i386.h (processor_type): Ditto.
5055 * doc/invoke.texi: Ditto.
5057 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5060 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5061 INTEGER_CST offset, add it together with bitpos / 8 and
5062 sign extend based on POINTER_SIZE.
5066 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5069 * config/i386/constraints.md (Yc): New register constraint.
5070 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5071 Use Yc constraint for alternative 2 of operand 0. Remove
5072 preferred_for_speed attribute.
5074 2018-03-14 Richard Biener <rguenther@suse.de>
5076 PR tree-optimization/84830
5077 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5078 with the old one to avoid oscillations.
5080 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5083 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5085 (assign_by_spills): Return a flag of reload assignment failure.
5086 Do not process the reload assignment failures. Do not spill other
5087 reload pseudos if they has the same reg class. Update n if
5089 (lra_assign): Add a return arg. Set up from the result of
5090 assign_by_spills call.
5091 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5092 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5093 usage_insns if it is not NULL.
5094 (spill_hard_reg_in_range): New function.
5095 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5096 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5097 function prototypes.
5098 (lra_assign): Change prototype.
5099 * lra.c (lra): Add code to deal with fails by splitting hard reg
5102 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5104 * config/riscv/riscv.opt (mrelax): New option.
5105 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5106 "norelax" when riscv_mrelax is disabled.
5107 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5109 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5112 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5113 reassociation for int modes.
5115 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5117 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5118 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5120 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5121 * config/aarch64/aarch64-sve.md
5122 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5123 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5124 (*extend<mode><Vwide>2): Rename to...
5125 (aarch64_sve_extend<mode><Vwide>2): ...this.
5126 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5127 renaming the old pattern to...
5128 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5130 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5131 define_expand, renaming the old pattern to...
5132 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5133 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5134 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5135 account when deciding which SVE instruction the optab should use.
5136 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5138 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5140 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5141 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5142 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5143 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5144 (tlsdesc_small_<mode>): Turn a define_expand and use
5145 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5146 (tlsdesc_small_advsimd_<mode>): ...this.
5147 (tlsdesc_small_sve_<mode>): New pattern.
5149 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5151 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5152 (UNSPEC_UMUL_HIGHPART): New constants.
5153 (MUL_HIGHPART): New int iteraor.
5154 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5155 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5157 (*<su>mul<mode>3_highpart): New define_insn.
5159 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5162 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5165 2018-03-13 Martin Liska <mliska@suse.cz>
5168 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5170 (sem_item_optimizer::~sem_item_optimizer): Release it.
5171 (sem_item_optimizer::merge_classes): Register variable aliases.
5172 (sem_item_optimizer::fixup_pt_set): New function.
5173 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5174 * ipa-icf.h: Declare new variables and functions.
5176 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5179 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5180 integer_pow2p@2 and test integer_pow2p in condition.
5181 (A < 0 ? C : 0): Similarly for @1.
5184 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5185 characters starting at p contain '\0' character, don't look beyond
5189 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5190 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5193 * reg-stack.c (change_stack): Change update_end var from int to
5194 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5195 also call set_block_for_insn on the newly added insns and rescan.
5198 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5199 on the last operand.
5202 * tree.c (stabilize_reference_1): Return save_expr (e) for
5203 STATEMENT_LIST even if it doesn't have side-effects.
5205 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5207 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5209 2018-03-12 Renlin Li <renlin.li@arm.com>
5211 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5212 aarch64_output_scalar_simd_mov_immediate.
5214 2018-03-12 Martin Sebor <msebor@redhat.com>
5216 PR tree-optimization/83456
5217 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5218 for perfectly overlapping calls to memcpy.
5219 (gimple_fold_builtin_memory_chk): Same.
5220 (gimple_fold_builtin_strcpy): Handle no-warning.
5221 (gimple_fold_builtin_stxcpy_chk): Same.
5222 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5224 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5226 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5227 parameter. Use it for SFmode.
5228 (rs6000_function_arg_advance_1): Adjust.
5229 (rs6000_function_arg): Adjust.
5230 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5232 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5234 PR rtl-optimization/84169
5235 PR rtl-optimization/84780
5236 * combine.c (can_combine_p): Check for a 2-insn combination whether
5237 the destination register is used between the two insns, too.
5239 2018-03-12 Richard Biener <rguenther@suse.de>
5241 PR tree-optimization/84803
5242 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5243 for refs DR analysis didn't process.
5245 2018-03-12 Richard Biener <rguenther@suse.de>
5247 PR tree-optimization/84777
5248 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5249 force-vectorize loops ignore whether we are optimizing for size.
5251 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5253 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5254 (TARGET_MD_ASM_ADJUST): Define.
5256 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5257 Kito Cheng <kito.cheng@gmail.com>
5258 Chung-Ju Wu <jasonwucj@gmail.com>
5260 * config/nds32/nds32.c (nds32_compute_stack_frame,
5261 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5262 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5263 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5264 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5265 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5266 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5267 * config/nds32/nds32.md (prologue, epilogue): Use macro
5268 NDS32_V3PUSH_AVAILABLE_P to do checking.
5270 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5273 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5274 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5275 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5276 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5277 addition of most attributes on !orig_type_die or the attribute not
5278 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5280 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5281 Chung-Ju Wu <jasonwucj@gmail.com>
5283 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5285 * config/nds32/nds32.opt (mvh): New option.
5287 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5288 Chung-Ju Wu <jasonwucj@gmail.com>
5290 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5292 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5293 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5296 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5297 Chung-Ju Wu <jasonwucj@gmail.com>
5299 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5301 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5302 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5304 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5305 Kito Cheng <kito.cheng@gmail.com>
5306 Chung-Ju Wu <jasonwucj@gmail.com>
5308 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5309 UNSPEC_FFMISM and UNSPEC_FLMISM.
5310 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5311 for ffb, ffmism and flmism.
5312 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5313 (unspec_ffmism): Ditto.
5314 (unspec_flmism): Ditto.
5315 (nds32_expand_builtin_impl): Check if string extension is available.
5316 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5317 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5319 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5322 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5325 * lra-assigns.c (assign_by_spills): Return a flag of reload
5326 assignment failure. Do not process the reload assignment
5327 failures. Do not spill other reload pseudos if they has the same
5329 (lra_assign): Add a return arg. Set up from the result of
5330 assign_by_spills call.
5331 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5332 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5333 usage_insns if it is not NULL.
5334 (spill_hard_reg_in_range): New function.
5335 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5336 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5337 function prototypes.
5338 (lra_assign): Change prototype.
5339 * lra.c (lra): Add code to deal with fails by splitting hard reg
5342 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5345 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5347 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5350 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5351 (dw_val_node): Add val_symbolic_view.
5352 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5353 (symview_upper_bound): New.
5354 (new_line_info_table): Initialize symviews_since_reset.
5355 (dwarf2out_source_line): Count symviews_since_reset and set
5356 symview_upper_bound.
5357 (dw_val_equal_p): Handle symview.
5358 (add_AT_symview): New.
5359 (print_dw_val): Handle symview.
5360 (attr_checksum, attr_checksum_ordered): Likewise.
5361 (same_dw_val_p, size_of_die): Likewise.
5362 (value_format, output_die): Likewise.
5363 (add_high_low_attributes): Use add_AT_symview for entry_view.
5364 (dwarf2out_finish): Reset symview_upper_bound, clear
5367 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5370 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5371 Add strict argument and use it.
5372 (rs6000_split_multireg_move): Update for new strict argument.
5373 (mem_operand_gpr): Disallow all non-offsettable addresses.
5374 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5376 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5379 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5380 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5381 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5384 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5385 decl, use remap_type if we want to use the type.
5387 2018-03-09 Martin Sebor <msebor@redhat.com>
5389 PR tree-optimization/84526
5390 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5392 (builtin_access::generic_overlap): Be prepared to handle non-array
5395 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5397 PR rtl-optimization/84682
5398 * lra-constraints.c (process_address_1): Check is_address flag
5399 for address constraints.
5400 (process_alt_operands): Likewise.
5401 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5402 preprocess_constraints.
5403 * recog.h (preprocess_constraints): Add oploc parameter.
5405 * recog.c (preprocess_constraints): Test address_operand for
5406 CT_ADDRESS constraints.
5408 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5411 * lra-assigns.c (assign_by_spills): Return a flag of reload
5412 assignment failure. Do not process the reload assignment
5413 failures. Do not spill other reload pseudos if they has the same
5415 (lra_assign): Add a return arg. Set up from the result of
5416 assign_by_spills call.
5417 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5418 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5419 usage_insns if it is not NULL.
5420 (spill_hard_reg_in_range): New function.
5421 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5422 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5423 function prototypes.
5424 (lra_assign): Change prototype.
5425 * lra.c (lra): Add code to deal with fails by splitting hard reg
5428 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5431 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5432 Accept complain bool parameter. Only emit errors if it is true.
5433 (arm_parse_cpu_option_name): Likewise.
5434 (arm_target_thumb_only): Adjust callers of the above.
5435 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5436 prototype to take a default true bool parameter.
5437 (arm_parse_arch_option_name): Likewise.
5439 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5440 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5444 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5445 * configure: Regenerate.
5446 * configure.ac ("linker --version-script option"): New.
5447 ("linker soname option"): New.
5449 2018-03-09 Richard Biener <rguenther@suse.de>
5451 PR tree-optimization/84775
5452 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5453 immediate uses of predicate stmts and mark them modified.
5456 PR tree-optimization/84178
5457 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5459 (version_loop_for_if_conversion): Delay update_ssa call.
5460 (tree_if_conversion): Delay update_ssa until after predicate
5463 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5466 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5467 when the function accesses prior frames.
5469 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5472 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5473 gen_llsym, otherwise call maybe_gen_llsym.
5476 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5477 has ',' character inside of it.
5479 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5482 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5483 as clobbering CC_REGNUM.
5485 2018-03-08 Richard Biener <rguenther@suse.de>
5488 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5489 (follow_copies_to_constant): Do not follow SSA names registered
5492 2018-03-08 Richard Biener <rguenther@suse.de>
5494 PR tree-optimization/84178
5495 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5497 (version_loop_for_if_conversion): Delay update_ssa call.
5498 (tree_if_conversion): Delay update_ssa until after predicate
5501 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5503 PR tree-optimization/84178
5504 * tree-if-conv.c (release_bb_predicate): Remove the
5505 the assertion that the stmts have NULL use_ops.
5506 Discard the statements, asserting that they haven't
5507 yet been added to a BB.
5509 2018-03-08 Richard Biener <rguenther@suse.de>
5511 PR tree-optimization/84746
5512 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5513 (phi_translate): Pass in destination ANTIC_OUT set.
5514 (phi_translate_1): Likewise. For a simplified result lookup
5515 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5516 (phi_translate_set): Adjust.
5517 (do_pre_regular_insertion): Likewise.
5518 (do_pre_partial_partial_insertion): Likewise.
5520 2018-03-08 Martin Liska <mliska@suse.cz>
5522 PR gcov-profile/84735
5523 * doc/gcov.texi: Document usage of profile files.
5524 * gcov-io.h: Document changes in the format.
5526 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5530 * dwarf2out.c (struct dw_line_info_table): Update comments for
5532 (FORCE_RESET_NEXT_VIEW): New.
5533 (FORCE_RESETTING_VIEW_P): New.
5534 (RESETTING_VIEW_P): Check for -1 too.
5535 (ZERO_VIEW_P): Likewise.
5536 (new_line_info_table): Force-reset next view.
5537 (dwarf2out_begin_function): Likewise.
5538 (dwarf2out_source_line): Simplify zero_view_p initialization.
5539 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5540 view directly. Omit view when omitting .loc at line 0.
5542 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5544 PR tree-optimization/84740
5545 * tree-switch-conversion.c (process_switch): Call build_constructors
5546 only if info.phi_count is non-zero.
5548 PR tree-optimization/84739
5549 * tree-tailcall.c (find_tail_calls): Check call arguments against
5550 DECL_ARGUMENTS (current_function_decl) rather than
5551 DECL_ARGUMENTS (func) when checking for tail recursion.
5553 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5555 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5556 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5557 Volker Reichelt's entry and add entries for people that perform
5558 GCC fuzzy testing and report numerous bugs.
5560 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5563 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5564 readonly data in sdata, if that is disabled.
5565 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5566 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5567 -mreadonly-in-sdata option.
5569 2018-03-07 Martin Sebor <msebor@redhat.com>
5571 PR tree-optimization/84468
5572 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5573 basic block when looking for nul assignment.
5575 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5578 * except.h (output_function_exception_table): Adjust prototype.
5579 * except.c (output_function_exception_table): Remove FNNAME parameter
5580 and add SECTION parameter. Ouput one part of the table at a time.
5581 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5582 the first part of the exception table and emit unwind directives.
5583 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5584 (i386_pe_seh_cold_init): Likewise.
5585 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5586 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5587 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5588 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5589 * config/i386/winnt.c: Include except.h.
5590 (struct seh_frame_state): Add reg_offset, after_prologue and
5591 in_cold_section fields.
5592 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5593 (i386_pe_seh_cold_init): New function.
5594 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5595 to seh->in_cold_section.
5596 (seh_emit_push): Record the offset of the push.
5597 (seh_emit_save): Record the offet of the save.
5598 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5599 Test seh->after_prologue to disregard the epilogue.
5600 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5601 (i386_pe_end_cold_function): New function.
5603 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5606 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5607 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5610 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5612 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5613 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5616 * multiple_target.c: Include tree-inline.h and intl.h.
5617 (expand_target_clones): Diagnose and fail if node->definition and
5618 !tree_versionable_function_p (node->decl).
5620 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5622 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5624 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5625 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5626 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5628 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5631 * combine.c (try_combine): Use reg_or_subregno instead of handling
5632 just paradoxical SUBREGs and REGs.
5634 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5636 * config/arc/arc.c (arc_finalize_pic): Remove function.
5637 (arc_must_save_register): We use single base PIC register, remove
5638 checks to save/restore the PIC register.
5639 (arc_expand_prologue): Likewise.
5640 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5642 (arc_verify_short): Likewise.
5643 (arc_attr_type): Likewise.
5644 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5645 (walk_stores): Likewise.
5646 (arc_address_cost): Make it static.
5647 (arc_verify_short): Likewise.
5648 (branch_dest): Likewise.
5649 (arc_attr_type): Likewise.
5650 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5651 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5652 (arc_final_prescan_insn): Remove inserting the nops due to
5653 hardware hazards. It is done in reorg step.
5654 (insn_length_variant_t): Remove.
5655 (insn_length_parameters_t): Likewise.
5656 (arc_insn_length_parameters): Likewise.
5657 (arc_get_insn_variants): Likewise.
5658 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5660 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5663 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5666 PR tree-optimization/84687
5667 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5669 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5671 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5673 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5674 Rename to ppc_speculation_barrier.
5675 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5676 __builtin_ppc_speculation_barrier.
5678 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5681 * combine.c (combine_simplify_rtx): Don't try to simplify if
5682 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5685 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5687 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5688 to 32 bytes when compiling for POWER9.
5690 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5693 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5694 regparm >= 3 with no arg reg available also for calls with
5695 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5698 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5700 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5702 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5705 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5707 2018-03-05 Richard Biener <rguenther@suse.de>
5709 PR tree-optimization/84486
5710 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5711 When inserting a __builtin_assume_aligned call set the LHS
5712 SSA name alignment info accordingly.
5714 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5716 PR tree-optimization/84114
5717 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5718 Avoid reassociation of FLOAT_MODE addition.
5720 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5722 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5723 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5724 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5725 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5726 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5727 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5728 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5730 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5731 __builtin_ia32_wbinvd): New builtins.
5732 (SPECIAL_ARGS2): New.
5733 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5734 (SPECIAL_ARGS2): New.
5735 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5736 (ix86_valid_target_attribute_inner_p): Ditto.
5737 (ix86_init_mmx_sse_builtins): Add special_args2.
5738 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5739 TARGET_WBNOINVD_P): New.
5740 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5741 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5742 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5743 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5744 * config/i386/pconfigintrin.h: New file.
5745 * config/i386/wbnoinvdintrin.h: Ditto.
5746 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5747 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5749 2018-03-05 Richard Biener <rguenther@suse.de>
5751 PR tree-optimization/84670
5752 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5754 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5755 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5756 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5757 doesn't grow if all successors (recursively) were visited at least
5760 2018-03-05 Richard Biener <rguenther@suse.de>
5762 PR tree-optimization/84650
5763 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5764 if executed in the loop pipeline.
5766 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5768 * doc/configfiles.texi (Configuration Files): Move info about
5769 conditionalizing $target-protos.h to...
5770 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5771 differs from $target-protos.h.
5773 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5774 Chung-Ju Wu <jasonwucj@gmail.com>
5776 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5777 * config/nds32/nds32-multiple.md (setmemsi): Define.
5778 * config/nds32/nds32-memory-manipulation.c
5779 (nds32_gen_dup_4_byte_to_word_value): New.
5780 (emit_setmem_word_loop): New.
5781 (emit_setmem_byte_loop): New.
5782 (nds32_expand_setmem_loop): New.
5783 (nds32_expand_setmem_loop_v3m): New.
5784 (nds32_expand_setmem_unroll): New.
5785 (nds32_expand_setmem): New.
5787 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5788 Chung-Ju Wu <jasonwucj@gmail.com>
5790 * config/nds32/nds32-memory-manipulation.c
5791 (nds32_emit_load_store): New.
5792 (nds32_emit_post_inc_load_store): New.
5793 (nds32_emit_mem_move): New.
5794 (nds32_emit_mem_move_block): New.
5795 (nds32_expand_movmemsi_loop_unknown_size): New.
5796 (nds32_expand_movmemsi_loop_known_size): New.
5797 (nds32_expand_movmemsi_loop): New.
5798 (nds32_expand_movmemsi_unroll): New.
5799 (nds32_expand_movmemqi): Rename ...
5800 (nds32_expand_movmemsi): ... to this.
5801 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5802 (movmemsi): ... to this.
5803 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5804 (nds32_expand_movmemsi): ... to this.
5806 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5807 Monk Chiang <sh.chiang04@gmail.com>
5808 Chung-Ju Wu <jasonwucj@gmail.com>
5810 * config/nds32/nds32-protos.h
5811 (nds32_expand_load_multiple): New arguments.
5812 (nds32_expand_store_multiple): Ditto.
5813 (nds32_valid_multiple_load_store): Rename ...
5814 (nds32_valid_multiple_load_store_p): ... to this.
5815 * config/nds32/nds32-memory-manipulation.c
5816 (nds32_expand_load_multiple): Refine implementation.
5817 (nds32_expand_store_multiple): Ditto.
5818 * config/nds32/nds32-multiple.md
5819 (load_multiple): Update nds32_expand_load_multiple interface.
5820 (store_multiple): Update nds32_expand_store_multiple interface.
5821 * config/nds32/nds32-predicates.c
5822 (nds32_valid_multiple_load_store): Rename ...
5823 (nds32_valid_multiple_load_store_p): ... to this and refine
5825 * config/nds32/predicates.md
5826 (nds32_load_multiple_and_update_address_operation): New predicate.
5827 (nds32_store_multiple_and_update_address_operation): New predicate.
5829 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5830 Chung-Ju Wu <jasonwucj@gmail.com>
5832 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5833 (combo): New attribute.
5834 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5836 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5838 * config/nds32/nds32.opt: Change -mcmodel= default value.
5840 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5841 Monk Chiang <sh.chiang04@gmail.com>
5842 Chung-Ju Wu <jasonwucj@gmail.com>
5844 * config/nds32/constants.md (unspec_element): New enum.
5845 * config/nds32/constraints.md (Umw): New constraint.
5846 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5847 * config/nds32/nds32-intrinsic.md: Likewise.
5848 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5849 (nds32_valid_smw_lwm_base_p): New.
5850 (nds32_output_smw_single_word): New.
5851 (nds32_output_lmw_single_word): New.
5852 (nds32_expand_unaligned_load): New.
5853 (nds32_expand_unaligned_store): New.
5854 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5855 (nds32_output_smw_single_word): Declare.
5856 (nds32_output_lmw_single_word): Declare.
5857 (nds32_expand_unaligned_load): Declare.
5858 (nds32_expand_unaligned_store): Declare.
5859 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5860 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5861 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5862 NDS32_BUILTIN_UASTORE_DW.
5863 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5866 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5867 Kito Cheng <kito.cheng@gmail.com>
5868 Chung-Ju Wu <jasonwucj@gmail.com>
5870 * config/nds32/nds32-intrinsic.c
5871 (nds32_expand_builtin_null_ftype_reg): Delete.
5872 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5873 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5874 (nds32_read_argument): New.
5875 (nds32_legitimize_target): Ditto.
5876 (nds32_legitimize_argument): Ditto.
5877 (nds32_check_constant_argument): Ditto.
5878 (nds32_expand_unop_builtin): Ditto.
5879 (nds32_expand_unopimm_builtin): Ditto.
5880 (nds32_expand_binop_builtin): Ditto.
5881 (nds32_builtin_decl_impl): Ditto.
5882 (builtin_description): Ditto.
5883 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5884 (nds32_init_builtins_impl): Ditto.
5885 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5886 (nds32_builtin_decl): New.
5887 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5888 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5890 2018-03-02 Jeff Law <law@redhat.com>
5892 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5893 (redundant_insn, fill_simple_delay_slots): Likewise.
5894 (fill_slots_from_thread): Likewise.
5895 * resource.c (mark_referenced_resources): Likewise.
5896 (mark_set_resources, find_dead_or_set_registers): Likewise.
5898 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5900 * substring-locations.h (format_warning_va): Formatting fix for
5902 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5904 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5905 * substring-locations.c: Include intl.h.
5906 (format_warning_va): Turned into small wrapper around
5907 format_warning_n_va, renamed to ...
5908 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5909 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5911 (format_warning_at_substring_n): New function.
5912 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5913 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5914 format_warning_at_substring with just a shorter name instead of
5915 const function pointer.
5916 (fmtwarn_n): New function.
5917 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5918 appropriate, get rid of all the fmtstr temporaries, move conditionals
5919 with G_() wrapped string literals directly into fmtwarn arguments,
5920 cast dir.len to (int), formatting fixes.
5922 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5924 * doc/invoke.texi: Remove "Cilk Plus" references.
5926 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5927 Richard Biener <rguenther@suse.de>
5930 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5931 for error or warning attributes if CALL_FROM_THUNK_P is set.
5934 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5937 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5938 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5941 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5942 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5944 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5945 instead of -1U in last predictors element's probability member.
5947 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5950 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5951 arguments if they are comparable.
5953 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5955 PR tree-optimization/84634
5956 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5957 masks and masked_loop_p with a single loop_masks, making sure it's
5958 null for bb vectorization.
5960 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5962 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5963 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5964 loop->force_vectorize to check whether there is no alias.
5966 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5969 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5971 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5973 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5974 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5975 instead of a loop around prev_real_insn.
5976 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5980 * config/i386/i386.c (ix86_print_operand): Use conditional
5981 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5984 2018-03-02 Richard Biener <rguenther@suse.de>
5986 PR tree-optimization/84427
5987 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5988 (bitmap_set_subtract_values): Rewrite to handle multiple
5991 (prune_clobbered_mems): Likewise.
5992 (phi_translate): Take edge instead of pred/phiblock.
5993 (phi_translate_1): Likewise.
5994 (phi_translate_set): Likewise. Insert all translated
5995 exprs for a value into the set, keeping possibly multiple
5996 expressions per value.
5997 (compute_antic_aux): Adjust for phi_translate changes.
5998 When intersecting union the expressions and prune those
5999 not in the final value set, keeping possibly multiple
6000 expressions per value. Do not use value-insertion
6001 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6002 all expressions. Add verification that the value-sets
6003 only shrink during iteration.
6004 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6005 (do_pre_regular_insertion): Likewise.
6006 (do_pre_partial_partial_insertion): Likewise.
6008 2018-03-02 Richard Biener <rguenther@suse.de>
6011 * config/darwin.c (saved_debug_info_level): New static global.
6012 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6013 (darwin_asm_lto_end): Restore debug info generation settings.
6015 2018-03-01 Martin Liska <mliska@suse.cz>
6018 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6021 2018-03-01 Richard Biener <rguenther@suse.de>
6024 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6025 type annotation with LTO when debug was disabled at compile-time.
6027 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6029 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6031 (mips_final_postscan_insn): Likewise.
6033 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6035 PR rtl-optimization/84528
6036 * alias.c (init_alias_target): Add commentary.
6037 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6038 a unique base value if the frame pointer is not eliminated
6039 to the stack pointer.
6041 2018-03-01 Tom de Vries <tom@codesourcery.com>
6043 PR rtl-optimization/83327
6044 * lra-int.h (hard_regs_spilled_into): Declare.
6045 * lra.c (hard_regs_spilled_into): Define.
6046 (init_reg_info): Init hard_regs_spilled_into.
6047 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6048 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6049 (process_bb_lives): Handle hard_regs_spilled_into.
6050 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6051 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6053 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6055 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6056 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6057 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6058 * config/rs6000/aix72.h: New file.
6060 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6062 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6063 instead of warning_at with conditional singular and plural messages
6067 * stor-layout.c (update_alignment_for_field): For
6068 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6069 && !DECL_PACKED (field), do the alignment update, just use
6070 only desired_align instead of MAX (type_align, desired_align)
6072 (place_field): Don't do known_align < desired_align handling
6073 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6074 is non-NULL, instead do it after rli->prev_field handling and
6075 only if not within a bitfield word. For DECL_PACKED (field)
6076 use type_align of BITS_PER_UNIT.
6078 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6080 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6081 superfluous parentheses and trailing spaces.
6083 2018-02-28 Richard Biener <rguenther@suse.de>
6085 PR tree-optimization/84584
6086 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6087 SCoPs with fake exit edge.
6089 2018-02-28 Martin Liska <mliska@suse.cz>
6092 * timevar.c (timer::print): Fix format to properly print 100%
6095 2018-02-28 Richard Biener <rguenther@suse.de>
6098 * genmatch.c (capture_info::walk_match): Do not mark
6099 captured expressions without operands as expr_p given
6100 they act more like predicates and should be subject to
6101 "lost tail" side-effect preserving.
6103 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6105 PR rtl-optimization/81611
6106 * auto-inc-dec.c (attempt_change): Move dead note from
6107 mem_insn if it's the next use of regno
6108 (find_address): Take address use of reg holding
6109 non-incremented value. Add parm to limit search to the named
6111 (merge_in_block): Attempt to use a mem insn that is the next
6112 use of the original regno.
6114 2018-02-27 Martin Sebor <msebor@redhat.com>
6117 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6118 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6120 2018-02-27 Martin Sebor <msebor@redhat.com>
6122 PR translation/84207
6123 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6124 n argument to unsigned HOST_WIDE_INT.
6125 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6126 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6127 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6128 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6130 2018-02-27 Richard Biener <rguenther@suse.de>
6132 PR tree-optimization/84512
6133 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6134 Do not use the estimate returned from record_stmt_cost for
6135 the scalar iteration cost but sum properly using add_stmt_cost.
6137 2018-02-27 Richard Biener <rguenther@suse.de>
6139 PR tree-optimization/84466
6140 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6141 Adjust last change to less strictly validate use operands.
6143 2018-02-27 Martin Liska <mliska@suse.cz>
6145 PR gcov-profile/84548
6146 * gcov.c (process_file): Allow partial overlap and consider it
6147 also as group functions.
6148 (output_lines): Properly calculate range of lines for a group.
6150 2018-02-27 Martin Liska <mliska@suse.cz>
6152 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6153 'ggc' suffixes. Change first column width.
6154 (timer::print): Fix formatting of the column.
6156 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6158 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6159 preserve inline entry blocks for the sake of debug inline
6160 entry point markers alone.
6161 (remove_unused_locals): Suggest in comments a better place to
6162 force the preservation of inline entry blocks that are
6163 otherwise unused, but do not preserve them.
6165 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6167 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6169 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6172 * config/i386/constraints.md (Bs): Replace
6173 ix86_indirect_branch_register with
6174 TARGET_INDIRECT_BRANCH_REGISTER.
6176 * config/i386/i386.md (indirect_jump): Likewise.
6177 (tablejump): Likewise.
6178 (*sibcall_memory): Likewise.
6179 (*sibcall_value_memory): Likewise.
6180 Peepholes of indirect call and jump via memory: Likewise.
6181 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6182 (*sibcall_value_GOT_32): Likewise.
6183 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6184 (GOT_memory_operand): Likewise.
6185 (call_insn_operand): Likewise.
6186 (sibcall_insn_operand): Likewise.
6187 (GOT32_symbol_operand): Likewise.
6188 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6190 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6192 PR rtl-optimization/83496
6193 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6194 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6195 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6196 redundant insn, if any.
6197 (relax_delay_slots): Likewise.
6198 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6200 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6202 PR tree-optimization/83965
6203 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6204 that grouped statements are part of a reduction chain. Return
6205 true if the statement is not marked as a reduction itself but
6207 (vect_recog_dot_prod_pattern): Don't check whether the statement
6208 is part of a group here.
6209 (vect_recog_sad_pattern): Likewise.
6210 (vect_recog_widen_sum_pattern): Likewise.
6212 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6215 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6217 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6220 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6222 (ix86_output_indirect_function_return): New prototype.
6223 (ix86_split_simple_return_pop_internal): Likewise.
6224 * config/i386/i386.c (indirect_return_via_cx): New.
6225 (indirect_return_via_cx_bnd): Likewise.
6226 (indirect_thunk_name): Handle return va CX_REG.
6227 (output_indirect_thunk_function): Create alias for
6228 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6229 (ix86_output_indirect_jmp): Remove the bool argument.
6230 (ix86_output_indirect_function_return): New function.
6231 (ix86_split_simple_return_pop_internal): Likewise.
6232 * config/i386/i386.md (*indirect_jump): Don't pass false
6233 to ix86_output_indirect_jmp.
6234 (*tablejump_1): Likewise.
6235 (simple_return_pop_internal): Change it to define_insn_and_split.
6236 Call ix86_split_simple_return_pop_internal to split it for
6238 (simple_return_indirect_internal): Call
6239 ix86_output_indirect_function_return instead of
6240 ix86_output_indirect_jmp.
6242 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6245 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6246 memset and value initialization afterwards.
6248 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6250 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6252 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6255 * common/config/aarch64/aarch64-common.c
6256 (aarch_option_optimization_table[]): Switch
6257 off fomit-frame-pointer
6259 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6260 Chung-Ju Wu <jasonwucj@gmail.com>
6262 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6264 (store_multiple): Ditto.
6266 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6268 * config.gcc: Add --with-cpu support for nds32 target.
6269 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6270 * config/nds32/nds32.opt: Add -mcpu= option.
6272 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6274 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6275 isel=yes): Warn for these deprecated options.
6277 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6279 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6280 ISA_2_5_MASKS_EMBEDDED.
6282 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6284 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6285 p->max as pointers rather than using iterative_hash_expr.
6287 2018-02-23 Carl Love <cel@us.ibm.com>
6289 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6290 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6292 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6293 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6294 P8V_BUILTIN_VEC_VUNSIGNED2.
6296 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6299 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6300 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6302 * lra-constraints.c (curr_insn_transform): Set up
6303 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6305 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6306 * lra-eliminations.c (spill_pseudos): Ditto.
6307 (process_insn_for_elimination): Ditto.
6308 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6309 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6310 LRA_NON_CLOBBERED_ALT.
6312 2018-02-22 Martin Sebor <msebor@redhat.com>
6314 PR tree-optimization/84480
6315 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6316 to maybe_diag_stxncpy_trunc. Call it.
6317 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6318 from gimple_fold_builtin_strcpy. Print inlining stack.
6319 (handle_builtin_stxncpy): Print inlining stack.
6320 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6322 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6325 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6326 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6327 and -fcheck-pointer-bounds are used together.
6328 (indirect_thunk_prefix): New enum.
6329 (indirect_thunk_need_prefix): New function.
6330 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6331 "_nt" instead of "_bnd" for NOTRACK prefix.
6332 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6333 (output_indirect_thunk_function): Likewise.
6335 (ix86_code_end): Update output_indirect_thunk_function calls.
6336 (ix86_output_indirect_branch_via_reg): Replace
6337 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6338 (ix86_output_indirect_branch_via_push): Likewise.
6339 (ix86_output_function_return): Likewise.
6340 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6341 incompatible with -fcf-protection=branch and
6342 -fcheck-pointer-bounds.
6344 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6347 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6348 Change gcc_assert call to output_operand_lossage.
6350 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6352 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6354 2018-02-22 DJ Delorie <dj@redhat.com>
6355 Sebastian Perta <sebastian.perta@renesas.com>
6356 Oleg Endo <olegendo@gcc.gnu.org>
6358 * config/rx/rx.c (rx_rtx_costs): New function.
6359 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6361 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6363 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6365 2018-02-22 Martin Liska <mliska@suse.cz>
6368 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6369 Add "native" as a possible value.
6371 2018-02-22 Martin Liska <mliska@suse.cz>
6374 * config/i386/i386.c (ix86_option_override_internal):
6375 Add "native" as a possible value for -march and -mtune.
6377 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6380 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6381 to all type variants.
6383 PR tree-optimization/84503
6384 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6385 width as info->bitpos + info->bitsize - start.
6386 (merged_store_group::merge_overlapping): Simplify width computation.
6387 (check_no_overlap): New function.
6388 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6389 start + width and last_order of the group, fail if check_no_overlap
6391 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6392 to group if check_no_overlap fails.
6394 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6396 * config/rs6000/altivec.md: Delete contraint arguments to
6397 define_expand, define_split, and define_peephole2, and in
6398 define_insn_and_split if always unused.
6399 * config/rs6000/darwin.md: Ditto.
6400 * config/rs6000/dfp.md: Ditto.
6401 * config/rs6000/rs6000.md: Ditto.
6402 * config/rs6000/sync.md: Ditto.
6403 * config/rs6000/vector.md: Ditto.
6404 * config/rs6000/vsx.md: Ditto.
6406 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6408 * config/rs6000/altivec.md: Write output control strings as braced
6409 blocks instead of double-quoted strings.
6410 * config/rs6000/darwin.md: Ditto.
6411 * config/rs6000/rs6000.md: Ditto.
6412 * config/rs6000/vector.md: Ditto.
6413 * config/rs6000/vsx.md: Ditto.
6415 2018-02-21 Jason Merrill <jason@redhat.com>
6417 PR c++/84314 - ICE with templates and fastcall attribute.
6418 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6420 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6422 * ipa-cp.c (determine_versionability): Fix comment typos.
6424 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6427 * ipa-cp.c (determine_versionability): Do not version functions caling
6430 2018-02-21 Martin Liska <mliska@suse.cz>
6433 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6434 Add "native" as a possible value.
6435 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6436 the macro when native cpu detection is available.
6438 2018-02-21 Martin Liska <mliska@suse.cz>
6441 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6442 Add "native" as a possible value.
6443 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6444 when native cpu detection is available.
6446 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6447 Martin Sebor <msebor@redhat.com>
6449 PR tree-optimization/84478
6450 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6452 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6453 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6454 support which is conservatively correct, for 2 only stay conservative
6455 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6456 argument to the 2 argument get_range_strlen, adjust 6 arg
6457 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6459 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6460 (gimple_fold_builtin_strlen): Pass true as last argument to
6463 2018-02-20 Martin Sebor <msebor@redhat.com>
6466 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6467 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6468 (builtin_memref::builtin_memref): Factor out parts into
6469 set_base_and_offset and call it.
6471 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6474 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6475 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6476 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6477 search at the associated MODE_INT.
6479 2018-02-20 Jeff Law <law@redhat.com>
6482 PR tree-optimization/81592
6484 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6485 for range data rather than using global data.
6486 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6487 range data rather than using global data.
6488 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6489 pass it to children as needed.
6490 (struct directive::fmtresult): Similarly.
6491 (struct directive::set_width): Similarly.
6492 (struct directive::set_precision): Similarly.
6493 (format_integer, format_directive, parse_directive): Similarly.
6494 (format_none): Accept unnamed vr_values parameter.
6495 (format_percent, format_floating, format_character): Similarly.
6496 (format_string, format_plain): Similarly.
6497 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6498 the EVRP range analyzer for range data rather than using global data.
6499 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6500 gimple-ssa-evrp-analyze.h
6501 (class sprintf_dom_walker): Add after_dom_children member function.
6502 Add evrp_range_analyzer member.
6503 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6504 range analyzer as needed.
6505 (sprintf_dom_walker::after_dom_children): New member function.
6506 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6508 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6509 (evrp_range_analyzer::pop_to_marker): Likewise.
6511 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6513 PR tree-optimization/84419
6514 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6515 with the required type if its current type is compatible but
6518 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6521 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6522 after vectorization.
6524 2018-02-20 Martin Liska <mliska@suse.cz>
6527 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6528 possible values if we don't have a hint.
6530 2018-02-20 Martin Liska <mliska@suse.cz>
6534 * final.c (shorten_branches): Build align_tab array with one
6536 * opts.c (finish_options): Add alignment option limit check.
6537 (MAX_CODE_ALIGN): Likewise.
6538 (MAX_CODE_ALIGN_VALUE): Likewise.
6539 * doc/invoke.texi: Document maximum allowed option value for
6540 all -falign-* options.
6542 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6545 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6546 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6547 * var-tracking.c (emit_note_insn_var_location): Remove all references
6548 to NOTE_INSN_CALL_ARG_LOCATION.
6549 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6550 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6551 Use copy_rtx_if_shared.
6552 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6553 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6554 (dwarf2out_var_location): Remove handling of
6555 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6557 * final.c (final_scan_insn): Remove all references to
6558 NOTE_INSN_CALL_ARG_LOCATION.
6559 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6560 before dumping final insns.
6561 * except.c (emit_note_eh_region_end): Remove all references to
6562 NOTE_INSN_CALL_ARG_LOCATION.
6563 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6564 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6565 * config/arc/arc.c (hwloop_optimize): Likewise.
6566 * config/arm/arm.c (create_fix_barrier): Likewise.
6567 * config/s390/s390.c (s390_chunkify_start): Likewise.
6568 * config/sh/sh.c (find_barrier): Likewise.
6569 * config/i386/i386.c (rest_of_insert_endbranch,
6570 ix86_seh_fixup_eh_fallthru): Likewise.
6571 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6572 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6573 * config/frv/frv.c (frv_function_prologue): Likewise.
6574 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6576 (note_outside_basic_block_p): Remove all references to
6577 NOTE_INSN_CALL_ARG_LOCATION.
6578 * gengtype.c (adjust_field_rtx_def): Likewise.
6579 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6581 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6582 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6585 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6588 PR tree-optimization/84452
6589 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6590 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6593 2018-02-19 Martin Liska <mliska@suse.cz>
6596 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6598 2018-02-19 Martin Liska <mliska@suse.cz>
6599 Richard Sandiford <richard.sandiford@linaro.org>
6601 PR tree-optimization/82491
6602 * gimple-fold.c (get_base_constructor): Make earlier bail out
6605 2018-02-19 Carl Love <cel@us.ibm.com>
6607 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6608 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6610 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6611 P8V_BUILTIN_VEC_NEG.
6613 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6615 * config/rl78/rl78.md (movdf): New define expand.
6617 2018-02-19 Martin Liska <mliska@suse.cz>
6620 * doc/invoke.texi: Fix typo.
6621 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6623 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6625 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6626 handle rs6000_single_float and rs6000_double_float specially for
6629 2018-02-16 Jeff Law <law@redhat.com>
6631 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6634 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6637 * ipa-inline.c (inline_small_functions): Fix a typo.
6639 2018-02-16 Nathan Sidwell <nathan@acm.org>
6641 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6643 2018-02-16 Carl Love <cel@us.ibm.com>
6645 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6646 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6647 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6648 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6649 expansion to P8V_BUILTIN_VEC_FLOAT2.
6651 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6653 PR rtl-optimization/70023
6654 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6655 src_regno into account.
6657 2018-02-16 Carl Love <cel@us.ibm.com>
6659 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6660 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6661 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6662 * config/rs6000/rs6000.c: Remove case statements for
6663 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6664 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6665 and P9V_BUILTIN_VEC_VINSERT4B.
6666 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6667 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6668 * config/rs6000/vsx.md:
6669 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6672 2018-02-16 Carl Love <cel@us.ibm.com>
6674 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6676 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6678 * config/rs6000/rs6000-c.c: Add the definitions for
6679 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6680 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6681 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6682 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6683 definition for insert4b and define insn *insert3b_internal.
6684 * doc/extend.texi: Add documentation for vec_extract4b.
6686 2018-02-16 Nathan Sidwell <nathan@acm.org>
6688 * doc/extend.texi (Backwards Compatibility): Mention friend
6689 injection. Note for-scope is deprecated.
6690 * doc/invoke.texi (-ffriend-injection): Deprecate.
6692 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6694 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6695 that moved to I2, also allow destinations that are a paradoxical
6696 subreg (instead of a normal reg).
6698 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6701 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6704 2018-02-16 Richard Biener <rguenther@suse.de>
6706 PR tree-optimization/84037
6707 PR tree-optimization/84016
6709 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6710 Adjust vec_construct for the fact we need additional higher latency
6711 128bit inserts for AVX256 and AVX512 vector builds.
6712 (ix86_add_stmt_cost): Scale vector construction cost for
6715 2018-02-16 Richard Biener <rguenther@suse.de>
6717 PR tree-optimization/84417
6718 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6719 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6720 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6722 2018-02-16 Martin Liska <mliska@suse.cz>
6725 * internal-fn.def (ASAN_CHECK): Set proper flags.
6726 (ASAN_MARK): Likewise.
6728 2018-02-16 Julia Koval <julia.koval@intel.com>
6730 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6731 from PTA_CANNONLAKE.
6733 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6736 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6737 Use ++iter rather than iter++ for std::list iterators.
6738 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6739 defer deleting them until all nodes in the forest are processed. Do
6740 free even leaf nodes. Change to_process into auto_vec.
6743 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6744 * vec.h (vec_default_construct): Use memset instead of placement new
6745 if BROKEN_VALUE_INITIALIZATION is defined.
6746 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6747 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6750 PR rtl-optimization/83723
6751 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6752 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6753 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6755 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6757 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6759 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6761 PR rtl-optimization/81443
6762 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6763 from inner REGs to paradoxical SUBREGs.
6765 2018-02-16 Richard Biener <rguenther@suse.de>
6767 PR tree-optimization/84399
6768 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6769 For operands we can analyze at their definition make sure we can
6770 analyze them at each use as well.
6772 2018-02-16 Richard Biener <rguenther@suse.de>
6774 PR tree-optimization/84190
6775 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6776 volatile accesses if the decl isn't volatile.
6778 2018-02-15 Jason Merrill <jason@redhat.com>
6780 PR c++/84314 - ICE with templates and fastcall attribute.
6781 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6782 TYPE_CANONICAL on an existing type.
6784 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6786 PR tree-optimization/84383
6787 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6788 dstoff nor call operand_equal_p if dstbase is NULL.
6790 PR tree-optimization/84334
6791 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6792 also a CONSTANT_CLASS_P, punt.
6794 2018-02-14 Jim Wilson <jimw@sifive.com>
6796 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6797 first SMALL_OPERAND check. New local min_second_step. Move assert
6798 to where locals are set. Add TARGET_RVC support.
6799 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6801 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6803 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6805 2018-02-14 Martin Sebor <msebor@redhat.com>
6807 PR tree-optimization/83698
6808 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6809 arrays constrain the offset range to their bounds.
6810 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6811 (builtin_access::overlap): Avoid setting the size of overlap if it's
6813 (maybe_diag_overlap): Also consider arrays when deciding what values
6814 of offsets to include in diagnostics.
6816 2018-02-14 Martin Sebor <msebor@redhat.com>
6819 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6820 that correspond to the kind of a declaration.
6822 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6825 * config/pa/pa.md: Load address of PIC label using the linkage table
6826 if the label is nonlocal.
6828 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6830 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6831 warning message if user requests -maltivec=be.
6832 * doc/invoke.texi: Document deprecation of -maltivec=be.
6834 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6837 * config/rs6000/rs6000-c.c: Update definitions for
6838 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6839 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6841 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6844 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6845 add _get_ssp intrinsics. Remove argument from
6846 __builtin_ia32_rdssp[d|q].
6847 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6848 * config/i386/i386-builtin.def: Remove argument from
6849 __builtin_ia32_rdssp[d|q].
6850 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6851 ix86_expand_special_args_builtin for _rdssp[d|q].
6852 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6853 Clear register before usage.
6854 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6855 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6857 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6859 PR tree-optimization/84357
6860 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6861 operand 1 of an ARRAY_REF too.
6863 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6866 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6867 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6869 (set_of_reg): New struct.
6870 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6871 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6872 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6874 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6875 Split into bitclr, bitset, bitinvert patterns if appropriate.
6876 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6877 use rx_fuse_in_memory_bitop.
6878 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6879 to named insn, correct maximum insn length.
6881 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6884 * machmode.def: Define a complex mode for PARTIAL_INT.
6885 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6887 * doc/rtl.texi: Document CSPImode.
6888 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6890 (msp430_hard_regno_nregs_with_padding): Likewise.
6892 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6895 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6897 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6899 PR rtl-optimization/84169
6900 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6901 we generated a parallel as new i3 and we split that to new i2 and i3
6902 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6903 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6904 those to i2, not i1. Partially rewrite this scan code.
6906 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6909 * stor-layout.c (place_field): For variable length fields, adjust
6910 offset_align afterwards not just based on the field's alignment,
6911 but also on the size.
6914 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6915 of exps and logs in the use_exp2 case.
6917 2018-02-13 Jeff Law <law@redhat.com>
6919 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6922 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6925 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6929 * common.opt (gas-loc-support, gas-locview-support): New.
6930 (ginline-points, ginternal-reset-location-views): New.
6931 * doc/invoke.texi: Document them. Use @itemx where intended.
6932 (gvariable-location-views): Adjust.
6933 * target.def (reset_location_view): New.
6934 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6935 (TARGET_RESET_LOCATION_VIEW): New.
6936 * doc/tm.texi: Rebuilt.
6937 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6938 (dwarf2out_default_as_locview_support): New.
6939 (output_asm_line_debug_info): Use option variables.
6940 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6941 (output_loc_list): Likewise.
6942 (add_high_low_attributes): Check option variables.
6943 Don't output entry view attribute in strict mode.
6944 (gen_inlined_subroutine_die): Check option variables.
6945 (dwarf2out_inline_entry): Likewise.
6946 (init_sections_and_labels): Likewise.
6947 (dwarf2out_early_finish): Likewise.
6948 (maybe_reset_location_view): New, from...
6949 (dwarf2out_var_location): ... here. Call it.
6950 * debug.h (dwarf2out_default_as_loc_support): Declare.
6951 (dwarf2out_default_as_locview_support): Declare.
6952 * hooks.c (hook_int_rtx_insn_0): New.
6953 * hooks.h (hook_int_rtx_insn_0): Declare.
6954 * toplev.c (process_options): Take -gas-loc-support and
6955 -gas-locview-support from dwarf2out. Enable
6956 -gvariable-location-views by default only with locview
6957 assembler support. Enable -ginternal-reset-location-views by
6958 default only if the target defines the corresponding hook.
6959 Enable -ginline-points by default if location views are
6960 enabled; force it disabled if statement frontiers are
6962 * tree-inline.c (expand_call_inline): Check option variables.
6963 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6965 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6967 PR tree-optimization/84321
6968 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6969 handling. Also check whether the anti-range contains any values
6970 that satisfy the mask; switch to a VR_RANGE if not.
6972 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6975 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6977 2018-02-13 Martin Jambor <mjambor@suse.cz>
6980 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6981 of call statements, also set location of a load to a temporary.
6983 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6985 * config/rl78/rl78.c (add_vector_labels): New function.
6986 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6987 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6988 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6989 which checks that no arguments are passed.
6990 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6991 * doc/extend.texi: Documentation for the new attribute.
6993 2018-02-13 Andreas Schwab <schwab@suse.de>
6995 * config/riscv/linux.h (CPP_SPEC): Define.
6997 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7000 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7001 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7002 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7003 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7004 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7005 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7006 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7007 temporarily for AES and PCLMUL builtins.
7009 PR tree-optimization/84339
7010 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7011 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7015 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7016 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7017 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7019 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7021 * omp-simd-clone.h: New file.
7022 * omp-simd-clone.c: Include omp-simd-clone.h.
7023 (expand_simd_clones): No longer static.
7024 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7025 cgraph.h and omp-simd-clone.h.
7026 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7027 (vect_recog_widen_shift_pattern): Formatting fix.
7028 (vect_pattern_recog_1): Don't check optab for calls.
7031 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7032 operands[2] into a REG before using gen_lowpart on it.
7034 2018-02-12 Jeff Law <law@redhat.com>
7037 * config/sh/sh.c (find_barrier): Consider a sibling call
7040 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7041 successfully back substituting a reg.
7043 2018-02-12 Richard Biener <rguenther@suse.de>
7045 PR tree-optimization/84037
7046 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7047 parameter, move visited init to caller.
7048 (vect_slp_analyze_operations): Separate cost from validity
7049 check, initialize visited once for all instances.
7050 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7052 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7054 (vect_model_promotion_demotion_cost): Likewise.
7055 (vectorizable_bswap): Guard cost modeling with !slp_node
7056 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7058 (vectorizable_call): Likewise.
7059 (vectorizable_conversion): Likewise.
7060 (vectorizable_assignment): Likewise.
7061 (vectorizable_shift): Likewise.
7062 (vectorizable_operation): Likewise.
7063 (vectorizable_store): Likewise.
7064 (vectorizable_load): Likewise.
7065 (vectorizable_condition): Likewise.
7066 (vectorizable_comparison): Likewise.
7068 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7071 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7072 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7073 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7075 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7078 * params.def (inline-min-speedup): Increase from 8 to 15.
7079 (max-inline-insns-auto): Decrease from 40 to 30.
7080 * ipa-split.c (consider_split): Add some buffer for function to
7081 be considered inlining candidate.
7082 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7085 2018-02-12 Richard Biener <rguenther@suse.de>
7087 PR tree-optimization/84037
7088 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7089 matched stmts if we cannot swap the non-matched ones.
7091 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7093 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7094 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7095 _mm_maskz_scalef_round_ss): New intrinsics.
7096 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7097 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7098 __builtin_ia32_scalefss_round): Remove.
7099 (__builtin_ia32_scalefsd_mask_round,
7100 __builtin_ia32_scalefss_mask_round): New intrinsics.
7101 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7102 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7103 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7104 "<round_constraint>")): Changed to ...
7105 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7106 "<round_scalar_constraint>")): ... this.
7107 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7108 %0, %1, %2<round_op3>}"): Changed to ...
7109 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7110 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7111 %2<round_scalar_mask_op3>}"): ... this.
7112 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7114 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7116 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7117 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7118 (_mm_maskz_sqrt_round_ss): New intrinsics.
7119 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7120 (__builtin_ia32_sqrtsd_mask_round)
7121 (__builtin_ia32_sqrtss_mask_round): New builtins.
7122 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7123 (__builtin_ia32_sqrtss_round): Remove.
7124 (__builtin_ia32_sqrtsd_mask_round)
7125 (__builtin_ia32_sqrtss_mask_round): New builtins.
7126 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7127 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7128 ((match_operand:VF_128 1 "vector_operand"
7129 "xBm,<round_constraint>")): Changed to ...
7130 ((match_operand:VF_128 1 "vector_operand"
7131 "xBm,<round_scalar_constraint>")): ... this.
7132 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7133 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7134 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7135 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7136 %<iptr>1<round_scalar_mask_op3>}): ... this.
7137 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7138 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7140 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7143 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7144 Cast vec_cmpeq result to correct type.
7145 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7146 Cast vec_cmpgt result to correct type.
7148 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7150 * final.c (final_scan_insn_1): Renamed from...
7151 (final_scan_insn): ... this. New wrapper, to recover
7152 seen from the outermost call in recursive ones.
7153 * config/sparc/sparc.c (output_return): Drop seen from call.
7154 (output_sibcall): Likewise.
7155 * config/visium/visium.c (output_branch): Likewise.
7157 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7159 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7162 2018-02-10 Alan Modra <amodra@gmail.com>
7165 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7166 Specify LR as an input.
7168 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7171 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7172 remove_member_access_dummy_vars): New functions.
7173 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7174 lower_omp_1, execute_lower_omp): Use them.
7176 PR rtl-optimization/84308
7177 * shrink-wrap.c (spread_components): Release todo vector.
7179 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7181 PR rtl-optimization/57193
7182 * ira-color.c (struct allocno_color_data): Add member
7183 conflict_allocno_hard_prefs.
7184 (update_conflict_allocno_hard_prefs): New.
7185 (bucket_allocno_compare_func): Add a preference based on
7186 conflict_allocno_hard_prefs.
7187 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7188 (color_allocnos): Remove a dead code. Initiate
7189 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7191 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7194 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7195 constraint from =wa to wa. Avoid a subreg on the output operand,
7196 instead use a pseudo and subreg it in a move.
7197 (p9_xxbrd_<mode>): Changed to ...
7198 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7199 (p9_xxbrd_v2df): New expander.
7200 (p9_xxbrw_<mode>): Changed to ...
7201 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7202 (p9_xxbrw_v4sf): New expander.
7204 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7206 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7208 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7211 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7212 multiply in 32-bit mode.
7213 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7214 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7217 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7219 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7220 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7221 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7222 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7224 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7227 * dwarf2out.c (is_trivial_indirect_ref): New function.
7228 (dwarf2out_late_global_decl): Do not generate a location
7229 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7230 and that are not defined in the current unit.
7232 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7234 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7235 instead of a libcall for UNORDERED.
7237 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7240 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7241 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7243 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7246 * config/s390/s390.c (s390_set_current_function): Invoke
7247 s390_indirect_branch_settings also if fndecl didn't change.
7249 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7251 * config/rs6000/rs6000.md (blockage): Set length to zero.
7253 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7255 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7257 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7260 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7261 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7265 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7266 PARALLEL incoming that failed vt_get_decl_and_offset check.
7269 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7270 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7272 (get_variable_section): For decls in named .bss* sections pass true as
7273 second argument to bss_initializer_p.
7275 2018-02-09 Marek Polacek <polacek@redhat.com>
7276 Jakub Jelinek <jakub@redhat.com>
7279 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7280 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7281 Sync some changes from cxx_fold_indirect_ref.
7283 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7285 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7287 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7288 (BLOCK_INLINE_ENTRY_LABEL): New.
7289 (dwarf2out_var_location): Disregard inline entry markers.
7290 (inline_entry_data): New struct.
7291 (inline_entry_data_hasher): New hashtable type.
7292 (inline_entry_data_hasher::hash): New.
7293 (inline_entry_data_hasher::equal): New.
7294 (inline_entry_data_table): New variable.
7295 (add_high_low_attributes): Add DW_AT_entry_pc and
7296 DW_AT_GNU_entry_view attributes if a pending entry is found
7297 in inline_entry_data_table. Add old entry_pc attribute only
7298 if debug nonbinding markers are disabled.
7299 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7300 markers are enabled.
7301 (block_within_block_p, dwarf2out_inline_entry): New.
7302 (dwarf2out_finish): Check that no entries remained in
7303 inline_entry_data_table.
7304 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7305 (final_scan_insn, notice_source_line): Likewise.
7306 (rest_of_clean_state): Skip inline entry markers.
7307 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7309 * gimple.c (gimple_build_debug_inline_entry): New.
7310 * gimple.h (enum gimple_debug_subcode): Add
7311 GIMPLE_DEBUG_INLINE_ENTRY.
7312 (gimple_build_debug_inline_entry): Declare.
7313 (gimple_debug_inline_entry_p): New.
7314 (gimple_debug_nonbind_marker_p): Adjust.
7315 * insn-notes.def (INLINE_ENTRY): New.
7316 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7317 inline entry marker notes.
7318 (print_insn): Likewise.
7319 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7320 (INSN_DEBUG_MARKER_KIND): Likewise.
7321 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7322 * tree-inline.c (expand_call_inline): Build and insert
7323 debug_inline_entry stmt.
7324 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7325 inline entry blocks early, if nonbind markers are enabled.
7326 (dump_scope_block): Dump fragment info.
7327 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7328 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7329 (gimple_build_debug_inline_entry): New.
7330 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7331 Enable/disable inline entry points too.
7332 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7333 (DEBUG_INSN): Describe inline entry markers.
7335 * common.opt (gvariable-location-views): New.
7336 (gvariable-location-views=incompat5): New.
7337 * config.in: Rebuilt.
7338 * configure: Rebuilt.
7339 * configure.ac: Test assembler for view support.
7340 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7341 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7342 * dwarf2out.c (var_loc_view): New typedef.
7343 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7344 (dwarf2out_locviews_in_attribute): New.
7345 (dwarf2out_locviews_in_loclist): New.
7346 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7347 (enum dw_line_info_opcode): Add LI_adv_address.
7348 (struct dw_line_info_table): Add view.
7349 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7350 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7351 (zero_view_p): New variable.
7352 (ZERO_VIEW_P): New macro.
7353 (output_asm_line_debug_info): New.
7354 (struct var_loc_node): Add view.
7355 (add_AT_view_list, AT_loc_list): New.
7356 (add_var_loc_to_decl): Add view param. Test it against last.
7357 (new_loc_list): Add view params. Record them.
7358 (AT_loc_list_ptr): Handle loc and view lists.
7359 (view_list_to_loc_list_val_node): New.
7360 (print_dw_val): Handle dw_val_class_view_list.
7361 (size_of_die): Likewise.
7362 (value_format): Likewise.
7363 (loc_list_has_views): New.
7364 (gen_llsym): Set vl_symbol too.
7365 (maybe_gen_llsym, skip_loc_list_entry): New.
7366 (dwarf2out_maybe_output_loclist_view_pair): New.
7367 (output_loc_list): Output view list or entries too.
7368 (output_view_list_offset): New.
7369 (output_die): Handle dw_val_class_view_list.
7370 (output_dwarf_version): New.
7371 (output_compilation_unit_header): Use it.
7372 (output_skeleton_debug_sections): Likewise.
7373 (output_rnglists, output_line_info): Likewise.
7374 (output_pubnames, output_aranges): Update version comments.
7375 (output_one_line_info_table): Output view numbers in asm comments.
7376 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7377 Call maybe_gen_llsym.
7378 (loc_list_from_tree_1): Adjust.
7379 (add_AT_location_description): Create view list attribute if
7380 needed, check it's absent otherwise.
7381 (convert_cfa_to_fb_loc_list): Adjust.
7382 (maybe_emit_file): Call output_asm_line_debug_info for test.
7383 (dwarf2out_var_location): Reset views as needed. Precompute
7384 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7385 attribute. Set view.
7386 (new_line_info_table): Reset next view.
7387 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7388 (dwarf2out_source_line): Likewise. Output view resets and labels to
7389 the assembler, or select appropriate line info opcodes.
7390 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7391 (optimize_string_length): Catch it. Adjust.
7392 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7393 dw_val_class_view_list, and remove it if no longer needed.
7394 (hash_loc_list): Hash view numbers.
7395 (loc_list_hasher::equal): Compare them.
7396 (optimize_location_lists): Check whether a view list symbol is
7397 needed, and whether the locview attribute is present, and
7398 whether they match. Remove the locview attribute if no longer
7400 (index_location_lists): Call skip_loc_list_entry for test.
7401 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7402 Use output_dwarf_version.
7403 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7404 (struct dw_val_node): Add val_view_list.
7405 * final.c (SEEN_NEXT_VIEW): New.
7406 (set_next_view_needed): New.
7407 (clear_next_view_needed): New.
7408 (maybe_output_next_view): New.
7409 (final_start_function): Rename to...
7410 (final_start_function_1): ... this. Take pointer to FIRST,
7411 add SEEN parameter. Emit param bindings in the initial view.
7412 (final_start_function): Reintroduce SEEN-less interface.
7413 (final): Rename to...
7414 (final_1): ... this. Take SEEN parameter. Output final pending
7415 next view at the end.
7416 (final): Reintroduce seen-less interface.
7417 (final_scan_insn): Output pending next view before switching
7418 sections or ending a block. Mark the next view as needed when
7419 outputting variable locations. Notify debug backend of section
7420 changes, and of location view changes.
7421 (rest_of_handle_final): Adjust.
7422 * toplev.c (process_options): Autodetect value for debug variable
7423 location views option. Warn on incompat5 without -gdwarf-5.
7424 * doc/invoke.texi (gvariable-location-views): New.
7425 (gvariable-location-views=incompat5): New.
7426 (gno-variable-location-views): New.
7428 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7430 PR tree-optimization/84136
7431 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7432 that the result of find_edge is non-NULL.
7434 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7437 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7438 storing integer register in SImode. Fix cost of 256 and 512
7439 byte aligned SSE register store.
7441 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7443 * config/i386/i386.c (ix86_multiplication_cost): Fix
7444 multiplication cost for TARGET_AVX512DQ.
7446 2018-02-08 Marek Polacek <polacek@redhat.com>
7448 PR tree-optimization/84238
7449 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7452 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7454 PR tree-optimization/84265
7455 * tree-vect-stmts.c (vectorizable_store): Don't treat
7456 VMAT_CONTIGUOUS accesses as grouped.
7457 (vectorizable_load): Likewise.
7459 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7461 PR tree-optimization/81635
7462 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7463 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7464 (test_round_for_mask): New functions.
7465 (wide_int_cc_tests): Call test_round_for_mask.
7466 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7467 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7468 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7469 range returned by get_range_info.
7471 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7474 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7475 * symtab.c: Include builtins.h
7476 (symtab_node::output_to_lto_symbol_table_p): Move here
7477 from lto-streamer-out.c:output_symbol_p.
7478 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7479 (output_symbol_p): Move all logic to symtab.c
7480 (produce_symtab): Update.
7482 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7484 * config/s390/s390-opts.h (enum indirect_branch): Define.
7485 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7486 (s390_indirect_branch_via_thunk)
7487 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7488 (enum s390_indirect_branch_type): Define.
7489 * config/s390/s390.c (struct s390_frame_layout, struct
7490 machine_function): Remove.
7491 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7492 (indirect_branch_table_label_no, indirect_branch_table_name):
7494 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7495 (enum s390_indirect_branch_option): Define.
7496 (s390_return_addr_from_memory): New function.
7497 (s390_handle_string_attribute): New function.
7498 (s390_attribute_table): Add new attribute handler.
7499 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7500 (s390_indirect_branch_via_thunk): New function.
7501 (s390_indirect_branch_via_inline_thunk): New function.
7502 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7503 sibling call optimization for non z10 compiles.
7504 (s390_emit_call): Force indirect branch target to be a single
7505 register. Add r1 clobber for non-z10 compiles.
7506 (s390_emit_epilogue): Emit return jump via return_use expander.
7507 (s390_reorg): Handle JUMP_INSNs as execute targets.
7508 (s390_option_override_internal): Perform validity checks for the
7509 new command line options.
7510 (s390_indirect_branch_attrvalue): New function.
7511 (s390_indirect_branch_settings): New function.
7512 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7513 (s390_output_indirect_thunk_function): New function.
7514 (s390_code_end): Implement target hook.
7515 (s390_case_values_threshold): Implement target hook.
7516 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7518 * config/s390/s390.h (struct s390_frame_layout)
7519 (struct machine_function): Move here from s390.c.
7520 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7521 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7522 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7523 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7524 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7525 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7526 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7527 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7528 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7529 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7530 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7531 (mnemonic attribute): Add values which aren't recognized
7533 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7534 pattern for branch conversion. Fix mnemonic attribute.
7535 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7536 indirect branch via thunk if requested.
7537 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7538 ("*indirect_jump"): Disable for branch conversion using out of
7540 ("indirect_jump_via_thunk<mode>_z10")
7541 ("indirect_jump_via_thunk<mode>")
7542 ("indirect_jump_via_inlinethunk<mode>_z10")
7543 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7544 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7545 ("casesi_jump_via_inlinethunk<mode>_z10")
7546 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7547 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7548 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7549 ("*indirect2_jump"): Disable for branch conversion.
7550 ("casesi_jump"): Turn into expander and expand patterns for branch
7552 ("return_use"): New expander.
7553 ("*return"): Emit return via thunk and rename it to ...
7554 ("*return<mode>"): ... this one.
7555 * config/s390/s390.opt: Add new options and and enum for the
7558 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7560 * lra-constraints.c (match_reload): Unconditionally use
7561 gen_lowpart_SUBREG, rather than selecting between that
7562 and equivalent gen_rtx_SUBREG code.
7564 2018-02-08 Richard Biener <rguenther@suse.de>
7566 PR tree-optimization/84233
7567 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7568 changed flag instead of boguously re-using phi_inserted.
7570 2018-02-08 Martin Jambor <mjambor@suse.cz>
7572 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7573 static local variables.
7575 2018-02-08 Richard Biener <rguenther@suse.de>
7577 PR tree-optimization/84278
7578 * tree-vect-stmts.c (vectorizable_store): When looking for
7579 smaller vector types to perform grouped strided loads/stores
7580 make sure the mode is supported by the target.
7581 (vectorizable_load): Likewise.
7583 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7585 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7586 Increase LDP/STP opportunities by adding adjacent callee-saves.
7588 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7590 PR rtl-optimization/84068
7591 PR rtl-optimization/83459
7592 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7594 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7596 PR tree-optimization/84224
7597 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7598 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7601 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7604 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7605 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7606 count, remove one USE.
7608 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7610 * doc/install.texi (Configuration): Document the
7611 --with-long-double-format={ibm,ieee} PowerPC configuration
7615 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7616 Convert from define_expand to be define_insn_and_split. Rework
7617 float/double/_Float128 conversions to QI/HI/SImode to work with
7618 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7619 conversions to QI/HImode types did a store and then a load to
7620 truncate the value. For conversions to VSX registers, don't split
7621 the insn, instead emit the code directly. Use the code iterator
7622 any_fix to combine signed and unsigned conversions.
7623 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7624 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7625 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7626 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7627 (fix_<mode>di2_hw): Likewise.
7628 (fixuns_<mode>di2_hw): Likewise.
7629 (fix_<mode>si2_hw): Likewise.
7630 (fixuns_<mode>si2_hw): Likewise.
7631 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7632 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7633 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7634 fix<uns>_trunc<SFDF:mode>si2_p8.
7635 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7637 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7638 (fix<uns>_<mode>_mem): Likewise.
7639 (fctiw<u>z_<mode>_mem): Likewise.
7640 (fix<uns>_<mode>_mem): Likewise.
7641 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7642 the register allocator from doing a direct move to the GPRs to do
7643 a store, and instead use the ISA 3.0 store byte/half-word from
7644 vector register instruction. For IEEE 128-bit floating point,
7645 also optimize stores of 32-bit ints.
7646 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7648 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7650 * genextract.c (push_pathstr_operand): New function to support
7652 (walk_rtx): Call push_pathstr_operand.
7653 (print_path): Support [a-zA-Z].
7655 2018-02-07 Richard Biener <rguenther@suse.de>
7657 PR tree-optimization/84037
7658 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7659 (cse_and_gimplify_to_preheader): Declare.
7660 (vect_get_place_in_interleaving_chain): Likewise.
7661 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7663 (_loop_vec_info::~_loop_vec_info): Delete it.
7664 (cse_and_gimplify_to_preheader): New function.
7665 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7666 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7667 (vectorizable_load): Likewise. For grouped stores always base
7668 the IV on the first element.
7669 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7670 condition before gimplifying.
7672 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7674 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7675 *DIV_EXPR and *MOD_EXPR.
7677 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7680 * config/i386/i386.c (ix86_option_override_internal): Mask out
7681 the CF_SET bit when checking -fcf-protection.
7683 2018-02-07 Tom de Vries <tom@codesourcery.com>
7686 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7689 2018-02-07 Richard Biener <rguenther@suse.de>
7691 PR tree-optimization/84204
7692 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7695 PR tree-optimization/84205
7696 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7697 special-case isl_ast_op_zdiv_r.
7699 PR tree-optimization/84223
7700 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7701 Only add conditions from within the region.
7702 (gather_bbs::after_dom_children): Adjust.
7704 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7707 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7708 * config/avr/avr.md: Only post-reload split REG-REG moves if
7709 either register is GENERAL_REG_P.
7711 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7713 PR tree-optimization/84235
7714 * tree-ssa-scopedtables.c
7715 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7716 if the subtraction is performed in floating point type where NaNs are
7717 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7718 build 1. Formatting fix.
7720 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7723 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7724 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7725 and skip it regardless of bb boundaries. Use CALL_P macro,
7726 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7727 unnecessarily, formatting fix.
7729 2018-02-06 Michael Collison <michael.collison@arm.com>
7731 * config/arm/thumb2.md:
7732 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7733 (*thumb_mov_notscc): Ditto.
7735 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7738 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7739 unsigned_fix, not "s".
7741 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7743 * configure.ac (gcc_fn_eh_frame_ro): New function.
7744 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7745 correct .eh_frame permissions.
7746 * configure: Regenerate.
7748 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7750 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7753 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7755 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7756 Display warning message for -mno-speculate-indirect-jumps.
7758 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7760 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7762 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7764 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7766 PR tree-optimization/84225
7767 * tree-eh.c (find_trapping_overflow): Only call
7768 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7770 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7773 * config/i386/i386.c: Reimplement the check of possible options
7774 -mibt/-mshstk conbination. Change error messages.
7775 * doc/invoke.texi: Fix a typo: remove extra '='.
7777 2018-02-06 Marek Polacek <polacek@redhat.com>
7779 PR tree-optimization/84228
7780 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7782 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7785 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7786 emitted arch directives.
7787 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7788 __ARM_FEATURE_COPROC before changing architectures.
7790 2018-02-06 Richard Biener <rguenther@suse.de>
7792 * config/i386/i386.c (print_reg): Fix typo.
7793 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7795 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7797 * configure: Regenerate.
7799 2018-02-05 Martin Sebor <msebor@redhat.com>
7801 PR tree-optimization/83369
7802 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7805 2018-02-05 Martin Liska <mliska@suse.cz>
7807 * doc/invoke.texi: Cherry-pick upstream r323995.
7809 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7811 * ira.c (ira_init_register_move_cost): Adjust comment.
7813 2018-02-05 Martin Liska <mliska@suse.cz>
7815 PR gcov-profile/84137
7816 * doc/gcov.texi: Fix typo in documentation.
7818 2018-02-05 Martin Liska <mliska@suse.cz>
7820 PR gcov-profile/83879
7821 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7822 using dlopen functionality.
7824 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7826 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7827 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7828 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7829 _mm_maskz_range_round_ss): New intrinsics.
7830 (__builtin_ia32_rangesd128_round)
7831 (__builtin_ia32_rangess128_round): Remove.
7832 (__builtin_ia32_rangesd128_mask_round,
7833 __builtin_ia32_rangess128_mask_round): New builtins.
7834 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7835 __builtin_ia32_rangess128_round): Remove.
7836 (__builtin_ia32_rangesd128_mask_round,
7837 __builtin_ia32_rangess128_mask_round): New builtins.
7838 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7839 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7840 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7841 "<round_saeonly_constraint>")): Changed to ...
7842 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7843 "<round_saeonly_scalar_constraint>")): ... this.
7844 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7845 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7846 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7847 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7848 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7850 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7852 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7854 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7855 Remove all values except native, 8540 and 8548.
7857 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7859 * config/i386/i386.c (ix86_output_function_return): Pass
7860 INVALID_REGNUM, instead of -1, as invalid register number to
7861 indirect_thunk_name and output_indirect_thunk.
7863 2018-02-02 Julia Koval <julia.koval@intel.com>
7865 * config.gcc: Add -march=icelake.
7866 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7867 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7868 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7869 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7870 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7871 (processor_target_table): Add icelake.
7872 (ix86_option_override_internal): Handle new PTAs.
7873 (get_builtin_code_for_version): Handle icelake.
7874 (M_INTEL_COREI7_ICELAKE): New.
7875 (fold_builtin_cpu): Handle icelake.
7876 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7877 * doc/invoke.texi: Add -march=icelake.
7879 2018-02-02 Julia Koval <julia.koval@intel.com>
7881 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7882 to wide_int_bitmask.
7883 * wide-int-bitmask.h: New.
7885 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7888 * config/i386/i386.md: Replace Pmode with word_mode in
7889 builtin_setjmp_setup and builtin_longjmp to support x32.
7891 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7895 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7897 (rs6000_supported_cpu_names): New static variable.
7898 (linux_cpu_translation_table): Likewise.
7899 (elf_platform) <cpu>: Define new static variable and use it.
7900 Translate kernel AT_PLATFORM name to canonical name if needed.
7901 Error if platform name is unknown.
7903 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7906 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7908 2018-02-01 Jeff Law <law@redhat.com>
7911 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7912 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7913 the scratch if RELEASE_VIA_POP is false.
7914 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7915 If we have to save a temporary register, decrement SIZE appropriately.
7916 Pass new arguments to release_scratch_register_on_entry.
7917 (ix86_adjust_stack_and_probe): Likewise.
7918 (ix86_emit_probe_stack_range): Pass new arguments to
7919 release_scratch_register_on_entry.
7921 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7923 PR rtl-optimization/84157
7924 * combine.c (change_zero_ext): Use REG_P predicate in
7925 front of HARD_REGISTER_P predicate.
7927 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7929 * config/avr/avr.c (avr_option_override): Move disabling of
7930 -fdelete-null-pointer-checks to...
7931 * common/config/avr/avr-common.c (avr_option_optimization_table):
7934 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7936 PR tree-optimization/81635
7937 * tree-data-ref.c (split_constant_offset_1): For types that
7938 wrap on overflow, try to use range info to prove that wrapping
7941 2018-02-01 Renlin Li <renlin.li@arm.com>
7944 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7946 (aarch64_register_move_cost): Likewise.
7947 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7949 (REG_CLASS_NAMES): Likewise.
7950 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7951 TAILCALL_ADDR_REGS. Remove IP registers.
7952 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7954 2018-02-01 Richard Biener <rguenther@suse.de>
7956 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7957 for specifying RPO order and allow NULL for that.
7958 * domwalk.c (dom_walker::dom_walker): Likewise.
7959 (dom_walker::walk): Handle NULL RPO order.
7960 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7962 (rewrite_update_dom_walker): Likewise.
7963 (mark_def_dom_walker): Likewise.
7965 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7967 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7968 (aarch64_maybe_expand_sve_subreg_move): Declare.
7969 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7970 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7972 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7973 that are semantically a reverse operation.
7974 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7975 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7976 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7978 (aarch64_can_change_mode_class): For big-endian, forbid changes
7979 between two SVE modes if they have different element sizes.
7981 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7983 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7984 the TImode handling for big-endian targets.
7986 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7988 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7989 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7991 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7992 Remove BSWAP handing for big-endian targets and use the form of
7993 LD1RQ appropariate for the mode.
7995 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7997 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7998 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8001 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8004 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8005 check for operands that need to go through aarch64_sve_reload_be.
8007 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8009 PR tree-optimization/81661
8010 PR tree-optimization/84117
8011 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8012 * tree-eh.c: Include gimplify.h.
8013 (find_trapping_overflow, replace_trapping_overflow,
8014 rewrite_to_non_trapping_overflow): New functions.
8015 * tree-vect-loop.c: Include tree-eh.h.
8016 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8017 * tree-data-ref.c: Include tree-eh.h.
8018 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8020 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8022 PR rtl-optimization/84123
8023 * combine.c (change_zero_ext): Check if hard register satisfies
8024 can_change_dest_mode before calling gen_lowpart_SUBREG.
8026 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8029 * ira.c (ira_init_register_move_cost): Remove assert.
8031 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8033 PR rtl-optimization/84071
8034 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8035 * doc/tm.texi: Regenerate.
8037 2018-01-31 Richard Biener <rguenther@suse.de>
8039 PR tree-optimization/84132
8040 * tree-data-ref.c (analyze_miv_subscript): Properly
8041 check whether evolution_function_is_affine_multivariate_p
8042 before calling gcd_of_steps_may_divide_p.
8044 2018-01-31 Julia Koval <julia.koval@intel.com>
8047 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8048 * config/i386/i386.md (rdpid_rex64) New.
8049 (rdpid): Make 32bit only.
8051 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8054 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8055 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8057 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8060 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8062 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8064 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8066 PR rtl-optimization/84071
8067 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8068 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8070 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8072 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8073 (arc_attribute_table): Add 'aux' attribute.
8074 (arc_in_small_data_p): Consider aux like variables.
8075 (arc_is_aux_reg_p): New function.
8076 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8077 (arc_get_aux_arg): New function.
8078 (prepare_move_operands): Handle aux-register access.
8079 (arc_handle_aux_attribute): New function.
8080 * doc/extend.texi (ARC Variable attributes): Add subsection.
8082 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8084 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8085 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8086 (arc_attribute_table): Add 'uncached' attribute.
8087 (arc_print_operand): Print '.di' flag for uncached memory
8089 (arc_in_small_data_p): Do not consider for small data the uncached
8091 (arc_is_uncached_mem_p): New function.
8092 * config/arc/predicates.md (compact_store_memory_operand): Check
8093 for uncached memory accesses.
8094 (nonvol_nonimm_operand): Likewise.
8095 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8097 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8100 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8101 falign-loops=): Add Optimization flag.
8103 2018-01-30 Jeff Law <law@redhat.com>
8106 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8107 INT_REGISTERS_SAVED. Check it prior to calling
8108 get_scratch_register_on_entry.
8109 (ix86_adjust_stack_and_probe): Similarly.
8110 (ix86_emit_probe_stack_range): Similarly.
8111 (ix86_expand_prologue): Corresponding changes.
8113 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8116 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8117 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8119 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8122 * lra-constraints.c (curr_insn_transform): Process AND in the
8125 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8127 PR rtl-optimization/83986
8128 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8129 dependence against last_pending_memory_flush in addition to
8132 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8134 PR tree-optimization/81611
8135 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8138 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8141 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8144 2018-01-30 Richard Biener <rguenther@suse.de>
8145 Jakub Jelinek <jakub@redhat.com>
8147 PR tree-optimization/84111
8148 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8149 inner loops added during recursion, as they don't have up-to-date
8152 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8155 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8156 (can_inline_edge_by_limits_p): ... here.
8157 (can_early_inline_edge_p, check_callers,
8158 update_caller_keys, update_callee_keys, recursive_inlining,
8159 add_new_edges_to_heap, speculation_useful_p,
8160 inline_small_functions,
8161 inline_small_functions, flatten_function,
8162 inline_to_all_callers_1): Update.
8164 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8166 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8169 2018-01-30 Richard Biener <rguenther@suse.de>
8171 PR tree-optimization/83008
8172 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8173 invariant and constant vector uses in stmts when they need
8176 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8179 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8180 * configure: Regenerate.
8182 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8184 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8186 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8187 Use gen_rtx_REG rather than gen_lowpart.
8189 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8191 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8192 rather than 0 when creating partial subregs.
8194 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8196 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8199 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8202 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8203 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8204 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8205 flags. This restores the settings used before the 2017-07-24.
8206 Turning off pre increment/decrement/modify allows IVOPTS to
8207 optimize DF/SF loops where the index is an int.
8209 2018-01-29 Richard Biener <rguenther@suse.de>
8210 Kelvin Nilsen <kelvin@gcc.gnu.org>
8213 * tree-vect-stmts.c (vectorizable_call): Don't call
8214 targetm.vectorize_builtin_md_vectorized_function if callee is
8217 2018-01-22 Carl Love <cel@us.ibm.com>
8219 * doc/extend.tex: Fix typo in second arg in
8220 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8222 2018-01-29 Richard Biener <rguenther@suse.de>
8224 PR tree-optimization/84086
8225 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8226 (flush_ssaname_freelist): When SSA names were released reset
8227 the SCEV hash table.
8229 2018-01-29 Richard Biener <rguenther@suse.de>
8231 PR tree-optimization/84057
8232 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8233 removed paths when removing edges.
8235 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8237 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8238 -mfunction-return=@var{choice}.
8240 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8243 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8245 (layout::print_source_line): Likewise.
8246 (test_get_line_width_without_trailing_whitespace): Add test cases.
8248 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8251 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8254 2018-01-26 Jim Wilson <jimw@sifive.com>
8256 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8258 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8261 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8263 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8264 and CMP + SUB-immediate -> SUBS.
8266 2018-01-26 Martin Sebor <msebor@redhat.com>
8268 PR tree-optimization/83896
8269 * tree-ssa-strlen.c (get_string_len): Rename...
8270 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8271 Avoid assuming length is constant.
8272 (handle_char_store): Use HOST_WIDE_INT for string length.
8274 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8277 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8278 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8280 2018-01-26 Richard Biener <rguenther@suse.de>
8282 PR rtl-optimization/84003
8283 * dse.c (record_store): Only record redundant stores when
8284 the earlier store aliases at least all accesses the later one does.
8286 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8288 PR rtl-optimization/83985
8289 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8290 REG_CFA_RESTORE insns.
8291 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8292 REG_CFA_RESTORE insns here.
8295 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8296 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8298 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8300 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8301 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8302 (arc_init): Likewise.
8303 (arc_override_options): Likewise.
8304 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8306 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8308 * config/arc/arc.h (TARGET_DBNZ): Define.
8309 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8310 properly set the tune attribute.
8311 (dbnz): Use TARGET_DBNZ guard.
8312 * config/arc/arc.opt (mtune): Add core3 option.
8314 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8316 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8317 recognize new pic like addresses.
8318 (arc_delegitimize_address): Clean up.
8320 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8322 * config/arc/arc-arches.def: Option mrf16 valid for all
8324 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8325 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8326 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8327 * config/arc/arc-tables.opt: Regenerate.
8328 * config/arc/arc.c (arc_conditional_register_usage): Handle
8329 reduced register file case.
8330 (arc_file_start): Set must have build attributes.
8331 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8333 * config/arc/arc.opt (mrf16): Add new option.
8334 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8335 * config/arc/genmultilib.awk: Handle new mrf16 option.
8336 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8337 * config/arc/t-multilib: Regenerate.
8338 * doc/invoke.texi (ARC Options): Document mrf16 option.
8340 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8342 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8343 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8344 (arc_attribute_table): Add 'secure_call' attribute.
8345 (arc_print_operand): Print secure call operand.
8346 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8348 (arc_is_secure_call_p): New function. * config/arc/arc.md
8349 (call_i): Add support for sjli instruction.
8350 (call_value_i): Likewise.
8351 * config/arc/constraints.md (Csc): New constraint.
8353 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8354 John Eric Martin <John.Martin@emmicro-us.com>
8356 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8357 * config/arc/arc.c (_arc_jli_section): New struct.
8358 (arc_jli_section): New type.
8359 (rc_jli_sections): New static variable.
8360 (arc_handle_jli_attribute): New function.
8361 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8362 (arc_file_end): New function.
8363 (TARGET_ASM_FILE_END): Define.
8364 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8365 (arc_add_jli_section): New function.
8366 (jli_call_scan): Likewise.
8367 (arc_reorg): Call jli_call_scan.
8368 (arc_output_addsi): Remove 'S' from printing asm operand.
8369 (arc_is_jli_call_p): New function.
8370 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8372 (movhi_insn): Likewise.
8373 (movsi_insn): Likewise.
8374 (movsi_set_cc_insn): Likewise.
8375 (loadqi_update): Likewise.
8376 (load_zeroextendqisi_update): Likewise.
8377 (load_signextendqisi_update): Likewise.
8378 (loadhi_update): Likewise.
8379 (load_zeroextendhisi_update): Likewise.
8380 (load_signextendhisi_update): Likewise.
8381 (loadsi_update): Likewise.
8382 (loadsf_update): Likewise.
8383 (movsicc_insn): Likewise.
8384 (bset_insn): Likewise.
8385 (bxor_insn): Likewise.
8386 (bclr_insn): Likewise.
8387 (bmsk_insn): Likewise.
8388 (bicsi3_insn): Likewise.
8389 (cmpsi_cc_c_insn): Likewise.
8390 (movsi_ne): Likewise.
8391 (movsi_cond_exec): Likewise.
8392 (clrsbsi2): Likewise.
8403 (call_i): Remove 'S' asm letter, add jli instruction.
8404 (call_value_i): Likewise.
8405 * config/arc/arc.op (mjli-always): New option.
8406 * config/arc/constraints.md (Cji): New constraint.
8407 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8409 (subsf3_fpx): Likewise.
8410 (mulsf3_fpx): Likewise.
8411 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8413 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8415 * doc/invoke.texi (ARC): Document mjli-always option.
8417 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8419 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8420 avoid addition with 0 and use incw and decw where possible.
8422 2018-01-26 Richard Biener <rguenther@suse.de>
8424 PR tree-optimization/81082
8425 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8426 association if it requires casting to unsigned.
8427 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8428 from fold_plusminus_mult_expr to catch important cases late when
8429 range info is available.
8431 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8433 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8434 * configure.ac (hidden_linkonce): New test.
8435 * configure: Regenerate.
8436 * config.in: Regenerate.
8438 2018-01-26 Julia Koval <julia.koval@intel.com>
8440 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8441 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8442 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8443 _mm_mask_bitshuffle_epi64_mask): Fix type.
8444 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8445 USI_FTYPE_V4DI_V4DI_USI): Remove.
8446 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8447 __builtin_ia32_vpshufbitqmb256_mask,
8448 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8449 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8450 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8452 2018-01-26 Alan Modra <amodra@gmail.com>
8455 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8456 UNSPEC_VBPERMQ. Sort other unspecs.
8458 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8460 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8462 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8465 * predict.c (drop_profile): Do not push/pop cfun; update also
8467 (handle_missing_profiles): Fix logic looking for zero profiles.
8469 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8472 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8473 on functions with #pragma omp declare simd or functions with simd
8475 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8476 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8477 Remove trailing \n from warning_at calls.
8479 2018-01-25 Tom de Vries <tom@codesourcery.com>
8482 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8483 for neutered workers.
8485 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8488 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8489 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8491 2018-01-24 Jeff Law <law@redhat.com>
8494 * i386.c (get_probe_interval): Move to earlier point.
8495 (ix86_compute_frame_layout): If -fstack-clash-protection and
8496 the frame is larger than the probe interval, then use pushes
8497 to save registers rather than reg->mem moves.
8498 (ix86_expand_prologue): Remove conditional for int_registers_saved
8501 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8504 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8505 min/max for never referenced object.
8507 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8510 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8512 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8513 attributes from DECL_ATTRIBUTES (decl) without affecting
8514 DECL_ATTRIBUTES (current_function_decl).
8515 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8516 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8518 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8520 PR tree-optimization/83979
8521 * fold-const.c (fold_comparison): Use constant_boolean_node
8522 instead of boolean_{true,false}_node.
8524 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8526 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8529 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8531 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8532 Simplify the clause that sets the length attribute.
8533 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8534 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8535 clause that sets the length attribute.
8536 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8538 2018-01-24 Tom de Vries <tom@codesourcery.com>
8541 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8542 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8543 Add strict parameter.
8544 (prevent_branch_around_nothing): Insert dummy insn between branch to
8545 label and label with no ptx insn inbetween.
8546 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8548 2018-01-24 Tom de Vries <tom@codesourcery.com>
8551 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8552 for neutered threads in warp.
8553 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8555 2018-01-24 Richard Biener <rguenther@suse.de>
8557 PR tree-optimization/83176
8558 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8561 2018-01-24 Richard Biener <rguenther@suse.de>
8563 PR tree-optimization/82819
8564 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8565 code generating pluses that are no-ops in the target precision.
8567 2018-01-24 Richard Biener <rguenther@suse.de>
8570 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8572 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8574 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8575 to merge probabilities.
8576 * predict.c (probably_never_executed): Also mark as cold functions
8577 with global 0 profile and guessed local profile.
8578 * profile-count.c (profile_probability::combine_with_count): New
8580 * profile-count.h (profile_probability::operator*,
8581 profile_probability::operator*=, profile_probability::operator/,
8582 profile_probability::operator/=): Reduce precision to adjusted
8583 and set value to guessed on contradictory divisions.
8584 (profile_probability::combine_with_freq): Remove.
8585 (profile_probability::combine_wiht_count): Declare.
8586 (profile_count::force_nonzero):: Set to adjusted.
8587 (profile_count::probability_in):: Set quality to adjusted.
8588 * tree-ssa-tail-merge.c (replace_block_by): Use
8591 2018-01-23 Andrew Waterman <andrew@sifive.com>
8592 Jim Wilson <jimw@sifive.com>
8594 * config/riscv/riscv.c (riscv_stack_boundary): New.
8595 (riscv_option_override): Set riscv_stack_boundary. Handle
8596 riscv_preferred_stack_boundary_arg.
8597 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8598 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8599 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8600 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8601 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8602 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8604 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8607 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8608 of struct ix86_frame.
8609 (ix86_expand_epilogue): Likewise. Add a local variable for
8610 the reg_save_offset field in struct ix86_frame.
8612 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8614 PR tree-optimization/82604
8615 * tree-loop-distribution.c (enum partition_kind): New enum item
8616 PKIND_PARTIAL_MEMSET.
8617 (partition_builtin_p): Support above new enum item.
8618 (generate_code_for_partition): Ditto.
8619 (compute_access_range): Differentiate cases that equality can be
8620 proven at all loops, the innermost loops or no loops.
8621 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8622 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8623 (finalize_partitions, distribute_loop): Don't fuse partition of
8624 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8625 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8628 2018-01-23 Martin Liska <mliska@suse.cz>
8630 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8631 order to ignore the predictor.
8632 (PRED_POLYMORPHIC_CALL): Likewise.
8633 (PRED_RECURSIVE_CALL): Likewise.
8635 2018-01-23 Martin Liska <mliska@suse.cz>
8637 * tree-profile.c (tree_profiling): Print function header to
8638 aware reader which function we are working on.
8639 * value-prof.c (gimple_find_values_to_profile): Do not print
8640 not interesting value histograms.
8642 2018-01-23 Martin Liska <mliska@suse.cz>
8644 * profile-count.h (enum profile_quality): Add
8645 profile_uninitialized as the first value. Do not number values
8646 as they are zero based.
8647 (profile_count::verify): Update sanity check.
8648 (profile_probability::verify): Likewise.
8650 2018-01-23 Nathan Sidwell <nathan@acm.org>
8652 * doc/invoke.texi (ffor-scope): Deprecate.
8654 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8656 PR tree-optimization/83510
8657 * domwalk.c (set_all_edges_as_executable): New function.
8658 (dom_walker::dom_walker): Convert bool param
8659 "skip_unreachable_blocks" to enum reachability. Move setup of
8660 edge flags to set_all_edges_as_executable and only do it when
8661 reachability is REACHABLE_BLOCKS.
8662 * domwalk.h (enum dom_walker::reachability): New enum.
8663 (dom_walker::dom_walker): Convert bool param
8664 "skip_unreachable_blocks" to enum reachability.
8665 (set_all_edges_as_executable): New decl.
8666 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8667 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8669 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8670 but converting true to REACHABLE_BLOCKS.
8671 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8673 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8674 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8675 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8677 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8678 if check_all_array_refs will be called.
8680 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8682 * tree.c (selftest::test_location_wrappers): Add more test
8685 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8687 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8688 (selftest::test_bit_in_range): Likewise.
8690 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8693 * doc/sourcebuild.texi (vect_float): Say that the selector
8694 only describes the situation when -funsafe-math-optimizations is on.
8695 (vect_float_strict): Document.
8697 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8699 PR tree-optimization/83965
8700 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8701 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8702 instead of checking only for a reduction.
8703 (vect_recog_widen_sum_pattern): Likewise.
8705 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8707 * predict.c (probably_never_executed): Only use precise profile info.
8708 (compute_function_frequency): Skip after inlining hack since we now
8709 have quality checking.
8711 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8713 * profile-count.h (profile_probability::very_unlikely,
8714 profile_probability::unlikely, profile_probability::even): Set
8715 precision to guessed.
8717 2018-01-23 Richard Biener <rguenther@suse.de>
8719 PR tree-optimization/83963
8720 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8721 Properly terminate dominator walk when crossing the exit edge not
8722 when visiting its source block.
8724 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8727 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8728 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8730 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8732 PR tree-optimization/83957
8733 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8734 semicolon after for body surrounded by braces.
8736 PR tree-optimization/83081
8737 * profile-count.h (profile_probability::split): New method.
8738 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8739 Use profile_probability::split.
8740 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8741 when splitting a single conditional jump into 2.
8743 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8745 PR tree-optimization/69452
8746 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8749 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8751 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8752 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8753 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8755 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8757 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8758 * config/rl78/rl78.md (movdi): New define_expand.
8759 * config/rl78/rl78.c (rl78_split_movdi): New function.
8761 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8764 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8766 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8767 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8768 128-bit to produce an UNSPEC move to get the double word with the
8769 signbit and then a shift directly to do signbit.
8770 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8771 implementation with a new version that just does either a direct
8772 move or a regular move. Move memory interface to separate insns.
8773 Move insns so they are next to the expander.
8774 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8775 with signbit move. Split big and little endian case.
8776 (signbit<mode>2_dm_mem_le): Likewise.
8777 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8778 (signbit<mode>2_dm2): Likewise.
8780 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8782 * config/rl78/rl78.md (anddi3): New define_expand.
8784 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8786 * config/rl78/rl78.md (umindi3): New define_expand.
8788 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8790 * config/rl78/rl78.md (smindi3): New define_expand.
8792 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8794 * config/rl78/rl78.md (smaxdi3): New define_expand.
8796 2018-01-22 Carl Love <cel@us.ibm.com>
8798 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8799 LVX_V1TI): Add macro expansion.
8800 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8801 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8802 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8803 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8804 Change check to determine if the instruction is a byte reversing
8805 entry. Fix typo in comment.
8806 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8807 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8808 Add def_builtin calls for new builtins.
8809 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8810 Add define_insn expansion.
8812 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8814 * config/rl78/rl78.md (umaxdi3): New define_expand.
8816 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8818 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8819 for non-QImode registers.
8821 2018-01-22 Richard Biener <rguenther@suse.de>
8823 PR tree-optimization/83963
8824 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8825 including the loop exit block.
8826 (scop_detection::merge_sese): Likewise.
8827 (scop_detection::add_scop): Do it here instead.
8829 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8831 * doc/sourcebuild.texi (arm_softfloat): Document.
8833 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8836 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8837 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8838 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8840 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8841 David Edelsohn <dje.gcc@gmail.com>
8844 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8845 Change "crset eq" to "crset 2".
8846 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8847 (*call_indirect_aix<mode>_nospec): Likewise.
8848 (*call_value_indirect_aix<mode>_nospec): Likewise.
8849 (*call_indirect_elfv2<mode>_nospec): Likewise.
8850 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8851 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8852 change assembly output from . to $.
8853 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8854 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8855 (*tablejump<mode>_internal1_nospec): Likewise.
8857 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8860 * config/sh/sh_optimize_sett_clrt.cc:
8861 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8863 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8865 PR tree-optimization/83940
8866 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8867 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8868 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8869 use it to pass back the definition type.
8870 (vect_check_store_rhs): Likewise rhs_dt_out.
8871 (vect_build_gather_load_calls): Add a mask_dt argument and use
8872 it instead of a call to vect_is_simple_use.
8873 (vectorizable_store): Update calls to vect_check_load_store_mask
8874 and vect_check_store_rhs. Use the dt returned by the latter instead
8875 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8876 instead of calls to vect_is_simple_use. Pass the scalar rather
8877 than the vector operand to vect_is_simple_use when handling
8878 second and subsequent copies of an rhs value.
8879 (vectorizable_load): Update calls to vect_check_load_store_mask
8880 and vect_build_gather_load_calls. Use the cached mask_dt and
8881 gs_info.offset_dt instead of calls to vect_is_simple_use.
8883 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8886 * tree-emutls.c: Include gimplify.h.
8887 (lower_emutls_2): New function.
8888 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8889 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8890 it before further processing.
8893 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8894 UINTVAL (trueop1) instead of INTVAL (op1).
8896 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8900 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8901 INCOMING_FRAME_SP_OFFSET if not defined.
8902 (scan_trace): Add ENTRY argument. If true and
8903 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8904 emit a note to adjust the CFA offset.
8905 (create_cfi_notes): Adjust scan_trace callers.
8906 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8907 INCOMING_FRAME_SP_OFFSET in the CIE.
8908 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8909 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8911 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8912 * doc/tm.texi: Regenerated.
8914 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8916 PR rtl-optimization/83147
8917 * lra-constraints.c (remove_inheritance_pseudos): Use
8918 lra_substitute_pseudo_within_insn.
8920 2018-01-19 Tom de Vries <tom@codesourcery.com>
8921 Cesar Philippidis <cesar@codesourcery.com>
8924 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8926 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8929 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8930 spaces for function labels.
8932 2018-01-19 Martin Liska <mliska@suse.cz>
8934 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8935 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8936 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8937 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8938 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8939 (PRED_CONST_RETURN): Change from 69 to 65.
8940 (PRED_NULL_RETURN): Change from 91 to 71.
8941 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8942 (PRED_LOOP_GUARD): Change from 66 to 73.
8944 2018-01-19 Martin Liska <mliska@suse.cz>
8946 * predict.c (predict_insn_def): Add new assert.
8947 (struct branch_predictor): Change type to signed integer.
8948 (test_prediction_value_range): Amend test to cover
8950 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8951 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8952 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8953 (PRED_LOOP_IV_COMPARE): Likewise.
8954 * predict.h (PROB_UNINITIALIZED): Define new constant.
8956 2018-01-19 Martin Liska <mliska@suse.cz>
8958 * predict.c (dump_prediction): Add new format for
8959 analyze_brprob.py script which is enabled with -details
8961 * profile-count.h (precise_p): New function.
8963 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8965 PR tree-optimization/83922
8966 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8967 there are no statements that need masking.
8968 (vect_active_double_reduction_p): New function.
8969 (vect_analyze_loop_operations): Use it when handling phis that
8970 are not in the loop header.
8972 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8974 PR tree-optimization/83914
8975 * tree-vect-loop.c (vectorizable_induction): Don't convert
8976 init_expr or apply the peeling adjustment for inductions
8977 that are nested within the vectorized loop.
8979 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8981 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8984 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8988 * function.h (gimplify_parameters): Add gimple_seq * argument.
8989 * function.c: Include gimple.h and options.h.
8990 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8991 for the added local temporaries if needed.
8992 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8993 if there are any parameter cleanups, wrap whole body into a
8994 try/finally with the cleanups.
8996 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8999 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9000 Use GET_MODE_CLASS for scalar floating point.
9002 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9006 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9007 Fix call of call_cgraph_insertion_hooks.
9009 2018-01-18 Martin Sebor <msebor@redhat.com>
9011 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9013 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9016 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9019 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9022 * common.opt: (-ffile-prefix-map): New option.
9023 * opts.c (common_handle_option): Defer it.
9024 * opts-global.c (handle_common_deferred_options): Handle it.
9025 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9026 * file-prefix-map.h: New file.
9027 (remap_debug_filename, add_debug_prefix_map): ...here.
9028 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9029 * final.c (debug_prefix_map, add_debug_prefix_map
9030 remap_debug_filename): Move to...
9031 * file-prefix-map.c: New file.
9032 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9033 generalize, get rid of alloca(), use strrchr() instead of strchr().
9034 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9035 Implement in terms of add_prefix_map().
9036 (remap_macro_filename, remap_debug_filename): Implement in term of
9038 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9039 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9040 * dbxout.c: Include file-prefix-map.h.
9041 * varasm.c: Likewise.
9042 * vmsdbgout.c: Likewise.
9043 * xcoffout.c: Likewise.
9044 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9045 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9046 * doc/invoke.texi (-ffile-prefix-map): Document.
9047 (-fdebug-prefix-map): Update description.
9049 2018-01-18 Martin Liska <mliska@suse.cz>
9051 * config/i386/i386.c (indirect_thunk_name): Document that also
9053 (output_indirect_thunk): Document why both instructions
9054 (pause and lfence) are generated.
9056 2018-01-18 Richard Biener <rguenther@suse.de>
9058 PR tree-optimization/83887
9059 * graphite-scop-detection.c
9060 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9061 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9062 (scop_detection::merge_sese): Re-implement with a flood-fill
9063 algorithm that properly finds a SESE region if it exists.
9065 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9068 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9069 pointer_diff optimizations use view_convert instead of convert.
9071 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9073 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9074 Generate different code for -mno-speculate-indirect-jumps.
9075 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9076 (*call_indirect_aix<mode>): Disable for
9077 -mno-speculate-indirect-jumps.
9078 (*call_indirect_aix<mode>_nospec): New define_insn.
9079 (*call_value_indirect_aix<mode>): Disable for
9080 -mno-speculate-indirect-jumps.
9081 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9082 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9083 -mno-speculate-indirect-jumps.
9084 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9086 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9088 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9089 long double type, set the flags for noting the default long double
9090 type, even if we don't pass or return a long double type.
9092 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9095 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9098 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9100 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9101 support for merge[hl].
9102 (fold_mergehl_helper): New helper function.
9103 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9104 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9105 (altivec_vmrglw_direct): Add xxmrglw insn.
9107 2018-01-17 Andrew Waterman <andrew@sifive.com>
9109 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9110 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9112 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9115 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9116 call the lto_location_cache before reading the
9117 DECL_SOURCE_LOCATION of the types.
9119 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9120 Richard Sandiford <richard.sandiford@linaro.org>
9122 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9123 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9124 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9125 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9126 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9128 * config/aarch64/constraints.md (aarch64_movti_operand):
9130 * config/aarch64/predicates.md (Uti): Add new constraint.
9132 2018-01-17 Carl Love <cel@us.ibm.com>
9134 * config/rs6000/vsx.md (define_expand xl_len_r,
9135 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9136 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9138 (define_expand, define_insn): Move the shift left from the
9139 define_insn to the define_expand for lxvl and stxvl instructions.
9140 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9141 and XL_LEN_R definitions to PURE.
9143 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9145 * config/i386/i386.c (indirect_thunk_name): Declare regno
9146 as unsigned int. Compare regno with INVALID_REGNUM.
9147 (output_indirect_thunk): Ditto.
9148 (output_indirect_thunk_function): Ditto.
9149 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9150 in the call to output_indirect_thunk_function.
9152 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9155 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9156 rather than the size of inner_type to determine the stack slot size
9157 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9159 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9162 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9165 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9167 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9168 endian Linux systems to optionally enable multilibs for selecting
9169 the long double type if the user configured an explicit type.
9170 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9171 have no long double multilibs if not defined.
9172 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9173 warn if the user used -mabi={ieee,ibm}longdouble and we built
9174 multilibs for long double.
9175 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9176 appropriate multilib option.
9177 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9179 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9180 for building long double multilibs.
9181 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9183 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9185 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9188 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9190 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9193 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9196 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9199 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9201 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9202 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9204 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9206 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9207 different rtl trees depending on TARGET_64BIT.
9208 (rs6000_gen_lvx): Likewise.
9210 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9212 * config/visium/visium.md (nop): Tweak comment.
9213 (hazard_nop): Likewise.
9215 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9217 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9218 -mspeculate-indirect-jumps.
9219 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9220 for -mno-speculate-indirect-jumps.
9221 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9222 (*call_value_indirect_elfv2<mode>): Disable for
9223 -mno-speculate-indirect-jumps.
9224 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9225 (indirect_jump): Emit different RTL for
9226 -mno-speculate-indirect-jumps.
9227 (*indirect_jump<mode>): Disable for
9228 -mno-speculate-indirect-jumps.
9229 (*indirect_jump<mode>_nospec): New define_insn.
9230 (tablejump): Emit different RTL for
9231 -mno-speculate-indirect-jumps.
9232 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9233 (tablejumpsi_nospec): New define_expand.
9234 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9235 (tablejumpdi_nospec): New define_expand.
9236 (*tablejump<mode>_internal1): Disable for
9237 -mno-speculate-indirect-jumps.
9238 (*tablejump<mode>_internal1_nospec): New define_insn.
9239 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9242 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9244 * caller-save.c (insert_save): Drop unnecessary parameter. All
9247 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9248 Richard Biener <rguenth@suse.de>
9251 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9252 return early, inline manually is_gimple_sizepos. Make sure if we
9253 call gimplify_expr we don't end up with a gimple constant.
9254 * tree.c (variably_modified_type_p): Don't return true for
9255 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9256 * gimplify.h (is_gimple_sizepos): Remove.
9258 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9260 PR tree-optimization/83857
9261 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9262 vectorizable_live_operation for pure SLP statements.
9263 (vectorizable_live_operation): Handle PHIs.
9265 2018-01-16 Richard Biener <rguenther@suse.de>
9267 PR tree-optimization/83867
9268 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9269 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9271 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9274 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9275 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9276 If off is not INTEGER_CST, issue a may not be aligned warning
9277 rather than isn't aligned. Use isn%'t rather than isn't.
9278 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9280 <case MULT_EXPR>: Improve the case when bottom and one of the
9281 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9282 operand, in that case check if the other operand is multiple of
9283 bottom divided by the INTEGER_CST operand.
9285 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9288 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9289 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9290 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9291 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9292 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9293 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9294 (pa_function_arg_size): New function.
9296 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9298 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9299 in a separate statement.
9301 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9303 PR tree-optimization/83847
9304 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9305 group gathers and scatters.
9307 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9309 PR rtl-optimization/86620
9310 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9312 PR rtl-optimization/83213
9313 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9314 to last if both are JUMP_INSNs.
9316 PR tree-optimization/83843
9317 * gimple-ssa-store-merging.c
9318 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9319 store_immediate_info for bswap/nop orig_stores.
9321 2018-01-15 Andrew Waterman <andrew@sifive.com>
9323 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9325 <UDIV>: Increase cost if !TARGET_DIV.
9327 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9329 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9330 (define_attr "cr_logical_3op"): New.
9331 (cceq_ior_compare): Adjust.
9332 (cceq_ior_compare_complement): Adjust.
9333 (*cceq_rev_compare): Adjust.
9334 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9335 (is_cracked_insn): Adjust.
9336 (insn_must_be_first_in_group): Adjust.
9337 * config/rs6000/40x.md: Adjust.
9338 * config/rs6000/440.md: Adjust.
9339 * config/rs6000/476.md: Adjust.
9340 * config/rs6000/601.md: Adjust.
9341 * config/rs6000/603.md: Adjust.
9342 * config/rs6000/6xx.md: Adjust.
9343 * config/rs6000/7450.md: Adjust.
9344 * config/rs6000/7xx.md: Adjust.
9345 * config/rs6000/8540.md: Adjust.
9346 * config/rs6000/cell.md: Adjust.
9347 * config/rs6000/e300c2c3.md: Adjust.
9348 * config/rs6000/e500mc.md: Adjust.
9349 * config/rs6000/e500mc64.md: Adjust.
9350 * config/rs6000/e5500.md: Adjust.
9351 * config/rs6000/e6500.md: Adjust.
9352 * config/rs6000/mpc.md: Adjust.
9353 * config/rs6000/power4.md: Adjust.
9354 * config/rs6000/power5.md: Adjust.
9355 * config/rs6000/power6.md: Adjust.
9356 * config/rs6000/power7.md: Adjust.
9357 * config/rs6000/power8.md: Adjust.
9358 * config/rs6000/power9.md: Adjust.
9359 * config/rs6000/rs64.md: Adjust.
9360 * config/rs6000/titan.md: Adjust.
9362 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9364 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9365 ix86_indirect_branch_register logic.
9367 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9369 * config/i386/constraints.md (Bs): Update
9370 ix86_indirect_branch_register check. Don't check
9371 ix86_indirect_branch_register with GOT_memory_operand.
9373 * config/i386/predicates.md (GOT_memory_operand): Don't check
9374 ix86_indirect_branch_register here.
9375 (GOT32_symbol_operand): Likewise.
9377 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9379 * config/i386/predicates.md (constant_call_address_operand):
9380 Rewrite ix86_indirect_branch_register logic.
9381 (sibcall_insn_operand): Likewise.
9383 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9385 * config/i386/constraints.md (Bs): Replace
9386 ix86_indirect_branch_thunk_register with
9387 ix86_indirect_branch_register.
9389 * config/i386/i386.md (indirect_jump): Likewise.
9390 (tablejump): Likewise.
9391 (*sibcall_memory): Likewise.
9392 (*sibcall_value_memory): Likewise.
9393 Peepholes of indirect call and jump via memory: Likewise.
9394 * config/i386/i386.opt: Likewise.
9395 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9396 (GOT_memory_operand): Likewise.
9397 (call_insn_operand): Likewise.
9398 (sibcall_insn_operand): Likewise.
9399 (GOT32_symbol_operand): Likewise.
9401 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9404 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9405 type rather than type addr's type points to.
9406 (expand_omp_atomic_mutex): Likewise.
9407 (expand_omp_atomic): Likewise.
9409 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9412 * config/i386/i386.c (output_indirect_thunk_function): Use
9413 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9414 for __x86_return_thunk.
9416 2018-01-15 Richard Biener <rguenther@suse.de>
9419 * expmed.c (extract_bit_field_1): Fix typo.
9421 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9424 * config/arm/iterators.md (VF): New mode iterator.
9425 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9426 Remove integer-related logic from pattern.
9427 (neon_vabd<mode>_3): Likewise.
9429 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9432 * common.opt (fstrict-overflow): No longer an alias.
9433 (fwrapv-pointer): New option.
9434 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9435 also for pointer types based on flag_wrapv_pointer.
9436 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9437 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9438 opts->x_flag_wrapv got set.
9439 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9440 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9441 POINTER_TYPE_OVERFLOW_UNDEFINED.
9442 * match.pd: Likewise in address comparison pattern.
9443 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9445 2018-01-15 Richard Biener <rguenther@suse.de>
9448 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9449 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9450 Reset type names to their identifier if their TYPE_DECL doesn't
9451 have linkage (and thus is used for ODR and devirt).
9452 (save_debug_info_for_decl): Remove.
9453 (save_debug_info_for_type): Likewise.
9454 (add_tree_to_fld_list): Adjust.
9455 * tree-pretty-print.c (dump_generic_node): Make dumping of
9456 type names more robust.
9458 2018-01-15 Richard Biener <rguenther@suse.de>
9460 * BASE-VER: Bump to 8.0.1.
9462 2018-01-14 Martin Sebor <msebor@redhat.com>
9465 * builtins.c (check_access): Avoid warning when the no-warning bit
9468 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9470 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9471 * ira-color (allocno_hard_regs_compare): Likewise.
9473 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9476 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9477 Use .pushsection/.popsection.
9479 2018-01-14 Martin Sebor <msebor@redhat.com>
9482 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9484 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9486 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9487 entry from extra_headers.
9488 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9489 extra_headers, make the list bitwise identical to the i?86-*-* one.
9491 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9493 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9494 -mcmodel=large with -mindirect-branch=thunk,
9495 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9496 -mfunction-return=thunk-extern.
9497 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9498 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9499 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9501 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9503 * config/i386/i386.c (print_reg): Print the name of the full
9504 integer register without '%'.
9505 (ix86_print_operand): Handle 'V'.
9506 * doc/extend.texi: Document 'V' modifier.
9508 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9510 * config/i386/constraints.md (Bs): Disallow memory operand for
9511 -mindirect-branch-register.
9513 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9514 (GOT_memory_operand): Likewise.
9515 (call_insn_operand): Likewise.
9516 (sibcall_insn_operand): Likewise.
9517 (GOT32_symbol_operand): Likewise.
9518 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9519 for -mindirect-branch-register.
9520 (tablejump): Likewise.
9521 (*sibcall_memory): Likewise.
9522 (*sibcall_value_memory): Likewise.
9523 Disallow peepholes of indirect call and jump via memory for
9524 -mindirect-branch-register.
9525 (*call_pop): Replace m with Bw.
9526 (*call_value_pop): Likewise.
9527 (*sibcall_pop_memory): Replace m with Bs.
9528 * config/i386/i386.opt (mindirect-branch-register): New option.
9529 * doc/invoke.texi: Document -mindirect-branch-register option.
9531 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9533 * config/i386/i386-protos.h (ix86_output_function_return): New.
9534 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9535 set function_return_type.
9536 (indirect_thunk_name): Add ret_p to indicate thunk for function
9538 (output_indirect_thunk_function): Pass false to
9539 indirect_thunk_name.
9540 (ix86_output_indirect_branch_via_reg): Likewise.
9541 (ix86_output_indirect_branch_via_push): Likewise.
9542 (output_indirect_thunk_function): Create alias for function
9543 return thunk if regno < 0.
9544 (ix86_output_function_return): New function.
9545 (ix86_handle_fndecl_attribute): Handle function_return.
9546 (ix86_attribute_table): Add function_return.
9547 * config/i386/i386.h (machine_function): Add
9548 function_return_type.
9549 * config/i386/i386.md (simple_return_internal): Use
9550 ix86_output_function_return.
9551 (simple_return_internal_long): Likewise.
9552 * config/i386/i386.opt (mfunction-return=): New option.
9553 (indirect_branch): Mention -mfunction-return=.
9554 * doc/extend.texi: Document function_return function attribute.
9555 * doc/invoke.texi: Document -mfunction-return= option.
9557 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9559 * config/i386/i386-opts.h (indirect_branch): New.
9560 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9561 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9562 with local indirect jump when converting indirect call and jump.
9563 (ix86_set_indirect_branch_type): New.
9564 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9565 (indirectlabelno): New.
9566 (indirect_thunk_needed): Likewise.
9567 (indirect_thunk_bnd_needed): Likewise.
9568 (indirect_thunks_used): Likewise.
9569 (indirect_thunks_bnd_used): Likewise.
9570 (INDIRECT_LABEL): Likewise.
9571 (indirect_thunk_name): Likewise.
9572 (output_indirect_thunk): Likewise.
9573 (output_indirect_thunk_function): Likewise.
9574 (ix86_output_indirect_branch_via_reg): Likewise.
9575 (ix86_output_indirect_branch_via_push): Likewise.
9576 (ix86_output_indirect_branch): Likewise.
9577 (ix86_output_indirect_jmp): Likewise.
9578 (ix86_code_end): Call output_indirect_thunk_function if needed.
9579 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9581 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9582 (ix86_attribute_table): Add indirect_branch.
9583 * config/i386/i386.h (machine_function): Add indirect_branch_type
9584 and has_local_indirect_jump.
9585 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9587 (tablejump): Likewise.
9588 (*indirect_jump): Use ix86_output_indirect_jmp.
9589 (*tablejump_1): Likewise.
9590 (simple_return_indirect_internal): Likewise.
9591 * config/i386/i386.opt (mindirect-branch=): New option.
9592 (indirect_branch): New.
9595 (thunk-inline): Likewise.
9596 (thunk-extern): Likewise.
9597 * doc/extend.texi: Document indirect_branch function attribute.
9598 * doc/invoke.texi: Document -mindirect-branch= option.
9600 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9603 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9605 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9607 * ipa-inline.c (want_inline_small_function_p): Return false if
9608 inlining has already failed with CIF_FINAL_ERROR.
9609 (update_caller_keys): Call want_inline_small_function_p before
9611 (update_callee_keys): Likewise.
9613 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9615 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9617 (rs6000_quadword_masked_address_p): Likewise.
9618 (quad_aligned_load_p): Likewise.
9619 (quad_aligned_store_p): Likewise.
9620 (const_load_sequence_p): Add comment to describe the outer-most loop.
9621 (mimic_memory_attributes_and_flags): New function.
9622 (rs6000_gen_stvx): Likewise.
9623 (replace_swapped_aligned_store): Likewise.
9624 (rs6000_gen_lvx): Likewise.
9625 (replace_swapped_aligned_load): Likewise.
9626 (replace_swapped_load_constant): Capitalize argument name in
9627 comment describing this function.
9628 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9629 and stores that access quad-word aligned addresses and replace
9630 with stvx or lvx instructions when appropriate.
9631 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9632 New function prototype.
9633 (rs6000_quadword_masked_address_p): Likewise.
9634 (rs6000_gen_lvx): Likewise.
9635 (rs6000_gen_stvx): Likewise.
9636 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9637 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9638 when memory address is aligned.
9639 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9640 this split to select lvx instruction when memory address is aligned.
9641 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9642 instruction when memory address is aligned.
9643 (*vsx_le_perm_load_v16qi): Likewise.
9644 (four unnamed splitters): Modify to select the stvx instruction
9645 when memory is aligned.
9647 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9649 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9650 which appears in the queue multiple times.
9652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9653 Alan Hayward <alan.hayward@arm.com>
9654 David Sherwood <david.sherwood@arm.com>
9656 * tree-vectorizer.h (vec_lower_bound): New structure.
9657 (_loop_vec_info): Add check_nonzero and lower_bounds.
9658 (LOOP_VINFO_CHECK_NONZERO): New macro.
9659 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9660 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9661 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9662 fields. Make seg_len the distance travelled, not including the
9664 (dr_direction_indicator): Declare.
9665 (dr_zero_step_indicator): Likewise.
9666 (dr_known_forward_stride_p): Likewise.
9667 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9669 (runtime_alias_check_p): Allow runtime alias checks with
9671 (operator ==): Compare access_size and align.
9672 (prune_runtime_alias_test_list): Rework for new distinction between
9673 the access_size and seg_len.
9674 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9676 (get_segment_min_max): New function.
9677 (create_intersect_range_checks): Use it.
9678 (dr_step_indicator): New function.
9679 (dr_direction_indicator): Likewise.
9680 (dr_zero_step_indicator): Likewise.
9681 (dr_known_forward_stride_p): Likewise.
9682 * tree-loop-distribution.c (data_ref_segment_size): Return
9683 DR_STEP * (niters - 1).
9684 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9686 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9687 (vect_preserves_scalar_order_p): New function, split out from...
9688 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9689 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9690 (vect_vfa_access_size): New function.
9691 (vect_vfa_align): Likewise.
9692 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9693 (dump_lower_bound): New function.
9694 (vect_check_lower_bound): Likewise.
9695 (vect_small_gap_p): Likewise.
9696 (vectorizable_with_step_bound_p): Likewise.
9697 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9698 depencies if the vectorization factor is 1. Convert the checks
9699 for nonzero steps into checks on the bounds of DR_STEP. Try using
9700 a bunds check for variable steps if the minimum required step is
9701 relatively small. Update calls to the dr_with_seg_len
9702 constructor and to vect_compile_time_alias.
9703 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9705 (vect_loop_versioning): Call it.
9706 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9708 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9710 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9711 Alan Hayward <alan.hayward@arm.com>
9712 David Sherwood <david.sherwood@arm.com>
9714 * doc/sourcebuild.texi (vect_scatter_store): Document.
9715 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9717 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9719 * genopinit.c (main): Add supports_vec_scatter_store and
9720 supports_vec_scatter_store_cached to target_optabs.
9721 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9722 IFN_MASK_SCATTER_STORE.
9723 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9725 * internal-fn.h (internal_store_fn_p): Declare.
9726 (internal_fn_stored_value_index): Likewise.
9727 * internal-fn.c (scatter_store_direct): New macro.
9728 (expand_scatter_store_optab_fn): New function.
9729 (direct_scatter_store_optab_supported_p): New macro.
9730 (internal_store_fn_p): New function.
9731 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9732 IFN_MASK_SCATTER_STORE.
9733 (internal_fn_mask_index): Likewise.
9734 (internal_fn_stored_value_index): New function.
9735 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9737 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9738 * optabs-query.c (supports_vec_scatter_store_p): New function.
9739 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9740 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9741 true for scatter stores.
9742 (vect_gather_scatter_fn_p): Handle scatter stores too.
9743 (vect_check_gather_scatter): Consider using scatter stores if
9744 supports_vec_scatter_store_p.
9745 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9747 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9748 internal_fn_stored_value_index.
9749 (check_load_store_masking): Handle scatter stores too.
9750 (vect_get_store_rhs): Make public.
9751 (vectorizable_call): Use internal_store_fn_p.
9752 (vectorizable_store): Handle scatter store internal functions.
9753 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9754 when deciding whether the end of the group has been reached.
9755 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9756 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9757 (mask_scatter_store<mode>): New insns.
9759 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9760 Alan Hayward <alan.hayward@arm.com>
9761 David Sherwood <david.sherwood@arm.com>
9763 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9764 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9765 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9767 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9768 Use vect_truncate_gather_scatter_offset if we can't treat the
9769 operation as a normal gather load or scatter store.
9770 (get_group_load_store_type): Take the gather_scatter_info
9771 as argument. Try using a gather load or scatter store for
9772 single-element groups.
9773 (get_load_store_type): Update calls to get_group_load_store_type
9774 and vect_use_strided_gather_scatters_p.
9776 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9777 Alan Hayward <alan.hayward@arm.com>
9778 David Sherwood <david.sherwood@arm.com>
9780 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9781 optional tree argument.
9782 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9784 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9785 but continue to use the current value as a fallback.
9786 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9787 to compare the updates.
9788 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9789 (get_load_store_type): Use it when handling a strided access.
9790 (vect_get_strided_load_store_ops): New function.
9791 (vect_get_data_ptr_increment): Likewise.
9792 (vectorizable_load): Handle strided gather loads. Always pass
9793 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9795 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9796 Alan Hayward <alan.hayward@arm.com>
9797 David Sherwood <david.sherwood@arm.com>
9799 * doc/md.texi (gather_load@var{m}): Document.
9800 (mask_gather_load@var{m}): Likewise.
9801 * genopinit.c (main): Add supports_vec_gather_load and
9802 supports_vec_gather_load_cached to target_optabs.
9803 * optabs-tree.c (init_tree_optimization_optabs): Use
9804 ggc_cleared_alloc to allocate target_optabs.
9805 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9806 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9808 * internal-fn.h (internal_load_fn_p): Declare.
9809 (internal_gather_scatter_fn_p): Likewise.
9810 (internal_fn_mask_index): Likewise.
9811 (internal_gather_scatter_fn_supported_p): Likewise.
9812 * internal-fn.c (gather_load_direct): New macro.
9813 (expand_gather_load_optab_fn): New function.
9814 (direct_gather_load_optab_supported_p): New macro.
9815 (direct_internal_fn_optab): New function.
9816 (internal_load_fn_p): Likewise.
9817 (internal_gather_scatter_fn_p): Likewise.
9818 (internal_fn_mask_index): Likewise.
9819 (internal_gather_scatter_fn_supported_p): Likewise.
9820 * optabs-query.c (supports_at_least_one_mode_p): New function.
9821 (supports_vec_gather_load_p): Likewise.
9822 * optabs-query.h (supports_vec_gather_load_p): Declare.
9823 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9824 and memory_type field.
9825 (NUM_PATTERNS): Bump to 15.
9826 * tree-vect-data-refs.c: Include internal-fn.h.
9827 (vect_gather_scatter_fn_p): New function.
9828 (vect_describe_gather_scatter_call): Likewise.
9829 (vect_check_gather_scatter): Try using internal functions for
9830 gather loads. Recognize existing calls to a gather load function.
9831 (vect_analyze_data_refs): Consider using gather loads if
9832 supports_vec_gather_load_p.
9833 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9834 (vect_get_gather_scatter_offset_type): Likewise.
9835 (vect_convert_mask_for_vectype): Likewise.
9836 (vect_add_conversion_to_patterm): Likewise.
9837 (vect_try_gather_scatter_pattern): Likewise.
9838 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9839 (vect_vect_recog_func_ptrs): Add it.
9840 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9841 internal_fn_mask_index and internal_gather_scatter_fn_p.
9842 (check_load_store_masking): Take the gather_scatter_info as an
9843 argument and handle gather loads.
9844 (vect_get_gather_scatter_ops): New function.
9845 (vectorizable_call): Check internal_load_fn_p.
9846 (vectorizable_load): Likewise. Handle gather load internal
9848 (vectorizable_store): Update call to check_load_store_masking.
9849 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9850 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9851 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9852 (aarch64_gather_scale_operand_d): New predicates.
9853 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9854 (mask_gather_load<mode>): New insns.
9856 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9857 Alan Hayward <alan.hayward@arm.com>
9858 David Sherwood <david.sherwood@arm.com>
9860 * optabs.def (fold_left_plus_optab): New optab.
9861 * doc/md.texi (fold_left_plus_@var{m}): Document.
9862 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9863 * internal-fn.c (fold_left_direct): Define.
9864 (expand_fold_left_optab_fn): Likewise.
9865 (direct_fold_left_optab_supported_p): Likewise.
9866 * fold-const-call.c (fold_const_fold_left): New function.
9867 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9868 * tree-parloops.c (valid_reduction_p): New function.
9869 (gather_scalar_reductions): Use it.
9870 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9871 (vect_finish_replace_stmt): Declare.
9872 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9873 (needs_fold_left_reduction_p): New function, split out from...
9874 (vect_is_simple_reduction): ...here. Accept reductions that
9875 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9876 (vect_force_simple_reduction): Also store the reduction type in
9877 the assignment's STMT_VINFO_REDUC_TYPE.
9878 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9879 (merge_with_identity): New function.
9880 (vect_expand_fold_left): Likewise.
9881 (vectorize_fold_left_reduction): Likewise.
9882 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9883 scalar phi in place for it. Check for target support and reject
9884 cases that would reassociate the operation. Defer the transform
9885 phase to vectorize_fold_left_reduction.
9886 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9887 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9888 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9892 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9893 call to ifc_temp_var.
9895 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9896 Alan Hayward <alan.hayward@arm.com>
9897 David Sherwood <david.sherwood@arm.com>
9899 * target.def (legitimize_address_displacement): Take the original
9900 offset as a poly_int.
9901 * targhooks.h (default_legitimize_address_displacement): Update
9903 * targhooks.c (default_legitimize_address_displacement): Likewise.
9904 * doc/tm.texi: Regenerate.
9905 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9906 as an argument, moving assert of ad->disp == ad->disp_term to...
9907 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9908 Try calling targetm.legitimize_address_displacement before expanding
9909 the address rather than afterwards, and adjust for the new interface.
9910 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9911 Match the new hook interface. Handle SVE addresses.
9912 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9915 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9917 * Makefile.in (OBJS): Add early-remat.o.
9918 * target.def (select_early_remat_modes): New hook.
9919 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9920 * doc/tm.texi: Regenerate.
9921 * targhooks.h (default_select_early_remat_modes): Declare.
9922 * targhooks.c (default_select_early_remat_modes): New function.
9923 * timevar.def (TV_EARLY_REMAT): New timevar.
9924 * passes.def (pass_early_remat): New pass.
9925 * tree-pass.h (make_pass_early_remat): Declare.
9926 * early-remat.c: New file.
9927 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9929 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9931 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9932 Alan Hayward <alan.hayward@arm.com>
9933 David Sherwood <david.sherwood@arm.com>
9935 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9936 vfm1 with a bound_epilog parameter.
9937 (vect_do_peeling): Update calls accordingly, and move the prologue
9938 call earlier in the function. Treat the base bound_epilog as 0 for
9939 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9940 this base when peeling for gaps.
9941 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9942 with fully-masked loops.
9943 (vect_estimate_min_profitable_iters): Handle the single peeled
9944 iteration in that case.
9946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9947 Alan Hayward <alan.hayward@arm.com>
9948 David Sherwood <david.sherwood@arm.com>
9950 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9951 single-element interleaving even if the size is not a power of 2.
9952 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9953 accesses for single-element interleaving if the group size is
9956 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9957 Alan Hayward <alan.hayward@arm.com>
9958 David Sherwood <david.sherwood@arm.com>
9960 * doc/md.texi (fold_extract_last_@var{m}): Document.
9961 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9962 * optabs.def (fold_extract_last_optab): New optab.
9963 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9964 * internal-fn.c (fold_extract_direct): New macro.
9965 (expand_fold_extract_optab_fn): Likewise.
9966 (direct_fold_extract_optab_supported_p): Likewise.
9967 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9968 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9969 EXTRACT_LAST_REDUCTION.
9970 (get_initial_def_for_reduction): Do not create an initial vector
9971 for EXTRACT_LAST_REDUCTION reductions.
9972 (vectorizable_reduction): Leave the scalar phi in place for
9973 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9974 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9975 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9976 transform phase to vectorizable_condition.
9977 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9979 (vect_finish_stmt_generation): ...here.
9980 (vect_finish_replace_stmt): New function.
9981 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9982 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9984 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9987 Alan Hayward <alan.hayward@arm.com>
9988 David Sherwood <david.sherwood@arm.com>
9990 * doc/md.texi (extract_last_@var{m}): Document.
9991 * optabs.def (extract_last_optab): New optab.
9992 * internal-fn.def (EXTRACT_LAST): New internal function.
9993 * internal-fn.c (cond_unary_direct): New macro.
9994 (expand_cond_unary_optab_fn): Likewise.
9995 (direct_cond_unary_optab_supported_p): Likewise.
9996 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9997 loops using EXTRACT_LAST.
9998 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9999 (extract_last_<mode>): ...this optab.
10000 (vec_extract<mode><Vel>): Update accordingly.
10002 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10003 Alan Hayward <alan.hayward@arm.com>
10004 David Sherwood <david.sherwood@arm.com>
10006 * target.def (empty_mask_is_expensive): New hook.
10007 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10008 * doc/tm.texi: Regenerate.
10009 * targhooks.h (default_empty_mask_is_expensive): Declare.
10010 * targhooks.c (default_empty_mask_is_expensive): New function.
10011 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10012 if the target says that empty masks are expensive.
10013 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10015 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10017 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10018 Alan Hayward <alan.hayward@arm.com>
10019 David Sherwood <david.sherwood@arm.com>
10021 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10022 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10023 (vect_use_loop_mask_for_alignment_p): New function.
10024 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10025 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10026 niters_skip argument. Make sure that the first niters_skip elements
10027 of the first iteration are inactive.
10028 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10029 Update call to vect_set_loop_masks_directly.
10030 (get_misalign_in_elems): New function, split out from...
10031 (vect_gen_prolog_loop_niters): ...here.
10032 (vect_update_init_of_dr): Take a code argument that specifies whether
10033 the adjustment should be added or subtracted.
10034 (vect_update_init_of_drs): Likewise.
10035 (vect_prepare_for_masked_peels): New function.
10036 (vect_do_peeling): Skip prologue peeling if we're using a mask
10037 instead. Update call to vect_update_inits_of_drs.
10038 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10040 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10041 alignment. Do not include the number of peeled iterations in
10042 the minimum threshold in that case.
10043 (vectorizable_induction): Adjust the start value down by
10044 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10045 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10046 Take the number of skipped iterations into account when calculating
10048 * tree-vect-stmts.c (vect_gen_while_not): New function.
10050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10051 Alan Hayward <alan.hayward@arm.com>
10052 David Sherwood <david.sherwood@arm.com>
10054 * doc/sourcebuild.texi (vect_fully_masked): Document.
10055 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10056 default value to 0.
10057 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10059 (vect_analyze_loop_2): ...here. Don't check the vectorization
10060 factor against the number of loop iterations if the loop is
10063 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10064 Alan Hayward <alan.hayward@arm.com>
10065 David Sherwood <david.sherwood@arm.com>
10067 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10068 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10069 (dump_groups): Update accordingly.
10070 (iv_use::mem_type): New member variable.
10071 (address_p): New function.
10072 (record_use): Add a mem_type argument and initialize the new
10074 (record_group_use): Add a mem_type argument. Use address_p.
10075 Remove obsolete null checks of base_object. Update call to record_use.
10076 (find_interesting_uses_op): Update call to record_group_use.
10077 (find_interesting_uses_cond): Likewise.
10078 (find_interesting_uses_address): Likewise.
10079 (get_mem_type_for_internal_fn): New function.
10080 (find_address_like_use): Likewise.
10081 (find_interesting_uses_stmt): Try find_address_like_use before
10082 calling find_interesting_uses_op.
10083 (addr_offset_valid_p): Use the iv mem_type field as the type
10084 of the addressed memory.
10085 (add_autoinc_candidates): Likewise.
10086 (get_address_cost): Likewise.
10087 (split_small_address_groups_p): Use address_p.
10088 (split_address_groups): Likewise.
10089 (add_iv_candidate_for_use): Likewise.
10090 (autoinc_possible_for_pair): Likewise.
10091 (rewrite_groups): Likewise.
10092 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10093 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10094 (get_alias_ptr_type_for_ptr_address): New function.
10095 (rewrite_use_address): Rewrite address uses in calls that were
10096 identified by find_address_like_use.
10098 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10099 Alan Hayward <alan.hayward@arm.com>
10100 David Sherwood <david.sherwood@arm.com>
10102 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10104 * gimple-expr.h (is_gimple_addressable: Likewise.
10105 * gimple-expr.c (is_gimple_address): Likewise.
10106 * internal-fn.c (expand_call_mem_ref): New function.
10107 (expand_mask_load_optab_fn): Use it.
10108 (expand_mask_store_optab_fn): Likewise.
10110 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10111 Alan Hayward <alan.hayward@arm.com>
10112 David Sherwood <david.sherwood@arm.com>
10114 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10115 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10116 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10117 (cond_umax@var{mode}): Document.
10118 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10119 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10120 (cond_umin_optab, cond_umax_optab): New optabs.
10121 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10122 (COND_IOR, COND_XOR): New internal functions.
10123 * internal-fn.h (get_conditional_internal_fn): Declare.
10124 * internal-fn.c (cond_binary_direct): New macro.
10125 (expand_cond_binary_optab_fn): Likewise.
10126 (direct_cond_binary_optab_supported_p): Likewise.
10127 (get_conditional_internal_fn): New function.
10128 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10129 Cope with reduction statements that are vectorized as calls rather
10131 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10132 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10133 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10134 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10135 (UNSPEC_COND_EOR): New unspecs.
10136 (optab): Add mappings for them.
10137 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10138 (sve_int_op, sve_fp_op): New int attributes.
10140 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10141 Alan Hayward <alan.hayward@arm.com>
10142 David Sherwood <david.sherwood@arm.com>
10144 * optabs.def (while_ult_optab): New optab.
10145 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10146 * internal-fn.def (WHILE_ULT): New internal function.
10147 * internal-fn.h (direct_internal_fn_supported_p): New override
10148 that takes two types as argument.
10149 * internal-fn.c (while_direct): New macro.
10150 (expand_while_optab_fn): New function.
10151 (convert_optab_supported_p): Likewise.
10152 (direct_while_optab_supported_p): New macro.
10153 * wide-int.h (wi::udiv_ceil): New function.
10154 * tree-vectorizer.h (rgroup_masks): New structure.
10155 (vec_loop_masks): New typedef.
10156 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10157 and fully_masked_p.
10158 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10159 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10160 (vect_max_vf): New function.
10161 (slpeel_make_loop_iterate_ntimes): Delete.
10162 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10163 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10164 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10165 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10166 internal-fn.h, stor-layout.h and optabs-query.h.
10167 (vect_set_loop_mask): New function.
10168 (add_preheader_seq): Likewise.
10169 (add_header_seq): Likewise.
10170 (interleave_supported_p): Likewise.
10171 (vect_maybe_permute_loop_masks): Likewise.
10172 (vect_set_loop_masks_directly): Likewise.
10173 (vect_set_loop_condition_masked): Likewise.
10174 (vect_set_loop_condition_unmasked): New function, split out from
10175 slpeel_make_loop_iterate_ntimes.
10176 (slpeel_make_loop_iterate_ntimes): Rename to..
10177 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10178 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10179 (vect_do_peeling): Update call accordingly.
10180 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10182 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10183 mask_compare_type, can_fully_mask_p and fully_masked_p.
10184 (release_vec_loop_masks): New function.
10185 (_loop_vec_info): Use it to free the loop masks.
10186 (can_produce_all_loop_masks_p): New function.
10187 (vect_get_max_nscalars_per_iter): Likewise.
10188 (vect_verify_full_masking): Likewise.
10189 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10190 retries, and free the mask rgroups before retrying. Check loop-wide
10191 reasons for disallowing fully-masked loops. Make the final decision
10192 about whether use a fully-masked loop or not.
10193 (vect_estimate_min_profitable_iters): Do not assume that peeling
10194 for the number of iterations will be needed for fully-masked loops.
10195 (vectorizable_reduction): Disable fully-masked loops.
10196 (vectorizable_live_operation): Likewise.
10197 (vect_halve_mask_nunits): New function.
10198 (vect_double_mask_nunits): Likewise.
10199 (vect_record_loop_mask): Likewise.
10200 (vect_get_loop_mask): Likewise.
10201 (vect_transform_loop): Handle the case in which the final loop
10202 iteration might handle a partial vector. Call vect_set_loop_condition
10203 instead of slpeel_make_loop_iterate_ntimes.
10204 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10205 (check_load_store_masking): New function.
10206 (prepare_load_store_mask): Likewise.
10207 (vectorizable_store): Handle fully-masked loops.
10208 (vectorizable_load): Likewise.
10209 (supportable_widening_operation): Use vect_halve_mask_nunits for
10211 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10212 (vect_gen_while): New function.
10213 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10214 (aarch64_uqdec<mode>): New insn.
10216 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10217 Alan Hayward <alan.hayward@arm.com>
10218 David Sherwood <david.sherwood@arm.com>
10220 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10221 (reduc_xor_scal_optab): New optabs.
10222 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10223 (reduc_xor_scal_@var{m}): Document.
10224 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10225 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10226 internal functions.
10227 * fold-const-call.c (fold_const_call): Handle them.
10228 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10229 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10230 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10231 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10232 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10233 (UNSPEC_XORV): New unspecs.
10234 (optab): Add entries for them.
10235 (BITWISEV): New int iterator.
10236 (bit_reduc_op): New int attributes.
10238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10239 Alan Hayward <alan.hayward@arm.com>
10240 David Sherwood <david.sherwood@arm.com>
10242 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10243 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10244 * optabs.def (vec_shl_insert_optab): New optab.
10245 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10246 (duplicate_and_interleave): Likewise.
10247 * tree-vect-loop.c: Include internal-fn.h.
10248 (neutral_op_for_slp_reduction): New function, split out from
10249 get_initial_defs_for_reduction.
10250 (get_initial_def_for_reduction): Handle option 2 for variable-length
10251 vectors by loading the neutral value into a vector and then shifting
10252 the initial value into element 0.
10253 (get_initial_defs_for_reduction): Replace the code argument with
10254 the neutral value calculated by neutral_op_for_slp_reduction.
10255 Use gimple_build_vector for constant-length vectors.
10256 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10257 but the first group_size elements have a neutral value.
10258 Use duplicate_and_interleave otherwise.
10259 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10260 Update call to get_initial_defs_for_reduction. Handle SLP
10261 reductions for variable-length vectors by creating one vector
10262 result for each scalar result, with the elements associated
10263 with other scalar results stubbed out with the neutral value.
10264 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10265 Require IFN_VEC_SHL_INSERT for double reductions on
10266 variable-length vectors, or SLP reductions that have
10267 a neutral value. Require can_duplicate_and_interleave_p
10268 support for variable-length unchained SLP reductions if there
10269 is no neutral value, such as for MIN/MAX reductions. Also require
10270 the number of vector elements to be a multiple of the number of
10271 SLP statements when doing variable-length unchained SLP reductions.
10272 Update call to vect_create_epilog_for_reduction.
10273 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10274 and remove initial values.
10275 (duplicate_and_interleave): Make public.
10276 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10277 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10279 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10280 Alan Hayward <alan.hayward@arm.com>
10281 David Sherwood <david.sherwood@arm.com>
10283 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10284 (can_duplicate_and_interleave_p): New function.
10285 (vect_get_and_check_slp_defs): Take the vector of statements
10286 rather than just the current one. Remove excess parentheses.
10287 Restriction rejectinon of vect_constant_def and vect_external_def
10288 for variable-length vectors to boolean types, or types for which
10289 can_duplicate_and_interleave_p is false.
10290 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10291 (duplicate_and_interleave): New function.
10292 (vect_get_constant_vectors): Use gimple_build_vector for
10293 constant-length vectors and suitable variable-length constant
10294 vectors. Use duplicate_and_interleave for other variable-length
10295 vectors. Don't defer the update when inserting new statements.
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 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10302 min_profitable_iters doesn't go negative.
10304 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10305 Alan Hayward <alan.hayward@arm.com>
10306 David Sherwood <david.sherwood@arm.com>
10308 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10309 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10310 * optabs.def (vec_mask_load_lanes_optab): New optab.
10311 (vec_mask_store_lanes_optab): Likewise.
10312 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10313 (MASK_STORE_LANES): Likewise.
10314 * internal-fn.c (mask_load_lanes_direct): New macro.
10315 (mask_store_lanes_direct): Likewise.
10316 (expand_mask_load_optab_fn): Handle masked operations.
10317 (expand_mask_load_lanes_optab_fn): New macro.
10318 (expand_mask_store_optab_fn): Handle masked operations.
10319 (expand_mask_store_lanes_optab_fn): New macro.
10320 (direct_mask_load_lanes_optab_supported_p): Likewise.
10321 (direct_mask_store_lanes_optab_supported_p): Likewise.
10322 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10324 (vect_load_lanes_supported): Likewise.
10325 * tree-vect-data-refs.c (strip_conversion): New function.
10326 (can_group_stmts_p): Likewise.
10327 (vect_analyze_data_ref_accesses): Use it instead of checking
10328 for a pair of assignments.
10329 (vect_store_lanes_supported): Take a masked_p parameter.
10330 (vect_load_lanes_supported): Likewise.
10331 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10332 vect_store_lanes_supported and vect_load_lanes_supported.
10333 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10334 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10335 parameter. Don't allow gaps for masked accesses.
10336 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10337 and vect_load_lanes_supported.
10338 (get_load_store_type): Take a masked_p parameter and update
10339 call to get_group_load_store_type.
10340 (vectorizable_store): Update call to get_load_store_type.
10341 Handle IFN_MASK_STORE_LANES.
10342 (vectorizable_load): Update call to get_load_store_type.
10343 Handle IFN_MASK_LOAD_LANES.
10345 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10346 Alan Hayward <alan.hayward@arm.com>
10347 David Sherwood <david.sherwood@arm.com>
10349 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10351 * config/aarch64/aarch64-protos.h
10352 (aarch64_sve_struct_memory_operand_p): Declare.
10353 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10354 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10355 (VPRED, vpred): Handle SVE structure modes.
10356 * config/aarch64/constraints.md (Utx): New constraint.
10357 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10358 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10359 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10360 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10361 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10362 structure modes. Split into pieces after RA.
10363 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10364 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10366 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10367 SVE structure modes.
10368 (aarch64_classify_address): Likewise.
10369 (sizetochar): Move earlier in file.
10370 (aarch64_print_operand): Handle SVE register lists.
10371 (aarch64_array_mode): New function.
10372 (aarch64_sve_struct_memory_operand_p): Likewise.
10373 (TARGET_ARRAY_MODE): Redefine.
10375 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10376 Alan Hayward <alan.hayward@arm.com>
10377 David Sherwood <david.sherwood@arm.com>
10379 * target.def (array_mode): New target hook.
10380 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10381 * doc/tm.texi: Regenerate.
10382 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10383 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10384 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10385 targetm.array_mode.
10386 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10389 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10390 Alan Hayward <alan.hayward@arm.com>
10391 David Sherwood <david.sherwood@arm.com>
10393 * fold-const.c (fold_binary_loc): Check the argument types
10394 rather than the result type when testing for a vector operation.
10396 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10398 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10399 * doc/tm.texi: Regenerate.
10401 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10402 Alan Hayward <alan.hayward@arm.com>
10403 David Sherwood <david.sherwood@arm.com>
10405 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10406 (sve): Document new AArch64 extension.
10407 * doc/md.texi (w): Extend the description of the AArch64
10408 constraint to include SVE vectors.
10409 (Upl, Upa): Document new AArch64 predicate constraints.
10410 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10412 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10413 (msve-vector-bits=): New option.
10414 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10415 SVE when these are disabled.
10416 (sve): New extension.
10417 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10418 modes. Adjust their number of units based on aarch64_sve_vg.
10419 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10420 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10421 aarch64_addr_query_type.
10422 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10423 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10424 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10425 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10426 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10427 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10428 (aarch64_simd_imm_zero_p): Delete.
10429 (aarch64_check_zero_based_sve_index_immediate): Declare.
10430 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10431 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10432 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10433 (aarch64_sve_float_mul_immediate_p): Likewise.
10434 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10435 rather than an rtx.
10436 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10437 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10438 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10439 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10440 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10441 (aarch64_regmode_natural_size): Likewise.
10442 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10443 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10445 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10446 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10447 for VG and the SVE predicate registers.
10448 (V_ALIASES): Add a "z"-prefixed alias.
10449 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10450 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10451 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10452 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10453 (REG_CLASS_NAMES): Add entries for them.
10454 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10455 and the predicate registers.
10456 (aarch64_sve_vg): Declare.
10457 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10458 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10459 (REGMODE_NATURAL_SIZE): Define.
10460 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10462 * config/aarch64/aarch64.c: Include cfgrtl.h.
10463 (simd_immediate_info): Add a constructor for series vectors,
10464 and an associated step field.
10465 (aarch64_sve_vg): New variable.
10466 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10467 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10468 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10469 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10470 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10471 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10472 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10473 (aarch64_get_mask_mode): New functions.
10474 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10475 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10476 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10477 predicate modes and predicate registers. Explicitly restrict
10478 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10479 to store a vector mode if it is recognized by
10480 aarch64_classify_vector_mode.
10481 (aarch64_regmode_natural_size): New function.
10482 (aarch64_hard_regno_caller_save_mode): Return the original mode
10484 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10485 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10486 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10487 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10489 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10490 does not overlap dest if the function is frame-related. Handle
10492 (aarch64_split_add_offset): New function.
10493 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10494 them aarch64_add_offset.
10495 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10496 and update call to aarch64_sub_sp.
10497 (aarch64_add_cfa_expression): New function.
10498 (aarch64_expand_prologue): Pass extra temporary registers to the
10499 functions above. Handle the case in which we need to emit new
10500 DW_CFA_expressions for registers that were originally saved
10501 relative to the stack pointer, but now have to be expressed
10502 relative to the frame pointer.
10503 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10505 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10506 IP0 and IP1 values for SVE frames.
10507 (aarch64_expand_vec_series): New function.
10508 (aarch64_expand_sve_widened_duplicate): Likewise.
10509 (aarch64_expand_sve_const_vector): Likewise.
10510 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10511 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10512 into the register, rather than emitting a SET directly.
10513 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10514 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10515 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10516 (offset_9bit_signed_scaled_p): New functions.
10517 (aarch64_replicate_bitmask_imm): New function.
10518 (aarch64_bitmask_imm): Use it.
10519 (aarch64_cannot_force_const_mem): Reject expressions involving
10520 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10521 (aarch64_classify_index): Handle SVE indices, by requiring
10522 a plain register index with a scale that matches the element size.
10523 (aarch64_classify_address): Handle SVE addresses. Assert that
10524 the mode of the address is VOIDmode or an integer mode.
10525 Update call to aarch64_classify_symbol.
10526 (aarch64_classify_symbolic_expression): Update call to
10527 aarch64_classify_symbol.
10528 (aarch64_const_vec_all_in_range_p): New function.
10529 (aarch64_print_vector_float_operand): Likewise.
10530 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10531 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10532 and the FP immediates 1.0 and 0.5.
10533 (aarch64_print_address_internal): Handle SVE addresses.
10534 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10535 (aarch64_regno_regclass): Handle predicate registers.
10536 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10538 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10539 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10540 (aarch64_convert_sve_vector_bits): New function.
10541 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10542 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10543 rather than an rtx.
10544 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10545 Handle SVE vector and predicate modes. Accept VL-based constants
10546 that need only one temporary register, and VL offsets that require
10547 no temporary registers.
10548 (aarch64_conditional_register_usage): Mark the predicate registers
10549 as fixed if SVE isn't available.
10550 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10551 Return true for SVE vector and predicate modes.
10552 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10553 rather than an unsigned int. Handle SVE modes.
10554 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10556 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10558 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10559 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10560 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10561 (aarch64_sve_float_mul_immediate_p): New functions.
10562 (aarch64_sve_valid_immediate): New function.
10563 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10564 Explicitly reject structure modes. Check for INDEX constants.
10565 Handle PTRUE and PFALSE constants.
10566 (aarch64_check_zero_based_sve_index_immediate): New function.
10567 (aarch64_simd_imm_zero_p): Delete.
10568 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10569 vector modes. Accept constants in the range of CNT[BHWD].
10570 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10571 ask for an Advanced SIMD mode.
10572 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10573 (aarch64_simd_vector_alignment): Handle SVE predicates.
10574 (aarch64_vectorize_preferred_vector_alignment): New function.
10575 (aarch64_simd_vector_alignment_reachable): Use it instead of
10577 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10578 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10580 (MAX_VECT_LEN): Delete.
10581 (expand_vec_perm_d): Add a vec_flags field.
10582 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10583 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10584 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10586 (aarch64_evpc_rev): Rename to...
10587 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10588 (aarch64_evpc_rev_global): New function.
10589 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10590 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10592 (aarch64_evpc_sve_tbl): New function.
10593 (aarch64_expand_vec_perm_const_1): Update after rename of
10594 aarch64_evpc_rev. Handle SVE permutes too, trying
10595 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10596 than aarch64_evpc_tbl.
10597 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10598 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10599 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10600 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10601 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10602 (aarch64_expand_sve_vcond): New functions.
10603 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10604 of aarch64_vector_mode_p.
10605 (aarch64_dwarf_poly_indeterminate_value): New function.
10606 (aarch64_compute_pressure_classes): Likewise.
10607 (aarch64_can_change_mode_class): Likewise.
10608 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10609 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10610 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10611 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10612 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10613 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10614 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10615 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10617 (Dn, Dl, Dr): Accept const as well as const_vector.
10618 (Dz): Likewise. Compare against CONST0_RTX.
10619 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10620 of "vector" where appropriate.
10621 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10622 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10623 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10624 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10625 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10626 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10627 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10628 (v_int_equiv): Extend to SVE modes.
10629 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10631 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10632 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10633 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10634 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10635 (SVE_COND_FP_CMP): New int iterators.
10636 (perm_hilo): Handle the new unpack unspecs.
10637 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10639 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10640 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10641 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10642 (aarch64_equality_operator, aarch64_constant_vector_operand)
10643 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10644 (aarch64_sve_nonimmediate_operand): Likewise.
10645 (aarch64_sve_general_operand): Likewise.
10646 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10647 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10648 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10649 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10650 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10651 (aarch64_sve_float_arith_immediate): Likewise.
10652 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10653 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10654 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10655 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10656 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10657 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10658 (aarch64_sve_float_arith_operand): Likewise.
10659 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10660 (aarch64_sve_float_mul_operand): Likewise.
10661 (aarch64_sve_vec_perm_operand): Likewise.
10662 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10663 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10664 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10665 as well as const_vector.
10666 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10667 in file. Use CONST0_RTX and CONSTM1_RTX.
10668 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10669 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10670 Use aarch64_simd_imm_zero.
10671 * config/aarch64/aarch64-sve.md: New file.
10672 * config/aarch64/aarch64.md: Include it.
10673 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10674 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10675 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10676 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10677 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10678 (sve): New attribute.
10679 (enabled): Disable instructions with the sve attribute unless
10681 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10682 aarch64_expand_mov_immediate.
10683 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10684 CNT[BHSD] immediates.
10685 (movti): Split CONST_POLY_INT moves into two halves.
10686 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10687 Split additions that need a temporary here if the destination
10688 is the stack pointer.
10689 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10690 (*add<mode>3_poly_1): New instruction.
10691 (set_clobber_cc): New expander.
10693 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10695 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10696 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10697 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10698 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10699 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10700 Change innermode from fixed_mode_size to machine_mode.
10701 (simplify_subreg): Update call accordingly. Handle a constant-sized
10702 subreg of a variable-length CONST_VECTOR.
10704 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10705 Alan Hayward <alan.hayward@arm.com>
10706 David Sherwood <david.sherwood@arm.com>
10708 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10709 (add_offset_to_base): New function, split out from...
10710 (create_mem_ref): ...here. When handling a scale other than 1,
10711 check first whether the address is valid without the offset.
10712 Add it into the base if so, leaving the index and scale as-is.
10714 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10717 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10718 fold_for_warn before checking if arg2 is INTEGER_CST.
10720 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10722 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10723 (store_multiple_operation): Delete.
10724 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10725 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10726 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10727 guarded by TARGET_STRING.
10728 (rs6000_output_load_multiple): Delete.
10729 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10730 OPTION_MASK_STRING / TARGET_STRING handling.
10731 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10732 (const rs6000_opt_masks) <"string">: Change mask to 0.
10733 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10734 (MASK_STRING): Delete.
10735 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10737 (load_multiple): Delete.
10744 (store_multiple): Delete.
10751 (movmemsi_8reg): Delete.
10752 (corresponding unnamed define_insn): Delete.
10753 (movmemsi_6reg): Delete.
10754 (corresponding unnamed define_insn): Delete.
10755 (movmemsi_4reg): Delete.
10756 (corresponding unnamed define_insn): Delete.
10757 (movmemsi_2reg): Delete.
10758 (corresponding unnamed define_insn): Delete.
10759 (movmemsi_1reg): Delete.
10760 (corresponding unnamed define_insn): Delete.
10761 * config/rs6000/rs6000.opt (mno-string): New.
10762 (mstring): Replace by deprecation warning stub.
10763 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10765 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10767 * regrename.c (regrename_do_replace): If replacing the same
10768 reg multiple times, try to reuse last created gen_raw_REG.
10771 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10772 main to workaround a bug in GDB.
10774 2018-01-12 Tom de Vries <tom@codesourcery.com>
10777 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10779 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10781 PR rtl-optimization/80481
10782 * ira-color.c (get_cap_member): New function.
10783 (allocnos_conflict_by_live_ranges_p): Use it.
10784 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10785 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10787 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10790 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10791 (*saddl_se_1): Ditto.
10792 (*ssubsi_1): Ditto.
10793 (*ssubl_se_1): Ditto.
10795 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10797 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10798 rather than wi::to_widest for DR_INITs.
10799 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10800 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10801 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10803 (vect_analyze_group_access_1): Note that here.
10805 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10807 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10808 polynomial type sizes.
10810 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10812 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10813 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10814 (gimple_add_tmp_var): Likewise.
10816 2018-01-12 Martin Liska <mliska@suse.cz>
10818 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10819 (gimple_alloc_sizes): Likewise.
10820 (dump_gimple_statistics): Use PRIu64 in printf format.
10821 * gimple.h: Change uint64_t to int.
10823 2018-01-12 Martin Liska <mliska@suse.cz>
10825 * tree-core.h: Use uint64_t instead of int.
10826 * tree.c (tree_node_counts): Likewise.
10827 (tree_node_sizes): Likewise.
10828 (dump_tree_statistics): Use PRIu64 in printf format.
10830 2018-01-12 Martin Liska <mliska@suse.cz>
10832 * Makefile.in: As qsort_chk is implemented in vec.c, add
10833 vec.o to linkage of gencfn-macros.
10834 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10835 passing the info to record_node_allocation_statistics.
10836 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10838 * ggc-common.c (struct ggc_usage): Add operator== and use
10839 it in operator< and compare function.
10840 * mem-stats.h (struct mem_usage): Likewise.
10841 * vec.c (struct vec_usage): Remove operator< and compare
10842 function. Can be simply inherited.
10844 2018-01-12 Martin Jambor <mjambor@suse.cz>
10847 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10848 * tree-ssa-math-opts.c: Include domwalk.h.
10849 (convert_mult_to_fma_1): New function.
10850 (fma_transformation_info): New type.
10851 (fma_deferring_state): Likewise.
10852 (cancel_fma_deferring): New function.
10853 (result_of_phi): Likewise.
10854 (last_fma_candidate_feeds_initial_phi): Likewise.
10855 (convert_mult_to_fma): Added deferring logic, split actual
10856 transformation to convert_mult_to_fma_1.
10857 (math_opts_dom_walker): New type.
10858 (math_opts_dom_walker::after_dom_children): New method, body moved
10859 here from pass_optimize_widening_mul::execute, added deferring logic
10861 (pass_optimize_widening_mul::execute): Moved most of code to
10862 math_opts_dom_walker::after_dom_children.
10863 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10864 * config/i386/i386.c (ix86_option_override_internal): Added
10865 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10867 2018-01-12 Richard Biener <rguenther@suse.de>
10870 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10871 inline instance vars.
10873 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10876 * config/rx/rx.c (rx_is_restricted_memory_address):
10877 Handle SUBREG case.
10879 2018-01-12 Richard Biener <rguenther@suse.de>
10881 PR tree-optimization/80846
10882 * target.def (split_reduction): New target hook.
10883 * targhooks.c (default_split_reduction): New function.
10884 * targhooks.h (default_split_reduction): Declare.
10885 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10886 target requests first reduce vectors by combining low and high
10888 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10889 (get_vectype_for_scalar_type_and_size): Export.
10890 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10891 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10892 * doc/tm.texi: Regenerate.
10893 * config/i386/i386.c (ix86_split_reduction): Implement
10894 TARGET_VECTORIZE_SPLIT_REDUCTION.
10896 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10899 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10900 in PIC mode except for TARGET_VXWORKS_RTP.
10901 * config/sparc/sparc.c: Include cfgrtl.h.
10902 (TARGET_INIT_PIC_REG): Define.
10903 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10904 (sparc_pic_register_p): New predicate.
10905 (sparc_legitimate_address_p): Use it.
10906 (sparc_legitimize_pic_address): Likewise.
10907 (sparc_delegitimize_address): Likewise.
10908 (sparc_mode_dependent_address_p): Likewise.
10909 (gen_load_pcrel_sym): Remove 4th parameter.
10910 (load_got_register): Adjust call to above. Remove obsolete stuff.
10911 (sparc_expand_prologue): Do not call load_got_register here.
10912 (sparc_flat_expand_prologue): Likewise.
10913 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10914 (sparc_use_pseudo_pic_reg): New function.
10915 (sparc_init_pic_reg): Likewise.
10916 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10917 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10919 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10921 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10922 Add item for branch_cost.
10924 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10926 PR rtl-optimization/83565
10927 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10928 not extend the result to a larger mode for rotate operations.
10929 (num_sign_bit_copies1): Likewise.
10931 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10934 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10936 Use values-Xc.o for -pedantic.
10937 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10939 2018-01-12 Martin Liska <mliska@suse.cz>
10942 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10944 (possible_polymorphic_call_targets): Use it.
10945 (ipa_devirt): Likewise.
10947 2018-01-12 Martin Liska <mliska@suse.cz>
10949 * profile-count.h (enum profile_quality): Use 0 as invalid
10950 enum value of profile_quality.
10952 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10954 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10955 -mext-string options.
10957 2018-01-12 Richard Biener <rguenther@suse.de>
10959 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10960 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10961 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10963 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10965 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10967 * configure.ac (--with-long-double-format): Add support for the
10968 configuration option to change the default long double format on
10970 * config.gcc (powerpc*-linux*-*): Likewise.
10971 * configure: Regenerate.
10972 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10973 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10974 used without modification.
10976 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10978 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10979 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10980 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10981 MISC_BUILTIN_SPEC_BARRIER.
10982 (rs6000_init_builtins): Likewise.
10983 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10985 (speculation_barrier): New define_insn.
10986 * doc/extend.texi: Document __builtin_speculation_barrier.
10988 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10991 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10992 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10993 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10995 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10996 integral modes instead of "ss" and "sd".
10997 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10998 vectors with 32-bit and 64-bit elements.
10999 (vecdupssescalarmodesuffix): New mode attribute.
11000 (vec_dup<mode>): Use it.
11002 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11005 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11006 frame if argument is passed on stack.
11008 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11011 * ree.c (combine_reaching_defs): Optimize also
11012 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11013 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11015 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11017 PR middle-end/83189
11018 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11020 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11022 PR middle-end/83718
11023 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11024 after they are computed.
11026 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11028 PR tree-optimization/83695
11029 * gimple-loop-linterchange.cc
11030 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11031 reset cached scev information after interchange.
11032 (pass_linterchange::execute): Remove call to scev_reset_htab.
11034 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11036 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11037 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11038 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11039 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11040 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11041 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11042 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11043 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11044 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11045 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11046 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11047 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11048 (V_lane_reg): Likewise.
11049 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11051 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11052 (vfmal_lane_low<mode>_intrinsic,
11053 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11054 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11055 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11056 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11057 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11058 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11060 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11062 * config/arm/arm-cpus.in (fp16fml): New feature.
11063 (ALL_SIMD): Add fp16fml.
11064 (armv8.2-a): Add fp16fml as an option.
11065 (armv8.3-a): Likewise.
11066 (armv8.4-a): Add fp16fml as part of fp16.
11067 * config/arm/arm.h (TARGET_FP16FML): Define.
11068 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11070 * config/arm/arm-modes.def (V2HF): Define.
11071 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11072 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11073 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11074 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11075 vfmsl_low, vfmsl_high): New set of builtins.
11076 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11077 (vfml_op): New code attribute.
11078 (VFMLHALVES): New int iterator.
11079 (VFML, VFMLSEL): New mode attributes.
11080 (V_reg): Define mapping for V2HF.
11081 (V_hi, V_lo): New mode attributes.
11082 (VF_constraint): Likewise.
11083 (vfml_half, vfml_half_selector): New int attributes.
11084 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11086 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11087 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11089 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11090 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11091 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11092 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11094 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11095 Document new effective target and option set.
11097 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11099 * config/arm/arm-cpus.in (armv8_4): New feature.
11100 (ARMv8_4a): New fgroup.
11101 (armv8.4-a): New arch.
11102 * config/arm/arm-tables.opt: Regenerate.
11103 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11104 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11105 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11106 Add matching rules for -march=armv8.4-a and extensions.
11107 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11109 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11112 * config/rx/rx.md (BW): New mode attribute.
11113 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11115 2018-01-11 Richard Biener <rguenther@suse.de>
11117 PR tree-optimization/83435
11118 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11119 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11120 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11122 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11123 Alan Hayward <alan.hayward@arm.com>
11124 David Sherwood <david.sherwood@arm.com>
11126 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11128 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11129 (aarch64_print_address_internal): Use it to check for a zero offset.
11131 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11132 Alan Hayward <alan.hayward@arm.com>
11133 David Sherwood <david.sherwood@arm.com>
11135 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11136 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11137 Return a poly_int64 rather than a HOST_WIDE_INT.
11138 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11139 rather than a HOST_WIDE_INT.
11140 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11141 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11142 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11143 final_offset from HOST_WIDE_INT to poly_int64.
11144 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11145 to_constant when getting the number of units in an Advanced SIMD
11147 (aarch64_builtin_vectorized_function): Check for a constant number
11149 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11151 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11152 attribute instead of GET_MODE_NUNITS.
11153 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11154 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11155 GET_MODE_SIZE for fixed-size registers.
11156 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11157 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11158 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11159 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11160 (aarch64_print_operand, aarch64_print_address_internal)
11161 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11162 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11163 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11164 Handle polynomial GET_MODE_SIZE.
11165 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11166 wider than SImode without modification.
11167 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11168 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11169 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11170 passing and returning SVE modes.
11171 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11172 rather than GEN_INT.
11173 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11174 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11175 (aarch64_allocate_and_probe_stack_space): Likewise.
11176 (aarch64_layout_frame): Cope with polynomial offsets.
11177 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11178 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11179 polynomial offsets.
11180 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11181 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11182 poly_int64 rather than a HOST_WIDE_INT.
11183 (aarch64_get_separate_components, aarch64_process_components)
11184 (aarch64_expand_prologue, aarch64_expand_epilogue)
11185 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11186 (aarch64_anchor_offset): New function, split out from...
11187 (aarch64_legitimize_address): ...here.
11188 (aarch64_builtin_vectorization_cost): Handle polynomial
11189 TYPE_VECTOR_SUBPARTS.
11190 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11192 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11193 number of elements from the PARALLEL rather than the mode.
11194 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11195 rather than GET_MODE_BITSIZE.
11196 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11197 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11198 (aarch64_expand_vec_perm_const_1): Handle polynomial
11199 d->perm.length () and d->perm elements.
11200 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11201 Apply to_constant to d->perm elements.
11202 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11203 polynomial CONST_VECTOR_NUNITS.
11204 (aarch64_move_pointer): Take amount as a poly_int64 rather
11206 (aarch64_progress_pointer): Avoid temporary variable.
11207 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11208 the mode attribute instead of GET_MODE.
11210 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11211 Alan Hayward <alan.hayward@arm.com>
11212 David Sherwood <david.sherwood@arm.com>
11214 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11215 x exists before using it.
11216 (aarch64_add_constant_internal): Rename to...
11217 (aarch64_add_offset_1): ...this. Replace regnum with separate
11218 src and dest rtxes. Handle the case in which they're different,
11219 including when the offset is zero. Replace scratchreg with an rtx.
11220 Use 2 additions if there is no spare register into which we can
11221 move a 16-bit constant.
11222 (aarch64_add_constant): Delete.
11223 (aarch64_add_offset): Replace reg with separate src and dest
11224 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11225 Use aarch64_add_offset_1.
11226 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11227 an rtx rather than an int. Take the delta as a poly_int64
11228 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11229 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11230 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11231 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11232 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11233 and aarch64_add_sp.
11234 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11235 aarch64_add_constant.
11237 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11239 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11240 Use scalar_float_mode.
11242 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11244 * config/aarch64/aarch64-simd.md
11245 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11246 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11247 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11248 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11249 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11250 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11251 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11252 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11253 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11254 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11256 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11259 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11260 targ_options->x_arm_arch_string is non NULL.
11262 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11264 * config/aarch64/aarch64.h
11265 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11267 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11270 * expmed.c (emit_store_flag_force): Swap if const op0
11271 and change VOIDmode to mode of op0.
11273 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11275 PR rtl-optimization/83761
11276 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11277 than bytes to mode_for_size.
11279 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11281 PR middle-end/83189
11282 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11283 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11286 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11288 PR middle-end/83575
11289 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11290 when in layout mode.
11291 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11292 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11295 2018-01-10 Michael Collison <michael.collison@arm.com>
11297 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11298 * config/aarch64/aarch64-option-extension.def: Add
11299 AARCH64_OPT_EXTENSION of 'fp16fml'.
11300 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11301 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11302 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11303 * config/aarch64/constraints.md (Ui7): New constraint.
11304 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11305 (VFMLA_SEL_W): Ditto.
11308 (VFMLA16_LOW): New int iterator.
11309 (VFMLA16_HIGH): Ditto.
11310 (UNSPEC_FMLAL): New unspec.
11311 (UNSPEC_FMLSL): Ditto.
11312 (UNSPEC_FMLAL2): Ditto.
11313 (UNSPEC_FMLSL2): Ditto.
11314 (f16mac): New code attribute.
11315 * config/aarch64/aarch64-simd-builtins.def
11316 (aarch64_fmlal_lowv2sf): Ditto.
11317 (aarch64_fmlsl_lowv2sf): Ditto.
11318 (aarch64_fmlalq_lowv4sf): Ditto.
11319 (aarch64_fmlslq_lowv4sf): Ditto.
11320 (aarch64_fmlal_highv2sf): Ditto.
11321 (aarch64_fmlsl_highv2sf): Ditto.
11322 (aarch64_fmlalq_highv4sf): Ditto.
11323 (aarch64_fmlslq_highv4sf): Ditto.
11324 (aarch64_fmlal_lane_lowv2sf): Ditto.
11325 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11326 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11327 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11328 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11329 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11330 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11331 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11332 (aarch64_fmlal_lane_highv2sf): Ditto.
11333 (aarch64_fmlsl_lane_highv2sf): Ditto.
11334 (aarch64_fmlal_laneq_highv2sf): Ditto.
11335 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11336 (aarch64_fmlalq_lane_highv4sf): Ditto.
11337 (aarch64_fmlsl_lane_highv4sf): Ditto.
11338 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11339 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11340 * config/aarch64/aarch64-simd.md:
11341 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11342 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11343 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11344 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11345 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11346 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11347 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11348 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11349 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11350 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11351 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11352 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11353 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11354 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11355 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11356 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11357 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11358 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11359 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11360 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11361 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11362 (vfmlsl_low_u32): Ditto.
11363 (vfmlalq_low_u32): Ditto.
11364 (vfmlslq_low_u32): Ditto.
11365 (vfmlal_high_u32): Ditto.
11366 (vfmlsl_high_u32): Ditto.
11367 (vfmlalq_high_u32): Ditto.
11368 (vfmlslq_high_u32): Ditto.
11369 (vfmlal_lane_low_u32): Ditto.
11370 (vfmlsl_lane_low_u32): Ditto.
11371 (vfmlal_laneq_low_u32): Ditto.
11372 (vfmlsl_laneq_low_u32): Ditto.
11373 (vfmlalq_lane_low_u32): Ditto.
11374 (vfmlslq_lane_low_u32): Ditto.
11375 (vfmlalq_laneq_low_u32): Ditto.
11376 (vfmlslq_laneq_low_u32): Ditto.
11377 (vfmlal_lane_high_u32): Ditto.
11378 (vfmlsl_lane_high_u32): Ditto.
11379 (vfmlal_laneq_high_u32): Ditto.
11380 (vfmlsl_laneq_high_u32): Ditto.
11381 (vfmlalq_lane_high_u32): Ditto.
11382 (vfmlslq_lane_high_u32): Ditto.
11383 (vfmlalq_laneq_high_u32): Ditto.
11384 (vfmlslq_laneq_high_u32): Ditto.
11385 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11386 (AARCH64_FL_FOR_ARCH8_4): New.
11387 (AARCH64_ISA_F16FML): New ISA flag.
11388 (TARGET_F16FML): New feature flag for fp16fml.
11389 (doc/invoke.texi): Document new fp16fml option.
11391 2018-01-10 Michael Collison <michael.collison@arm.com>
11393 * config/aarch64/aarch64-builtins.c:
11394 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11395 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11396 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11397 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11398 (AARCH64_ISA_SHA3): New ISA flag.
11399 (TARGET_SHA3): New feature flag for sha3.
11400 * config/aarch64/iterators.md (sha512_op): New int attribute.
11401 (CRYPTO_SHA512): New int iterator.
11402 (UNSPEC_SHA512H): New unspec.
11403 (UNSPEC_SHA512H2): Ditto.
11404 (UNSPEC_SHA512SU0): Ditto.
11405 (UNSPEC_SHA512SU1): Ditto.
11406 * config/aarch64/aarch64-simd-builtins.def
11407 (aarch64_crypto_sha512hqv2di): New builtin.
11408 (aarch64_crypto_sha512h2qv2di): Ditto.
11409 (aarch64_crypto_sha512su0qv2di): Ditto.
11410 (aarch64_crypto_sha512su1qv2di): Ditto.
11411 (aarch64_eor3qv8hi): Ditto.
11412 (aarch64_rax1qv2di): Ditto.
11413 (aarch64_xarqv2di): Ditto.
11414 (aarch64_bcaxqv8hi): Ditto.
11415 * config/aarch64/aarch64-simd.md:
11416 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11417 (aarch64_crypto_sha512su0qv2di): Ditto.
11418 (aarch64_crypto_sha512su1qv2di): Ditto.
11419 (aarch64_eor3qv8hi): Ditto.
11420 (aarch64_rax1qv2di): Ditto.
11421 (aarch64_xarqv2di): Ditto.
11422 (aarch64_bcaxqv8hi): Ditto.
11423 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11424 (vsha512h2q_u64): Ditto.
11425 (vsha512su0q_u64): Ditto.
11426 (vsha512su1q_u64): Ditto.
11427 (veor3q_u16): Ditto.
11428 (vrax1q_u64): Ditto.
11429 (vxarq_u64): Ditto.
11430 (vbcaxq_u16): Ditto.
11431 * config/arm/types.md (crypto_sha512): New type attribute.
11432 (crypto_sha3): Ditto.
11433 (doc/invoke.texi): Document new sha3 option.
11435 2018-01-10 Michael Collison <michael.collison@arm.com>
11437 * config/aarch64/aarch64-builtins.c:
11438 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11439 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11440 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11441 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11442 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11443 (AARCH64_ISA_SM4): New ISA flag.
11444 (TARGET_SM4): New feature flag for sm4.
11445 * config/aarch64/aarch64-simd-builtins.def
11446 (aarch64_sm3ss1qv4si): Ditto.
11447 (aarch64_sm3tt1aq4si): Ditto.
11448 (aarch64_sm3tt1bq4si): Ditto.
11449 (aarch64_sm3tt2aq4si): Ditto.
11450 (aarch64_sm3tt2bq4si): Ditto.
11451 (aarch64_sm3partw1qv4si): Ditto.
11452 (aarch64_sm3partw2qv4si): Ditto.
11453 (aarch64_sm4eqv4si): Ditto.
11454 (aarch64_sm4ekeyqv4si): Ditto.
11455 * config/aarch64/aarch64-simd.md:
11456 (aarch64_sm3ss1qv4si): Ditto.
11457 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11458 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11459 (aarch64_sm4eqv4si): Ditto.
11460 (aarch64_sm4ekeyqv4si): Ditto.
11461 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11462 (sm3part_op): Ditto.
11463 (CRYPTO_SM3TT): Ditto.
11464 (CRYPTO_SM3PART): Ditto.
11465 (UNSPEC_SM3SS1): New unspec.
11466 (UNSPEC_SM3TT1A): Ditto.
11467 (UNSPEC_SM3TT1B): Ditto.
11468 (UNSPEC_SM3TT2A): Ditto.
11469 (UNSPEC_SM3TT2B): Ditto.
11470 (UNSPEC_SM3PARTW1): Ditto.
11471 (UNSPEC_SM3PARTW2): Ditto.
11472 (UNSPEC_SM4E): Ditto.
11473 (UNSPEC_SM4EKEY): Ditto.
11474 * config/aarch64/constraints.md (Ui2): New constraint.
11475 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11476 * config/arm/types.md (crypto_sm3): New type attribute.
11477 (crypto_sm4): Ditto.
11478 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11479 (vsm3tt1aq_u32): Ditto.
11480 (vsm3tt1bq_u32): Ditto.
11481 (vsm3tt2aq_u32): Ditto.
11482 (vsm3tt2bq_u32): Ditto.
11483 (vsm3partw1q_u32): Ditto.
11484 (vsm3partw2q_u32): Ditto.
11485 (vsm4eq_u32): Ditto.
11486 (vsm4ekeyq_u32): Ditto.
11487 (doc/invoke.texi): Document new sm4 option.
11489 2018-01-10 Michael Collison <michael.collison@arm.com>
11491 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11492 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11493 (AARCH64_FL_FOR_ARCH8_4): New.
11494 (AARCH64_FL_V8_4): New flag.
11495 (doc/invoke.texi): Document new armv8.4-a option.
11497 2018-01-10 Michael Collison <michael.collison@arm.com>
11499 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11500 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11501 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11502 * config/aarch64/aarch64-option-extension.def: Add
11503 AARCH64_OPT_EXTENSION of 'sha2'.
11504 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11505 (crypto): Disable sha2 and aes if crypto disabled.
11506 (crypto): Enable aes and sha2 if enabled.
11507 (simd): Disable sha2 and aes if simd disabled.
11508 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11510 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11511 (TARGET_SHA2): New feature flag for sha2.
11512 (TARGET_AES): New feature flag for aes.
11513 * config/aarch64/aarch64-simd.md:
11514 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11515 conditional on TARGET_AES.
11516 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11517 (aarch64_crypto_sha1hsi): Make pattern conditional
11519 (aarch64_crypto_sha1hv4si): Ditto.
11520 (aarch64_be_crypto_sha1hv4si): Ditto.
11521 (aarch64_crypto_sha1su1v4si): Ditto.
11522 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11523 (aarch64_crypto_sha1su0v4si): Ditto.
11524 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11525 (aarch64_crypto_sha256su0v4si): Ditto.
11526 (aarch64_crypto_sha256su1v4si): Ditto.
11527 (doc/invoke.texi): Document new aes and sha2 options.
11529 2018-01-10 Martin Sebor <msebor@redhat.com>
11531 PR tree-optimization/83781
11532 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11535 2018-01-11 Martin Sebor <msebor@gmail.com>
11536 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11538 PR tree-optimization/83501
11539 PR tree-optimization/81703
11541 * tree-ssa-strlen.c (get_string_cst): Rename...
11542 (get_string_len): ...to this. Handle global constants.
11543 (handle_char_store): Adjust.
11545 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11546 Jim Wilson <jimw@sifive.com>
11548 * config/riscv/riscv-protos.h (riscv_output_return): New.
11549 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11550 (riscv_attribute_table, riscv_output_return),
11551 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11552 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11553 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11555 (riscv_expand_prologue): Add early return for naked function.
11556 (riscv_expand_epilogue): Likewise.
11557 (riscv_function_ok_for_sibcall): Return false for naked function.
11558 (riscv_set_current_function): New.
11559 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11560 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11561 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11562 * doc/extend.texi (RISC-V Function Attributes): New.
11564 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11566 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11567 check for 128-bit long double before checking TCmode.
11568 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11569 128-bit long doubles before checking TFmode or TCmode.
11570 (FLOAT128_IBM_P): Likewise.
11572 2018-01-10 Martin Sebor <msebor@redhat.com>
11574 PR tree-optimization/83671
11575 * builtins.c (c_strlen): Unconditionally return zero for the empty
11577 Use -Warray-bounds for warnings.
11578 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11579 for non-constant array indices with COMPONENT_REF, arrays of
11580 arrays, and pointers to arrays.
11581 (gimple_fold_builtin_strlen): Determine and set length range for
11582 non-constant character arrays.
11584 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11586 PR middle-end/81897
11587 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11590 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11592 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11594 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11597 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11598 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11599 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11600 indexed_or_indirect_operand predicate.
11601 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11602 (*vsx_le_perm_load_v8hi): Likewise.
11603 (*vsx_le_perm_load_v16qi): Likewise.
11604 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11605 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11606 (*vsx_le_perm_store_v8hi): Likewise.
11607 (*vsx_le_perm_store_v16qi): Likewise.
11608 (eight unnamed splitters): Likewise.
11610 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11612 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11613 * config/rs6000/emmintrin.h: Likewise.
11614 * config/rs6000/mmintrin.h: Likewise.
11615 * config/rs6000/xmmintrin.h: Likewise.
11617 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11620 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11622 * tree.c (tree_nop_conversion): Return true for location wrapper
11624 (maybe_wrap_with_location): New function.
11625 (selftest::check_strip_nops): New function.
11626 (selftest::test_location_wrappers): New function.
11627 (selftest::tree_c_tests): Call it.
11628 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11629 (maybe_wrap_with_location): New decl.
11630 (EXPR_LOCATION_WRAPPER_P): New macro.
11631 (location_wrapper_p): New inline function.
11632 (tree_strip_any_location_wrapper): New inline function.
11634 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11637 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11638 stack_realign_offset for the largest alignment of stack slot
11640 (ix86_find_max_used_stack_alignment): New function.
11641 (ix86_finalize_stack_frame_flags): Use it. Set
11642 max_used_stack_alignment if we don't realign stack.
11643 * config/i386/i386.h (machine_function): Add
11644 max_used_stack_alignment.
11646 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11648 * config/arm/arm.opt (-mbranch-cost): New option.
11649 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11652 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11655 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11656 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11658 2018-01-10 Richard Biener <rguenther@suse.de>
11661 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11662 early out so it also covers the case where we have a non-NULL
11665 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11667 PR tree-optimization/83753
11668 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11669 for non-strided grouped accesses if the number of elements is 1.
11671 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11674 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11675 * i386.h (TARGET_USE_GATHER): Define.
11676 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11678 2018-01-10 Martin Liska <mliska@suse.cz>
11681 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11682 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11684 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11685 CLEANUP_NO_PARTITIONING is not set.
11687 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11689 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11690 for vectors, as a partial revert of r254296.
11691 * rtl.h (const_vec_p): Delete.
11692 (const_vec_duplicate_p): Don't test for vector CONSTs.
11693 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11694 * expmed.c (make_tree): Likewise.
11697 * common.md (E, F): Use CONSTANT_P instead of checking for
11699 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11700 checking for CONST_VECTOR.
11702 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11704 PR middle-end/83575
11705 * predict.c (force_edge_cold): Handle in more sane way edges
11706 with no prediction.
11708 2018-01-09 Carl Love <cel@us.ibm.com>
11710 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11712 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11713 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11714 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11715 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11716 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11717 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11718 * config/rs6000/rs6000-protos.h: Add extern defition for
11719 rs6000_generate_float2_double_code.
11720 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11722 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11723 (float2_v2df): Add define_expand.
11725 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11728 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11729 op_mode in the force_to_mode call.
11731 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11733 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11734 instead of checking each element individually.
11735 (aarch64_evpc_uzp): Likewise.
11736 (aarch64_evpc_zip): Likewise.
11737 (aarch64_evpc_ext): Likewise.
11738 (aarch64_evpc_rev): Likewise.
11739 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11740 instead of checking each element individually. Return true without
11742 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11743 whether all selected elements come from the same input, instead of
11744 checking each element individually. Remove calls to gen_rtx_REG,
11745 start_sequence and end_sequence and instead assert that no rtl is
11748 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11750 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11751 order of HIGH and CONST checks.
11753 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11755 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11756 if the destination isn't an SSA_NAME.
11758 2018-01-09 Richard Biener <rguenther@suse.de>
11760 PR tree-optimization/83668
11761 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11763 (canonicalize_loop_form): ... here, renamed from ...
11764 (canonicalize_loop_closed_ssa_form): ... this and amended to
11765 swap successor edges for loop exit blocks to make us use
11766 the RPO order we need for initial schedule generation.
11768 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11770 PR tree-optimization/64811
11771 * match.pd: When optimizing comparisons with Inf, avoid
11772 introducing or losing exceptions from comparisons with NaN.
11774 2018-01-09 Martin Liska <mliska@suse.cz>
11777 * asan.c (shadow_mem_size): Add gcc_assert.
11779 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11781 Don't save registers in main().
11784 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11785 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11786 * config/avr/avr.c (avr_set_current_function): Don't error if
11787 naked, OS_task or OS_main are specified at the same time.
11788 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11790 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11792 * common/config/avr/avr-common.c (avr_option_optimization_table):
11793 Switch on -mmain-is-OS_task for optimizing compilations.
11795 2018-01-09 Richard Biener <rguenther@suse.de>
11797 PR tree-optimization/83572
11798 * graphite.c: Include cfganal.h.
11799 (graphite_transform_loops): Connect infinite loops to exit
11800 and remove fake edges at the end.
11802 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11804 * ipa-inline.c (edge_badness): Revert accidental checkin.
11806 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11809 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11810 symbols; not inline clones.
11812 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11815 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11816 hard registers. Formatting fixes.
11818 PR preprocessor/83722
11819 * gcc.c (try_generate_repro): Pass
11820 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11821 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11824 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11825 Kito Cheng <kito.cheng@gmail.com>
11827 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11828 (riscv_leaf_function_p): Delete.
11829 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11831 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11833 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11835 (do_ifelse): New function.
11836 (do_isel): New function.
11837 (do_sub3): New function.
11838 (do_add3): New function.
11839 (do_load_mask_compare): New function.
11840 (do_overlap_load_compare): New function.
11841 (expand_compare_loop): New function.
11842 (expand_block_compare): Call expand_compare_loop() when appropriate.
11843 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11844 option description.
11845 (-mblock-compare-inline-loop-limit): New option.
11847 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11850 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11851 Reverse order of second and third operands in first alternative.
11852 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11853 of first and second elements in UNSPEC_VPERMR vector.
11854 (altivec_expand_vec_perm_le): Likewise.
11856 2018-01-08 Jeff Law <law@redhat.com>
11858 PR rtl-optimizatin/81308
11859 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11860 (process_switch): If group_case_labels makes a change, then set
11862 (pass_convert_switch::execute): If a switch is converted, then
11863 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11865 PR rtl-optimization/81308
11866 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11869 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11871 PR target/83663 - Revert r255946
11872 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11873 generation for cases where splatting a value is not useful.
11874 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11875 across a vec_duplicate and a paradoxical subreg forming a vector
11876 mode to a vec_concat.
11878 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11880 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11881 -march=armv8.3-a variants.
11882 * config/arm/t-multilib: Likewise.
11883 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11885 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11887 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11889 (cceq_ior_compare_complement): Give it a name so I can use it, and
11890 change boolean_or_operator predicate to boolean_operator so it can
11891 be used to generate a crand.
11892 (eqne): New code iterator.
11893 (bd/bd_neg): New code_attrs.
11894 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11895 a single define_insn.
11896 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11897 decrement (bdnzt/bdnzf/bdzt/bdzf).
11898 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11899 with the new names of the branch decrement patterns, and added the
11900 names of the branch decrement conditional patterns.
11902 2018-01-08 Richard Biener <rguenther@suse.de>
11904 PR tree-optimization/83563
11905 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11908 2018-01-08 Richard Biener <rguenther@suse.de>
11910 PR middle-end/83713
11911 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11913 2018-01-08 Richard Biener <rguenther@suse.de>
11915 PR tree-optimization/83685
11916 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11917 references to abnormals.
11919 2018-01-08 Richard Biener <rguenther@suse.de>
11922 * dwarf2out.c (output_indirect_strings): Handle empty
11923 skeleton_debug_str_hash.
11924 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11926 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11928 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11929 (emit_store_direct): Likewise.
11930 (arc_trampoline_adjust_address): Likewise.
11931 (arc_asm_trampoline_template): New function.
11932 (arc_initialize_trampoline): Use asm_trampoline_template.
11933 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11934 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11935 * config/arc/arc.md (flush_icache): Delete pattern.
11937 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11939 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11940 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11943 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11946 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11947 by not USED_FOR_TARGET.
11948 (make_pass_resolve_sw_modes): Likewise.
11950 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11952 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11955 2018-01-08 Richard Biener <rguenther@suse.de>
11957 PR middle-end/83580
11958 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11960 2018-01-08 Richard Biener <rguenther@suse.de>
11962 PR middle-end/83517
11963 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11965 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11967 PR middle-end/81897
11968 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11969 basic blocks with a small number of successors.
11970 (convert_control_dep_chain_into_preds): Improve handling of
11972 (dump_predicates): Split apart into...
11973 (dump_pred_chain): ...here...
11974 (dump_pred_info): ...and here.
11975 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11976 (can_chain_union_be_invalidated_p): Improve check for invalidation
11978 (uninit_uses_cannot_happen): Avoid unnecessary if
11979 convert_control_dep_chain_into_preds yielded nothing.
11981 2018-01-06 Martin Sebor <msebor@redhat.com>
11983 PR tree-optimization/83640
11984 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11985 subtracting negative offset from size.
11986 (builtin_access::overlap): Adjust offset bounds of the access to fall
11987 within the size of the object if possible.
11989 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11991 PR rtl-optimization/83699
11992 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11993 extract_bit_field_as_subreg to cases in which the extracted
11994 value is also a vector.
11996 * lra-constraints.c (process_alt_operands): Test for the equivalence
11997 substitutions when detecting a possible reload cycle.
11999 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12002 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12003 by default if flag_selective_schedling{,2}. Formatting fixes.
12005 PR rtl-optimization/83682
12006 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12007 if it has non-VECTOR_MODE element mode.
12008 (vec_duplicate_p): Likewise.
12010 PR middle-end/83694
12011 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12012 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12014 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12017 * config/i386/i386-builtin.def
12018 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12019 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12020 Require also OPTION_MASK_ISA_AVX512F in addition to
12021 OPTION_MASK_ISA_GFNI.
12022 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12023 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12024 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12025 to OPTION_MASK_ISA_GFNI.
12026 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12027 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12028 OPTION_MASK_ISA_AVX512BW.
12029 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12030 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12031 addition to OPTION_MASK_ISA_GFNI.
12032 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12033 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12034 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12035 to OPTION_MASK_ISA_GFNI.
12036 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12037 a requirement for all ISAs rather than any of them with a few
12039 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12041 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12042 bitmasks to be enabled with 3 exceptions, instead of requiring any
12043 enabled ISA with lots of exceptions.
12044 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12045 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12046 Change avx512bw in isa attribute to avx512f.
12047 * config/i386/sgxintrin.h: Add license boilerplate.
12048 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12049 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12050 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12051 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12053 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12054 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12055 temporarily sse2 rather than sse if not enabled already.
12058 * config/i386/sse.md (VI248_VLBW): Rename to ...
12059 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12060 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12061 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12062 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12063 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12064 mode iterator instead of VI248_VLBW.
12066 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12068 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12069 (record_modified): Skip clobbers; add debug output.
12070 (param_change_prob): Use sreal frequencies.
12072 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12074 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12075 punt for user-aligned variables.
12077 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12079 * tree-chrec.c (chrec_contains_symbols): Return true for
12082 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12085 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12086 of (x|y) == x for BICS pattern.
12088 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12090 PR tree-optimization/83605
12091 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12092 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12095 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12097 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12098 * config/epiphany/rtems.h: New file.
12100 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12101 Uros Bizjak <ubizjak@gmail.com>
12104 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12105 QIreg_operand instead of register_operand predicate.
12106 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12107 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12108 comments instead of -fmitigate[-_]rop.
12110 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12113 * cgraphunit.c (symbol_table::compile): Switch to text_section
12114 before calling assembly_start debug hook.
12115 * run-rtl-passes.c (run_rtl_passes): Likewise.
12118 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12120 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12121 range_int_cst_p rather than !symbolic_range_p before calling
12122 extract_range_from_multiplicative_op_1.
12124 2018-01-04 Jeff Law <law@redhat.com>
12126 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12127 redundant test in assertion.
12129 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12131 * doc/rtl.texi: Document machine_mode wrapper classes.
12133 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12135 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12136 using tree_to_uhwi.
12138 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12140 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12141 the VEC_PERM_EXPR fold to fail.
12143 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12146 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12147 to switched_sections.
12149 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12152 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12153 test for d.testing.
12155 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12158 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12159 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12161 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12164 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12165 is BLKmode and bitpos not zero or mode change is needed.
12167 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12170 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12173 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12176 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12177 instead of MULT rtx. Update all corresponding splitters.
12178 (*saddl_se): Ditto.
12179 (*ssub<modesuffix>): Ditto.
12180 (*ssubl_se): Ditto.
12181 (*cmp_sadd_di): Update split patterns.
12182 (*cmp_sadd_si): Ditto.
12183 (*cmp_sadd_sidi): Ditto.
12184 (*cmp_ssub_di): Ditto.
12185 (*cmp_ssub_si): Ditto.
12186 (*cmp_ssub_sidi): Ditto.
12187 * config/alpha/predicates.md (const23_operand): New predicate.
12188 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12189 Look for ASHIFT, not MULT inner operand.
12190 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12192 2018-01-04 Martin Liska <mliska@suse.cz>
12194 PR gcov-profile/83669
12195 * gcov.c (output_intermediate_file): Add version to intermediate
12197 * doc/gcov.texi: Document new field 'version' in intermediate
12198 file format. Fix location of '-k' option of gcov command.
12200 2018-01-04 Martin Liska <mliska@suse.cz>
12203 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12205 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12207 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12209 2018-01-03 Martin Sebor <msebor@redhat.com>
12211 PR tree-optimization/83655
12212 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12213 checking calls with invalid arguments.
12215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12217 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12218 (vectorizable_mask_load_store): Delete.
12219 (vectorizable_call): Return false for masked loads and stores.
12220 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12221 instead of gimple_assign_rhs1.
12222 (vectorizable_load): Handle IFN_MASK_LOAD.
12223 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12227 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12229 (vectorizable_mask_load_store): ...here.
12230 (vectorizable_load): ...and here.
12232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12234 * tree-vect-stmts.c (vect_build_all_ones_mask)
12235 (vect_build_zero_merge_argument): New functions, split out from...
12236 (vectorizable_load): ...here.
12238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12240 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12242 (vectorizable_mask_load_store): ...here.
12243 (vectorizable_store): ...and here.
12245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12247 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12249 (vectorizable_mask_load_store): ...here.
12251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12253 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12254 (vect_model_store_cost): Take a vec_load_store_type instead of a
12256 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12257 (vect_model_store_cost): Take a vec_load_store_type instead of a
12259 (vectorizable_mask_load_store): Update accordingly.
12260 (vectorizable_store): Likewise.
12261 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12265 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12266 IFN_MASK_LOAD calls here rather than...
12267 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12270 Alan Hayward <alan.hayward@arm.com>
12271 David Sherwood <david.sherwood@arm.com>
12273 * expmed.c (extract_bit_field_1): For vector extracts,
12274 fall back to extract_bit_field_as_subreg if vec_extract
12277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12278 Alan Hayward <alan.hayward@arm.com>
12279 David Sherwood <david.sherwood@arm.com>
12281 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12282 they are variable or constant sized.
12283 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12284 slots for constant-sized data.
12286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12287 Alan Hayward <alan.hayward@arm.com>
12288 David Sherwood <david.sherwood@arm.com>
12290 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12291 handling COND_EXPRs with boolean comparisons, try to find a better
12292 basis for the mask type than the boolean itself.
12294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12296 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12297 is calculated and how it can be overridden.
12298 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12299 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12301 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12305 Alan Hayward <alan.hayward@arm.com>
12306 David Sherwood <david.sherwood@arm.com>
12308 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12309 Remove the mode argument.
12310 (aarch64_simd_valid_immediate): Remove the mode and inverse
12312 * config/aarch64/iterators.md (bitsize): New iterator.
12313 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12314 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12315 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12316 aarch64_simd_valid_immediate.
12317 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12318 (aarch64_reg_or_bic_imm): Likewise.
12319 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12320 with an insn_type enum and msl with a modifier_type enum.
12321 Replace element_width with a scalar_mode. Change the shift
12322 to unsigned int. Add constructors for scalar_float_mode and
12323 scalar_int_mode elements.
12324 (aarch64_vect_float_const_representable_p): Delete.
12325 (aarch64_can_const_movi_rtx_p)
12326 (aarch64_simd_scalar_immediate_valid_for_move)
12327 (aarch64_simd_make_constant): Update call to
12328 aarch64_simd_valid_immediate.
12329 (aarch64_advsimd_valid_immediate_hs): New function.
12330 (aarch64_advsimd_valid_immediate): Likewise.
12331 (aarch64_simd_valid_immediate): Remove mode and inverse
12332 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12333 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12334 and aarch64_float_const_representable_p on the result.
12335 (aarch64_output_simd_mov_immediate): Remove mode argument.
12336 Update call to aarch64_simd_valid_immediate and use of
12337 simd_immediate_info.
12338 (aarch64_output_scalar_simd_mov_immediate): Update call
12341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12342 Alan Hayward <alan.hayward@arm.com>
12343 David Sherwood <david.sherwood@arm.com>
12345 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12346 (mode_nunits): Likewise CONST_MODE_NUNITS.
12347 * machmode.def (ADJUST_NUNITS): Document.
12348 * genmodes.c (mode_data::need_nunits_adj): New field.
12349 (blank_mode): Update accordingly.
12350 (adj_nunits): New variable.
12351 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12353 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12354 listed in adj_nunits.
12355 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12356 listed in adj_nunits. Don't emit case statements for such modes.
12357 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12358 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12359 nothing if adj_nunits is nonnull.
12360 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12361 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12362 (emit_mode_fbit): Update use of print_maybe_const_decl.
12363 (emit_move_size): Likewise. Treat the array as non-const
12365 (emit_mode_adjustments): Handle adj_nunits.
12367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12369 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12370 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12371 (VECTOR_MODES): Use it.
12372 (make_vector_modes): Take the prefix as an argument.
12374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12375 Alan Hayward <alan.hayward@arm.com>
12376 David Sherwood <david.sherwood@arm.com>
12378 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12379 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12380 for MODE_VECTOR_BOOL.
12381 * machmode.def (VECTOR_BOOL_MODE): Document.
12382 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12383 (make_vector_bool_mode): New function.
12384 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12386 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12387 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12389 * stor-layout.c (int_mode_for_mode): Likewise.
12390 * tree.c (build_vector_type_for_mode): Likewise.
12391 * varasm.c (output_constant_pool_2): Likewise.
12392 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12393 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12394 for MODE_VECTOR_BOOL.
12395 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12396 of mode class checks.
12397 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12398 instead of a list of mode class checks.
12399 (expand_vector_scalar_condition): Likewise.
12400 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12403 Alan Hayward <alan.hayward@arm.com>
12404 David Sherwood <david.sherwood@arm.com>
12406 * machmode.h (mode_size): Change from unsigned short to
12408 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12409 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12410 or if measurement_type is not polynomial.
12411 (fixed_size_mode::includes_p): Check for constant-sized modes.
12412 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12413 return a poly_uint16 rather than an unsigned short.
12414 (emit_mode_size): Change the type of mode_size from unsigned short
12415 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12416 (emit_mode_adjustments): Cope with polynomial vector sizes.
12417 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12419 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12421 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12422 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12423 * caller-save.c (setup_save_areas): Likewise.
12424 (replace_reg_with_saved_mem): Likewise.
12425 * calls.c (emit_library_call_value_1): Likewise.
12426 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12427 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12428 (gen_lowpart_for_combine): Likewise.
12429 * convert.c (convert_to_integer_1): Likewise.
12430 * cse.c (equiv_constant, cse_insn): Likewise.
12431 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12432 (cselib_subst_to_values): Likewise.
12433 * dce.c (word_dce_process_block): Likewise.
12434 * df-problems.c (df_word_lr_mark_ref): Likewise.
12435 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12436 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12437 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12438 (rtl_for_decl_location): Likewise.
12439 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12440 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12441 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12442 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12443 (expand_expr_real_1): Likewise.
12444 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12445 (pad_below): Likewise.
12446 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12447 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12448 * ira.c (get_subreg_tracking_sizes): Likewise.
12449 * ira-build.c (ira_create_allocno_objects): Likewise.
12450 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12451 (ira_sort_regnos_for_alter_reg): Likewise.
12452 * ira-costs.c (record_operand_costs): Likewise.
12453 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12454 (resolve_simple_move): Likewise.
12455 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12456 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12457 (lra_constraints): Likewise.
12458 (CONST_POOL_OK_P): Reject variable-sized modes.
12459 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12460 (add_pseudo_to_slot, lra_spill): Likewise.
12461 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12462 * optabs-query.c (get_best_extraction_insn): Likewise.
12463 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12464 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12465 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12466 * recog.c (offsettable_address_addr_space_p): Likewise.
12467 * regcprop.c (maybe_mode_change): Likewise.
12468 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12469 * regrename.c (build_def_use): Likewise.
12470 * regstat.c (dump_reg_info): Likewise.
12471 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12472 (find_reloads, find_reloads_subreg_address): Likewise.
12473 * reload1.c (eliminate_regs_1): Likewise.
12474 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12475 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12476 (simplify_binary_operation_1, simplify_subreg): Likewise.
12477 * targhooks.c (default_function_arg_padding): Likewise.
12478 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12479 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12480 (verify_gimple_assign_ternary): Likewise.
12481 * tree-inline.c (estimate_move_cost): Likewise.
12482 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12483 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12484 (get_address_cost_ainc): Likewise.
12485 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12486 (vect_supportable_dr_alignment): Likewise.
12487 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12488 (vectorizable_reduction): Likewise.
12489 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12490 (vectorizable_operation, vectorizable_load): Likewise.
12491 * tree.c (build_same_sized_truth_vector_type): Likewise.
12492 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12493 * var-tracking.c (emit_note_insn_var_location): Likewise.
12494 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12495 (ADDR_VEC_ALIGN): Likewise.
12497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12498 Alan Hayward <alan.hayward@arm.com>
12499 David Sherwood <david.sherwood@arm.com>
12501 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12503 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12504 or if measurement_type is polynomial.
12505 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12506 * combine.c (make_extraction): Likewise.
12507 * dse.c (find_shift_sequence): Likewise.
12508 * dwarf2out.c (mem_loc_descriptor): Likewise.
12509 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12510 (extract_bit_field, extract_low_bits): Likewise.
12511 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12512 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12513 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12514 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12515 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12516 * reload.c (find_reloads): Likewise.
12517 * reload1.c (alter_reg): Likewise.
12518 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12519 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12520 * tree-if-conv.c (predicate_mem_writes): Likewise.
12521 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12522 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12523 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12524 * valtrack.c (dead_debug_insert_temp): Likewise.
12525 * varasm.c (mergeable_constant_section): Likewise.
12526 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12529 Alan Hayward <alan.hayward@arm.com>
12530 David Sherwood <david.sherwood@arm.com>
12532 * expr.c (expand_assignment): Cope with polynomial mode sizes
12533 when assigning to a CONCAT.
12535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12536 Alan Hayward <alan.hayward@arm.com>
12537 David Sherwood <david.sherwood@arm.com>
12539 * machmode.h (mode_precision): Change from unsigned short to
12541 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12543 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12544 or if measurement_type is not polynomial.
12545 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12546 in which the mode is already known to be a scalar_int_mode.
12547 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12548 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12550 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12551 for GET_MODE_PRECISION.
12552 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12553 for GET_MODE_PRECISION.
12554 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12556 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12557 (expand_field_assignment, make_extraction): Likewise.
12558 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12559 (get_last_value): Likewise.
12560 * convert.c (convert_to_integer_1): Likewise.
12561 * cse.c (cse_insn): Likewise.
12562 * expr.c (expand_expr_real_1): Likewise.
12563 * lra-constraints.c (simplify_operand_subreg): Likewise.
12564 * optabs-query.c (can_atomic_load_p): Likewise.
12565 * optabs.c (expand_atomic_load): Likewise.
12566 (expand_atomic_store): Likewise.
12567 * ree.c (combine_reaching_defs): Likewise.
12568 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12569 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12570 * tree.h (type_has_mode_precision_p): Likewise.
12571 * ubsan.c (instrument_si_overflow): Likewise.
12573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12574 Alan Hayward <alan.hayward@arm.com>
12575 David Sherwood <david.sherwood@arm.com>
12577 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12578 polynomial numbers of units.
12579 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12580 (valid_vector_subparts_p): New function.
12581 (build_vector_type): Remove temporary shim and take the number
12582 of units as a poly_uint64 rather than an int.
12583 (build_opaque_vector_type): Take the number of units as a
12584 poly_uint64 rather than an int.
12585 * tree.c (build_vector_from_ctor): Handle polynomial
12586 TYPE_VECTOR_SUBPARTS.
12587 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12588 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12589 (build_vector_from_val): If the number of units is variable,
12590 use build_vec_duplicate_cst for constant operands and
12591 VEC_DUPLICATE_EXPR otherwise.
12592 (make_vector_type): Remove temporary is_constant ().
12593 (build_vector_type, build_opaque_vector_type): Take the number of
12594 units as a poly_uint64 rather than an int.
12595 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12597 * cfgexpand.c (expand_debug_expr): Likewise.
12598 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12599 (store_constructor, expand_expr_real_1): Likewise.
12600 (const_scalar_mask_from_tree): Likewise.
12601 * fold-const-call.c (fold_const_reduction): Likewise.
12602 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12603 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12604 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12605 (fold_relational_const): Likewise.
12606 (native_interpret_vector): Likewise. Change the size from an
12607 int to an unsigned int.
12608 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12609 TYPE_VECTOR_SUBPARTS.
12610 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12611 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12612 duplicating a non-constant operand into a variable-length vector.
12613 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12614 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12615 * ipa-icf.c (sem_variable::equals): Likewise.
12616 * match.pd: Likewise.
12617 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12618 * print-tree.c (print_node): Likewise.
12619 * stor-layout.c (layout_type): Likewise.
12620 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12621 * tree-cfg.c (verify_gimple_comparison): Likewise.
12622 (verify_gimple_assign_binary): Likewise.
12623 (verify_gimple_assign_ternary): Likewise.
12624 (verify_gimple_assign_single): Likewise.
12625 * tree-pretty-print.c (dump_generic_node): Likewise.
12626 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12627 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12628 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12629 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12630 (vect_shift_permute_load_chain): Likewise.
12631 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12632 (expand_vector_condition, optimize_vector_constructor): Likewise.
12633 (lower_vec_perm, get_compute_type): Likewise.
12634 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12635 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12636 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12637 (vect_recog_mask_conversion_pattern): Likewise.
12638 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12639 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12640 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12641 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12642 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12643 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12644 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12645 (supportable_widening_operation): Likewise.
12646 (supportable_narrowing_operation): Likewise.
12647 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12649 * varasm.c (output_constant): Likewise.
12651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12652 Alan Hayward <alan.hayward@arm.com>
12653 David Sherwood <david.sherwood@arm.com>
12655 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12656 so that both the length == 3 and length != 3 cases set up their
12657 own permute vectors. Add comments explaining why we know the
12658 number of elements is constant.
12659 (vect_permute_load_chain): Likewise.
12661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12662 Alan Hayward <alan.hayward@arm.com>
12663 David Sherwood <david.sherwood@arm.com>
12665 * machmode.h (mode_nunits): Change from unsigned char to
12667 (ONLY_FIXED_SIZE_MODES): New macro.
12668 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12669 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12670 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12672 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12673 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12674 or if measurement_type is not polynomial.
12675 * genmodes.c (ZERO_COEFFS): New macro.
12676 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12678 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12679 Use ZERO_COEFFS when emitting initializers.
12680 * data-streamer.h (bp_pack_poly_value): New function.
12681 (bp_unpack_poly_value): Likewise.
12682 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12683 for GET_MODE_NUNITS.
12684 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12685 for GET_MODE_NUNITS.
12686 * tree.c (make_vector_type): Remove temporary shim and make
12687 the real function take the number of units as a poly_uint64
12688 rather than an int.
12689 (build_vector_type_for_mode): Handle polynomial nunits.
12690 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12691 * emit-rtl.c (const_vec_series_p_1): Likewise.
12692 (gen_rtx_CONST_VECTOR): Likewise.
12693 * fold-const.c (test_vec_duplicate_folding): Likewise.
12694 * genrecog.c (validate_pattern): Likewise.
12695 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12696 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12697 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12698 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12699 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12700 * rtlanal.c (subreg_get_info): Likewise.
12701 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12702 (vect_grouped_load_supported): Likewise.
12703 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12704 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12705 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12706 (simplify_const_unary_operation, simplify_binary_operation_1)
12707 (simplify_const_binary_operation, simplify_ternary_operation)
12708 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12709 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12710 instead of CONST_VECTOR_NUNITS.
12711 * varasm.c (output_constant_pool_2): Likewise.
12712 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12713 explicit-encoded elements in the XVEC for variable-length vectors.
12715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12717 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12720 Alan Hayward <alan.hayward@arm.com>
12721 David Sherwood <david.sherwood@arm.com>
12723 * coretypes.h (fixed_size_mode): Declare.
12724 (fixed_size_mode_pod): New typedef.
12725 * builtins.h (target_builtins::x_apply_args_mode)
12726 (target_builtins::x_apply_result_mode): Change type to
12727 fixed_size_mode_pod.
12728 * builtins.c (apply_args_size, apply_result_size, result_vector)
12729 (expand_builtin_apply_args_1, expand_builtin_apply)
12730 (expand_builtin_return): Update accordingly.
12732 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12734 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12735 * cselib.c (cselib_hash_rtx): Likewise.
12736 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12737 CONST_VECTOR encoding.
12739 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12740 Jeff Law <law@redhat.com>
12743 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12744 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12745 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12746 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12749 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12750 explicitly probe *sp in a noreturn function if there were any callee
12751 register saves or frame pointer is needed.
12753 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12756 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12757 BLKmode for ternary, binary or unary expressions.
12760 * var-tracking.c (delete_vta_debug_insn): New inline function.
12761 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12762 insns from get_insns () to NULL instead of each bb separately.
12763 Use delete_vta_debug_insn. No longer static.
12764 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12765 delete_vta_debug_insns callers.
12766 * rtl.h (delete_vta_debug_insns): Declare.
12767 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12768 instead of variable_tracking_main.
12770 2018-01-03 Martin Sebor <msebor@redhat.com>
12772 PR tree-optimization/83603
12773 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12774 arguments past the endof the argument list in functions declared
12775 without a prototype.
12776 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12777 Avoid checking when arguments are null.
12779 2018-01-03 Martin Sebor <msebor@redhat.com>
12782 * doc/extend.texi (attribute const): Fix a typo.
12783 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12784 issuing -Wsuggest-attribute for void functions.
12786 2018-01-03 Martin Sebor <msebor@redhat.com>
12788 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12789 offset_int::from instead of wide_int::to_shwi.
12790 (maybe_diag_overlap): Remove assertion.
12791 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12792 * gimple-ssa-sprintf.c (format_directive): Same.
12793 (parse_directive): Same.
12794 (sprintf_dom_walker::compute_format_length): Same.
12795 (try_substitute_return_value): Same.
12797 2018-01-03 Jeff Law <law@redhat.com>
12799 PR middle-end/83654
12800 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12801 non-constant residual for zero at runtime and avoid probing in
12802 that case. Reorganize code for trailing problem to mirror handling
12805 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12807 PR tree-optimization/83501
12808 * tree-ssa-strlen.c (get_string_cst): New.
12809 (handle_char_store): Call get_string_cst.
12811 2018-01-03 Martin Liska <mliska@suse.cz>
12813 PR tree-optimization/83593
12814 * tree-ssa-strlen.c: Include tree-cfg.h.
12815 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12816 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12817 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12819 (strlen_dom_walker::before_dom_children): Call
12820 gimple_purge_dead_eh_edges. Dump tranformation with details
12822 (strlen_dom_walker::before_dom_children): Update call by adding
12823 new argument cleanup_eh.
12824 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12826 2018-01-03 Martin Liska <mliska@suse.cz>
12829 * cif-code.def (VARIADIC_THUNK): New enum value.
12830 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12833 2018-01-03 Jan Beulich <jbeulich@suse.com>
12835 * sse.md (mov<mode>_internal): Tighten condition for when to use
12836 vmovdqu<ssescalarsize> for TI and OI modes.
12838 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12840 Update copyright years.
12842 2018-01-03 Martin Liska <mliska@suse.cz>
12845 * ipa-visibility.c (function_and_variable_visibility): Skip
12846 functions with noipa attribure.
12848 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12850 * gcc.c (process_command): Update copyright notice dates.
12851 * gcov-dump.c (print_version): Ditto.
12852 * gcov.c (print_version): Ditto.
12853 * gcov-tool.c (print_version): Ditto.
12854 * gengtype.c (create_file): Ditto.
12855 * doc/cpp.texi: Bump @copying's copyright year.
12856 * doc/cppinternals.texi: Ditto.
12857 * doc/gcc.texi: Ditto.
12858 * doc/gccint.texi: Ditto.
12859 * doc/gcov.texi: Ditto.
12860 * doc/install.texi: Ditto.
12861 * doc/invoke.texi: Ditto.
12863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12865 * vector-builder.h (vector_builder::m_full_nelts): Change from
12866 unsigned int to poly_uint64.
12867 (vector_builder::full_nelts): Update prototype accordingly.
12868 (vector_builder::new_vector): Likewise.
12869 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12870 (vector_builder::operator ==): Likewise.
12871 (vector_builder::finalize): Likewise.
12872 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12873 Take the number of elements as a poly_uint64 rather than an
12875 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12876 from unsigned int to poly_uint64.
12877 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12878 (vec_perm_indices::new_vector): Likewise.
12879 (vec_perm_indices::length): Likewise.
12880 (vec_perm_indices::nelts_per_input): Likewise.
12881 (vec_perm_indices::input_nelts): Likewise.
12882 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12883 number of elements per input as a poly_uint64 rather than an
12884 unsigned int. Use the original encoding for variable-length
12885 vectors, rather than clamping each individual element.
12886 For the second and subsequent elements in each pattern,
12887 clamp the step and base before clamping their sum.
12888 (vec_perm_indices::series_p): Handle polynomial element counts.
12889 (vec_perm_indices::all_in_range_p): Likewise.
12890 (vec_perm_indices_to_tree): Likewise.
12891 (vec_perm_indices_to_rtx): Likewise.
12892 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12893 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12894 (tree_vector_builder::new_binary_operation): Handle polynomial
12895 element counts. Return false if we need to know the number
12896 of elements at compile time.
12897 * fold-const.c (fold_vec_perm): Punt if the number of elements
12898 isn't known at compile time.
12900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12902 * vec-perm-indices.h (vec_perm_builder): Change element type
12903 from HOST_WIDE_INT to poly_int64.
12904 (vec_perm_indices::element_type): Update accordingly.
12905 (vec_perm_indices::clamp): Handle polynomial element_types.
12906 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12907 (vec_perm_indices::all_in_range_p): Likewise.
12908 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12910 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12911 polynomial vec_perm_indices element types.
12912 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12913 * fold-const.c (fold_vec_perm): Likewise.
12914 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12915 * tree-vect-generic.c (lower_vec_perm): Likewise.
12916 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12917 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12918 element type to HOST_WIDE_INT.
12920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12921 Alan Hayward <alan.hayward@arm.com>
12922 David Sherwood <david.sherwood@arm.com>
12924 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12925 rather than an int. Use plus_constant.
12926 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12927 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12930 Alan Hayward <alan.hayward@arm.com>
12931 David Sherwood <david.sherwood@arm.com>
12933 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12934 a HOST_WIDE_INT to a poly_int64.
12936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12937 Alan Hayward <alan.hayward@arm.com>
12938 David Sherwood <david.sherwood@arm.com>
12940 * calls.c (load_register_parameters): Cope with polynomial
12941 mode sizes. Require a constant size for BLKmode parameters
12942 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12943 forces a parameter to be padded at the lsb end in order to
12944 fill a complete number of words, require the parameter size
12945 to be ordered wrt UNITS_PER_WORD.
12947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12948 Alan Hayward <alan.hayward@arm.com>
12949 David Sherwood <david.sherwood@arm.com>
12951 * reload1.c (spill_stack_slot_width): Change element type
12952 from unsigned int to poly_uint64_pod.
12953 (alter_reg): Treat mode sizes as polynomial.
12955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12956 Alan Hayward <alan.hayward@arm.com>
12957 David Sherwood <david.sherwood@arm.com>
12959 * reload.c (complex_word_subreg_p): New function.
12960 (reload_inner_reg_of_subreg, push_reload): Use it.
12962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12963 Alan Hayward <alan.hayward@arm.com>
12964 David Sherwood <david.sherwood@arm.com>
12966 * lra-constraints.c (process_alt_operands): Reject matched
12967 operands whose sizes aren't ordered.
12968 (match_reload): Refer to this check here.
12970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12971 Alan Hayward <alan.hayward@arm.com>
12972 David Sherwood <david.sherwood@arm.com>
12974 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12975 that the mode size is in the set {1, 2, 4, 8, 16}.
12977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12978 Alan Hayward <alan.hayward@arm.com>
12979 David Sherwood <david.sherwood@arm.com>
12981 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12982 Use plus_constant instead of gen_rtx_PLUS.
12984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12985 Alan Hayward <alan.hayward@arm.com>
12986 David Sherwood <david.sherwood@arm.com>
12988 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12989 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12990 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12991 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12992 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12993 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12994 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12995 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12996 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12997 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12999 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13000 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13001 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13002 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13003 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13004 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13005 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13006 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13007 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13008 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13010 * expr.c (emit_move_resolve_push): Treat the input and result
13011 of PUSH_ROUNDING as a poly_int64.
13012 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13013 (emit_push_insn): Likewise.
13014 * lra-eliminations.c (mark_not_eliminable): Likewise.
13015 * recog.c (push_operand): Likewise.
13016 * reload1.c (elimination_effects): Likewise.
13017 * rtlanal.c (nonzero_bits1): Likewise.
13018 * calls.c (store_one_arg): Likewise. Require the padding to be
13019 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13026 Use plus_constant instead of gen_rtx_PLUS.
13028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13029 Alan Hayward <alan.hayward@arm.com>
13030 David Sherwood <david.sherwood@arm.com>
13032 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13033 rather than an int.
13035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13036 Alan Hayward <alan.hayward@arm.com>
13037 David Sherwood <david.sherwood@arm.com>
13039 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13040 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13041 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
13048 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13049 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13050 as a poly_uint64 too.
13052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13053 Alan Hayward <alan.hayward@arm.com>
13054 David Sherwood <david.sherwood@arm.com>
13056 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13059 Alan Hayward <alan.hayward@arm.com>
13060 David Sherwood <david.sherwood@arm.com>
13062 * combine.c (can_change_dest_mode): Handle polynomial
13063 REGMODE_NATURAL_SIZE.
13064 * expmed.c (store_bit_field_1): Likewise.
13065 * expr.c (store_constructor): Likewise.
13066 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13067 and polynomial REGMODE_NATURAL_SIZE.
13068 (gen_lowpart_common): Likewise.
13069 * reginfo.c (record_subregs_of_mode): Likewise.
13070 * rtlanal.c (read_modify_subreg_p): Likewise.
13072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13073 Alan Hayward <alan.hayward@arm.com>
13074 David Sherwood <david.sherwood@arm.com>
13076 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13077 numbers of elements.
13079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13080 Alan Hayward <alan.hayward@arm.com>
13081 David Sherwood <david.sherwood@arm.com>
13083 * match.pd: Cope with polynomial numbers of vector elements.
13085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13086 Alan Hayward <alan.hayward@arm.com>
13087 David Sherwood <david.sherwood@arm.com>
13089 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13090 in a POINTER_PLUS_EXPR.
13092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13093 Alan Hayward <alan.hayward@arm.com>
13094 David Sherwood <david.sherwood@arm.com>
13096 * omp-simd-clone.c (simd_clone_subparts): New function.
13097 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13098 (ipa_simd_modify_function_body): Likewise.
13100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13101 Alan Hayward <alan.hayward@arm.com>
13102 David Sherwood <david.sherwood@arm.com>
13104 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13105 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13106 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13107 (expand_vector_condition, vector_element): Likewise.
13108 (subparts_gt): New function.
13109 (get_compute_type): Use subparts_gt.
13110 (count_type_subparts): Delete.
13111 (expand_vector_operations_1): Use subparts_gt instead of
13112 count_type_subparts.
13114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13115 Alan Hayward <alan.hayward@arm.com>
13116 David Sherwood <david.sherwood@arm.com>
13118 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13119 (vect_compile_time_alias): ...this new function. Do the calculation
13120 on poly_ints rather than trees.
13121 (vect_prune_runtime_alias_test_list): Update call accordingly.
13123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13124 Alan Hayward <alan.hayward@arm.com>
13125 David Sherwood <david.sherwood@arm.com>
13127 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13129 (vect_schedule_slp_instance): 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 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13136 constant and extern definitions for variable-length vectors.
13137 (vect_get_constant_vectors): Note that the number of units
13138 is known to be constant.
13140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13141 Alan Hayward <alan.hayward@arm.com>
13142 David Sherwood <david.sherwood@arm.com>
13144 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13145 of units as polynomial. Choose between WIDE and NARROW based
13148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13149 Alan Hayward <alan.hayward@arm.com>
13150 David Sherwood <david.sherwood@arm.com>
13152 * tree-vect-stmts.c (simd_clone_subparts): New function.
13153 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13156 Alan Hayward <alan.hayward@arm.com>
13157 David Sherwood <david.sherwood@arm.com>
13159 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13160 vectors as polynomial. Use build_index_vector for
13161 IFN_GOMP_SIMD_LANE.
13163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13164 Alan Hayward <alan.hayward@arm.com>
13165 David Sherwood <david.sherwood@arm.com>
13167 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13168 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13169 for variable-length vectors.
13170 (vectorizable_mask_load_store): Treat the number of units as
13171 polynomial, asserting that it is constant if the condition has
13172 already been enforced.
13173 (vectorizable_store, vectorizable_load): Likewise.
13175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13176 Alan Hayward <alan.hayward@arm.com>
13177 David Sherwood <david.sherwood@arm.com>
13179 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13180 of units as polynomial. Punt if we can't tell at compile time
13181 which vector contains the final result.
13183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13184 Alan Hayward <alan.hayward@arm.com>
13185 David Sherwood <david.sherwood@arm.com>
13187 * tree-vect-loop.c (vectorizable_induction): Treat the number
13188 of units as polynomial. Punt on SLP inductions. Use an integer
13189 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13190 cast of such a series for variable-length floating-point
13193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13194 Alan Hayward <alan.hayward@arm.com>
13195 David Sherwood <david.sherwood@arm.com>
13197 * tree.h (build_index_vector): Declare.
13198 * tree.c (build_index_vector): New function.
13199 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13200 of units as polynomial, forcibly converting it to a constant if
13201 vectorizable_reduction has already enforced the condition.
13202 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13203 to create a {1,2,3,...} vector.
13204 (vectorizable_reduction): Treat the number of units as polynomial.
13205 Choose vectype_in based on the largest scalar element size rather
13206 than the smallest number of units. Enforce the restrictions
13209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13210 Alan Hayward <alan.hayward@arm.com>
13211 David Sherwood <david.sherwood@arm.com>
13213 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13214 number of units as polynomial.
13216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13217 Alan Hayward <alan.hayward@arm.com>
13218 David Sherwood <david.sherwood@arm.com>
13220 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13221 * target.def (autovectorize_vector_sizes): Return the vector sizes
13222 by pointer, using vector_sizes rather than a bitmask.
13223 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13224 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13225 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13227 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13228 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13229 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13230 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13231 * omp-general.c (omp_max_vf): Likewise.
13232 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13233 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13234 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13235 * tree-vect-slp.c (vect_slp_bb): Likewise.
13236 * doc/tm.texi: Regenerate.
13237 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13239 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13240 the vector size as a poly_uint64 rather than an unsigned int.
13241 (current_vector_size): Change from an unsigned int to a poly_uint64.
13242 (get_vectype_for_scalar_type): Update accordingly.
13243 * tree.h (build_truth_vector_type): Take the size and number of
13244 units as a poly_uint64 rather than an unsigned int.
13245 (build_vector_type): Add a temporary overload that takes
13246 the number of units as a poly_uint64 rather than an unsigned int.
13247 * tree.c (make_vector_type): Likewise.
13248 (build_truth_vector_type): Take the number of units as a poly_uint64
13249 rather than an unsigned int.
13251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13252 Alan Hayward <alan.hayward@arm.com>
13253 David Sherwood <david.sherwood@arm.com>
13255 * target.def (get_mask_mode): Take the number of units and length
13256 as poly_uint64s rather than unsigned ints.
13257 * targhooks.h (default_get_mask_mode): Update accordingly.
13258 * targhooks.c (default_get_mask_mode): Likewise.
13259 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13260 * doc/tm.texi: Regenerate.
13262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13263 Alan Hayward <alan.hayward@arm.com>
13264 David Sherwood <david.sherwood@arm.com>
13266 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13267 * omp-general.c (omp_max_vf): Likewise.
13268 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13269 (expand_omp_simd): Handle polynomial safelen.
13270 * omp-low.c (omplow_simd_context): Add a default constructor.
13271 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13272 (lower_rec_simd_input_clauses): Update accordingly.
13273 (lower_rec_input_clauses): Likewise.
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 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13280 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13281 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13282 (vect_analyze_slp_cost): Likewise.
13283 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13284 (vect_model_load_cost): Likewise.
13286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13287 Alan Hayward <alan.hayward@arm.com>
13288 David Sherwood <david.sherwood@arm.com>
13290 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13291 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13292 from an unsigned int * to a poly_uint64_pod *.
13293 (calculate_unrolling_factor): New function.
13294 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13297 Alan Hayward <alan.hayward@arm.com>
13298 David Sherwood <david.sherwood@arm.com>
13300 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13301 from an unsigned int to a poly_uint64.
13302 (_loop_vec_info::slp_unrolling_factor): Likewise.
13303 (_loop_vec_info::vectorization_factor): Change from an int
13305 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13306 (vect_get_num_vectors): New function.
13307 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13308 (vect_get_num_copies): Use vect_get_num_vectors.
13309 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13310 to an unsigned int *.
13311 (vect_analyze_data_refs): Change min_vf from an int * to a
13313 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13314 than an unsigned HOST_WIDE_INT.
13315 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13316 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13317 to an unsigned int *.
13318 (vect_analyze_data_ref_dependences): Likewise.
13319 (vect_compute_data_ref_alignment): Handle polynomial vf.
13320 (vect_enhance_data_refs_alignment): Likewise.
13321 (vect_prune_runtime_alias_test_list): Likewise.
13322 (vect_shift_permute_load_chain): Likewise.
13323 (vect_supportable_dr_alignment): Likewise.
13324 (dependence_distance_ge_vf): Take the vectorization factor as a
13325 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13326 (vect_analyze_data_refs): Change min_vf from an int * to a
13328 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13329 vfm1 as a poly_uint64 rather than an int. Make the same change
13330 for the returned bound_scalar.
13331 (vect_gen_vector_loop_niters): Handle polynomial vf.
13332 (vect_do_peeling): Likewise. Update call to
13333 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13334 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13336 * tree-vect-loop.c (vect_determine_vectorization_factor)
13337 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13338 (vect_get_known_peeling_cost): Likewise.
13339 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13340 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13341 (vect_transform_loop): Likewise. Use the lowest possible VF when
13342 updating the upper bounds of the loop.
13343 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13344 rather than an int.
13345 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13346 polynomial unroll factors.
13347 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13348 (vect_make_slp_decision): Likewise.
13349 (vect_supported_load_permutation_p): Likewise, and polynomial
13351 (vect_analyze_slp_cost): Handle polynomial vf.
13352 (vect_slp_analyze_node_operations): Likewise.
13353 (vect_slp_analyze_bb_1): Likewise.
13354 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13355 than an unsigned HOST_WIDE_INT.
13356 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13357 (vectorizable_load): Handle polynomial vf.
13358 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13360 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13363 Alan Hayward <alan.hayward@arm.com>
13364 David Sherwood <david.sherwood@arm.com>
13366 * match.pd: Handle bit operations involving three constants
13367 and try to fold one pair.
13369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13371 * tree-vect-loop-manip.c: Include gimple-fold.h.
13372 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13373 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13374 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13375 Add a path that uses a step of VF instead of 1, but disable it
13377 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13378 and niters_no_overflow parameters. Update calls to
13379 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13380 Create a new SSA name if the latter choses to use a ste other
13381 than zero, and return it via niters_vector_mult_vf_var.
13382 * tree-vect-loop.c (vect_transform_loop): Update calls to
13383 vect_do_peeling, vect_gen_vector_loop_niters and
13384 slpeel_make_loop_iterate_ntimes.
13385 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13386 (vect_gen_vector_loop_niters): Update declarations after above changes.
13388 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13390 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13391 128-bit round to integer instructions.
13392 (ceil<mode>2): Likewise.
13393 (btrunc<mode>2): Likewise.
13394 (round<mode>2): Likewise.
13396 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13398 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13399 unaligned VSX load/store on P8/P9.
13400 (expand_block_clear): Allow the use of unaligned VSX
13401 load/store on P8/P9.
13403 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13405 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13407 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13408 swap associated with both a load and a store.
13410 2018-01-02 Andrew Waterman <andrew@sifive.com>
13412 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13413 * config/riscv/riscv.md (clear_cache): Use it.
13415 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13417 * web.c: Remove out-of-date comment.
13419 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13421 * expr.c (fixup_args_size_notes): Check that any existing
13422 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13423 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13424 (emit_single_push_insn): ...here.
13426 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13428 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13429 (const_vector_encoded_nelts): New function.
13430 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13431 (const_vector_int_elt, const_vector_elt): Declare.
13432 * emit-rtl.c (const_vector_int_elt_1): New function.
13433 (const_vector_elt): Likewise.
13434 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13435 of CONST_VECTOR_ELT.
13437 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13439 * expr.c: Include rtx-vector-builder.h.
13440 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13441 directly on the tree encoding.
13442 (const_vector_from_tree): Likewise.
13443 * optabs.c: Include rtx-vector-builder.h.
13444 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13445 sequence of "u" values.
13446 * vec-perm-indices.c: Include rtx-vector-builder.h.
13447 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13448 directly on the vec_perm_indices encoding.
13450 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13452 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13453 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13454 * rtx-vector-builder.h: New file.
13455 * rtx-vector-builder.c: Likewise.
13456 * rtl.h (rtx_def::u2): Add a const_vector field.
13457 (CONST_VECTOR_NPATTERNS): New macro.
13458 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13459 (CONST_VECTOR_DUPLICATE_P): Likewise.
13460 (CONST_VECTOR_STEPPED_P): Likewise.
13461 (CONST_VECTOR_ENCODED_ELT): Likewise.
13462 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13463 (unwrap_const_vec_duplicate): Likewise.
13464 (const_vec_series_p): Check for a non-duplicated vector encoding.
13465 Say that the function only returns true for integer vectors.
13466 * emit-rtl.c: Include rtx-vector-builder.h.
13467 (gen_const_vec_duplicate_1): Delete.
13468 (gen_const_vector): Call gen_const_vec_duplicate instead of
13469 gen_const_vec_duplicate_1.
13470 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13471 (gen_const_vec_duplicate): Use rtx_vector_builder.
13472 (gen_const_vec_series): Likewise.
13473 (gen_rtx_CONST_VECTOR): Likewise.
13474 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13475 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13476 Build a new vector rather than modifying a CONST_VECTOR in-place.
13477 (handle_special_swappables): Update call accordingly.
13478 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13479 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13480 Build a new vector rather than modifying a CONST_VECTOR in-place.
13481 (handle_special_swappables): Update call accordingly.
13483 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13485 * simplify-rtx.c (simplify_const_binary_operation): Use
13486 CONST_VECTOR_ELT instead of XVECEXP.
13488 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13490 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13491 the selector elements to be different from the data elements
13492 if the selector is a VECTOR_CST.
13493 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13494 ssizetype for the selector.
13496 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13498 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13499 before testing each element individually.
13500 * tree-vect-generic.c (lower_vec_perm): Likewise.
13502 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13504 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13505 * selftest-run-tests.c (selftest::run_tests): Call it.
13506 * vector-builder.h (vector_builder::operator ==): New function.
13507 (vector_builder::operator !=): Likewise.
13508 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13509 (vec_perm_indices::all_from_input_p): New function.
13510 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13511 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13512 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13513 instead of reading the VECTOR_CST directly. Detect whether both
13514 vector inputs are the same before constructing the vec_perm_indices,
13515 and update the number of inputs argument accordingly. Use the
13516 utility functions added above. Only construct sel2 if we need to.
13518 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13520 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13521 the broadcast of the low byte.
13522 (expand_mult_highpart): Use an explicit encoding for the permutes.
13523 * optabs-query.c (can_mult_highpart_p): Likewise.
13524 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13525 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13526 (vectorizable_bswap): Likewise.
13527 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13528 explicit encoding for the power-of-2 permutes.
13529 (vect_permute_store_chain): Likewise.
13530 (vect_grouped_load_supported): Likewise.
13531 (vect_permute_load_chain): Likewise.
13533 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13535 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13536 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13537 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13538 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13539 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13540 (vect_gen_perm_mask_any): Likewise.
13542 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13544 * int-vector-builder.h: New file.
13545 * vec-perm-indices.h: Include int-vector-builder.h.
13546 (vec_perm_indices): Redefine as an int_vector_builder.
13547 (auto_vec_perm_indices): Delete.
13548 (vec_perm_builder): Redefine as a stand-alone class.
13549 (vec_perm_indices::vec_perm_indices): New function.
13550 (vec_perm_indices::clamp): Likewise.
13551 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13552 (vec_perm_indices::new_vector): New function.
13553 (vec_perm_indices::new_expanded_vector): Update for new
13554 vec_perm_indices class.
13555 (vec_perm_indices::rotate_inputs): New function.
13556 (vec_perm_indices::all_in_range_p): Operate directly on the
13557 encoded form, without computing elided elements.
13558 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13559 encoding. Update for new vec_perm_indices class.
13560 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13561 the given vec_perm_builder.
13562 (expand_vec_perm_var): Update vec_perm_builder constructor.
13563 (expand_mult_highpart): Use vec_perm_builder instead of
13564 auto_vec_perm_indices.
13565 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13566 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13567 or double series encoding as appropriate.
13568 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13569 vec_perm_indices instead of auto_vec_perm_indices.
13570 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13571 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13572 (vect_permute_store_chain): Likewise.
13573 (vect_grouped_load_supported): Likewise.
13574 (vect_permute_load_chain): Likewise.
13575 (vect_shift_permute_load_chain): Likewise.
13576 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13577 (vect_transform_slp_perm_load): Likewise.
13578 (vect_schedule_slp_instance): Likewise.
13579 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13580 (vectorizable_mask_load_store): Likewise.
13581 (vectorizable_bswap): Likewise.
13582 (vectorizable_store): Likewise.
13583 (vectorizable_load): Likewise.
13584 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13585 vec_perm_indices instead of auto_vec_perm_indices. Use
13586 tree_to_vec_perm_builder to read the vector from a tree.
13587 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13588 vec_perm_builder instead of a vec_perm_indices.
13589 (have_whole_vector_shift): Use vec_perm_builder and
13590 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13591 truncation to calc_vec_perm_mask_for_shift.
13592 (vect_create_epilog_for_reduction): Likewise.
13593 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13594 from auto_vec_perm_indices to vec_perm_indices.
13595 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13596 instead of changing individual elements.
13597 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13598 the vector in d.perm.
13599 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13600 from auto_vec_perm_indices to vec_perm_indices.
13601 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13602 instead of changing individual elements.
13603 (arm_vectorize_vec_perm_const): Use new_vector to install
13604 the vector in d.perm.
13605 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13606 Update vec_perm_builder constructor.
13607 (rs6000_expand_interleave): Likewise.
13608 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13609 (rs6000_expand_interleave): Likewise.
13611 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13613 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13614 to qimode could truncate the indices.
13615 * optabs.c (expand_vec_perm_var): Likewise.
13617 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13619 * Makefile.in (OBJS): Add vec-perm-indices.o.
13620 * vec-perm-indices.h: New file.
13621 * vec-perm-indices.c: Likewise.
13622 * target.h (vec_perm_indices): Replace with a forward class
13624 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13625 * optabs.h: Include vec-perm-indices.h.
13626 (expand_vec_perm): Delete.
13627 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13628 (expand_vec_perm_const): Declare.
13629 * target.def (vec_perm_const_ok): Replace with...
13630 (vec_perm_const): ...this new hook.
13631 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13632 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13633 * doc/tm.texi: Regenerate.
13634 * optabs.def (vec_perm_const): Delete.
13635 * doc/md.texi (vec_perm_const): Likewise.
13636 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13637 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13638 expand_vec_perm for constant permutation vectors. Assert that
13639 the mode of variable permutation vectors is the integer equivalent
13640 of the mode that is being permuted.
13641 * optabs-query.h (selector_fits_mode_p): Declare.
13642 * optabs-query.c: Include vec-perm-indices.h.
13643 (selector_fits_mode_p): New function.
13644 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13645 is defined, instead of checking whether the vec_perm_const_optab
13646 exists. Use targetm.vectorize.vec_perm_const instead of
13647 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13648 fit in the vector mode before using a variable permute.
13649 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13650 vec_perm_indices instead of an rtx.
13651 (expand_vec_perm): Replace with...
13652 (expand_vec_perm_const): ...this new function. Take the selector
13653 as a vec_perm_indices rather than an rtx. Also take the mode of
13654 the selector. Update call to shift_amt_for_vec_perm_mask.
13655 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13656 Use vec_perm_indices::new_expanded_vector to expand the original
13657 selector into bytes. Check whether the indices fit in the vector
13658 mode before using a variable permute.
13659 (expand_vec_perm_var): Make global.
13660 (expand_mult_highpart): Use expand_vec_perm_const.
13661 * fold-const.c: Includes vec-perm-indices.h.
13662 * tree-ssa-forwprop.c: Likewise.
13663 * tree-vect-data-refs.c: Likewise.
13664 * tree-vect-generic.c: Likewise.
13665 * tree-vect-loop.c: Likewise.
13666 * tree-vect-slp.c: Likewise.
13667 * tree-vect-stmts.c: Likewise.
13668 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13670 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13671 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13672 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13673 (aarch64_vectorize_vec_perm_const): ...this new function.
13674 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13675 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13676 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13677 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13678 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13679 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13680 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13682 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13683 check for NEON modes.
13684 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13685 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13686 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13687 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13689 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13690 the old VEC_PERM_CONST conditions.
13691 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13692 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13693 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13694 (ia64_vectorize_vec_perm_const_ok): Merge into...
13695 (ia64_vectorize_vec_perm_const): ...this new function.
13696 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13697 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13698 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13699 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13700 * config/mips/mips.c (mips_expand_vec_perm_const)
13701 (mips_vectorize_vec_perm_const_ok): Merge into...
13702 (mips_vectorize_vec_perm_const): ...this new function.
13703 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13704 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13705 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13706 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13707 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13708 (rs6000_expand_vec_perm_const): Delete.
13709 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13711 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13712 (altivec_expand_vec_perm_const_le): Take each operand individually.
13713 Operate on constant selectors rather than rtxes.
13714 (altivec_expand_vec_perm_const): Likewise. Update call to
13715 altivec_expand_vec_perm_const_le.
13716 (rs6000_expand_vec_perm_const): Delete.
13717 (rs6000_vectorize_vec_perm_const_ok): Delete.
13718 (rs6000_vectorize_vec_perm_const): New function.
13719 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13720 an element count and rtx array.
13721 (rs6000_expand_extract_even): Update call accordingly.
13722 (rs6000_expand_interleave): Likewise.
13723 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13724 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13725 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13726 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13727 (rs6000_expand_vec_perm_const): Delete.
13728 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13729 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13730 (altivec_expand_vec_perm_const_le): Take each operand individually.
13731 Operate on constant selectors rather than rtxes.
13732 (altivec_expand_vec_perm_const): Likewise. Update call to
13733 altivec_expand_vec_perm_const_le.
13734 (rs6000_expand_vec_perm_const): Delete.
13735 (rs6000_vectorize_vec_perm_const_ok): Delete.
13736 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13737 reference to the SPE evmerge intructions.
13738 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13739 an element count and rtx array.
13740 (rs6000_expand_extract_even): Update call accordingly.
13741 (rs6000_expand_interleave): Likewise.
13742 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13743 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13745 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13747 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13749 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13750 vector mode and that that mode matches the mode of the data
13752 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13753 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13754 directly using expand_vec_perm_1 when forcing selectors into
13756 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13758 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13760 * optabs-query.h (can_vec_perm_p): Delete.
13761 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13762 * optabs-query.c (can_vec_perm_p): Split into...
13763 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13764 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13765 particular selector is valid.
13766 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13767 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13768 (vect_grouped_load_supported): Likewise.
13769 (vect_shift_permute_load_chain): Likewise.
13770 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13771 (vect_transform_slp_perm_load): Likewise.
13772 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13773 (vectorizable_bswap): Likewise.
13774 (vect_gen_perm_mask_checked): Likewise.
13775 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13776 implementations of variable permutation vectors into account
13777 when deciding which selector to use.
13778 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13779 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13780 with a false third argument.
13781 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13782 to test whether the constant selector is valid and can_vec_perm_var_p
13783 to test whether a variable selector is valid.
13785 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13787 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13788 * optabs-query.c (can_vec_perm_p): Likewise.
13789 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13790 instead of vec_perm_indices.
13791 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13792 (vect_gen_perm_mask_checked): Likewise,
13793 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13794 (vect_gen_perm_mask_checked): Likewise,
13796 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13798 * optabs-query.h (qimode_for_vec_perm): Declare.
13799 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13800 (qimode_for_vec_perm): ...this new function.
13801 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13803 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13805 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13806 does not have a conditional at the top.
13808 2018-01-02 Richard Biener <rguenther@suse.de>
13810 * ipa-inline.c (big_speedup_p): Fix expression.
13812 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13815 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13818 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13822 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13823 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13824 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13825 cond_taken_branch_cost 3->4.
13827 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13829 PR tree-optimization/83581
13830 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13831 TODO_cleanup_cfg if any changes have been made.
13833 PR middle-end/83608
13834 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13835 convert_modes if target mode has the right side, but different mode
13838 PR middle-end/83609
13839 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13840 last argument when extracting from CONCAT. If either from_real or
13841 from_imag is NULL, use expansion through memory. If result is not
13842 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13843 the parts directly to inner mode, if even that fails, use expansion
13846 PR middle-end/83623
13847 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13848 check for bswap in mode rather than HImode and use that in expand_unop
13851 Copyright (C) 2018 Free Software Foundation, Inc.
13853 Copying and distribution of this file, with or without modification,
13854 are permitted in any medium without royalty provided the copyright
13855 notice and this notice are preserved.