1 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4 * config/aarch64/aarch64-ldpstp.md: Replace uses of
5 aarch64_mem_pair_operand with memory_operand and delete operand swapping
7 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
8 Add check for legitimate_address.
9 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
10 (aarch64_swap_ldrstr_operands): New.
11 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
14 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17 * config/aarch64/aarch64.md: New patterns to generate stp
19 (store_pair_sw, store_pair_dw): New patterns to generate stp for
20 single words and double words.
21 (load_pair_sw, load_pair_dw): Likewise.
22 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
24 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
26 * config/aarch64/aarch64-ldpstp.md: Modify peephole
27 for different mode ldpstp and add peephole for merged zero stores.
29 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
31 (aarch64_gen_store_pair): Rename calls to match new patterns.
32 (aarch64_gen_load_pair): Rename calls to match new patterns.
33 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
34 (load_pair<DREG:mode><DREG2:mode>): ... This.
35 (store_pair<mode>): Rename to...
36 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
37 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
39 (V_INT_EQUIV): Handle SImode.
40 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
43 2018-05-22 Martin Sebor <msebor@redhat.com>
46 * calls.c (maybe_warn_nonstring_arg): Use string length to set
47 or ajust the presumed bound on an operation to avoid unnecessary
50 2018-05-22 Martin Sebor <msebor@redhat.com>
52 PR tree-optimization/85826
53 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
54 assuming that a DECL necesarily has a constant size.
56 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
59 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
61 2018-05-22 Richard Biener <rguenther@suse.de>
63 PR tree-optimization/85834
64 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
65 non-constant and non-zero memset arguments.
67 2018-05-22 Martin Liska <mliska@suse.cz>
70 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
72 2018-05-22 Richard Biener <rguenther@suse.de>
74 PR tree-optimization/85863
75 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
76 comparisons when vectype is specified.
77 (vectorizable_condition): Do not specify vectype for
78 vect_is_simple_cond when SLP vectorizing.
80 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
83 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
84 define __ibm128 as long double.
85 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
86 as a distinct type when IEEE 128-bit support is enabled.
87 (init_float128_ieee): Fix up conversions between IFmode and IEEE
88 128-bit types to use the correct functions.
89 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
90 convert between 128-bit floating point types that have different
91 modes but the same representation, instead of using gen_lowpart to
93 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
95 (IFKF_reg): New attributes to give the register constraints for
97 (extend<mode>tf2_internal): New insns to mark an explicit
98 conversion between 128-bit floating point types that have a
99 different mode but share the same representation.
101 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
103 PR tree-optimization/85814
104 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
105 a null return from get_strinfo when unsharing the next
106 strinfo in the chain.
108 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
111 * varasm.c (weak_finish): Clean up weak_decls.
113 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
115 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
116 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
117 UNSPEC_UADALP values.
118 * config/aarch64/iterators.md (ABAL): New int iterator.
121 (sur): Add mappings for the above.
122 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
124 (aarch64_<sur>abal<mode>_4): Likewise.
125 (aarch64_<sur>adalp<mode>_3): Likewise.
126 (<sur>sadv16qi): New define_expand.
128 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
130 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
131 (*movdf_internal): Ditto.
132 (*rcpsf2_sse): Ditto.
133 (*rsqrtsf2_sse): Ditto.
134 (*sqrt<mode>2_sse): Ditto.
136 2018-05-21 Tamar Christina <tamar.christina@arm.com>
138 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
140 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
141 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
142 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
143 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
145 * config/aarch64/arm_neon.h: Likewise.
146 * config/aarch64/iterators.md (VQ_I): New.
148 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
150 * config.gcc: Add arc/t-multilib-linux to tmake_file for
152 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
155 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
157 * config/nds32/constraints.md (S): New constraint.
158 * config/nds32/nds32.md (call_internal): Use constraint S.
159 (call_value_internal): Likewise.
160 (sibcall_internal): Likewise.
161 (sibcall_value_internal): Likewise.
163 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
164 Chung-Ju Wu <jasonwucj@gmail.com>
166 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
169 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
170 Chung-Ju Wu <jasonwucj@gmail.com>
172 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
173 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
174 (nds32_rtx_costs_impl): Simplify.
175 (nds32_address_cost_impl): Simplify.
176 (nds32_init_rtx_costs): New function.
177 (nds32_rtx_costs_speed_prefer): Likewise.
178 (nds32_rtx_costs_size_prefer): Likewise.
179 (nds32_address_cost_speed_prefer): Likewise.
180 (nds32_address_cost_speed_fwprop): Likewise.
181 (nds32_address_cost_size_prefer): Likewise.
182 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
183 * config/nds32/nds32.c (nds32_option_override): Use
184 nds32_init_rtx_costs function.
186 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
188 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
189 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
190 (TARGET_PIPELINE_N8): Likewise.
191 (TARGET_PIPELINE_N10): Likewise.
192 (TARGET_PIPELINE_N13): Likewise.
193 (TARGET_PIPELINE_GRAYWOLF): Likewise.
195 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
197 * config/nds32/nds32-fpu.md: Update copyright year.
199 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
201 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
203 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
205 * config/nds32/nds32.c
206 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
207 * config/nds32/nds32.opt (minline-asm-r15): New option.
209 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
211 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
213 * config/nds32/nds32.md (abssi2): New pattern.
215 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
217 * config/i386/i386.md (rex64namesuffix): New mode attribute.
218 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
219 Merge insn pattern from sse_cvtsi2ss<round_name> and
220 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
221 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
222 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
223 using SWI48 mode iterator.
224 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
225 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
226 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
227 pattern from sse_cvttss2si<round_saeonly_name>
228 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
229 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
230 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
231 using SWI48 mode iterator.
232 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
233 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
234 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
235 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
236 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
237 using SWI48 mode iterator.
238 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
239 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
240 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
241 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
242 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
244 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
245 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
246 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
247 pattern from sse_cvttsd2si<round_saeonly_name>
248 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
250 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
252 * config/nds32/nds32-md-auxiliary.c
253 (nds32_valid_smw_lwm_base_p): Refine.
254 (nds32_output_smw_single_word): Refine.
255 (nds32_output_smw_double_word): New.
256 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
258 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
260 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
261 (nds32_output_stack_pop): Refine.
262 (nds32_expand_unaligned_load): Refine.
263 (nds32_expand_unaligned_store): Refine.
265 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
266 Chung-Ju Wu <jasonwucj@gmail.com>
268 * config/nds32/constants.md: Add TP_REGNUM constant.
269 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
270 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
272 * config/nds32/nds32-doubleword.md: Consider flag_pic.
273 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
274 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
275 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
276 and PIC code generation.
277 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
279 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
281 * config/nds32/nds32.md: Support TLS and PIC.
282 * config/nds32/nds32.c: Support TLS and PIC.
283 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
284 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
287 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
289 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
292 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
293 Chung-Ju Wu <jasonwucj@gmail.com>
295 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
296 * config/nds32/nds32-md-auxiliary.c
297 (symbolic_reference_mentioned_p): New.
298 (nds32_legitimize_ict_address): New.
299 (nds32_expand_ict_move): New.
300 (nds32_indirect_call_referenced_p): New.
301 (nds32_symbol_binds_local_p): Delete.
302 (nds32_long_call_p): Modify.
303 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
304 * config/nds32/nds32-protos.h
305 (symbolic_reference_mentioned_p): Declare.
306 (nds32_legitimize_ict_address): Declare.
307 (nds32_expand_ict_move): Declare.
308 (nds32_indirect_call_referenced_p): Declare.
309 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
310 (nds32_relax_group): Use nds32_ict_const_p as condition.
311 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
312 (nds32_asm_file_start): Output ict_model directive in asm code.
313 (nds32_legitimate_address_p): Consider indirect call.
314 (nds32_print_operand): Consider indirect call.
315 (nds32_print_operand_address): Consider indirect call.
316 (nds32_insert_attributes): Handle "indirect_call" attribute.
317 (TARGET_LEGITIMATE_ADDRESS_P): Define.
318 (TARGET_LEGITIMATE_CONSTANT_P): Define.
319 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
320 (TARGET_DELEGITIMIZE_ADDRESS): Define.
321 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
322 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
323 (TARGET_ICT_MODEL_SMALL): Define.
324 (TARGET_ICT_MODEL_LARGE): Define.
325 * config/nds32/nds32.md (movsi): Consider ict model.
326 (call, call_value): Consider ict model.
327 (sibcall, sibcall_value): Consider ict model.
328 * config/nds32/nds32.opt (mict-model): New option.
329 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
332 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
333 Monk Chiang <sh.chiang04@gmail.com>
334 Jim Wilson <jimw@sifive.com>
336 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
337 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
338 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
339 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
340 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
341 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
342 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
343 compute save_libcall_adjustment properly.
344 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
345 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
346 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
347 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
348 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
349 (ABI_SPEC): Handle mabi=ilp32e.
350 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
352 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
353 <-march>: Add rv32e as an example.
355 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
358 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
359 (intra_create_variable_infos): Handle C++ constructors.
361 2018-05-18 Martin Liska <mliska@suse.cz>
363 * passes.def: Remove a redundant pass.
365 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
368 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
370 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
372 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
374 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
376 (armv2, armv2a, armv3, armv3m): Delete architectures.
377 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
378 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
379 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
381 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
382 (*mulsidi3adddi): Likewise.
383 (mulsidi3): Likewise.
384 (*mulsidi3_nov6): Likewise.
385 (umulsidi3): Likewise.
386 (umulsidi3_nov6): Likewise.
387 (umaddsidi4): Likewise.
388 (*umulsidi3adddi): Likewise.
389 (smulsi3_highpart): Likewise.
390 (*smulsi3_highpart_nov6): Likewise.
391 (umulsi3_highpart): Likewise.
392 (*umulsi3_highpart_nov6): Likewise.
393 * config/arm/arm.h (arm_arch3m): Delete.
394 * config/arm/arm.c (arm_arch3m): Delete.
395 (arm_option_override_internal): Update armv3-related comment.
396 (arm_configure_build_target): Delete use of isa_bit_mode32.
397 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
398 (arm_rtx_costs_internal): Delete check of arm_arch3m.
399 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
402 * config/arm/arm-protos.h (arm_arch3m): Delete.
403 * config/arm/arm-tables.opt: Regenerate.
404 * config/arm/arm-tune.md: Likewise.
405 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
406 deleted architectures.
408 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
410 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
411 (armv5t, armv5te): New features.
412 (ARMv5, ARMv5e): Delete fgroups.
413 (ARMv5t, ARMv5te): Adjust for above changes.
415 (armv5, armv5e): Delete arches.
416 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
418 (*call_reg_arm): Likewise.
419 (*call_value_reg_armv5): Likewise.
420 (*call_value_reg_arm): Likewise.
421 (*call_symbol): Likewise.
422 (*call_value_symbol): Likewise.
423 (*sibcall_insn): Likewise.
424 (*sibcall_value_insn): Likewise.
426 (prefetch): Likewise.
427 (define_split and define_peephole2 dependent on arm_arch5):
429 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
431 (TARGET_ARM_QBIT): Likewise.
432 (TARGET_DSP_MULTIPLY): Likewise.
433 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
434 (arm_arch5, arm_arch5e): Delete.
435 (arm_arch5t, arm_arch5te): Declare.
436 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
437 (arm_arch5t): Declare.
438 (arm_option_reconfigure_globals): Update for the above.
439 (arm_options_perform_arch_sanity_checks): Update comment, replace
440 use of arm_arch5 with arm_arch5t.
441 (use_return_insn): Likewise.
442 (arm_emit_call_insn): Likewise.
443 (output_return_instruction): Likewise.
444 (arm_final_prescan_insn): Likewise.
445 (arm_coproc_builtin_available): Likewise.
446 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
447 arm_arch5e with arm_arch5t and arm_arch5te.
448 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
449 (arm_arch5t, arm_arch5te): Declare.
450 * config/arm/arm-tables.opt: Regenerate.
451 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
452 * config/arm/t-multilib: Likewise.
453 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
454 instead of arm_arch5.
455 (*call_reg_thumb1): Likewise.
456 (*call_value_reg_thumb1_v5): Likewise.
457 (*call_value_reg_thumb1): Likewise.
458 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
460 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
462 2018-05-18 Martin Liska <mliska@suse.cz>
464 PR gcov-profile/84846
465 * doc/gcov.texi: Document -t option of gcov tool.
467 2018-05-18 Martin Liska <mliska@suse.cz>
469 PR gcov-profile/84846
470 * gcov.c (print_usage): Add new -t option.
471 (process_args): Handle the option.
472 (generate_results): Use stdout as output when requested by
475 2018-05-18 Martin Liska <mliska@suse.cz>
477 PR gcov-profile/84846
478 * coverage.c (coverage_init): Write PWD to .gcno file.
479 * doc/gcov.texi: Document how working directory is printed.
480 * gcov-dump.c (dump_gcov_file): Print PWD.
481 * gcov.c (output_intermediate_file): Likewise.
482 (read_graph_file): Read PWD string.
483 (output_lines): Print PWD.
485 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
488 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
489 for retval and return false if all args to phi are zero.
491 2018-05-18 Richard Biener <rguenther@suse.de>
493 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
495 (evrp_dom_walker::before_dom_children): Call it.
497 2018-05-18 Richard Biener <rguenther@suse.de>
499 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
500 results when processing array refs with variable index.
502 2018-05-18 Toon Moene <toon@moene.org>
504 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
505 directly after that of -floop-interchange. Indicate that both
506 options are enabled by default when specifying -O3.
508 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
510 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
511 iterator. Delete separate integer-mode vec_set<mode> expander.
512 (aarch64_simd_vec_setv2di): Delete.
513 (vec_setv2di): Delete.
514 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
515 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
516 the "w, r" alternative.
518 2018-05-18 Martin Liska <mliska@suse.cz>
520 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
521 * tree-pass.h (make_pass_lower_switch_O0): New function.
522 * tree-switch-conversion.c (node_has_low_bound): Remove.
523 (node_has_high_bound): Likewise.
524 (node_is_bounded): Likewise.
525 (class pass_lower_switch): Make it a template type and create
527 (pass_lower_switch::execute): Add template argument.
528 (make_pass_lower_switch): New function.
529 (make_pass_lower_switch_O0): New function.
530 (do_jump_if_equal): Remove.
531 (emit_case_nodes): Simplify to just handle all 3 cases and leave
532 all the hard work to tree optimization passes.
534 2018-05-18 Martin Liska <mliska@suse.cz>
536 * dbgcnt.c (limit_low): Renamed from limit.
537 (limit_high): New variable.
538 (dbg_cnt_is_enabled): Check for upper limit.
539 (dbg_cnt): Adjust dumping.
540 (dbg_cnt_set_limit_by_index): Add new argument for high
542 (dbg_cnt_set_limit_by_name): Likewise.
543 (dbg_cnt_process_single_pair): Parse new format.
544 (dbg_cnt_process_opt): Use strtok.
545 (dbg_cnt_list_all_counters): Remove 'value' and add
547 * doc/invoke.texi: Document changes.
549 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
551 * doc/sourcebuild.texi (scalar_all_fma): Document.
552 * tree.def (FMA_EXPR): Delete.
553 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
554 * internal-fn.c (ternary_direct): New macro.
555 (expand_ternary_optab_fn): Likewise.
556 (direct_ternary_optab_supported_p): Likewise.
557 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
558 * builtins.c (fold_builtin_fma): Delete.
559 (fold_builtin_3): Don't call it.
560 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
561 * expr.c (expand_expr_real_2): Likewise.
562 * fold-const.c (operand_equal_p): Likewise.
563 (fold_ternary_loc): Likewise.
564 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
565 * gimple.c (DEFTREECODE): Likewise.
566 * gimplify.c (gimplify_expr): Likewise.
567 * optabs-tree.c (optab_for_tree_code): Likewise.
568 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
569 * tree-eh.c (operation_could_trap_p): Likewise.
570 (stmt_could_throw_1_p): Likewise.
571 * tree-inline.c (estimate_operator_cost): Likewise.
572 * tree-pretty-print.c (dump_generic_node): Likewise.
573 (op_code_prio): Likewise.
574 * tree-ssa-loop-im.c (stmt_cost): Likewise.
575 * tree-ssa-operands.c (get_expr_operands): Likewise.
576 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
577 * fold-const-call.h (fold_fma): Delete.
578 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
579 CFN_FNMA and CFN_FNMS.
581 * genmatch.c (combined_fn): New enum.
582 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
583 (commutative_op): New function.
584 (commutate): Use it. Handle more than 2 operands.
585 (dt_operand::gen_gimple_expr): Use commutative_op.
586 (parser::parse_expr): Allow :c to be used with non-binary
587 operators if the commutative operand is known.
588 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
589 CFN_FMS, CFN_FNMA and CFN_FNMS.
590 (backprop::process_assign_use): Remove FMA_EXPR handling.
591 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
592 (gen_hsa_fma): New function.
593 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
594 IFN_FNMA and IFN_FNMS.
595 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
596 * gimple-fold.h (follow_all_ssa_edges): Declare.
597 * gimple-fold.c (follow_all_ssa_edges): New function.
598 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
599 gimple_build interface and use follow_all_ssa_edges to fold the result.
600 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
601 instead of checking for optabs directly.
602 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
603 rather than FMA_EXPRs.
604 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
605 call to IFN_FMA instead of an FMA_EXPR.
607 2018-05-17 Jim Wilson <jimw@sifive.com>
609 * expr.c (do_tablejump): When converting index to Pmode, if we have a
610 sign extended promoted subreg, and the range does not have the sign bit
611 set, then do a sign extend.
613 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
614 test, check for sign extended subreg and/or constant operands, and
615 do a sign extend in that case.
617 2018-05-17 Steve Ellcey <sellcey@cavium.com>
619 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
620 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
622 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
623 Change logics_shift_reg to logics_shift_imm.
624 (thunderx2t99_fp_loadpair_basic): Delete.
625 (thunderx2t99_fp_storepair_basic): Delete.
626 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
627 (thunderx2t99_asimd_polynomial): Delete.
628 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
629 and neon_fp_mul_d_scalar_q.
630 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
631 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
632 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
633 (thunderx2t99_asimd_lut): Add missing tbl types.
634 (thunderx2t99_asimd_ext): Delete.
635 (thunderx2t99_asimd_load1_1_mult): Delete.
636 (thunderx2t99_asimd_load1_2_mult): Delete.
637 (thunderx2t99_asimd_load1_ldp): New.
638 (thunderx2t99_asimd_load1): New.
639 (thunderx2t99_asimd_load2): Add missing *load2* types.
640 (thunderx2t99_asimd_load3): New.
641 (thunderx2t99_asimd_load4): New.
642 (thunderx2t99_asimd_store1_1_mult): Delete.
643 (thunderx2t99_asimd_store1_2_mult): Delete.
644 (thunderx2t99_asimd_store2_mult): Delete.
645 (thunderx2t99_asimd_store2_onelane): Delete.
646 (thunderx2t99_asimd_store_stp): New.
647 (thunderx2t99_asimd_store1): New.
648 (thunderx2t99_asimd_store2): New.
649 (thunderx2t99_asimd_store3): New.
650 (thunderx2t99_asimd_store4): New.
652 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
654 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
655 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
657 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
658 Segher Boessenkool <segher@kernel.crashing.org>
661 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
664 2018-05-17 Richard Biener <rguenther@suse.de>
666 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
667 for pruning loop and prune defs feeding only already visited PHIs.
669 2018-05-17 Richard Biener <rguenther@suse.de>
671 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
673 2018-05-17 Bin Cheng <bin.cheng@arm.com>
674 Richard Biener <rguenther@suse.de>
676 PR tree-optimization/85793
677 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
678 for VMAT_ELEMENTWISE.
680 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
682 * internal-fn.h (lookup_internal_fn): Declare
683 * internal-fn.c (lookup_internal_fn): New function.
684 * gimple.c (gimple_build_call_from_tree): Handle calls to
686 * gimple-pretty-print.c (dump_gimple_call): Print "." before
687 internal function names.
688 * tree-pretty-print.c (dump_generic_node): Likewise.
689 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
691 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
693 * gimple-fold.h (gimple_build): Make the function forms take
694 combined_fn rather than built_in_function.
695 (gimple_simplify): Likewise.
696 * gimple-match-head.c (gimple_simplify): Likewise.
697 * gimple-fold.c (gimple_build): Likewise.
698 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
699 rather than gimple_build_call_internal.
700 (get_initial_defs_for_reduction): Likewise.
701 (vect_create_epilog_for_reduction): Likewise.
702 (vectorizable_live_operation): Likewise.
704 2018-05-17 Martin Liska <mliska@suse.cz>
706 * gimple-ssa-sprintf.c (format_directive): Do not use
707 space in between 'G_' and '('.
709 2018-05-17 Jakub Jelinek <jakub@redhat.com>
712 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
713 even if the mask is not all ones.
716 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
718 (ix86_gimple_fold_builtin): Likewise.
721 * config/i386/i386.c: Include tree-vector-builder.h.
722 (ix86_vector_shift_count): New function.
723 (ix86_fold_builtin): Fold shift builtins by scalar count.
724 (ix86_gimple_fold_builtin): Likewise.
726 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
727 _mm512_setzero): New intrinsics.
729 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
730 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
732 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
733 code generation for cases where splatting a value is not useful.
734 * simplify-rtx.c (simplify_ternary_operation): Simplify
735 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
736 mode to a vec_concat.
738 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
740 * config.gcc: Support "goldmont-plus".
741 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
743 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
744 PROCESSOR_GOLDMONT_PLUS.
745 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
746 (processor_target_table): Add "goldmont-plus".
747 (PTA_GOLDMONT_PLUS): Define.
748 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
749 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
750 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
751 (fold_builtin_cpu): Add "goldmont-plus".
752 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
753 (ix86_option_override_internal): Add "goldmont-plus".
754 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
755 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
756 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
757 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
759 2018-05-17 Richard Biener <rguenther@suse.de>
761 PR tree-optimization/85757
762 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
763 remove defs that only feed that PHI from further processing.
765 2018-05-16 Jim Wilson <jimw@sifive.com>
767 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
769 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
771 2018-05-16 Mark Wielaard <mark@klomp.org>
773 * dwarf2out.c (count_index_strings): New function.
774 (output_indirect_strings): Call count_index_strings and generate
775 header for dwarf_version >= 5.
777 2018-05-16 Mark Wielaard <mark@klomp.org>
779 * dwarf2out.c (dwarf_FORM): New function.
780 (set_indirect_string): Use dwarf_FORM.
781 (reset_indirect_string): Likewise.
782 (size_of_die): Likewise.
783 (value_format): Likewise.
784 (output_die): Likewise.
785 (add_skeleton_AT_string): Likewise.
786 (output_macinfo_op): Likewise.
787 (index_string): Likewise.
788 (output_index_string_offset): Likewise.
789 (output_index_string): Likewise.
790 (count_index_strings): Likewise.
792 2018-05-16 Carl Love <cel@us.ibm.com>
794 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
795 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
797 2018-05-16 Martin Jambor <mjambor@suse.cz>
799 * ipa-prop.c (ipa_free_all_edge_args): Remove.
800 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
802 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
804 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
805 (fnma<mode>4): Likewise.
806 (fms<mode>4): Likewise.
807 (fnms<mode>4): Likewise.
808 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
809 (aarch64_fnma<mode>4): Likewise.
810 (aarch64_fms<mode>4): Likewise.
811 (aarch64_fnms<mode>4): Likewise.
812 (aarch64_fnmadd<mode>4): Likewise.
814 2018-05-16 Jason Merrill <jason@redhat.com>
816 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
818 2018-05-16 Richard Biener <rguenther@suse.de>
820 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
821 (dump_stmt_cost): Declare.
822 (add_stmt_cost): Dump cost we add.
823 (add_stmt_costs): New function.
824 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
826 (vect_analyze_stmt): Adjust prototype.
827 (vectorizable_condition): Likewise.
828 (vectorizable_live_operation): Likewise.
829 (vectorizable_reduction): Likewise.
830 (vectorizable_induction): Likewise.
831 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
832 cost vector to pass to vectorizable_ and record afterwards.
833 (vect_model_reduction_cost): Take cost vector argument and adjust.
834 (vect_model_induction_cost): Likewise.
835 (vectorizable_reduction): Likewise.
836 (vectorizable_induction): Likewise.
837 (vectorizable_live_operation): Likewise.
838 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
839 SLP_TREE_NUMBER_OF_VEC_STMTS.
840 (vect_analyze_slp_cost_1): Remove.
841 (vect_analyze_slp_cost): Likewise.
842 (vect_slp_analyze_node_operations): Take visited args and
843 a target cost vector. Avoid processing already visited stmt sets.
844 (vect_slp_analyze_operations): Use a local cost vector to gather
845 costs and register those of non-discarded instances.
846 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
847 (vect_schedule_slp_instance): Remove copying of
848 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
850 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
851 adding cost. Record cost entry location.
852 (vect_prologue_cost_for_slp_op): Function to compute cost of
853 a constant or invariant generated for SLP vect in the prologue,
854 split out from vect_analyze_slp_cost_1.
855 (vect_model_simple_cost): Make static. Adjust for SLP costing.
856 (vect_model_promotion_demotion_cost): Likewise.
857 (vect_model_store_cost): Likewise, make static.
858 (vect_model_load_cost): Likewise.
859 (vectorizable_bswap): Add cost vector arg and adjust.
860 (vectorizable_call): Likewise.
861 (vectorizable_simd_clone_call): Likewise.
862 (vectorizable_conversion): Likewise.
863 (vectorizable_assignment): Likewise.
864 (vectorizable_shift): Likewise.
865 (vectorizable_operation): Likewise.
866 (vectorizable_store): Likewise.
867 (vectorizable_load): Likewise.
868 (vectorizable_condition): Likewise.
869 (vectorizable_comparison): Likewise.
870 (can_vectorize_live_stmts): Likewise.
871 (vect_analyze_stmt): Likewise.
872 (vect_transform_stmt): Adjust calls to vectorizable_*.
873 * tree-vectorizer.c: Include gimple-pretty-print.h.
874 (dump_stmt_cost): New function.
876 2018-05-16 Richard Biener <rguenther@suse.de>
878 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
879 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
880 * tree-ssa-dse.c: Include tree-ssa-loop.h.
881 (check_name): New callback.
882 (dse_classify_store): Track cycles via a visited bitmap of PHI
883 defs and simplify handling of in-loop and across loop dead stores
884 and properly fail for loop-variant refs. Handle byte-tracking with
885 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
888 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
890 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
891 (vect_get_mask_type_for_stmt): Likewise.
892 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
894 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
895 to determine the statement's vector type and the vector type that
896 should be used for calculating nunits. Deal with cases in which
897 the type has to be deferred.
898 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
899 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
900 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
901 (vect_determine_vf_for_stmt): New functions, split out from...
902 (vect_determine_vectorization_factor): ...here.
903 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
904 (vect_get_mask_type_for_stmt): New functions, split out from
905 vect_determine_vectorization_factor.
907 2018-05-16 Richard Biener <rguenther@suse.de>
909 * tree-cfg.c (verify_gimple_assign_ternary): Properly
910 verify the [VEC_]COND_EXPR embedded comparison.
912 2018-05-15 Martin Sebor <msebor@redhat.com>
914 PR tree-optimization/85753
915 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
916 RECORD_TYPE in addition to ARRAY_TYPE.
918 2018-05-15 Martin Sebor <msebor@redhat.com>
921 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
923 2018-05-15 Richard Biener <rguenther@suse.de>
925 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
926 add by_clobber_p one. Change algorithm to collect all defs
927 representing uses we need to walk and try reducing them to
928 a single one before failing.
929 (dse_dom_walker::dse_optimize_stmt): Adjust.
931 2018-05-13 Mark Wielaard <mark@klomp.org>
933 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
934 (size_of_loc_descr): Likewise.
935 (output_loc_operands): Likewise.
936 (output_loc_operands_raw): Likewise.
937 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
938 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
939 (hash_loc_operands): Likewise.
940 (compare_loc_operands): Likewise.
942 2018-05-14 Mark Wielaard <mark@klomp.org>
944 * dwarf2out.c (count_index_addrs): New function.
945 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
947 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
949 PR tree-optimization/83648
950 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
951 return value as malloc candidate.
953 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
956 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
957 as true in call to suggest_attribute.
959 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
961 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
964 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
966 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
968 (aarch64_crypto_aesd_fused): Likewise.
970 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
972 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
973 (movsi_aarch64): Likewise.
974 (load_pairsi): Likewise.
975 (load_pairdi): Likewise.
976 (store_pairsi): Likewise.
977 (store_pairdi): Likewise.
978 (load_pairsf): Likewise.
979 (load_pairdf): Likewise.
980 (store_pairsf): Likewise.
981 (store_pairdf): Likewise.
982 (zero_extend): Likewise.
983 (trunc): Swap alternatives.
984 (fcvt_target): Add '?' to prefer w over r.
986 2018-05-14 Jakub Jelinek <jakub@redhat.com>
989 * config/i386/i386.md: Disallow non-commutative arithmetics in
990 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
991 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
992 in the peephole2 before it.
994 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
996 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
997 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
998 (ix86_handle_option): Handle -mcldemote.
999 * config.gcc: New header.
1000 * config/i386/cldemoteintrin.h: New file.
1001 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1002 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1004 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1005 OPTION_MASK_ISA_CLDEMOTE.
1006 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1007 (ix86_valid_target_attribute_inner_p): Ditto.
1008 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1009 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1010 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1011 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1012 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1014 * config/i386/i386.opt: Add -mcldemote.
1015 * config/i386/x86intrin.h: New header.
1016 * doc/invoke.texi: Add -mcldemote.
1018 2018-05-14 Richard Biener <rguenther@suse.de>
1020 * doc/match-and-simplify.texi: Adjust :s documentation.
1022 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1024 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1025 intended memcpy size.
1026 (REORDER_45): Likewise.
1028 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1030 * sort.cc: New file.
1031 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1032 * vec.c (qsort_chk): Use gcc_qsort.
1033 * Makefile.in (OBJS-libcommon): Add sort.o.
1034 (build/sort.o): New target. Use it...
1035 (BUILD_RTL): ... here, and...
1036 (build/gencfn-macros): ... here, and...
1037 (build/genmatch): ... here.
1039 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1040 Chung-Ju Wu <jasonwucj@gmail.com>
1042 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1043 * config/nds32/nds32-graywolf.md: New file.
1044 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1045 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1047 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1048 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1049 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1050 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1051 * config/nds32/pipelines.md: Include n15 settings.
1053 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1054 Chung-Ju Wu <jasonwucj@gmail.com>
1056 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1057 * config/nds32/nds32-n13.md: New file.
1058 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1059 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1061 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1062 * config/nds32/nds32.md (pipeline_model): Add n13.
1063 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1064 * config/nds32/pipelines.md: Include n13 settings.
1066 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1067 Chung-Ju Wu <jasonwucj@gmail.com>
1069 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1070 * config/nds32/nds32-n10.md: New file.
1071 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1072 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1074 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1075 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1076 * config/nds32/nds32.md (pipeline_model): Add n10.
1077 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1078 * config/nds32/pipelines.md: Include n10 settings.
1080 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1081 Kito Cheng <kito.cheng@gmail.com>
1082 Chung-Ju Wu <jasonwucj@gmail.com>
1084 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1085 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1086 Add enum values for DSP extension instructions.
1087 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1089 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1090 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1092 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1093 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1094 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1095 * config/nds32/nds32-intrinsic.md: Likewise.
1096 * config/nds32/nds32_intrinsic.h: Likewise.
1097 * config/nds32/nds32-md-auxiliary.c: Likewise.
1098 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1099 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1100 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1101 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1102 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1103 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1104 TYPE_DMAC in switch statement.
1105 * config/nds32/nds32.c: New checking and implementation for DSP
1106 extension instructions.
1107 * config/nds32/nds32.h: Likewise.
1108 * config/nds32/nds32.md: Likewise.
1109 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1110 * config/nds32/predicates.md: Implement new predicates for DSP
1113 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1115 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1116 Reformat alternatives and attributes so it is easier to identify
1117 which constraints/attributes go with which instruction.
1118 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1119 (mov<mode>_softfloat32, FMOVE64): Likewise.
1120 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1121 (mov<mode>_softfloat64, FMOVE64): Likewise.
1123 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1125 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1127 (Basic PowerPC Built-in Functions): The new name of the
1128 subsection previously known as "PowerPC Built-in Functions".
1129 (Basic PowerPC Built-in Functions Available on all Configurations):
1131 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1132 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1133 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1134 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1136 2018-05-11 Martin Jambor <mjambor@suse.cz>
1139 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1142 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1145 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1147 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1149 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1150 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1151 (ix86_handle_option): Handle -mwaitpkg.
1152 * config.gcc: New header.
1153 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1154 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1155 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1157 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1158 OPTION_MASK_ISA_WAITPKG.
1159 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1160 (ix86_option_override_internal): Add PTA_WAITPKG.
1161 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1162 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1163 IX86_BUILTIN_TPAUSE.
1164 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1165 __builtin_ia32_umwait and __builtin_ia32_tpause.
1166 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1167 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1168 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1169 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1170 UNSPECV_TPAUSE): New.
1171 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1172 * config/i386/i386.opt: Add -mwaitpkg.
1173 * config/i386/waitpkgintrin.h: New file.
1174 * config/i386/x86intrin.h: New header.
1175 * doc/invoke.texi: Add -mwaitpkg.
1177 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1180 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1182 (cortex-m0): Use armv6s-m isa.
1183 (cortex-m0plus): Likewise.
1184 (cortex-m1): Likewise.
1185 (cortex-m0.small-multiply): Likewise.
1186 (cortex-m0plus.small-multiply): Likewise.
1187 (cortex-m1.small-multiply): Likewise.
1189 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1190 Jakub Jelinek <jakub@redhat.com>
1192 PR tree-optimization/85692
1193 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1194 source permute as well.
1196 2018-05-11 Martin Liska <mliska@suse.cz>
1199 * doc/extend.texi: Document LLVM style format for no_sanitize
1202 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1204 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1205 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1206 (mode_supports_vsx_dform_quad): Likewise.
1207 (mode_supports_vmx_dform): Move these functions to be next to the
1208 other mode_supports functions.
1209 (mode_supports_dq_form): Likewise.
1210 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1211 mode_supports_dq_form.
1212 (reg_offset_addressing_ok_p): Likewise.
1213 (offsettable_ok_by_alignment): Likewise.
1214 (rs6000_legitimate_offset_address_p): Likewise.
1215 (legitimate_lo_sum_address_p): Likewise.
1216 (rs6000_legitimize_address): Likewise.
1217 (rs6000_legitimize_reload_address): Likewise.
1218 (rs6000_secondary_reload_inner): Likewise.
1219 (rs6000_preferred_reload_class): Likewise.
1220 (rs6000_output_move_128bit): Likewise.
1222 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1224 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1225 Generate SImode target register for null target.
1226 <case IX86_BUILTIN_XGETBV>: Ditto.
1227 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1228 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1230 2018-05-10 Carl Love <cel@us.ibm.com>
1232 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1233 dcbtt and dcbtstt if operands[2] is 0.
1235 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1238 * config/i386/sse.md (usadv64qi): New expander.
1240 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1242 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1243 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1244 -maltivec=be support.
1245 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1246 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1247 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1248 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1249 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1250 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1251 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1252 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1253 altivec_vsumsws): Adjust.
1254 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1255 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1256 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1258 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1259 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1260 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1261 (altivec_lve<VI_char>x): Delete expand.
1262 (*altivec_lve<VI_char>x_internal): Rename to...
1263 (altivec_lve<VI_char>x): ... this.
1264 (altivec_lvxl_<mode>): Delete expand.
1265 (*altivec_lvxl_<mode>_internal): Rename to ...
1266 (altivec_lvxl_<mode>): ... this.
1267 (altivec_stvxl_<mode>): Delete expand.
1268 (*altivec_stvxl_<mode>_internal): Rename to ...
1269 (altivec_stvxl_<mode>): ... this.
1270 (altivec_stve<VI_char>x): Delete expand.
1271 (*altivec_stve<VI_char>x_internal): Rename to ...
1272 (altivec_stve<VI_char>x): ... this.
1273 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1274 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1275 reduc_plus_scal_<mode>): Adjust.
1276 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1278 (rs6000_cpu_cpp_builtins): Adjust.
1279 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1280 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1281 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1282 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1283 -maltivec=be support.
1284 (rs6000_split_vec_extract_var): Adjust.
1285 (rs6000_split_v4si_init): Adjust.
1286 (swap_selector_for_mode): Delete.
1287 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1288 altivec_expand_stvex_be): Delete.
1289 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1290 -maltivec=be support.
1291 (rs6000_gimple_fold_builtin): Ditto.
1292 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1294 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1295 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1296 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1297 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1298 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1299 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1300 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1301 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1302 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1303 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1304 anonymous split): Adjust.
1305 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1306 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1308 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1310 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1311 when --with-gxx-include-dir is also specified.
1312 * configure: Regenerate.
1314 2018-05-09 Jim Wilson <jimw@sifive.com>
1317 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1318 * config/riscv/t-withmultilib: New.
1319 * config/riscv/withmultilib.h: New.
1320 * doc/install.texi: Document RISC-V --with-multilib-list support.
1322 2018-05-09 Richard Biener <rguenther@suse.de>
1324 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1326 (vect_bb_vectorization_profitable_p): Adjust. Compute
1327 actual scalar cost using the cost vector and the add_stmt_cost
1330 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1332 PR rtl-optimization/85645
1333 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1334 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1336 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1338 PR rtl-optimization/85645
1339 * shrink-wrap.c (spread_components): Return a boolean saying if
1340 anything was changed.
1341 (try_shrink_wrapping_separate): Iterate spread_components until
1342 nothing changes anymore.
1344 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1346 PR rtl-optimization/85645
1347 * regrename.c (build_def_use): Also kill the chains that include the
1348 destination of a REG_CFA_REGISTER note.
1350 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1352 PR rtl-optimization/85645
1353 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1354 insn that has a REG_CFA_REGISTER note.
1356 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1358 * cfgexpand.c (expand_clobber): New function.
1359 (expand_gimple_stmt_1): Use it.
1360 * tree-vect-stmts.c (vect_clobber_variable): New function,
1362 (vectorizable_simd_clone_call): ...here.
1363 (vectorizable_store): Emit a clobber either side of an
1364 IFN_STORE_LANES sequence.
1365 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1367 2018-05-09 Tom de Vries <tom@codesourcery.com>
1370 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1371 (define_insn "trap_if_false"): Add exit after trap.
1373 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1375 PR rtl-optimization/85638
1376 * bb-reorder.c: Include common/common-target.h.
1377 (create_forwarder_block): New function extracted from...
1378 (fix_up_crossing_landing_pad): ...here. Rename into...
1379 (dw2_fix_up_crossing_landing_pad): ...this.
1380 (sjlj_fix_up_crossing_landing_pad): New function.
1381 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1382 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1383 from both partitions and exit the loop after one iteration.
1385 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1388 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1390 (Basic PowerPC Built-in Functions): The new name of the
1391 subsection previously known as "PowerPC Built-in Functions".
1392 (Basic PowerPC Built-in Functions Available on all Configurations):
1394 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1396 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1398 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1400 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1403 2018-05-08 Jim Wilson <jimw@sifive.com>
1405 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1406 (LD_EMUL_SUFFIX): New.
1407 (LINK_SPEC): Use it.
1409 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1411 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1413 (Basic PowerPC Built-in Functions): The new name of the
1414 subsection previously known as "PowerPC Built-in Functions".
1415 (Basic PowerPC Built-in Functions Available on all Configurations):
1417 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1419 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1421 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1423 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1426 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1429 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1430 after cmpelim optimization.
1432 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1434 * config.gcc: Support "goldmont".
1435 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1436 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1438 * config/i386/i386.c (m_GOLDMONT): Define.
1439 (processor_target_table): Add "goldmont".
1440 (PTA_GOLDMONT): Define.
1441 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1442 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1443 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1444 (fold_builtin_cpu): Add "goldmont".
1445 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1446 (ix86_option_override_internal): Add "goldmont".
1447 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1448 (processor_type): Add PROCESSOR_GOLDMONT.
1449 * config/i386/i386.md: Add CPU "glm".
1450 * config/i386/glm.md: New file.
1451 * config/i386/x86-tune.def: Add m_GOLDMONT.
1452 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1454 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1457 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1459 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1460 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1461 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1464 * config/i386/i386.c (ix86_fold_builtin): Handle
1465 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1468 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1469 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1471 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1474 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1475 (check_arch): Likewise.
1476 (check_fpu): Return the result rather than printing it.
1477 (end arch): Fix operator precedence.
1478 (end cpu): Likewise.
1479 (END): Print the result from check_fpu.
1481 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1482 Alan Hayward <alan.hayward@arm.com>
1483 David Sherwood <david.sherwood@arm.com>
1485 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1486 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1487 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1488 (*fcmuo<mode>_and): New patterns.
1490 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1492 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1493 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1494 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1495 (cmp_op, sve_imm_con): New code attributes.
1496 (SVE_COND_INT_CMP, imm_con): Delete.
1497 (cmp_op): Remove above unspecs from int attribute.
1498 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1500 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1501 comparison-specific unspecs.
1502 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1503 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1504 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1505 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1506 (*vec_fcm<cmp_op><mode>): Rename to...
1507 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1508 (*vec_fcmuo<mode>): Rename to...
1509 (*fcmuo<mode>): ...this and adjust likewise.
1510 (*pred_fcm<cmp_op><mode>): New pattern.
1511 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1512 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1514 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1516 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1517 (aarch64_emit_sve_predicated_cond): New function.
1518 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1519 (aarch64_emit_unspec_cond_or): Replace with...
1520 (aarch64_emit_sve_or_conds): ...this new function. Use
1521 aarch64_emit_sve_ptrue_op for the individual comparisons and
1522 aarch64_emit_binop to OR them together.
1523 (aarch64_emit_inverted_unspec_cond): Replace with...
1524 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1525 aarch64_emit_sve_ptrue_op for the comparison and
1526 aarch64_emit_unop to invert the result.
1527 (aarch64_expand_sve_vec_cmp_float): Update after the above
1528 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1530 2018-05-07 Nathan Sidwell <nathan@acm.org>
1532 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1533 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1534 (Backwards Compatibility): Likewise.
1536 2018-05-07 Luis Machado <luis.machado@linaro.org>
1540 2018-05-07 Luis Machado <luis.machado@linaro.org>
1542 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1543 <prefetch_dynamic_strides>: New const bool field.
1544 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1545 prefetch_dynamic_strides.
1546 (exynosm1_prefetch_tune): Likewise.
1547 (thunderxt88_prefetch_tune): Likewise.
1548 (thunderx_prefetch_tune): Likewise.
1549 (thunderx2t99_prefetch_tune): Likewise.
1550 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1551 (aarch64_override_options_internal): Update to set
1552 PARAM_PREFETCH_DYNAMIC_STRIDES.
1553 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1554 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1555 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1556 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1557 prefetch-dynamic-strides setting.
1559 2018-05-07 Luis Machado <luis.machado@linaro.org>
1561 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1562 <minimum_stride>: New const int field.
1563 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1564 minimum_stride field.
1565 (exynosm1_prefetch_tune): Likewise.
1566 (thunderxt88_prefetch_tune): Likewise.
1567 (thunderx_prefetch_tune): Likewise.
1568 (thunderx2t99_prefetch_tune): Likewise.
1569 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1570 (aarch64_override_options_internal): Update to set
1571 PARAM_PREFETCH_MINIMUM_STRIDE.
1572 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1573 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1574 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1575 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1576 stride is constant and is below the minimum stride threshold.
1578 2018-05-07 Luis Machado <luis.machado@linaro.org>
1580 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1583 2018-05-07 Luis Machado <luis.machado@linaro.org>
1585 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1586 <prefetch_dynamic_strides>: New const bool field.
1587 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1588 prefetch_dynamic_strides.
1589 (exynosm1_prefetch_tune): Likewise.
1590 (thunderxt88_prefetch_tune): Likewise.
1591 (thunderx_prefetch_tune): Likewise.
1592 (thunderx2t99_prefetch_tune): Likewise.
1593 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1594 (aarch64_override_options_internal): Update to set
1595 PARAM_PREFETCH_DYNAMIC_STRIDES.
1596 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1597 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1598 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1599 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1600 prefetch-dynamic-strides setting.
1602 2018-05-07 Luis Machado <luis.machado@linaro.org>
1604 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1605 <minimum_stride>: New const int field.
1606 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1607 minimum_stride field.
1608 (exynosm1_prefetch_tune): Likewise.
1609 (thunderxt88_prefetch_tune): Likewise.
1610 (thunderx_prefetch_tune): Likewise.
1611 (thunderx2t99_prefetch_tune): Likewise.
1612 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1613 (aarch64_override_options_internal): Update to set
1614 PARAM_PREFETCH_MINIMUM_STRIDE.
1615 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1616 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1617 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1618 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1619 stride is constant and is below the minimum stride threshold.
1621 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1624 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1625 the type is addressable. Don't force op into register if it has
1628 2018-05-05 Roland McGrath <mcgrathr@google.com>
1631 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1632 any section for which we don't know a specific type it should have,
1633 regardless of name. Previously this was done only for the exact
1634 names ".init_array", ".fini_array", and ".preinit_array".
1635 (default_elf_asm_named_section): Add comment about
1636 relationship with default_section_type_flags and SECTION_NOTYPE.
1637 (get_section): Don't consider it a type conflict if one side has
1638 SECTION_NOTYPE and the other doesn't, as long as neither has the
1639 SECTION_BSS et al used in the default_section_type_flags logic.
1641 2018-05-05 Tom de Vries <tom@codesourcery.com>
1644 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1645 (workaround_barsyncs): New function.
1646 (nvptx_reorg): Use workaround_barsyncs.
1647 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1648 (define_expand "nvptx_membar_cta"): New define_expand.
1649 (define_insn "*nvptx_membar_cta"): New insn.
1651 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1653 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1654 To improve optimization opportunities.
1655 * builtin-types.def: The new needed builtin types for the above.
1657 2018-05-04 Richard Biener <rguenther@suse.de>
1659 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1660 * gimple-ssa-store-merging.c
1661 (imm_store_chain_info::output_merged_store): Remove redundant create,
1662 release split_store vector contents on failure.
1663 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1664 scalar stmt vector on cache hit.
1666 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1668 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1670 * config.gcc (powerpc-xilinx-eabi*): Remove.
1671 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1673 (fusion_addis_mem_combo_load): Ditto.
1674 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1676 (rs6000_cpu_cpp_builtins): Ditto.
1677 * config/rs6000/rs6000-linux.c
1678 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1679 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1680 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1682 (rs6000_setup_reg_addr_masks): Ditto.
1683 (rs6000_init_hard_regno_mode_ok): Ditto.
1684 (rs6000_option_override_internal): Ditto.
1685 (legitimate_lo_sum_address_p): Ditto.
1686 (rs6000_legitimize_address): Ditto.
1687 (rs6000_legitimize_reload_address): Ditto.
1688 (rs6000_legitimate_address_p): Ditto.
1689 (abi_v4_pass_in_fpr): Ditto.
1690 (setup_incoming_varargs): Ditto.
1691 (rs6000_gimplify_va_arg): Ditto.
1692 (rs6000_split_multireg_move): Ditto.
1693 (rs6000_savres_strategy): Ditto.
1694 (rs6000_emit_prologue_components): Ditto.
1695 (rs6000_emit_epilogue_components): Ditto.
1696 (rs6000_emit_prologue): Ditto.
1697 (rs6000_emit_epilogue): Ditto.
1698 (rs6000_elf_file_end): Ditto.
1699 (rs6000_function_value): Ditto.
1700 (rs6000_libcall_value): Ditto.
1701 * config/rs6000/rs6000.h: Ditto.
1702 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1703 (TARGET_MINMAX): ... this. New.
1704 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1705 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1706 (*movsi_internal1_single): Delete.
1707 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1708 mfpu=, mxilinx-fpu): Delete.
1709 * config/rs6000/singlefp.h: Delete.
1710 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1711 * config/rs6000/t-rs6000: Ditto.
1712 * config/rs6000/t-xilinx: Delete.
1713 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1714 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1715 (VStype_simple): Delete.
1716 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1717 * config/rs6000/xfpu.h: Delete.
1718 * config/rs6000/xfpu.md: Delete.
1719 * config/rs6000/xilinx.h: Delete.
1720 * config/rs6000/xilinx.opt: Delete.
1721 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1722 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1724 2018-05-04 Tom de Vries <tom@codesourcery.com>
1727 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1730 2018-05-04 Richard Biener <rguenther@suse.de>
1733 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1734 (expand_complex_div_wide): Likewise.
1735 (expand_complex_operations_1): Likewise.
1736 (expand_complex_libcall): Preserve EH info of the original stmt.
1737 (tree_lower_complex): Handle removed blocks.
1738 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1739 on complex multiplication and division libcall builtins.
1741 2018-05-04 Richard Biener <rguenther@suse.de>
1744 * fold-const.c (negate_expr_p): Restrict negation of operand
1745 zero of a division to when we know that can happen without
1747 (fold_negate_expr_1): Likewise.
1749 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1752 * real.h (real_nextafter): Declare.
1753 * real.c (real_nextafter): New function.
1754 * fold-const-call.c (fold_const_nextafter): New function.
1755 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1756 CASE_CFN_NEXTTOWARD.
1757 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1758 even when arg1_mode is different from arg0_mode.
1760 2018-05-03 Nathan Sidwell <nathan@acm.org>
1762 * doc/extend.texi (Deprecated Features): Remove
1764 (Backwards Compatibility): Likewise.
1765 * doc/invoke.texi (C++ Language Options): Likewise.
1766 (C++ Dialect Options): Likewise.
1768 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1771 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1772 _mm512_mask_mullox_epi64): New intrinsics.
1774 2018-05-03 Tom de Vries <tom@codesourcery.com>
1777 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1778 dump files): Add offload-tree.
1780 2018-05-03 Richard Biener <rguenther@suse.de>
1782 PR tree-optimization/85615
1783 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1784 to loops not nested in BBs loop father to avoid creating multi-entry
1787 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1789 PR tree-optimization/70291
1790 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1791 arguments. Change return type to tree. Emit libcall as a new
1792 statement rather than replacing existing one when inplace_p is true.
1793 (expand_complex_multiplication_components): New function.
1794 (expand_complex_multiplication): Expand floating-point complex
1795 multiplication using the above.
1796 (expand_complex_division): Rename inner_type parameter to type.
1797 Update expand_complex_libcall call-site.
1798 (expand_complex_operations_1): Update expand_complex_multiplication
1799 and expand_complex_division call-sites.
1801 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1804 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1805 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1806 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1807 the highest significant bit of the shift count mask is clear. In
1808 check whether and[sq]i3 is needed verify that all significant bits
1809 of the shift count other than the highest are set.
1811 2018-05-02 Tom de Vries <tom@codesourcery.com>
1814 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1815 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1816 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1817 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1818 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1819 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1820 __builtin_goacc_parlevel_size.
1822 2018-05-02 Richard Biener <rguenther@suse.de>
1824 PR tree-optimization/85597
1825 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1826 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1829 2018-05-02 Tom de Vries <tom@codesourcery.com>
1832 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1833 dump files): Add ltrans-tree.
1835 2018-05-02 Tom de Vries <tom@codesourcery.com>
1838 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1839 dump files): Add wpa-ipa.
1841 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1843 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1844 powerpc*-*-linux*paired* target.
1845 * config/rs6000/750cl.h: Delete.
1846 * config/rs6000/paired.h: Delete.
1847 * config/rs6000/paired.md: Delete.
1848 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1850 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1851 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1852 comment. Remove paired float support.
1853 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1854 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1856 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1857 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1859 * config/rs6000/rs6000.c: Remove paired float support.
1860 (paired_expand_vector_init, paired_expand_vector_move,
1861 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1862 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1863 paired_expand_builtin, paired_expand_predicate_builtin,
1864 paired_init_builtins): Delete.
1865 * config/rs6000/rs6000.h: Remove paired float support.
1866 * config/rs6000/rs6000.md: Remove paired float support.
1867 (move_from_CR_ov_bit): Delete.
1868 * config/rs6000/rs6000.opt (mpaired): Delete.
1869 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1870 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1872 2018-05-02 Richard Biener <rguenther@suse.de>
1875 * gimplify.c (gimplify_save_expr): When in SSA form allow
1876 SAVE_EXPRs to compute to SSA vars.
1878 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1881 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1882 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1883 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1884 clobber operands[2], instead use a new pseudo. Formatting fixes.
1886 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1888 PR tree-optimization/85586
1889 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1890 exit early for statements in the same group if the accesses are
1893 2018-05-02 Tom de Vries <tom@codesourcery.com>
1896 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1899 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1901 PR tree-optimization/85143
1902 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1904 2018-05-01 Tom de Vries <tom@codesourcery.com>
1907 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1908 not found" error message.
1910 2018-05-01 Tom de Vries <tom@codesourcery.com>
1913 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1914 * vec.c (test_ordered_remove_if): New function.
1915 (vec_c_tests): Call test_ordered_remove_if.
1916 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1917 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1918 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1919 VEC_ORDERED_REMOVE_IF.
1921 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1923 PR tree-optimization/82665
1924 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1925 pointer subtraction where arguments come from a memchr call.
1927 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1929 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1930 --push-state --as-needed and --pop-state instead of --as-needed and
1931 --no-as-needed if ld supports it.
1932 * configure: Regenerated.
1935 * doc/install.texi2html: Replace _002d with - and _002a with * in
1936 generated html files using sed.
1938 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1941 * gcc-rich-location.c (blank_line_before_p): New function.
1942 (use_new_line): New function.
1943 (gcc_rich_location::add_fixit_insert_formatted): New function.
1944 * gcc-rich-location.h
1945 (gcc_rich_location::add_fixit_insert_formatted): New function.
1947 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1949 * selftest.c (assert_streq): Rename "expected" and "actual" to
1950 "val1" and "val2". Extend NULL-handling to cover both inputs
1951 symmetrically, while still requiring both to be non-NULL for a pass.
1952 * selftest.h (assert_streq): Rename "expected" and "actual" to
1954 (ASSERT_EQ): Likewise.
1955 (ASSERT_EQ_AT): Likewise.
1956 (ASSERT_KNOWN_EQ): Likewise.
1957 (ASSERT_KNOWN_EQ_AT): Likewise.
1958 (ASSERT_NE): Likewise.
1959 (ASSERT_MAYBE_NE): Likewise.
1960 (ASSERT_MAYBE_NE_AT): Likewise.
1961 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1962 the assertion to pass.
1963 (ASSERT_STREQ_AT): Likewise.
1965 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1967 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1968 interaction with -pie.
1970 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1972 * selftest.h: Fix alphabetization of per-source-file selftest
1975 2018-04-30 Jason Merrill <jason@redhat.com>
1977 PR c++/61982 - dead stores to destroyed objects.
1978 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1981 2018-04-30 Jason Merrill <jason@redhat.com>
1983 * tree.c (build_clobber): New.
1984 * tree.h: Declare it.
1985 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1987 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1989 * diagnostic-show-locus.c (layout::layout): Update for
1990 location_get_source_line returning a char_span.
1991 (struct char_span): Move to input.h.
1992 (struct correction): Update for fields in char_span becoming
1994 (struct source_line): Update for location_get_source_line
1995 returning a char_span.
1996 (layout::print_line): Likewise.
1997 * edit-context.c (edited_file::print_content): Likewise.
1998 (edited_file::print_diff_hunk): Likewise.
1999 (edited_file::print_run_of_changed_lines): Likewise.
2000 (edited_file::get_num_lines): Likewise.
2001 (edited_line::edited_line): Likewise.
2002 * final.c (asm_show_source): Likewise.
2003 * input.c (location_get_source_line): Convert return type
2004 from const char * to char_span, losing the final "line_len"
2006 (dump_location_info): Update for the above.
2007 (get_substring_ranges_for_loc): Likewise. Use a char_span
2008 when handling the literal within the line.
2009 (test_reading_source_line): Update for location_get_source_line
2010 returning a char_span.
2011 * input.h (class char_span): Move here from
2012 diagnostic-show-locus.c, converting from a struct to a class.
2013 Make data members private.
2014 (char_span::operator bool): New.
2015 (char_span::length): New.
2016 (char_span::get_buffer): New.
2017 (char_span::operator[]): New.
2018 (char_span::subspan): Make const.
2019 (char_span::xstrdup): New.
2020 (location_get_source_line): Convert return type from const char *
2021 to char_span, losing the final "line_size" param.
2023 2018-04-30 Jan Hubicka <jh@suse.cz>
2025 * lto-wrapper.c (ltrans_priorities): New static var.
2026 (cmp_priority): New.
2027 (run_gcc): Read priorities and if doing parallel build order
2028 the Makefile by them.
2030 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2032 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2034 2018-04-30 Richard Biener <rguenther@suse.de>
2036 * tree-cfg.c (verify_address): Remove base argument, add
2037 flag whether to check TREE_ADDRESSABLE and do that.
2038 (verify_expr): Remove.
2039 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2040 (verify_gimple_assign_single): Likewise.
2041 (verify_gimple_switch): Likewise.
2042 (verify_expr_location_1): Dereference tp once. Add (disabled)
2043 piece from verify_expr.
2044 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2046 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2048 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2050 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2052 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2053 (small_data_pattern): Likewise.
2054 (arc_rewrite_small_data): Likewise.
2055 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2056 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2057 (get_symbol_alignment): New function.
2058 (legitimate_small_data_address_p): Likewise.
2059 (legitimate_scaled_address): Update, call
2060 legitimate_small_data_address_p.
2061 (output_sdata): New static variable.
2062 (arc_print_operand): Update how we handle small data operands.
2063 (arc_print_operand_address): Likewise.
2064 (arc_legitimate_address_p): Update, use
2065 legitimate_small_data_address_p.
2066 (arc_rewrite_small_data_p): Remove.
2067 (arc_rewrite_small_data_1): Likewise.
2068 (arc_rewrite_small_data): Likewise.
2069 (small_data_pattern): Likewise.
2070 (compact_sda_memory_operand): Update to use
2071 legitimate_small_data_address_p and get_symbol_alignment.
2072 (prepare_move_operands): Don't rewite sdata pattern.
2073 (prepare_extend_operands): Remove.
2074 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2076 (zero_extendqisi2): Likewise.
2077 (zero_extendhisi2): Likewise.
2078 (extendqihi2): Likewise.
2079 (extendqisi2): Likewise.
2080 (extendhisi2): Likewise.
2084 * config/arc/constraints.md (Usd): Change it to memory constraint.
2086 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2088 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2089 as source of std instructions.
2090 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2091 allow 6-bit constants as source for store instructions.
2092 (movdi_insn): Update instruction pattern to allow 6-bit constants
2093 as source for store instructions.
2095 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2097 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2099 2018-04-30 Nathan Sidwell <nathan@acm.org>
2100 Sandra Loosemore <sandra@codesourcery.com>
2102 * dumpfile.c (dump_open): Allow '-' for stdout.
2103 * doc/invoke.texi (Developer Options): Document dump filename
2104 determination early. Document stdin/stdout selection.
2106 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2108 Microblaze Target: PIC data text relative
2110 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2111 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2113 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2114 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2115 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2116 New addressing mode for data-text relative position indepenedent code.
2117 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2118 'ADDRESS_SYMBOLIC_TXT_REL'.
2119 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2120 (microblaze_legitimate_pic_operand): Exclude function calls from
2121 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2122 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2124 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2125 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2126 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2127 for 'address + offset'.
2128 (microblaze_expand_prologue): Add new function prologue call for
2130 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2131 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2132 table in case of TARGET_PIC_DATA_TEXT_REL.
2133 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2134 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2135 Add new macros 'UNSPEC_TEXT',
2136 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2137 + exclude function calls from 'UNSPEC_PLT' in case of data text
2139 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2140 new target hook for generating address diff vector tables in case of
2142 * doc/tm.texi : Regenerate.
2143 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2144 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2145 of addr diff vector generation.
2146 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2147 target hook definition.
2148 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2149 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2150 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2151 option for data text relative.
2153 2018-04-30 Richard Biener <rguenther@suse.de>
2155 * tree-chrec.h (evolution_function_is_constant_p): Remove
2157 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2159 2018-04-30 Richard Biener <rguenther@suse.de>
2162 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2164 2018-04-30 Richard Biener <rguenther@suse.de>
2166 PR tree-optimization/28364
2167 PR tree-optimization/85275
2168 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2169 copying first exit test.
2171 2018-04-28 Mark Wielaard <mark@klomp.org>
2173 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2175 (dwarf_AT): Handle DW_AT_addr_base.
2176 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2178 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2181 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2182 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2183 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2184 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2186 2018-04-28 Richard Biener <rguenther@suse.de>
2188 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2189 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2190 to reflect use. Only add interesting stmts.
2192 2018-04-27 Martin Jambor <mjambor@suse.cz>
2195 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2196 the jump function allows for passing through aggregate values.
2198 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2200 * input.h (in_system_header_at): Convert from macro to inline
2202 (from_macro_expansion_at): Likewise.
2203 (from_macro_definition_at): Likewise.
2205 2018-04-27 Jeff Law <law@redhat.com>
2207 * config.gcc: Mark tile* targets as deprecated/obsolete.
2209 2018-04-27 Richard Biener <rguenther@suse.de>
2211 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2214 2018-04-27 Richard Biener <rguenther@suse.de>
2216 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2218 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2220 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2221 with Yd constraint. Set "preferred_for_speed" attribute from
2222 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2224 (*movdi_internal): Ditto.
2225 (movti_interunit splitters): Remove
2226 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2227 (movdi_interunit splitters): Ditto.
2228 * config/i386/constraints.md (Ye): Remove.
2229 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2231 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2234 * config/aarch64/constraints.md (Usg): Limit to 31.
2237 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2239 PR tree-optimization/85529
2240 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2241 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2242 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2243 zero extension or masking of the MSB bit.
2244 (optimize_range_tests): Add FIRST_BB argument, pass it through
2245 to optimize_range_tests_var_bound.
2246 (maybe_optimize_range_tests, reassociate_bb): Adjust
2247 optimize_range_tests callers.
2249 2018-04-26 Richard Biener <rguenther@suse.de>
2250 Jakub Jelinek <jakub@redhat.com>
2252 * cgraph.h (symbol_table): Just declare debug method here.
2253 * symtab.c (symbol_table::debug): Define.
2255 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2257 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2259 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2261 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2262 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2263 (*movdi_internal): Substitute Yi and Yj constraint with x
2264 and Ym and Yn constraint with y constraint. Update "isa"
2265 attribute and set "preferred_for_speed" attribute from
2266 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2267 (*movsi_internal): Ditto.
2268 (*movdf_internal): Ditto.
2269 (*movsf_internal): Ditto.
2270 (*zero_extendsidi2): Ditto.
2271 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2272 (sse2_loadld): Ditto.
2273 (*vec_extract<ssevecmodelower>_0): Ditto.
2274 (*vec_extractv4si_0_zext_sse4): Ditto.
2275 (vec_concatv2di): Ditto.
2276 (*vec_dup<mode>): Ditto.
2277 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2278 * config/i386/constraints.md (Yi): Remove.
2283 2018-04-26 Nathan Sidwell <nathan@acm.org>
2285 * dumpfile.c (dump_open): New.
2286 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2287 (dump_finish): Detect stdio/stderr by value not name.
2289 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2291 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2293 2018-04-26 Tom de Vries <tom@codesourcery.com>
2296 * config/nvptx/nvptx.c (verify_neutering_jumps)
2297 (verify_neutering_labels): New function
2298 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2300 2018-04-26 Tom de Vries <tom@codesourcery.com>
2303 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2304 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2307 2018-04-26 Richard Biener <rguenther@suse.de>
2308 Tom de Vries <tom@codesourcery.com>
2311 * lto-streamer-out.c (output_function): Fixup loops if required to match
2312 discovery done in the reader.
2314 2018-04-26 Richard Biener <rguenther@suse.de>
2316 PR tree-optimization/85116
2317 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2318 have a loop exit from the single latch predecessor. Remove
2319 case of header with just condition.
2320 (ch_base::copy_headers): Exclude infinite loops from any
2322 (pass_ch::execute): Record exits.
2324 2018-04-26 Richard Biener <rguenther@suse.de>
2326 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2327 prologue cost vector and pass it to vect_get_load_cost.
2328 (vect_get_peeling_costs_all_drs): Likewise.
2329 (vect_peeling_hash_get_lowest_cost): Likewise.
2330 (vect_enhance_data_refs_alignment): Likewise.
2332 2018-04-26 Richard Biener <rguenther@suse.de>
2335 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2336 checking of integer<->pointer conversions.
2337 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2338 sign-/zero-extending pointer types.
2339 (expand_omp_for_static_chunk): Likewise.
2341 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2342 Jean Lee <xiaoyur347@gmail.com>
2344 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2345 (TARGET_ASAN_SHADOW_OFFSET): Define.
2346 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2347 true for -fsanitize=address.
2349 2018-04-25 Mark Wielaard <mark@klomp.org>
2351 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2354 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2356 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2357 than "alu", remove explicit "memory" and "imm_disp" attributes.
2358 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2361 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2362 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2363 gen_lowpart_no_emit.
2365 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2368 * config/i386/i386.c (ix86_expand_builtin): Change memory
2369 operand to XI, extend p0 to Pmode.
2370 * config/i386/i386.md: Change unspec volatile and operand
2371 1 mode to XI, change operand 0 mode to P.
2373 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2375 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2376 GET_MODE_MASK before any checking.
2377 (nds32_can_use_bset_p): Likewise.
2378 (nds32_can_use_btgl_p): Likewise.
2380 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2382 * config/nds32/nds32-doubleword.md: New define_split pattern for
2383 illegal register number.
2385 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2387 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2389 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2391 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2393 2018-04-25 Richard Biener <rguenther@suse.de>
2395 * lto-streamer.h (LTO_major_version): Bump to 8.
2397 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2399 * BASE-VER: Set to 9.0.0.
2401 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2403 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2404 in __abskf2 and __powikf2.
2406 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2409 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2410 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2411 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2412 Use the above on operand 2. Reindent.
2413 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2415 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2418 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2420 * config/i386/i386.opt (mcet): Removed.
2421 * doc/install.texi: Remove -mcet documentation.
2422 * doc/invoke.texi: Likewise.
2424 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2427 * doc/install.texi: Remove -mcet from bootstrap-cet.
2429 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2432 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2433 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2437 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2438 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2439 containing a CONST_VECTOR.
2441 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2443 * doc/install.texi: Update newlib dependency for nvptx.
2445 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2448 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2449 instead of INTVAL when shifting x left.
2451 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2453 PR tree-optimization/85478
2454 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2455 vect_grouped_store_supported for single element vectors.
2457 2018-04-24 Richard Biener <rguenther@suse.de>
2460 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2461 load cost increase to the case of non-constant step.
2463 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2466 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2467 destination if any_malformed_asm.
2469 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2472 * expr.c (store_field): In the bitfield case, if the value comes from
2473 a function call and is returned in registers by means of a PARALLEL,
2474 do not change the mode of the temporary unless BLKmode and VOIDmode.
2476 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2478 PR rtl-optimization/85423
2479 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2480 dependencies to debug insns when the previous insn is non-debug.
2482 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2484 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2485 enums into a single definition.
2486 (fls): Fix predicates and printing.
2489 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2491 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2492 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2493 and short u6 immediate.
2494 (check_if_valid_sleep_operand): Remove.
2495 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2497 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2499 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2500 flag_always_save_lp condition.
2501 * config/nds32/nds32.opt (malways-save-lp): New option.
2503 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2505 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2506 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2507 * config/nds32/nds32.h
2508 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2509 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2511 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2513 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2514 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2516 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2517 Chung-Ju Wu <jasonwucj@gmail.com>
2519 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2521 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2522 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2524 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2526 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2528 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2530 * config/nds32/nds32-protos.h (nds32_data_alignment,
2531 nds32_local_alignment): Declare.
2532 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2533 nds32_local_alignment): New functions.
2534 (TARGET_CONSTANT_ALIGNMENT): Define.
2535 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2537 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2539 * config/nds32/nds32.c
2540 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2541 (TARGET_MODES_TIEABLE_P): Likewise.
2543 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2545 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2548 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2549 Chung-Ju Wu <jasonwucj@gmail.com>
2551 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2552 for unaligned access.
2553 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2555 * config/nds32/nds32-intrinsic.md: Likewise.
2556 * config/nds32/nds32_intrinsic.h: Likewise.
2557 * config/nds32/nds32.h (nds32_builtins): Likewise.
2558 * config/nds32/nds32.opt (munaligned-access): New option.
2559 * config/nds32/nds32.c (nds32_asm_file_start): Display
2560 flag_unaligned_access status.
2562 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2564 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2565 -mno-relax is present.
2566 * config/riscv/linux.h (LINK_SPEC): Ditto.
2568 2018-04-20 Martin Sebor <msebor@redhat.com>
2571 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2573 (gimple_fold_builtin_stxcpy_chk): Same.
2574 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2576 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2579 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2580 __powikf2 when long double is IEEE 128-bit.
2582 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2584 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2585 step to make sure stack always aligned.
2587 2018-04-20 Carl Love <cel@us.ibm.com>
2590 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2591 size check for arg0.
2593 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2594 Tom de Vries <tom@codesourcery.com>
2597 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2598 Emit insns for calls too.
2599 (nvptx_find_par): Always look for worker-level predecessor insn.
2600 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2602 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2603 (nvptx_process_pars): Propagate frames for calls.
2605 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2608 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2610 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2611 (ix86_handle_option): Don't handle OPT_mibt.
2612 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2614 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2616 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2617 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2618 (ix86_target_macros): Define __CET__ with flag_cf_protection
2619 for -fcf-protection.
2620 * config/i386/i386.c (isa2_opts): Remove -mibt.
2621 * config/i386/i386.h (TARGET_IBT): Removed.
2622 (TARGET_IBT_P): Likewise.
2623 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2624 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2625 * config/i386/i386.opt (mcet): Update help message.
2628 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2629 -mcet as an alias for -mshstk.
2631 2018-04-20 Richard Biener <rguenther@suse.de>
2634 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2635 complexity by forcing a single use of the multiply operand.
2637 2018-04-20 Martin Jambor <mjambor@suse.cz>
2640 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2641 recursion dependency to only apply to non-clones.
2643 2018-04-20 Martin Jambor <mjambor@suse.cz>
2646 * ipa-cp.c (create_specialized_node): Check that clones of
2647 self-recursive edges exist during IPA-CP.
2649 2018-04-19 Toon Moene <toon@moene.org>
2651 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2654 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2656 PR tree-optimization/85467
2657 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2658 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2659 VECTOR_CST element to type.
2661 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2664 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2665 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2666 (builtin_longjmp): Likewise.
2667 (save_stack_nonlocal): New pattern.
2668 (restore_stack_nonlocal): Likewise.
2670 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2673 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2674 Replace ASM_OUTPUT_LABEL with fprintf.
2676 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2679 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2680 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2681 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2682 define __IBT__ and __SHSTK__ for -fcf-protection.
2683 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2685 (ix86_trampoline_init): Likewise.
2686 (x86_output_mi_thunk): Likewise.
2687 (ix86_notrack_prefixed_insn_p): Likewise.
2688 (ix86_option_override_internal): Don't disallow -fcf-protection.
2689 * config/i386/i386.md (rdssp<mode>): Also enable for
2691 (incssp<mode>): Likewise.
2692 (nop_endbr): Likewise.
2693 * config/i386/i386.opt (mcet): Change help message to built-in
2697 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2698 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2699 enable CET built-in functions.
2701 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2703 * common/config/i386/i386-common.c
2704 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2705 OPTION_MASK_ISA_MOVDIRI_UNSET,
2706 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2707 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2708 * config.gcc (movdirintrin.h): New header.
2709 * config/i386/cpuid.h (bit_MOVDIRI,
2710 bit_MOVDIR64B): New bits.
2711 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2713 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2714 (VOID, PVOID, PCVOID)): New function types.
2715 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2716 __builtin_ia32_directstoreu_u64,
2717 __builtin_ia32_movdir64b): New builtins.
2718 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2719 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2721 (ix86_valid_target_attribute_inner_p): Ditto.
2722 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2723 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2724 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2725 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2726 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2727 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2728 (movdiri<mode>, movdir64b_<mode>): New.
2729 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2730 * config/i386/immintrin.h: Include movdirintrin.h.
2731 * config/i386/movdirintrin.h: New file.
2732 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2734 2018-04-19 Richard Biener <rguenther@suse.de>
2737 * cfg.c (clear_bb_flags): When loop state says we have
2738 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2740 2018-04-19 Richard Biener <rguenther@suse.de>
2742 PR tree-optimization/84737
2743 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2744 copying restrict info.
2745 (vect_setup_realignment): Use it.
2746 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2747 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2748 the first DR to all generated stores.
2749 (vectorizable_load): Likewise for loads.
2751 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2753 PR tree-optimization/85446
2754 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2755 the integral and pointer types to have the same precision.
2757 * doc/install.texi: Document --disable-cet being the default and
2760 2018-04-18 Martin Liska <mliska@suse.cz>
2762 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2765 2018-04-18 Martin Liska <mliska@suse.cz>
2768 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2771 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2772 arguments if they are comparable.
2774 2018-04-18 Martin Liska <mliska@suse.cz>
2777 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2780 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2783 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2786 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2787 ENDBR after calling __morestack.
2789 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2792 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2793 by using gcc_base_ver to generate a gcc_driver_version, and use
2794 it when generating GCC_DRIVER_NAME.
2795 * configure: Regenerate.
2797 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2800 * config.gcc: Obsolete powerpc*-*-*spe*.
2802 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2805 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2806 (stabstr_D): Change type of unum from unsigned int to
2807 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2810 2018-04-17 Jim Wilson <jimw@sifive.com>
2813 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2814 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2815 Set arg_pointer_offset after using pretend_args_size.
2817 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2819 PR rtl-optimization/85431
2820 * dse.c (record_store): Ignore zero width stores.
2823 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2824 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2825 __builtin_stack_restore rather than after it.
2826 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2827 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2828 argument instead of virtual_dynamic_stack_rtx.
2830 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2832 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2834 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2835 Add note to error message to explain internal mapping of overloaded
2836 built-in function name to non-overloaded built-in function name.
2837 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2840 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2843 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2844 where the inputs overlap with the output.
2846 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2848 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2849 (=v, v) alternative and explicit "memory" attribute.
2850 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2851 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2853 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2854 "sselog1" type instead of "sselog".
2855 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2856 "sselog". Remove explicit "memory" attribute.
2857 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2858 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2860 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2861 "sselog1" type instead of "sselog". Remove explicit "memory"
2863 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2864 use "sselog1" type instead of "sselog". Remove explicit "memory"
2866 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2867 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2869 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2870 "sselog1" type instead of "sselog". Remove explicit "memory"
2872 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2873 use "sselog1" type instead of "sselog". Remove explicit "memory"
2877 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2880 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2883 2018-04-17 Martin Jambor <mjambor@suse.cz>
2886 * ipa-cp.c (create_specialized_node): Call
2887 expand_all_artificial_thunks if necessary.
2889 2018-04-17 Martin Liska <mliska@suse.cz>
2892 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2893 in message, remote space in between '_G' and '('.
2895 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2898 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2899 avx512f_vmcmp<mode>3<round_saeonly_name>,
2900 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2901 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2902 avx512f_rndscale<mode><round_saeonly_name>,
2903 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2904 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2905 Use %<iptr>2 instead of %2 for -masm=intel.
2906 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2907 avx512f_vcvttss2usi<round_saeonly_name>,
2908 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2910 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2911 avx512f_vcvttsd2usi<round_saeonly_name>,
2912 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2913 Use %q1 instead of %1 for -masm=intel.
2914 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2915 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2916 of %3 for -masm=intel.
2917 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2919 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2921 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2923 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2924 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2926 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2928 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2929 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2930 %g1 and one with %0 and %1.
2931 (avx512er_vmrcp28<mode><round_saeonly_name>,
2932 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2934 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2935 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2936 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2937 of %0 and %{%4%} for -masm=intel.
2938 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2939 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2940 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2941 order of %0 and %{%5%}%{z%} for -masm=intel.
2943 2018-04-17 Jan Hubicka <jh@suse.cz>
2946 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2948 2018-04-17 Martin Liska <mliska@suse.cz>
2951 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2952 for target_clone error message. Make default implementation
2953 clone to be a local declaration.
2954 (separate_attrs): Add new argument and check for an empty
2956 (expand_target_clones): Handle it.
2957 (ipa_target_clone): Make redirection just for target_clones
2960 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2961 Tom de Vries <tom@codesourcery.com>
2964 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2965 tiled basic blocks without omp continue statements.
2967 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2970 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2971 vec_extract expression as having side effects to make sure it gets
2974 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2977 * config/i386/i386.c (get_builtin_code_for_version): Check
2980 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2983 * gcc/config.gcc: Support "skylake".
2984 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2986 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2987 (processor_target_table): Add "skylake".
2988 (ix86_option_override_internal): Add "skylake".
2989 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2990 PROCESSOR_CANNONLAKE.
2991 (get_builtin_code_for_version): Fix priority for
2992 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2993 PROCESSOR_SKYLAKE-AVX512.
2994 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2995 (processor_type): Add PROCESSOR_SKYLAKE.
2997 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2998 Jason Merrill <jason@redhat.com>
3001 * convert.c (convert_to_integer_1): Use direct recursion for
3002 enumeral types and types with a precision less than the number
3003 of bits in their mode.
3005 2018-04-16 Julia Koval <julia.koval@intel.com>
3008 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3009 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3011 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3014 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3015 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3016 and -mno-direct-move.
3018 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3021 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3022 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3023 Ensure negative shifts result in {0}.
3025 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3027 PR rtl-optimization/79916
3028 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3029 regs (if any) to define how to gnerate SD moves when LRA is in
3032 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3034 PR rtl-optimization/85393
3035 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3036 * except.c (expand_dw2_landing_pad_for_region): Make static.
3037 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3038 a label and unconditional jump to old_bb, rather than
3039 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3042 PR rtl-optimization/85376
3043 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3044 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3045 instead of a specific value.
3047 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3048 Bin Cheng <bin.cheng@arm.com>
3050 PR tree-optimization/82965
3051 PR tree-optimization/83991
3052 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3053 by_profile_only parameter.
3054 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3055 information if the loop was predicted to iterate too many times.
3056 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3058 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3061 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3064 2018-04-13 Martin Liska <mliska@suse.cz>
3065 Jakub Jelinek <jakub@redhat.com>
3068 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3069 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3070 * builtins.c (expand_builtin_memory_copy_args): Use
3071 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3072 handle dest_addr == pc_rtx.
3074 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3077 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3078 asked to not generate direct moves.
3079 (fix_trunc<mode>si2_stfiwx): Similar.
3080 (fix_trunc<mode>si2_internal): Similar.
3082 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3085 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3086 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3087 lookup if dest in some wider mode is known to be const0_rtx and
3088 if so, record permanent equivalence for it to be ZERO_EXTEND of
3089 the narrower mode destination.
3091 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3093 * lto-streamer-out.c (output_function): Revert 259346.
3094 * omp-expand.c (expand_oacc_for): Likewise.
3096 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3098 PR rtl-optimization/85354
3099 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3100 * sel-sched.c (sel_global_init): ... here.
3102 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3105 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3106 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3107 mode for PE-COFF targets.
3108 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3109 (i386_pe_asm_lto_end): Likewise.
3110 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3111 (TARGET_ASM_LTO_END): Likewise.
3112 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3113 (i386_pe_asm_lto_start): New function.
3114 (i386_pe_asm_lto_end): Likewise.
3116 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3117 Richard Biener <rguenther@suse.de>
3120 * lto-streamer-out.c (output_function): Fix CFG loop state before
3122 * omp-expand.c (expand_oacc_for): Handle calls to internal
3123 functions like regular functions.
3125 2018-04-12 Richard Biener <rguenther@suse.de>
3128 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3129 for the early LTO debug to properly generate references to it
3130 during DIE emission. Do not re-use that for the skeleton for
3132 (dwarf2out_early_finish): Likewise.
3134 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3137 * config/i386/sse.md
3138 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3139 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3140 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3141 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3142 and output is a reg, avoid creating invalid lowpart subreg, but
3143 instead split into a 512-bit move. Don't split if not AVX512VL,
3144 input is xmm16+ reg and output is a mem.
3145 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3146 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3147 xmm16+ reg and output is a mem.
3149 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3151 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3152 also for flag_dwarf2_cfi_asm.
3154 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3156 PR rtl-optimization/85342
3157 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3158 a bool scalar var inside of the loop instead. Don't try to update
3159 recog_data.operand after failed apply_change_group.
3161 2018-04-12 Tom de Vries <tom@codesourcery.com>
3164 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3165 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3166 array with flexible array member as array without given dimension.
3167 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3168 argument for undefined param to true.
3170 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3173 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3174 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3175 from PowerPC section.
3176 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3177 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3178 help text that is too long.
3179 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3180 help text that is too long.
3181 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3182 help text that is too long.
3184 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3186 * config/alpha/alpha.md (stack_probe_internal): Rename
3187 from "probe_stack". Update all callers.
3189 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3191 PR rtl-optimization/84566
3192 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3193 sched_macro_fuse_insns.
3195 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3198 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3199 (compute_block_dependences): ... from here.
3201 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3203 PR tree-optimization/85331
3204 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3205 from int to HOST_WIDE_INT.
3207 2018-04-11 Martin Jambor <mjambor@suse.cz>
3210 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3211 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3212 not the same as the source val.
3213 (cgraph_edge_brings_value_p): New parameter.
3214 (gather_edges_for_value): Pass destination value to
3215 cgraph_edge_brings_value_p.
3216 (perhaps_add_new_callers): Likewise.
3217 (get_info_about_necessary_edges): Likewise and exclude values brought
3218 only by self-recursive edges.
3219 (create_specialized_node): Redirect only clones of self-calling edges.
3220 (+self_recursive_pass_through_p): New function.
3221 (find_more_scalar_values_for_callers_subset): Use it.
3222 (find_aggregate_values_for_callers_subset): Likewise.
3223 (known_aggs_to_agg_replacement_list): Removed.
3224 (decide_whether_version_node): Re-calculate known constants for all
3225 remaining context clones.
3227 2018-04-11 Richard Biener <rguenther@suse.de>
3230 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3231 from early DWARF output.
3232 (dwarf2out_early_finish): Output line info unconditionally into
3233 early DWARF and add reference to it.
3235 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3238 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3239 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3240 other than V2DFmode using iptr mode attribute.
3241 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3243 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3245 PR rtl-optimization/84659
3246 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3248 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3251 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3253 (output_loc_list): Pass address of a dummy size variable even in the
3254 locview handling loop.
3255 (index_location_lists): Add comment on why skip_loc_list_entry can't
3258 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3261 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3264 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3267 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3268 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3269 and -mstring-compare-inline-limit.
3271 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3274 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3275 for stack clash protection in a register whenever we need it to be in
3278 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3280 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3281 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3283 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3286 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3288 (mlong-double-): Ditto.
3289 * config/rs6000/sysv4.opt (msdata=): Ditto.
3290 (mtls-size=): Ditto.
3292 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3294 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3295 erroneous entries for
3296 "vector int vec_ldl (int, long int *)", and
3297 "vector unsigned int vec_ldl (int, unsigned long int *)".
3298 Add comments and entries for
3299 "vector bool char vec_ldl (int, bool char *)",
3300 "vector bool short vec_ldl (int, bool short *)",
3301 "vector bool int vec_ldl (int, bool int *)",
3302 "vector bool long long vec_ldl (int, bool long long *)",
3303 "vector pixel vec_ldl (int, pixel *)",
3304 "vector long long vec_ldl (int, long long *)",
3305 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3306 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3307 type tree bool_long_long_type_node and correct definition of
3308 bool_V2DI_type_node to make reference to this new type tree.
3309 (rs6000_mangle_type): Replace erroneous reference to
3310 bool_long_type_node with bool_long_long_type_node.
3311 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3312 comments to emphasize sign distinctions for char and int types and
3313 replace RS6000_BTI_bool_long constant with
3314 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3315 use of RS6000_BTI_pixel.
3316 (bool_long_type_node): Remove this macro definition.
3317 (bool_long_long_type_node): New macro definition
3319 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3321 PR rtl-optimization/85300
3322 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3323 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3324 simplify_unary_operation fails.
3326 2018-04-10 Martin Liska <mliska@suse.cz>
3328 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3329 cgraph_edge and ipa_ref.
3331 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3335 * config/i386/sse.md
3336 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3337 computation of the VEC_MERGE selector from mask.
3338 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3339 Fix decoding of the VEC_MERGE selector into mask.
3341 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3343 PR tree-optimization/85286
3344 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3346 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3348 * final.c (final_1): Set insn_last_address as well as
3349 insn_current_address.
3351 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3354 * explow.c (emit_stack_probe): Call validize_mem on memory location
3355 before passing it to gen_probe_stack. Create address operand and
3356 legitimize it for the probe_stack_address case.
3358 2018-04-09 Jan Hubicka <jh@suse.cz>
3361 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3362 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3363 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3365 (free_lang_data): Rebuild type inheritance graph.
3367 2018-04-09 Martin Sebor <msebor@redhat.com>
3369 * invoke.texi (-finline-small-functions): Mention other optimization
3371 (-findirect-inlining, -fpartial-inlining): Same.
3372 (-finline-functions-called-once): Same.
3373 (-freorder-blocks-and-partition): Same.
3375 2018-04-09 Jan Hubicka <jh@suse.cz>
3378 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3379 jumps; choose last target that matches the criteria (i.e.
3380 no partition changes for non-crossing jumps).
3381 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3382 support for redirecting crossing jumps to non-crossing.
3384 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3386 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3387 also for naked functions.
3389 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3391 * config/arc/arc.md (add_shift): New pattern.
3392 (add_shift2): Likewise.
3393 (sub_shift): Likewise.
3394 (sub_shift_cmp0_noout): Likewise.
3395 (compare_si_ashiftsi): Likewise.
3396 (xbfu_cmp0_noout): New combine pattern.
3397 (xbfu_cmp0"): Likewise.
3398 (movsi_set_cc_insn): Place the predicable variant first.
3399 (commutative_binary_cmp0_noout): Remove clobber.
3400 (commutative_binary_cmp0): New pattern.
3401 (noncommutative_binary_cmp0): Likewise.
3402 (noncommutative_binary_cmp0_noout): Likewise.
3403 (noncommutative_binary_comparison_result_used): Removed.
3404 (rsub_cmp0): New pattern.
3405 (rsub_cmp0_noout): Likewise.
3406 (extzvsi): Changed, keep only meaningful variants.
3407 (SQH, SEZ): New iterators.
3408 (SQH_postfix): New mode attribute.
3409 (SEZ_prefix): New code attribute.
3410 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3411 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3412 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3414 (noncommutative_operator): Check the availability of barrel
3417 2018-04-09 Richard Biener <rguenther@suse.de>
3419 PR tree-optimization/85284
3420 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3421 Only use the niter constraining form of simple_iv when the exit
3424 2018-04-09 Tom de Vries <tom@codesourcery.com>
3427 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3428 (define_expand "*memory_barrier"): New define_expand.
3429 (define_insn "memory_barrier"): New insn.
3431 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3433 PR rtl-optimization/80463
3434 PR rtl-optimization/83972
3435 PR rtl-optimization/83480
3437 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3438 correct producer for the insn.
3439 (tidy_control_flow): Fixup seqnos in case of debug insns.
3441 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3443 PR rtl-optimization/83913
3445 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3446 different sched-times when merging exprs.
3448 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3450 PR rtl-optimization/83962
3452 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3453 tidy_fallthru_edge and tidy_control_flow.
3455 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3457 PR rtl-optimization/83530
3459 * sel-sched.c (force_next_insn): New global variable.
3460 (remove_insn_for_debug): When force_next_insn is true, also leave only
3461 next insn in the ready list.
3462 (sel_sched_region): When the region wasn't scheduled, make another pass
3463 over it with force_next_insn set to 1.
3465 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3467 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3469 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3470 for interrupt control.
3471 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3472 functions for interrupt control.
3473 * config/nds32/nds32-intrinsic.md: Likewise.
3474 * config/nds32/nds32_intrinsic.h: Likewise.
3475 * config/nds32/nds32.h (nds32_builtins): Likewise.
3477 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3479 * config/nds32/nds32.c (nds32_init_machine_status,
3480 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3481 strict_aligned_p field.
3482 (nds32_expand_to_rtl_hook): New function.
3483 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3484 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3486 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3487 Chung-Ju Wu <jasonwucj@gmail.com>
3489 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3490 * config/nds32/nds32-n7.md: New file.
3491 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3492 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3494 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3495 * config/nds32/nds32.md (pipeline_model): Add n7.
3496 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3497 * config/nds32/pipelines.md: Include n7 settings.
3499 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3500 Chung-Ju Wu <jasonwucj@gmail.com>
3502 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3503 * config/nds32/nds32-e8.md: New file.
3504 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3505 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3507 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3508 * config/nds32/nds32.md (pipeline_model): Add e8.
3509 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3510 * config/nds32/pipelines.md: Include e8 settings.
3512 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3513 Chung-Ju Wu <jasonwucj@gmail.com>
3515 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3516 * config/nds32/nds32-n8.md: New file.
3517 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3518 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3520 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3521 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3522 * config/nds32/nds32.md (pipeline_model): Add n8.
3523 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3524 * config/nds32/pipelines.md: Include n8 settings.
3526 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3527 Chung-Ju Wu <jasonwucj@gmail.com>
3529 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3530 * config/nds32/nds32-n9-2r1w.md: New file.
3531 * config/nds32/nds32-n9-3r2w.md: New file.
3532 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3533 nds32_register_ports): New or modify for cpu n9.
3534 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3536 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3537 * config/nds32/nds32-utils.c: New file.
3538 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3539 TARGET_MUL_SLOW): Define.
3540 * config/nds32/nds32.md (pipeline_model): New attribute.
3541 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3542 New options that support cpu n9.
3543 * config/nds32/pipelines.md: Include n9 settings.
3544 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3546 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3548 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3549 information if necessary.
3550 (output_cond_branch_compare_zero): Likewise.
3551 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3552 (nds32_target_alignment): Refine for alignment.
3553 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3554 (FUNCTION_BOUNDARY): Modify.
3555 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3557 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3559 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3561 * config/nds32/constants.md (unspec_volatile_element): Add values for
3562 TLB operation and data prefetch.
3563 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3564 functions for TLB operation and data prefetch.
3565 * config/nds32/nds32-intrinsic.md: Likewise.
3566 * config/nds32/nds32_intrinsic.h: Likewise.
3567 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3568 (nds32_print_operand): Likewise.
3569 * config/nds32/nds32.h (nds32_builtins): Likewise.
3571 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3572 Andrew Pinski <pinsika@gcc.gnu.org>
3575 * match.pd: Use constant_boolean_node of correct type instead of
3576 boolean_true_node or boolean_false_node for simplifying
3577 pointer comparisons to zero.
3579 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3581 PR tree-optimization/80021
3582 * tree.c (verify_type_variant): Make error call in verify_variant_match
3583 translatable and remove final full stop.
3585 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3587 * config/nds32/constants.md (unspec_volatile_element): Add
3588 UNSPEC_VOLATILE_EH_RETURN.
3589 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3590 nds32_output_stack_pop): Support dwarf exception handling process.
3591 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3592 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3593 exception handling process.
3594 (nds32_compute_stack_frame): Likewise.
3595 (nds32_return_addr_rtx): Likewise.
3596 (nds32_initial_elimination_offset): Likewise.
3597 (nds32_expand_prologue): Likewise.
3598 (nds32_expand_epilogue): Likewise.
3599 (nds32_dynamic_chain_address): New function.
3600 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3602 (DYNAMIC_CHAIN_ADDRESS): Define.
3603 (EH_RETURN_DATA_REGNO): Define.
3604 (EH_RETURN_STACKADJ_RTX): Define.
3605 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3606 patterns for dwarf exception handling.
3608 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3610 * config/nds32/nds32.h: Clean up obsolete macros.
3612 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3614 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3615 Add enum values for particular instructions.
3616 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3617 particular intrinsic functions.
3618 * config/nds32/nds32-intrinsic.md: Likewise.
3619 * config/nds32/nds32_intrinsic.h: Likewise.
3620 * config/nds32/nds32.h (nds32_builtins): Likewise.
3621 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3622 (btst, ave): New patterns for particular instructions.
3624 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3626 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3627 Add enum values for atomic load/store and memory sync.
3628 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3630 * config/nds32/nds32-intrinsic.md: Likewise.
3631 * config/nds32/nds32_intrinsic.h: Likewise.
3632 * config/nds32/nds32.h (nds32_builtins): Likewise.
3634 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3636 PR tree-optimization/85257
3637 * fold-const.c (native_encode_vector): If not all elts could fit
3638 and off is -1, return 0 rather than offset.
3639 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3640 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3641 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3642 adjust buffer in native_interpret_expr call.
3644 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3646 * config/nds32/constants.md (unspec_volatile_element): Add cache
3647 control enum values.
3648 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3649 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3650 * config/nds32/nds32.c (nds32_cctl_names): New.
3651 (nds32_print_operand): Handle cache control register names.
3652 * config/nds32/nds32.h (nds32_builtins): New enum values.
3653 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3655 * config/nds32/nds32.md (type): Add mmu.
3656 * config/nds32/pipelines.md (simple_insn): Add mmu.
3658 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3660 * config/nds32/nds32.md (type): Remove call.
3661 * config/nds32/pipelines.md (simple_insn): Likewise.
3663 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3665 * config/nds32/constants.md (unspec_volatile_element): Add
3666 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3667 UNSPEC_VOLATILE_FMFCFG.
3668 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3669 description for fmfcfg and fmfcsr.
3670 (bdesc_1arg): Add fmtcsr.
3671 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3672 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3673 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3674 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3675 unspec_fmfcfg): New patterns.
3676 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3677 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3678 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3679 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3680 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3681 __nds32__fmfcfg): Define.
3683 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3685 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3686 intrinsic register names.
3687 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3688 intrinsic register enum values and macros.
3690 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3692 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3693 for load/store addressing form.
3694 (nds32_print_operand_address): Likewise.
3696 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3699 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3700 based on LABEL_REF. Remove useless assertion.
3701 (pic_address_needs_scratch): Fix formatting.
3702 (sparc_legitimize_pic_address): Minor tweaks.
3703 (sparc_delegitimize_address): Adjust assertion accordingly.
3704 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3705 into symbolic_operand.
3706 (movsi_high_pic_label_ref): Likewise.
3707 (movsi_lo_sum_pic_label_ref): Likewise.
3708 (movdi_pic_label_ref): Likewise.
3709 (movdi_high_pic_label_ref): Likewise.
3710 (movdi_lo_sum_pic_label_ref): Likewise.
3712 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3714 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3715 custom LIB_SPEC setup.
3717 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3718 Kito Cheng <kito.cheng@gmail.com>
3720 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3721 * config/riscv/freebsd.h: New.
3723 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3725 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3726 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3729 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3730 Kito Cheng <kito.cheng@gmail.com>
3732 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3733 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3734 * config/nds32/nds32-protos.h (nds32_output_call,
3735 nds32_output_return): Declare.
3736 * config/nds32/nds32.md: Refine all the call and return patterns.
3738 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3741 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3742 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3744 PR rtl-optimization/84872
3745 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3746 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3749 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3751 * expr.c (copy_blkmode_to_reg): Revert 254862.
3752 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3754 2018-04-06 Richard Biener <rguenther@suse.de>
3757 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3758 after seeing a component reference with an adjacent field. Treat
3759 refs to arrays at struct end of external decls similar to
3760 refs to unconstrained commons.
3762 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3765 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3766 look through SAVE_EXPRs with non-side-effects argument. Adjust
3768 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3771 2018-04-06 Richard Biener <rguenther@suse.de>
3774 * alias.c (find_base_term): New wrapper around find_base_term
3775 unwinding CSELIB_VAL_PTR changes.
3776 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3779 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3781 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3783 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3784 constant definitions.
3785 ("nop"): lr 0,0 -> nopr r0
3786 ("nop_lr0", "nop_lr1"): New insn definitions.
3788 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3790 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3791 NDS32_V3PUSH_AVAILABLE_P macro.
3793 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3794 Chung-Ju Wu <jasonwucj@gmail.com>
3796 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3797 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3798 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3799 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3800 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3801 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3802 * config/nds32/constraints.md: New constraints and checking for hard
3803 float configuration.
3804 * config/nds32/iterators.md: New mode iterator and attribute for hard
3805 float configuration.
3806 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3808 * config/nds32/nds32-fpu.md: New file.
3809 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3810 deal with hard float code generation.
3811 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3813 (abi_type, float_reg_number): New enum type.
3814 * config/nds32/nds32-predicates.c: New predicates for hard float.
3815 * config/nds32/nds32-protos.h: Declare functions for hard float.
3816 * config/nds32/nds32.c: Implementation for hard float configuration.
3817 * config/nds32/nds32.h: Definitions for hard float configuration.
3818 * config/nds32/nds32.md: Include hard float machine description and
3819 modify patterns for hard float configuration.
3820 * config/nds32/nds32.opt: New options for hard float configuration.
3821 * config/nds32/predicates.md: New predicates for hard float
3824 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3826 * common/config/nds32/nds32-common.c
3827 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3829 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3832 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3833 CONSTRUCTOR_ELT (ctor, ...)->value.
3835 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3838 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3840 2018-04-05 Tom de Vries <tom@codesourcery.com>
3843 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3846 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3847 Kito Cheng <kito.cheng@gmail.com>
3849 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3850 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3851 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3852 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3854 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3855 Kito Cheng <kito.cheng@gmail.com>
3857 * config/nds32/constraints.md (Ufe): New memory constraint.
3858 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3859 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3860 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3862 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3863 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3865 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3867 * config/nds32/nds32.md: Use optimize_size in the condition for
3868 alu-shift instructions.
3870 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3872 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3874 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3876 * config/nds32/nds32.md (negsi2): Refine pattern.
3878 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3879 Chung-Ju Wu <jasonwucj@gmail.com>
3881 * config/nds32/iterators.md (shift_rotate): New code iterator.
3882 (shift): New code attribute.
3883 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3884 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3885 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3886 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3887 bit-wise operations.
3888 (andsi3, *andsi3): Ditto.
3889 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3890 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3891 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3892 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3893 nds32_ior_operand, nds32_xor_operand): New predicates.
3895 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3897 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3898 (addsi3, subsi3): ... this.
3900 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3902 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3904 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3906 * config/nds32/nds32.md: Adjust indention.
3908 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3910 * config/nds32/nds32.md (feature): New attribute.
3912 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3914 * config/nds32/nds32.md (subtype): New attribute.
3916 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3919 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3920 expansion to perform a bitwise AND of the argument followed by a
3921 boolean negation of the result.
3923 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3925 PR rtl-optimization/84878
3926 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3927 the basic block. Assert the use reference is not artificial and that
3928 it has an associated insn.
3930 2018-04-04 Michael Matz <matz@suse.de>
3932 * builtins.c (compute_objsize): Pass correct operand
3933 to array_at_struct_end_p.
3935 2018-04-04 Richard Biener <rguenther@suse.de>
3938 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3939 from contexts for DINFO_LEVEL_TERSE and below.
3941 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3943 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3944 resiter_operand condition.
3945 * config/nds32/nds32.md (*move<mode>): Ditto.
3947 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3948 Monk Chiang <sh.chiang04@gmail.com>
3950 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3952 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3954 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3956 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3957 Kito Cheng <kito.cheng@gmail.com>
3959 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3960 nds32_cond_code_str, output_cond_branch,
3961 output_cond_branch_compare_zero, nds32_expand_cbranch,
3962 nds32_expand_cstore, nds32_expand_movcc,
3963 nds32_output_cbranchsi4_equality_zero,
3964 nds32_output_cbranchsi4_equality_reg,
3965 nds32_output_cbranchsi4_equality_reg_or_const_int,
3966 nds32_output_cbranchsi4_greater_less_zero: New functions.
3967 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3968 nds32_expand_cstore, nds32_expand_movcc,
3969 nds32_output_cbranchsi4_equality_zero,
3970 nds32_output_cbranchsi4_equality_reg,
3971 nds32_output_cbranchsi4_equality_reg_or_const_int,
3972 nds32_output_cbranchsi4_greater_less_zero): Declare.
3973 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3974 nds32_rimm11s_operand): New predicates.
3975 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3976 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3979 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3981 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3982 * config/nds32/nds32.md: Ditto.
3983 * config/nds32/pipelines.md: Ditto.
3985 2018-04-04 Richard Biener <rguenther@suse.de>
3987 PR tree-optimization/85168
3988 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3989 propagating abnormals.
3991 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3993 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3995 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3996 Kito Cheng <kito.cheng@gmail.com>
3998 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3999 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4000 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4001 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4002 * config/nds32/nds32.md (sibcall_internal): New.
4003 (sibcall_register): Remove.
4004 (sibcall_immediate): Remove.
4005 (sibcall_value_internal): New.
4006 (sibcall_value_register): Remove.
4007 (sibcall_value_immediate): Remove.
4008 * config/nds32/predicates.md (nds32_general_register_operand): New.
4009 (nds32_call_address_operand): New.
4011 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4013 PR rtl-optimization/85167
4014 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4015 bb_defs if *split_p, instead preinitialize it to NULL.
4017 PR tree-optimization/85156
4018 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4019 evaluating the argument multiple times.
4021 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4023 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4025 (_mm_cvtpd_ps): Likewise.
4026 (_mm_cvttpd_epi32): Likewise.
4027 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4028 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4029 vector, pixel, and bool following altivec.h include.
4031 2018-04-03 Martin Sebor <msebor@redhat.com>
4033 * doc/extend.texi (Common Function Attributes): Clarify.
4034 (const attribute): Likewise.
4035 (pure attribute): Likewise.
4037 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4040 * config/i386/i386.c (ix86_expand_vector_set): Use
4041 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4043 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4045 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4046 instructions when changing rounding bits to preserve precision bits
4047 in the x87 control word.
4049 2018-04-03 Martin Liska <mliska@suse.cz>
4051 PR tree-optimization/82491
4052 * rtl.h (strip_offset_and_add): Replace += suboffset with
4053 poly_uint64 () + suboffset.
4055 2018-03-29 Martin Liska <mliska@suse.cz>
4056 Martin Jambor <mjambor@suse.cz>
4059 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4060 param_type is not an integral or pointer type.
4062 2018-04-03 Richard Biener <rguenther@suse.de>
4064 * sese.h (recompute_all_dominators): Remove.
4066 2018-04-02 Martin Sebor <msebor@redhat.com>
4068 * doc/invoke.texi (-Wrestrict): Fix typos.
4070 2018-04-02 Jim Wilson <jimw@sifive.com>
4072 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4073 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4074 (<optab>di3, <optab>si3_extend): Likewise.
4075 (<optab>si3_mask, <optab>si3_mask_1): New.
4076 (<optab>di3_mask, <optab>di3_mask_1): New.
4077 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4078 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4079 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4081 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4083 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4086 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4088 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4089 (nds32_canonicalize_comparison): New function.
4091 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4092 Kito Cheng <kito.cheng@gmail.com>
4093 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4095 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4096 * config/nds32/constants.md (unspec_volatile_element): Add
4097 UNSPEC_VOLATILE_RELAX_GROUP.
4098 * config/nds32/nds32-relax-opt.c: New file.
4099 * config/nds32/nds32-predicates.c
4100 (nds32_symbol_load_store_p): New function.
4101 * config/nds32/nds32-protos.h
4102 (nds32_symbol_load_store_p): Declare function.
4103 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4104 * config/nds32/nds32.c
4105 (nds32_register_pass): New function to register pass.
4106 (nds32_register_passes): New function to register passes.
4107 * config/nds32/nds32.md (relax_group): New pattern.
4108 * config/nds32/nds32.opt (mrelax-hint): New option.
4109 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4111 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4113 * config/nds32/t-nds32: Modify files dependency.
4115 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4117 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4118 (PROFILE_HOOK): Define its implementation.
4120 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4122 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4123 type and 32-bit size.
4125 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4128 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4129 (V_128_256): New mode iterator.
4130 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4131 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4132 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4134 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4135 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4137 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4140 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4141 NaN inputs correctly.
4143 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4146 * config/rs6000/vsx.md (??r): New mode attribute.
4147 (*vsx_mov<mode>_64bit): Use it.
4148 (*vsx_mov<mode>_32bit): Likewise.
4150 2018-03-30 Martin Sebor <msebor@redhat.com>
4152 PR tree-optimization/84818
4153 * builtins.c (check_access): Use warning_n.
4155 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4158 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4160 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4163 2018-03-30 Julia Koval <julia.koval@intel.com>
4166 * x86-tune.def (movx, partial_reg_dependency): Enable for
4169 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4172 * lra-constraints.c (process_alt_operands): Move setting
4173 this_alternative_matches below.
4175 2018-03-29 Martin Liska <mliska@suse.cz>
4178 * doc/invoke.texi: Document how LTO works with debug info.
4179 Describe auto-load support of binutils. Mention 'x86-64'
4180 as valid option value of -march option.
4182 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4184 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4187 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4188 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4189 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4192 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4195 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4196 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4197 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4198 for RS6000_BTM_POWERPC64.
4199 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4200 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4201 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4206 2018-03-28 Carl Love <cel@us.ibm.com>
4209 2017-09-27 Carl Love <cel@us.ibm.com>
4211 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4212 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4213 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4216 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4218 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4219 instead of __vector bool.
4220 (_mm_max_pu8): Likewise.
4221 (_mm_min_pi16): Likewise.
4223 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4226 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4227 (DIVWEUO): Likewise.
4229 (DIVDEUO): Likewise.
4230 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4231 DIVWEUO and DIVDEUO.
4232 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4233 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4234 (div_extend): Likewise.
4235 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4237 (__builtin_divweuo): Likewise.
4238 (__builtin_divdeo): Likewise.
4239 (__builtin_divdeuo): Likewise.
4241 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4244 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4245 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4247 PR tree-optimization/82004
4248 * gimple-match-head.c (optimize_pow_to_exp): New function.
4249 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4250 Don't fold to exp if optimize_pow_to_exp is false.
4252 2018-03-28 Martin Liska <mliska@suse.cz>
4255 * calls.c (initialize_argument_information): Fix trailing space.
4256 * common.opt: Fix typo and provide better explanation for
4257 -fsanitize-coverage option.
4258 * config/i386/i386.opt: Fix typo.
4260 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4261 Martin Liska <mliska@suse.cz>
4264 * gimplify.c (asan_poison_variable): Don't do the check for
4265 gimplify_omp_ctxp here.
4266 (gimplify_decl_expr): Do it here.
4267 (gimplify_target_expr): Likewise.
4269 2018-03-28 Martin Liska <mliska@suse.cz>
4272 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4273 chkp_type_bounds_count if MPX is not enabled.
4275 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4277 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4279 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4282 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4283 function to create the function decl for complex long double
4284 multiply and divide for -mabi=ieeelongdouble.
4285 (init_float128_ieee): Call it.
4287 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4290 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4291 -fcf-protection=branch -mibt.
4292 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4294 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4297 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4299 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4302 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4303 extern array declarations.
4305 2018-03-27 Richard Biener <rguenther@suse.de>
4308 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4309 explicit single_use checks.
4311 2018-03-27 Richard Biener <rguenther@suse.de>
4313 PR tree-optimization/85082
4314 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4317 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4319 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4320 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4321 Turn on fasynchronous-unwind-tables and funwind-tables.
4323 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4326 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4327 (*bmi_blsr_<mode>_ccz): Ditto.
4329 2018-03-26 Tom de Vries <tom@codesourcery.com>
4331 PR tree-optimization/85063
4332 * omp-general.c (offloading_function_p): New function. Factor out
4334 * omp-offload.c (pass_omp_target_link::gate): ... here.
4335 * omp-general.h (offloading_function_p): Declare.
4336 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4337 with attribute omp declare target for offloading functions.
4339 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4341 PR tree-optimization/84005
4342 * tree-data-ref.h (get_base_for_alignment): Declare.
4343 * tree-data-ref.c (get_base_for_alignment_1): New function.
4344 (get_base_for_alignment): Likewise.
4345 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4346 get_base_for_alignment to find a suitable base object, instead
4347 of always using drb->base_address.
4349 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4352 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4353 known size by default.
4355 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4358 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4359 and non BLKmode operands.
4361 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4364 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4365 Clean up attributes.
4367 2018-03-23 Richard Biener <rguenther@suse.de>
4370 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4371 we are going to emit early debug for LTO.
4373 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4376 * function.c (match_asm_constraints_1): Don't optimize if input
4377 doesn't satisfy general_operand predicate for output's mode.
4380 * alias.c (write_dependence_p): Don't require for x_canonicalized
4381 non-VOIDmode if x has VOIDmode.
4384 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4385 just don't try to optimize it rather than assert it never happens.
4387 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4389 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4390 macro expansions for definition of ST_INTERNAL_<mode> and
4391 LD_INTERNAL_<mode> builtins.
4392 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4394 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4396 (altivec_expand_st_builtin): Likewise.
4397 (altivec_expand_builtin): Remove calls to deleted functions.
4398 (rs6000_address_for_altivec): Delete this function.
4399 * config/rs6000/vector.md: Remove expands for
4400 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4402 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4405 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4406 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4407 re-computing once computed.
4408 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4409 (arm_init_machine_status): Initialize
4410 machine->static_chain_stack_bytes.
4412 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4415 * doc/extend.texi: Add four new prototypes for vec_ld.
4416 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4417 definitions for more logical presentation.
4418 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4419 entries for V1TI variants of __builtin_altivec_ld builtin.
4420 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4421 handling of V1TI variant of LVX icode pattern.
4422 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4423 (rs6000_gimple_fold_builtin): Likewise.
4424 (altivec_init_builtins): Add code to define
4425 __builtin_altivec_lvx_v1ti function.
4427 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4430 * function.c (match_asm_constraints_1): Don't do the optimization
4431 if input isn't a REG, SUBREG, MEM or constant.
4433 2018-03-22 Tom de Vries <tom@codesourcery.com>
4435 PR tree-optimization/84956
4436 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4437 bb_has_abnormal_pred.
4439 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4442 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4443 DECL_INITIAL (decl) to decl at the end.
4444 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4447 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4449 * doc/extend.texi (__builtin_tgmath): Document when complex
4450 integer types are treated as _Complex _Float64.
4452 2018-03-21 Tom de Vries <tom@codesourcery.com>
4454 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4456 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4458 PR tree-optimization/84960
4459 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4460 if it is ENTRY block, move them into single succ of ENTRY in that case.
4462 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4464 PR tree-optimization/84811
4465 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4466 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4467 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4468 (known_subrange_p): Update accordingly. Cast each value involved
4469 in the size comparison, rather than casting the result of the
4472 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4474 PR tree-optimization/84982
4475 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4476 by flipping the least significant bit rather than all bits from
4477 bitpos to bitpos + bitsize - 1.
4479 2018-03-21 Nathan Sidwell <nathan@acm.org>
4481 * doc/extend.texi (Deprecated Features): Remove mention of
4482 long-deleted deprecations.
4484 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4487 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4488 * configure: Regenerate.
4490 2018-03-21 Tom de Vries <tom@codesourcery.com>
4492 PR tree-optimization/83126
4493 * tree-parloops.c (num_phis): New function.
4494 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4496 2018-03-21 Nathan Sidwell <nathan@acm.org>
4498 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4499 mention anon-struct/union members and trailing attributes.
4501 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4503 PR tree-optimization/84969
4504 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4505 builtin memset partitions if they set different rhs values.
4507 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4509 PR rtl-optimization/84989
4510 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4511 VEC_DUPLICATE with scalar result mode.
4513 2018-03-21 Martin Liska <mliska@suse.cz>
4516 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4517 not intended return statement.
4519 2018-03-21 Martin Liska <mliska@suse.cz>
4522 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4523 (chkp_find_bound_slots_1): Limit number of iterations.
4525 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4528 * Minor grammar fixes for x86 options.
4530 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4533 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4534 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4536 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4539 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4540 (altivec_lvx_<mode>_1op): Likewise.
4541 (altivec_stvx_<mode>_2op): Likewise.
4542 (altivec_stvx_<mode>_1op): Likewise.
4543 (altivec_lvx_<VM2:mode>): New define_expand.
4544 (altivec_stvx_<VM2:mode>): Likewise.
4545 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4546 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4547 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4548 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4549 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4550 (rs6000_gen_lvx): Likewise.
4551 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4552 (altivec_expand_stv_builtin): Likewise.
4553 (altivec_expand_builtin): Likewise.
4554 * config/rs6000/vector.md: Likewise.
4556 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4559 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4562 2018-03-20 Richard Biener <rguenther@suse.de>
4565 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4566 sign-conversions as zero, fall back to standard scalar_stmt
4569 2018-03-20 Martin Liska <mliska@suse.cz>
4572 * predict.c (rebuild_frequencies): Handle case when we have
4573 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4575 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4578 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4579 flag_section_anchors.
4580 * varasm.c (use_blocks_for_decl_p): Remove hack for
4581 dw2_force_const_mem.
4584 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4586 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4587 be created, use lowpart_subreg of operands[0] rather than operands[0]
4589 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4590 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4591 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4592 and n constraint instead of aarch64_shift_imm_di and Usd.
4593 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4594 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4596 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4599 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4600 to favor GPR over NEON registers.
4601 (<shift>di3_neon): Likewise.
4603 2018-03-20 Tom de Vries <tom@codesourcery.com>
4606 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4607 (nvptx_process_pars): Emit bar.sync asap and alap.
4609 2018-03-20 Tom de Vries <tom@codesourcery.com>
4612 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4613 seen_label if seen_label is already set.
4615 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4618 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4619 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4620 Use 1U instead of 1. Formatting fixes.
4623 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4624 instead of TREE_TYPE (s1) for the return value.
4626 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4628 PR tree-optimization/84946
4629 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4630 bitsize + bitsize in poly_uint64 rather than poly_int64.
4633 * dwarf2asm.c: Include fold-const.c.
4634 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4635 of decl rather than decl itself.
4637 PR rtl-optimization/84643
4638 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4640 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4643 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4644 calling assemble_variable.
4646 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4649 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4650 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4652 2018-03-19 Jim Wilson <jimw@sifive.com>
4655 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4656 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4657 (riscv_first_stack_step): Likewise.
4658 (riscv_option_override): Use STACK_BOUNDARY instead of
4660 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4662 (BIGGEST_ALIGNMENT): Set to 128.
4663 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4664 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4667 2018-03-19 Richard Biener <rguenther@suse.de>
4669 PR tree-optimization/84933
4670 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4671 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4673 2018-03-19 Richard Biener <rguenther@suse.de>
4675 PR tree-optimization/84859
4676 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4677 (cond_if_else_store_replacement): Perform sinking operation on
4678 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4679 Generalize what a BB with a single eligible store is.
4681 2018-03-19 Richard Biener <rguenther@suse.de>
4683 PR tree-optimization/84929
4684 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4685 chrec_is_positive against non-chrec arg.
4687 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4690 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4692 2018-03-18 Martin Liska <mliska@suse.cz>
4694 PR rtl-optimization/84635
4695 * regrename.c (build_def_use): Use matches_mode only when
4698 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4700 PR tree-optimization/84913
4701 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4702 vectorize chains of COND_EXPRs.
4704 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4706 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4708 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4710 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4712 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4714 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4716 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4717 Kito Cheng <kito.cheng@gmail.com>
4719 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4720 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4721 (nds32_adjust_reg_alloc_order): New function.
4722 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4724 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4726 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4727 nds32_print_operand, nds32_print_operand_address): Use
4728 HOST_WIDE_INT_PRINT_DEC instead.
4730 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4732 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4734 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4737 * config/i386/i386.c (initial_ix86_tune_features,
4738 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4740 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4741 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4742 rather than 1u << ix86_tune. Formatting fix.
4743 (ix86_option_override_internal): Change ix86_arch_mask from
4744 unsigned int to unsigned HOST_WIDE_INT, initialize to
4745 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4746 (ix86_function_specific_restore): Likewise.
4748 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4751 * postreload.c (reload_combine_recognize_pattern): Perform
4752 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4753 truncate_int_for_mode the result for the destination's mode.
4756 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4757 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4759 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4762 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4765 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4766 regno_allocno_class_array and sorted_pseudos.
4767 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4768 insns where regno is used.
4770 2018-03-16 Martin Liska <mliska@suse.cz>
4773 * multiple_target.c (create_dispatcher_calls): Redirect
4774 reference in the symbol table.
4776 2018-03-16 Martin Liska <mliska@suse.cz>
4779 * multiple_target.c (create_dispatcher_calls): Redirect also
4782 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4786 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4787 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4788 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4790 2018-03-16 Julia Koval <julia.koval@intel.com>
4792 * doc/invoke.texi (Skylake Server): Add CLWB.
4793 Cannonlake): Remove CLWB.
4795 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4797 PR tree-optimization/84841
4798 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4800 (FLOAT_ONE_CONST_TYPE): Define.
4801 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4802 (sort_by_operand_rank): Put entries with higher constant_type last
4803 rather than first to match comments.
4805 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4807 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4810 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4813 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4814 check and use address of target always.
4816 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4819 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4820 (indirect_thunk_bnd_needed): Likewise.
4821 (indirect_thunks_used): Likewise.
4822 (indirect_thunks_bnd_used): Likewise.
4823 (indirect_return_needed): New.
4824 (indirect_return_bnd_needed): Likewise.
4825 (output_indirect_thunk_function): Add a bool argument for
4827 (output_indirect_thunk_function): Don't generate alias for
4828 function return thunk.
4829 (ix86_code_end): Call output_indirect_thunk_function to generate
4830 function return thunks.
4831 (ix86_output_function_return): Set indirect_return_bnd_needed
4832 and indirect_return_needed instead of indirect_thunk_bnd_needed
4833 and indirect_thunk_needed.
4835 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4837 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4838 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4839 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4841 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4842 Paul Hua <paul.hua.gm@gmail.com>
4845 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4847 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4849 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4850 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4853 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4856 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4857 instead of GET_MODE_SIZE when comparing Units.
4859 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4862 * varasm.c (hash_section): Return an unchangeble hash value
4863 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4864 Return !aarch64_can_use_per_function_literal_pools_p ().
4866 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4869 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4870 rather than address of cmode as last argument to prepare_cmp_insn.
4872 2018-03-15 Julia Koval <julia.koval@intel.com>
4874 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4875 F_AVX512VNNI, F_AVX512BITALG): New.
4877 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4880 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4881 insn for floating-point loads and stores.
4883 2018-03-14 Carl Love <cel@us.ibm.com>
4885 * config/rs6000/rs6000-c.c: Add macro definitions for
4886 ALTIVEC_BUILTIN_VEC_PERMXOR.
4887 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4888 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4889 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4890 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4892 * config/doc/extend.texi: Add prototypes for vec_permxor.
4894 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4897 * diagnostic-show-locus.c (class layout_point): Convert m_line
4898 from int to linenum_type.
4899 (line_span::comparator): Use linenum "compare" function when
4900 comparing line numbers.
4901 (test_line_span): New function.
4902 (layout_range::contains_point): Convert param "row" from int to
4904 (layout_range::intersects_line_p): Likewise.
4905 (layout::will_show_line_p): Likewise.
4906 (layout::print_source_line): Likewise.
4907 (layout::should_print_annotation_line_p): Likewise.
4908 (layout::print_annotation_line): Likewise.
4909 (layout::print_leading_fixits): Likewise.
4910 (layout::annotation_line_showed_range_p): Likewise.
4911 (struct line_corrections): Likewise for field m_row.
4912 (line_corrections::line_corrections): Likewise for param "row".
4913 (layout::print_trailing_fixits): Likewise.
4914 (layout::get_state_at_point): Likewise.
4915 (layout::get_x_bound_for_row): Likewise.
4916 (layout::print_line): Likewise.
4917 (diagnostic_show_locus): Likewise for locals "last_line" and
4919 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4920 * input.c (selftest::test_linenum_comparisons): New function.
4921 (selftest::input_c_tests): Call it.
4922 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4923 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4924 * selftest.h (ASSERT_GT): New macro.
4925 (ASSERT_GT_AT): New macro.
4926 (ASSERT_LT): New macro.
4927 (ASSERT_LT_AT): New macro.
4929 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4931 PR rtl-optimization/84780
4932 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4934 2018-03-14 Martin Liska <mliska@suse.cz>
4936 * tree.c (record_node_allocation_statistics): Use
4937 get_stats_node_kind.
4938 (get_stats_node_kind): New function extracted from
4939 record_node_allocation_statistics.
4940 (free_node): Use get_stats_node_kind.
4942 2018-03-14 Richard Biener <rguenther@suse.de>
4944 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4945 that the value-set of ANTIC_IN doesn't grow.
4948 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4950 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4951 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4953 2018-03-14 Julia Koval <julia.koval@intel.com>
4955 * config.gcc (icelake-client, icelake-server): New.
4957 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4958 (initial_ix86_arch_features): Ditto.
4959 (PTA_SKYLAKE): Add SGX.
4960 (PTA_ICELAKE): Remove.
4961 (PTA_ICELAKE_CLIENT): New.
4962 (PTA_ICELAKE_SERVER): New.
4963 (ix86_option_override_internal): Split up icelake on icelake client and
4965 (get_builtin_code_for_version): Ditto.
4966 (fold_builtin_cpu): Ditto.
4967 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4968 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4969 * config/i386/i386.h (processor_type): Ditto.
4970 * doc/invoke.texi: Ditto.
4972 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4975 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4976 INTEGER_CST offset, add it together with bitpos / 8 and
4977 sign extend based on POINTER_SIZE.
4981 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4984 * config/i386/constraints.md (Yc): New register constraint.
4985 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4986 Use Yc constraint for alternative 2 of operand 0. Remove
4987 preferred_for_speed attribute.
4989 2018-03-14 Richard Biener <rguenther@suse.de>
4991 PR tree-optimization/84830
4992 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4993 with the old one to avoid oscillations.
4995 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4998 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5000 (assign_by_spills): Return a flag of reload assignment failure.
5001 Do not process the reload assignment failures. Do not spill other
5002 reload pseudos if they has the same reg class. Update n if
5004 (lra_assign): Add a return arg. Set up from the result of
5005 assign_by_spills call.
5006 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5007 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5008 usage_insns if it is not NULL.
5009 (spill_hard_reg_in_range): New function.
5010 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5011 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5012 function prototypes.
5013 (lra_assign): Change prototype.
5014 * lra.c (lra): Add code to deal with fails by splitting hard reg
5017 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5019 * config/riscv/riscv.opt (mrelax): New option.
5020 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5021 "norelax" when riscv_mrelax is disabled.
5022 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5024 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5027 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5028 reassociation for int modes.
5030 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5032 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5033 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5035 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5036 * config/aarch64/aarch64-sve.md
5037 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5038 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5039 (*extend<mode><Vwide>2): Rename to...
5040 (aarch64_sve_extend<mode><Vwide>2): ...this.
5041 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5042 renaming the old pattern to...
5043 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5045 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5046 define_expand, renaming the old pattern to...
5047 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5048 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5049 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5050 account when deciding which SVE instruction the optab should use.
5051 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5053 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5055 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5056 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5057 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5058 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5059 (tlsdesc_small_<mode>): Turn a define_expand and use
5060 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5061 (tlsdesc_small_advsimd_<mode>): ...this.
5062 (tlsdesc_small_sve_<mode>): New pattern.
5064 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5066 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5067 (UNSPEC_UMUL_HIGHPART): New constants.
5068 (MUL_HIGHPART): New int iteraor.
5069 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5070 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5072 (*<su>mul<mode>3_highpart): New define_insn.
5074 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5077 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5080 2018-03-13 Martin Liska <mliska@suse.cz>
5083 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5085 (sem_item_optimizer::~sem_item_optimizer): Release it.
5086 (sem_item_optimizer::merge_classes): Register variable aliases.
5087 (sem_item_optimizer::fixup_pt_set): New function.
5088 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5089 * ipa-icf.h: Declare new variables and functions.
5091 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5094 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5095 integer_pow2p@2 and test integer_pow2p in condition.
5096 (A < 0 ? C : 0): Similarly for @1.
5099 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5100 characters starting at p contain '\0' character, don't look beyond
5104 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5105 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5108 * reg-stack.c (change_stack): Change update_end var from int to
5109 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5110 also call set_block_for_insn on the newly added insns and rescan.
5113 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5114 on the last operand.
5117 * tree.c (stabilize_reference_1): Return save_expr (e) for
5118 STATEMENT_LIST even if it doesn't have side-effects.
5120 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5122 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5124 2018-03-12 Renlin Li <renlin.li@arm.com>
5126 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5127 aarch64_output_scalar_simd_mov_immediate.
5129 2018-03-12 Martin Sebor <msebor@redhat.com>
5131 PR tree-optimization/83456
5132 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5133 for perfectly overlapping calls to memcpy.
5134 (gimple_fold_builtin_memory_chk): Same.
5135 (gimple_fold_builtin_strcpy): Handle no-warning.
5136 (gimple_fold_builtin_stxcpy_chk): Same.
5137 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5139 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5141 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5142 parameter. Use it for SFmode.
5143 (rs6000_function_arg_advance_1): Adjust.
5144 (rs6000_function_arg): Adjust.
5145 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5147 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5149 PR rtl-optimization/84169
5150 PR rtl-optimization/84780
5151 * combine.c (can_combine_p): Check for a 2-insn combination whether
5152 the destination register is used between the two insns, too.
5154 2018-03-12 Richard Biener <rguenther@suse.de>
5156 PR tree-optimization/84803
5157 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5158 for refs DR analysis didn't process.
5160 2018-03-12 Richard Biener <rguenther@suse.de>
5162 PR tree-optimization/84777
5163 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5164 force-vectorize loops ignore whether we are optimizing for size.
5166 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5168 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5169 (TARGET_MD_ASM_ADJUST): Define.
5171 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5172 Kito Cheng <kito.cheng@gmail.com>
5173 Chung-Ju Wu <jasonwucj@gmail.com>
5175 * config/nds32/nds32.c (nds32_compute_stack_frame,
5176 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5177 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5178 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5179 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5180 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5181 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5182 * config/nds32/nds32.md (prologue, epilogue): Use macro
5183 NDS32_V3PUSH_AVAILABLE_P to do checking.
5185 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5188 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5189 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5190 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5191 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5192 addition of most attributes on !orig_type_die or the attribute not
5193 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5195 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5196 Chung-Ju Wu <jasonwucj@gmail.com>
5198 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5200 * config/nds32/nds32.opt (mvh): New option.
5202 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5203 Chung-Ju Wu <jasonwucj@gmail.com>
5205 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5207 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5208 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5211 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5212 Chung-Ju Wu <jasonwucj@gmail.com>
5214 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5216 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5217 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5219 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5220 Kito Cheng <kito.cheng@gmail.com>
5221 Chung-Ju Wu <jasonwucj@gmail.com>
5223 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5224 UNSPEC_FFMISM and UNSPEC_FLMISM.
5225 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5226 for ffb, ffmism and flmism.
5227 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5228 (unspec_ffmism): Ditto.
5229 (unspec_flmism): Ditto.
5230 (nds32_expand_builtin_impl): Check if string extension is available.
5231 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5232 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5234 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5237 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5240 * lra-assigns.c (assign_by_spills): Return a flag of reload
5241 assignment failure. Do not process the reload assignment
5242 failures. Do not spill other reload pseudos if they has the same
5244 (lra_assign): Add a return arg. Set up from the result of
5245 assign_by_spills call.
5246 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5247 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5248 usage_insns if it is not NULL.
5249 (spill_hard_reg_in_range): New function.
5250 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5251 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5252 function prototypes.
5253 (lra_assign): Change prototype.
5254 * lra.c (lra): Add code to deal with fails by splitting hard reg
5257 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5260 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5262 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5265 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5266 (dw_val_node): Add val_symbolic_view.
5267 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5268 (symview_upper_bound): New.
5269 (new_line_info_table): Initialize symviews_since_reset.
5270 (dwarf2out_source_line): Count symviews_since_reset and set
5271 symview_upper_bound.
5272 (dw_val_equal_p): Handle symview.
5273 (add_AT_symview): New.
5274 (print_dw_val): Handle symview.
5275 (attr_checksum, attr_checksum_ordered): Likewise.
5276 (same_dw_val_p, size_of_die): Likewise.
5277 (value_format, output_die): Likewise.
5278 (add_high_low_attributes): Use add_AT_symview for entry_view.
5279 (dwarf2out_finish): Reset symview_upper_bound, clear
5282 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5285 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5286 Add strict argument and use it.
5287 (rs6000_split_multireg_move): Update for new strict argument.
5288 (mem_operand_gpr): Disallow all non-offsettable addresses.
5289 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5291 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5294 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5295 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5296 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5299 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5300 decl, use remap_type if we want to use the type.
5302 2018-03-09 Martin Sebor <msebor@redhat.com>
5304 PR tree-optimization/84526
5305 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5307 (builtin_access::generic_overlap): Be prepared to handle non-array
5310 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5312 PR rtl-optimization/84682
5313 * lra-constraints.c (process_address_1): Check is_address flag
5314 for address constraints.
5315 (process_alt_operands): Likewise.
5316 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5317 preprocess_constraints.
5318 * recog.h (preprocess_constraints): Add oploc parameter.
5320 * recog.c (preprocess_constraints): Test address_operand for
5321 CT_ADDRESS constraints.
5323 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5326 * lra-assigns.c (assign_by_spills): Return a flag of reload
5327 assignment failure. Do not process the reload assignment
5328 failures. Do not spill other reload pseudos if they has the same
5330 (lra_assign): Add a return arg. Set up from the result of
5331 assign_by_spills call.
5332 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5333 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5334 usage_insns if it is not NULL.
5335 (spill_hard_reg_in_range): New function.
5336 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5337 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5338 function prototypes.
5339 (lra_assign): Change prototype.
5340 * lra.c (lra): Add code to deal with fails by splitting hard reg
5343 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5346 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5347 Accept complain bool parameter. Only emit errors if it is true.
5348 (arm_parse_cpu_option_name): Likewise.
5349 (arm_target_thumb_only): Adjust callers of the above.
5350 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5351 prototype to take a default true bool parameter.
5352 (arm_parse_arch_option_name): Likewise.
5354 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5355 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5359 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5360 * configure: Regenerate.
5361 * configure.ac ("linker --version-script option"): New.
5362 ("linker soname option"): New.
5364 2018-03-09 Richard Biener <rguenther@suse.de>
5366 PR tree-optimization/84775
5367 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5368 immediate uses of predicate stmts and mark them modified.
5371 PR tree-optimization/84178
5372 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5374 (version_loop_for_if_conversion): Delay update_ssa call.
5375 (tree_if_conversion): Delay update_ssa until after predicate
5378 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5381 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5382 when the function accesses prior frames.
5384 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5387 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5388 gen_llsym, otherwise call maybe_gen_llsym.
5391 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5392 has ',' character inside of it.
5394 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5397 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5398 as clobbering CC_REGNUM.
5400 2018-03-08 Richard Biener <rguenther@suse.de>
5403 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5404 (follow_copies_to_constant): Do not follow SSA names registered
5407 2018-03-08 Richard Biener <rguenther@suse.de>
5409 PR tree-optimization/84178
5410 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5412 (version_loop_for_if_conversion): Delay update_ssa call.
5413 (tree_if_conversion): Delay update_ssa until after predicate
5416 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5418 PR tree-optimization/84178
5419 * tree-if-conv.c (release_bb_predicate): Remove the
5420 the assertion that the stmts have NULL use_ops.
5421 Discard the statements, asserting that they haven't
5422 yet been added to a BB.
5424 2018-03-08 Richard Biener <rguenther@suse.de>
5426 PR tree-optimization/84746
5427 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5428 (phi_translate): Pass in destination ANTIC_OUT set.
5429 (phi_translate_1): Likewise. For a simplified result lookup
5430 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5431 (phi_translate_set): Adjust.
5432 (do_pre_regular_insertion): Likewise.
5433 (do_pre_partial_partial_insertion): Likewise.
5435 2018-03-08 Martin Liska <mliska@suse.cz>
5437 PR gcov-profile/84735
5438 * doc/gcov.texi: Document usage of profile files.
5439 * gcov-io.h: Document changes in the format.
5441 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5445 * dwarf2out.c (struct dw_line_info_table): Update comments for
5447 (FORCE_RESET_NEXT_VIEW): New.
5448 (FORCE_RESETTING_VIEW_P): New.
5449 (RESETTING_VIEW_P): Check for -1 too.
5450 (ZERO_VIEW_P): Likewise.
5451 (new_line_info_table): Force-reset next view.
5452 (dwarf2out_begin_function): Likewise.
5453 (dwarf2out_source_line): Simplify zero_view_p initialization.
5454 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5455 view directly. Omit view when omitting .loc at line 0.
5457 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5459 PR tree-optimization/84740
5460 * tree-switch-conversion.c (process_switch): Call build_constructors
5461 only if info.phi_count is non-zero.
5463 PR tree-optimization/84739
5464 * tree-tailcall.c (find_tail_calls): Check call arguments against
5465 DECL_ARGUMENTS (current_function_decl) rather than
5466 DECL_ARGUMENTS (func) when checking for tail recursion.
5468 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5470 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5471 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5472 Volker Reichelt's entry and add entries for people that perform
5473 GCC fuzzy testing and report numerous bugs.
5475 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5478 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5479 readonly data in sdata, if that is disabled.
5480 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5481 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5482 -mreadonly-in-sdata option.
5484 2018-03-07 Martin Sebor <msebor@redhat.com>
5486 PR tree-optimization/84468
5487 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5488 basic block when looking for nul assignment.
5490 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5493 * except.h (output_function_exception_table): Adjust prototype.
5494 * except.c (output_function_exception_table): Remove FNNAME parameter
5495 and add SECTION parameter. Ouput one part of the table at a time.
5496 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5497 the first part of the exception table and emit unwind directives.
5498 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5499 (i386_pe_seh_cold_init): Likewise.
5500 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5501 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5502 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5503 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5504 * config/i386/winnt.c: Include except.h.
5505 (struct seh_frame_state): Add reg_offset, after_prologue and
5506 in_cold_section fields.
5507 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5508 (i386_pe_seh_cold_init): New function.
5509 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5510 to seh->in_cold_section.
5511 (seh_emit_push): Record the offset of the push.
5512 (seh_emit_save): Record the offet of the save.
5513 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5514 Test seh->after_prologue to disregard the epilogue.
5515 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5516 (i386_pe_end_cold_function): New function.
5518 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5521 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5522 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5525 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5527 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5528 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5531 * multiple_target.c: Include tree-inline.h and intl.h.
5532 (expand_target_clones): Diagnose and fail if node->definition and
5533 !tree_versionable_function_p (node->decl).
5535 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5537 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5539 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5540 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5541 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5543 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5546 * combine.c (try_combine): Use reg_or_subregno instead of handling
5547 just paradoxical SUBREGs and REGs.
5549 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5551 * config/arc/arc.c (arc_finalize_pic): Remove function.
5552 (arc_must_save_register): We use single base PIC register, remove
5553 checks to save/restore the PIC register.
5554 (arc_expand_prologue): Likewise.
5555 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5557 (arc_verify_short): Likewise.
5558 (arc_attr_type): Likewise.
5559 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5560 (walk_stores): Likewise.
5561 (arc_address_cost): Make it static.
5562 (arc_verify_short): Likewise.
5563 (branch_dest): Likewise.
5564 (arc_attr_type): Likewise.
5565 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5566 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5567 (arc_final_prescan_insn): Remove inserting the nops due to
5568 hardware hazards. It is done in reorg step.
5569 (insn_length_variant_t): Remove.
5570 (insn_length_parameters_t): Likewise.
5571 (arc_insn_length_parameters): Likewise.
5572 (arc_get_insn_variants): Likewise.
5573 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5575 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5578 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5581 PR tree-optimization/84687
5582 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5584 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5586 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5588 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5589 Rename to ppc_speculation_barrier.
5590 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5591 __builtin_ppc_speculation_barrier.
5593 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5596 * combine.c (combine_simplify_rtx): Don't try to simplify if
5597 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5600 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5602 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5603 to 32 bytes when compiling for POWER9.
5605 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5608 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5609 regparm >= 3 with no arg reg available also for calls with
5610 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5613 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5615 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5617 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5620 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5622 2018-03-05 Richard Biener <rguenther@suse.de>
5624 PR tree-optimization/84486
5625 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5626 When inserting a __builtin_assume_aligned call set the LHS
5627 SSA name alignment info accordingly.
5629 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5631 PR tree-optimization/84114
5632 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5633 Avoid reassociation of FLOAT_MODE addition.
5635 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5637 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5638 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5639 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5640 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5641 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5642 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5643 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5645 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5646 __builtin_ia32_wbinvd): New builtins.
5647 (SPECIAL_ARGS2): New.
5648 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5649 (SPECIAL_ARGS2): New.
5650 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5651 (ix86_valid_target_attribute_inner_p): Ditto.
5652 (ix86_init_mmx_sse_builtins): Add special_args2.
5653 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5654 TARGET_WBNOINVD_P): New.
5655 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5656 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5657 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5658 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5659 * config/i386/pconfigintrin.h: New file.
5660 * config/i386/wbnoinvdintrin.h: Ditto.
5661 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5662 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5664 2018-03-05 Richard Biener <rguenther@suse.de>
5666 PR tree-optimization/84670
5667 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5669 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5670 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5671 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5672 doesn't grow if all successors (recursively) were visited at least
5675 2018-03-05 Richard Biener <rguenther@suse.de>
5677 PR tree-optimization/84650
5678 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5679 if executed in the loop pipeline.
5681 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5683 * doc/configfiles.texi (Configuration Files): Move info about
5684 conditionalizing $target-protos.h to...
5685 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5686 differs from $target-protos.h.
5688 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5689 Chung-Ju Wu <jasonwucj@gmail.com>
5691 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5692 * config/nds32/nds32-multiple.md (setmemsi): Define.
5693 * config/nds32/nds32-memory-manipulation.c
5694 (nds32_gen_dup_4_byte_to_word_value): New.
5695 (emit_setmem_word_loop): New.
5696 (emit_setmem_byte_loop): New.
5697 (nds32_expand_setmem_loop): New.
5698 (nds32_expand_setmem_loop_v3m): New.
5699 (nds32_expand_setmem_unroll): New.
5700 (nds32_expand_setmem): New.
5702 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5703 Chung-Ju Wu <jasonwucj@gmail.com>
5705 * config/nds32/nds32-memory-manipulation.c
5706 (nds32_emit_load_store): New.
5707 (nds32_emit_post_inc_load_store): New.
5708 (nds32_emit_mem_move): New.
5709 (nds32_emit_mem_move_block): New.
5710 (nds32_expand_movmemsi_loop_unknown_size): New.
5711 (nds32_expand_movmemsi_loop_known_size): New.
5712 (nds32_expand_movmemsi_loop): New.
5713 (nds32_expand_movmemsi_unroll): New.
5714 (nds32_expand_movmemqi): Rename ...
5715 (nds32_expand_movmemsi): ... to this.
5716 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5717 (movmemsi): ... to this.
5718 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5719 (nds32_expand_movmemsi): ... to this.
5721 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5722 Monk Chiang <sh.chiang04@gmail.com>
5723 Chung-Ju Wu <jasonwucj@gmail.com>
5725 * config/nds32/nds32-protos.h
5726 (nds32_expand_load_multiple): New arguments.
5727 (nds32_expand_store_multiple): Ditto.
5728 (nds32_valid_multiple_load_store): Rename ...
5729 (nds32_valid_multiple_load_store_p): ... to this.
5730 * config/nds32/nds32-memory-manipulation.c
5731 (nds32_expand_load_multiple): Refine implementation.
5732 (nds32_expand_store_multiple): Ditto.
5733 * config/nds32/nds32-multiple.md
5734 (load_multiple): Update nds32_expand_load_multiple interface.
5735 (store_multiple): Update nds32_expand_store_multiple interface.
5736 * config/nds32/nds32-predicates.c
5737 (nds32_valid_multiple_load_store): Rename ...
5738 (nds32_valid_multiple_load_store_p): ... to this and refine
5740 * config/nds32/predicates.md
5741 (nds32_load_multiple_and_update_address_operation): New predicate.
5742 (nds32_store_multiple_and_update_address_operation): New predicate.
5744 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5745 Chung-Ju Wu <jasonwucj@gmail.com>
5747 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5748 (combo): New attribute.
5749 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5751 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5753 * config/nds32/nds32.opt: Change -mcmodel= default value.
5755 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5756 Monk Chiang <sh.chiang04@gmail.com>
5757 Chung-Ju Wu <jasonwucj@gmail.com>
5759 * config/nds32/constants.md (unspec_element): New enum.
5760 * config/nds32/constraints.md (Umw): New constraint.
5761 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5762 * config/nds32/nds32-intrinsic.md: Likewise.
5763 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5764 (nds32_valid_smw_lwm_base_p): New.
5765 (nds32_output_smw_single_word): New.
5766 (nds32_output_lmw_single_word): New.
5767 (nds32_expand_unaligned_load): New.
5768 (nds32_expand_unaligned_store): New.
5769 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5770 (nds32_output_smw_single_word): Declare.
5771 (nds32_output_lmw_single_word): Declare.
5772 (nds32_expand_unaligned_load): Declare.
5773 (nds32_expand_unaligned_store): Declare.
5774 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5775 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5776 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5777 NDS32_BUILTIN_UASTORE_DW.
5778 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5781 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5782 Kito Cheng <kito.cheng@gmail.com>
5783 Chung-Ju Wu <jasonwucj@gmail.com>
5785 * config/nds32/nds32-intrinsic.c
5786 (nds32_expand_builtin_null_ftype_reg): Delete.
5787 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5788 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5789 (nds32_read_argument): New.
5790 (nds32_legitimize_target): Ditto.
5791 (nds32_legitimize_argument): Ditto.
5792 (nds32_check_constant_argument): Ditto.
5793 (nds32_expand_unop_builtin): Ditto.
5794 (nds32_expand_unopimm_builtin): Ditto.
5795 (nds32_expand_binop_builtin): Ditto.
5796 (nds32_builtin_decl_impl): Ditto.
5797 (builtin_description): Ditto.
5798 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5799 (nds32_init_builtins_impl): Ditto.
5800 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5801 (nds32_builtin_decl): New.
5802 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5803 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5805 2018-03-02 Jeff Law <law@redhat.com>
5807 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5808 (redundant_insn, fill_simple_delay_slots): Likewise.
5809 (fill_slots_from_thread): Likewise.
5810 * resource.c (mark_referenced_resources): Likewise.
5811 (mark_set_resources, find_dead_or_set_registers): Likewise.
5813 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5815 * substring-locations.h (format_warning_va): Formatting fix for
5817 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5819 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5820 * substring-locations.c: Include intl.h.
5821 (format_warning_va): Turned into small wrapper around
5822 format_warning_n_va, renamed to ...
5823 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5824 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5826 (format_warning_at_substring_n): New function.
5827 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5828 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5829 format_warning_at_substring with just a shorter name instead of
5830 const function pointer.
5831 (fmtwarn_n): New function.
5832 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5833 appropriate, get rid of all the fmtstr temporaries, move conditionals
5834 with G_() wrapped string literals directly into fmtwarn arguments,
5835 cast dir.len to (int), formatting fixes.
5837 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5839 * doc/invoke.texi: Remove "Cilk Plus" references.
5841 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5842 Richard Biener <rguenther@suse.de>
5845 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5846 for error or warning attributes if CALL_FROM_THUNK_P is set.
5849 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5852 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5853 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5856 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5857 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5859 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5860 instead of -1U in last predictors element's probability member.
5862 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5865 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5866 arguments if they are comparable.
5868 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5870 PR tree-optimization/84634
5871 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5872 masks and masked_loop_p with a single loop_masks, making sure it's
5873 null for bb vectorization.
5875 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5877 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5878 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5879 loop->force_vectorize to check whether there is no alias.
5881 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5884 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5886 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5888 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5889 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5890 instead of a loop around prev_real_insn.
5891 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5895 * config/i386/i386.c (ix86_print_operand): Use conditional
5896 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5899 2018-03-02 Richard Biener <rguenther@suse.de>
5901 PR tree-optimization/84427
5902 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5903 (bitmap_set_subtract_values): Rewrite to handle multiple
5906 (prune_clobbered_mems): Likewise.
5907 (phi_translate): Take edge instead of pred/phiblock.
5908 (phi_translate_1): Likewise.
5909 (phi_translate_set): Likewise. Insert all translated
5910 exprs for a value into the set, keeping possibly multiple
5911 expressions per value.
5912 (compute_antic_aux): Adjust for phi_translate changes.
5913 When intersecting union the expressions and prune those
5914 not in the final value set, keeping possibly multiple
5915 expressions per value. Do not use value-insertion
5916 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5917 all expressions. Add verification that the value-sets
5918 only shrink during iteration.
5919 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5920 (do_pre_regular_insertion): Likewise.
5921 (do_pre_partial_partial_insertion): Likewise.
5923 2018-03-02 Richard Biener <rguenther@suse.de>
5926 * config/darwin.c (saved_debug_info_level): New static global.
5927 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5928 (darwin_asm_lto_end): Restore debug info generation settings.
5930 2018-03-01 Martin Liska <mliska@suse.cz>
5933 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5936 2018-03-01 Richard Biener <rguenther@suse.de>
5939 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5940 type annotation with LTO when debug was disabled at compile-time.
5942 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5944 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5946 (mips_final_postscan_insn): Likewise.
5948 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5950 PR rtl-optimization/84528
5951 * alias.c (init_alias_target): Add commentary.
5952 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5953 a unique base value if the frame pointer is not eliminated
5954 to the stack pointer.
5956 2018-03-01 Tom de Vries <tom@codesourcery.com>
5958 PR rtl-optimization/83327
5959 * lra-int.h (hard_regs_spilled_into): Declare.
5960 * lra.c (hard_regs_spilled_into): Define.
5961 (init_reg_info): Init hard_regs_spilled_into.
5962 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5963 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5964 (process_bb_lives): Handle hard_regs_spilled_into.
5965 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5966 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5968 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5970 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5971 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5972 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5973 * config/rs6000/aix72.h: New file.
5975 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5977 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5978 instead of warning_at with conditional singular and plural messages
5982 * stor-layout.c (update_alignment_for_field): For
5983 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5984 && !DECL_PACKED (field), do the alignment update, just use
5985 only desired_align instead of MAX (type_align, desired_align)
5987 (place_field): Don't do known_align < desired_align handling
5988 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5989 is non-NULL, instead do it after rli->prev_field handling and
5990 only if not within a bitfield word. For DECL_PACKED (field)
5991 use type_align of BITS_PER_UNIT.
5993 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5995 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5996 superfluous parentheses and trailing spaces.
5998 2018-02-28 Richard Biener <rguenther@suse.de>
6000 PR tree-optimization/84584
6001 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6002 SCoPs with fake exit edge.
6004 2018-02-28 Martin Liska <mliska@suse.cz>
6007 * timevar.c (timer::print): Fix format to properly print 100%
6010 2018-02-28 Richard Biener <rguenther@suse.de>
6013 * genmatch.c (capture_info::walk_match): Do not mark
6014 captured expressions without operands as expr_p given
6015 they act more like predicates and should be subject to
6016 "lost tail" side-effect preserving.
6018 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6020 PR rtl-optimization/81611
6021 * auto-inc-dec.c (attempt_change): Move dead note from
6022 mem_insn if it's the next use of regno
6023 (find_address): Take address use of reg holding
6024 non-incremented value. Add parm to limit search to the named
6026 (merge_in_block): Attempt to use a mem insn that is the next
6027 use of the original regno.
6029 2018-02-27 Martin Sebor <msebor@redhat.com>
6032 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6033 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6035 2018-02-27 Martin Sebor <msebor@redhat.com>
6037 PR translation/84207
6038 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6039 n argument to unsigned HOST_WIDE_INT.
6040 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6041 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6042 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6043 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6045 2018-02-27 Richard Biener <rguenther@suse.de>
6047 PR tree-optimization/84512
6048 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6049 Do not use the estimate returned from record_stmt_cost for
6050 the scalar iteration cost but sum properly using add_stmt_cost.
6052 2018-02-27 Richard Biener <rguenther@suse.de>
6054 PR tree-optimization/84466
6055 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6056 Adjust last change to less strictly validate use operands.
6058 2018-02-27 Martin Liska <mliska@suse.cz>
6060 PR gcov-profile/84548
6061 * gcov.c (process_file): Allow partial overlap and consider it
6062 also as group functions.
6063 (output_lines): Properly calculate range of lines for a group.
6065 2018-02-27 Martin Liska <mliska@suse.cz>
6067 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6068 'ggc' suffixes. Change first column width.
6069 (timer::print): Fix formatting of the column.
6071 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6073 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6074 preserve inline entry blocks for the sake of debug inline
6075 entry point markers alone.
6076 (remove_unused_locals): Suggest in comments a better place to
6077 force the preservation of inline entry blocks that are
6078 otherwise unused, but do not preserve them.
6080 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6082 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6084 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6087 * config/i386/constraints.md (Bs): Replace
6088 ix86_indirect_branch_register with
6089 TARGET_INDIRECT_BRANCH_REGISTER.
6091 * config/i386/i386.md (indirect_jump): Likewise.
6092 (tablejump): Likewise.
6093 (*sibcall_memory): Likewise.
6094 (*sibcall_value_memory): Likewise.
6095 Peepholes of indirect call and jump via memory: Likewise.
6096 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6097 (*sibcall_value_GOT_32): Likewise.
6098 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6099 (GOT_memory_operand): Likewise.
6100 (call_insn_operand): Likewise.
6101 (sibcall_insn_operand): Likewise.
6102 (GOT32_symbol_operand): Likewise.
6103 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6105 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6107 PR rtl-optimization/83496
6108 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6109 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6110 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6111 redundant insn, if any.
6112 (relax_delay_slots): Likewise.
6113 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6115 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6117 PR tree-optimization/83965
6118 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6119 that grouped statements are part of a reduction chain. Return
6120 true if the statement is not marked as a reduction itself but
6122 (vect_recog_dot_prod_pattern): Don't check whether the statement
6123 is part of a group here.
6124 (vect_recog_sad_pattern): Likewise.
6125 (vect_recog_widen_sum_pattern): Likewise.
6127 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6130 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6132 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6135 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6137 (ix86_output_indirect_function_return): New prototype.
6138 (ix86_split_simple_return_pop_internal): Likewise.
6139 * config/i386/i386.c (indirect_return_via_cx): New.
6140 (indirect_return_via_cx_bnd): Likewise.
6141 (indirect_thunk_name): Handle return va CX_REG.
6142 (output_indirect_thunk_function): Create alias for
6143 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6144 (ix86_output_indirect_jmp): Remove the bool argument.
6145 (ix86_output_indirect_function_return): New function.
6146 (ix86_split_simple_return_pop_internal): Likewise.
6147 * config/i386/i386.md (*indirect_jump): Don't pass false
6148 to ix86_output_indirect_jmp.
6149 (*tablejump_1): Likewise.
6150 (simple_return_pop_internal): Change it to define_insn_and_split.
6151 Call ix86_split_simple_return_pop_internal to split it for
6153 (simple_return_indirect_internal): Call
6154 ix86_output_indirect_function_return instead of
6155 ix86_output_indirect_jmp.
6157 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6160 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6161 memset and value initialization afterwards.
6163 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6165 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6167 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6170 * common/config/aarch64/aarch64-common.c
6171 (aarch_option_optimization_table[]): Switch
6172 off fomit-frame-pointer
6174 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6175 Chung-Ju Wu <jasonwucj@gmail.com>
6177 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6179 (store_multiple): Ditto.
6181 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6183 * config.gcc: Add --with-cpu support for nds32 target.
6184 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6185 * config/nds32/nds32.opt: Add -mcpu= option.
6187 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6189 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6190 isel=yes): Warn for these deprecated options.
6192 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6194 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6195 ISA_2_5_MASKS_EMBEDDED.
6197 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6199 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6200 p->max as pointers rather than using iterative_hash_expr.
6202 2018-02-23 Carl Love <cel@us.ibm.com>
6204 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6205 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6207 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6208 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6209 P8V_BUILTIN_VEC_VUNSIGNED2.
6211 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6214 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6215 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6217 * lra-constraints.c (curr_insn_transform): Set up
6218 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6220 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6221 * lra-eliminations.c (spill_pseudos): Ditto.
6222 (process_insn_for_elimination): Ditto.
6223 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6224 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6225 LRA_NON_CLOBBERED_ALT.
6227 2018-02-22 Martin Sebor <msebor@redhat.com>
6229 PR tree-optimization/84480
6230 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6231 to maybe_diag_stxncpy_trunc. Call it.
6232 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6233 from gimple_fold_builtin_strcpy. Print inlining stack.
6234 (handle_builtin_stxncpy): Print inlining stack.
6235 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6237 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6240 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6241 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6242 and -fcheck-pointer-bounds are used together.
6243 (indirect_thunk_prefix): New enum.
6244 (indirect_thunk_need_prefix): New function.
6245 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6246 "_nt" instead of "_bnd" for NOTRACK prefix.
6247 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6248 (output_indirect_thunk_function): Likewise.
6250 (ix86_code_end): Update output_indirect_thunk_function calls.
6251 (ix86_output_indirect_branch_via_reg): Replace
6252 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6253 (ix86_output_indirect_branch_via_push): Likewise.
6254 (ix86_output_function_return): Likewise.
6255 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6256 incompatible with -fcf-protection=branch and
6257 -fcheck-pointer-bounds.
6259 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6262 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6263 Change gcc_assert call to output_operand_lossage.
6265 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6267 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6269 2018-02-22 DJ Delorie <dj@redhat.com>
6270 Sebastian Perta <sebastian.perta@renesas.com>
6271 Oleg Endo <olegendo@gcc.gnu.org>
6273 * config/rx/rx.c (rx_rtx_costs): New function.
6274 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6276 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6278 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6280 2018-02-22 Martin Liska <mliska@suse.cz>
6283 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6284 Add "native" as a possible value.
6286 2018-02-22 Martin Liska <mliska@suse.cz>
6289 * config/i386/i386.c (ix86_option_override_internal):
6290 Add "native" as a possible value for -march and -mtune.
6292 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6295 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6296 to all type variants.
6298 PR tree-optimization/84503
6299 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6300 width as info->bitpos + info->bitsize - start.
6301 (merged_store_group::merge_overlapping): Simplify width computation.
6302 (check_no_overlap): New function.
6303 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6304 start + width and last_order of the group, fail if check_no_overlap
6306 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6307 to group if check_no_overlap fails.
6309 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6311 * config/rs6000/altivec.md: Delete contraint arguments to
6312 define_expand, define_split, and define_peephole2, and in
6313 define_insn_and_split if always unused.
6314 * config/rs6000/darwin.md: Ditto.
6315 * config/rs6000/dfp.md: Ditto.
6316 * config/rs6000/rs6000.md: Ditto.
6317 * config/rs6000/sync.md: Ditto.
6318 * config/rs6000/vector.md: Ditto.
6319 * config/rs6000/vsx.md: Ditto.
6321 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6323 * config/rs6000/altivec.md: Write output control strings as braced
6324 blocks instead of double-quoted strings.
6325 * config/rs6000/darwin.md: Ditto.
6326 * config/rs6000/rs6000.md: Ditto.
6327 * config/rs6000/vector.md: Ditto.
6328 * config/rs6000/vsx.md: Ditto.
6330 2018-02-21 Jason Merrill <jason@redhat.com>
6332 PR c++/84314 - ICE with templates and fastcall attribute.
6333 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6335 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6337 * ipa-cp.c (determine_versionability): Fix comment typos.
6339 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6342 * ipa-cp.c (determine_versionability): Do not version functions caling
6345 2018-02-21 Martin Liska <mliska@suse.cz>
6348 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6349 Add "native" as a possible value.
6350 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6351 the macro when native cpu detection is available.
6353 2018-02-21 Martin Liska <mliska@suse.cz>
6356 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6357 Add "native" as a possible value.
6358 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6359 when native cpu detection is available.
6361 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6362 Martin Sebor <msebor@redhat.com>
6364 PR tree-optimization/84478
6365 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6367 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6368 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6369 support which is conservatively correct, for 2 only stay conservative
6370 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6371 argument to the 2 argument get_range_strlen, adjust 6 arg
6372 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6374 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6375 (gimple_fold_builtin_strlen): Pass true as last argument to
6378 2018-02-20 Martin Sebor <msebor@redhat.com>
6381 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6382 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6383 (builtin_memref::builtin_memref): Factor out parts into
6384 set_base_and_offset and call it.
6386 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6389 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6390 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6391 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6392 search at the associated MODE_INT.
6394 2018-02-20 Jeff Law <law@redhat.com>
6397 PR tree-optimization/81592
6399 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6400 for range data rather than using global data.
6401 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6402 range data rather than using global data.
6403 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6404 pass it to children as needed.
6405 (struct directive::fmtresult): Similarly.
6406 (struct directive::set_width): Similarly.
6407 (struct directive::set_precision): Similarly.
6408 (format_integer, format_directive, parse_directive): Similarly.
6409 (format_none): Accept unnamed vr_values parameter.
6410 (format_percent, format_floating, format_character): Similarly.
6411 (format_string, format_plain): Similarly.
6412 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6413 the EVRP range analyzer for range data rather than using global data.
6414 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6415 gimple-ssa-evrp-analyze.h
6416 (class sprintf_dom_walker): Add after_dom_children member function.
6417 Add evrp_range_analyzer member.
6418 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6419 range analyzer as needed.
6420 (sprintf_dom_walker::after_dom_children): New member function.
6421 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6423 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6424 (evrp_range_analyzer::pop_to_marker): Likewise.
6426 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6428 PR tree-optimization/84419
6429 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6430 with the required type if its current type is compatible but
6433 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6436 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6437 after vectorization.
6439 2018-02-20 Martin Liska <mliska@suse.cz>
6442 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6443 possible values if we don't have a hint.
6445 2018-02-20 Martin Liska <mliska@suse.cz>
6449 * final.c (shorten_branches): Build align_tab array with one
6451 * opts.c (finish_options): Add alignment option limit check.
6452 (MAX_CODE_ALIGN): Likewise.
6453 (MAX_CODE_ALIGN_VALUE): Likewise.
6454 * doc/invoke.texi: Document maximum allowed option value for
6455 all -falign-* options.
6457 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6460 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6461 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6462 * var-tracking.c (emit_note_insn_var_location): Remove all references
6463 to NOTE_INSN_CALL_ARG_LOCATION.
6464 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6465 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6466 Use copy_rtx_if_shared.
6467 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6468 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6469 (dwarf2out_var_location): Remove handling of
6470 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6472 * final.c (final_scan_insn): Remove all references to
6473 NOTE_INSN_CALL_ARG_LOCATION.
6474 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6475 before dumping final insns.
6476 * except.c (emit_note_eh_region_end): Remove all references to
6477 NOTE_INSN_CALL_ARG_LOCATION.
6478 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6479 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6480 * config/arc/arc.c (hwloop_optimize): Likewise.
6481 * config/arm/arm.c (create_fix_barrier): Likewise.
6482 * config/s390/s390.c (s390_chunkify_start): Likewise.
6483 * config/sh/sh.c (find_barrier): Likewise.
6484 * config/i386/i386.c (rest_of_insert_endbranch,
6485 ix86_seh_fixup_eh_fallthru): Likewise.
6486 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6487 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6488 * config/frv/frv.c (frv_function_prologue): Likewise.
6489 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6491 (note_outside_basic_block_p): Remove all references to
6492 NOTE_INSN_CALL_ARG_LOCATION.
6493 * gengtype.c (adjust_field_rtx_def): Likewise.
6494 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6496 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6497 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6500 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6503 PR tree-optimization/84452
6504 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6505 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6508 2018-02-19 Martin Liska <mliska@suse.cz>
6511 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6513 2018-02-19 Martin Liska <mliska@suse.cz>
6514 Richard Sandiford <richard.sandiford@linaro.org>
6516 PR tree-optimization/82491
6517 * gimple-fold.c (get_base_constructor): Make earlier bail out
6520 2018-02-19 Carl Love <cel@us.ibm.com>
6522 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6523 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6525 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6526 P8V_BUILTIN_VEC_NEG.
6528 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6530 * config/rl78/rl78.md (movdf): New define expand.
6532 2018-02-19 Martin Liska <mliska@suse.cz>
6535 * doc/invoke.texi: Fix typo.
6536 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6538 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6540 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6541 handle rs6000_single_float and rs6000_double_float specially for
6544 2018-02-16 Jeff Law <law@redhat.com>
6546 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6549 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6552 * ipa-inline.c (inline_small_functions): Fix a typo.
6554 2018-02-16 Nathan Sidwell <nathan@acm.org>
6556 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6558 2018-02-16 Carl Love <cel@us.ibm.com>
6560 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6561 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6562 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6563 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6564 expansion to P8V_BUILTIN_VEC_FLOAT2.
6566 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6568 PR rtl-optimization/70023
6569 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6570 src_regno into account.
6572 2018-02-16 Carl Love <cel@us.ibm.com>
6574 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6575 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6576 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6577 * config/rs6000/rs6000.c: Remove case statements for
6578 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6579 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6580 and P9V_BUILTIN_VEC_VINSERT4B.
6581 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6582 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6583 * config/rs6000/vsx.md:
6584 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6587 2018-02-16 Carl Love <cel@us.ibm.com>
6589 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6591 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6593 * config/rs6000/rs6000-c.c: Add the definitions for
6594 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6595 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6596 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6597 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6598 definition for insert4b and define insn *insert3b_internal.
6599 * doc/extend.texi: Add documentation for vec_extract4b.
6601 2018-02-16 Nathan Sidwell <nathan@acm.org>
6603 * doc/extend.texi (Backwards Compatibility): Mention friend
6604 injection. Note for-scope is deprecated.
6605 * doc/invoke.texi (-ffriend-injection): Deprecate.
6607 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6609 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6610 that moved to I2, also allow destinations that are a paradoxical
6611 subreg (instead of a normal reg).
6613 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6616 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6619 2018-02-16 Richard Biener <rguenther@suse.de>
6621 PR tree-optimization/84037
6622 PR tree-optimization/84016
6624 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6625 Adjust vec_construct for the fact we need additional higher latency
6626 128bit inserts for AVX256 and AVX512 vector builds.
6627 (ix86_add_stmt_cost): Scale vector construction cost for
6630 2018-02-16 Richard Biener <rguenther@suse.de>
6632 PR tree-optimization/84417
6633 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6634 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6635 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6637 2018-02-16 Martin Liska <mliska@suse.cz>
6640 * internal-fn.def (ASAN_CHECK): Set proper flags.
6641 (ASAN_MARK): Likewise.
6643 2018-02-16 Julia Koval <julia.koval@intel.com>
6645 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6646 from PTA_CANNONLAKE.
6648 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6651 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6652 Use ++iter rather than iter++ for std::list iterators.
6653 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6654 defer deleting them until all nodes in the forest are processed. Do
6655 free even leaf nodes. Change to_process into auto_vec.
6658 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6659 * vec.h (vec_default_construct): Use memset instead of placement new
6660 if BROKEN_VALUE_INITIALIZATION is defined.
6661 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6662 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6665 PR rtl-optimization/83723
6666 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6667 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6668 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6670 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6672 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6674 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6676 PR rtl-optimization/81443
6677 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6678 from inner REGs to paradoxical SUBREGs.
6680 2018-02-16 Richard Biener <rguenther@suse.de>
6682 PR tree-optimization/84399
6683 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6684 For operands we can analyze at their definition make sure we can
6685 analyze them at each use as well.
6687 2018-02-16 Richard Biener <rguenther@suse.de>
6689 PR tree-optimization/84190
6690 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6691 volatile accesses if the decl isn't volatile.
6693 2018-02-15 Jason Merrill <jason@redhat.com>
6695 PR c++/84314 - ICE with templates and fastcall attribute.
6696 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6697 TYPE_CANONICAL on an existing type.
6699 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6701 PR tree-optimization/84383
6702 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6703 dstoff nor call operand_equal_p if dstbase is NULL.
6705 PR tree-optimization/84334
6706 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6707 also a CONSTANT_CLASS_P, punt.
6709 2018-02-14 Jim Wilson <jimw@sifive.com>
6711 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6712 first SMALL_OPERAND check. New local min_second_step. Move assert
6713 to where locals are set. Add TARGET_RVC support.
6714 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6716 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6718 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6720 2018-02-14 Martin Sebor <msebor@redhat.com>
6722 PR tree-optimization/83698
6723 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6724 arrays constrain the offset range to their bounds.
6725 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6726 (builtin_access::overlap): Avoid setting the size of overlap if it's
6728 (maybe_diag_overlap): Also consider arrays when deciding what values
6729 of offsets to include in diagnostics.
6731 2018-02-14 Martin Sebor <msebor@redhat.com>
6734 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6735 that correspond to the kind of a declaration.
6737 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6740 * config/pa/pa.md: Load address of PIC label using the linkage table
6741 if the label is nonlocal.
6743 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6745 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6746 warning message if user requests -maltivec=be.
6747 * doc/invoke.texi: Document deprecation of -maltivec=be.
6749 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6752 * config/rs6000/rs6000-c.c: Update definitions for
6753 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6754 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6756 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6759 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6760 add _get_ssp intrinsics. Remove argument from
6761 __builtin_ia32_rdssp[d|q].
6762 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6763 * config/i386/i386-builtin.def: Remove argument from
6764 __builtin_ia32_rdssp[d|q].
6765 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6766 ix86_expand_special_args_builtin for _rdssp[d|q].
6767 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6768 Clear register before usage.
6769 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6770 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6772 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6774 PR tree-optimization/84357
6775 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6776 operand 1 of an ARRAY_REF too.
6778 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6781 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6782 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6784 (set_of_reg): New struct.
6785 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6786 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6787 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6789 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6790 Split into bitclr, bitset, bitinvert patterns if appropriate.
6791 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6792 use rx_fuse_in_memory_bitop.
6793 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6794 to named insn, correct maximum insn length.
6796 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6799 * machmode.def: Define a complex mode for PARTIAL_INT.
6800 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6802 * doc/rtl.texi: Document CSPImode.
6803 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6805 (msp430_hard_regno_nregs_with_padding): Likewise.
6807 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6810 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6812 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6814 PR rtl-optimization/84169
6815 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6816 we generated a parallel as new i3 and we split that to new i2 and i3
6817 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6818 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6819 those to i2, not i1. Partially rewrite this scan code.
6821 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6824 * stor-layout.c (place_field): For variable length fields, adjust
6825 offset_align afterwards not just based on the field's alignment,
6826 but also on the size.
6829 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6830 of exps and logs in the use_exp2 case.
6832 2018-02-13 Jeff Law <law@redhat.com>
6834 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6837 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6840 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6844 * common.opt (gas-loc-support, gas-locview-support): New.
6845 (ginline-points, ginternal-reset-location-views): New.
6846 * doc/invoke.texi: Document them. Use @itemx where intended.
6847 (gvariable-location-views): Adjust.
6848 * target.def (reset_location_view): New.
6849 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6850 (TARGET_RESET_LOCATION_VIEW): New.
6851 * doc/tm.texi: Rebuilt.
6852 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6853 (dwarf2out_default_as_locview_support): New.
6854 (output_asm_line_debug_info): Use option variables.
6855 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6856 (output_loc_list): Likewise.
6857 (add_high_low_attributes): Check option variables.
6858 Don't output entry view attribute in strict mode.
6859 (gen_inlined_subroutine_die): Check option variables.
6860 (dwarf2out_inline_entry): Likewise.
6861 (init_sections_and_labels): Likewise.
6862 (dwarf2out_early_finish): Likewise.
6863 (maybe_reset_location_view): New, from...
6864 (dwarf2out_var_location): ... here. Call it.
6865 * debug.h (dwarf2out_default_as_loc_support): Declare.
6866 (dwarf2out_default_as_locview_support): Declare.
6867 * hooks.c (hook_int_rtx_insn_0): New.
6868 * hooks.h (hook_int_rtx_insn_0): Declare.
6869 * toplev.c (process_options): Take -gas-loc-support and
6870 -gas-locview-support from dwarf2out. Enable
6871 -gvariable-location-views by default only with locview
6872 assembler support. Enable -ginternal-reset-location-views by
6873 default only if the target defines the corresponding hook.
6874 Enable -ginline-points by default if location views are
6875 enabled; force it disabled if statement frontiers are
6877 * tree-inline.c (expand_call_inline): Check option variables.
6878 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6880 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6882 PR tree-optimization/84321
6883 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6884 handling. Also check whether the anti-range contains any values
6885 that satisfy the mask; switch to a VR_RANGE if not.
6887 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6890 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6892 2018-02-13 Martin Jambor <mjambor@suse.cz>
6895 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6896 of call statements, also set location of a load to a temporary.
6898 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6900 * config/rl78/rl78.c (add_vector_labels): New function.
6901 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6902 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6903 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6904 which checks that no arguments are passed.
6905 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6906 * doc/extend.texi: Documentation for the new attribute.
6908 2018-02-13 Andreas Schwab <schwab@suse.de>
6910 * config/riscv/linux.h (CPP_SPEC): Define.
6912 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6915 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6916 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6917 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6918 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6919 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6920 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6921 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6922 temporarily for AES and PCLMUL builtins.
6924 PR tree-optimization/84339
6925 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6926 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6930 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6931 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6932 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6934 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6936 * omp-simd-clone.h: New file.
6937 * omp-simd-clone.c: Include omp-simd-clone.h.
6938 (expand_simd_clones): No longer static.
6939 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6940 cgraph.h and omp-simd-clone.h.
6941 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6942 (vect_recog_widen_shift_pattern): Formatting fix.
6943 (vect_pattern_recog_1): Don't check optab for calls.
6946 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6947 operands[2] into a REG before using gen_lowpart on it.
6949 2018-02-12 Jeff Law <law@redhat.com>
6952 * config/sh/sh.c (find_barrier): Consider a sibling call
6955 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6956 successfully back substituting a reg.
6958 2018-02-12 Richard Biener <rguenther@suse.de>
6960 PR tree-optimization/84037
6961 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6962 parameter, move visited init to caller.
6963 (vect_slp_analyze_operations): Separate cost from validity
6964 check, initialize visited once for all instances.
6965 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6967 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6969 (vect_model_promotion_demotion_cost): Likewise.
6970 (vectorizable_bswap): Guard cost modeling with !slp_node
6971 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6973 (vectorizable_call): Likewise.
6974 (vectorizable_conversion): Likewise.
6975 (vectorizable_assignment): Likewise.
6976 (vectorizable_shift): Likewise.
6977 (vectorizable_operation): Likewise.
6978 (vectorizable_store): Likewise.
6979 (vectorizable_load): Likewise.
6980 (vectorizable_condition): Likewise.
6981 (vectorizable_comparison): Likewise.
6983 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6986 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6987 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6988 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6990 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6993 * params.def (inline-min-speedup): Increase from 8 to 15.
6994 (max-inline-insns-auto): Decrease from 40 to 30.
6995 * ipa-split.c (consider_split): Add some buffer for function to
6996 be considered inlining candidate.
6997 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7000 2018-02-12 Richard Biener <rguenther@suse.de>
7002 PR tree-optimization/84037
7003 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7004 matched stmts if we cannot swap the non-matched ones.
7006 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7008 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7009 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7010 _mm_maskz_scalef_round_ss): New intrinsics.
7011 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7012 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7013 __builtin_ia32_scalefss_round): Remove.
7014 (__builtin_ia32_scalefsd_mask_round,
7015 __builtin_ia32_scalefss_mask_round): New intrinsics.
7016 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7017 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7018 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7019 "<round_constraint>")): Changed to ...
7020 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7021 "<round_scalar_constraint>")): ... this.
7022 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7023 %0, %1, %2<round_op3>}"): Changed to ...
7024 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7025 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7026 %2<round_scalar_mask_op3>}"): ... this.
7027 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7029 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7031 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7032 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7033 (_mm_maskz_sqrt_round_ss): New intrinsics.
7034 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7035 (__builtin_ia32_sqrtsd_mask_round)
7036 (__builtin_ia32_sqrtss_mask_round): New builtins.
7037 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7038 (__builtin_ia32_sqrtss_round): Remove.
7039 (__builtin_ia32_sqrtsd_mask_round)
7040 (__builtin_ia32_sqrtss_mask_round): New builtins.
7041 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7042 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7043 ((match_operand:VF_128 1 "vector_operand"
7044 "xBm,<round_constraint>")): Changed to ...
7045 ((match_operand:VF_128 1 "vector_operand"
7046 "xBm,<round_scalar_constraint>")): ... this.
7047 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7048 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7049 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7050 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7051 %<iptr>1<round_scalar_mask_op3>}): ... this.
7052 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7053 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7055 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7058 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7059 Cast vec_cmpeq result to correct type.
7060 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7061 Cast vec_cmpgt result to correct type.
7063 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7065 * final.c (final_scan_insn_1): Renamed from...
7066 (final_scan_insn): ... this. New wrapper, to recover
7067 seen from the outermost call in recursive ones.
7068 * config/sparc/sparc.c (output_return): Drop seen from call.
7069 (output_sibcall): Likewise.
7070 * config/visium/visium.c (output_branch): Likewise.
7072 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7074 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7077 2018-02-10 Alan Modra <amodra@gmail.com>
7080 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7081 Specify LR as an input.
7083 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7086 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7087 remove_member_access_dummy_vars): New functions.
7088 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7089 lower_omp_1, execute_lower_omp): Use them.
7091 PR rtl-optimization/84308
7092 * shrink-wrap.c (spread_components): Release todo vector.
7094 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7096 PR rtl-optimization/57193
7097 * ira-color.c (struct allocno_color_data): Add member
7098 conflict_allocno_hard_prefs.
7099 (update_conflict_allocno_hard_prefs): New.
7100 (bucket_allocno_compare_func): Add a preference based on
7101 conflict_allocno_hard_prefs.
7102 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7103 (color_allocnos): Remove a dead code. Initiate
7104 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7106 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7109 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7110 constraint from =wa to wa. Avoid a subreg on the output operand,
7111 instead use a pseudo and subreg it in a move.
7112 (p9_xxbrd_<mode>): Changed to ...
7113 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7114 (p9_xxbrd_v2df): New expander.
7115 (p9_xxbrw_<mode>): Changed to ...
7116 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7117 (p9_xxbrw_v4sf): New expander.
7119 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7121 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7123 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7126 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7127 multiply in 32-bit mode.
7128 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7129 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7132 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7134 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7135 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7136 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7137 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7139 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7142 * dwarf2out.c (is_trivial_indirect_ref): New function.
7143 (dwarf2out_late_global_decl): Do not generate a location
7144 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7145 and that are not defined in the current unit.
7147 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7149 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7150 instead of a libcall for UNORDERED.
7152 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7155 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7156 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7158 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7161 * config/s390/s390.c (s390_set_current_function): Invoke
7162 s390_indirect_branch_settings also if fndecl didn't change.
7164 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7166 * config/rs6000/rs6000.md (blockage): Set length to zero.
7168 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7170 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7172 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7175 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7176 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7180 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7181 PARALLEL incoming that failed vt_get_decl_and_offset check.
7184 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7185 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7187 (get_variable_section): For decls in named .bss* sections pass true as
7188 second argument to bss_initializer_p.
7190 2018-02-09 Marek Polacek <polacek@redhat.com>
7191 Jakub Jelinek <jakub@redhat.com>
7194 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7195 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7196 Sync some changes from cxx_fold_indirect_ref.
7198 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7200 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7202 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7203 (BLOCK_INLINE_ENTRY_LABEL): New.
7204 (dwarf2out_var_location): Disregard inline entry markers.
7205 (inline_entry_data): New struct.
7206 (inline_entry_data_hasher): New hashtable type.
7207 (inline_entry_data_hasher::hash): New.
7208 (inline_entry_data_hasher::equal): New.
7209 (inline_entry_data_table): New variable.
7210 (add_high_low_attributes): Add DW_AT_entry_pc and
7211 DW_AT_GNU_entry_view attributes if a pending entry is found
7212 in inline_entry_data_table. Add old entry_pc attribute only
7213 if debug nonbinding markers are disabled.
7214 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7215 markers are enabled.
7216 (block_within_block_p, dwarf2out_inline_entry): New.
7217 (dwarf2out_finish): Check that no entries remained in
7218 inline_entry_data_table.
7219 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7220 (final_scan_insn, notice_source_line): Likewise.
7221 (rest_of_clean_state): Skip inline entry markers.
7222 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7224 * gimple.c (gimple_build_debug_inline_entry): New.
7225 * gimple.h (enum gimple_debug_subcode): Add
7226 GIMPLE_DEBUG_INLINE_ENTRY.
7227 (gimple_build_debug_inline_entry): Declare.
7228 (gimple_debug_inline_entry_p): New.
7229 (gimple_debug_nonbind_marker_p): Adjust.
7230 * insn-notes.def (INLINE_ENTRY): New.
7231 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7232 inline entry marker notes.
7233 (print_insn): Likewise.
7234 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7235 (INSN_DEBUG_MARKER_KIND): Likewise.
7236 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7237 * tree-inline.c (expand_call_inline): Build and insert
7238 debug_inline_entry stmt.
7239 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7240 inline entry blocks early, if nonbind markers are enabled.
7241 (dump_scope_block): Dump fragment info.
7242 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7243 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7244 (gimple_build_debug_inline_entry): New.
7245 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7246 Enable/disable inline entry points too.
7247 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7248 (DEBUG_INSN): Describe inline entry markers.
7250 * common.opt (gvariable-location-views): New.
7251 (gvariable-location-views=incompat5): New.
7252 * config.in: Rebuilt.
7253 * configure: Rebuilt.
7254 * configure.ac: Test assembler for view support.
7255 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7256 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7257 * dwarf2out.c (var_loc_view): New typedef.
7258 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7259 (dwarf2out_locviews_in_attribute): New.
7260 (dwarf2out_locviews_in_loclist): New.
7261 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7262 (enum dw_line_info_opcode): Add LI_adv_address.
7263 (struct dw_line_info_table): Add view.
7264 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7265 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7266 (zero_view_p): New variable.
7267 (ZERO_VIEW_P): New macro.
7268 (output_asm_line_debug_info): New.
7269 (struct var_loc_node): Add view.
7270 (add_AT_view_list, AT_loc_list): New.
7271 (add_var_loc_to_decl): Add view param. Test it against last.
7272 (new_loc_list): Add view params. Record them.
7273 (AT_loc_list_ptr): Handle loc and view lists.
7274 (view_list_to_loc_list_val_node): New.
7275 (print_dw_val): Handle dw_val_class_view_list.
7276 (size_of_die): Likewise.
7277 (value_format): Likewise.
7278 (loc_list_has_views): New.
7279 (gen_llsym): Set vl_symbol too.
7280 (maybe_gen_llsym, skip_loc_list_entry): New.
7281 (dwarf2out_maybe_output_loclist_view_pair): New.
7282 (output_loc_list): Output view list or entries too.
7283 (output_view_list_offset): New.
7284 (output_die): Handle dw_val_class_view_list.
7285 (output_dwarf_version): New.
7286 (output_compilation_unit_header): Use it.
7287 (output_skeleton_debug_sections): Likewise.
7288 (output_rnglists, output_line_info): Likewise.
7289 (output_pubnames, output_aranges): Update version comments.
7290 (output_one_line_info_table): Output view numbers in asm comments.
7291 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7292 Call maybe_gen_llsym.
7293 (loc_list_from_tree_1): Adjust.
7294 (add_AT_location_description): Create view list attribute if
7295 needed, check it's absent otherwise.
7296 (convert_cfa_to_fb_loc_list): Adjust.
7297 (maybe_emit_file): Call output_asm_line_debug_info for test.
7298 (dwarf2out_var_location): Reset views as needed. Precompute
7299 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7300 attribute. Set view.
7301 (new_line_info_table): Reset next view.
7302 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7303 (dwarf2out_source_line): Likewise. Output view resets and labels to
7304 the assembler, or select appropriate line info opcodes.
7305 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7306 (optimize_string_length): Catch it. Adjust.
7307 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7308 dw_val_class_view_list, and remove it if no longer needed.
7309 (hash_loc_list): Hash view numbers.
7310 (loc_list_hasher::equal): Compare them.
7311 (optimize_location_lists): Check whether a view list symbol is
7312 needed, and whether the locview attribute is present, and
7313 whether they match. Remove the locview attribute if no longer
7315 (index_location_lists): Call skip_loc_list_entry for test.
7316 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7317 Use output_dwarf_version.
7318 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7319 (struct dw_val_node): Add val_view_list.
7320 * final.c (SEEN_NEXT_VIEW): New.
7321 (set_next_view_needed): New.
7322 (clear_next_view_needed): New.
7323 (maybe_output_next_view): New.
7324 (final_start_function): Rename to...
7325 (final_start_function_1): ... this. Take pointer to FIRST,
7326 add SEEN parameter. Emit param bindings in the initial view.
7327 (final_start_function): Reintroduce SEEN-less interface.
7328 (final): Rename to...
7329 (final_1): ... this. Take SEEN parameter. Output final pending
7330 next view at the end.
7331 (final): Reintroduce seen-less interface.
7332 (final_scan_insn): Output pending next view before switching
7333 sections or ending a block. Mark the next view as needed when
7334 outputting variable locations. Notify debug backend of section
7335 changes, and of location view changes.
7336 (rest_of_handle_final): Adjust.
7337 * toplev.c (process_options): Autodetect value for debug variable
7338 location views option. Warn on incompat5 without -gdwarf-5.
7339 * doc/invoke.texi (gvariable-location-views): New.
7340 (gvariable-location-views=incompat5): New.
7341 (gno-variable-location-views): New.
7343 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7345 PR tree-optimization/84136
7346 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7347 that the result of find_edge is non-NULL.
7349 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7352 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7353 storing integer register in SImode. Fix cost of 256 and 512
7354 byte aligned SSE register store.
7356 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7358 * config/i386/i386.c (ix86_multiplication_cost): Fix
7359 multiplication cost for TARGET_AVX512DQ.
7361 2018-02-08 Marek Polacek <polacek@redhat.com>
7363 PR tree-optimization/84238
7364 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7367 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7369 PR tree-optimization/84265
7370 * tree-vect-stmts.c (vectorizable_store): Don't treat
7371 VMAT_CONTIGUOUS accesses as grouped.
7372 (vectorizable_load): Likewise.
7374 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7376 PR tree-optimization/81635
7377 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7378 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7379 (test_round_for_mask): New functions.
7380 (wide_int_cc_tests): Call test_round_for_mask.
7381 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7382 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7383 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7384 range returned by get_range_info.
7386 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7389 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7390 * symtab.c: Include builtins.h
7391 (symtab_node::output_to_lto_symbol_table_p): Move here
7392 from lto-streamer-out.c:output_symbol_p.
7393 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7394 (output_symbol_p): Move all logic to symtab.c
7395 (produce_symtab): Update.
7397 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7399 * config/s390/s390-opts.h (enum indirect_branch): Define.
7400 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7401 (s390_indirect_branch_via_thunk)
7402 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7403 (enum s390_indirect_branch_type): Define.
7404 * config/s390/s390.c (struct s390_frame_layout, struct
7405 machine_function): Remove.
7406 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7407 (indirect_branch_table_label_no, indirect_branch_table_name):
7409 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7410 (enum s390_indirect_branch_option): Define.
7411 (s390_return_addr_from_memory): New function.
7412 (s390_handle_string_attribute): New function.
7413 (s390_attribute_table): Add new attribute handler.
7414 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7415 (s390_indirect_branch_via_thunk): New function.
7416 (s390_indirect_branch_via_inline_thunk): New function.
7417 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7418 sibling call optimization for non z10 compiles.
7419 (s390_emit_call): Force indirect branch target to be a single
7420 register. Add r1 clobber for non-z10 compiles.
7421 (s390_emit_epilogue): Emit return jump via return_use expander.
7422 (s390_reorg): Handle JUMP_INSNs as execute targets.
7423 (s390_option_override_internal): Perform validity checks for the
7424 new command line options.
7425 (s390_indirect_branch_attrvalue): New function.
7426 (s390_indirect_branch_settings): New function.
7427 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7428 (s390_output_indirect_thunk_function): New function.
7429 (s390_code_end): Implement target hook.
7430 (s390_case_values_threshold): Implement target hook.
7431 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7433 * config/s390/s390.h (struct s390_frame_layout)
7434 (struct machine_function): Move here from s390.c.
7435 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7436 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7437 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7438 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7439 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7440 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7441 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7442 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7443 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7444 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7445 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7446 (mnemonic attribute): Add values which aren't recognized
7448 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7449 pattern for branch conversion. Fix mnemonic attribute.
7450 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7451 indirect branch via thunk if requested.
7452 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7453 ("*indirect_jump"): Disable for branch conversion using out of
7455 ("indirect_jump_via_thunk<mode>_z10")
7456 ("indirect_jump_via_thunk<mode>")
7457 ("indirect_jump_via_inlinethunk<mode>_z10")
7458 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7459 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7460 ("casesi_jump_via_inlinethunk<mode>_z10")
7461 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7462 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7463 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7464 ("*indirect2_jump"): Disable for branch conversion.
7465 ("casesi_jump"): Turn into expander and expand patterns for branch
7467 ("return_use"): New expander.
7468 ("*return"): Emit return via thunk and rename it to ...
7469 ("*return<mode>"): ... this one.
7470 * config/s390/s390.opt: Add new options and and enum for the
7473 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7475 * lra-constraints.c (match_reload): Unconditionally use
7476 gen_lowpart_SUBREG, rather than selecting between that
7477 and equivalent gen_rtx_SUBREG code.
7479 2018-02-08 Richard Biener <rguenther@suse.de>
7481 PR tree-optimization/84233
7482 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7483 changed flag instead of boguously re-using phi_inserted.
7485 2018-02-08 Martin Jambor <mjambor@suse.cz>
7487 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7488 static local variables.
7490 2018-02-08 Richard Biener <rguenther@suse.de>
7492 PR tree-optimization/84278
7493 * tree-vect-stmts.c (vectorizable_store): When looking for
7494 smaller vector types to perform grouped strided loads/stores
7495 make sure the mode is supported by the target.
7496 (vectorizable_load): Likewise.
7498 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7500 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7501 Increase LDP/STP opportunities by adding adjacent callee-saves.
7503 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7505 PR rtl-optimization/84068
7506 PR rtl-optimization/83459
7507 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7509 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7511 PR tree-optimization/84224
7512 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7513 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7516 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7519 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7520 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7521 count, remove one USE.
7523 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7525 * doc/install.texi (Configuration): Document the
7526 --with-long-double-format={ibm,ieee} PowerPC configuration
7530 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7531 Convert from define_expand to be define_insn_and_split. Rework
7532 float/double/_Float128 conversions to QI/HI/SImode to work with
7533 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7534 conversions to QI/HImode types did a store and then a load to
7535 truncate the value. For conversions to VSX registers, don't split
7536 the insn, instead emit the code directly. Use the code iterator
7537 any_fix to combine signed and unsigned conversions.
7538 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7539 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7540 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7541 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7542 (fix_<mode>di2_hw): Likewise.
7543 (fixuns_<mode>di2_hw): Likewise.
7544 (fix_<mode>si2_hw): Likewise.
7545 (fixuns_<mode>si2_hw): Likewise.
7546 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7547 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7548 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7549 fix<uns>_trunc<SFDF:mode>si2_p8.
7550 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7552 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7553 (fix<uns>_<mode>_mem): Likewise.
7554 (fctiw<u>z_<mode>_mem): Likewise.
7555 (fix<uns>_<mode>_mem): Likewise.
7556 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7557 the register allocator from doing a direct move to the GPRs to do
7558 a store, and instead use the ISA 3.0 store byte/half-word from
7559 vector register instruction. For IEEE 128-bit floating point,
7560 also optimize stores of 32-bit ints.
7561 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7563 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7565 * genextract.c (push_pathstr_operand): New function to support
7567 (walk_rtx): Call push_pathstr_operand.
7568 (print_path): Support [a-zA-Z].
7570 2018-02-07 Richard Biener <rguenther@suse.de>
7572 PR tree-optimization/84037
7573 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7574 (cse_and_gimplify_to_preheader): Declare.
7575 (vect_get_place_in_interleaving_chain): Likewise.
7576 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7578 (_loop_vec_info::~_loop_vec_info): Delete it.
7579 (cse_and_gimplify_to_preheader): New function.
7580 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7581 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7582 (vectorizable_load): Likewise. For grouped stores always base
7583 the IV on the first element.
7584 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7585 condition before gimplifying.
7587 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7589 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7590 *DIV_EXPR and *MOD_EXPR.
7592 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7595 * config/i386/i386.c (ix86_option_override_internal): Mask out
7596 the CF_SET bit when checking -fcf-protection.
7598 2018-02-07 Tom de Vries <tom@codesourcery.com>
7601 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7604 2018-02-07 Richard Biener <rguenther@suse.de>
7606 PR tree-optimization/84204
7607 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7610 PR tree-optimization/84205
7611 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7612 special-case isl_ast_op_zdiv_r.
7614 PR tree-optimization/84223
7615 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7616 Only add conditions from within the region.
7617 (gather_bbs::after_dom_children): Adjust.
7619 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7622 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7623 * config/avr/avr.md: Only post-reload split REG-REG moves if
7624 either register is GENERAL_REG_P.
7626 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7628 PR tree-optimization/84235
7629 * tree-ssa-scopedtables.c
7630 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7631 if the subtraction is performed in floating point type where NaNs are
7632 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7633 build 1. Formatting fix.
7635 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7638 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7639 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7640 and skip it regardless of bb boundaries. Use CALL_P macro,
7641 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7642 unnecessarily, formatting fix.
7644 2018-02-06 Michael Collison <michael.collison@arm.com>
7646 * config/arm/thumb2.md:
7647 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7648 (*thumb_mov_notscc): Ditto.
7650 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7653 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7654 unsigned_fix, not "s".
7656 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7658 * configure.ac (gcc_fn_eh_frame_ro): New function.
7659 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7660 correct .eh_frame permissions.
7661 * configure: Regenerate.
7663 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7665 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7668 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7670 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7671 Display warning message for -mno-speculate-indirect-jumps.
7673 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7675 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7677 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7679 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7681 PR tree-optimization/84225
7682 * tree-eh.c (find_trapping_overflow): Only call
7683 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7685 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7688 * config/i386/i386.c: Reimplement the check of possible options
7689 -mibt/-mshstk conbination. Change error messages.
7690 * doc/invoke.texi: Fix a typo: remove extra '='.
7692 2018-02-06 Marek Polacek <polacek@redhat.com>
7694 PR tree-optimization/84228
7695 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7697 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7700 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7701 emitted arch directives.
7702 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7703 __ARM_FEATURE_COPROC before changing architectures.
7705 2018-02-06 Richard Biener <rguenther@suse.de>
7707 * config/i386/i386.c (print_reg): Fix typo.
7708 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7710 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7712 * configure: Regenerate.
7714 2018-02-05 Martin Sebor <msebor@redhat.com>
7716 PR tree-optimization/83369
7717 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7720 2018-02-05 Martin Liska <mliska@suse.cz>
7722 * doc/invoke.texi: Cherry-pick upstream r323995.
7724 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7726 * ira.c (ira_init_register_move_cost): Adjust comment.
7728 2018-02-05 Martin Liska <mliska@suse.cz>
7730 PR gcov-profile/84137
7731 * doc/gcov.texi: Fix typo in documentation.
7733 2018-02-05 Martin Liska <mliska@suse.cz>
7735 PR gcov-profile/83879
7736 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7737 using dlopen functionality.
7739 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7741 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7742 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7743 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7744 _mm_maskz_range_round_ss): New intrinsics.
7745 (__builtin_ia32_rangesd128_round)
7746 (__builtin_ia32_rangess128_round): Remove.
7747 (__builtin_ia32_rangesd128_mask_round,
7748 __builtin_ia32_rangess128_mask_round): New builtins.
7749 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7750 __builtin_ia32_rangess128_round): Remove.
7751 (__builtin_ia32_rangesd128_mask_round,
7752 __builtin_ia32_rangess128_mask_round): New builtins.
7753 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7754 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7755 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7756 "<round_saeonly_constraint>")): Changed to ...
7757 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7758 "<round_saeonly_scalar_constraint>")): ... this.
7759 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7760 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7761 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7762 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7763 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7765 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7767 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7769 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7770 Remove all values except native, 8540 and 8548.
7772 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7774 * config/i386/i386.c (ix86_output_function_return): Pass
7775 INVALID_REGNUM, instead of -1, as invalid register number to
7776 indirect_thunk_name and output_indirect_thunk.
7778 2018-02-02 Julia Koval <julia.koval@intel.com>
7780 * config.gcc: Add -march=icelake.
7781 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7782 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7783 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7784 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7785 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7786 (processor_target_table): Add icelake.
7787 (ix86_option_override_internal): Handle new PTAs.
7788 (get_builtin_code_for_version): Handle icelake.
7789 (M_INTEL_COREI7_ICELAKE): New.
7790 (fold_builtin_cpu): Handle icelake.
7791 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7792 * doc/invoke.texi: Add -march=icelake.
7794 2018-02-02 Julia Koval <julia.koval@intel.com>
7796 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7797 to wide_int_bitmask.
7798 * wide-int-bitmask.h: New.
7800 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7803 * config/i386/i386.md: Replace Pmode with word_mode in
7804 builtin_setjmp_setup and builtin_longjmp to support x32.
7806 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7810 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7812 (rs6000_supported_cpu_names): New static variable.
7813 (linux_cpu_translation_table): Likewise.
7814 (elf_platform) <cpu>: Define new static variable and use it.
7815 Translate kernel AT_PLATFORM name to canonical name if needed.
7816 Error if platform name is unknown.
7818 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7821 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7823 2018-02-01 Jeff Law <law@redhat.com>
7826 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7827 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7828 the scratch if RELEASE_VIA_POP is false.
7829 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7830 If we have to save a temporary register, decrement SIZE appropriately.
7831 Pass new arguments to release_scratch_register_on_entry.
7832 (ix86_adjust_stack_and_probe): Likewise.
7833 (ix86_emit_probe_stack_range): Pass new arguments to
7834 release_scratch_register_on_entry.
7836 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7838 PR rtl-optimization/84157
7839 * combine.c (change_zero_ext): Use REG_P predicate in
7840 front of HARD_REGISTER_P predicate.
7842 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7844 * config/avr/avr.c (avr_option_override): Move disabling of
7845 -fdelete-null-pointer-checks to...
7846 * common/config/avr/avr-common.c (avr_option_optimization_table):
7849 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7851 PR tree-optimization/81635
7852 * tree-data-ref.c (split_constant_offset_1): For types that
7853 wrap on overflow, try to use range info to prove that wrapping
7856 2018-02-01 Renlin Li <renlin.li@arm.com>
7859 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7861 (aarch64_register_move_cost): Likewise.
7862 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7864 (REG_CLASS_NAMES): Likewise.
7865 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7866 TAILCALL_ADDR_REGS. Remove IP registers.
7867 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7869 2018-02-01 Richard Biener <rguenther@suse.de>
7871 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7872 for specifying RPO order and allow NULL for that.
7873 * domwalk.c (dom_walker::dom_walker): Likewise.
7874 (dom_walker::walk): Handle NULL RPO order.
7875 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7877 (rewrite_update_dom_walker): Likewise.
7878 (mark_def_dom_walker): Likewise.
7880 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7882 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7883 (aarch64_maybe_expand_sve_subreg_move): Declare.
7884 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7885 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7887 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7888 that are semantically a reverse operation.
7889 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7890 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7891 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7893 (aarch64_can_change_mode_class): For big-endian, forbid changes
7894 between two SVE modes if they have different element sizes.
7896 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7898 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7899 the TImode handling for big-endian targets.
7901 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7903 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7904 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7906 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7907 Remove BSWAP handing for big-endian targets and use the form of
7908 LD1RQ appropariate for the mode.
7910 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7912 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7913 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7916 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7919 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7920 check for operands that need to go through aarch64_sve_reload_be.
7922 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7924 PR tree-optimization/81661
7925 PR tree-optimization/84117
7926 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7927 * tree-eh.c: Include gimplify.h.
7928 (find_trapping_overflow, replace_trapping_overflow,
7929 rewrite_to_non_trapping_overflow): New functions.
7930 * tree-vect-loop.c: Include tree-eh.h.
7931 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7932 * tree-data-ref.c: Include tree-eh.h.
7933 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7935 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7937 PR rtl-optimization/84123
7938 * combine.c (change_zero_ext): Check if hard register satisfies
7939 can_change_dest_mode before calling gen_lowpart_SUBREG.
7941 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7944 * ira.c (ira_init_register_move_cost): Remove assert.
7946 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7948 PR rtl-optimization/84071
7949 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7950 * doc/tm.texi: Regenerate.
7952 2018-01-31 Richard Biener <rguenther@suse.de>
7954 PR tree-optimization/84132
7955 * tree-data-ref.c (analyze_miv_subscript): Properly
7956 check whether evolution_function_is_affine_multivariate_p
7957 before calling gcd_of_steps_may_divide_p.
7959 2018-01-31 Julia Koval <julia.koval@intel.com>
7962 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7963 * config/i386/i386.md (rdpid_rex64) New.
7964 (rdpid): Make 32bit only.
7966 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7969 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7970 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7972 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7975 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7977 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7979 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7981 PR rtl-optimization/84071
7982 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7983 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7985 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7987 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7988 (arc_attribute_table): Add 'aux' attribute.
7989 (arc_in_small_data_p): Consider aux like variables.
7990 (arc_is_aux_reg_p): New function.
7991 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7992 (arc_get_aux_arg): New function.
7993 (prepare_move_operands): Handle aux-register access.
7994 (arc_handle_aux_attribute): New function.
7995 * doc/extend.texi (ARC Variable attributes): Add subsection.
7997 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7999 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8000 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8001 (arc_attribute_table): Add 'uncached' attribute.
8002 (arc_print_operand): Print '.di' flag for uncached memory
8004 (arc_in_small_data_p): Do not consider for small data the uncached
8006 (arc_is_uncached_mem_p): New function.
8007 * config/arc/predicates.md (compact_store_memory_operand): Check
8008 for uncached memory accesses.
8009 (nonvol_nonimm_operand): Likewise.
8010 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8012 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8015 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8016 falign-loops=): Add Optimization flag.
8018 2018-01-30 Jeff Law <law@redhat.com>
8021 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8022 INT_REGISTERS_SAVED. Check it prior to calling
8023 get_scratch_register_on_entry.
8024 (ix86_adjust_stack_and_probe): Similarly.
8025 (ix86_emit_probe_stack_range): Similarly.
8026 (ix86_expand_prologue): Corresponding changes.
8028 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8031 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8032 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8034 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8037 * lra-constraints.c (curr_insn_transform): Process AND in the
8040 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8042 PR rtl-optimization/83986
8043 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8044 dependence against last_pending_memory_flush in addition to
8047 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8049 PR tree-optimization/81611
8050 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8053 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8056 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8059 2018-01-30 Richard Biener <rguenther@suse.de>
8060 Jakub Jelinek <jakub@redhat.com>
8062 PR tree-optimization/84111
8063 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8064 inner loops added during recursion, as they don't have up-to-date
8067 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8070 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8071 (can_inline_edge_by_limits_p): ... here.
8072 (can_early_inline_edge_p, check_callers,
8073 update_caller_keys, update_callee_keys, recursive_inlining,
8074 add_new_edges_to_heap, speculation_useful_p,
8075 inline_small_functions,
8076 inline_small_functions, flatten_function,
8077 inline_to_all_callers_1): Update.
8079 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8081 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8084 2018-01-30 Richard Biener <rguenther@suse.de>
8086 PR tree-optimization/83008
8087 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8088 invariant and constant vector uses in stmts when they need
8091 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8094 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8095 * configure: Regenerate.
8097 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8099 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8101 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8102 Use gen_rtx_REG rather than gen_lowpart.
8104 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8106 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8107 rather than 0 when creating partial subregs.
8109 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8111 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8114 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8117 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8118 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8119 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8120 flags. This restores the settings used before the 2017-07-24.
8121 Turning off pre increment/decrement/modify allows IVOPTS to
8122 optimize DF/SF loops where the index is an int.
8124 2018-01-29 Richard Biener <rguenther@suse.de>
8125 Kelvin Nilsen <kelvin@gcc.gnu.org>
8128 * tree-vect-stmts.c (vectorizable_call): Don't call
8129 targetm.vectorize_builtin_md_vectorized_function if callee is
8132 2018-01-22 Carl Love <cel@us.ibm.com>
8134 * doc/extend.tex: Fix typo in second arg in
8135 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8137 2018-01-29 Richard Biener <rguenther@suse.de>
8139 PR tree-optimization/84086
8140 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8141 (flush_ssaname_freelist): When SSA names were released reset
8142 the SCEV hash table.
8144 2018-01-29 Richard Biener <rguenther@suse.de>
8146 PR tree-optimization/84057
8147 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8148 removed paths when removing edges.
8150 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8152 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8153 -mfunction-return=@var{choice}.
8155 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8158 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8160 (layout::print_source_line): Likewise.
8161 (test_get_line_width_without_trailing_whitespace): Add test cases.
8163 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8166 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8169 2018-01-26 Jim Wilson <jimw@sifive.com>
8171 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8173 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8176 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8178 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8179 and CMP + SUB-immediate -> SUBS.
8181 2018-01-26 Martin Sebor <msebor@redhat.com>
8183 PR tree-optimization/83896
8184 * tree-ssa-strlen.c (get_string_len): Rename...
8185 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8186 Avoid assuming length is constant.
8187 (handle_char_store): Use HOST_WIDE_INT for string length.
8189 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8192 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8193 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8195 2018-01-26 Richard Biener <rguenther@suse.de>
8197 PR rtl-optimization/84003
8198 * dse.c (record_store): Only record redundant stores when
8199 the earlier store aliases at least all accesses the later one does.
8201 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8203 PR rtl-optimization/83985
8204 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8205 REG_CFA_RESTORE insns.
8206 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8207 REG_CFA_RESTORE insns here.
8210 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8211 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8213 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8215 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8216 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8217 (arc_init): Likewise.
8218 (arc_override_options): Likewise.
8219 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8221 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8223 * config/arc/arc.h (TARGET_DBNZ): Define.
8224 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8225 properly set the tune attribute.
8226 (dbnz): Use TARGET_DBNZ guard.
8227 * config/arc/arc.opt (mtune): Add core3 option.
8229 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8231 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8232 recognize new pic like addresses.
8233 (arc_delegitimize_address): Clean up.
8235 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8237 * config/arc/arc-arches.def: Option mrf16 valid for all
8239 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8240 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8241 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8242 * config/arc/arc-tables.opt: Regenerate.
8243 * config/arc/arc.c (arc_conditional_register_usage): Handle
8244 reduced register file case.
8245 (arc_file_start): Set must have build attributes.
8246 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8248 * config/arc/arc.opt (mrf16): Add new option.
8249 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8250 * config/arc/genmultilib.awk: Handle new mrf16 option.
8251 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8252 * config/arc/t-multilib: Regenerate.
8253 * doc/invoke.texi (ARC Options): Document mrf16 option.
8255 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8257 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8258 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8259 (arc_attribute_table): Add 'secure_call' attribute.
8260 (arc_print_operand): Print secure call operand.
8261 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8263 (arc_is_secure_call_p): New function. * config/arc/arc.md
8264 (call_i): Add support for sjli instruction.
8265 (call_value_i): Likewise.
8266 * config/arc/constraints.md (Csc): New constraint.
8268 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8269 John Eric Martin <John.Martin@emmicro-us.com>
8271 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8272 * config/arc/arc.c (_arc_jli_section): New struct.
8273 (arc_jli_section): New type.
8274 (rc_jli_sections): New static variable.
8275 (arc_handle_jli_attribute): New function.
8276 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8277 (arc_file_end): New function.
8278 (TARGET_ASM_FILE_END): Define.
8279 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8280 (arc_add_jli_section): New function.
8281 (jli_call_scan): Likewise.
8282 (arc_reorg): Call jli_call_scan.
8283 (arc_output_addsi): Remove 'S' from printing asm operand.
8284 (arc_is_jli_call_p): New function.
8285 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8287 (movhi_insn): Likewise.
8288 (movsi_insn): Likewise.
8289 (movsi_set_cc_insn): Likewise.
8290 (loadqi_update): Likewise.
8291 (load_zeroextendqisi_update): Likewise.
8292 (load_signextendqisi_update): Likewise.
8293 (loadhi_update): Likewise.
8294 (load_zeroextendhisi_update): Likewise.
8295 (load_signextendhisi_update): Likewise.
8296 (loadsi_update): Likewise.
8297 (loadsf_update): Likewise.
8298 (movsicc_insn): Likewise.
8299 (bset_insn): Likewise.
8300 (bxor_insn): Likewise.
8301 (bclr_insn): Likewise.
8302 (bmsk_insn): Likewise.
8303 (bicsi3_insn): Likewise.
8304 (cmpsi_cc_c_insn): Likewise.
8305 (movsi_ne): Likewise.
8306 (movsi_cond_exec): Likewise.
8307 (clrsbsi2): Likewise.
8318 (call_i): Remove 'S' asm letter, add jli instruction.
8319 (call_value_i): Likewise.
8320 * config/arc/arc.op (mjli-always): New option.
8321 * config/arc/constraints.md (Cji): New constraint.
8322 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8324 (subsf3_fpx): Likewise.
8325 (mulsf3_fpx): Likewise.
8326 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8328 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8330 * doc/invoke.texi (ARC): Document mjli-always option.
8332 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8334 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8335 avoid addition with 0 and use incw and decw where possible.
8337 2018-01-26 Richard Biener <rguenther@suse.de>
8339 PR tree-optimization/81082
8340 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8341 association if it requires casting to unsigned.
8342 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8343 from fold_plusminus_mult_expr to catch important cases late when
8344 range info is available.
8346 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8348 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8349 * configure.ac (hidden_linkonce): New test.
8350 * configure: Regenerate.
8351 * config.in: Regenerate.
8353 2018-01-26 Julia Koval <julia.koval@intel.com>
8355 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8356 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8357 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8358 _mm_mask_bitshuffle_epi64_mask): Fix type.
8359 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8360 USI_FTYPE_V4DI_V4DI_USI): Remove.
8361 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8362 __builtin_ia32_vpshufbitqmb256_mask,
8363 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8364 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8365 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8367 2018-01-26 Alan Modra <amodra@gmail.com>
8370 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8371 UNSPEC_VBPERMQ. Sort other unspecs.
8373 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8375 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8377 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8380 * predict.c (drop_profile): Do not push/pop cfun; update also
8382 (handle_missing_profiles): Fix logic looking for zero profiles.
8384 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8387 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8388 on functions with #pragma omp declare simd or functions with simd
8390 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8391 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8392 Remove trailing \n from warning_at calls.
8394 2018-01-25 Tom de Vries <tom@codesourcery.com>
8397 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8398 for neutered workers.
8400 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8403 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8404 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8406 2018-01-24 Jeff Law <law@redhat.com>
8409 * i386.c (get_probe_interval): Move to earlier point.
8410 (ix86_compute_frame_layout): If -fstack-clash-protection and
8411 the frame is larger than the probe interval, then use pushes
8412 to save registers rather than reg->mem moves.
8413 (ix86_expand_prologue): Remove conditional for int_registers_saved
8416 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8419 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8420 min/max for never referenced object.
8422 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8425 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8427 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8428 attributes from DECL_ATTRIBUTES (decl) without affecting
8429 DECL_ATTRIBUTES (current_function_decl).
8430 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8431 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8433 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8435 PR tree-optimization/83979
8436 * fold-const.c (fold_comparison): Use constant_boolean_node
8437 instead of boolean_{true,false}_node.
8439 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8441 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8444 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8446 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8447 Simplify the clause that sets the length attribute.
8448 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8449 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8450 clause that sets the length attribute.
8451 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8453 2018-01-24 Tom de Vries <tom@codesourcery.com>
8456 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8457 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8458 Add strict parameter.
8459 (prevent_branch_around_nothing): Insert dummy insn between branch to
8460 label and label with no ptx insn inbetween.
8461 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8463 2018-01-24 Tom de Vries <tom@codesourcery.com>
8466 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8467 for neutered threads in warp.
8468 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8470 2018-01-24 Richard Biener <rguenther@suse.de>
8472 PR tree-optimization/83176
8473 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8476 2018-01-24 Richard Biener <rguenther@suse.de>
8478 PR tree-optimization/82819
8479 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8480 code generating pluses that are no-ops in the target precision.
8482 2018-01-24 Richard Biener <rguenther@suse.de>
8485 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8487 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8489 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8490 to merge probabilities.
8491 * predict.c (probably_never_executed): Also mark as cold functions
8492 with global 0 profile and guessed local profile.
8493 * profile-count.c (profile_probability::combine_with_count): New
8495 * profile-count.h (profile_probability::operator*,
8496 profile_probability::operator*=, profile_probability::operator/,
8497 profile_probability::operator/=): Reduce precision to adjusted
8498 and set value to guessed on contradictory divisions.
8499 (profile_probability::combine_with_freq): Remove.
8500 (profile_probability::combine_wiht_count): Declare.
8501 (profile_count::force_nonzero):: Set to adjusted.
8502 (profile_count::probability_in):: Set quality to adjusted.
8503 * tree-ssa-tail-merge.c (replace_block_by): Use
8506 2018-01-23 Andrew Waterman <andrew@sifive.com>
8507 Jim Wilson <jimw@sifive.com>
8509 * config/riscv/riscv.c (riscv_stack_boundary): New.
8510 (riscv_option_override): Set riscv_stack_boundary. Handle
8511 riscv_preferred_stack_boundary_arg.
8512 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8513 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8514 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8515 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8516 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8517 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8519 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8522 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8523 of struct ix86_frame.
8524 (ix86_expand_epilogue): Likewise. Add a local variable for
8525 the reg_save_offset field in struct ix86_frame.
8527 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8529 PR tree-optimization/82604
8530 * tree-loop-distribution.c (enum partition_kind): New enum item
8531 PKIND_PARTIAL_MEMSET.
8532 (partition_builtin_p): Support above new enum item.
8533 (generate_code_for_partition): Ditto.
8534 (compute_access_range): Differentiate cases that equality can be
8535 proven at all loops, the innermost loops or no loops.
8536 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8537 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8538 (finalize_partitions, distribute_loop): Don't fuse partition of
8539 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8540 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8543 2018-01-23 Martin Liska <mliska@suse.cz>
8545 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8546 order to ignore the predictor.
8547 (PRED_POLYMORPHIC_CALL): Likewise.
8548 (PRED_RECURSIVE_CALL): Likewise.
8550 2018-01-23 Martin Liska <mliska@suse.cz>
8552 * tree-profile.c (tree_profiling): Print function header to
8553 aware reader which function we are working on.
8554 * value-prof.c (gimple_find_values_to_profile): Do not print
8555 not interesting value histograms.
8557 2018-01-23 Martin Liska <mliska@suse.cz>
8559 * profile-count.h (enum profile_quality): Add
8560 profile_uninitialized as the first value. Do not number values
8561 as they are zero based.
8562 (profile_count::verify): Update sanity check.
8563 (profile_probability::verify): Likewise.
8565 2018-01-23 Nathan Sidwell <nathan@acm.org>
8567 * doc/invoke.texi (ffor-scope): Deprecate.
8569 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8571 PR tree-optimization/83510
8572 * domwalk.c (set_all_edges_as_executable): New function.
8573 (dom_walker::dom_walker): Convert bool param
8574 "skip_unreachable_blocks" to enum reachability. Move setup of
8575 edge flags to set_all_edges_as_executable and only do it when
8576 reachability is REACHABLE_BLOCKS.
8577 * domwalk.h (enum dom_walker::reachability): New enum.
8578 (dom_walker::dom_walker): Convert bool param
8579 "skip_unreachable_blocks" to enum reachability.
8580 (set_all_edges_as_executable): New decl.
8581 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8582 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8584 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8585 but converting true to REACHABLE_BLOCKS.
8586 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8588 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8589 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8590 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8592 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8593 if check_all_array_refs will be called.
8595 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8597 * tree.c (selftest::test_location_wrappers): Add more test
8600 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8602 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8603 (selftest::test_bit_in_range): Likewise.
8605 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8608 * doc/sourcebuild.texi (vect_float): Say that the selector
8609 only describes the situation when -funsafe-math-optimizations is on.
8610 (vect_float_strict): Document.
8612 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8614 PR tree-optimization/83965
8615 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8616 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8617 instead of checking only for a reduction.
8618 (vect_recog_widen_sum_pattern): Likewise.
8620 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8622 * predict.c (probably_never_executed): Only use precise profile info.
8623 (compute_function_frequency): Skip after inlining hack since we now
8624 have quality checking.
8626 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8628 * profile-count.h (profile_probability::very_unlikely,
8629 profile_probability::unlikely, profile_probability::even): Set
8630 precision to guessed.
8632 2018-01-23 Richard Biener <rguenther@suse.de>
8634 PR tree-optimization/83963
8635 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8636 Properly terminate dominator walk when crossing the exit edge not
8637 when visiting its source block.
8639 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8642 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8643 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8645 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8647 PR tree-optimization/83957
8648 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8649 semicolon after for body surrounded by braces.
8651 PR tree-optimization/83081
8652 * profile-count.h (profile_probability::split): New method.
8653 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8654 Use profile_probability::split.
8655 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8656 when splitting a single conditional jump into 2.
8658 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8660 PR tree-optimization/69452
8661 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8664 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8666 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8667 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8668 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8670 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8672 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8673 * config/rl78/rl78.md (movdi): New define_expand.
8674 * config/rl78/rl78.c (rl78_split_movdi): New function.
8676 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8679 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8681 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8682 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8683 128-bit to produce an UNSPEC move to get the double word with the
8684 signbit and then a shift directly to do signbit.
8685 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8686 implementation with a new version that just does either a direct
8687 move or a regular move. Move memory interface to separate insns.
8688 Move insns so they are next to the expander.
8689 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8690 with signbit move. Split big and little endian case.
8691 (signbit<mode>2_dm_mem_le): Likewise.
8692 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8693 (signbit<mode>2_dm2): Likewise.
8695 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8697 * config/rl78/rl78.md (anddi3): New define_expand.
8699 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8701 * config/rl78/rl78.md (umindi3): New define_expand.
8703 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8705 * config/rl78/rl78.md (smindi3): New define_expand.
8707 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8709 * config/rl78/rl78.md (smaxdi3): New define_expand.
8711 2018-01-22 Carl Love <cel@us.ibm.com>
8713 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8714 LVX_V1TI): Add macro expansion.
8715 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8716 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8717 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8718 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8719 Change check to determine if the instruction is a byte reversing
8720 entry. Fix typo in comment.
8721 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8722 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8723 Add def_builtin calls for new builtins.
8724 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8725 Add define_insn expansion.
8727 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8729 * config/rl78/rl78.md (umaxdi3): New define_expand.
8731 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8733 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8734 for non-QImode registers.
8736 2018-01-22 Richard Biener <rguenther@suse.de>
8738 PR tree-optimization/83963
8739 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8740 including the loop exit block.
8741 (scop_detection::merge_sese): Likewise.
8742 (scop_detection::add_scop): Do it here instead.
8744 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8746 * doc/sourcebuild.texi (arm_softfloat): Document.
8748 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8751 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8752 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8753 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8755 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8756 David Edelsohn <dje.gcc@gmail.com>
8759 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8760 Change "crset eq" to "crset 2".
8761 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8762 (*call_indirect_aix<mode>_nospec): Likewise.
8763 (*call_value_indirect_aix<mode>_nospec): Likewise.
8764 (*call_indirect_elfv2<mode>_nospec): Likewise.
8765 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8766 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8767 change assembly output from . to $.
8768 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8769 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8770 (*tablejump<mode>_internal1_nospec): Likewise.
8772 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8775 * config/sh/sh_optimize_sett_clrt.cc:
8776 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8778 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8780 PR tree-optimization/83940
8781 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8782 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8783 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8784 use it to pass back the definition type.
8785 (vect_check_store_rhs): Likewise rhs_dt_out.
8786 (vect_build_gather_load_calls): Add a mask_dt argument and use
8787 it instead of a call to vect_is_simple_use.
8788 (vectorizable_store): Update calls to vect_check_load_store_mask
8789 and vect_check_store_rhs. Use the dt returned by the latter instead
8790 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8791 instead of calls to vect_is_simple_use. Pass the scalar rather
8792 than the vector operand to vect_is_simple_use when handling
8793 second and subsequent copies of an rhs value.
8794 (vectorizable_load): Update calls to vect_check_load_store_mask
8795 and vect_build_gather_load_calls. Use the cached mask_dt and
8796 gs_info.offset_dt instead of calls to vect_is_simple_use.
8798 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8801 * tree-emutls.c: Include gimplify.h.
8802 (lower_emutls_2): New function.
8803 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8804 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8805 it before further processing.
8808 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8809 UINTVAL (trueop1) instead of INTVAL (op1).
8811 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8815 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8816 INCOMING_FRAME_SP_OFFSET if not defined.
8817 (scan_trace): Add ENTRY argument. If true and
8818 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8819 emit a note to adjust the CFA offset.
8820 (create_cfi_notes): Adjust scan_trace callers.
8821 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8822 INCOMING_FRAME_SP_OFFSET in the CIE.
8823 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8824 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8826 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8827 * doc/tm.texi: Regenerated.
8829 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8831 PR rtl-optimization/83147
8832 * lra-constraints.c (remove_inheritance_pseudos): Use
8833 lra_substitute_pseudo_within_insn.
8835 2018-01-19 Tom de Vries <tom@codesourcery.com>
8836 Cesar Philippidis <cesar@codesourcery.com>
8839 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8841 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8844 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8845 spaces for function labels.
8847 2018-01-19 Martin Liska <mliska@suse.cz>
8849 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8850 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8851 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8852 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8853 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8854 (PRED_CONST_RETURN): Change from 69 to 65.
8855 (PRED_NULL_RETURN): Change from 91 to 71.
8856 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8857 (PRED_LOOP_GUARD): Change from 66 to 73.
8859 2018-01-19 Martin Liska <mliska@suse.cz>
8861 * predict.c (predict_insn_def): Add new assert.
8862 (struct branch_predictor): Change type to signed integer.
8863 (test_prediction_value_range): Amend test to cover
8865 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8866 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8867 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8868 (PRED_LOOP_IV_COMPARE): Likewise.
8869 * predict.h (PROB_UNINITIALIZED): Define new constant.
8871 2018-01-19 Martin Liska <mliska@suse.cz>
8873 * predict.c (dump_prediction): Add new format for
8874 analyze_brprob.py script which is enabled with -details
8876 * profile-count.h (precise_p): New function.
8878 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8880 PR tree-optimization/83922
8881 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8882 there are no statements that need masking.
8883 (vect_active_double_reduction_p): New function.
8884 (vect_analyze_loop_operations): Use it when handling phis that
8885 are not in the loop header.
8887 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8889 PR tree-optimization/83914
8890 * tree-vect-loop.c (vectorizable_induction): Don't convert
8891 init_expr or apply the peeling adjustment for inductions
8892 that are nested within the vectorized loop.
8894 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8896 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8899 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8903 * function.h (gimplify_parameters): Add gimple_seq * argument.
8904 * function.c: Include gimple.h and options.h.
8905 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8906 for the added local temporaries if needed.
8907 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8908 if there are any parameter cleanups, wrap whole body into a
8909 try/finally with the cleanups.
8911 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8914 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8915 Use GET_MODE_CLASS for scalar floating point.
8917 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8921 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8922 Fix call of call_cgraph_insertion_hooks.
8924 2018-01-18 Martin Sebor <msebor@redhat.com>
8926 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8928 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8931 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8934 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8937 * common.opt: (-ffile-prefix-map): New option.
8938 * opts.c (common_handle_option): Defer it.
8939 * opts-global.c (handle_common_deferred_options): Handle it.
8940 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8941 * file-prefix-map.h: New file.
8942 (remap_debug_filename, add_debug_prefix_map): ...here.
8943 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8944 * final.c (debug_prefix_map, add_debug_prefix_map
8945 remap_debug_filename): Move to...
8946 * file-prefix-map.c: New file.
8947 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8948 generalize, get rid of alloca(), use strrchr() instead of strchr().
8949 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8950 Implement in terms of add_prefix_map().
8951 (remap_macro_filename, remap_debug_filename): Implement in term of
8953 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8954 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8955 * dbxout.c: Include file-prefix-map.h.
8956 * varasm.c: Likewise.
8957 * vmsdbgout.c: Likewise.
8958 * xcoffout.c: Likewise.
8959 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8960 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8961 * doc/invoke.texi (-ffile-prefix-map): Document.
8962 (-fdebug-prefix-map): Update description.
8964 2018-01-18 Martin Liska <mliska@suse.cz>
8966 * config/i386/i386.c (indirect_thunk_name): Document that also
8968 (output_indirect_thunk): Document why both instructions
8969 (pause and lfence) are generated.
8971 2018-01-18 Richard Biener <rguenther@suse.de>
8973 PR tree-optimization/83887
8974 * graphite-scop-detection.c
8975 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8976 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8977 (scop_detection::merge_sese): Re-implement with a flood-fill
8978 algorithm that properly finds a SESE region if it exists.
8980 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8983 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8984 pointer_diff optimizations use view_convert instead of convert.
8986 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8988 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8989 Generate different code for -mno-speculate-indirect-jumps.
8990 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8991 (*call_indirect_aix<mode>): Disable for
8992 -mno-speculate-indirect-jumps.
8993 (*call_indirect_aix<mode>_nospec): New define_insn.
8994 (*call_value_indirect_aix<mode>): Disable for
8995 -mno-speculate-indirect-jumps.
8996 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8997 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8998 -mno-speculate-indirect-jumps.
8999 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9001 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9003 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9004 long double type, set the flags for noting the default long double
9005 type, even if we don't pass or return a long double type.
9007 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9010 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9013 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9015 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9016 support for merge[hl].
9017 (fold_mergehl_helper): New helper function.
9018 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9019 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9020 (altivec_vmrglw_direct): Add xxmrglw insn.
9022 2018-01-17 Andrew Waterman <andrew@sifive.com>
9024 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9025 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9027 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9030 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9031 call the lto_location_cache before reading the
9032 DECL_SOURCE_LOCATION of the types.
9034 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9035 Richard Sandiford <richard.sandiford@linaro.org>
9037 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9038 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9039 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9040 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9041 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9043 * config/aarch64/constraints.md (aarch64_movti_operand):
9045 * config/aarch64/predicates.md (Uti): Add new constraint.
9047 2018-01-17 Carl Love <cel@us.ibm.com>
9049 * config/rs6000/vsx.md (define_expand xl_len_r,
9050 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9051 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9053 (define_expand, define_insn): Move the shift left from the
9054 define_insn to the define_expand for lxvl and stxvl instructions.
9055 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9056 and XL_LEN_R definitions to PURE.
9058 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9060 * config/i386/i386.c (indirect_thunk_name): Declare regno
9061 as unsigned int. Compare regno with INVALID_REGNUM.
9062 (output_indirect_thunk): Ditto.
9063 (output_indirect_thunk_function): Ditto.
9064 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9065 in the call to output_indirect_thunk_function.
9067 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9070 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9071 rather than the size of inner_type to determine the stack slot size
9072 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9074 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9077 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9080 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9082 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9083 endian Linux systems to optionally enable multilibs for selecting
9084 the long double type if the user configured an explicit type.
9085 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9086 have no long double multilibs if not defined.
9087 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9088 warn if the user used -mabi={ieee,ibm}longdouble and we built
9089 multilibs for long double.
9090 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9091 appropriate multilib option.
9092 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9094 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9095 for building long double multilibs.
9096 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9098 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9100 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9103 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9105 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9108 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9111 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9114 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9116 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9117 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9119 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9121 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9122 different rtl trees depending on TARGET_64BIT.
9123 (rs6000_gen_lvx): Likewise.
9125 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9127 * config/visium/visium.md (nop): Tweak comment.
9128 (hazard_nop): Likewise.
9130 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9132 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9133 -mspeculate-indirect-jumps.
9134 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9135 for -mno-speculate-indirect-jumps.
9136 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9137 (*call_value_indirect_elfv2<mode>): Disable for
9138 -mno-speculate-indirect-jumps.
9139 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9140 (indirect_jump): Emit different RTL for
9141 -mno-speculate-indirect-jumps.
9142 (*indirect_jump<mode>): Disable for
9143 -mno-speculate-indirect-jumps.
9144 (*indirect_jump<mode>_nospec): New define_insn.
9145 (tablejump): Emit different RTL for
9146 -mno-speculate-indirect-jumps.
9147 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9148 (tablejumpsi_nospec): New define_expand.
9149 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9150 (tablejumpdi_nospec): New define_expand.
9151 (*tablejump<mode>_internal1): Disable for
9152 -mno-speculate-indirect-jumps.
9153 (*tablejump<mode>_internal1_nospec): New define_insn.
9154 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9157 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9159 * caller-save.c (insert_save): Drop unnecessary parameter. All
9162 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9163 Richard Biener <rguenth@suse.de>
9166 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9167 return early, inline manually is_gimple_sizepos. Make sure if we
9168 call gimplify_expr we don't end up with a gimple constant.
9169 * tree.c (variably_modified_type_p): Don't return true for
9170 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9171 * gimplify.h (is_gimple_sizepos): Remove.
9173 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9175 PR tree-optimization/83857
9176 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9177 vectorizable_live_operation for pure SLP statements.
9178 (vectorizable_live_operation): Handle PHIs.
9180 2018-01-16 Richard Biener <rguenther@suse.de>
9182 PR tree-optimization/83867
9183 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9184 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9186 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9189 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9190 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9191 If off is not INTEGER_CST, issue a may not be aligned warning
9192 rather than isn't aligned. Use isn%'t rather than isn't.
9193 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9195 <case MULT_EXPR>: Improve the case when bottom and one of the
9196 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9197 operand, in that case check if the other operand is multiple of
9198 bottom divided by the INTEGER_CST operand.
9200 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9203 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9204 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9205 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9206 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9207 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9208 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9209 (pa_function_arg_size): New function.
9211 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9213 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9214 in a separate statement.
9216 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9218 PR tree-optimization/83847
9219 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9220 group gathers and scatters.
9222 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9224 PR rtl-optimization/86620
9225 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9227 PR rtl-optimization/83213
9228 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9229 to last if both are JUMP_INSNs.
9231 PR tree-optimization/83843
9232 * gimple-ssa-store-merging.c
9233 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9234 store_immediate_info for bswap/nop orig_stores.
9236 2018-01-15 Andrew Waterman <andrew@sifive.com>
9238 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9240 <UDIV>: Increase cost if !TARGET_DIV.
9242 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9244 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9245 (define_attr "cr_logical_3op"): New.
9246 (cceq_ior_compare): Adjust.
9247 (cceq_ior_compare_complement): Adjust.
9248 (*cceq_rev_compare): Adjust.
9249 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9250 (is_cracked_insn): Adjust.
9251 (insn_must_be_first_in_group): Adjust.
9252 * config/rs6000/40x.md: Adjust.
9253 * config/rs6000/440.md: Adjust.
9254 * config/rs6000/476.md: Adjust.
9255 * config/rs6000/601.md: Adjust.
9256 * config/rs6000/603.md: Adjust.
9257 * config/rs6000/6xx.md: Adjust.
9258 * config/rs6000/7450.md: Adjust.
9259 * config/rs6000/7xx.md: Adjust.
9260 * config/rs6000/8540.md: Adjust.
9261 * config/rs6000/cell.md: Adjust.
9262 * config/rs6000/e300c2c3.md: Adjust.
9263 * config/rs6000/e500mc.md: Adjust.
9264 * config/rs6000/e500mc64.md: Adjust.
9265 * config/rs6000/e5500.md: Adjust.
9266 * config/rs6000/e6500.md: Adjust.
9267 * config/rs6000/mpc.md: Adjust.
9268 * config/rs6000/power4.md: Adjust.
9269 * config/rs6000/power5.md: Adjust.
9270 * config/rs6000/power6.md: Adjust.
9271 * config/rs6000/power7.md: Adjust.
9272 * config/rs6000/power8.md: Adjust.
9273 * config/rs6000/power9.md: Adjust.
9274 * config/rs6000/rs64.md: Adjust.
9275 * config/rs6000/titan.md: Adjust.
9277 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9279 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9280 ix86_indirect_branch_register logic.
9282 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9284 * config/i386/constraints.md (Bs): Update
9285 ix86_indirect_branch_register check. Don't check
9286 ix86_indirect_branch_register with GOT_memory_operand.
9288 * config/i386/predicates.md (GOT_memory_operand): Don't check
9289 ix86_indirect_branch_register here.
9290 (GOT32_symbol_operand): Likewise.
9292 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9294 * config/i386/predicates.md (constant_call_address_operand):
9295 Rewrite ix86_indirect_branch_register logic.
9296 (sibcall_insn_operand): Likewise.
9298 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9300 * config/i386/constraints.md (Bs): Replace
9301 ix86_indirect_branch_thunk_register with
9302 ix86_indirect_branch_register.
9304 * config/i386/i386.md (indirect_jump): Likewise.
9305 (tablejump): Likewise.
9306 (*sibcall_memory): Likewise.
9307 (*sibcall_value_memory): Likewise.
9308 Peepholes of indirect call and jump via memory: Likewise.
9309 * config/i386/i386.opt: Likewise.
9310 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9311 (GOT_memory_operand): Likewise.
9312 (call_insn_operand): Likewise.
9313 (sibcall_insn_operand): Likewise.
9314 (GOT32_symbol_operand): Likewise.
9316 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9319 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9320 type rather than type addr's type points to.
9321 (expand_omp_atomic_mutex): Likewise.
9322 (expand_omp_atomic): Likewise.
9324 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9327 * config/i386/i386.c (output_indirect_thunk_function): Use
9328 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9329 for __x86_return_thunk.
9331 2018-01-15 Richard Biener <rguenther@suse.de>
9334 * expmed.c (extract_bit_field_1): Fix typo.
9336 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9339 * config/arm/iterators.md (VF): New mode iterator.
9340 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9341 Remove integer-related logic from pattern.
9342 (neon_vabd<mode>_3): Likewise.
9344 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9347 * common.opt (fstrict-overflow): No longer an alias.
9348 (fwrapv-pointer): New option.
9349 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9350 also for pointer types based on flag_wrapv_pointer.
9351 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9352 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9353 opts->x_flag_wrapv got set.
9354 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9355 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9356 POINTER_TYPE_OVERFLOW_UNDEFINED.
9357 * match.pd: Likewise in address comparison pattern.
9358 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9360 2018-01-15 Richard Biener <rguenther@suse.de>
9363 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9364 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9365 Reset type names to their identifier if their TYPE_DECL doesn't
9366 have linkage (and thus is used for ODR and devirt).
9367 (save_debug_info_for_decl): Remove.
9368 (save_debug_info_for_type): Likewise.
9369 (add_tree_to_fld_list): Adjust.
9370 * tree-pretty-print.c (dump_generic_node): Make dumping of
9371 type names more robust.
9373 2018-01-15 Richard Biener <rguenther@suse.de>
9375 * BASE-VER: Bump to 8.0.1.
9377 2018-01-14 Martin Sebor <msebor@redhat.com>
9380 * builtins.c (check_access): Avoid warning when the no-warning bit
9383 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9385 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9386 * ira-color (allocno_hard_regs_compare): Likewise.
9388 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9391 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9392 Use .pushsection/.popsection.
9394 2018-01-14 Martin Sebor <msebor@redhat.com>
9397 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9399 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9401 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9402 entry from extra_headers.
9403 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9404 extra_headers, make the list bitwise identical to the i?86-*-* one.
9406 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9408 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9409 -mcmodel=large with -mindirect-branch=thunk,
9410 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9411 -mfunction-return=thunk-extern.
9412 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9413 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9414 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9416 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9418 * config/i386/i386.c (print_reg): Print the name of the full
9419 integer register without '%'.
9420 (ix86_print_operand): Handle 'V'.
9421 * doc/extend.texi: Document 'V' modifier.
9423 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9425 * config/i386/constraints.md (Bs): Disallow memory operand for
9426 -mindirect-branch-register.
9428 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9429 (GOT_memory_operand): Likewise.
9430 (call_insn_operand): Likewise.
9431 (sibcall_insn_operand): Likewise.
9432 (GOT32_symbol_operand): Likewise.
9433 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9434 for -mindirect-branch-register.
9435 (tablejump): Likewise.
9436 (*sibcall_memory): Likewise.
9437 (*sibcall_value_memory): Likewise.
9438 Disallow peepholes of indirect call and jump via memory for
9439 -mindirect-branch-register.
9440 (*call_pop): Replace m with Bw.
9441 (*call_value_pop): Likewise.
9442 (*sibcall_pop_memory): Replace m with Bs.
9443 * config/i386/i386.opt (mindirect-branch-register): New option.
9444 * doc/invoke.texi: Document -mindirect-branch-register option.
9446 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9448 * config/i386/i386-protos.h (ix86_output_function_return): New.
9449 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9450 set function_return_type.
9451 (indirect_thunk_name): Add ret_p to indicate thunk for function
9453 (output_indirect_thunk_function): Pass false to
9454 indirect_thunk_name.
9455 (ix86_output_indirect_branch_via_reg): Likewise.
9456 (ix86_output_indirect_branch_via_push): Likewise.
9457 (output_indirect_thunk_function): Create alias for function
9458 return thunk if regno < 0.
9459 (ix86_output_function_return): New function.
9460 (ix86_handle_fndecl_attribute): Handle function_return.
9461 (ix86_attribute_table): Add function_return.
9462 * config/i386/i386.h (machine_function): Add
9463 function_return_type.
9464 * config/i386/i386.md (simple_return_internal): Use
9465 ix86_output_function_return.
9466 (simple_return_internal_long): Likewise.
9467 * config/i386/i386.opt (mfunction-return=): New option.
9468 (indirect_branch): Mention -mfunction-return=.
9469 * doc/extend.texi: Document function_return function attribute.
9470 * doc/invoke.texi: Document -mfunction-return= option.
9472 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9474 * config/i386/i386-opts.h (indirect_branch): New.
9475 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9476 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9477 with local indirect jump when converting indirect call and jump.
9478 (ix86_set_indirect_branch_type): New.
9479 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9480 (indirectlabelno): New.
9481 (indirect_thunk_needed): Likewise.
9482 (indirect_thunk_bnd_needed): Likewise.
9483 (indirect_thunks_used): Likewise.
9484 (indirect_thunks_bnd_used): Likewise.
9485 (INDIRECT_LABEL): Likewise.
9486 (indirect_thunk_name): Likewise.
9487 (output_indirect_thunk): Likewise.
9488 (output_indirect_thunk_function): Likewise.
9489 (ix86_output_indirect_branch_via_reg): Likewise.
9490 (ix86_output_indirect_branch_via_push): Likewise.
9491 (ix86_output_indirect_branch): Likewise.
9492 (ix86_output_indirect_jmp): Likewise.
9493 (ix86_code_end): Call output_indirect_thunk_function if needed.
9494 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9496 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9497 (ix86_attribute_table): Add indirect_branch.
9498 * config/i386/i386.h (machine_function): Add indirect_branch_type
9499 and has_local_indirect_jump.
9500 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9502 (tablejump): Likewise.
9503 (*indirect_jump): Use ix86_output_indirect_jmp.
9504 (*tablejump_1): Likewise.
9505 (simple_return_indirect_internal): Likewise.
9506 * config/i386/i386.opt (mindirect-branch=): New option.
9507 (indirect_branch): New.
9510 (thunk-inline): Likewise.
9511 (thunk-extern): Likewise.
9512 * doc/extend.texi: Document indirect_branch function attribute.
9513 * doc/invoke.texi: Document -mindirect-branch= option.
9515 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9518 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9520 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9522 * ipa-inline.c (want_inline_small_function_p): Return false if
9523 inlining has already failed with CIF_FINAL_ERROR.
9524 (update_caller_keys): Call want_inline_small_function_p before
9526 (update_callee_keys): Likewise.
9528 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9530 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9532 (rs6000_quadword_masked_address_p): Likewise.
9533 (quad_aligned_load_p): Likewise.
9534 (quad_aligned_store_p): Likewise.
9535 (const_load_sequence_p): Add comment to describe the outer-most loop.
9536 (mimic_memory_attributes_and_flags): New function.
9537 (rs6000_gen_stvx): Likewise.
9538 (replace_swapped_aligned_store): Likewise.
9539 (rs6000_gen_lvx): Likewise.
9540 (replace_swapped_aligned_load): Likewise.
9541 (replace_swapped_load_constant): Capitalize argument name in
9542 comment describing this function.
9543 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9544 and stores that access quad-word aligned addresses and replace
9545 with stvx or lvx instructions when appropriate.
9546 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9547 New function prototype.
9548 (rs6000_quadword_masked_address_p): Likewise.
9549 (rs6000_gen_lvx): Likewise.
9550 (rs6000_gen_stvx): Likewise.
9551 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9552 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9553 when memory address is aligned.
9554 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9555 this split to select lvx instruction when memory address is aligned.
9556 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9557 instruction when memory address is aligned.
9558 (*vsx_le_perm_load_v16qi): Likewise.
9559 (four unnamed splitters): Modify to select the stvx instruction
9560 when memory is aligned.
9562 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9564 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9565 which appears in the queue multiple times.
9567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9568 Alan Hayward <alan.hayward@arm.com>
9569 David Sherwood <david.sherwood@arm.com>
9571 * tree-vectorizer.h (vec_lower_bound): New structure.
9572 (_loop_vec_info): Add check_nonzero and lower_bounds.
9573 (LOOP_VINFO_CHECK_NONZERO): New macro.
9574 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9575 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9576 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9577 fields. Make seg_len the distance travelled, not including the
9579 (dr_direction_indicator): Declare.
9580 (dr_zero_step_indicator): Likewise.
9581 (dr_known_forward_stride_p): Likewise.
9582 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9584 (runtime_alias_check_p): Allow runtime alias checks with
9586 (operator ==): Compare access_size and align.
9587 (prune_runtime_alias_test_list): Rework for new distinction between
9588 the access_size and seg_len.
9589 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9591 (get_segment_min_max): New function.
9592 (create_intersect_range_checks): Use it.
9593 (dr_step_indicator): New function.
9594 (dr_direction_indicator): Likewise.
9595 (dr_zero_step_indicator): Likewise.
9596 (dr_known_forward_stride_p): Likewise.
9597 * tree-loop-distribution.c (data_ref_segment_size): Return
9598 DR_STEP * (niters - 1).
9599 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9601 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9602 (vect_preserves_scalar_order_p): New function, split out from...
9603 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9604 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9605 (vect_vfa_access_size): New function.
9606 (vect_vfa_align): Likewise.
9607 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9608 (dump_lower_bound): New function.
9609 (vect_check_lower_bound): Likewise.
9610 (vect_small_gap_p): Likewise.
9611 (vectorizable_with_step_bound_p): Likewise.
9612 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9613 depencies if the vectorization factor is 1. Convert the checks
9614 for nonzero steps into checks on the bounds of DR_STEP. Try using
9615 a bunds check for variable steps if the minimum required step is
9616 relatively small. Update calls to the dr_with_seg_len
9617 constructor and to vect_compile_time_alias.
9618 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9620 (vect_loop_versioning): Call it.
9621 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9623 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9626 Alan Hayward <alan.hayward@arm.com>
9627 David Sherwood <david.sherwood@arm.com>
9629 * doc/sourcebuild.texi (vect_scatter_store): Document.
9630 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9632 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9634 * genopinit.c (main): Add supports_vec_scatter_store and
9635 supports_vec_scatter_store_cached to target_optabs.
9636 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9637 IFN_MASK_SCATTER_STORE.
9638 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9640 * internal-fn.h (internal_store_fn_p): Declare.
9641 (internal_fn_stored_value_index): Likewise.
9642 * internal-fn.c (scatter_store_direct): New macro.
9643 (expand_scatter_store_optab_fn): New function.
9644 (direct_scatter_store_optab_supported_p): New macro.
9645 (internal_store_fn_p): New function.
9646 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9647 IFN_MASK_SCATTER_STORE.
9648 (internal_fn_mask_index): Likewise.
9649 (internal_fn_stored_value_index): New function.
9650 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9652 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9653 * optabs-query.c (supports_vec_scatter_store_p): New function.
9654 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9655 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9656 true for scatter stores.
9657 (vect_gather_scatter_fn_p): Handle scatter stores too.
9658 (vect_check_gather_scatter): Consider using scatter stores if
9659 supports_vec_scatter_store_p.
9660 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9662 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9663 internal_fn_stored_value_index.
9664 (check_load_store_masking): Handle scatter stores too.
9665 (vect_get_store_rhs): Make public.
9666 (vectorizable_call): Use internal_store_fn_p.
9667 (vectorizable_store): Handle scatter store internal functions.
9668 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9669 when deciding whether the end of the group has been reached.
9670 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9671 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9672 (mask_scatter_store<mode>): New insns.
9674 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9675 Alan Hayward <alan.hayward@arm.com>
9676 David Sherwood <david.sherwood@arm.com>
9678 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9679 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9680 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9682 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9683 Use vect_truncate_gather_scatter_offset if we can't treat the
9684 operation as a normal gather load or scatter store.
9685 (get_group_load_store_type): Take the gather_scatter_info
9686 as argument. Try using a gather load or scatter store for
9687 single-element groups.
9688 (get_load_store_type): Update calls to get_group_load_store_type
9689 and vect_use_strided_gather_scatters_p.
9691 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9692 Alan Hayward <alan.hayward@arm.com>
9693 David Sherwood <david.sherwood@arm.com>
9695 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9696 optional tree argument.
9697 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9699 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9700 but continue to use the current value as a fallback.
9701 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9702 to compare the updates.
9703 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9704 (get_load_store_type): Use it when handling a strided access.
9705 (vect_get_strided_load_store_ops): New function.
9706 (vect_get_data_ptr_increment): Likewise.
9707 (vectorizable_load): Handle strided gather loads. Always pass
9708 a step to vect_create_data_ref_ptr and bump_vector_ptr.
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/md.texi (gather_load@var{m}): Document.
9715 (mask_gather_load@var{m}): Likewise.
9716 * genopinit.c (main): Add supports_vec_gather_load and
9717 supports_vec_gather_load_cached to target_optabs.
9718 * optabs-tree.c (init_tree_optimization_optabs): Use
9719 ggc_cleared_alloc to allocate target_optabs.
9720 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9721 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9723 * internal-fn.h (internal_load_fn_p): Declare.
9724 (internal_gather_scatter_fn_p): Likewise.
9725 (internal_fn_mask_index): Likewise.
9726 (internal_gather_scatter_fn_supported_p): Likewise.
9727 * internal-fn.c (gather_load_direct): New macro.
9728 (expand_gather_load_optab_fn): New function.
9729 (direct_gather_load_optab_supported_p): New macro.
9730 (direct_internal_fn_optab): New function.
9731 (internal_load_fn_p): Likewise.
9732 (internal_gather_scatter_fn_p): Likewise.
9733 (internal_fn_mask_index): Likewise.
9734 (internal_gather_scatter_fn_supported_p): Likewise.
9735 * optabs-query.c (supports_at_least_one_mode_p): New function.
9736 (supports_vec_gather_load_p): Likewise.
9737 * optabs-query.h (supports_vec_gather_load_p): Declare.
9738 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9739 and memory_type field.
9740 (NUM_PATTERNS): Bump to 15.
9741 * tree-vect-data-refs.c: Include internal-fn.h.
9742 (vect_gather_scatter_fn_p): New function.
9743 (vect_describe_gather_scatter_call): Likewise.
9744 (vect_check_gather_scatter): Try using internal functions for
9745 gather loads. Recognize existing calls to a gather load function.
9746 (vect_analyze_data_refs): Consider using gather loads if
9747 supports_vec_gather_load_p.
9748 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9749 (vect_get_gather_scatter_offset_type): Likewise.
9750 (vect_convert_mask_for_vectype): Likewise.
9751 (vect_add_conversion_to_patterm): Likewise.
9752 (vect_try_gather_scatter_pattern): Likewise.
9753 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9754 (vect_vect_recog_func_ptrs): Add it.
9755 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9756 internal_fn_mask_index and internal_gather_scatter_fn_p.
9757 (check_load_store_masking): Take the gather_scatter_info as an
9758 argument and handle gather loads.
9759 (vect_get_gather_scatter_ops): New function.
9760 (vectorizable_call): Check internal_load_fn_p.
9761 (vectorizable_load): Likewise. Handle gather load internal
9763 (vectorizable_store): Update call to check_load_store_masking.
9764 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9765 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9766 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9767 (aarch64_gather_scale_operand_d): New predicates.
9768 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9769 (mask_gather_load<mode>): New insns.
9771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9772 Alan Hayward <alan.hayward@arm.com>
9773 David Sherwood <david.sherwood@arm.com>
9775 * optabs.def (fold_left_plus_optab): New optab.
9776 * doc/md.texi (fold_left_plus_@var{m}): Document.
9777 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9778 * internal-fn.c (fold_left_direct): Define.
9779 (expand_fold_left_optab_fn): Likewise.
9780 (direct_fold_left_optab_supported_p): Likewise.
9781 * fold-const-call.c (fold_const_fold_left): New function.
9782 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9783 * tree-parloops.c (valid_reduction_p): New function.
9784 (gather_scalar_reductions): Use it.
9785 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9786 (vect_finish_replace_stmt): Declare.
9787 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9788 (needs_fold_left_reduction_p): New function, split out from...
9789 (vect_is_simple_reduction): ...here. Accept reductions that
9790 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9791 (vect_force_simple_reduction): Also store the reduction type in
9792 the assignment's STMT_VINFO_REDUC_TYPE.
9793 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9794 (merge_with_identity): New function.
9795 (vect_expand_fold_left): Likewise.
9796 (vectorize_fold_left_reduction): Likewise.
9797 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9798 scalar phi in place for it. Check for target support and reject
9799 cases that would reassociate the operation. Defer the transform
9800 phase to vectorize_fold_left_reduction.
9801 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9802 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9803 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9805 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9807 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9808 call to ifc_temp_var.
9810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9811 Alan Hayward <alan.hayward@arm.com>
9812 David Sherwood <david.sherwood@arm.com>
9814 * target.def (legitimize_address_displacement): Take the original
9815 offset as a poly_int.
9816 * targhooks.h (default_legitimize_address_displacement): Update
9818 * targhooks.c (default_legitimize_address_displacement): Likewise.
9819 * doc/tm.texi: Regenerate.
9820 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9821 as an argument, moving assert of ad->disp == ad->disp_term to...
9822 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9823 Try calling targetm.legitimize_address_displacement before expanding
9824 the address rather than afterwards, and adjust for the new interface.
9825 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9826 Match the new hook interface. Handle SVE addresses.
9827 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9830 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9832 * Makefile.in (OBJS): Add early-remat.o.
9833 * target.def (select_early_remat_modes): New hook.
9834 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9835 * doc/tm.texi: Regenerate.
9836 * targhooks.h (default_select_early_remat_modes): Declare.
9837 * targhooks.c (default_select_early_remat_modes): New function.
9838 * timevar.def (TV_EARLY_REMAT): New timevar.
9839 * passes.def (pass_early_remat): New pass.
9840 * tree-pass.h (make_pass_early_remat): Declare.
9841 * early-remat.c: New file.
9842 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9844 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9847 Alan Hayward <alan.hayward@arm.com>
9848 David Sherwood <david.sherwood@arm.com>
9850 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9851 vfm1 with a bound_epilog parameter.
9852 (vect_do_peeling): Update calls accordingly, and move the prologue
9853 call earlier in the function. Treat the base bound_epilog as 0 for
9854 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9855 this base when peeling for gaps.
9856 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9857 with fully-masked loops.
9858 (vect_estimate_min_profitable_iters): Handle the single peeled
9859 iteration in that case.
9861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9862 Alan Hayward <alan.hayward@arm.com>
9863 David Sherwood <david.sherwood@arm.com>
9865 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9866 single-element interleaving even if the size is not a power of 2.
9867 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9868 accesses for single-element interleaving if the group size is
9871 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9872 Alan Hayward <alan.hayward@arm.com>
9873 David Sherwood <david.sherwood@arm.com>
9875 * doc/md.texi (fold_extract_last_@var{m}): Document.
9876 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9877 * optabs.def (fold_extract_last_optab): New optab.
9878 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9879 * internal-fn.c (fold_extract_direct): New macro.
9880 (expand_fold_extract_optab_fn): Likewise.
9881 (direct_fold_extract_optab_supported_p): Likewise.
9882 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9883 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9884 EXTRACT_LAST_REDUCTION.
9885 (get_initial_def_for_reduction): Do not create an initial vector
9886 for EXTRACT_LAST_REDUCTION reductions.
9887 (vectorizable_reduction): Leave the scalar phi in place for
9888 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9889 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9890 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9891 transform phase to vectorizable_condition.
9892 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9894 (vect_finish_stmt_generation): ...here.
9895 (vect_finish_replace_stmt): New function.
9896 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9897 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9899 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9901 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9902 Alan Hayward <alan.hayward@arm.com>
9903 David Sherwood <david.sherwood@arm.com>
9905 * doc/md.texi (extract_last_@var{m}): Document.
9906 * optabs.def (extract_last_optab): New optab.
9907 * internal-fn.def (EXTRACT_LAST): New internal function.
9908 * internal-fn.c (cond_unary_direct): New macro.
9909 (expand_cond_unary_optab_fn): Likewise.
9910 (direct_cond_unary_optab_supported_p): Likewise.
9911 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9912 loops using EXTRACT_LAST.
9913 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9914 (extract_last_<mode>): ...this optab.
9915 (vec_extract<mode><Vel>): Update accordingly.
9917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9918 Alan Hayward <alan.hayward@arm.com>
9919 David Sherwood <david.sherwood@arm.com>
9921 * target.def (empty_mask_is_expensive): New hook.
9922 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9923 * doc/tm.texi: Regenerate.
9924 * targhooks.h (default_empty_mask_is_expensive): Declare.
9925 * targhooks.c (default_empty_mask_is_expensive): New function.
9926 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9927 if the target says that empty masks are expensive.
9928 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9930 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9932 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9933 Alan Hayward <alan.hayward@arm.com>
9934 David Sherwood <david.sherwood@arm.com>
9936 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9937 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9938 (vect_use_loop_mask_for_alignment_p): New function.
9939 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9940 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9941 niters_skip argument. Make sure that the first niters_skip elements
9942 of the first iteration are inactive.
9943 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9944 Update call to vect_set_loop_masks_directly.
9945 (get_misalign_in_elems): New function, split out from...
9946 (vect_gen_prolog_loop_niters): ...here.
9947 (vect_update_init_of_dr): Take a code argument that specifies whether
9948 the adjustment should be added or subtracted.
9949 (vect_update_init_of_drs): Likewise.
9950 (vect_prepare_for_masked_peels): New function.
9951 (vect_do_peeling): Skip prologue peeling if we're using a mask
9952 instead. Update call to vect_update_inits_of_drs.
9953 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9955 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9956 alignment. Do not include the number of peeled iterations in
9957 the minimum threshold in that case.
9958 (vectorizable_induction): Adjust the start value down by
9959 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9960 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9961 Take the number of skipped iterations into account when calculating
9963 * tree-vect-stmts.c (vect_gen_while_not): New function.
9965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9966 Alan Hayward <alan.hayward@arm.com>
9967 David Sherwood <david.sherwood@arm.com>
9969 * doc/sourcebuild.texi (vect_fully_masked): Document.
9970 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9972 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9974 (vect_analyze_loop_2): ...here. Don't check the vectorization
9975 factor against the number of loop iterations if the loop is
9978 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9979 Alan Hayward <alan.hayward@arm.com>
9980 David Sherwood <david.sherwood@arm.com>
9982 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9983 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9984 (dump_groups): Update accordingly.
9985 (iv_use::mem_type): New member variable.
9986 (address_p): New function.
9987 (record_use): Add a mem_type argument and initialize the new
9989 (record_group_use): Add a mem_type argument. Use address_p.
9990 Remove obsolete null checks of base_object. Update call to record_use.
9991 (find_interesting_uses_op): Update call to record_group_use.
9992 (find_interesting_uses_cond): Likewise.
9993 (find_interesting_uses_address): Likewise.
9994 (get_mem_type_for_internal_fn): New function.
9995 (find_address_like_use): Likewise.
9996 (find_interesting_uses_stmt): Try find_address_like_use before
9997 calling find_interesting_uses_op.
9998 (addr_offset_valid_p): Use the iv mem_type field as the type
9999 of the addressed memory.
10000 (add_autoinc_candidates): Likewise.
10001 (get_address_cost): Likewise.
10002 (split_small_address_groups_p): Use address_p.
10003 (split_address_groups): Likewise.
10004 (add_iv_candidate_for_use): Likewise.
10005 (autoinc_possible_for_pair): Likewise.
10006 (rewrite_groups): Likewise.
10007 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10008 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10009 (get_alias_ptr_type_for_ptr_address): New function.
10010 (rewrite_use_address): Rewrite address uses in calls that were
10011 identified by find_address_like_use.
10013 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10014 Alan Hayward <alan.hayward@arm.com>
10015 David Sherwood <david.sherwood@arm.com>
10017 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10019 * gimple-expr.h (is_gimple_addressable: Likewise.
10020 * gimple-expr.c (is_gimple_address): Likewise.
10021 * internal-fn.c (expand_call_mem_ref): New function.
10022 (expand_mask_load_optab_fn): Use it.
10023 (expand_mask_store_optab_fn): Likewise.
10025 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10026 Alan Hayward <alan.hayward@arm.com>
10027 David Sherwood <david.sherwood@arm.com>
10029 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10030 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10031 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10032 (cond_umax@var{mode}): Document.
10033 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10034 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10035 (cond_umin_optab, cond_umax_optab): New optabs.
10036 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10037 (COND_IOR, COND_XOR): New internal functions.
10038 * internal-fn.h (get_conditional_internal_fn): Declare.
10039 * internal-fn.c (cond_binary_direct): New macro.
10040 (expand_cond_binary_optab_fn): Likewise.
10041 (direct_cond_binary_optab_supported_p): Likewise.
10042 (get_conditional_internal_fn): New function.
10043 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10044 Cope with reduction statements that are vectorized as calls rather
10046 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10047 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10048 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10049 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10050 (UNSPEC_COND_EOR): New unspecs.
10051 (optab): Add mappings for them.
10052 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10053 (sve_int_op, sve_fp_op): New int attributes.
10055 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10056 Alan Hayward <alan.hayward@arm.com>
10057 David Sherwood <david.sherwood@arm.com>
10059 * optabs.def (while_ult_optab): New optab.
10060 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10061 * internal-fn.def (WHILE_ULT): New internal function.
10062 * internal-fn.h (direct_internal_fn_supported_p): New override
10063 that takes two types as argument.
10064 * internal-fn.c (while_direct): New macro.
10065 (expand_while_optab_fn): New function.
10066 (convert_optab_supported_p): Likewise.
10067 (direct_while_optab_supported_p): New macro.
10068 * wide-int.h (wi::udiv_ceil): New function.
10069 * tree-vectorizer.h (rgroup_masks): New structure.
10070 (vec_loop_masks): New typedef.
10071 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10072 and fully_masked_p.
10073 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10074 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10075 (vect_max_vf): New function.
10076 (slpeel_make_loop_iterate_ntimes): Delete.
10077 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10078 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10079 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10080 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10081 internal-fn.h, stor-layout.h and optabs-query.h.
10082 (vect_set_loop_mask): New function.
10083 (add_preheader_seq): Likewise.
10084 (add_header_seq): Likewise.
10085 (interleave_supported_p): Likewise.
10086 (vect_maybe_permute_loop_masks): Likewise.
10087 (vect_set_loop_masks_directly): Likewise.
10088 (vect_set_loop_condition_masked): Likewise.
10089 (vect_set_loop_condition_unmasked): New function, split out from
10090 slpeel_make_loop_iterate_ntimes.
10091 (slpeel_make_loop_iterate_ntimes): Rename to..
10092 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10093 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10094 (vect_do_peeling): Update call accordingly.
10095 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10097 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10098 mask_compare_type, can_fully_mask_p and fully_masked_p.
10099 (release_vec_loop_masks): New function.
10100 (_loop_vec_info): Use it to free the loop masks.
10101 (can_produce_all_loop_masks_p): New function.
10102 (vect_get_max_nscalars_per_iter): Likewise.
10103 (vect_verify_full_masking): Likewise.
10104 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10105 retries, and free the mask rgroups before retrying. Check loop-wide
10106 reasons for disallowing fully-masked loops. Make the final decision
10107 about whether use a fully-masked loop or not.
10108 (vect_estimate_min_profitable_iters): Do not assume that peeling
10109 for the number of iterations will be needed for fully-masked loops.
10110 (vectorizable_reduction): Disable fully-masked loops.
10111 (vectorizable_live_operation): Likewise.
10112 (vect_halve_mask_nunits): New function.
10113 (vect_double_mask_nunits): Likewise.
10114 (vect_record_loop_mask): Likewise.
10115 (vect_get_loop_mask): Likewise.
10116 (vect_transform_loop): Handle the case in which the final loop
10117 iteration might handle a partial vector. Call vect_set_loop_condition
10118 instead of slpeel_make_loop_iterate_ntimes.
10119 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10120 (check_load_store_masking): New function.
10121 (prepare_load_store_mask): Likewise.
10122 (vectorizable_store): Handle fully-masked loops.
10123 (vectorizable_load): Likewise.
10124 (supportable_widening_operation): Use vect_halve_mask_nunits for
10126 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10127 (vect_gen_while): New function.
10128 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10129 (aarch64_uqdec<mode>): New insn.
10131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10132 Alan Hayward <alan.hayward@arm.com>
10133 David Sherwood <david.sherwood@arm.com>
10135 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10136 (reduc_xor_scal_optab): New optabs.
10137 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10138 (reduc_xor_scal_@var{m}): Document.
10139 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10140 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10141 internal functions.
10142 * fold-const-call.c (fold_const_call): Handle them.
10143 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10144 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10145 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10146 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10147 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10148 (UNSPEC_XORV): New unspecs.
10149 (optab): Add entries for them.
10150 (BITWISEV): New int iterator.
10151 (bit_reduc_op): New int attributes.
10153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10154 Alan Hayward <alan.hayward@arm.com>
10155 David Sherwood <david.sherwood@arm.com>
10157 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10158 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10159 * optabs.def (vec_shl_insert_optab): New optab.
10160 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10161 (duplicate_and_interleave): Likewise.
10162 * tree-vect-loop.c: Include internal-fn.h.
10163 (neutral_op_for_slp_reduction): New function, split out from
10164 get_initial_defs_for_reduction.
10165 (get_initial_def_for_reduction): Handle option 2 for variable-length
10166 vectors by loading the neutral value into a vector and then shifting
10167 the initial value into element 0.
10168 (get_initial_defs_for_reduction): Replace the code argument with
10169 the neutral value calculated by neutral_op_for_slp_reduction.
10170 Use gimple_build_vector for constant-length vectors.
10171 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10172 but the first group_size elements have a neutral value.
10173 Use duplicate_and_interleave otherwise.
10174 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10175 Update call to get_initial_defs_for_reduction. Handle SLP
10176 reductions for variable-length vectors by creating one vector
10177 result for each scalar result, with the elements associated
10178 with other scalar results stubbed out with the neutral value.
10179 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10180 Require IFN_VEC_SHL_INSERT for double reductions on
10181 variable-length vectors, or SLP reductions that have
10182 a neutral value. Require can_duplicate_and_interleave_p
10183 support for variable-length unchained SLP reductions if there
10184 is no neutral value, such as for MIN/MAX reductions. Also require
10185 the number of vector elements to be a multiple of the number of
10186 SLP statements when doing variable-length unchained SLP reductions.
10187 Update call to vect_create_epilog_for_reduction.
10188 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10189 and remove initial values.
10190 (duplicate_and_interleave): Make public.
10191 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10192 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10194 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10195 Alan Hayward <alan.hayward@arm.com>
10196 David Sherwood <david.sherwood@arm.com>
10198 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10199 (can_duplicate_and_interleave_p): New function.
10200 (vect_get_and_check_slp_defs): Take the vector of statements
10201 rather than just the current one. Remove excess parentheses.
10202 Restriction rejectinon of vect_constant_def and vect_external_def
10203 for variable-length vectors to boolean types, or types for which
10204 can_duplicate_and_interleave_p is false.
10205 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10206 (duplicate_and_interleave): New function.
10207 (vect_get_constant_vectors): Use gimple_build_vector for
10208 constant-length vectors and suitable variable-length constant
10209 vectors. Use duplicate_and_interleave for other variable-length
10210 vectors. Don't defer the update when inserting new statements.
10212 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10213 Alan Hayward <alan.hayward@arm.com>
10214 David Sherwood <david.sherwood@arm.com>
10216 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10217 min_profitable_iters doesn't go negative.
10219 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10220 Alan Hayward <alan.hayward@arm.com>
10221 David Sherwood <david.sherwood@arm.com>
10223 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10224 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10225 * optabs.def (vec_mask_load_lanes_optab): New optab.
10226 (vec_mask_store_lanes_optab): Likewise.
10227 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10228 (MASK_STORE_LANES): Likewise.
10229 * internal-fn.c (mask_load_lanes_direct): New macro.
10230 (mask_store_lanes_direct): Likewise.
10231 (expand_mask_load_optab_fn): Handle masked operations.
10232 (expand_mask_load_lanes_optab_fn): New macro.
10233 (expand_mask_store_optab_fn): Handle masked operations.
10234 (expand_mask_store_lanes_optab_fn): New macro.
10235 (direct_mask_load_lanes_optab_supported_p): Likewise.
10236 (direct_mask_store_lanes_optab_supported_p): Likewise.
10237 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10239 (vect_load_lanes_supported): Likewise.
10240 * tree-vect-data-refs.c (strip_conversion): New function.
10241 (can_group_stmts_p): Likewise.
10242 (vect_analyze_data_ref_accesses): Use it instead of checking
10243 for a pair of assignments.
10244 (vect_store_lanes_supported): Take a masked_p parameter.
10245 (vect_load_lanes_supported): Likewise.
10246 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10247 vect_store_lanes_supported and vect_load_lanes_supported.
10248 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10249 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10250 parameter. Don't allow gaps for masked accesses.
10251 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10252 and vect_load_lanes_supported.
10253 (get_load_store_type): Take a masked_p parameter and update
10254 call to get_group_load_store_type.
10255 (vectorizable_store): Update call to get_load_store_type.
10256 Handle IFN_MASK_STORE_LANES.
10257 (vectorizable_load): Update call to get_load_store_type.
10258 Handle IFN_MASK_LOAD_LANES.
10260 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10261 Alan Hayward <alan.hayward@arm.com>
10262 David Sherwood <david.sherwood@arm.com>
10264 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10266 * config/aarch64/aarch64-protos.h
10267 (aarch64_sve_struct_memory_operand_p): Declare.
10268 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10269 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10270 (VPRED, vpred): Handle SVE structure modes.
10271 * config/aarch64/constraints.md (Utx): New constraint.
10272 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10273 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10274 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10275 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10276 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10277 structure modes. Split into pieces after RA.
10278 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10279 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10281 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10282 SVE structure modes.
10283 (aarch64_classify_address): Likewise.
10284 (sizetochar): Move earlier in file.
10285 (aarch64_print_operand): Handle SVE register lists.
10286 (aarch64_array_mode): New function.
10287 (aarch64_sve_struct_memory_operand_p): Likewise.
10288 (TARGET_ARRAY_MODE): Redefine.
10290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10291 Alan Hayward <alan.hayward@arm.com>
10292 David Sherwood <david.sherwood@arm.com>
10294 * target.def (array_mode): New target hook.
10295 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10296 * doc/tm.texi: Regenerate.
10297 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10298 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10299 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10300 targetm.array_mode.
10301 * stor-layout.c (mode_for_array): Likewise. Support polynomial
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 * fold-const.c (fold_binary_loc): Check the argument types
10309 rather than the result type when testing for a vector operation.
10311 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10313 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10314 * doc/tm.texi: Regenerate.
10316 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10317 Alan Hayward <alan.hayward@arm.com>
10318 David Sherwood <david.sherwood@arm.com>
10320 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10321 (sve): Document new AArch64 extension.
10322 * doc/md.texi (w): Extend the description of the AArch64
10323 constraint to include SVE vectors.
10324 (Upl, Upa): Document new AArch64 predicate constraints.
10325 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10327 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10328 (msve-vector-bits=): New option.
10329 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10330 SVE when these are disabled.
10331 (sve): New extension.
10332 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10333 modes. Adjust their number of units based on aarch64_sve_vg.
10334 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10335 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10336 aarch64_addr_query_type.
10337 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10338 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10339 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10340 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10341 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10342 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10343 (aarch64_simd_imm_zero_p): Delete.
10344 (aarch64_check_zero_based_sve_index_immediate): Declare.
10345 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10346 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10347 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10348 (aarch64_sve_float_mul_immediate_p): Likewise.
10349 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10350 rather than an rtx.
10351 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10352 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10353 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10354 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10355 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10356 (aarch64_regmode_natural_size): Likewise.
10357 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10358 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10360 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10361 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10362 for VG and the SVE predicate registers.
10363 (V_ALIASES): Add a "z"-prefixed alias.
10364 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10365 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10366 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10367 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10368 (REG_CLASS_NAMES): Add entries for them.
10369 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10370 and the predicate registers.
10371 (aarch64_sve_vg): Declare.
10372 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10373 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10374 (REGMODE_NATURAL_SIZE): Define.
10375 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10377 * config/aarch64/aarch64.c: Include cfgrtl.h.
10378 (simd_immediate_info): Add a constructor for series vectors,
10379 and an associated step field.
10380 (aarch64_sve_vg): New variable.
10381 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10382 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10383 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10384 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10385 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10386 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10387 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10388 (aarch64_get_mask_mode): New functions.
10389 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10390 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10391 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10392 predicate modes and predicate registers. Explicitly restrict
10393 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10394 to store a vector mode if it is recognized by
10395 aarch64_classify_vector_mode.
10396 (aarch64_regmode_natural_size): New function.
10397 (aarch64_hard_regno_caller_save_mode): Return the original mode
10399 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10400 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10401 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10402 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10404 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10405 does not overlap dest if the function is frame-related. Handle
10407 (aarch64_split_add_offset): New function.
10408 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10409 them aarch64_add_offset.
10410 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10411 and update call to aarch64_sub_sp.
10412 (aarch64_add_cfa_expression): New function.
10413 (aarch64_expand_prologue): Pass extra temporary registers to the
10414 functions above. Handle the case in which we need to emit new
10415 DW_CFA_expressions for registers that were originally saved
10416 relative to the stack pointer, but now have to be expressed
10417 relative to the frame pointer.
10418 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10420 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10421 IP0 and IP1 values for SVE frames.
10422 (aarch64_expand_vec_series): New function.
10423 (aarch64_expand_sve_widened_duplicate): Likewise.
10424 (aarch64_expand_sve_const_vector): Likewise.
10425 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10426 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10427 into the register, rather than emitting a SET directly.
10428 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10429 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10430 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10431 (offset_9bit_signed_scaled_p): New functions.
10432 (aarch64_replicate_bitmask_imm): New function.
10433 (aarch64_bitmask_imm): Use it.
10434 (aarch64_cannot_force_const_mem): Reject expressions involving
10435 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10436 (aarch64_classify_index): Handle SVE indices, by requiring
10437 a plain register index with a scale that matches the element size.
10438 (aarch64_classify_address): Handle SVE addresses. Assert that
10439 the mode of the address is VOIDmode or an integer mode.
10440 Update call to aarch64_classify_symbol.
10441 (aarch64_classify_symbolic_expression): Update call to
10442 aarch64_classify_symbol.
10443 (aarch64_const_vec_all_in_range_p): New function.
10444 (aarch64_print_vector_float_operand): Likewise.
10445 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10446 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10447 and the FP immediates 1.0 and 0.5.
10448 (aarch64_print_address_internal): Handle SVE addresses.
10449 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10450 (aarch64_regno_regclass): Handle predicate registers.
10451 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10453 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10454 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10455 (aarch64_convert_sve_vector_bits): New function.
10456 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10457 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10458 rather than an rtx.
10459 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10460 Handle SVE vector and predicate modes. Accept VL-based constants
10461 that need only one temporary register, and VL offsets that require
10462 no temporary registers.
10463 (aarch64_conditional_register_usage): Mark the predicate registers
10464 as fixed if SVE isn't available.
10465 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10466 Return true for SVE vector and predicate modes.
10467 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10468 rather than an unsigned int. Handle SVE modes.
10469 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10471 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10473 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10474 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10475 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10476 (aarch64_sve_float_mul_immediate_p): New functions.
10477 (aarch64_sve_valid_immediate): New function.
10478 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10479 Explicitly reject structure modes. Check for INDEX constants.
10480 Handle PTRUE and PFALSE constants.
10481 (aarch64_check_zero_based_sve_index_immediate): New function.
10482 (aarch64_simd_imm_zero_p): Delete.
10483 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10484 vector modes. Accept constants in the range of CNT[BHWD].
10485 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10486 ask for an Advanced SIMD mode.
10487 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10488 (aarch64_simd_vector_alignment): Handle SVE predicates.
10489 (aarch64_vectorize_preferred_vector_alignment): New function.
10490 (aarch64_simd_vector_alignment_reachable): Use it instead of
10492 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10493 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10495 (MAX_VECT_LEN): Delete.
10496 (expand_vec_perm_d): Add a vec_flags field.
10497 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10498 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10499 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10501 (aarch64_evpc_rev): Rename to...
10502 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10503 (aarch64_evpc_rev_global): New function.
10504 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10505 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10507 (aarch64_evpc_sve_tbl): New function.
10508 (aarch64_expand_vec_perm_const_1): Update after rename of
10509 aarch64_evpc_rev. Handle SVE permutes too, trying
10510 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10511 than aarch64_evpc_tbl.
10512 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10513 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10514 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10515 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10516 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10517 (aarch64_expand_sve_vcond): New functions.
10518 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10519 of aarch64_vector_mode_p.
10520 (aarch64_dwarf_poly_indeterminate_value): New function.
10521 (aarch64_compute_pressure_classes): Likewise.
10522 (aarch64_can_change_mode_class): Likewise.
10523 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10524 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10525 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10526 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10527 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10528 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10529 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10530 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10532 (Dn, Dl, Dr): Accept const as well as const_vector.
10533 (Dz): Likewise. Compare against CONST0_RTX.
10534 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10535 of "vector" where appropriate.
10536 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10537 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10538 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10539 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10540 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10541 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10542 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10543 (v_int_equiv): Extend to SVE modes.
10544 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10546 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10547 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10548 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10549 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10550 (SVE_COND_FP_CMP): New int iterators.
10551 (perm_hilo): Handle the new unpack unspecs.
10552 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10554 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10555 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10556 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10557 (aarch64_equality_operator, aarch64_constant_vector_operand)
10558 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10559 (aarch64_sve_nonimmediate_operand): Likewise.
10560 (aarch64_sve_general_operand): Likewise.
10561 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10562 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10563 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10564 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10565 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10566 (aarch64_sve_float_arith_immediate): Likewise.
10567 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10568 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10569 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10570 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10571 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10572 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10573 (aarch64_sve_float_arith_operand): Likewise.
10574 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10575 (aarch64_sve_float_mul_operand): Likewise.
10576 (aarch64_sve_vec_perm_operand): Likewise.
10577 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10578 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10579 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10580 as well as const_vector.
10581 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10582 in file. Use CONST0_RTX and CONSTM1_RTX.
10583 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10584 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10585 Use aarch64_simd_imm_zero.
10586 * config/aarch64/aarch64-sve.md: New file.
10587 * config/aarch64/aarch64.md: Include it.
10588 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10589 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10590 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10591 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10592 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10593 (sve): New attribute.
10594 (enabled): Disable instructions with the sve attribute unless
10596 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10597 aarch64_expand_mov_immediate.
10598 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10599 CNT[BHSD] immediates.
10600 (movti): Split CONST_POLY_INT moves into two halves.
10601 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10602 Split additions that need a temporary here if the destination
10603 is the stack pointer.
10604 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10605 (*add<mode>3_poly_1): New instruction.
10606 (set_clobber_cc): New expander.
10608 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10610 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10611 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10612 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10613 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10614 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10615 Change innermode from fixed_mode_size to machine_mode.
10616 (simplify_subreg): Update call accordingly. Handle a constant-sized
10617 subreg of a variable-length CONST_VECTOR.
10619 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10620 Alan Hayward <alan.hayward@arm.com>
10621 David Sherwood <david.sherwood@arm.com>
10623 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10624 (add_offset_to_base): New function, split out from...
10625 (create_mem_ref): ...here. When handling a scale other than 1,
10626 check first whether the address is valid without the offset.
10627 Add it into the base if so, leaving the index and scale as-is.
10629 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10632 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10633 fold_for_warn before checking if arg2 is INTEGER_CST.
10635 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10637 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10638 (store_multiple_operation): Delete.
10639 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10640 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10641 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10642 guarded by TARGET_STRING.
10643 (rs6000_output_load_multiple): Delete.
10644 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10645 OPTION_MASK_STRING / TARGET_STRING handling.
10646 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10647 (const rs6000_opt_masks) <"string">: Change mask to 0.
10648 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10649 (MASK_STRING): Delete.
10650 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10652 (load_multiple): Delete.
10659 (store_multiple): Delete.
10666 (movmemsi_8reg): Delete.
10667 (corresponding unnamed define_insn): Delete.
10668 (movmemsi_6reg): Delete.
10669 (corresponding unnamed define_insn): Delete.
10670 (movmemsi_4reg): Delete.
10671 (corresponding unnamed define_insn): Delete.
10672 (movmemsi_2reg): Delete.
10673 (corresponding unnamed define_insn): Delete.
10674 (movmemsi_1reg): Delete.
10675 (corresponding unnamed define_insn): Delete.
10676 * config/rs6000/rs6000.opt (mno-string): New.
10677 (mstring): Replace by deprecation warning stub.
10678 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10680 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10682 * regrename.c (regrename_do_replace): If replacing the same
10683 reg multiple times, try to reuse last created gen_raw_REG.
10686 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10687 main to workaround a bug in GDB.
10689 2018-01-12 Tom de Vries <tom@codesourcery.com>
10692 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10694 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10696 PR rtl-optimization/80481
10697 * ira-color.c (get_cap_member): New function.
10698 (allocnos_conflict_by_live_ranges_p): Use it.
10699 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10700 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10702 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10705 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10706 (*saddl_se_1): Ditto.
10707 (*ssubsi_1): Ditto.
10708 (*ssubl_se_1): Ditto.
10710 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10712 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10713 rather than wi::to_widest for DR_INITs.
10714 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10715 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10716 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10718 (vect_analyze_group_access_1): Note that here.
10720 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10722 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10723 polynomial type sizes.
10725 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10727 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10728 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10729 (gimple_add_tmp_var): Likewise.
10731 2018-01-12 Martin Liska <mliska@suse.cz>
10733 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10734 (gimple_alloc_sizes): Likewise.
10735 (dump_gimple_statistics): Use PRIu64 in printf format.
10736 * gimple.h: Change uint64_t to int.
10738 2018-01-12 Martin Liska <mliska@suse.cz>
10740 * tree-core.h: Use uint64_t instead of int.
10741 * tree.c (tree_node_counts): Likewise.
10742 (tree_node_sizes): Likewise.
10743 (dump_tree_statistics): Use PRIu64 in printf format.
10745 2018-01-12 Martin Liska <mliska@suse.cz>
10747 * Makefile.in: As qsort_chk is implemented in vec.c, add
10748 vec.o to linkage of gencfn-macros.
10749 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10750 passing the info to record_node_allocation_statistics.
10751 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10753 * ggc-common.c (struct ggc_usage): Add operator== and use
10754 it in operator< and compare function.
10755 * mem-stats.h (struct mem_usage): Likewise.
10756 * vec.c (struct vec_usage): Remove operator< and compare
10757 function. Can be simply inherited.
10759 2018-01-12 Martin Jambor <mjambor@suse.cz>
10762 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10763 * tree-ssa-math-opts.c: Include domwalk.h.
10764 (convert_mult_to_fma_1): New function.
10765 (fma_transformation_info): New type.
10766 (fma_deferring_state): Likewise.
10767 (cancel_fma_deferring): New function.
10768 (result_of_phi): Likewise.
10769 (last_fma_candidate_feeds_initial_phi): Likewise.
10770 (convert_mult_to_fma): Added deferring logic, split actual
10771 transformation to convert_mult_to_fma_1.
10772 (math_opts_dom_walker): New type.
10773 (math_opts_dom_walker::after_dom_children): New method, body moved
10774 here from pass_optimize_widening_mul::execute, added deferring logic
10776 (pass_optimize_widening_mul::execute): Moved most of code to
10777 math_opts_dom_walker::after_dom_children.
10778 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10779 * config/i386/i386.c (ix86_option_override_internal): Added
10780 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10782 2018-01-12 Richard Biener <rguenther@suse.de>
10785 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10786 inline instance vars.
10788 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10791 * config/rx/rx.c (rx_is_restricted_memory_address):
10792 Handle SUBREG case.
10794 2018-01-12 Richard Biener <rguenther@suse.de>
10796 PR tree-optimization/80846
10797 * target.def (split_reduction): New target hook.
10798 * targhooks.c (default_split_reduction): New function.
10799 * targhooks.h (default_split_reduction): Declare.
10800 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10801 target requests first reduce vectors by combining low and high
10803 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10804 (get_vectype_for_scalar_type_and_size): Export.
10805 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10806 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10807 * doc/tm.texi: Regenerate.
10808 * config/i386/i386.c (ix86_split_reduction): Implement
10809 TARGET_VECTORIZE_SPLIT_REDUCTION.
10811 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10814 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10815 in PIC mode except for TARGET_VXWORKS_RTP.
10816 * config/sparc/sparc.c: Include cfgrtl.h.
10817 (TARGET_INIT_PIC_REG): Define.
10818 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10819 (sparc_pic_register_p): New predicate.
10820 (sparc_legitimate_address_p): Use it.
10821 (sparc_legitimize_pic_address): Likewise.
10822 (sparc_delegitimize_address): Likewise.
10823 (sparc_mode_dependent_address_p): Likewise.
10824 (gen_load_pcrel_sym): Remove 4th parameter.
10825 (load_got_register): Adjust call to above. Remove obsolete stuff.
10826 (sparc_expand_prologue): Do not call load_got_register here.
10827 (sparc_flat_expand_prologue): Likewise.
10828 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10829 (sparc_use_pseudo_pic_reg): New function.
10830 (sparc_init_pic_reg): Likewise.
10831 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10832 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10834 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10836 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10837 Add item for branch_cost.
10839 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10841 PR rtl-optimization/83565
10842 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10843 not extend the result to a larger mode for rotate operations.
10844 (num_sign_bit_copies1): Likewise.
10846 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10849 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10851 Use values-Xc.o for -pedantic.
10852 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10854 2018-01-12 Martin Liska <mliska@suse.cz>
10857 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10859 (possible_polymorphic_call_targets): Use it.
10860 (ipa_devirt): Likewise.
10862 2018-01-12 Martin Liska <mliska@suse.cz>
10864 * profile-count.h (enum profile_quality): Use 0 as invalid
10865 enum value of profile_quality.
10867 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10869 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10870 -mext-string options.
10872 2018-01-12 Richard Biener <rguenther@suse.de>
10874 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10875 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10876 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10878 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10880 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10882 * configure.ac (--with-long-double-format): Add support for the
10883 configuration option to change the default long double format on
10885 * config.gcc (powerpc*-linux*-*): Likewise.
10886 * configure: Regenerate.
10887 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10888 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10889 used without modification.
10891 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10893 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10894 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10895 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10896 MISC_BUILTIN_SPEC_BARRIER.
10897 (rs6000_init_builtins): Likewise.
10898 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10900 (speculation_barrier): New define_insn.
10901 * doc/extend.texi: Document __builtin_speculation_barrier.
10903 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10906 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10907 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10908 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10910 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10911 integral modes instead of "ss" and "sd".
10912 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10913 vectors with 32-bit and 64-bit elements.
10914 (vecdupssescalarmodesuffix): New mode attribute.
10915 (vec_dup<mode>): Use it.
10917 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10920 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10921 frame if argument is passed on stack.
10923 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10926 * ree.c (combine_reaching_defs): Optimize also
10927 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10928 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10930 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10932 PR middle-end/83189
10933 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10935 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10937 PR middle-end/83718
10938 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10939 after they are computed.
10941 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10943 PR tree-optimization/83695
10944 * gimple-loop-linterchange.cc
10945 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10946 reset cached scev information after interchange.
10947 (pass_linterchange::execute): Remove call to scev_reset_htab.
10949 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10951 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10952 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10953 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10954 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10955 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10956 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10957 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10958 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10959 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10960 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10961 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10962 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10963 (V_lane_reg): Likewise.
10964 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10966 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10967 (vfmal_lane_low<mode>_intrinsic,
10968 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10969 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10970 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10971 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10972 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10973 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10975 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10977 * config/arm/arm-cpus.in (fp16fml): New feature.
10978 (ALL_SIMD): Add fp16fml.
10979 (armv8.2-a): Add fp16fml as an option.
10980 (armv8.3-a): Likewise.
10981 (armv8.4-a): Add fp16fml as part of fp16.
10982 * config/arm/arm.h (TARGET_FP16FML): Define.
10983 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10985 * config/arm/arm-modes.def (V2HF): Define.
10986 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10987 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10988 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10989 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10990 vfmsl_low, vfmsl_high): New set of builtins.
10991 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10992 (vfml_op): New code attribute.
10993 (VFMLHALVES): New int iterator.
10994 (VFML, VFMLSEL): New mode attributes.
10995 (V_reg): Define mapping for V2HF.
10996 (V_hi, V_lo): New mode attributes.
10997 (VF_constraint): Likewise.
10998 (vfml_half, vfml_half_selector): New int attributes.
10999 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11001 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11002 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11004 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11005 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11006 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11007 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11009 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11010 Document new effective target and option set.
11012 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11014 * config/arm/arm-cpus.in (armv8_4): New feature.
11015 (ARMv8_4a): New fgroup.
11016 (armv8.4-a): New arch.
11017 * config/arm/arm-tables.opt: Regenerate.
11018 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11019 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11020 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11021 Add matching rules for -march=armv8.4-a and extensions.
11022 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11024 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11027 * config/rx/rx.md (BW): New mode attribute.
11028 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11030 2018-01-11 Richard Biener <rguenther@suse.de>
11032 PR tree-optimization/83435
11033 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11034 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11035 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11037 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11038 Alan Hayward <alan.hayward@arm.com>
11039 David Sherwood <david.sherwood@arm.com>
11041 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11043 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11044 (aarch64_print_address_internal): Use it to check for a zero offset.
11046 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11047 Alan Hayward <alan.hayward@arm.com>
11048 David Sherwood <david.sherwood@arm.com>
11050 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11051 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11052 Return a poly_int64 rather than a HOST_WIDE_INT.
11053 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11054 rather than a HOST_WIDE_INT.
11055 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11056 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11057 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11058 final_offset from HOST_WIDE_INT to poly_int64.
11059 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11060 to_constant when getting the number of units in an Advanced SIMD
11062 (aarch64_builtin_vectorized_function): Check for a constant number
11064 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11066 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11067 attribute instead of GET_MODE_NUNITS.
11068 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11069 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11070 GET_MODE_SIZE for fixed-size registers.
11071 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11072 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11073 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11074 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11075 (aarch64_print_operand, aarch64_print_address_internal)
11076 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11077 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11078 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11079 Handle polynomial GET_MODE_SIZE.
11080 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11081 wider than SImode without modification.
11082 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11083 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11084 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11085 passing and returning SVE modes.
11086 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11087 rather than GEN_INT.
11088 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11089 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11090 (aarch64_allocate_and_probe_stack_space): Likewise.
11091 (aarch64_layout_frame): Cope with polynomial offsets.
11092 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11093 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11094 polynomial offsets.
11095 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11096 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11097 poly_int64 rather than a HOST_WIDE_INT.
11098 (aarch64_get_separate_components, aarch64_process_components)
11099 (aarch64_expand_prologue, aarch64_expand_epilogue)
11100 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11101 (aarch64_anchor_offset): New function, split out from...
11102 (aarch64_legitimize_address): ...here.
11103 (aarch64_builtin_vectorization_cost): Handle polynomial
11104 TYPE_VECTOR_SUBPARTS.
11105 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11107 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11108 number of elements from the PARALLEL rather than the mode.
11109 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11110 rather than GET_MODE_BITSIZE.
11111 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11112 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11113 (aarch64_expand_vec_perm_const_1): Handle polynomial
11114 d->perm.length () and d->perm elements.
11115 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11116 Apply to_constant to d->perm elements.
11117 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11118 polynomial CONST_VECTOR_NUNITS.
11119 (aarch64_move_pointer): Take amount as a poly_int64 rather
11121 (aarch64_progress_pointer): Avoid temporary variable.
11122 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11123 the mode attribute instead of GET_MODE.
11125 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11126 Alan Hayward <alan.hayward@arm.com>
11127 David Sherwood <david.sherwood@arm.com>
11129 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11130 x exists before using it.
11131 (aarch64_add_constant_internal): Rename to...
11132 (aarch64_add_offset_1): ...this. Replace regnum with separate
11133 src and dest rtxes. Handle the case in which they're different,
11134 including when the offset is zero. Replace scratchreg with an rtx.
11135 Use 2 additions if there is no spare register into which we can
11136 move a 16-bit constant.
11137 (aarch64_add_constant): Delete.
11138 (aarch64_add_offset): Replace reg with separate src and dest
11139 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11140 Use aarch64_add_offset_1.
11141 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11142 an rtx rather than an int. Take the delta as a poly_int64
11143 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11144 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11145 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11146 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11147 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11148 and aarch64_add_sp.
11149 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11150 aarch64_add_constant.
11152 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11154 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11155 Use scalar_float_mode.
11157 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11159 * config/aarch64/aarch64-simd.md
11160 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11161 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11162 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11163 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11164 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11165 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11166 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11167 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11168 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11169 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11171 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11174 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11175 targ_options->x_arm_arch_string is non NULL.
11177 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11179 * config/aarch64/aarch64.h
11180 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11182 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11185 * expmed.c (emit_store_flag_force): Swap if const op0
11186 and change VOIDmode to mode of op0.
11188 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11190 PR rtl-optimization/83761
11191 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11192 than bytes to mode_for_size.
11194 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11196 PR middle-end/83189
11197 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11198 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11201 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11203 PR middle-end/83575
11204 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11205 when in layout mode.
11206 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11207 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11210 2018-01-10 Michael Collison <michael.collison@arm.com>
11212 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11213 * config/aarch64/aarch64-option-extension.def: Add
11214 AARCH64_OPT_EXTENSION of 'fp16fml'.
11215 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11216 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11217 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11218 * config/aarch64/constraints.md (Ui7): New constraint.
11219 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11220 (VFMLA_SEL_W): Ditto.
11223 (VFMLA16_LOW): New int iterator.
11224 (VFMLA16_HIGH): Ditto.
11225 (UNSPEC_FMLAL): New unspec.
11226 (UNSPEC_FMLSL): Ditto.
11227 (UNSPEC_FMLAL2): Ditto.
11228 (UNSPEC_FMLSL2): Ditto.
11229 (f16mac): New code attribute.
11230 * config/aarch64/aarch64-simd-builtins.def
11231 (aarch64_fmlal_lowv2sf): Ditto.
11232 (aarch64_fmlsl_lowv2sf): Ditto.
11233 (aarch64_fmlalq_lowv4sf): Ditto.
11234 (aarch64_fmlslq_lowv4sf): Ditto.
11235 (aarch64_fmlal_highv2sf): Ditto.
11236 (aarch64_fmlsl_highv2sf): Ditto.
11237 (aarch64_fmlalq_highv4sf): Ditto.
11238 (aarch64_fmlslq_highv4sf): Ditto.
11239 (aarch64_fmlal_lane_lowv2sf): Ditto.
11240 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11241 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11242 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11243 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11244 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11245 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11246 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11247 (aarch64_fmlal_lane_highv2sf): Ditto.
11248 (aarch64_fmlsl_lane_highv2sf): Ditto.
11249 (aarch64_fmlal_laneq_highv2sf): Ditto.
11250 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11251 (aarch64_fmlalq_lane_highv4sf): Ditto.
11252 (aarch64_fmlsl_lane_highv4sf): Ditto.
11253 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11254 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11255 * config/aarch64/aarch64-simd.md:
11256 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11257 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11258 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11259 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11260 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11261 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11262 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11263 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11264 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11265 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11266 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11267 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11268 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11269 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11270 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11271 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11272 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11273 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11274 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11275 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11276 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11277 (vfmlsl_low_u32): Ditto.
11278 (vfmlalq_low_u32): Ditto.
11279 (vfmlslq_low_u32): Ditto.
11280 (vfmlal_high_u32): Ditto.
11281 (vfmlsl_high_u32): Ditto.
11282 (vfmlalq_high_u32): Ditto.
11283 (vfmlslq_high_u32): Ditto.
11284 (vfmlal_lane_low_u32): Ditto.
11285 (vfmlsl_lane_low_u32): Ditto.
11286 (vfmlal_laneq_low_u32): Ditto.
11287 (vfmlsl_laneq_low_u32): Ditto.
11288 (vfmlalq_lane_low_u32): Ditto.
11289 (vfmlslq_lane_low_u32): Ditto.
11290 (vfmlalq_laneq_low_u32): Ditto.
11291 (vfmlslq_laneq_low_u32): Ditto.
11292 (vfmlal_lane_high_u32): Ditto.
11293 (vfmlsl_lane_high_u32): Ditto.
11294 (vfmlal_laneq_high_u32): Ditto.
11295 (vfmlsl_laneq_high_u32): Ditto.
11296 (vfmlalq_lane_high_u32): Ditto.
11297 (vfmlslq_lane_high_u32): Ditto.
11298 (vfmlalq_laneq_high_u32): Ditto.
11299 (vfmlslq_laneq_high_u32): Ditto.
11300 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11301 (AARCH64_FL_FOR_ARCH8_4): New.
11302 (AARCH64_ISA_F16FML): New ISA flag.
11303 (TARGET_F16FML): New feature flag for fp16fml.
11304 (doc/invoke.texi): Document new fp16fml option.
11306 2018-01-10 Michael Collison <michael.collison@arm.com>
11308 * config/aarch64/aarch64-builtins.c:
11309 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11310 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11311 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11312 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11313 (AARCH64_ISA_SHA3): New ISA flag.
11314 (TARGET_SHA3): New feature flag for sha3.
11315 * config/aarch64/iterators.md (sha512_op): New int attribute.
11316 (CRYPTO_SHA512): New int iterator.
11317 (UNSPEC_SHA512H): New unspec.
11318 (UNSPEC_SHA512H2): Ditto.
11319 (UNSPEC_SHA512SU0): Ditto.
11320 (UNSPEC_SHA512SU1): Ditto.
11321 * config/aarch64/aarch64-simd-builtins.def
11322 (aarch64_crypto_sha512hqv2di): New builtin.
11323 (aarch64_crypto_sha512h2qv2di): Ditto.
11324 (aarch64_crypto_sha512su0qv2di): Ditto.
11325 (aarch64_crypto_sha512su1qv2di): Ditto.
11326 (aarch64_eor3qv8hi): Ditto.
11327 (aarch64_rax1qv2di): Ditto.
11328 (aarch64_xarqv2di): Ditto.
11329 (aarch64_bcaxqv8hi): Ditto.
11330 * config/aarch64/aarch64-simd.md:
11331 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11332 (aarch64_crypto_sha512su0qv2di): Ditto.
11333 (aarch64_crypto_sha512su1qv2di): Ditto.
11334 (aarch64_eor3qv8hi): Ditto.
11335 (aarch64_rax1qv2di): Ditto.
11336 (aarch64_xarqv2di): Ditto.
11337 (aarch64_bcaxqv8hi): Ditto.
11338 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11339 (vsha512h2q_u64): Ditto.
11340 (vsha512su0q_u64): Ditto.
11341 (vsha512su1q_u64): Ditto.
11342 (veor3q_u16): Ditto.
11343 (vrax1q_u64): Ditto.
11344 (vxarq_u64): Ditto.
11345 (vbcaxq_u16): Ditto.
11346 * config/arm/types.md (crypto_sha512): New type attribute.
11347 (crypto_sha3): Ditto.
11348 (doc/invoke.texi): Document new sha3 option.
11350 2018-01-10 Michael Collison <michael.collison@arm.com>
11352 * config/aarch64/aarch64-builtins.c:
11353 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11354 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11355 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11356 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11357 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11358 (AARCH64_ISA_SM4): New ISA flag.
11359 (TARGET_SM4): New feature flag for sm4.
11360 * config/aarch64/aarch64-simd-builtins.def
11361 (aarch64_sm3ss1qv4si): Ditto.
11362 (aarch64_sm3tt1aq4si): Ditto.
11363 (aarch64_sm3tt1bq4si): Ditto.
11364 (aarch64_sm3tt2aq4si): Ditto.
11365 (aarch64_sm3tt2bq4si): Ditto.
11366 (aarch64_sm3partw1qv4si): Ditto.
11367 (aarch64_sm3partw2qv4si): Ditto.
11368 (aarch64_sm4eqv4si): Ditto.
11369 (aarch64_sm4ekeyqv4si): Ditto.
11370 * config/aarch64/aarch64-simd.md:
11371 (aarch64_sm3ss1qv4si): Ditto.
11372 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11373 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11374 (aarch64_sm4eqv4si): Ditto.
11375 (aarch64_sm4ekeyqv4si): Ditto.
11376 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11377 (sm3part_op): Ditto.
11378 (CRYPTO_SM3TT): Ditto.
11379 (CRYPTO_SM3PART): Ditto.
11380 (UNSPEC_SM3SS1): New unspec.
11381 (UNSPEC_SM3TT1A): Ditto.
11382 (UNSPEC_SM3TT1B): Ditto.
11383 (UNSPEC_SM3TT2A): Ditto.
11384 (UNSPEC_SM3TT2B): Ditto.
11385 (UNSPEC_SM3PARTW1): Ditto.
11386 (UNSPEC_SM3PARTW2): Ditto.
11387 (UNSPEC_SM4E): Ditto.
11388 (UNSPEC_SM4EKEY): Ditto.
11389 * config/aarch64/constraints.md (Ui2): New constraint.
11390 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11391 * config/arm/types.md (crypto_sm3): New type attribute.
11392 (crypto_sm4): Ditto.
11393 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11394 (vsm3tt1aq_u32): Ditto.
11395 (vsm3tt1bq_u32): Ditto.
11396 (vsm3tt2aq_u32): Ditto.
11397 (vsm3tt2bq_u32): Ditto.
11398 (vsm3partw1q_u32): Ditto.
11399 (vsm3partw2q_u32): Ditto.
11400 (vsm4eq_u32): Ditto.
11401 (vsm4ekeyq_u32): Ditto.
11402 (doc/invoke.texi): Document new sm4 option.
11404 2018-01-10 Michael Collison <michael.collison@arm.com>
11406 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11407 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11408 (AARCH64_FL_FOR_ARCH8_4): New.
11409 (AARCH64_FL_V8_4): New flag.
11410 (doc/invoke.texi): Document new armv8.4-a option.
11412 2018-01-10 Michael Collison <michael.collison@arm.com>
11414 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11415 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11416 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11417 * config/aarch64/aarch64-option-extension.def: Add
11418 AARCH64_OPT_EXTENSION of 'sha2'.
11419 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11420 (crypto): Disable sha2 and aes if crypto disabled.
11421 (crypto): Enable aes and sha2 if enabled.
11422 (simd): Disable sha2 and aes if simd disabled.
11423 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11425 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11426 (TARGET_SHA2): New feature flag for sha2.
11427 (TARGET_AES): New feature flag for aes.
11428 * config/aarch64/aarch64-simd.md:
11429 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11430 conditional on TARGET_AES.
11431 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11432 (aarch64_crypto_sha1hsi): Make pattern conditional
11434 (aarch64_crypto_sha1hv4si): Ditto.
11435 (aarch64_be_crypto_sha1hv4si): Ditto.
11436 (aarch64_crypto_sha1su1v4si): Ditto.
11437 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11438 (aarch64_crypto_sha1su0v4si): Ditto.
11439 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11440 (aarch64_crypto_sha256su0v4si): Ditto.
11441 (aarch64_crypto_sha256su1v4si): Ditto.
11442 (doc/invoke.texi): Document new aes and sha2 options.
11444 2018-01-10 Martin Sebor <msebor@redhat.com>
11446 PR tree-optimization/83781
11447 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11450 2018-01-11 Martin Sebor <msebor@gmail.com>
11451 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11453 PR tree-optimization/83501
11454 PR tree-optimization/81703
11456 * tree-ssa-strlen.c (get_string_cst): Rename...
11457 (get_string_len): ...to this. Handle global constants.
11458 (handle_char_store): Adjust.
11460 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11461 Jim Wilson <jimw@sifive.com>
11463 * config/riscv/riscv-protos.h (riscv_output_return): New.
11464 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11465 (riscv_attribute_table, riscv_output_return),
11466 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11467 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11468 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11470 (riscv_expand_prologue): Add early return for naked function.
11471 (riscv_expand_epilogue): Likewise.
11472 (riscv_function_ok_for_sibcall): Return false for naked function.
11473 (riscv_set_current_function): New.
11474 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11475 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11476 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11477 * doc/extend.texi (RISC-V Function Attributes): New.
11479 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11481 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11482 check for 128-bit long double before checking TCmode.
11483 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11484 128-bit long doubles before checking TFmode or TCmode.
11485 (FLOAT128_IBM_P): Likewise.
11487 2018-01-10 Martin Sebor <msebor@redhat.com>
11489 PR tree-optimization/83671
11490 * builtins.c (c_strlen): Unconditionally return zero for the empty
11492 Use -Warray-bounds for warnings.
11493 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11494 for non-constant array indices with COMPONENT_REF, arrays of
11495 arrays, and pointers to arrays.
11496 (gimple_fold_builtin_strlen): Determine and set length range for
11497 non-constant character arrays.
11499 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11501 PR middle-end/81897
11502 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11505 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11507 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11509 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11512 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11513 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11514 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11515 indexed_or_indirect_operand predicate.
11516 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11517 (*vsx_le_perm_load_v8hi): Likewise.
11518 (*vsx_le_perm_load_v16qi): Likewise.
11519 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11520 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11521 (*vsx_le_perm_store_v8hi): Likewise.
11522 (*vsx_le_perm_store_v16qi): Likewise.
11523 (eight unnamed splitters): Likewise.
11525 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11527 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11528 * config/rs6000/emmintrin.h: Likewise.
11529 * config/rs6000/mmintrin.h: Likewise.
11530 * config/rs6000/xmmintrin.h: Likewise.
11532 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11535 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11537 * tree.c (tree_nop_conversion): Return true for location wrapper
11539 (maybe_wrap_with_location): New function.
11540 (selftest::check_strip_nops): New function.
11541 (selftest::test_location_wrappers): New function.
11542 (selftest::tree_c_tests): Call it.
11543 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11544 (maybe_wrap_with_location): New decl.
11545 (EXPR_LOCATION_WRAPPER_P): New macro.
11546 (location_wrapper_p): New inline function.
11547 (tree_strip_any_location_wrapper): New inline function.
11549 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11552 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11553 stack_realign_offset for the largest alignment of stack slot
11555 (ix86_find_max_used_stack_alignment): New function.
11556 (ix86_finalize_stack_frame_flags): Use it. Set
11557 max_used_stack_alignment if we don't realign stack.
11558 * config/i386/i386.h (machine_function): Add
11559 max_used_stack_alignment.
11561 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11563 * config/arm/arm.opt (-mbranch-cost): New option.
11564 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11567 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11570 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11571 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11573 2018-01-10 Richard Biener <rguenther@suse.de>
11576 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11577 early out so it also covers the case where we have a non-NULL
11580 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11582 PR tree-optimization/83753
11583 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11584 for non-strided grouped accesses if the number of elements is 1.
11586 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11589 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11590 * i386.h (TARGET_USE_GATHER): Define.
11591 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11593 2018-01-10 Martin Liska <mliska@suse.cz>
11596 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11597 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11599 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11600 CLEANUP_NO_PARTITIONING is not set.
11602 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11604 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11605 for vectors, as a partial revert of r254296.
11606 * rtl.h (const_vec_p): Delete.
11607 (const_vec_duplicate_p): Don't test for vector CONSTs.
11608 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11609 * expmed.c (make_tree): Likewise.
11612 * common.md (E, F): Use CONSTANT_P instead of checking for
11614 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11615 checking for CONST_VECTOR.
11617 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11619 PR middle-end/83575
11620 * predict.c (force_edge_cold): Handle in more sane way edges
11621 with no prediction.
11623 2018-01-09 Carl Love <cel@us.ibm.com>
11625 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11627 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11628 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11629 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11630 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11631 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11632 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11633 * config/rs6000/rs6000-protos.h: Add extern defition for
11634 rs6000_generate_float2_double_code.
11635 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11637 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11638 (float2_v2df): Add define_expand.
11640 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11643 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11644 op_mode in the force_to_mode call.
11646 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11648 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11649 instead of checking each element individually.
11650 (aarch64_evpc_uzp): Likewise.
11651 (aarch64_evpc_zip): Likewise.
11652 (aarch64_evpc_ext): Likewise.
11653 (aarch64_evpc_rev): Likewise.
11654 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11655 instead of checking each element individually. Return true without
11657 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11658 whether all selected elements come from the same input, instead of
11659 checking each element individually. Remove calls to gen_rtx_REG,
11660 start_sequence and end_sequence and instead assert that no rtl is
11663 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11665 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11666 order of HIGH and CONST checks.
11668 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11670 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11671 if the destination isn't an SSA_NAME.
11673 2018-01-09 Richard Biener <rguenther@suse.de>
11675 PR tree-optimization/83668
11676 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11678 (canonicalize_loop_form): ... here, renamed from ...
11679 (canonicalize_loop_closed_ssa_form): ... this and amended to
11680 swap successor edges for loop exit blocks to make us use
11681 the RPO order we need for initial schedule generation.
11683 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11685 PR tree-optimization/64811
11686 * match.pd: When optimizing comparisons with Inf, avoid
11687 introducing or losing exceptions from comparisons with NaN.
11689 2018-01-09 Martin Liska <mliska@suse.cz>
11692 * asan.c (shadow_mem_size): Add gcc_assert.
11694 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11696 Don't save registers in main().
11699 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11700 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11701 * config/avr/avr.c (avr_set_current_function): Don't error if
11702 naked, OS_task or OS_main are specified at the same time.
11703 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11705 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11707 * common/config/avr/avr-common.c (avr_option_optimization_table):
11708 Switch on -mmain-is-OS_task for optimizing compilations.
11710 2018-01-09 Richard Biener <rguenther@suse.de>
11712 PR tree-optimization/83572
11713 * graphite.c: Include cfganal.h.
11714 (graphite_transform_loops): Connect infinite loops to exit
11715 and remove fake edges at the end.
11717 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11719 * ipa-inline.c (edge_badness): Revert accidental checkin.
11721 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11724 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11725 symbols; not inline clones.
11727 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11730 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11731 hard registers. Formatting fixes.
11733 PR preprocessor/83722
11734 * gcc.c (try_generate_repro): Pass
11735 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11736 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11739 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11740 Kito Cheng <kito.cheng@gmail.com>
11742 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11743 (riscv_leaf_function_p): Delete.
11744 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11746 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11748 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11750 (do_ifelse): New function.
11751 (do_isel): New function.
11752 (do_sub3): New function.
11753 (do_add3): New function.
11754 (do_load_mask_compare): New function.
11755 (do_overlap_load_compare): New function.
11756 (expand_compare_loop): New function.
11757 (expand_block_compare): Call expand_compare_loop() when appropriate.
11758 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11759 option description.
11760 (-mblock-compare-inline-loop-limit): New option.
11762 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11765 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11766 Reverse order of second and third operands in first alternative.
11767 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11768 of first and second elements in UNSPEC_VPERMR vector.
11769 (altivec_expand_vec_perm_le): Likewise.
11771 2018-01-08 Jeff Law <law@redhat.com>
11773 PR rtl-optimizatin/81308
11774 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11775 (process_switch): If group_case_labels makes a change, then set
11777 (pass_convert_switch::execute): If a switch is converted, then
11778 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11780 PR rtl-optimization/81308
11781 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11784 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11786 PR target/83663 - Revert r255946
11787 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11788 generation for cases where splatting a value is not useful.
11789 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11790 across a vec_duplicate and a paradoxical subreg forming a vector
11791 mode to a vec_concat.
11793 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11795 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11796 -march=armv8.3-a variants.
11797 * config/arm/t-multilib: Likewise.
11798 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11800 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11802 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11804 (cceq_ior_compare_complement): Give it a name so I can use it, and
11805 change boolean_or_operator predicate to boolean_operator so it can
11806 be used to generate a crand.
11807 (eqne): New code iterator.
11808 (bd/bd_neg): New code_attrs.
11809 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11810 a single define_insn.
11811 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11812 decrement (bdnzt/bdnzf/bdzt/bdzf).
11813 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11814 with the new names of the branch decrement patterns, and added the
11815 names of the branch decrement conditional patterns.
11817 2018-01-08 Richard Biener <rguenther@suse.de>
11819 PR tree-optimization/83563
11820 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11823 2018-01-08 Richard Biener <rguenther@suse.de>
11825 PR middle-end/83713
11826 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11828 2018-01-08 Richard Biener <rguenther@suse.de>
11830 PR tree-optimization/83685
11831 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11832 references to abnormals.
11834 2018-01-08 Richard Biener <rguenther@suse.de>
11837 * dwarf2out.c (output_indirect_strings): Handle empty
11838 skeleton_debug_str_hash.
11839 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11841 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11843 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11844 (emit_store_direct): Likewise.
11845 (arc_trampoline_adjust_address): Likewise.
11846 (arc_asm_trampoline_template): New function.
11847 (arc_initialize_trampoline): Use asm_trampoline_template.
11848 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11849 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11850 * config/arc/arc.md (flush_icache): Delete pattern.
11852 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11854 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11855 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11858 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11861 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11862 by not USED_FOR_TARGET.
11863 (make_pass_resolve_sw_modes): Likewise.
11865 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11867 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11870 2018-01-08 Richard Biener <rguenther@suse.de>
11872 PR middle-end/83580
11873 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11875 2018-01-08 Richard Biener <rguenther@suse.de>
11877 PR middle-end/83517
11878 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11880 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11882 PR middle-end/81897
11883 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11884 basic blocks with a small number of successors.
11885 (convert_control_dep_chain_into_preds): Improve handling of
11887 (dump_predicates): Split apart into...
11888 (dump_pred_chain): ...here...
11889 (dump_pred_info): ...and here.
11890 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11891 (can_chain_union_be_invalidated_p): Improve check for invalidation
11893 (uninit_uses_cannot_happen): Avoid unnecessary if
11894 convert_control_dep_chain_into_preds yielded nothing.
11896 2018-01-06 Martin Sebor <msebor@redhat.com>
11898 PR tree-optimization/83640
11899 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11900 subtracting negative offset from size.
11901 (builtin_access::overlap): Adjust offset bounds of the access to fall
11902 within the size of the object if possible.
11904 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11906 PR rtl-optimization/83699
11907 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11908 extract_bit_field_as_subreg to cases in which the extracted
11909 value is also a vector.
11911 * lra-constraints.c (process_alt_operands): Test for the equivalence
11912 substitutions when detecting a possible reload cycle.
11914 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11917 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11918 by default if flag_selective_schedling{,2}. Formatting fixes.
11920 PR rtl-optimization/83682
11921 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11922 if it has non-VECTOR_MODE element mode.
11923 (vec_duplicate_p): Likewise.
11925 PR middle-end/83694
11926 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11927 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11929 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11932 * config/i386/i386-builtin.def
11933 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11934 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11935 Require also OPTION_MASK_ISA_AVX512F in addition to
11936 OPTION_MASK_ISA_GFNI.
11937 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11938 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11939 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11940 to OPTION_MASK_ISA_GFNI.
11941 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11942 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11943 OPTION_MASK_ISA_AVX512BW.
11944 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11945 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11946 addition to OPTION_MASK_ISA_GFNI.
11947 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11948 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11949 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11950 to OPTION_MASK_ISA_GFNI.
11951 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11952 a requirement for all ISAs rather than any of them with a few
11954 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11956 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11957 bitmasks to be enabled with 3 exceptions, instead of requiring any
11958 enabled ISA with lots of exceptions.
11959 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11960 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11961 Change avx512bw in isa attribute to avx512f.
11962 * config/i386/sgxintrin.h: Add license boilerplate.
11963 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11964 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11965 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11966 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11968 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11969 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11970 temporarily sse2 rather than sse if not enabled already.
11973 * config/i386/sse.md (VI248_VLBW): Rename to ...
11974 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11975 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11976 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11977 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11978 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11979 mode iterator instead of VI248_VLBW.
11981 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11983 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11984 (record_modified): Skip clobbers; add debug output.
11985 (param_change_prob): Use sreal frequencies.
11987 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11989 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11990 punt for user-aligned variables.
11992 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11994 * tree-chrec.c (chrec_contains_symbols): Return true for
11997 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12000 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12001 of (x|y) == x for BICS pattern.
12003 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12005 PR tree-optimization/83605
12006 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12007 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12010 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12012 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12013 * config/epiphany/rtems.h: New file.
12015 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12016 Uros Bizjak <ubizjak@gmail.com>
12019 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12020 QIreg_operand instead of register_operand predicate.
12021 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12022 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12023 comments instead of -fmitigate[-_]rop.
12025 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12028 * cgraphunit.c (symbol_table::compile): Switch to text_section
12029 before calling assembly_start debug hook.
12030 * run-rtl-passes.c (run_rtl_passes): Likewise.
12033 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12035 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12036 range_int_cst_p rather than !symbolic_range_p before calling
12037 extract_range_from_multiplicative_op_1.
12039 2018-01-04 Jeff Law <law@redhat.com>
12041 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12042 redundant test in assertion.
12044 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12046 * doc/rtl.texi: Document machine_mode wrapper classes.
12048 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12050 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12051 using tree_to_uhwi.
12053 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12055 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12056 the VEC_PERM_EXPR fold to fail.
12058 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12061 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12062 to switched_sections.
12064 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12067 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12068 test for d.testing.
12070 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12073 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12074 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12076 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12079 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12080 is BLKmode and bitpos not zero or mode change is needed.
12082 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12085 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12088 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12091 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12092 instead of MULT rtx. Update all corresponding splitters.
12093 (*saddl_se): Ditto.
12094 (*ssub<modesuffix>): Ditto.
12095 (*ssubl_se): Ditto.
12096 (*cmp_sadd_di): Update split patterns.
12097 (*cmp_sadd_si): Ditto.
12098 (*cmp_sadd_sidi): Ditto.
12099 (*cmp_ssub_di): Ditto.
12100 (*cmp_ssub_si): Ditto.
12101 (*cmp_ssub_sidi): Ditto.
12102 * config/alpha/predicates.md (const23_operand): New predicate.
12103 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12104 Look for ASHIFT, not MULT inner operand.
12105 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12107 2018-01-04 Martin Liska <mliska@suse.cz>
12109 PR gcov-profile/83669
12110 * gcov.c (output_intermediate_file): Add version to intermediate
12112 * doc/gcov.texi: Document new field 'version' in intermediate
12113 file format. Fix location of '-k' option of gcov command.
12115 2018-01-04 Martin Liska <mliska@suse.cz>
12118 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12120 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12122 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12124 2018-01-03 Martin Sebor <msebor@redhat.com>
12126 PR tree-optimization/83655
12127 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12128 checking calls with invalid arguments.
12130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12132 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12133 (vectorizable_mask_load_store): Delete.
12134 (vectorizable_call): Return false for masked loads and stores.
12135 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12136 instead of gimple_assign_rhs1.
12137 (vectorizable_load): Handle IFN_MASK_LOAD.
12138 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12142 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12144 (vectorizable_mask_load_store): ...here.
12145 (vectorizable_load): ...and here.
12147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12149 * tree-vect-stmts.c (vect_build_all_ones_mask)
12150 (vect_build_zero_merge_argument): New functions, split out from...
12151 (vectorizable_load): ...here.
12153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12155 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12157 (vectorizable_mask_load_store): ...here.
12158 (vectorizable_store): ...and here.
12160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12162 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12164 (vectorizable_mask_load_store): ...here.
12166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12168 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12169 (vect_model_store_cost): Take a vec_load_store_type instead of a
12171 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12172 (vect_model_store_cost): Take a vec_load_store_type instead of a
12174 (vectorizable_mask_load_store): Update accordingly.
12175 (vectorizable_store): Likewise.
12176 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12180 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12181 IFN_MASK_LOAD calls here rather than...
12182 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12185 Alan Hayward <alan.hayward@arm.com>
12186 David Sherwood <david.sherwood@arm.com>
12188 * expmed.c (extract_bit_field_1): For vector extracts,
12189 fall back to extract_bit_field_as_subreg if vec_extract
12192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12193 Alan Hayward <alan.hayward@arm.com>
12194 David Sherwood <david.sherwood@arm.com>
12196 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12197 they are variable or constant sized.
12198 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12199 slots for constant-sized data.
12201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12202 Alan Hayward <alan.hayward@arm.com>
12203 David Sherwood <david.sherwood@arm.com>
12205 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12206 handling COND_EXPRs with boolean comparisons, try to find a better
12207 basis for the mask type than the boolean itself.
12209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12211 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12212 is calculated and how it can be overridden.
12213 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12214 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12216 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12220 Alan Hayward <alan.hayward@arm.com>
12221 David Sherwood <david.sherwood@arm.com>
12223 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12224 Remove the mode argument.
12225 (aarch64_simd_valid_immediate): Remove the mode and inverse
12227 * config/aarch64/iterators.md (bitsize): New iterator.
12228 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12229 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12230 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12231 aarch64_simd_valid_immediate.
12232 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12233 (aarch64_reg_or_bic_imm): Likewise.
12234 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12235 with an insn_type enum and msl with a modifier_type enum.
12236 Replace element_width with a scalar_mode. Change the shift
12237 to unsigned int. Add constructors for scalar_float_mode and
12238 scalar_int_mode elements.
12239 (aarch64_vect_float_const_representable_p): Delete.
12240 (aarch64_can_const_movi_rtx_p)
12241 (aarch64_simd_scalar_immediate_valid_for_move)
12242 (aarch64_simd_make_constant): Update call to
12243 aarch64_simd_valid_immediate.
12244 (aarch64_advsimd_valid_immediate_hs): New function.
12245 (aarch64_advsimd_valid_immediate): Likewise.
12246 (aarch64_simd_valid_immediate): Remove mode and inverse
12247 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12248 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12249 and aarch64_float_const_representable_p on the result.
12250 (aarch64_output_simd_mov_immediate): Remove mode argument.
12251 Update call to aarch64_simd_valid_immediate and use of
12252 simd_immediate_info.
12253 (aarch64_output_scalar_simd_mov_immediate): Update call
12256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12257 Alan Hayward <alan.hayward@arm.com>
12258 David Sherwood <david.sherwood@arm.com>
12260 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12261 (mode_nunits): Likewise CONST_MODE_NUNITS.
12262 * machmode.def (ADJUST_NUNITS): Document.
12263 * genmodes.c (mode_data::need_nunits_adj): New field.
12264 (blank_mode): Update accordingly.
12265 (adj_nunits): New variable.
12266 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12268 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12269 listed in adj_nunits.
12270 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12271 listed in adj_nunits. Don't emit case statements for such modes.
12272 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12273 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12274 nothing if adj_nunits is nonnull.
12275 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12276 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12277 (emit_mode_fbit): Update use of print_maybe_const_decl.
12278 (emit_move_size): Likewise. Treat the array as non-const
12280 (emit_mode_adjustments): Handle adj_nunits.
12282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12284 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12285 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12286 (VECTOR_MODES): Use it.
12287 (make_vector_modes): Take the prefix as an argument.
12289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12290 Alan Hayward <alan.hayward@arm.com>
12291 David Sherwood <david.sherwood@arm.com>
12293 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12294 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12295 for MODE_VECTOR_BOOL.
12296 * machmode.def (VECTOR_BOOL_MODE): Document.
12297 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12298 (make_vector_bool_mode): New function.
12299 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12301 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12302 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12304 * stor-layout.c (int_mode_for_mode): Likewise.
12305 * tree.c (build_vector_type_for_mode): Likewise.
12306 * varasm.c (output_constant_pool_2): Likewise.
12307 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12308 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12309 for MODE_VECTOR_BOOL.
12310 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12311 of mode class checks.
12312 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12313 instead of a list of mode class checks.
12314 (expand_vector_scalar_condition): Likewise.
12315 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12318 Alan Hayward <alan.hayward@arm.com>
12319 David Sherwood <david.sherwood@arm.com>
12321 * machmode.h (mode_size): Change from unsigned short to
12323 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12324 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12325 or if measurement_type is not polynomial.
12326 (fixed_size_mode::includes_p): Check for constant-sized modes.
12327 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12328 return a poly_uint16 rather than an unsigned short.
12329 (emit_mode_size): Change the type of mode_size from unsigned short
12330 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12331 (emit_mode_adjustments): Cope with polynomial vector sizes.
12332 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12334 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12336 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12337 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12338 * caller-save.c (setup_save_areas): Likewise.
12339 (replace_reg_with_saved_mem): Likewise.
12340 * calls.c (emit_library_call_value_1): Likewise.
12341 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12342 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12343 (gen_lowpart_for_combine): Likewise.
12344 * convert.c (convert_to_integer_1): Likewise.
12345 * cse.c (equiv_constant, cse_insn): Likewise.
12346 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12347 (cselib_subst_to_values): Likewise.
12348 * dce.c (word_dce_process_block): Likewise.
12349 * df-problems.c (df_word_lr_mark_ref): Likewise.
12350 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12351 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12352 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12353 (rtl_for_decl_location): Likewise.
12354 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12355 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12356 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12357 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12358 (expand_expr_real_1): Likewise.
12359 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12360 (pad_below): Likewise.
12361 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12362 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12363 * ira.c (get_subreg_tracking_sizes): Likewise.
12364 * ira-build.c (ira_create_allocno_objects): Likewise.
12365 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12366 (ira_sort_regnos_for_alter_reg): Likewise.
12367 * ira-costs.c (record_operand_costs): Likewise.
12368 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12369 (resolve_simple_move): Likewise.
12370 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12371 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12372 (lra_constraints): Likewise.
12373 (CONST_POOL_OK_P): Reject variable-sized modes.
12374 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12375 (add_pseudo_to_slot, lra_spill): Likewise.
12376 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12377 * optabs-query.c (get_best_extraction_insn): Likewise.
12378 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12379 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12380 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12381 * recog.c (offsettable_address_addr_space_p): Likewise.
12382 * regcprop.c (maybe_mode_change): Likewise.
12383 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12384 * regrename.c (build_def_use): Likewise.
12385 * regstat.c (dump_reg_info): Likewise.
12386 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12387 (find_reloads, find_reloads_subreg_address): Likewise.
12388 * reload1.c (eliminate_regs_1): Likewise.
12389 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12390 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12391 (simplify_binary_operation_1, simplify_subreg): Likewise.
12392 * targhooks.c (default_function_arg_padding): Likewise.
12393 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12394 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12395 (verify_gimple_assign_ternary): Likewise.
12396 * tree-inline.c (estimate_move_cost): Likewise.
12397 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12398 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12399 (get_address_cost_ainc): Likewise.
12400 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12401 (vect_supportable_dr_alignment): Likewise.
12402 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12403 (vectorizable_reduction): Likewise.
12404 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12405 (vectorizable_operation, vectorizable_load): Likewise.
12406 * tree.c (build_same_sized_truth_vector_type): Likewise.
12407 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12408 * var-tracking.c (emit_note_insn_var_location): Likewise.
12409 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12410 (ADDR_VEC_ALIGN): Likewise.
12412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12413 Alan Hayward <alan.hayward@arm.com>
12414 David Sherwood <david.sherwood@arm.com>
12416 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12418 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12419 or if measurement_type is polynomial.
12420 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12421 * combine.c (make_extraction): Likewise.
12422 * dse.c (find_shift_sequence): Likewise.
12423 * dwarf2out.c (mem_loc_descriptor): Likewise.
12424 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12425 (extract_bit_field, extract_low_bits): Likewise.
12426 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12427 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12428 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12429 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12430 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12431 * reload.c (find_reloads): Likewise.
12432 * reload1.c (alter_reg): Likewise.
12433 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12434 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12435 * tree-if-conv.c (predicate_mem_writes): Likewise.
12436 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12437 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12438 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12439 * valtrack.c (dead_debug_insert_temp): Likewise.
12440 * varasm.c (mergeable_constant_section): Likewise.
12441 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12444 Alan Hayward <alan.hayward@arm.com>
12445 David Sherwood <david.sherwood@arm.com>
12447 * expr.c (expand_assignment): Cope with polynomial mode sizes
12448 when assigning to a CONCAT.
12450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12451 Alan Hayward <alan.hayward@arm.com>
12452 David Sherwood <david.sherwood@arm.com>
12454 * machmode.h (mode_precision): Change from unsigned short to
12456 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12458 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12459 or if measurement_type is not polynomial.
12460 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12461 in which the mode is already known to be a scalar_int_mode.
12462 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12463 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12465 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12466 for GET_MODE_PRECISION.
12467 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12468 for GET_MODE_PRECISION.
12469 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12471 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12472 (expand_field_assignment, make_extraction): Likewise.
12473 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12474 (get_last_value): Likewise.
12475 * convert.c (convert_to_integer_1): Likewise.
12476 * cse.c (cse_insn): Likewise.
12477 * expr.c (expand_expr_real_1): Likewise.
12478 * lra-constraints.c (simplify_operand_subreg): Likewise.
12479 * optabs-query.c (can_atomic_load_p): Likewise.
12480 * optabs.c (expand_atomic_load): Likewise.
12481 (expand_atomic_store): Likewise.
12482 * ree.c (combine_reaching_defs): Likewise.
12483 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12484 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12485 * tree.h (type_has_mode_precision_p): Likewise.
12486 * ubsan.c (instrument_si_overflow): Likewise.
12488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12489 Alan Hayward <alan.hayward@arm.com>
12490 David Sherwood <david.sherwood@arm.com>
12492 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12493 polynomial numbers of units.
12494 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12495 (valid_vector_subparts_p): New function.
12496 (build_vector_type): Remove temporary shim and take the number
12497 of units as a poly_uint64 rather than an int.
12498 (build_opaque_vector_type): Take the number of units as a
12499 poly_uint64 rather than an int.
12500 * tree.c (build_vector_from_ctor): Handle polynomial
12501 TYPE_VECTOR_SUBPARTS.
12502 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12503 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12504 (build_vector_from_val): If the number of units is variable,
12505 use build_vec_duplicate_cst for constant operands and
12506 VEC_DUPLICATE_EXPR otherwise.
12507 (make_vector_type): Remove temporary is_constant ().
12508 (build_vector_type, build_opaque_vector_type): Take the number of
12509 units as a poly_uint64 rather than an int.
12510 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12512 * cfgexpand.c (expand_debug_expr): Likewise.
12513 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12514 (store_constructor, expand_expr_real_1): Likewise.
12515 (const_scalar_mask_from_tree): Likewise.
12516 * fold-const-call.c (fold_const_reduction): Likewise.
12517 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12518 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12519 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12520 (fold_relational_const): Likewise.
12521 (native_interpret_vector): Likewise. Change the size from an
12522 int to an unsigned int.
12523 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12524 TYPE_VECTOR_SUBPARTS.
12525 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12526 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12527 duplicating a non-constant operand into a variable-length vector.
12528 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12529 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12530 * ipa-icf.c (sem_variable::equals): Likewise.
12531 * match.pd: Likewise.
12532 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12533 * print-tree.c (print_node): Likewise.
12534 * stor-layout.c (layout_type): Likewise.
12535 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12536 * tree-cfg.c (verify_gimple_comparison): Likewise.
12537 (verify_gimple_assign_binary): Likewise.
12538 (verify_gimple_assign_ternary): Likewise.
12539 (verify_gimple_assign_single): Likewise.
12540 * tree-pretty-print.c (dump_generic_node): Likewise.
12541 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12542 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12543 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12544 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12545 (vect_shift_permute_load_chain): Likewise.
12546 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12547 (expand_vector_condition, optimize_vector_constructor): Likewise.
12548 (lower_vec_perm, get_compute_type): Likewise.
12549 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12550 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12551 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12552 (vect_recog_mask_conversion_pattern): Likewise.
12553 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12554 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12555 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12556 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12557 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12558 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12559 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12560 (supportable_widening_operation): Likewise.
12561 (supportable_narrowing_operation): Likewise.
12562 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12564 * varasm.c (output_constant): Likewise.
12566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12567 Alan Hayward <alan.hayward@arm.com>
12568 David Sherwood <david.sherwood@arm.com>
12570 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12571 so that both the length == 3 and length != 3 cases set up their
12572 own permute vectors. Add comments explaining why we know the
12573 number of elements is constant.
12574 (vect_permute_load_chain): Likewise.
12576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12577 Alan Hayward <alan.hayward@arm.com>
12578 David Sherwood <david.sherwood@arm.com>
12580 * machmode.h (mode_nunits): Change from unsigned char to
12582 (ONLY_FIXED_SIZE_MODES): New macro.
12583 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12584 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12585 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12587 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12588 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12589 or if measurement_type is not polynomial.
12590 * genmodes.c (ZERO_COEFFS): New macro.
12591 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12593 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12594 Use ZERO_COEFFS when emitting initializers.
12595 * data-streamer.h (bp_pack_poly_value): New function.
12596 (bp_unpack_poly_value): Likewise.
12597 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12598 for GET_MODE_NUNITS.
12599 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12600 for GET_MODE_NUNITS.
12601 * tree.c (make_vector_type): Remove temporary shim and make
12602 the real function take the number of units as a poly_uint64
12603 rather than an int.
12604 (build_vector_type_for_mode): Handle polynomial nunits.
12605 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12606 * emit-rtl.c (const_vec_series_p_1): Likewise.
12607 (gen_rtx_CONST_VECTOR): Likewise.
12608 * fold-const.c (test_vec_duplicate_folding): Likewise.
12609 * genrecog.c (validate_pattern): Likewise.
12610 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12611 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12612 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12613 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12614 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12615 * rtlanal.c (subreg_get_info): Likewise.
12616 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12617 (vect_grouped_load_supported): Likewise.
12618 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12619 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12620 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12621 (simplify_const_unary_operation, simplify_binary_operation_1)
12622 (simplify_const_binary_operation, simplify_ternary_operation)
12623 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12624 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12625 instead of CONST_VECTOR_NUNITS.
12626 * varasm.c (output_constant_pool_2): Likewise.
12627 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12628 explicit-encoded elements in the XVEC for variable-length vectors.
12630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12632 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12635 Alan Hayward <alan.hayward@arm.com>
12636 David Sherwood <david.sherwood@arm.com>
12638 * coretypes.h (fixed_size_mode): Declare.
12639 (fixed_size_mode_pod): New typedef.
12640 * builtins.h (target_builtins::x_apply_args_mode)
12641 (target_builtins::x_apply_result_mode): Change type to
12642 fixed_size_mode_pod.
12643 * builtins.c (apply_args_size, apply_result_size, result_vector)
12644 (expand_builtin_apply_args_1, expand_builtin_apply)
12645 (expand_builtin_return): Update accordingly.
12647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12649 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12650 * cselib.c (cselib_hash_rtx): Likewise.
12651 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12652 CONST_VECTOR encoding.
12654 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12655 Jeff Law <law@redhat.com>
12658 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12659 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12660 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12661 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12664 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12665 explicitly probe *sp in a noreturn function if there were any callee
12666 register saves or frame pointer is needed.
12668 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12671 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12672 BLKmode for ternary, binary or unary expressions.
12675 * var-tracking.c (delete_vta_debug_insn): New inline function.
12676 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12677 insns from get_insns () to NULL instead of each bb separately.
12678 Use delete_vta_debug_insn. No longer static.
12679 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12680 delete_vta_debug_insns callers.
12681 * rtl.h (delete_vta_debug_insns): Declare.
12682 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12683 instead of variable_tracking_main.
12685 2018-01-03 Martin Sebor <msebor@redhat.com>
12687 PR tree-optimization/83603
12688 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12689 arguments past the endof the argument list in functions declared
12690 without a prototype.
12691 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12692 Avoid checking when arguments are null.
12694 2018-01-03 Martin Sebor <msebor@redhat.com>
12697 * doc/extend.texi (attribute const): Fix a typo.
12698 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12699 issuing -Wsuggest-attribute for void functions.
12701 2018-01-03 Martin Sebor <msebor@redhat.com>
12703 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12704 offset_int::from instead of wide_int::to_shwi.
12705 (maybe_diag_overlap): Remove assertion.
12706 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12707 * gimple-ssa-sprintf.c (format_directive): Same.
12708 (parse_directive): Same.
12709 (sprintf_dom_walker::compute_format_length): Same.
12710 (try_substitute_return_value): Same.
12712 2018-01-03 Jeff Law <law@redhat.com>
12714 PR middle-end/83654
12715 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12716 non-constant residual for zero at runtime and avoid probing in
12717 that case. Reorganize code for trailing problem to mirror handling
12720 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12722 PR tree-optimization/83501
12723 * tree-ssa-strlen.c (get_string_cst): New.
12724 (handle_char_store): Call get_string_cst.
12726 2018-01-03 Martin Liska <mliska@suse.cz>
12728 PR tree-optimization/83593
12729 * tree-ssa-strlen.c: Include tree-cfg.h.
12730 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12731 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12732 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12734 (strlen_dom_walker::before_dom_children): Call
12735 gimple_purge_dead_eh_edges. Dump tranformation with details
12737 (strlen_dom_walker::before_dom_children): Update call by adding
12738 new argument cleanup_eh.
12739 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12741 2018-01-03 Martin Liska <mliska@suse.cz>
12744 * cif-code.def (VARIADIC_THUNK): New enum value.
12745 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12748 2018-01-03 Jan Beulich <jbeulich@suse.com>
12750 * sse.md (mov<mode>_internal): Tighten condition for when to use
12751 vmovdqu<ssescalarsize> for TI and OI modes.
12753 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12755 Update copyright years.
12757 2018-01-03 Martin Liska <mliska@suse.cz>
12760 * ipa-visibility.c (function_and_variable_visibility): Skip
12761 functions with noipa attribure.
12763 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12765 * gcc.c (process_command): Update copyright notice dates.
12766 * gcov-dump.c (print_version): Ditto.
12767 * gcov.c (print_version): Ditto.
12768 * gcov-tool.c (print_version): Ditto.
12769 * gengtype.c (create_file): Ditto.
12770 * doc/cpp.texi: Bump @copying's copyright year.
12771 * doc/cppinternals.texi: Ditto.
12772 * doc/gcc.texi: Ditto.
12773 * doc/gccint.texi: Ditto.
12774 * doc/gcov.texi: Ditto.
12775 * doc/install.texi: Ditto.
12776 * doc/invoke.texi: Ditto.
12778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12780 * vector-builder.h (vector_builder::m_full_nelts): Change from
12781 unsigned int to poly_uint64.
12782 (vector_builder::full_nelts): Update prototype accordingly.
12783 (vector_builder::new_vector): Likewise.
12784 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12785 (vector_builder::operator ==): Likewise.
12786 (vector_builder::finalize): Likewise.
12787 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12788 Take the number of elements as a poly_uint64 rather than an
12790 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12791 from unsigned int to poly_uint64.
12792 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12793 (vec_perm_indices::new_vector): Likewise.
12794 (vec_perm_indices::length): Likewise.
12795 (vec_perm_indices::nelts_per_input): Likewise.
12796 (vec_perm_indices::input_nelts): Likewise.
12797 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12798 number of elements per input as a poly_uint64 rather than an
12799 unsigned int. Use the original encoding for variable-length
12800 vectors, rather than clamping each individual element.
12801 For the second and subsequent elements in each pattern,
12802 clamp the step and base before clamping their sum.
12803 (vec_perm_indices::series_p): Handle polynomial element counts.
12804 (vec_perm_indices::all_in_range_p): Likewise.
12805 (vec_perm_indices_to_tree): Likewise.
12806 (vec_perm_indices_to_rtx): Likewise.
12807 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12808 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12809 (tree_vector_builder::new_binary_operation): Handle polynomial
12810 element counts. Return false if we need to know the number
12811 of elements at compile time.
12812 * fold-const.c (fold_vec_perm): Punt if the number of elements
12813 isn't known at compile time.
12815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12817 * vec-perm-indices.h (vec_perm_builder): Change element type
12818 from HOST_WIDE_INT to poly_int64.
12819 (vec_perm_indices::element_type): Update accordingly.
12820 (vec_perm_indices::clamp): Handle polynomial element_types.
12821 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12822 (vec_perm_indices::all_in_range_p): Likewise.
12823 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12825 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12826 polynomial vec_perm_indices element types.
12827 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12828 * fold-const.c (fold_vec_perm): Likewise.
12829 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12830 * tree-vect-generic.c (lower_vec_perm): Likewise.
12831 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12832 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12833 element type to HOST_WIDE_INT.
12835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12836 Alan Hayward <alan.hayward@arm.com>
12837 David Sherwood <david.sherwood@arm.com>
12839 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12840 rather than an int. Use plus_constant.
12841 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12842 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12845 Alan Hayward <alan.hayward@arm.com>
12846 David Sherwood <david.sherwood@arm.com>
12848 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12849 a HOST_WIDE_INT to a poly_int64.
12851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12852 Alan Hayward <alan.hayward@arm.com>
12853 David Sherwood <david.sherwood@arm.com>
12855 * calls.c (load_register_parameters): Cope with polynomial
12856 mode sizes. Require a constant size for BLKmode parameters
12857 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12858 forces a parameter to be padded at the lsb end in order to
12859 fill a complete number of words, require the parameter size
12860 to be ordered wrt UNITS_PER_WORD.
12862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12863 Alan Hayward <alan.hayward@arm.com>
12864 David Sherwood <david.sherwood@arm.com>
12866 * reload1.c (spill_stack_slot_width): Change element type
12867 from unsigned int to poly_uint64_pod.
12868 (alter_reg): Treat mode sizes as polynomial.
12870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12871 Alan Hayward <alan.hayward@arm.com>
12872 David Sherwood <david.sherwood@arm.com>
12874 * reload.c (complex_word_subreg_p): New function.
12875 (reload_inner_reg_of_subreg, push_reload): Use it.
12877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12878 Alan Hayward <alan.hayward@arm.com>
12879 David Sherwood <david.sherwood@arm.com>
12881 * lra-constraints.c (process_alt_operands): Reject matched
12882 operands whose sizes aren't ordered.
12883 (match_reload): Refer to this check here.
12885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12886 Alan Hayward <alan.hayward@arm.com>
12887 David Sherwood <david.sherwood@arm.com>
12889 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12890 that the mode size is in the set {1, 2, 4, 8, 16}.
12892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12893 Alan Hayward <alan.hayward@arm.com>
12894 David Sherwood <david.sherwood@arm.com>
12896 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12897 Use plus_constant instead of gen_rtx_PLUS.
12899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12900 Alan Hayward <alan.hayward@arm.com>
12901 David Sherwood <david.sherwood@arm.com>
12903 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12904 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12905 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12906 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12907 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12908 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12909 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12910 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12911 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12912 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12914 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12915 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12916 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12917 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12918 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12919 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12920 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12921 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12922 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12923 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12925 * expr.c (emit_move_resolve_push): Treat the input and result
12926 of PUSH_ROUNDING as a poly_int64.
12927 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12928 (emit_push_insn): Likewise.
12929 * lra-eliminations.c (mark_not_eliminable): Likewise.
12930 * recog.c (push_operand): Likewise.
12931 * reload1.c (elimination_effects): Likewise.
12932 * rtlanal.c (nonzero_bits1): Likewise.
12933 * calls.c (store_one_arg): Likewise. Require the padding to be
12934 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12941 Use plus_constant instead of gen_rtx_PLUS.
12943 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12944 Alan Hayward <alan.hayward@arm.com>
12945 David Sherwood <david.sherwood@arm.com>
12947 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12948 rather than an int.
12950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12951 Alan Hayward <alan.hayward@arm.com>
12952 David Sherwood <david.sherwood@arm.com>
12954 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12955 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12956 via stack temporaries. Treat the mode size as polynomial too.
12958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12959 Alan Hayward <alan.hayward@arm.com>
12960 David Sherwood <david.sherwood@arm.com>
12962 * expr.c (expand_expr_real_2): When handling conversions involving
12963 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12964 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12965 as a poly_uint64 too.
12967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12968 Alan Hayward <alan.hayward@arm.com>
12969 David Sherwood <david.sherwood@arm.com>
12971 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12974 Alan Hayward <alan.hayward@arm.com>
12975 David Sherwood <david.sherwood@arm.com>
12977 * combine.c (can_change_dest_mode): Handle polynomial
12978 REGMODE_NATURAL_SIZE.
12979 * expmed.c (store_bit_field_1): Likewise.
12980 * expr.c (store_constructor): Likewise.
12981 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12982 and polynomial REGMODE_NATURAL_SIZE.
12983 (gen_lowpart_common): Likewise.
12984 * reginfo.c (record_subregs_of_mode): Likewise.
12985 * rtlanal.c (read_modify_subreg_p): Likewise.
12987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12988 Alan Hayward <alan.hayward@arm.com>
12989 David Sherwood <david.sherwood@arm.com>
12991 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12992 numbers of elements.
12994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12995 Alan Hayward <alan.hayward@arm.com>
12996 David Sherwood <david.sherwood@arm.com>
12998 * match.pd: Cope with polynomial numbers of vector elements.
13000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13001 Alan Hayward <alan.hayward@arm.com>
13002 David Sherwood <david.sherwood@arm.com>
13004 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13005 in a POINTER_PLUS_EXPR.
13007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13008 Alan Hayward <alan.hayward@arm.com>
13009 David Sherwood <david.sherwood@arm.com>
13011 * omp-simd-clone.c (simd_clone_subparts): New function.
13012 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13013 (ipa_simd_modify_function_body): Likewise.
13015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13016 Alan Hayward <alan.hayward@arm.com>
13017 David Sherwood <david.sherwood@arm.com>
13019 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13020 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13021 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13022 (expand_vector_condition, vector_element): Likewise.
13023 (subparts_gt): New function.
13024 (get_compute_type): Use subparts_gt.
13025 (count_type_subparts): Delete.
13026 (expand_vector_operations_1): Use subparts_gt instead of
13027 count_type_subparts.
13029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13030 Alan Hayward <alan.hayward@arm.com>
13031 David Sherwood <david.sherwood@arm.com>
13033 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13034 (vect_compile_time_alias): ...this new function. Do the calculation
13035 on poly_ints rather than trees.
13036 (vect_prune_runtime_alias_test_list): Update call accordingly.
13038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13039 Alan Hayward <alan.hayward@arm.com>
13040 David Sherwood <david.sherwood@arm.com>
13042 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13044 (vect_schedule_slp_instance): Likewise.
13046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13047 Alan Hayward <alan.hayward@arm.com>
13048 David Sherwood <david.sherwood@arm.com>
13050 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13051 constant and extern definitions for variable-length vectors.
13052 (vect_get_constant_vectors): Note that the number of units
13053 is known to be constant.
13055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13056 Alan Hayward <alan.hayward@arm.com>
13057 David Sherwood <david.sherwood@arm.com>
13059 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13060 of units as polynomial. Choose between WIDE and NARROW based
13063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13064 Alan Hayward <alan.hayward@arm.com>
13065 David Sherwood <david.sherwood@arm.com>
13067 * tree-vect-stmts.c (simd_clone_subparts): New function.
13068 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13071 Alan Hayward <alan.hayward@arm.com>
13072 David Sherwood <david.sherwood@arm.com>
13074 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13075 vectors as polynomial. Use build_index_vector for
13076 IFN_GOMP_SIMD_LANE.
13078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13079 Alan Hayward <alan.hayward@arm.com>
13080 David Sherwood <david.sherwood@arm.com>
13082 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13083 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13084 for variable-length vectors.
13085 (vectorizable_mask_load_store): Treat the number of units as
13086 polynomial, asserting that it is constant if the condition has
13087 already been enforced.
13088 (vectorizable_store, vectorizable_load): Likewise.
13090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13091 Alan Hayward <alan.hayward@arm.com>
13092 David Sherwood <david.sherwood@arm.com>
13094 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13095 of units as polynomial. Punt if we can't tell at compile time
13096 which vector contains the final result.
13098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13099 Alan Hayward <alan.hayward@arm.com>
13100 David Sherwood <david.sherwood@arm.com>
13102 * tree-vect-loop.c (vectorizable_induction): Treat the number
13103 of units as polynomial. Punt on SLP inductions. Use an integer
13104 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13105 cast of such a series for variable-length floating-point
13108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13109 Alan Hayward <alan.hayward@arm.com>
13110 David Sherwood <david.sherwood@arm.com>
13112 * tree.h (build_index_vector): Declare.
13113 * tree.c (build_index_vector): New function.
13114 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13115 of units as polynomial, forcibly converting it to a constant if
13116 vectorizable_reduction has already enforced the condition.
13117 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13118 to create a {1,2,3,...} vector.
13119 (vectorizable_reduction): Treat the number of units as polynomial.
13120 Choose vectype_in based on the largest scalar element size rather
13121 than the smallest number of units. Enforce the restrictions
13124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13125 Alan Hayward <alan.hayward@arm.com>
13126 David Sherwood <david.sherwood@arm.com>
13128 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13129 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13136 * target.def (autovectorize_vector_sizes): Return the vector sizes
13137 by pointer, using vector_sizes rather than a bitmask.
13138 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13139 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13140 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13142 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13143 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13144 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13145 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13146 * omp-general.c (omp_max_vf): Likewise.
13147 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13148 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13149 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13150 * tree-vect-slp.c (vect_slp_bb): Likewise.
13151 * doc/tm.texi: Regenerate.
13152 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13154 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13155 the vector size as a poly_uint64 rather than an unsigned int.
13156 (current_vector_size): Change from an unsigned int to a poly_uint64.
13157 (get_vectype_for_scalar_type): Update accordingly.
13158 * tree.h (build_truth_vector_type): Take the size and number of
13159 units as a poly_uint64 rather than an unsigned int.
13160 (build_vector_type): Add a temporary overload that takes
13161 the number of units as a poly_uint64 rather than an unsigned int.
13162 * tree.c (make_vector_type): Likewise.
13163 (build_truth_vector_type): Take the number of units as a poly_uint64
13164 rather than an unsigned int.
13166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13167 Alan Hayward <alan.hayward@arm.com>
13168 David Sherwood <david.sherwood@arm.com>
13170 * target.def (get_mask_mode): Take the number of units and length
13171 as poly_uint64s rather than unsigned ints.
13172 * targhooks.h (default_get_mask_mode): Update accordingly.
13173 * targhooks.c (default_get_mask_mode): Likewise.
13174 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13175 * doc/tm.texi: Regenerate.
13177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13178 Alan Hayward <alan.hayward@arm.com>
13179 David Sherwood <david.sherwood@arm.com>
13181 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13182 * omp-general.c (omp_max_vf): Likewise.
13183 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13184 (expand_omp_simd): Handle polynomial safelen.
13185 * omp-low.c (omplow_simd_context): Add a default constructor.
13186 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13187 (lower_rec_simd_input_clauses): Update accordingly.
13188 (lower_rec_input_clauses): Likewise.
13190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13191 Alan Hayward <alan.hayward@arm.com>
13192 David Sherwood <david.sherwood@arm.com>
13194 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13195 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13196 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13197 (vect_analyze_slp_cost): Likewise.
13198 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13199 (vect_model_load_cost): Likewise.
13201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13202 Alan Hayward <alan.hayward@arm.com>
13203 David Sherwood <david.sherwood@arm.com>
13205 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13206 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13207 from an unsigned int * to a poly_uint64_pod *.
13208 (calculate_unrolling_factor): New function.
13209 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13212 Alan Hayward <alan.hayward@arm.com>
13213 David Sherwood <david.sherwood@arm.com>
13215 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13216 from an unsigned int to a poly_uint64.
13217 (_loop_vec_info::slp_unrolling_factor): Likewise.
13218 (_loop_vec_info::vectorization_factor): Change from an int
13220 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13221 (vect_get_num_vectors): New function.
13222 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13223 (vect_get_num_copies): Use vect_get_num_vectors.
13224 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13225 to an unsigned int *.
13226 (vect_analyze_data_refs): Change min_vf from an int * to a
13228 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13229 than an unsigned HOST_WIDE_INT.
13230 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13231 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13232 to an unsigned int *.
13233 (vect_analyze_data_ref_dependences): Likewise.
13234 (vect_compute_data_ref_alignment): Handle polynomial vf.
13235 (vect_enhance_data_refs_alignment): Likewise.
13236 (vect_prune_runtime_alias_test_list): Likewise.
13237 (vect_shift_permute_load_chain): Likewise.
13238 (vect_supportable_dr_alignment): Likewise.
13239 (dependence_distance_ge_vf): Take the vectorization factor as a
13240 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13241 (vect_analyze_data_refs): Change min_vf from an int * to a
13243 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13244 vfm1 as a poly_uint64 rather than an int. Make the same change
13245 for the returned bound_scalar.
13246 (vect_gen_vector_loop_niters): Handle polynomial vf.
13247 (vect_do_peeling): Likewise. Update call to
13248 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13249 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13251 * tree-vect-loop.c (vect_determine_vectorization_factor)
13252 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13253 (vect_get_known_peeling_cost): Likewise.
13254 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13255 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13256 (vect_transform_loop): Likewise. Use the lowest possible VF when
13257 updating the upper bounds of the loop.
13258 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13259 rather than an int.
13260 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13261 polynomial unroll factors.
13262 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13263 (vect_make_slp_decision): Likewise.
13264 (vect_supported_load_permutation_p): Likewise, and polynomial
13266 (vect_analyze_slp_cost): Handle polynomial vf.
13267 (vect_slp_analyze_node_operations): Likewise.
13268 (vect_slp_analyze_bb_1): Likewise.
13269 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13270 than an unsigned HOST_WIDE_INT.
13271 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13272 (vectorizable_load): Handle polynomial vf.
13273 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13275 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13278 Alan Hayward <alan.hayward@arm.com>
13279 David Sherwood <david.sherwood@arm.com>
13281 * match.pd: Handle bit operations involving three constants
13282 and try to fold one pair.
13284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13286 * tree-vect-loop-manip.c: Include gimple-fold.h.
13287 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13288 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13289 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13290 Add a path that uses a step of VF instead of 1, but disable it
13292 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13293 and niters_no_overflow parameters. Update calls to
13294 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13295 Create a new SSA name if the latter choses to use a ste other
13296 than zero, and return it via niters_vector_mult_vf_var.
13297 * tree-vect-loop.c (vect_transform_loop): Update calls to
13298 vect_do_peeling, vect_gen_vector_loop_niters and
13299 slpeel_make_loop_iterate_ntimes.
13300 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13301 (vect_gen_vector_loop_niters): Update declarations after above changes.
13303 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13305 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13306 128-bit round to integer instructions.
13307 (ceil<mode>2): Likewise.
13308 (btrunc<mode>2): Likewise.
13309 (round<mode>2): Likewise.
13311 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13313 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13314 unaligned VSX load/store on P8/P9.
13315 (expand_block_clear): Allow the use of unaligned VSX
13316 load/store on P8/P9.
13318 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13320 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13322 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13323 swap associated with both a load and a store.
13325 2018-01-02 Andrew Waterman <andrew@sifive.com>
13327 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13328 * config/riscv/riscv.md (clear_cache): Use it.
13330 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13332 * web.c: Remove out-of-date comment.
13334 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13336 * expr.c (fixup_args_size_notes): Check that any existing
13337 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13338 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13339 (emit_single_push_insn): ...here.
13341 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13343 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13344 (const_vector_encoded_nelts): New function.
13345 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13346 (const_vector_int_elt, const_vector_elt): Declare.
13347 * emit-rtl.c (const_vector_int_elt_1): New function.
13348 (const_vector_elt): Likewise.
13349 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13350 of CONST_VECTOR_ELT.
13352 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13354 * expr.c: Include rtx-vector-builder.h.
13355 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13356 directly on the tree encoding.
13357 (const_vector_from_tree): Likewise.
13358 * optabs.c: Include rtx-vector-builder.h.
13359 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13360 sequence of "u" values.
13361 * vec-perm-indices.c: Include rtx-vector-builder.h.
13362 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13363 directly on the vec_perm_indices encoding.
13365 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13367 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13368 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13369 * rtx-vector-builder.h: New file.
13370 * rtx-vector-builder.c: Likewise.
13371 * rtl.h (rtx_def::u2): Add a const_vector field.
13372 (CONST_VECTOR_NPATTERNS): New macro.
13373 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13374 (CONST_VECTOR_DUPLICATE_P): Likewise.
13375 (CONST_VECTOR_STEPPED_P): Likewise.
13376 (CONST_VECTOR_ENCODED_ELT): Likewise.
13377 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13378 (unwrap_const_vec_duplicate): Likewise.
13379 (const_vec_series_p): Check for a non-duplicated vector encoding.
13380 Say that the function only returns true for integer vectors.
13381 * emit-rtl.c: Include rtx-vector-builder.h.
13382 (gen_const_vec_duplicate_1): Delete.
13383 (gen_const_vector): Call gen_const_vec_duplicate instead of
13384 gen_const_vec_duplicate_1.
13385 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13386 (gen_const_vec_duplicate): Use rtx_vector_builder.
13387 (gen_const_vec_series): Likewise.
13388 (gen_rtx_CONST_VECTOR): Likewise.
13389 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13390 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13391 Build a new vector rather than modifying a CONST_VECTOR in-place.
13392 (handle_special_swappables): Update call accordingly.
13393 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13394 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13395 Build a new vector rather than modifying a CONST_VECTOR in-place.
13396 (handle_special_swappables): Update call accordingly.
13398 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13400 * simplify-rtx.c (simplify_const_binary_operation): Use
13401 CONST_VECTOR_ELT instead of XVECEXP.
13403 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13405 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13406 the selector elements to be different from the data elements
13407 if the selector is a VECTOR_CST.
13408 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13409 ssizetype for the selector.
13411 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13413 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13414 before testing each element individually.
13415 * tree-vect-generic.c (lower_vec_perm): Likewise.
13417 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13419 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13420 * selftest-run-tests.c (selftest::run_tests): Call it.
13421 * vector-builder.h (vector_builder::operator ==): New function.
13422 (vector_builder::operator !=): Likewise.
13423 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13424 (vec_perm_indices::all_from_input_p): New function.
13425 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13426 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13427 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13428 instead of reading the VECTOR_CST directly. Detect whether both
13429 vector inputs are the same before constructing the vec_perm_indices,
13430 and update the number of inputs argument accordingly. Use the
13431 utility functions added above. Only construct sel2 if we need to.
13433 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13435 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13436 the broadcast of the low byte.
13437 (expand_mult_highpart): Use an explicit encoding for the permutes.
13438 * optabs-query.c (can_mult_highpart_p): Likewise.
13439 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13440 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13441 (vectorizable_bswap): Likewise.
13442 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13443 explicit encoding for the power-of-2 permutes.
13444 (vect_permute_store_chain): Likewise.
13445 (vect_grouped_load_supported): Likewise.
13446 (vect_permute_load_chain): Likewise.
13448 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13450 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13451 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13452 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13453 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13454 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13455 (vect_gen_perm_mask_any): Likewise.
13457 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13459 * int-vector-builder.h: New file.
13460 * vec-perm-indices.h: Include int-vector-builder.h.
13461 (vec_perm_indices): Redefine as an int_vector_builder.
13462 (auto_vec_perm_indices): Delete.
13463 (vec_perm_builder): Redefine as a stand-alone class.
13464 (vec_perm_indices::vec_perm_indices): New function.
13465 (vec_perm_indices::clamp): Likewise.
13466 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13467 (vec_perm_indices::new_vector): New function.
13468 (vec_perm_indices::new_expanded_vector): Update for new
13469 vec_perm_indices class.
13470 (vec_perm_indices::rotate_inputs): New function.
13471 (vec_perm_indices::all_in_range_p): Operate directly on the
13472 encoded form, without computing elided elements.
13473 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13474 encoding. Update for new vec_perm_indices class.
13475 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13476 the given vec_perm_builder.
13477 (expand_vec_perm_var): Update vec_perm_builder constructor.
13478 (expand_mult_highpart): Use vec_perm_builder instead of
13479 auto_vec_perm_indices.
13480 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13481 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13482 or double series encoding as appropriate.
13483 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13484 vec_perm_indices instead of auto_vec_perm_indices.
13485 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13486 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13487 (vect_permute_store_chain): Likewise.
13488 (vect_grouped_load_supported): Likewise.
13489 (vect_permute_load_chain): Likewise.
13490 (vect_shift_permute_load_chain): Likewise.
13491 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13492 (vect_transform_slp_perm_load): Likewise.
13493 (vect_schedule_slp_instance): Likewise.
13494 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13495 (vectorizable_mask_load_store): Likewise.
13496 (vectorizable_bswap): Likewise.
13497 (vectorizable_store): Likewise.
13498 (vectorizable_load): Likewise.
13499 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13500 vec_perm_indices instead of auto_vec_perm_indices. Use
13501 tree_to_vec_perm_builder to read the vector from a tree.
13502 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13503 vec_perm_builder instead of a vec_perm_indices.
13504 (have_whole_vector_shift): Use vec_perm_builder and
13505 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13506 truncation to calc_vec_perm_mask_for_shift.
13507 (vect_create_epilog_for_reduction): Likewise.
13508 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13509 from auto_vec_perm_indices to vec_perm_indices.
13510 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13511 instead of changing individual elements.
13512 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13513 the vector in d.perm.
13514 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13515 from auto_vec_perm_indices to vec_perm_indices.
13516 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13517 instead of changing individual elements.
13518 (arm_vectorize_vec_perm_const): Use new_vector to install
13519 the vector in d.perm.
13520 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13521 Update vec_perm_builder constructor.
13522 (rs6000_expand_interleave): Likewise.
13523 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13524 (rs6000_expand_interleave): Likewise.
13526 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13528 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13529 to qimode could truncate the indices.
13530 * optabs.c (expand_vec_perm_var): Likewise.
13532 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13534 * Makefile.in (OBJS): Add vec-perm-indices.o.
13535 * vec-perm-indices.h: New file.
13536 * vec-perm-indices.c: Likewise.
13537 * target.h (vec_perm_indices): Replace with a forward class
13539 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13540 * optabs.h: Include vec-perm-indices.h.
13541 (expand_vec_perm): Delete.
13542 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13543 (expand_vec_perm_const): Declare.
13544 * target.def (vec_perm_const_ok): Replace with...
13545 (vec_perm_const): ...this new hook.
13546 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13547 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13548 * doc/tm.texi: Regenerate.
13549 * optabs.def (vec_perm_const): Delete.
13550 * doc/md.texi (vec_perm_const): Likewise.
13551 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13552 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13553 expand_vec_perm for constant permutation vectors. Assert that
13554 the mode of variable permutation vectors is the integer equivalent
13555 of the mode that is being permuted.
13556 * optabs-query.h (selector_fits_mode_p): Declare.
13557 * optabs-query.c: Include vec-perm-indices.h.
13558 (selector_fits_mode_p): New function.
13559 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13560 is defined, instead of checking whether the vec_perm_const_optab
13561 exists. Use targetm.vectorize.vec_perm_const instead of
13562 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13563 fit in the vector mode before using a variable permute.
13564 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13565 vec_perm_indices instead of an rtx.
13566 (expand_vec_perm): Replace with...
13567 (expand_vec_perm_const): ...this new function. Take the selector
13568 as a vec_perm_indices rather than an rtx. Also take the mode of
13569 the selector. Update call to shift_amt_for_vec_perm_mask.
13570 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13571 Use vec_perm_indices::new_expanded_vector to expand the original
13572 selector into bytes. Check whether the indices fit in the vector
13573 mode before using a variable permute.
13574 (expand_vec_perm_var): Make global.
13575 (expand_mult_highpart): Use expand_vec_perm_const.
13576 * fold-const.c: Includes vec-perm-indices.h.
13577 * tree-ssa-forwprop.c: Likewise.
13578 * tree-vect-data-refs.c: Likewise.
13579 * tree-vect-generic.c: Likewise.
13580 * tree-vect-loop.c: Likewise.
13581 * tree-vect-slp.c: Likewise.
13582 * tree-vect-stmts.c: Likewise.
13583 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13585 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13586 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13587 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13588 (aarch64_vectorize_vec_perm_const): ...this new function.
13589 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13590 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13591 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13592 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13593 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13594 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13595 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13597 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13598 check for NEON modes.
13599 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13600 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13601 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13602 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13604 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13605 the old VEC_PERM_CONST conditions.
13606 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13607 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13608 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13609 (ia64_vectorize_vec_perm_const_ok): Merge into...
13610 (ia64_vectorize_vec_perm_const): ...this new function.
13611 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13612 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13613 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13614 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13615 * config/mips/mips.c (mips_expand_vec_perm_const)
13616 (mips_vectorize_vec_perm_const_ok): Merge into...
13617 (mips_vectorize_vec_perm_const): ...this new function.
13618 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13619 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13620 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13621 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13622 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13623 (rs6000_expand_vec_perm_const): Delete.
13624 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13626 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13627 (altivec_expand_vec_perm_const_le): Take each operand individually.
13628 Operate on constant selectors rather than rtxes.
13629 (altivec_expand_vec_perm_const): Likewise. Update call to
13630 altivec_expand_vec_perm_const_le.
13631 (rs6000_expand_vec_perm_const): Delete.
13632 (rs6000_vectorize_vec_perm_const_ok): Delete.
13633 (rs6000_vectorize_vec_perm_const): New function.
13634 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13635 an element count and rtx array.
13636 (rs6000_expand_extract_even): Update call accordingly.
13637 (rs6000_expand_interleave): Likewise.
13638 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13639 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13640 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13641 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13642 (rs6000_expand_vec_perm_const): Delete.
13643 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13644 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13645 (altivec_expand_vec_perm_const_le): Take each operand individually.
13646 Operate on constant selectors rather than rtxes.
13647 (altivec_expand_vec_perm_const): Likewise. Update call to
13648 altivec_expand_vec_perm_const_le.
13649 (rs6000_expand_vec_perm_const): Delete.
13650 (rs6000_vectorize_vec_perm_const_ok): Delete.
13651 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13652 reference to the SPE evmerge intructions.
13653 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13654 an element count and rtx array.
13655 (rs6000_expand_extract_even): Update call accordingly.
13656 (rs6000_expand_interleave): Likewise.
13657 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13658 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13660 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13662 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13664 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13665 vector mode and that that mode matches the mode of the data
13667 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13668 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13669 directly using expand_vec_perm_1 when forcing selectors into
13671 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13673 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13675 * optabs-query.h (can_vec_perm_p): Delete.
13676 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13677 * optabs-query.c (can_vec_perm_p): Split into...
13678 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13679 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13680 particular selector is valid.
13681 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13682 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13683 (vect_grouped_load_supported): Likewise.
13684 (vect_shift_permute_load_chain): Likewise.
13685 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13686 (vect_transform_slp_perm_load): Likewise.
13687 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13688 (vectorizable_bswap): Likewise.
13689 (vect_gen_perm_mask_checked): Likewise.
13690 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13691 implementations of variable permutation vectors into account
13692 when deciding which selector to use.
13693 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13694 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13695 with a false third argument.
13696 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13697 to test whether the constant selector is valid and can_vec_perm_var_p
13698 to test whether a variable selector is valid.
13700 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13702 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13703 * optabs-query.c (can_vec_perm_p): Likewise.
13704 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13705 instead of vec_perm_indices.
13706 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13707 (vect_gen_perm_mask_checked): Likewise,
13708 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13709 (vect_gen_perm_mask_checked): Likewise,
13711 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13713 * optabs-query.h (qimode_for_vec_perm): Declare.
13714 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13715 (qimode_for_vec_perm): ...this new function.
13716 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13718 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13720 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13721 does not have a conditional at the top.
13723 2018-01-02 Richard Biener <rguenther@suse.de>
13725 * ipa-inline.c (big_speedup_p): Fix expression.
13727 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13730 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13733 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13737 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13738 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13739 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13740 cond_taken_branch_cost 3->4.
13742 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13744 PR tree-optimization/83581
13745 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13746 TODO_cleanup_cfg if any changes have been made.
13748 PR middle-end/83608
13749 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13750 convert_modes if target mode has the right side, but different mode
13753 PR middle-end/83609
13754 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13755 last argument when extracting from CONCAT. If either from_real or
13756 from_imag is NULL, use expansion through memory. If result is not
13757 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13758 the parts directly to inner mode, if even that fails, use expansion
13761 PR middle-end/83623
13762 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13763 check for bswap in mode rather than HImode and use that in expand_unop
13766 Copyright (C) 2018 Free Software Foundation, Inc.
13768 Copying and distribution of this file, with or without modification,
13769 are permitted in any medium without royalty provided the copyright
13770 notice and this notice are preserved.