1 2018-05-22 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/85834
4 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
5 non-constant and non-zero memset arguments.
7 2018-05-22 Martin Liska <mliska@suse.cz>
10 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
12 2018-05-22 Richard Biener <rguenther@suse.de>
14 PR tree-optimization/85863
15 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
16 comparisons when vectype is specified.
17 (vectorizable_condition): Do not specify vectype for
18 vect_is_simple_cond when SLP vectorizing.
20 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
23 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
24 define __ibm128 as long double.
25 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
26 as a distinct type when IEEE 128-bit support is enabled.
27 (init_float128_ieee): Fix up conversions between IFmode and IEEE
28 128-bit types to use the correct functions.
29 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
30 convert between 128-bit floating point types that have different
31 modes but the same representation, instead of using gen_lowpart to
33 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
35 (IFKF_reg): New attributes to give the register constraints for
37 (extend<mode>tf2_internal): New insns to mark an explicit
38 conversion between 128-bit floating point types that have a
39 different mode but share the same representation.
41 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
43 PR tree-optimization/85814
44 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
45 a null return from get_strinfo when unsharing the next
48 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
51 * varasm.c (weak_finish): Clean up weak_decls.
53 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
55 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
56 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
58 * config/aarch64/iterators.md (ABAL): New int iterator.
61 (sur): Add mappings for the above.
62 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
64 (aarch64_<sur>abal<mode>_4): Likewise.
65 (aarch64_<sur>adalp<mode>_3): Likewise.
66 (<sur>sadv16qi): New define_expand.
68 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
70 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
71 (*movdf_internal): Ditto.
73 (*rsqrtsf2_sse): Ditto.
74 (*sqrt<mode>2_sse): Ditto.
76 2018-05-21 Tamar Christina <tamar.christina@arm.com>
78 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
80 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
81 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
82 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
83 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
85 * config/aarch64/arm_neon.h: Likewise.
86 * config/aarch64/iterators.md (VQ_I): New.
88 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
90 * config.gcc: Add arc/t-multilib-linux to tmake_file for
92 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
95 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
97 * config/nds32/constraints.md (S): New constraint.
98 * config/nds32/nds32.md (call_internal): Use constraint S.
99 (call_value_internal): Likewise.
100 (sibcall_internal): Likewise.
101 (sibcall_value_internal): Likewise.
103 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
104 Chung-Ju Wu <jasonwucj@gmail.com>
106 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
109 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
110 Chung-Ju Wu <jasonwucj@gmail.com>
112 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
113 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
114 (nds32_rtx_costs_impl): Simplify.
115 (nds32_address_cost_impl): Simplify.
116 (nds32_init_rtx_costs): New function.
117 (nds32_rtx_costs_speed_prefer): Likewise.
118 (nds32_rtx_costs_size_prefer): Likewise.
119 (nds32_address_cost_speed_prefer): Likewise.
120 (nds32_address_cost_speed_fwprop): Likewise.
121 (nds32_address_cost_size_prefer): Likewise.
122 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
123 * config/nds32/nds32.c (nds32_option_override): Use
124 nds32_init_rtx_costs function.
126 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
128 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
129 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
130 (TARGET_PIPELINE_N8): Likewise.
131 (TARGET_PIPELINE_N10): Likewise.
132 (TARGET_PIPELINE_N13): Likewise.
133 (TARGET_PIPELINE_GRAYWOLF): Likewise.
135 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
137 * config/nds32/nds32-fpu.md: Update copyright year.
139 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
141 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
143 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
145 * config/nds32/nds32.c
146 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
147 * config/nds32/nds32.opt (minline-asm-r15): New option.
149 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
151 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
153 * config/nds32/nds32.md (abssi2): New pattern.
155 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
157 * config/i386/i386.md (rex64namesuffix): New mode attribute.
158 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
159 Merge insn pattern from sse_cvtsi2ss<round_name> and
160 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
161 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
162 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
163 using SWI48 mode iterator.
164 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
165 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
166 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
167 pattern from sse_cvttss2si<round_saeonly_name>
168 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
169 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
170 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
171 using SWI48 mode iterator.
172 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
173 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
174 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
175 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
176 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
177 using SWI48 mode iterator.
178 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
179 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
180 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
181 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
182 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
184 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
185 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
186 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
187 pattern from sse_cvttsd2si<round_saeonly_name>
188 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
190 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
192 * config/nds32/nds32-md-auxiliary.c
193 (nds32_valid_smw_lwm_base_p): Refine.
194 (nds32_output_smw_single_word): Refine.
195 (nds32_output_smw_double_word): New.
196 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
198 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
200 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
201 (nds32_output_stack_pop): Refine.
202 (nds32_expand_unaligned_load): Refine.
203 (nds32_expand_unaligned_store): Refine.
205 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
206 Chung-Ju Wu <jasonwucj@gmail.com>
208 * config/nds32/constants.md: Add TP_REGNUM constant.
209 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
210 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
212 * config/nds32/nds32-doubleword.md: Consider flag_pic.
213 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
214 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
215 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
216 and PIC code generation.
217 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
219 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
221 * config/nds32/nds32.md: Support TLS and PIC.
222 * config/nds32/nds32.c: Support TLS and PIC.
223 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
224 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
227 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
229 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
232 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
233 Chung-Ju Wu <jasonwucj@gmail.com>
235 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
236 * config/nds32/nds32-md-auxiliary.c
237 (symbolic_reference_mentioned_p): New.
238 (nds32_legitimize_ict_address): New.
239 (nds32_expand_ict_move): New.
240 (nds32_indirect_call_referenced_p): New.
241 (nds32_symbol_binds_local_p): Delete.
242 (nds32_long_call_p): Modify.
243 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
244 * config/nds32/nds32-protos.h
245 (symbolic_reference_mentioned_p): Declare.
246 (nds32_legitimize_ict_address): Declare.
247 (nds32_expand_ict_move): Declare.
248 (nds32_indirect_call_referenced_p): Declare.
249 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
250 (nds32_relax_group): Use nds32_ict_const_p as condition.
251 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
252 (nds32_asm_file_start): Output ict_model directive in asm code.
253 (nds32_legitimate_address_p): Consider indirect call.
254 (nds32_print_operand): Consider indirect call.
255 (nds32_print_operand_address): Consider indirect call.
256 (nds32_insert_attributes): Handle "indirect_call" attribute.
257 (TARGET_LEGITIMATE_ADDRESS_P): Define.
258 (TARGET_LEGITIMATE_CONSTANT_P): Define.
259 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
260 (TARGET_DELEGITIMIZE_ADDRESS): Define.
261 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
262 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
263 (TARGET_ICT_MODEL_SMALL): Define.
264 (TARGET_ICT_MODEL_LARGE): Define.
265 * config/nds32/nds32.md (movsi): Consider ict model.
266 (call, call_value): Consider ict model.
267 (sibcall, sibcall_value): Consider ict model.
268 * config/nds32/nds32.opt (mict-model): New option.
269 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
272 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
273 Monk Chiang <sh.chiang04@gmail.com>
274 Jim Wilson <jimw@sifive.com>
276 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
277 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
278 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
279 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
280 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
281 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
282 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
283 compute save_libcall_adjustment properly.
284 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
285 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
286 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
287 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
288 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
289 (ABI_SPEC): Handle mabi=ilp32e.
290 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
292 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
293 <-march>: Add rv32e as an example.
295 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
298 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
299 (intra_create_variable_infos): Handle C++ constructors.
301 2018-05-18 Martin Liska <mliska@suse.cz>
303 * passes.def: Remove a redundant pass.
305 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
308 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
310 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
312 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
314 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
316 (armv2, armv2a, armv3, armv3m): Delete architectures.
317 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
318 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
319 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
321 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
322 (*mulsidi3adddi): Likewise.
323 (mulsidi3): Likewise.
324 (*mulsidi3_nov6): Likewise.
325 (umulsidi3): Likewise.
326 (umulsidi3_nov6): Likewise.
327 (umaddsidi4): Likewise.
328 (*umulsidi3adddi): Likewise.
329 (smulsi3_highpart): Likewise.
330 (*smulsi3_highpart_nov6): Likewise.
331 (umulsi3_highpart): Likewise.
332 (*umulsi3_highpart_nov6): Likewise.
333 * config/arm/arm.h (arm_arch3m): Delete.
334 * config/arm/arm.c (arm_arch3m): Delete.
335 (arm_option_override_internal): Update armv3-related comment.
336 (arm_configure_build_target): Delete use of isa_bit_mode32.
337 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
338 (arm_rtx_costs_internal): Delete check of arm_arch3m.
339 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
342 * config/arm/arm-protos.h (arm_arch3m): Delete.
343 * config/arm/arm-tables.opt: Regenerate.
344 * config/arm/arm-tune.md: Likewise.
345 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
346 deleted architectures.
348 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
350 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
351 (armv5t, armv5te): New features.
352 (ARMv5, ARMv5e): Delete fgroups.
353 (ARMv5t, ARMv5te): Adjust for above changes.
355 (armv5, armv5e): Delete arches.
356 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
358 (*call_reg_arm): Likewise.
359 (*call_value_reg_armv5): Likewise.
360 (*call_value_reg_arm): Likewise.
361 (*call_symbol): Likewise.
362 (*call_value_symbol): Likewise.
363 (*sibcall_insn): Likewise.
364 (*sibcall_value_insn): Likewise.
366 (prefetch): Likewise.
367 (define_split and define_peephole2 dependent on arm_arch5):
369 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
371 (TARGET_ARM_QBIT): Likewise.
372 (TARGET_DSP_MULTIPLY): Likewise.
373 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
374 (arm_arch5, arm_arch5e): Delete.
375 (arm_arch5t, arm_arch5te): Declare.
376 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
377 (arm_arch5t): Declare.
378 (arm_option_reconfigure_globals): Update for the above.
379 (arm_options_perform_arch_sanity_checks): Update comment, replace
380 use of arm_arch5 with arm_arch5t.
381 (use_return_insn): Likewise.
382 (arm_emit_call_insn): Likewise.
383 (output_return_instruction): Likewise.
384 (arm_final_prescan_insn): Likewise.
385 (arm_coproc_builtin_available): Likewise.
386 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
387 arm_arch5e with arm_arch5t and arm_arch5te.
388 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
389 (arm_arch5t, arm_arch5te): Declare.
390 * config/arm/arm-tables.opt: Regenerate.
391 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
392 * config/arm/t-multilib: Likewise.
393 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
394 instead of arm_arch5.
395 (*call_reg_thumb1): Likewise.
396 (*call_value_reg_thumb1_v5): Likewise.
397 (*call_value_reg_thumb1): Likewise.
398 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
400 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
402 2018-05-18 Martin Liska <mliska@suse.cz>
404 PR gcov-profile/84846
405 * doc/gcov.texi: Document -t option of gcov tool.
407 2018-05-18 Martin Liska <mliska@suse.cz>
409 PR gcov-profile/84846
410 * gcov.c (print_usage): Add new -t option.
411 (process_args): Handle the option.
412 (generate_results): Use stdout as output when requested by
415 2018-05-18 Martin Liska <mliska@suse.cz>
417 PR gcov-profile/84846
418 * coverage.c (coverage_init): Write PWD to .gcno file.
419 * doc/gcov.texi: Document how working directory is printed.
420 * gcov-dump.c (dump_gcov_file): Print PWD.
421 * gcov.c (output_intermediate_file): Likewise.
422 (read_graph_file): Read PWD string.
423 (output_lines): Print PWD.
425 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
428 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
429 for retval and return false if all args to phi are zero.
431 2018-05-18 Richard Biener <rguenther@suse.de>
433 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
435 (evrp_dom_walker::before_dom_children): Call it.
437 2018-05-18 Richard Biener <rguenther@suse.de>
439 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
440 results when processing array refs with variable index.
442 2018-05-18 Toon Moene <toon@moene.org>
444 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
445 directly after that of -floop-interchange. Indicate that both
446 options are enabled by default when specifying -O3.
448 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
450 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
451 iterator. Delete separate integer-mode vec_set<mode> expander.
452 (aarch64_simd_vec_setv2di): Delete.
453 (vec_setv2di): Delete.
454 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
455 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
456 the "w, r" alternative.
458 2018-05-18 Martin Liska <mliska@suse.cz>
460 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
461 * tree-pass.h (make_pass_lower_switch_O0): New function.
462 * tree-switch-conversion.c (node_has_low_bound): Remove.
463 (node_has_high_bound): Likewise.
464 (node_is_bounded): Likewise.
465 (class pass_lower_switch): Make it a template type and create
467 (pass_lower_switch::execute): Add template argument.
468 (make_pass_lower_switch): New function.
469 (make_pass_lower_switch_O0): New function.
470 (do_jump_if_equal): Remove.
471 (emit_case_nodes): Simplify to just handle all 3 cases and leave
472 all the hard work to tree optimization passes.
474 2018-05-18 Martin Liska <mliska@suse.cz>
476 * dbgcnt.c (limit_low): Renamed from limit.
477 (limit_high): New variable.
478 (dbg_cnt_is_enabled): Check for upper limit.
479 (dbg_cnt): Adjust dumping.
480 (dbg_cnt_set_limit_by_index): Add new argument for high
482 (dbg_cnt_set_limit_by_name): Likewise.
483 (dbg_cnt_process_single_pair): Parse new format.
484 (dbg_cnt_process_opt): Use strtok.
485 (dbg_cnt_list_all_counters): Remove 'value' and add
487 * doc/invoke.texi: Document changes.
489 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
491 * doc/sourcebuild.texi (scalar_all_fma): Document.
492 * tree.def (FMA_EXPR): Delete.
493 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
494 * internal-fn.c (ternary_direct): New macro.
495 (expand_ternary_optab_fn): Likewise.
496 (direct_ternary_optab_supported_p): Likewise.
497 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
498 * builtins.c (fold_builtin_fma): Delete.
499 (fold_builtin_3): Don't call it.
500 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
501 * expr.c (expand_expr_real_2): Likewise.
502 * fold-const.c (operand_equal_p): Likewise.
503 (fold_ternary_loc): Likewise.
504 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
505 * gimple.c (DEFTREECODE): Likewise.
506 * gimplify.c (gimplify_expr): Likewise.
507 * optabs-tree.c (optab_for_tree_code): Likewise.
508 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
509 * tree-eh.c (operation_could_trap_p): Likewise.
510 (stmt_could_throw_1_p): Likewise.
511 * tree-inline.c (estimate_operator_cost): Likewise.
512 * tree-pretty-print.c (dump_generic_node): Likewise.
513 (op_code_prio): Likewise.
514 * tree-ssa-loop-im.c (stmt_cost): Likewise.
515 * tree-ssa-operands.c (get_expr_operands): Likewise.
516 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
517 * fold-const-call.h (fold_fma): Delete.
518 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
519 CFN_FNMA and CFN_FNMS.
521 * genmatch.c (combined_fn): New enum.
522 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
523 (commutative_op): New function.
524 (commutate): Use it. Handle more than 2 operands.
525 (dt_operand::gen_gimple_expr): Use commutative_op.
526 (parser::parse_expr): Allow :c to be used with non-binary
527 operators if the commutative operand is known.
528 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
529 CFN_FMS, CFN_FNMA and CFN_FNMS.
530 (backprop::process_assign_use): Remove FMA_EXPR handling.
531 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
532 (gen_hsa_fma): New function.
533 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
534 IFN_FNMA and IFN_FNMS.
535 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
536 * gimple-fold.h (follow_all_ssa_edges): Declare.
537 * gimple-fold.c (follow_all_ssa_edges): New function.
538 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
539 gimple_build interface and use follow_all_ssa_edges to fold the result.
540 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
541 instead of checking for optabs directly.
542 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
543 rather than FMA_EXPRs.
544 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
545 call to IFN_FMA instead of an FMA_EXPR.
547 2018-05-17 Jim Wilson <jimw@sifive.com>
549 * expr.c (do_tablejump): When converting index to Pmode, if we have a
550 sign extended promoted subreg, and the range does not have the sign bit
551 set, then do a sign extend.
553 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
554 test, check for sign extended subreg and/or constant operands, and
555 do a sign extend in that case.
557 2018-05-17 Steve Ellcey <sellcey@cavium.com>
559 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
560 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
562 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
563 Change logics_shift_reg to logics_shift_imm.
564 (thunderx2t99_fp_loadpair_basic): Delete.
565 (thunderx2t99_fp_storepair_basic): Delete.
566 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
567 (thunderx2t99_asimd_polynomial): Delete.
568 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
569 and neon_fp_mul_d_scalar_q.
570 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
571 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
572 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
573 (thunderx2t99_asimd_lut): Add missing tbl types.
574 (thunderx2t99_asimd_ext): Delete.
575 (thunderx2t99_asimd_load1_1_mult): Delete.
576 (thunderx2t99_asimd_load1_2_mult): Delete.
577 (thunderx2t99_asimd_load1_ldp): New.
578 (thunderx2t99_asimd_load1): New.
579 (thunderx2t99_asimd_load2): Add missing *load2* types.
580 (thunderx2t99_asimd_load3): New.
581 (thunderx2t99_asimd_load4): New.
582 (thunderx2t99_asimd_store1_1_mult): Delete.
583 (thunderx2t99_asimd_store1_2_mult): Delete.
584 (thunderx2t99_asimd_store2_mult): Delete.
585 (thunderx2t99_asimd_store2_onelane): Delete.
586 (thunderx2t99_asimd_store_stp): New.
587 (thunderx2t99_asimd_store1): New.
588 (thunderx2t99_asimd_store2): New.
589 (thunderx2t99_asimd_store3): New.
590 (thunderx2t99_asimd_store4): New.
592 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
594 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
595 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
597 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
598 Segher Boessenkool <segher@kernel.crashing.org>
601 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
604 2018-05-17 Richard Biener <rguenther@suse.de>
606 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
607 for pruning loop and prune defs feeding only already visited PHIs.
609 2018-05-17 Richard Biener <rguenther@suse.de>
611 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
613 2018-05-17 Bin Cheng <bin.cheng@arm.com>
614 Richard Biener <rguenther@suse.de>
616 PR tree-optimization/85793
617 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
618 for VMAT_ELEMENTWISE.
620 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
622 * internal-fn.h (lookup_internal_fn): Declare
623 * internal-fn.c (lookup_internal_fn): New function.
624 * gimple.c (gimple_build_call_from_tree): Handle calls to
626 * gimple-pretty-print.c (dump_gimple_call): Print "." before
627 internal function names.
628 * tree-pretty-print.c (dump_generic_node): Likewise.
629 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
631 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
633 * gimple-fold.h (gimple_build): Make the function forms take
634 combined_fn rather than built_in_function.
635 (gimple_simplify): Likewise.
636 * gimple-match-head.c (gimple_simplify): Likewise.
637 * gimple-fold.c (gimple_build): Likewise.
638 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
639 rather than gimple_build_call_internal.
640 (get_initial_defs_for_reduction): Likewise.
641 (vect_create_epilog_for_reduction): Likewise.
642 (vectorizable_live_operation): Likewise.
644 2018-05-17 Martin Liska <mliska@suse.cz>
646 * gimple-ssa-sprintf.c (format_directive): Do not use
647 space in between 'G_' and '('.
649 2018-05-17 Jakub Jelinek <jakub@redhat.com>
652 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
653 even if the mask is not all ones.
656 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
658 (ix86_gimple_fold_builtin): Likewise.
661 * config/i386/i386.c: Include tree-vector-builder.h.
662 (ix86_vector_shift_count): New function.
663 (ix86_fold_builtin): Fold shift builtins by scalar count.
664 (ix86_gimple_fold_builtin): Likewise.
666 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
667 _mm512_setzero): New intrinsics.
669 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
670 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
672 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
673 code generation for cases where splatting a value is not useful.
674 * simplify-rtx.c (simplify_ternary_operation): Simplify
675 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
676 mode to a vec_concat.
678 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
680 * config.gcc: Support "goldmont-plus".
681 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
683 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
684 PROCESSOR_GOLDMONT_PLUS.
685 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
686 (processor_target_table): Add "goldmont-plus".
687 (PTA_GOLDMONT_PLUS): Define.
688 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
689 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
690 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
691 (fold_builtin_cpu): Add "goldmont-plus".
692 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
693 (ix86_option_override_internal): Add "goldmont-plus".
694 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
695 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
696 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
697 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
699 2018-05-17 Richard Biener <rguenther@suse.de>
701 PR tree-optimization/85757
702 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
703 remove defs that only feed that PHI from further processing.
705 2018-05-16 Jim Wilson <jimw@sifive.com>
707 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
709 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
711 2018-05-16 Mark Wielaard <mark@klomp.org>
713 * dwarf2out.c (count_index_strings): New function.
714 (output_indirect_strings): Call count_index_strings and generate
715 header for dwarf_version >= 5.
717 2018-05-16 Mark Wielaard <mark@klomp.org>
719 * dwarf2out.c (dwarf_FORM): New function.
720 (set_indirect_string): Use dwarf_FORM.
721 (reset_indirect_string): Likewise.
722 (size_of_die): Likewise.
723 (value_format): Likewise.
724 (output_die): Likewise.
725 (add_skeleton_AT_string): Likewise.
726 (output_macinfo_op): Likewise.
727 (index_string): Likewise.
728 (output_index_string_offset): Likewise.
729 (output_index_string): Likewise.
730 (count_index_strings): Likewise.
732 2018-05-16 Carl Love <cel@us.ibm.com>
734 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
735 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
737 2018-05-16 Martin Jambor <mjambor@suse.cz>
739 * ipa-prop.c (ipa_free_all_edge_args): Remove.
740 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
742 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
744 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
745 (fnma<mode>4): Likewise.
746 (fms<mode>4): Likewise.
747 (fnms<mode>4): Likewise.
748 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
749 (aarch64_fnma<mode>4): Likewise.
750 (aarch64_fms<mode>4): Likewise.
751 (aarch64_fnms<mode>4): Likewise.
752 (aarch64_fnmadd<mode>4): Likewise.
754 2018-05-16 Jason Merrill <jason@redhat.com>
756 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
758 2018-05-16 Richard Biener <rguenther@suse.de>
760 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
761 (dump_stmt_cost): Declare.
762 (add_stmt_cost): Dump cost we add.
763 (add_stmt_costs): New function.
764 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
766 (vect_analyze_stmt): Adjust prototype.
767 (vectorizable_condition): Likewise.
768 (vectorizable_live_operation): Likewise.
769 (vectorizable_reduction): Likewise.
770 (vectorizable_induction): Likewise.
771 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
772 cost vector to pass to vectorizable_ and record afterwards.
773 (vect_model_reduction_cost): Take cost vector argument and adjust.
774 (vect_model_induction_cost): Likewise.
775 (vectorizable_reduction): Likewise.
776 (vectorizable_induction): Likewise.
777 (vectorizable_live_operation): Likewise.
778 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
779 SLP_TREE_NUMBER_OF_VEC_STMTS.
780 (vect_analyze_slp_cost_1): Remove.
781 (vect_analyze_slp_cost): Likewise.
782 (vect_slp_analyze_node_operations): Take visited args and
783 a target cost vector. Avoid processing already visited stmt sets.
784 (vect_slp_analyze_operations): Use a local cost vector to gather
785 costs and register those of non-discarded instances.
786 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
787 (vect_schedule_slp_instance): Remove copying of
788 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
790 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
791 adding cost. Record cost entry location.
792 (vect_prologue_cost_for_slp_op): Function to compute cost of
793 a constant or invariant generated for SLP vect in the prologue,
794 split out from vect_analyze_slp_cost_1.
795 (vect_model_simple_cost): Make static. Adjust for SLP costing.
796 (vect_model_promotion_demotion_cost): Likewise.
797 (vect_model_store_cost): Likewise, make static.
798 (vect_model_load_cost): Likewise.
799 (vectorizable_bswap): Add cost vector arg and adjust.
800 (vectorizable_call): Likewise.
801 (vectorizable_simd_clone_call): Likewise.
802 (vectorizable_conversion): Likewise.
803 (vectorizable_assignment): Likewise.
804 (vectorizable_shift): Likewise.
805 (vectorizable_operation): Likewise.
806 (vectorizable_store): Likewise.
807 (vectorizable_load): Likewise.
808 (vectorizable_condition): Likewise.
809 (vectorizable_comparison): Likewise.
810 (can_vectorize_live_stmts): Likewise.
811 (vect_analyze_stmt): Likewise.
812 (vect_transform_stmt): Adjust calls to vectorizable_*.
813 * tree-vectorizer.c: Include gimple-pretty-print.h.
814 (dump_stmt_cost): New function.
816 2018-05-16 Richard Biener <rguenther@suse.de>
818 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
819 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
820 * tree-ssa-dse.c: Include tree-ssa-loop.h.
821 (check_name): New callback.
822 (dse_classify_store): Track cycles via a visited bitmap of PHI
823 defs and simplify handling of in-loop and across loop dead stores
824 and properly fail for loop-variant refs. Handle byte-tracking with
825 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
828 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
830 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
831 (vect_get_mask_type_for_stmt): Likewise.
832 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
834 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
835 to determine the statement's vector type and the vector type that
836 should be used for calculating nunits. Deal with cases in which
837 the type has to be deferred.
838 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
839 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
840 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
841 (vect_determine_vf_for_stmt): New functions, split out from...
842 (vect_determine_vectorization_factor): ...here.
843 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
844 (vect_get_mask_type_for_stmt): New functions, split out from
845 vect_determine_vectorization_factor.
847 2018-05-16 Richard Biener <rguenther@suse.de>
849 * tree-cfg.c (verify_gimple_assign_ternary): Properly
850 verify the [VEC_]COND_EXPR embedded comparison.
852 2018-05-15 Martin Sebor <msebor@redhat.com>
854 PR tree-optimization/85753
855 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
856 RECORD_TYPE in addition to ARRAY_TYPE.
858 2018-05-15 Martin Sebor <msebor@redhat.com>
861 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
863 2018-05-15 Richard Biener <rguenther@suse.de>
865 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
866 add by_clobber_p one. Change algorithm to collect all defs
867 representing uses we need to walk and try reducing them to
868 a single one before failing.
869 (dse_dom_walker::dse_optimize_stmt): Adjust.
871 2018-05-13 Mark Wielaard <mark@klomp.org>
873 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
874 (size_of_loc_descr): Likewise.
875 (output_loc_operands): Likewise.
876 (output_loc_operands_raw): Likewise.
877 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
878 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
879 (hash_loc_operands): Likewise.
880 (compare_loc_operands): Likewise.
882 2018-05-14 Mark Wielaard <mark@klomp.org>
884 * dwarf2out.c (count_index_addrs): New function.
885 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
887 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
889 PR tree-optimization/83648
890 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
891 return value as malloc candidate.
893 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
896 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
897 as true in call to suggest_attribute.
899 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
901 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
904 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
906 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
908 (aarch64_crypto_aesd_fused): Likewise.
910 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
912 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
913 (movsi_aarch64): Likewise.
914 (load_pairsi): Likewise.
915 (load_pairdi): Likewise.
916 (store_pairsi): Likewise.
917 (store_pairdi): Likewise.
918 (load_pairsf): Likewise.
919 (load_pairdf): Likewise.
920 (store_pairsf): Likewise.
921 (store_pairdf): Likewise.
922 (zero_extend): Likewise.
923 (trunc): Swap alternatives.
924 (fcvt_target): Add '?' to prefer w over r.
926 2018-05-14 Jakub Jelinek <jakub@redhat.com>
929 * config/i386/i386.md: Disallow non-commutative arithmetics in
930 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
931 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
932 in the peephole2 before it.
934 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
936 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
937 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
938 (ix86_handle_option): Handle -mcldemote.
939 * config.gcc: New header.
940 * config/i386/cldemoteintrin.h: New file.
941 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
942 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
944 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
945 OPTION_MASK_ISA_CLDEMOTE.
946 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
947 (ix86_valid_target_attribute_inner_p): Ditto.
948 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
949 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
950 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
951 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
952 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
954 * config/i386/i386.opt: Add -mcldemote.
955 * config/i386/x86intrin.h: New header.
956 * doc/invoke.texi: Add -mcldemote.
958 2018-05-14 Richard Biener <rguenther@suse.de>
960 * doc/match-and-simplify.texi: Adjust :s documentation.
962 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
964 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
965 intended memcpy size.
966 (REORDER_45): Likewise.
968 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
971 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
972 * vec.c (qsort_chk): Use gcc_qsort.
973 * Makefile.in (OBJS-libcommon): Add sort.o.
974 (build/sort.o): New target. Use it...
975 (BUILD_RTL): ... here, and...
976 (build/gencfn-macros): ... here, and...
977 (build/genmatch): ... here.
979 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
980 Chung-Ju Wu <jasonwucj@gmail.com>
982 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
983 * config/nds32/nds32-graywolf.md: New file.
984 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
985 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
987 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
988 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
989 * config/nds32/nds32.md (pipeline_model): Add graywolf.
990 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
991 * config/nds32/pipelines.md: Include n15 settings.
993 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
994 Chung-Ju Wu <jasonwucj@gmail.com>
996 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
997 * config/nds32/nds32-n13.md: New file.
998 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
999 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1001 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1002 * config/nds32/nds32.md (pipeline_model): Add n13.
1003 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1004 * config/nds32/pipelines.md: Include n13 settings.
1006 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1007 Chung-Ju Wu <jasonwucj@gmail.com>
1009 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1010 * config/nds32/nds32-n10.md: New file.
1011 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1012 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1014 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1015 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1016 * config/nds32/nds32.md (pipeline_model): Add n10.
1017 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1018 * config/nds32/pipelines.md: Include n10 settings.
1020 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1021 Kito Cheng <kito.cheng@gmail.com>
1022 Chung-Ju Wu <jasonwucj@gmail.com>
1024 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1025 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1026 Add enum values for DSP extension instructions.
1027 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1029 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1030 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1032 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1033 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1034 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1035 * config/nds32/nds32-intrinsic.md: Likewise.
1036 * config/nds32/nds32_intrinsic.h: Likewise.
1037 * config/nds32/nds32-md-auxiliary.c: Likewise.
1038 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1039 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1040 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1041 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1042 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1043 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1044 TYPE_DMAC in switch statement.
1045 * config/nds32/nds32.c: New checking and implementation for DSP
1046 extension instructions.
1047 * config/nds32/nds32.h: Likewise.
1048 * config/nds32/nds32.md: Likewise.
1049 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1050 * config/nds32/predicates.md: Implement new predicates for DSP
1053 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1055 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1056 Reformat alternatives and attributes so it is easier to identify
1057 which constraints/attributes go with which instruction.
1058 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1059 (mov<mode>_softfloat32, FMOVE64): Likewise.
1060 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1061 (mov<mode>_softfloat64, FMOVE64): Likewise.
1063 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1065 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1067 (Basic PowerPC Built-in Functions): The new name of the
1068 subsection previously known as "PowerPC Built-in Functions".
1069 (Basic PowerPC Built-in Functions Available on all Configurations):
1071 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1072 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1073 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1074 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1076 2018-05-11 Martin Jambor <mjambor@suse.cz>
1079 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1082 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1085 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1087 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1089 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1090 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1091 (ix86_handle_option): Handle -mwaitpkg.
1092 * config.gcc: New header.
1093 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1094 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1095 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1097 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1098 OPTION_MASK_ISA_WAITPKG.
1099 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1100 (ix86_option_override_internal): Add PTA_WAITPKG.
1101 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1102 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1103 IX86_BUILTIN_TPAUSE.
1104 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1105 __builtin_ia32_umwait and __builtin_ia32_tpause.
1106 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1107 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1108 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1109 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1110 UNSPECV_TPAUSE): New.
1111 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1112 * config/i386/i386.opt: Add -mwaitpkg.
1113 * config/i386/waitpkgintrin.h: New file.
1114 * config/i386/x86intrin.h: New header.
1115 * doc/invoke.texi: Add -mwaitpkg.
1117 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1120 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1122 (cortex-m0): Use armv6s-m isa.
1123 (cortex-m0plus): Likewise.
1124 (cortex-m1): Likewise.
1125 (cortex-m0.small-multiply): Likewise.
1126 (cortex-m0plus.small-multiply): Likewise.
1127 (cortex-m1.small-multiply): Likewise.
1129 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1130 Jakub Jelinek <jakub@redhat.com>
1132 PR tree-optimization/85692
1133 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1134 source permute as well.
1136 2018-05-11 Martin Liska <mliska@suse.cz>
1139 * doc/extend.texi: Document LLVM style format for no_sanitize
1142 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1144 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1145 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1146 (mode_supports_vsx_dform_quad): Likewise.
1147 (mode_supports_vmx_dform): Move these functions to be next to the
1148 other mode_supports functions.
1149 (mode_supports_dq_form): Likewise.
1150 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1151 mode_supports_dq_form.
1152 (reg_offset_addressing_ok_p): Likewise.
1153 (offsettable_ok_by_alignment): Likewise.
1154 (rs6000_legitimate_offset_address_p): Likewise.
1155 (legitimate_lo_sum_address_p): Likewise.
1156 (rs6000_legitimize_address): Likewise.
1157 (rs6000_legitimize_reload_address): Likewise.
1158 (rs6000_secondary_reload_inner): Likewise.
1159 (rs6000_preferred_reload_class): Likewise.
1160 (rs6000_output_move_128bit): Likewise.
1162 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1164 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1165 Generate SImode target register for null target.
1166 <case IX86_BUILTIN_XGETBV>: Ditto.
1167 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1168 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1170 2018-05-10 Carl Love <cel@us.ibm.com>
1172 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1173 dcbtt and dcbtstt if operands[2] is 0.
1175 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1178 * config/i386/sse.md (usadv64qi): New expander.
1180 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1182 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1183 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1184 -maltivec=be support.
1185 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1186 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1187 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1188 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1189 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1190 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1191 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1192 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1193 altivec_vsumsws): Adjust.
1194 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1195 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1196 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1198 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1199 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1200 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1201 (altivec_lve<VI_char>x): Delete expand.
1202 (*altivec_lve<VI_char>x_internal): Rename to...
1203 (altivec_lve<VI_char>x): ... this.
1204 (altivec_lvxl_<mode>): Delete expand.
1205 (*altivec_lvxl_<mode>_internal): Rename to ...
1206 (altivec_lvxl_<mode>): ... this.
1207 (altivec_stvxl_<mode>): Delete expand.
1208 (*altivec_stvxl_<mode>_internal): Rename to ...
1209 (altivec_stvxl_<mode>): ... this.
1210 (altivec_stve<VI_char>x): Delete expand.
1211 (*altivec_stve<VI_char>x_internal): Rename to ...
1212 (altivec_stve<VI_char>x): ... this.
1213 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1214 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1215 reduc_plus_scal_<mode>): Adjust.
1216 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1218 (rs6000_cpu_cpp_builtins): Adjust.
1219 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1220 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1221 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1222 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1223 -maltivec=be support.
1224 (rs6000_split_vec_extract_var): Adjust.
1225 (rs6000_split_v4si_init): Adjust.
1226 (swap_selector_for_mode): Delete.
1227 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1228 altivec_expand_stvex_be): Delete.
1229 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1230 -maltivec=be support.
1231 (rs6000_gimple_fold_builtin): Ditto.
1232 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1234 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1235 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1236 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1237 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1238 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1239 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1240 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1241 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1242 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1243 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1244 anonymous split): Adjust.
1245 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1246 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1248 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1250 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1251 when --with-gxx-include-dir is also specified.
1252 * configure: Regenerate.
1254 2018-05-09 Jim Wilson <jimw@sifive.com>
1257 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1258 * config/riscv/t-withmultilib: New.
1259 * config/riscv/withmultilib.h: New.
1260 * doc/install.texi: Document RISC-V --with-multilib-list support.
1262 2018-05-09 Richard Biener <rguenther@suse.de>
1264 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1266 (vect_bb_vectorization_profitable_p): Adjust. Compute
1267 actual scalar cost using the cost vector and the add_stmt_cost
1270 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1272 PR rtl-optimization/85645
1273 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1274 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1276 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1278 PR rtl-optimization/85645
1279 * shrink-wrap.c (spread_components): Return a boolean saying if
1280 anything was changed.
1281 (try_shrink_wrapping_separate): Iterate spread_components until
1282 nothing changes anymore.
1284 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1286 PR rtl-optimization/85645
1287 * regrename.c (build_def_use): Also kill the chains that include the
1288 destination of a REG_CFA_REGISTER note.
1290 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1292 PR rtl-optimization/85645
1293 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1294 insn that has a REG_CFA_REGISTER note.
1296 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1298 * cfgexpand.c (expand_clobber): New function.
1299 (expand_gimple_stmt_1): Use it.
1300 * tree-vect-stmts.c (vect_clobber_variable): New function,
1302 (vectorizable_simd_clone_call): ...here.
1303 (vectorizable_store): Emit a clobber either side of an
1304 IFN_STORE_LANES sequence.
1305 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1307 2018-05-09 Tom de Vries <tom@codesourcery.com>
1310 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1311 (define_insn "trap_if_false"): Add exit after trap.
1313 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1315 PR rtl-optimization/85638
1316 * bb-reorder.c: Include common/common-target.h.
1317 (create_forwarder_block): New function extracted from...
1318 (fix_up_crossing_landing_pad): ...here. Rename into...
1319 (dw2_fix_up_crossing_landing_pad): ...this.
1320 (sjlj_fix_up_crossing_landing_pad): New function.
1321 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1322 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1323 from both partitions and exit the loop after one iteration.
1325 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1328 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1330 (Basic PowerPC Built-in Functions): The new name of the
1331 subsection previously known as "PowerPC Built-in Functions".
1332 (Basic PowerPC Built-in Functions Available on all Configurations):
1334 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1336 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1338 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1340 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1343 2018-05-08 Jim Wilson <jimw@sifive.com>
1345 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1346 (LD_EMUL_SUFFIX): New.
1347 (LINK_SPEC): Use it.
1349 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1351 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1353 (Basic PowerPC Built-in Functions): The new name of the
1354 subsection previously known as "PowerPC Built-in Functions".
1355 (Basic PowerPC Built-in Functions Available on all Configurations):
1357 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1359 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1361 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1363 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1366 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1369 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1370 after cmpelim optimization.
1372 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1374 * config.gcc: Support "goldmont".
1375 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1376 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1378 * config/i386/i386.c (m_GOLDMONT): Define.
1379 (processor_target_table): Add "goldmont".
1380 (PTA_GOLDMONT): Define.
1381 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1382 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1383 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1384 (fold_builtin_cpu): Add "goldmont".
1385 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1386 (ix86_option_override_internal): Add "goldmont".
1387 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1388 (processor_type): Add PROCESSOR_GOLDMONT.
1389 * config/i386/i386.md: Add CPU "glm".
1390 * config/i386/glm.md: New file.
1391 * config/i386/x86-tune.def: Add m_GOLDMONT.
1392 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1394 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1397 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1399 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1400 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1401 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1404 * config/i386/i386.c (ix86_fold_builtin): Handle
1405 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1408 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1409 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1411 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1414 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1415 (check_arch): Likewise.
1416 (check_fpu): Return the result rather than printing it.
1417 (end arch): Fix operator precedence.
1418 (end cpu): Likewise.
1419 (END): Print the result from check_fpu.
1421 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1422 Alan Hayward <alan.hayward@arm.com>
1423 David Sherwood <david.sherwood@arm.com>
1425 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1426 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1427 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1428 (*fcmuo<mode>_and): New patterns.
1430 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1432 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1433 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1434 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1435 (cmp_op, sve_imm_con): New code attributes.
1436 (SVE_COND_INT_CMP, imm_con): Delete.
1437 (cmp_op): Remove above unspecs from int attribute.
1438 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1440 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1441 comparison-specific unspecs.
1442 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1443 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1444 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1445 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1446 (*vec_fcm<cmp_op><mode>): Rename to...
1447 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1448 (*vec_fcmuo<mode>): Rename to...
1449 (*fcmuo<mode>): ...this and adjust likewise.
1450 (*pred_fcm<cmp_op><mode>): New pattern.
1451 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1452 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1454 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1456 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1457 (aarch64_emit_sve_predicated_cond): New function.
1458 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1459 (aarch64_emit_unspec_cond_or): Replace with...
1460 (aarch64_emit_sve_or_conds): ...this new function. Use
1461 aarch64_emit_sve_ptrue_op for the individual comparisons and
1462 aarch64_emit_binop to OR them together.
1463 (aarch64_emit_inverted_unspec_cond): Replace with...
1464 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1465 aarch64_emit_sve_ptrue_op for the comparison and
1466 aarch64_emit_unop to invert the result.
1467 (aarch64_expand_sve_vec_cmp_float): Update after the above
1468 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1470 2018-05-07 Nathan Sidwell <nathan@acm.org>
1472 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1473 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1474 (Backwards Compatibility): Likewise.
1476 2018-05-07 Luis Machado <luis.machado@linaro.org>
1480 2018-05-07 Luis Machado <luis.machado@linaro.org>
1482 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1483 <prefetch_dynamic_strides>: New const bool field.
1484 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1485 prefetch_dynamic_strides.
1486 (exynosm1_prefetch_tune): Likewise.
1487 (thunderxt88_prefetch_tune): Likewise.
1488 (thunderx_prefetch_tune): Likewise.
1489 (thunderx2t99_prefetch_tune): Likewise.
1490 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1491 (aarch64_override_options_internal): Update to set
1492 PARAM_PREFETCH_DYNAMIC_STRIDES.
1493 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1494 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1495 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1496 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1497 prefetch-dynamic-strides setting.
1499 2018-05-07 Luis Machado <luis.machado@linaro.org>
1501 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1502 <minimum_stride>: New const int field.
1503 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1504 minimum_stride field.
1505 (exynosm1_prefetch_tune): Likewise.
1506 (thunderxt88_prefetch_tune): Likewise.
1507 (thunderx_prefetch_tune): Likewise.
1508 (thunderx2t99_prefetch_tune): Likewise.
1509 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1510 (aarch64_override_options_internal): Update to set
1511 PARAM_PREFETCH_MINIMUM_STRIDE.
1512 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1513 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1514 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1515 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1516 stride is constant and is below the minimum stride threshold.
1518 2018-05-07 Luis Machado <luis.machado@linaro.org>
1520 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1523 2018-05-07 Luis Machado <luis.machado@linaro.org>
1525 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1526 <prefetch_dynamic_strides>: New const bool field.
1527 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1528 prefetch_dynamic_strides.
1529 (exynosm1_prefetch_tune): Likewise.
1530 (thunderxt88_prefetch_tune): Likewise.
1531 (thunderx_prefetch_tune): Likewise.
1532 (thunderx2t99_prefetch_tune): Likewise.
1533 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1534 (aarch64_override_options_internal): Update to set
1535 PARAM_PREFETCH_DYNAMIC_STRIDES.
1536 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1537 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1538 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1539 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1540 prefetch-dynamic-strides setting.
1542 2018-05-07 Luis Machado <luis.machado@linaro.org>
1544 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1545 <minimum_stride>: New const int field.
1546 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1547 minimum_stride field.
1548 (exynosm1_prefetch_tune): Likewise.
1549 (thunderxt88_prefetch_tune): Likewise.
1550 (thunderx_prefetch_tune): Likewise.
1551 (thunderx2t99_prefetch_tune): Likewise.
1552 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1553 (aarch64_override_options_internal): Update to set
1554 PARAM_PREFETCH_MINIMUM_STRIDE.
1555 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1556 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1557 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1558 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1559 stride is constant and is below the minimum stride threshold.
1561 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1564 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1565 the type is addressable. Don't force op into register if it has
1568 2018-05-05 Roland McGrath <mcgrathr@google.com>
1571 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1572 any section for which we don't know a specific type it should have,
1573 regardless of name. Previously this was done only for the exact
1574 names ".init_array", ".fini_array", and ".preinit_array".
1575 (default_elf_asm_named_section): Add comment about
1576 relationship with default_section_type_flags and SECTION_NOTYPE.
1577 (get_section): Don't consider it a type conflict if one side has
1578 SECTION_NOTYPE and the other doesn't, as long as neither has the
1579 SECTION_BSS et al used in the default_section_type_flags logic.
1581 2018-05-05 Tom de Vries <tom@codesourcery.com>
1584 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1585 (workaround_barsyncs): New function.
1586 (nvptx_reorg): Use workaround_barsyncs.
1587 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1588 (define_expand "nvptx_membar_cta"): New define_expand.
1589 (define_insn "*nvptx_membar_cta"): New insn.
1591 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1593 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1594 To improve optimization opportunities.
1595 * builtin-types.def: The new needed builtin types for the above.
1597 2018-05-04 Richard Biener <rguenther@suse.de>
1599 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1600 * gimple-ssa-store-merging.c
1601 (imm_store_chain_info::output_merged_store): Remove redundant create,
1602 release split_store vector contents on failure.
1603 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1604 scalar stmt vector on cache hit.
1606 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1608 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1610 * config.gcc (powerpc-xilinx-eabi*): Remove.
1611 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1613 (fusion_addis_mem_combo_load): Ditto.
1614 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1616 (rs6000_cpu_cpp_builtins): Ditto.
1617 * config/rs6000/rs6000-linux.c
1618 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1619 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1620 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1622 (rs6000_setup_reg_addr_masks): Ditto.
1623 (rs6000_init_hard_regno_mode_ok): Ditto.
1624 (rs6000_option_override_internal): Ditto.
1625 (legitimate_lo_sum_address_p): Ditto.
1626 (rs6000_legitimize_address): Ditto.
1627 (rs6000_legitimize_reload_address): Ditto.
1628 (rs6000_legitimate_address_p): Ditto.
1629 (abi_v4_pass_in_fpr): Ditto.
1630 (setup_incoming_varargs): Ditto.
1631 (rs6000_gimplify_va_arg): Ditto.
1632 (rs6000_split_multireg_move): Ditto.
1633 (rs6000_savres_strategy): Ditto.
1634 (rs6000_emit_prologue_components): Ditto.
1635 (rs6000_emit_epilogue_components): Ditto.
1636 (rs6000_emit_prologue): Ditto.
1637 (rs6000_emit_epilogue): Ditto.
1638 (rs6000_elf_file_end): Ditto.
1639 (rs6000_function_value): Ditto.
1640 (rs6000_libcall_value): Ditto.
1641 * config/rs6000/rs6000.h: Ditto.
1642 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1643 (TARGET_MINMAX): ... this. New.
1644 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1645 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1646 (*movsi_internal1_single): Delete.
1647 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1648 mfpu=, mxilinx-fpu): Delete.
1649 * config/rs6000/singlefp.h: Delete.
1650 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1651 * config/rs6000/t-rs6000: Ditto.
1652 * config/rs6000/t-xilinx: Delete.
1653 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1654 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1655 (VStype_simple): Delete.
1656 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1657 * config/rs6000/xfpu.h: Delete.
1658 * config/rs6000/xfpu.md: Delete.
1659 * config/rs6000/xilinx.h: Delete.
1660 * config/rs6000/xilinx.opt: Delete.
1661 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1662 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1664 2018-05-04 Tom de Vries <tom@codesourcery.com>
1667 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1670 2018-05-04 Richard Biener <rguenther@suse.de>
1673 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1674 (expand_complex_div_wide): Likewise.
1675 (expand_complex_operations_1): Likewise.
1676 (expand_complex_libcall): Preserve EH info of the original stmt.
1677 (tree_lower_complex): Handle removed blocks.
1678 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1679 on complex multiplication and division libcall builtins.
1681 2018-05-04 Richard Biener <rguenther@suse.de>
1684 * fold-const.c (negate_expr_p): Restrict negation of operand
1685 zero of a division to when we know that can happen without
1687 (fold_negate_expr_1): Likewise.
1689 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1692 * real.h (real_nextafter): Declare.
1693 * real.c (real_nextafter): New function.
1694 * fold-const-call.c (fold_const_nextafter): New function.
1695 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1696 CASE_CFN_NEXTTOWARD.
1697 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1698 even when arg1_mode is different from arg0_mode.
1700 2018-05-03 Nathan Sidwell <nathan@acm.org>
1702 * doc/extend.texi (Deprecated Features): Remove
1704 (Backwards Compatibility): Likewise.
1705 * doc/invoke.texi (C++ Language Options): Likewise.
1706 (C++ Dialect Options): Likewise.
1708 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1711 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1712 _mm512_mask_mullox_epi64): New intrinsics.
1714 2018-05-03 Tom de Vries <tom@codesourcery.com>
1717 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1718 dump files): Add offload-tree.
1720 2018-05-03 Richard Biener <rguenther@suse.de>
1722 PR tree-optimization/85615
1723 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1724 to loops not nested in BBs loop father to avoid creating multi-entry
1727 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1729 PR tree-optimization/70291
1730 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1731 arguments. Change return type to tree. Emit libcall as a new
1732 statement rather than replacing existing one when inplace_p is true.
1733 (expand_complex_multiplication_components): New function.
1734 (expand_complex_multiplication): Expand floating-point complex
1735 multiplication using the above.
1736 (expand_complex_division): Rename inner_type parameter to type.
1737 Update expand_complex_libcall call-site.
1738 (expand_complex_operations_1): Update expand_complex_multiplication
1739 and expand_complex_division call-sites.
1741 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1744 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1745 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1746 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1747 the highest significant bit of the shift count mask is clear. In
1748 check whether and[sq]i3 is needed verify that all significant bits
1749 of the shift count other than the highest are set.
1751 2018-05-02 Tom de Vries <tom@codesourcery.com>
1754 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1755 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1756 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1757 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1758 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1759 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1760 __builtin_goacc_parlevel_size.
1762 2018-05-02 Richard Biener <rguenther@suse.de>
1764 PR tree-optimization/85597
1765 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1766 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1769 2018-05-02 Tom de Vries <tom@codesourcery.com>
1772 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1773 dump files): Add ltrans-tree.
1775 2018-05-02 Tom de Vries <tom@codesourcery.com>
1778 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1779 dump files): Add wpa-ipa.
1781 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1783 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1784 powerpc*-*-linux*paired* target.
1785 * config/rs6000/750cl.h: Delete.
1786 * config/rs6000/paired.h: Delete.
1787 * config/rs6000/paired.md: Delete.
1788 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1790 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1791 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1792 comment. Remove paired float support.
1793 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1794 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1796 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1797 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1799 * config/rs6000/rs6000.c: Remove paired float support.
1800 (paired_expand_vector_init, paired_expand_vector_move,
1801 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1802 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1803 paired_expand_builtin, paired_expand_predicate_builtin,
1804 paired_init_builtins): Delete.
1805 * config/rs6000/rs6000.h: Remove paired float support.
1806 * config/rs6000/rs6000.md: Remove paired float support.
1807 (move_from_CR_ov_bit): Delete.
1808 * config/rs6000/rs6000.opt (mpaired): Delete.
1809 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1810 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1812 2018-05-02 Richard Biener <rguenther@suse.de>
1815 * gimplify.c (gimplify_save_expr): When in SSA form allow
1816 SAVE_EXPRs to compute to SSA vars.
1818 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1821 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1822 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1823 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1824 clobber operands[2], instead use a new pseudo. Formatting fixes.
1826 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1828 PR tree-optimization/85586
1829 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1830 exit early for statements in the same group if the accesses are
1833 2018-05-02 Tom de Vries <tom@codesourcery.com>
1836 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1839 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1841 PR tree-optimization/85143
1842 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1844 2018-05-01 Tom de Vries <tom@codesourcery.com>
1847 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1848 not found" error message.
1850 2018-05-01 Tom de Vries <tom@codesourcery.com>
1853 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1854 * vec.c (test_ordered_remove_if): New function.
1855 (vec_c_tests): Call test_ordered_remove_if.
1856 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1857 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1858 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1859 VEC_ORDERED_REMOVE_IF.
1861 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1863 PR tree-optimization/82665
1864 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1865 pointer subtraction where arguments come from a memchr call.
1867 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1869 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1870 --push-state --as-needed and --pop-state instead of --as-needed and
1871 --no-as-needed if ld supports it.
1872 * configure: Regenerated.
1875 * doc/install.texi2html: Replace _002d with - and _002a with * in
1876 generated html files using sed.
1878 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1881 * gcc-rich-location.c (blank_line_before_p): New function.
1882 (use_new_line): New function.
1883 (gcc_rich_location::add_fixit_insert_formatted): New function.
1884 * gcc-rich-location.h
1885 (gcc_rich_location::add_fixit_insert_formatted): New function.
1887 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1889 * selftest.c (assert_streq): Rename "expected" and "actual" to
1890 "val1" and "val2". Extend NULL-handling to cover both inputs
1891 symmetrically, while still requiring both to be non-NULL for a pass.
1892 * selftest.h (assert_streq): Rename "expected" and "actual" to
1894 (ASSERT_EQ): Likewise.
1895 (ASSERT_EQ_AT): Likewise.
1896 (ASSERT_KNOWN_EQ): Likewise.
1897 (ASSERT_KNOWN_EQ_AT): Likewise.
1898 (ASSERT_NE): Likewise.
1899 (ASSERT_MAYBE_NE): Likewise.
1900 (ASSERT_MAYBE_NE_AT): Likewise.
1901 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1902 the assertion to pass.
1903 (ASSERT_STREQ_AT): Likewise.
1905 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1907 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1908 interaction with -pie.
1910 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1912 * selftest.h: Fix alphabetization of per-source-file selftest
1915 2018-04-30 Jason Merrill <jason@redhat.com>
1917 PR c++/61982 - dead stores to destroyed objects.
1918 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1921 2018-04-30 Jason Merrill <jason@redhat.com>
1923 * tree.c (build_clobber): New.
1924 * tree.h: Declare it.
1925 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1927 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1929 * diagnostic-show-locus.c (layout::layout): Update for
1930 location_get_source_line returning a char_span.
1931 (struct char_span): Move to input.h.
1932 (struct correction): Update for fields in char_span becoming
1934 (struct source_line): Update for location_get_source_line
1935 returning a char_span.
1936 (layout::print_line): Likewise.
1937 * edit-context.c (edited_file::print_content): Likewise.
1938 (edited_file::print_diff_hunk): Likewise.
1939 (edited_file::print_run_of_changed_lines): Likewise.
1940 (edited_file::get_num_lines): Likewise.
1941 (edited_line::edited_line): Likewise.
1942 * final.c (asm_show_source): Likewise.
1943 * input.c (location_get_source_line): Convert return type
1944 from const char * to char_span, losing the final "line_len"
1946 (dump_location_info): Update for the above.
1947 (get_substring_ranges_for_loc): Likewise. Use a char_span
1948 when handling the literal within the line.
1949 (test_reading_source_line): Update for location_get_source_line
1950 returning a char_span.
1951 * input.h (class char_span): Move here from
1952 diagnostic-show-locus.c, converting from a struct to a class.
1953 Make data members private.
1954 (char_span::operator bool): New.
1955 (char_span::length): New.
1956 (char_span::get_buffer): New.
1957 (char_span::operator[]): New.
1958 (char_span::subspan): Make const.
1959 (char_span::xstrdup): New.
1960 (location_get_source_line): Convert return type from const char *
1961 to char_span, losing the final "line_size" param.
1963 2018-04-30 Jan Hubicka <jh@suse.cz>
1965 * lto-wrapper.c (ltrans_priorities): New static var.
1966 (cmp_priority): New.
1967 (run_gcc): Read priorities and if doing parallel build order
1968 the Makefile by them.
1970 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1972 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1974 2018-04-30 Richard Biener <rguenther@suse.de>
1976 * tree-cfg.c (verify_address): Remove base argument, add
1977 flag whether to check TREE_ADDRESSABLE and do that.
1978 (verify_expr): Remove.
1979 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1980 (verify_gimple_assign_single): Likewise.
1981 (verify_gimple_switch): Likewise.
1982 (verify_expr_location_1): Dereference tp once. Add (disabled)
1983 piece from verify_expr.
1984 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1986 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1988 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1990 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1992 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1993 (small_data_pattern): Likewise.
1994 (arc_rewrite_small_data): Likewise.
1995 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1996 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1997 (get_symbol_alignment): New function.
1998 (legitimate_small_data_address_p): Likewise.
1999 (legitimate_scaled_address): Update, call
2000 legitimate_small_data_address_p.
2001 (output_sdata): New static variable.
2002 (arc_print_operand): Update how we handle small data operands.
2003 (arc_print_operand_address): Likewise.
2004 (arc_legitimate_address_p): Update, use
2005 legitimate_small_data_address_p.
2006 (arc_rewrite_small_data_p): Remove.
2007 (arc_rewrite_small_data_1): Likewise.
2008 (arc_rewrite_small_data): Likewise.
2009 (small_data_pattern): Likewise.
2010 (compact_sda_memory_operand): Update to use
2011 legitimate_small_data_address_p and get_symbol_alignment.
2012 (prepare_move_operands): Don't rewite sdata pattern.
2013 (prepare_extend_operands): Remove.
2014 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2016 (zero_extendqisi2): Likewise.
2017 (zero_extendhisi2): Likewise.
2018 (extendqihi2): Likewise.
2019 (extendqisi2): Likewise.
2020 (extendhisi2): Likewise.
2024 * config/arc/constraints.md (Usd): Change it to memory constraint.
2026 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2028 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2029 as source of std instructions.
2030 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2031 allow 6-bit constants as source for store instructions.
2032 (movdi_insn): Update instruction pattern to allow 6-bit constants
2033 as source for store instructions.
2035 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2037 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2039 2018-04-30 Nathan Sidwell <nathan@acm.org>
2040 Sandra Loosemore <sandra@codesourcery.com>
2042 * dumpfile.c (dump_open): Allow '-' for stdout.
2043 * doc/invoke.texi (Developer Options): Document dump filename
2044 determination early. Document stdin/stdout selection.
2046 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2048 Microblaze Target: PIC data text relative
2050 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2051 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2053 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2054 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2055 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2056 New addressing mode for data-text relative position indepenedent code.
2057 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2058 'ADDRESS_SYMBOLIC_TXT_REL'.
2059 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2060 (microblaze_legitimate_pic_operand): Exclude function calls from
2061 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2062 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2064 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2065 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2066 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2067 for 'address + offset'.
2068 (microblaze_expand_prologue): Add new function prologue call for
2070 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2071 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2072 table in case of TARGET_PIC_DATA_TEXT_REL.
2073 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2074 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2075 Add new macros 'UNSPEC_TEXT',
2076 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2077 + exclude function calls from 'UNSPEC_PLT' in case of data text
2079 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2080 new target hook for generating address diff vector tables in case of
2082 * doc/tm.texi : Regenerate.
2083 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2084 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2085 of addr diff vector generation.
2086 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2087 target hook definition.
2088 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2089 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2090 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2091 option for data text relative.
2093 2018-04-30 Richard Biener <rguenther@suse.de>
2095 * tree-chrec.h (evolution_function_is_constant_p): Remove
2097 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2099 2018-04-30 Richard Biener <rguenther@suse.de>
2102 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2104 2018-04-30 Richard Biener <rguenther@suse.de>
2106 PR tree-optimization/28364
2107 PR tree-optimization/85275
2108 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2109 copying first exit test.
2111 2018-04-28 Mark Wielaard <mark@klomp.org>
2113 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2115 (dwarf_AT): Handle DW_AT_addr_base.
2116 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2118 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2121 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2122 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2123 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2124 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2126 2018-04-28 Richard Biener <rguenther@suse.de>
2128 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2129 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2130 to reflect use. Only add interesting stmts.
2132 2018-04-27 Martin Jambor <mjambor@suse.cz>
2135 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2136 the jump function allows for passing through aggregate values.
2138 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2140 * input.h (in_system_header_at): Convert from macro to inline
2142 (from_macro_expansion_at): Likewise.
2143 (from_macro_definition_at): Likewise.
2145 2018-04-27 Jeff Law <law@redhat.com>
2147 * config.gcc: Mark tile* targets as deprecated/obsolete.
2149 2018-04-27 Richard Biener <rguenther@suse.de>
2151 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2154 2018-04-27 Richard Biener <rguenther@suse.de>
2156 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2158 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2160 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2161 with Yd constraint. Set "preferred_for_speed" attribute from
2162 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2164 (*movdi_internal): Ditto.
2165 (movti_interunit splitters): Remove
2166 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2167 (movdi_interunit splitters): Ditto.
2168 * config/i386/constraints.md (Ye): Remove.
2169 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2171 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2174 * config/aarch64/constraints.md (Usg): Limit to 31.
2177 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2179 PR tree-optimization/85529
2180 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2181 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2182 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2183 zero extension or masking of the MSB bit.
2184 (optimize_range_tests): Add FIRST_BB argument, pass it through
2185 to optimize_range_tests_var_bound.
2186 (maybe_optimize_range_tests, reassociate_bb): Adjust
2187 optimize_range_tests callers.
2189 2018-04-26 Richard Biener <rguenther@suse.de>
2190 Jakub Jelinek <jakub@redhat.com>
2192 * cgraph.h (symbol_table): Just declare debug method here.
2193 * symtab.c (symbol_table::debug): Define.
2195 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2197 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2199 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2201 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2202 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2203 (*movdi_internal): Substitute Yi and Yj constraint with x
2204 and Ym and Yn constraint with y constraint. Update "isa"
2205 attribute and set "preferred_for_speed" attribute from
2206 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2207 (*movsi_internal): Ditto.
2208 (*movdf_internal): Ditto.
2209 (*movsf_internal): Ditto.
2210 (*zero_extendsidi2): Ditto.
2211 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2212 (sse2_loadld): Ditto.
2213 (*vec_extract<ssevecmodelower>_0): Ditto.
2214 (*vec_extractv4si_0_zext_sse4): Ditto.
2215 (vec_concatv2di): Ditto.
2216 (*vec_dup<mode>): Ditto.
2217 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2218 * config/i386/constraints.md (Yi): Remove.
2223 2018-04-26 Nathan Sidwell <nathan@acm.org>
2225 * dumpfile.c (dump_open): New.
2226 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2227 (dump_finish): Detect stdio/stderr by value not name.
2229 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2231 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2233 2018-04-26 Tom de Vries <tom@codesourcery.com>
2236 * config/nvptx/nvptx.c (verify_neutering_jumps)
2237 (verify_neutering_labels): New function
2238 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2240 2018-04-26 Tom de Vries <tom@codesourcery.com>
2243 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2244 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2247 2018-04-26 Richard Biener <rguenther@suse.de>
2248 Tom de Vries <tom@codesourcery.com>
2251 * lto-streamer-out.c (output_function): Fixup loops if required to match
2252 discovery done in the reader.
2254 2018-04-26 Richard Biener <rguenther@suse.de>
2256 PR tree-optimization/85116
2257 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2258 have a loop exit from the single latch predecessor. Remove
2259 case of header with just condition.
2260 (ch_base::copy_headers): Exclude infinite loops from any
2262 (pass_ch::execute): Record exits.
2264 2018-04-26 Richard Biener <rguenther@suse.de>
2266 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2267 prologue cost vector and pass it to vect_get_load_cost.
2268 (vect_get_peeling_costs_all_drs): Likewise.
2269 (vect_peeling_hash_get_lowest_cost): Likewise.
2270 (vect_enhance_data_refs_alignment): Likewise.
2272 2018-04-26 Richard Biener <rguenther@suse.de>
2275 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2276 checking of integer<->pointer conversions.
2277 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2278 sign-/zero-extending pointer types.
2279 (expand_omp_for_static_chunk): Likewise.
2281 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2282 Jean Lee <xiaoyur347@gmail.com>
2284 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2285 (TARGET_ASAN_SHADOW_OFFSET): Define.
2286 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2287 true for -fsanitize=address.
2289 2018-04-25 Mark Wielaard <mark@klomp.org>
2291 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2294 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2296 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2297 than "alu", remove explicit "memory" and "imm_disp" attributes.
2298 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2301 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2302 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2303 gen_lowpart_no_emit.
2305 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2308 * config/i386/i386.c (ix86_expand_builtin): Change memory
2309 operand to XI, extend p0 to Pmode.
2310 * config/i386/i386.md: Change unspec volatile and operand
2311 1 mode to XI, change operand 0 mode to P.
2313 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2315 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2316 GET_MODE_MASK before any checking.
2317 (nds32_can_use_bset_p): Likewise.
2318 (nds32_can_use_btgl_p): Likewise.
2320 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2322 * config/nds32/nds32-doubleword.md: New define_split pattern for
2323 illegal register number.
2325 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2327 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2329 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2331 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2333 2018-04-25 Richard Biener <rguenther@suse.de>
2335 * lto-streamer.h (LTO_major_version): Bump to 8.
2337 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2339 * BASE-VER: Set to 9.0.0.
2341 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2343 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2344 in __abskf2 and __powikf2.
2346 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2349 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2350 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2351 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2352 Use the above on operand 2. Reindent.
2353 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2355 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2358 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2360 * config/i386/i386.opt (mcet): Removed.
2361 * doc/install.texi: Remove -mcet documentation.
2362 * doc/invoke.texi: Likewise.
2364 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2367 * doc/install.texi: Remove -mcet from bootstrap-cet.
2369 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2372 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2373 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2377 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2378 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2379 containing a CONST_VECTOR.
2381 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2383 * doc/install.texi: Update newlib dependency for nvptx.
2385 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2388 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2389 instead of INTVAL when shifting x left.
2391 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2393 PR tree-optimization/85478
2394 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2395 vect_grouped_store_supported for single element vectors.
2397 2018-04-24 Richard Biener <rguenther@suse.de>
2400 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2401 load cost increase to the case of non-constant step.
2403 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2406 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2407 destination if any_malformed_asm.
2409 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2412 * expr.c (store_field): In the bitfield case, if the value comes from
2413 a function call and is returned in registers by means of a PARALLEL,
2414 do not change the mode of the temporary unless BLKmode and VOIDmode.
2416 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2418 PR rtl-optimization/85423
2419 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2420 dependencies to debug insns when the previous insn is non-debug.
2422 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2424 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2425 enums into a single definition.
2426 (fls): Fix predicates and printing.
2429 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2431 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2432 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2433 and short u6 immediate.
2434 (check_if_valid_sleep_operand): Remove.
2435 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2437 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2439 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2440 flag_always_save_lp condition.
2441 * config/nds32/nds32.opt (malways-save-lp): New option.
2443 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2445 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2446 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2447 * config/nds32/nds32.h
2448 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2449 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2451 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2453 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2454 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2456 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2457 Chung-Ju Wu <jasonwucj@gmail.com>
2459 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2461 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2462 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2464 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2466 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2468 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2470 * config/nds32/nds32-protos.h (nds32_data_alignment,
2471 nds32_local_alignment): Declare.
2472 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2473 nds32_local_alignment): New functions.
2474 (TARGET_CONSTANT_ALIGNMENT): Define.
2475 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2477 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2479 * config/nds32/nds32.c
2480 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2481 (TARGET_MODES_TIEABLE_P): Likewise.
2483 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2485 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2488 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2489 Chung-Ju Wu <jasonwucj@gmail.com>
2491 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2492 for unaligned access.
2493 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2495 * config/nds32/nds32-intrinsic.md: Likewise.
2496 * config/nds32/nds32_intrinsic.h: Likewise.
2497 * config/nds32/nds32.h (nds32_builtins): Likewise.
2498 * config/nds32/nds32.opt (munaligned-access): New option.
2499 * config/nds32/nds32.c (nds32_asm_file_start): Display
2500 flag_unaligned_access status.
2502 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2504 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2505 -mno-relax is present.
2506 * config/riscv/linux.h (LINK_SPEC): Ditto.
2508 2018-04-20 Martin Sebor <msebor@redhat.com>
2511 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2513 (gimple_fold_builtin_stxcpy_chk): Same.
2514 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2516 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2519 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2520 __powikf2 when long double is IEEE 128-bit.
2522 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2524 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2525 step to make sure stack always aligned.
2527 2018-04-20 Carl Love <cel@us.ibm.com>
2530 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2531 size check for arg0.
2533 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2534 Tom de Vries <tom@codesourcery.com>
2537 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2538 Emit insns for calls too.
2539 (nvptx_find_par): Always look for worker-level predecessor insn.
2540 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2542 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2543 (nvptx_process_pars): Propagate frames for calls.
2545 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2548 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2550 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2551 (ix86_handle_option): Don't handle OPT_mibt.
2552 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2554 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2556 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2557 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2558 (ix86_target_macros): Define __CET__ with flag_cf_protection
2559 for -fcf-protection.
2560 * config/i386/i386.c (isa2_opts): Remove -mibt.
2561 * config/i386/i386.h (TARGET_IBT): Removed.
2562 (TARGET_IBT_P): Likewise.
2563 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2564 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2565 * config/i386/i386.opt (mcet): Update help message.
2568 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2569 -mcet as an alias for -mshstk.
2571 2018-04-20 Richard Biener <rguenther@suse.de>
2574 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2575 complexity by forcing a single use of the multiply operand.
2577 2018-04-20 Martin Jambor <mjambor@suse.cz>
2580 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2581 recursion dependency to only apply to non-clones.
2583 2018-04-20 Martin Jambor <mjambor@suse.cz>
2586 * ipa-cp.c (create_specialized_node): Check that clones of
2587 self-recursive edges exist during IPA-CP.
2589 2018-04-19 Toon Moene <toon@moene.org>
2591 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2594 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2596 PR tree-optimization/85467
2597 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2598 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2599 VECTOR_CST element to type.
2601 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2604 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2605 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2606 (builtin_longjmp): Likewise.
2607 (save_stack_nonlocal): New pattern.
2608 (restore_stack_nonlocal): Likewise.
2610 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2613 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2614 Replace ASM_OUTPUT_LABEL with fprintf.
2616 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2619 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2620 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2621 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2622 define __IBT__ and __SHSTK__ for -fcf-protection.
2623 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2625 (ix86_trampoline_init): Likewise.
2626 (x86_output_mi_thunk): Likewise.
2627 (ix86_notrack_prefixed_insn_p): Likewise.
2628 (ix86_option_override_internal): Don't disallow -fcf-protection.
2629 * config/i386/i386.md (rdssp<mode>): Also enable for
2631 (incssp<mode>): Likewise.
2632 (nop_endbr): Likewise.
2633 * config/i386/i386.opt (mcet): Change help message to built-in
2637 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2638 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2639 enable CET built-in functions.
2641 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2643 * common/config/i386/i386-common.c
2644 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2645 OPTION_MASK_ISA_MOVDIRI_UNSET,
2646 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2647 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2648 * config.gcc (movdirintrin.h): New header.
2649 * config/i386/cpuid.h (bit_MOVDIRI,
2650 bit_MOVDIR64B): New bits.
2651 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2653 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2654 (VOID, PVOID, PCVOID)): New function types.
2655 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2656 __builtin_ia32_directstoreu_u64,
2657 __builtin_ia32_movdir64b): New builtins.
2658 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2659 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2661 (ix86_valid_target_attribute_inner_p): Ditto.
2662 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2663 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2664 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2665 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2666 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2667 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2668 (movdiri<mode>, movdir64b_<mode>): New.
2669 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2670 * config/i386/immintrin.h: Include movdirintrin.h.
2671 * config/i386/movdirintrin.h: New file.
2672 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2674 2018-04-19 Richard Biener <rguenther@suse.de>
2677 * cfg.c (clear_bb_flags): When loop state says we have
2678 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2680 2018-04-19 Richard Biener <rguenther@suse.de>
2682 PR tree-optimization/84737
2683 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2684 copying restrict info.
2685 (vect_setup_realignment): Use it.
2686 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2687 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2688 the first DR to all generated stores.
2689 (vectorizable_load): Likewise for loads.
2691 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2693 PR tree-optimization/85446
2694 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2695 the integral and pointer types to have the same precision.
2697 * doc/install.texi: Document --disable-cet being the default and
2700 2018-04-18 Martin Liska <mliska@suse.cz>
2702 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2705 2018-04-18 Martin Liska <mliska@suse.cz>
2708 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2711 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2712 arguments if they are comparable.
2714 2018-04-18 Martin Liska <mliska@suse.cz>
2717 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2720 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2723 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2726 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2727 ENDBR after calling __morestack.
2729 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2732 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2733 by using gcc_base_ver to generate a gcc_driver_version, and use
2734 it when generating GCC_DRIVER_NAME.
2735 * configure: Regenerate.
2737 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2740 * config.gcc: Obsolete powerpc*-*-*spe*.
2742 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2745 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2746 (stabstr_D): Change type of unum from unsigned int to
2747 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2750 2018-04-17 Jim Wilson <jimw@sifive.com>
2753 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2754 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2755 Set arg_pointer_offset after using pretend_args_size.
2757 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2759 PR rtl-optimization/85431
2760 * dse.c (record_store): Ignore zero width stores.
2763 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2764 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2765 __builtin_stack_restore rather than after it.
2766 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2767 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2768 argument instead of virtual_dynamic_stack_rtx.
2770 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2772 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2774 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2775 Add note to error message to explain internal mapping of overloaded
2776 built-in function name to non-overloaded built-in function name.
2777 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2780 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2783 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2784 where the inputs overlap with the output.
2786 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2788 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2789 (=v, v) alternative and explicit "memory" attribute.
2790 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2791 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2793 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2794 "sselog1" type instead of "sselog".
2795 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2796 "sselog". Remove explicit "memory" attribute.
2797 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2798 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2800 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2801 "sselog1" type instead of "sselog". Remove explicit "memory"
2803 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2804 use "sselog1" type instead of "sselog". Remove explicit "memory"
2806 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2807 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2809 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2810 "sselog1" type instead of "sselog". Remove explicit "memory"
2812 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2813 use "sselog1" type instead of "sselog". Remove explicit "memory"
2817 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2820 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2823 2018-04-17 Martin Jambor <mjambor@suse.cz>
2826 * ipa-cp.c (create_specialized_node): Call
2827 expand_all_artificial_thunks if necessary.
2829 2018-04-17 Martin Liska <mliska@suse.cz>
2832 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2833 in message, remote space in between '_G' and '('.
2835 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2838 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2839 avx512f_vmcmp<mode>3<round_saeonly_name>,
2840 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2841 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2842 avx512f_rndscale<mode><round_saeonly_name>,
2843 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2844 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2845 Use %<iptr>2 instead of %2 for -masm=intel.
2846 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2847 avx512f_vcvttss2usi<round_saeonly_name>,
2848 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2850 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2851 avx512f_vcvttsd2usi<round_saeonly_name>,
2852 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2853 Use %q1 instead of %1 for -masm=intel.
2854 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2855 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2856 of %3 for -masm=intel.
2857 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2859 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2861 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2863 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2864 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2866 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2868 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2869 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2870 %g1 and one with %0 and %1.
2871 (avx512er_vmrcp28<mode><round_saeonly_name>,
2872 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2874 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2875 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2876 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2877 of %0 and %{%4%} for -masm=intel.
2878 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2879 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2880 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2881 order of %0 and %{%5%}%{z%} for -masm=intel.
2883 2018-04-17 Jan Hubicka <jh@suse.cz>
2886 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2888 2018-04-17 Martin Liska <mliska@suse.cz>
2891 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2892 for target_clone error message. Make default implementation
2893 clone to be a local declaration.
2894 (separate_attrs): Add new argument and check for an empty
2896 (expand_target_clones): Handle it.
2897 (ipa_target_clone): Make redirection just for target_clones
2900 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2901 Tom de Vries <tom@codesourcery.com>
2904 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2905 tiled basic blocks without omp continue statements.
2907 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2910 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2911 vec_extract expression as having side effects to make sure it gets
2914 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2917 * config/i386/i386.c (get_builtin_code_for_version): Check
2920 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2923 * gcc/config.gcc: Support "skylake".
2924 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2926 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2927 (processor_target_table): Add "skylake".
2928 (ix86_option_override_internal): Add "skylake".
2929 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2930 PROCESSOR_CANNONLAKE.
2931 (get_builtin_code_for_version): Fix priority for
2932 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2933 PROCESSOR_SKYLAKE-AVX512.
2934 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2935 (processor_type): Add PROCESSOR_SKYLAKE.
2937 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2938 Jason Merrill <jason@redhat.com>
2941 * convert.c (convert_to_integer_1): Use direct recursion for
2942 enumeral types and types with a precision less than the number
2943 of bits in their mode.
2945 2018-04-16 Julia Koval <julia.koval@intel.com>
2948 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2949 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2951 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2954 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2955 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2956 and -mno-direct-move.
2958 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2961 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2962 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2963 Ensure negative shifts result in {0}.
2965 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2967 PR rtl-optimization/79916
2968 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2969 regs (if any) to define how to gnerate SD moves when LRA is in
2972 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2974 PR rtl-optimization/85393
2975 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2976 * except.c (expand_dw2_landing_pad_for_region): Make static.
2977 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2978 a label and unconditional jump to old_bb, rather than
2979 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2982 PR rtl-optimization/85376
2983 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2984 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2985 instead of a specific value.
2987 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2988 Bin Cheng <bin.cheng@arm.com>
2990 PR tree-optimization/82965
2991 PR tree-optimization/83991
2992 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2993 by_profile_only parameter.
2994 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2995 information if the loop was predicted to iterate too many times.
2996 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2998 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3001 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3004 2018-04-13 Martin Liska <mliska@suse.cz>
3005 Jakub Jelinek <jakub@redhat.com>
3008 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3009 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3010 * builtins.c (expand_builtin_memory_copy_args): Use
3011 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3012 handle dest_addr == pc_rtx.
3014 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3017 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3018 asked to not generate direct moves.
3019 (fix_trunc<mode>si2_stfiwx): Similar.
3020 (fix_trunc<mode>si2_internal): Similar.
3022 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3025 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3026 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3027 lookup if dest in some wider mode is known to be const0_rtx and
3028 if so, record permanent equivalence for it to be ZERO_EXTEND of
3029 the narrower mode destination.
3031 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3033 * lto-streamer-out.c (output_function): Revert 259346.
3034 * omp-expand.c (expand_oacc_for): Likewise.
3036 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3038 PR rtl-optimization/85354
3039 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3040 * sel-sched.c (sel_global_init): ... here.
3042 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3045 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3046 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3047 mode for PE-COFF targets.
3048 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3049 (i386_pe_asm_lto_end): Likewise.
3050 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3051 (TARGET_ASM_LTO_END): Likewise.
3052 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3053 (i386_pe_asm_lto_start): New function.
3054 (i386_pe_asm_lto_end): Likewise.
3056 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3057 Richard Biener <rguenther@suse.de>
3060 * lto-streamer-out.c (output_function): Fix CFG loop state before
3062 * omp-expand.c (expand_oacc_for): Handle calls to internal
3063 functions like regular functions.
3065 2018-04-12 Richard Biener <rguenther@suse.de>
3068 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3069 for the early LTO debug to properly generate references to it
3070 during DIE emission. Do not re-use that for the skeleton for
3072 (dwarf2out_early_finish): Likewise.
3074 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3077 * config/i386/sse.md
3078 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3079 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3080 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3081 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3082 and output is a reg, avoid creating invalid lowpart subreg, but
3083 instead split into a 512-bit move. Don't split if not AVX512VL,
3084 input is xmm16+ reg and output is a mem.
3085 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3086 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3087 xmm16+ reg and output is a mem.
3089 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3091 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3092 also for flag_dwarf2_cfi_asm.
3094 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3096 PR rtl-optimization/85342
3097 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3098 a bool scalar var inside of the loop instead. Don't try to update
3099 recog_data.operand after failed apply_change_group.
3101 2018-04-12 Tom de Vries <tom@codesourcery.com>
3104 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3105 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3106 array with flexible array member as array without given dimension.
3107 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3108 argument for undefined param to true.
3110 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3113 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3114 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3115 from PowerPC section.
3116 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3117 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3118 help text that is too long.
3119 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3120 help text that is too long.
3121 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3122 help text that is too long.
3124 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3126 * config/alpha/alpha.md (stack_probe_internal): Rename
3127 from "probe_stack". Update all callers.
3129 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3131 PR rtl-optimization/84566
3132 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3133 sched_macro_fuse_insns.
3135 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3138 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3139 (compute_block_dependences): ... from here.
3141 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3143 PR tree-optimization/85331
3144 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3145 from int to HOST_WIDE_INT.
3147 2018-04-11 Martin Jambor <mjambor@suse.cz>
3150 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3151 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3152 not the same as the source val.
3153 (cgraph_edge_brings_value_p): New parameter.
3154 (gather_edges_for_value): Pass destination value to
3155 cgraph_edge_brings_value_p.
3156 (perhaps_add_new_callers): Likewise.
3157 (get_info_about_necessary_edges): Likewise and exclude values brought
3158 only by self-recursive edges.
3159 (create_specialized_node): Redirect only clones of self-calling edges.
3160 (+self_recursive_pass_through_p): New function.
3161 (find_more_scalar_values_for_callers_subset): Use it.
3162 (find_aggregate_values_for_callers_subset): Likewise.
3163 (known_aggs_to_agg_replacement_list): Removed.
3164 (decide_whether_version_node): Re-calculate known constants for all
3165 remaining context clones.
3167 2018-04-11 Richard Biener <rguenther@suse.de>
3170 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3171 from early DWARF output.
3172 (dwarf2out_early_finish): Output line info unconditionally into
3173 early DWARF and add reference to it.
3175 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3178 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3179 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3180 other than V2DFmode using iptr mode attribute.
3181 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3183 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3185 PR rtl-optimization/84659
3186 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3188 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3191 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3193 (output_loc_list): Pass address of a dummy size variable even in the
3194 locview handling loop.
3195 (index_location_lists): Add comment on why skip_loc_list_entry can't
3198 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3201 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3204 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3207 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3208 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3209 and -mstring-compare-inline-limit.
3211 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3214 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3215 for stack clash protection in a register whenever we need it to be in
3218 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3220 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3221 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3223 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3226 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3228 (mlong-double-): Ditto.
3229 * config/rs6000/sysv4.opt (msdata=): Ditto.
3230 (mtls-size=): Ditto.
3232 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3234 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3235 erroneous entries for
3236 "vector int vec_ldl (int, long int *)", and
3237 "vector unsigned int vec_ldl (int, unsigned long int *)".
3238 Add comments and entries for
3239 "vector bool char vec_ldl (int, bool char *)",
3240 "vector bool short vec_ldl (int, bool short *)",
3241 "vector bool int vec_ldl (int, bool int *)",
3242 "vector bool long long vec_ldl (int, bool long long *)",
3243 "vector pixel vec_ldl (int, pixel *)",
3244 "vector long long vec_ldl (int, long long *)",
3245 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3246 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3247 type tree bool_long_long_type_node and correct definition of
3248 bool_V2DI_type_node to make reference to this new type tree.
3249 (rs6000_mangle_type): Replace erroneous reference to
3250 bool_long_type_node with bool_long_long_type_node.
3251 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3252 comments to emphasize sign distinctions for char and int types and
3253 replace RS6000_BTI_bool_long constant with
3254 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3255 use of RS6000_BTI_pixel.
3256 (bool_long_type_node): Remove this macro definition.
3257 (bool_long_long_type_node): New macro definition
3259 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3261 PR rtl-optimization/85300
3262 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3263 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3264 simplify_unary_operation fails.
3266 2018-04-10 Martin Liska <mliska@suse.cz>
3268 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3269 cgraph_edge and ipa_ref.
3271 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3275 * config/i386/sse.md
3276 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3277 computation of the VEC_MERGE selector from mask.
3278 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3279 Fix decoding of the VEC_MERGE selector into mask.
3281 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3283 PR tree-optimization/85286
3284 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3286 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3288 * final.c (final_1): Set insn_last_address as well as
3289 insn_current_address.
3291 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3294 * explow.c (emit_stack_probe): Call validize_mem on memory location
3295 before passing it to gen_probe_stack. Create address operand and
3296 legitimize it for the probe_stack_address case.
3298 2018-04-09 Jan Hubicka <jh@suse.cz>
3301 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3302 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3303 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3305 (free_lang_data): Rebuild type inheritance graph.
3307 2018-04-09 Martin Sebor <msebor@redhat.com>
3309 * invoke.texi (-finline-small-functions): Mention other optimization
3311 (-findirect-inlining, -fpartial-inlining): Same.
3312 (-finline-functions-called-once): Same.
3313 (-freorder-blocks-and-partition): Same.
3315 2018-04-09 Jan Hubicka <jh@suse.cz>
3318 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3319 jumps; choose last target that matches the criteria (i.e.
3320 no partition changes for non-crossing jumps).
3321 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3322 support for redirecting crossing jumps to non-crossing.
3324 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3326 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3327 also for naked functions.
3329 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3331 * config/arc/arc.md (add_shift): New pattern.
3332 (add_shift2): Likewise.
3333 (sub_shift): Likewise.
3334 (sub_shift_cmp0_noout): Likewise.
3335 (compare_si_ashiftsi): Likewise.
3336 (xbfu_cmp0_noout): New combine pattern.
3337 (xbfu_cmp0"): Likewise.
3338 (movsi_set_cc_insn): Place the predicable variant first.
3339 (commutative_binary_cmp0_noout): Remove clobber.
3340 (commutative_binary_cmp0): New pattern.
3341 (noncommutative_binary_cmp0): Likewise.
3342 (noncommutative_binary_cmp0_noout): Likewise.
3343 (noncommutative_binary_comparison_result_used): Removed.
3344 (rsub_cmp0): New pattern.
3345 (rsub_cmp0_noout): Likewise.
3346 (extzvsi): Changed, keep only meaningful variants.
3347 (SQH, SEZ): New iterators.
3348 (SQH_postfix): New mode attribute.
3349 (SEZ_prefix): New code attribute.
3350 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3351 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3352 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3354 (noncommutative_operator): Check the availability of barrel
3357 2018-04-09 Richard Biener <rguenther@suse.de>
3359 PR tree-optimization/85284
3360 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3361 Only use the niter constraining form of simple_iv when the exit
3364 2018-04-09 Tom de Vries <tom@codesourcery.com>
3367 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3368 (define_expand "*memory_barrier"): New define_expand.
3369 (define_insn "memory_barrier"): New insn.
3371 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3373 PR rtl-optimization/80463
3374 PR rtl-optimization/83972
3375 PR rtl-optimization/83480
3377 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3378 correct producer for the insn.
3379 (tidy_control_flow): Fixup seqnos in case of debug insns.
3381 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3383 PR rtl-optimization/83913
3385 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3386 different sched-times when merging exprs.
3388 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3390 PR rtl-optimization/83962
3392 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3393 tidy_fallthru_edge and tidy_control_flow.
3395 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3397 PR rtl-optimization/83530
3399 * sel-sched.c (force_next_insn): New global variable.
3400 (remove_insn_for_debug): When force_next_insn is true, also leave only
3401 next insn in the ready list.
3402 (sel_sched_region): When the region wasn't scheduled, make another pass
3403 over it with force_next_insn set to 1.
3405 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3407 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3409 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3410 for interrupt control.
3411 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3412 functions for interrupt control.
3413 * config/nds32/nds32-intrinsic.md: Likewise.
3414 * config/nds32/nds32_intrinsic.h: Likewise.
3415 * config/nds32/nds32.h (nds32_builtins): Likewise.
3417 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3419 * config/nds32/nds32.c (nds32_init_machine_status,
3420 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3421 strict_aligned_p field.
3422 (nds32_expand_to_rtl_hook): New function.
3423 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3424 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3426 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3427 Chung-Ju Wu <jasonwucj@gmail.com>
3429 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3430 * config/nds32/nds32-n7.md: New file.
3431 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3432 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3434 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3435 * config/nds32/nds32.md (pipeline_model): Add n7.
3436 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3437 * config/nds32/pipelines.md: Include n7 settings.
3439 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3440 Chung-Ju Wu <jasonwucj@gmail.com>
3442 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3443 * config/nds32/nds32-e8.md: New file.
3444 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3445 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3447 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3448 * config/nds32/nds32.md (pipeline_model): Add e8.
3449 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3450 * config/nds32/pipelines.md: Include e8 settings.
3452 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3453 Chung-Ju Wu <jasonwucj@gmail.com>
3455 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3456 * config/nds32/nds32-n8.md: New file.
3457 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3458 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3460 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3461 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3462 * config/nds32/nds32.md (pipeline_model): Add n8.
3463 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3464 * config/nds32/pipelines.md: Include n8 settings.
3466 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3467 Chung-Ju Wu <jasonwucj@gmail.com>
3469 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3470 * config/nds32/nds32-n9-2r1w.md: New file.
3471 * config/nds32/nds32-n9-3r2w.md: New file.
3472 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3473 nds32_register_ports): New or modify for cpu n9.
3474 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3476 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3477 * config/nds32/nds32-utils.c: New file.
3478 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3479 TARGET_MUL_SLOW): Define.
3480 * config/nds32/nds32.md (pipeline_model): New attribute.
3481 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3482 New options that support cpu n9.
3483 * config/nds32/pipelines.md: Include n9 settings.
3484 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3486 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3488 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3489 information if necessary.
3490 (output_cond_branch_compare_zero): Likewise.
3491 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3492 (nds32_target_alignment): Refine for alignment.
3493 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3494 (FUNCTION_BOUNDARY): Modify.
3495 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3497 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3499 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3501 * config/nds32/constants.md (unspec_volatile_element): Add values for
3502 TLB operation and data prefetch.
3503 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3504 functions for TLB operation and data prefetch.
3505 * config/nds32/nds32-intrinsic.md: Likewise.
3506 * config/nds32/nds32_intrinsic.h: Likewise.
3507 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3508 (nds32_print_operand): Likewise.
3509 * config/nds32/nds32.h (nds32_builtins): Likewise.
3511 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3512 Andrew Pinski <pinsika@gcc.gnu.org>
3515 * match.pd: Use constant_boolean_node of correct type instead of
3516 boolean_true_node or boolean_false_node for simplifying
3517 pointer comparisons to zero.
3519 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3521 PR tree-optimization/80021
3522 * tree.c (verify_type_variant): Make error call in verify_variant_match
3523 translatable and remove final full stop.
3525 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3527 * config/nds32/constants.md (unspec_volatile_element): Add
3528 UNSPEC_VOLATILE_EH_RETURN.
3529 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3530 nds32_output_stack_pop): Support dwarf exception handling process.
3531 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3532 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3533 exception handling process.
3534 (nds32_compute_stack_frame): Likewise.
3535 (nds32_return_addr_rtx): Likewise.
3536 (nds32_initial_elimination_offset): Likewise.
3537 (nds32_expand_prologue): Likewise.
3538 (nds32_expand_epilogue): Likewise.
3539 (nds32_dynamic_chain_address): New function.
3540 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3542 (DYNAMIC_CHAIN_ADDRESS): Define.
3543 (EH_RETURN_DATA_REGNO): Define.
3544 (EH_RETURN_STACKADJ_RTX): Define.
3545 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3546 patterns for dwarf exception handling.
3548 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3550 * config/nds32/nds32.h: Clean up obsolete macros.
3552 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3554 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3555 Add enum values for particular instructions.
3556 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3557 particular intrinsic functions.
3558 * config/nds32/nds32-intrinsic.md: Likewise.
3559 * config/nds32/nds32_intrinsic.h: Likewise.
3560 * config/nds32/nds32.h (nds32_builtins): Likewise.
3561 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3562 (btst, ave): New patterns for particular instructions.
3564 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3566 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3567 Add enum values for atomic load/store and memory sync.
3568 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3570 * config/nds32/nds32-intrinsic.md: Likewise.
3571 * config/nds32/nds32_intrinsic.h: Likewise.
3572 * config/nds32/nds32.h (nds32_builtins): Likewise.
3574 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3576 PR tree-optimization/85257
3577 * fold-const.c (native_encode_vector): If not all elts could fit
3578 and off is -1, return 0 rather than offset.
3579 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3580 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3581 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3582 adjust buffer in native_interpret_expr call.
3584 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3586 * config/nds32/constants.md (unspec_volatile_element): Add cache
3587 control enum values.
3588 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3589 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3590 * config/nds32/nds32.c (nds32_cctl_names): New.
3591 (nds32_print_operand): Handle cache control register names.
3592 * config/nds32/nds32.h (nds32_builtins): New enum values.
3593 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3595 * config/nds32/nds32.md (type): Add mmu.
3596 * config/nds32/pipelines.md (simple_insn): Add mmu.
3598 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3600 * config/nds32/nds32.md (type): Remove call.
3601 * config/nds32/pipelines.md (simple_insn): Likewise.
3603 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3605 * config/nds32/constants.md (unspec_volatile_element): Add
3606 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3607 UNSPEC_VOLATILE_FMFCFG.
3608 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3609 description for fmfcfg and fmfcsr.
3610 (bdesc_1arg): Add fmtcsr.
3611 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3612 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3613 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3614 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3615 unspec_fmfcfg): New patterns.
3616 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3617 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3618 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3619 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3620 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3621 __nds32__fmfcfg): Define.
3623 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3625 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3626 intrinsic register names.
3627 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3628 intrinsic register enum values and macros.
3630 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3632 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3633 for load/store addressing form.
3634 (nds32_print_operand_address): Likewise.
3636 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3639 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3640 based on LABEL_REF. Remove useless assertion.
3641 (pic_address_needs_scratch): Fix formatting.
3642 (sparc_legitimize_pic_address): Minor tweaks.
3643 (sparc_delegitimize_address): Adjust assertion accordingly.
3644 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3645 into symbolic_operand.
3646 (movsi_high_pic_label_ref): Likewise.
3647 (movsi_lo_sum_pic_label_ref): Likewise.
3648 (movdi_pic_label_ref): Likewise.
3649 (movdi_high_pic_label_ref): Likewise.
3650 (movdi_lo_sum_pic_label_ref): Likewise.
3652 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3654 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3655 custom LIB_SPEC setup.
3657 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3658 Kito Cheng <kito.cheng@gmail.com>
3660 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3661 * config/riscv/freebsd.h: New.
3663 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3665 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3666 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3669 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3670 Kito Cheng <kito.cheng@gmail.com>
3672 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3673 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3674 * config/nds32/nds32-protos.h (nds32_output_call,
3675 nds32_output_return): Declare.
3676 * config/nds32/nds32.md: Refine all the call and return patterns.
3678 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3681 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3682 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3684 PR rtl-optimization/84872
3685 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3686 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3689 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3691 * expr.c (copy_blkmode_to_reg): Revert 254862.
3692 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3694 2018-04-06 Richard Biener <rguenther@suse.de>
3697 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3698 after seeing a component reference with an adjacent field. Treat
3699 refs to arrays at struct end of external decls similar to
3700 refs to unconstrained commons.
3702 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3705 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3706 look through SAVE_EXPRs with non-side-effects argument. Adjust
3708 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3711 2018-04-06 Richard Biener <rguenther@suse.de>
3714 * alias.c (find_base_term): New wrapper around find_base_term
3715 unwinding CSELIB_VAL_PTR changes.
3716 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3719 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3721 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3723 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3724 constant definitions.
3725 ("nop"): lr 0,0 -> nopr r0
3726 ("nop_lr0", "nop_lr1"): New insn definitions.
3728 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3730 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3731 NDS32_V3PUSH_AVAILABLE_P macro.
3733 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3734 Chung-Ju Wu <jasonwucj@gmail.com>
3736 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3737 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3738 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3739 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3740 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3741 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3742 * config/nds32/constraints.md: New constraints and checking for hard
3743 float configuration.
3744 * config/nds32/iterators.md: New mode iterator and attribute for hard
3745 float configuration.
3746 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3748 * config/nds32/nds32-fpu.md: New file.
3749 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3750 deal with hard float code generation.
3751 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3753 (abi_type, float_reg_number): New enum type.
3754 * config/nds32/nds32-predicates.c: New predicates for hard float.
3755 * config/nds32/nds32-protos.h: Declare functions for hard float.
3756 * config/nds32/nds32.c: Implementation for hard float configuration.
3757 * config/nds32/nds32.h: Definitions for hard float configuration.
3758 * config/nds32/nds32.md: Include hard float machine description and
3759 modify patterns for hard float configuration.
3760 * config/nds32/nds32.opt: New options for hard float configuration.
3761 * config/nds32/predicates.md: New predicates for hard float
3764 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3766 * common/config/nds32/nds32-common.c
3767 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3769 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3772 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3773 CONSTRUCTOR_ELT (ctor, ...)->value.
3775 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3778 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3780 2018-04-05 Tom de Vries <tom@codesourcery.com>
3783 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3786 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3787 Kito Cheng <kito.cheng@gmail.com>
3789 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3790 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3791 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3792 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3794 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3795 Kito Cheng <kito.cheng@gmail.com>
3797 * config/nds32/constraints.md (Ufe): New memory constraint.
3798 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3799 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3800 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3802 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3803 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3805 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3807 * config/nds32/nds32.md: Use optimize_size in the condition for
3808 alu-shift instructions.
3810 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3812 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3814 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3816 * config/nds32/nds32.md (negsi2): Refine pattern.
3818 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3819 Chung-Ju Wu <jasonwucj@gmail.com>
3821 * config/nds32/iterators.md (shift_rotate): New code iterator.
3822 (shift): New code attribute.
3823 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3824 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3825 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3826 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3827 bit-wise operations.
3828 (andsi3, *andsi3): Ditto.
3829 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3830 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3831 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3832 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3833 nds32_ior_operand, nds32_xor_operand): New predicates.
3835 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3837 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3838 (addsi3, subsi3): ... this.
3840 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3842 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3844 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3846 * config/nds32/nds32.md: Adjust indention.
3848 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3850 * config/nds32/nds32.md (feature): New attribute.
3852 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3854 * config/nds32/nds32.md (subtype): New attribute.
3856 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3859 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3860 expansion to perform a bitwise AND of the argument followed by a
3861 boolean negation of the result.
3863 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3865 PR rtl-optimization/84878
3866 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3867 the basic block. Assert the use reference is not artificial and that
3868 it has an associated insn.
3870 2018-04-04 Michael Matz <matz@suse.de>
3872 * builtins.c (compute_objsize): Pass correct operand
3873 to array_at_struct_end_p.
3875 2018-04-04 Richard Biener <rguenther@suse.de>
3878 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3879 from contexts for DINFO_LEVEL_TERSE and below.
3881 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3883 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3884 resiter_operand condition.
3885 * config/nds32/nds32.md (*move<mode>): Ditto.
3887 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3888 Monk Chiang <sh.chiang04@gmail.com>
3890 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3892 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3894 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3896 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3897 Kito Cheng <kito.cheng@gmail.com>
3899 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3900 nds32_cond_code_str, output_cond_branch,
3901 output_cond_branch_compare_zero, nds32_expand_cbranch,
3902 nds32_expand_cstore, nds32_expand_movcc,
3903 nds32_output_cbranchsi4_equality_zero,
3904 nds32_output_cbranchsi4_equality_reg,
3905 nds32_output_cbranchsi4_equality_reg_or_const_int,
3906 nds32_output_cbranchsi4_greater_less_zero: New functions.
3907 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3908 nds32_expand_cstore, nds32_expand_movcc,
3909 nds32_output_cbranchsi4_equality_zero,
3910 nds32_output_cbranchsi4_equality_reg,
3911 nds32_output_cbranchsi4_equality_reg_or_const_int,
3912 nds32_output_cbranchsi4_greater_less_zero): Declare.
3913 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3914 nds32_rimm11s_operand): New predicates.
3915 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3916 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3919 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3921 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3922 * config/nds32/nds32.md: Ditto.
3923 * config/nds32/pipelines.md: Ditto.
3925 2018-04-04 Richard Biener <rguenther@suse.de>
3927 PR tree-optimization/85168
3928 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3929 propagating abnormals.
3931 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3933 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3935 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3936 Kito Cheng <kito.cheng@gmail.com>
3938 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3939 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3940 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3941 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3942 * config/nds32/nds32.md (sibcall_internal): New.
3943 (sibcall_register): Remove.
3944 (sibcall_immediate): Remove.
3945 (sibcall_value_internal): New.
3946 (sibcall_value_register): Remove.
3947 (sibcall_value_immediate): Remove.
3948 * config/nds32/predicates.md (nds32_general_register_operand): New.
3949 (nds32_call_address_operand): New.
3951 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3953 PR rtl-optimization/85167
3954 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3955 bb_defs if *split_p, instead preinitialize it to NULL.
3957 PR tree-optimization/85156
3958 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3959 evaluating the argument multiple times.
3961 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3963 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3965 (_mm_cvtpd_ps): Likewise.
3966 (_mm_cvttpd_epi32): Likewise.
3967 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3968 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3969 vector, pixel, and bool following altivec.h include.
3971 2018-04-03 Martin Sebor <msebor@redhat.com>
3973 * doc/extend.texi (Common Function Attributes): Clarify.
3974 (const attribute): Likewise.
3975 (pure attribute): Likewise.
3977 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3980 * config/i386/i386.c (ix86_expand_vector_set): Use
3981 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3983 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3985 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3986 instructions when changing rounding bits to preserve precision bits
3987 in the x87 control word.
3989 2018-04-03 Martin Liska <mliska@suse.cz>
3991 PR tree-optimization/82491
3992 * rtl.h (strip_offset_and_add): Replace += suboffset with
3993 poly_uint64 () + suboffset.
3995 2018-03-29 Martin Liska <mliska@suse.cz>
3996 Martin Jambor <mjambor@suse.cz>
3999 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4000 param_type is not an integral or pointer type.
4002 2018-04-03 Richard Biener <rguenther@suse.de>
4004 * sese.h (recompute_all_dominators): Remove.
4006 2018-04-02 Martin Sebor <msebor@redhat.com>
4008 * doc/invoke.texi (-Wrestrict): Fix typos.
4010 2018-04-02 Jim Wilson <jimw@sifive.com>
4012 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4013 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4014 (<optab>di3, <optab>si3_extend): Likewise.
4015 (<optab>si3_mask, <optab>si3_mask_1): New.
4016 (<optab>di3_mask, <optab>di3_mask_1): New.
4017 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4018 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4019 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4021 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4023 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4026 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4028 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4029 (nds32_canonicalize_comparison): New function.
4031 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4032 Kito Cheng <kito.cheng@gmail.com>
4033 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4035 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4036 * config/nds32/constants.md (unspec_volatile_element): Add
4037 UNSPEC_VOLATILE_RELAX_GROUP.
4038 * config/nds32/nds32-relax-opt.c: New file.
4039 * config/nds32/nds32-predicates.c
4040 (nds32_symbol_load_store_p): New function.
4041 * config/nds32/nds32-protos.h
4042 (nds32_symbol_load_store_p): Declare function.
4043 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4044 * config/nds32/nds32.c
4045 (nds32_register_pass): New function to register pass.
4046 (nds32_register_passes): New function to register passes.
4047 * config/nds32/nds32.md (relax_group): New pattern.
4048 * config/nds32/nds32.opt (mrelax-hint): New option.
4049 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4051 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4053 * config/nds32/t-nds32: Modify files dependency.
4055 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4057 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4058 (PROFILE_HOOK): Define its implementation.
4060 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4062 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4063 type and 32-bit size.
4065 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4068 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4069 (V_128_256): New mode iterator.
4070 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4071 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4072 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4074 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4075 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4077 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4080 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4081 NaN inputs correctly.
4083 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4086 * config/rs6000/vsx.md (??r): New mode attribute.
4087 (*vsx_mov<mode>_64bit): Use it.
4088 (*vsx_mov<mode>_32bit): Likewise.
4090 2018-03-30 Martin Sebor <msebor@redhat.com>
4092 PR tree-optimization/84818
4093 * builtins.c (check_access): Use warning_n.
4095 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4098 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4100 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4103 2018-03-30 Julia Koval <julia.koval@intel.com>
4106 * x86-tune.def (movx, partial_reg_dependency): Enable for
4109 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4112 * lra-constraints.c (process_alt_operands): Move setting
4113 this_alternative_matches below.
4115 2018-03-29 Martin Liska <mliska@suse.cz>
4118 * doc/invoke.texi: Document how LTO works with debug info.
4119 Describe auto-load support of binutils. Mention 'x86-64'
4120 as valid option value of -march option.
4122 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4124 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4127 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4128 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4129 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4132 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4135 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4136 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4137 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4138 for RS6000_BTM_POWERPC64.
4139 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4140 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4141 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4146 2018-03-28 Carl Love <cel@us.ibm.com>
4149 2017-09-27 Carl Love <cel@us.ibm.com>
4151 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4152 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4153 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4156 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4158 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4159 instead of __vector bool.
4160 (_mm_max_pu8): Likewise.
4161 (_mm_min_pi16): Likewise.
4163 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4166 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4167 (DIVWEUO): Likewise.
4169 (DIVDEUO): Likewise.
4170 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4171 DIVWEUO and DIVDEUO.
4172 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4173 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4174 (div_extend): Likewise.
4175 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4177 (__builtin_divweuo): Likewise.
4178 (__builtin_divdeo): Likewise.
4179 (__builtin_divdeuo): Likewise.
4181 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4184 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4185 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4187 PR tree-optimization/82004
4188 * gimple-match-head.c (optimize_pow_to_exp): New function.
4189 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4190 Don't fold to exp if optimize_pow_to_exp is false.
4192 2018-03-28 Martin Liska <mliska@suse.cz>
4195 * calls.c (initialize_argument_information): Fix trailing space.
4196 * common.opt: Fix typo and provide better explanation for
4197 -fsanitize-coverage option.
4198 * config/i386/i386.opt: Fix typo.
4200 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4201 Martin Liska <mliska@suse.cz>
4204 * gimplify.c (asan_poison_variable): Don't do the check for
4205 gimplify_omp_ctxp here.
4206 (gimplify_decl_expr): Do it here.
4207 (gimplify_target_expr): Likewise.
4209 2018-03-28 Martin Liska <mliska@suse.cz>
4212 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4213 chkp_type_bounds_count if MPX is not enabled.
4215 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4217 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4219 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4222 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4223 function to create the function decl for complex long double
4224 multiply and divide for -mabi=ieeelongdouble.
4225 (init_float128_ieee): Call it.
4227 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4230 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4231 -fcf-protection=branch -mibt.
4232 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4234 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4237 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4239 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4242 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4243 extern array declarations.
4245 2018-03-27 Richard Biener <rguenther@suse.de>
4248 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4249 explicit single_use checks.
4251 2018-03-27 Richard Biener <rguenther@suse.de>
4253 PR tree-optimization/85082
4254 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4257 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4259 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4260 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4261 Turn on fasynchronous-unwind-tables and funwind-tables.
4263 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4266 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4267 (*bmi_blsr_<mode>_ccz): Ditto.
4269 2018-03-26 Tom de Vries <tom@codesourcery.com>
4271 PR tree-optimization/85063
4272 * omp-general.c (offloading_function_p): New function. Factor out
4274 * omp-offload.c (pass_omp_target_link::gate): ... here.
4275 * omp-general.h (offloading_function_p): Declare.
4276 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4277 with attribute omp declare target for offloading functions.
4279 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4281 PR tree-optimization/84005
4282 * tree-data-ref.h (get_base_for_alignment): Declare.
4283 * tree-data-ref.c (get_base_for_alignment_1): New function.
4284 (get_base_for_alignment): Likewise.
4285 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4286 get_base_for_alignment to find a suitable base object, instead
4287 of always using drb->base_address.
4289 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4292 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4293 known size by default.
4295 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4298 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4299 and non BLKmode operands.
4301 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4304 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4305 Clean up attributes.
4307 2018-03-23 Richard Biener <rguenther@suse.de>
4310 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4311 we are going to emit early debug for LTO.
4313 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4316 * function.c (match_asm_constraints_1): Don't optimize if input
4317 doesn't satisfy general_operand predicate for output's mode.
4320 * alias.c (write_dependence_p): Don't require for x_canonicalized
4321 non-VOIDmode if x has VOIDmode.
4324 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4325 just don't try to optimize it rather than assert it never happens.
4327 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4329 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4330 macro expansions for definition of ST_INTERNAL_<mode> and
4331 LD_INTERNAL_<mode> builtins.
4332 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4334 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4336 (altivec_expand_st_builtin): Likewise.
4337 (altivec_expand_builtin): Remove calls to deleted functions.
4338 (rs6000_address_for_altivec): Delete this function.
4339 * config/rs6000/vector.md: Remove expands for
4340 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4342 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4345 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4346 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4347 re-computing once computed.
4348 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4349 (arm_init_machine_status): Initialize
4350 machine->static_chain_stack_bytes.
4352 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4355 * doc/extend.texi: Add four new prototypes for vec_ld.
4356 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4357 definitions for more logical presentation.
4358 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4359 entries for V1TI variants of __builtin_altivec_ld builtin.
4360 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4361 handling of V1TI variant of LVX icode pattern.
4362 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4363 (rs6000_gimple_fold_builtin): Likewise.
4364 (altivec_init_builtins): Add code to define
4365 __builtin_altivec_lvx_v1ti function.
4367 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4370 * function.c (match_asm_constraints_1): Don't do the optimization
4371 if input isn't a REG, SUBREG, MEM or constant.
4373 2018-03-22 Tom de Vries <tom@codesourcery.com>
4375 PR tree-optimization/84956
4376 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4377 bb_has_abnormal_pred.
4379 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4382 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4383 DECL_INITIAL (decl) to decl at the end.
4384 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4387 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4389 * doc/extend.texi (__builtin_tgmath): Document when complex
4390 integer types are treated as _Complex _Float64.
4392 2018-03-21 Tom de Vries <tom@codesourcery.com>
4394 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4396 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4398 PR tree-optimization/84960
4399 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4400 if it is ENTRY block, move them into single succ of ENTRY in that case.
4402 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4404 PR tree-optimization/84811
4405 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4406 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4407 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4408 (known_subrange_p): Update accordingly. Cast each value involved
4409 in the size comparison, rather than casting the result of the
4412 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4414 PR tree-optimization/84982
4415 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4416 by flipping the least significant bit rather than all bits from
4417 bitpos to bitpos + bitsize - 1.
4419 2018-03-21 Nathan Sidwell <nathan@acm.org>
4421 * doc/extend.texi (Deprecated Features): Remove mention of
4422 long-deleted deprecations.
4424 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4427 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4428 * configure: Regenerate.
4430 2018-03-21 Tom de Vries <tom@codesourcery.com>
4432 PR tree-optimization/83126
4433 * tree-parloops.c (num_phis): New function.
4434 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4436 2018-03-21 Nathan Sidwell <nathan@acm.org>
4438 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4439 mention anon-struct/union members and trailing attributes.
4441 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4443 PR tree-optimization/84969
4444 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4445 builtin memset partitions if they set different rhs values.
4447 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4449 PR rtl-optimization/84989
4450 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4451 VEC_DUPLICATE with scalar result mode.
4453 2018-03-21 Martin Liska <mliska@suse.cz>
4456 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4457 not intended return statement.
4459 2018-03-21 Martin Liska <mliska@suse.cz>
4462 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4463 (chkp_find_bound_slots_1): Limit number of iterations.
4465 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4468 * Minor grammar fixes for x86 options.
4470 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4473 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4474 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4476 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4479 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4480 (altivec_lvx_<mode>_1op): Likewise.
4481 (altivec_stvx_<mode>_2op): Likewise.
4482 (altivec_stvx_<mode>_1op): Likewise.
4483 (altivec_lvx_<VM2:mode>): New define_expand.
4484 (altivec_stvx_<VM2:mode>): Likewise.
4485 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4486 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4487 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4488 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4489 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4490 (rs6000_gen_lvx): Likewise.
4491 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4492 (altivec_expand_stv_builtin): Likewise.
4493 (altivec_expand_builtin): Likewise.
4494 * config/rs6000/vector.md: Likewise.
4496 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4499 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4502 2018-03-20 Richard Biener <rguenther@suse.de>
4505 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4506 sign-conversions as zero, fall back to standard scalar_stmt
4509 2018-03-20 Martin Liska <mliska@suse.cz>
4512 * predict.c (rebuild_frequencies): Handle case when we have
4513 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4515 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4518 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4519 flag_section_anchors.
4520 * varasm.c (use_blocks_for_decl_p): Remove hack for
4521 dw2_force_const_mem.
4524 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4526 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4527 be created, use lowpart_subreg of operands[0] rather than operands[0]
4529 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4530 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4531 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4532 and n constraint instead of aarch64_shift_imm_di and Usd.
4533 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4534 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4536 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4539 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4540 to favor GPR over NEON registers.
4541 (<shift>di3_neon): Likewise.
4543 2018-03-20 Tom de Vries <tom@codesourcery.com>
4546 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4547 (nvptx_process_pars): Emit bar.sync asap and alap.
4549 2018-03-20 Tom de Vries <tom@codesourcery.com>
4552 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4553 seen_label if seen_label is already set.
4555 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4558 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4559 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4560 Use 1U instead of 1. Formatting fixes.
4563 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4564 instead of TREE_TYPE (s1) for the return value.
4566 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4568 PR tree-optimization/84946
4569 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4570 bitsize + bitsize in poly_uint64 rather than poly_int64.
4573 * dwarf2asm.c: Include fold-const.c.
4574 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4575 of decl rather than decl itself.
4577 PR rtl-optimization/84643
4578 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4580 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4583 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4584 calling assemble_variable.
4586 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4589 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4590 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4592 2018-03-19 Jim Wilson <jimw@sifive.com>
4595 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4596 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4597 (riscv_first_stack_step): Likewise.
4598 (riscv_option_override): Use STACK_BOUNDARY instead of
4600 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4602 (BIGGEST_ALIGNMENT): Set to 128.
4603 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4604 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4607 2018-03-19 Richard Biener <rguenther@suse.de>
4609 PR tree-optimization/84933
4610 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4611 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4613 2018-03-19 Richard Biener <rguenther@suse.de>
4615 PR tree-optimization/84859
4616 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4617 (cond_if_else_store_replacement): Perform sinking operation on
4618 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4619 Generalize what a BB with a single eligible store is.
4621 2018-03-19 Richard Biener <rguenther@suse.de>
4623 PR tree-optimization/84929
4624 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4625 chrec_is_positive against non-chrec arg.
4627 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4630 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4632 2018-03-18 Martin Liska <mliska@suse.cz>
4634 PR rtl-optimization/84635
4635 * regrename.c (build_def_use): Use matches_mode only when
4638 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4640 PR tree-optimization/84913
4641 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4642 vectorize chains of COND_EXPRs.
4644 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4646 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4648 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4650 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4652 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4654 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4656 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4657 Kito Cheng <kito.cheng@gmail.com>
4659 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4660 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4661 (nds32_adjust_reg_alloc_order): New function.
4662 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4664 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4666 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4667 nds32_print_operand, nds32_print_operand_address): Use
4668 HOST_WIDE_INT_PRINT_DEC instead.
4670 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4672 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4674 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4677 * config/i386/i386.c (initial_ix86_tune_features,
4678 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4680 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4681 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4682 rather than 1u << ix86_tune. Formatting fix.
4683 (ix86_option_override_internal): Change ix86_arch_mask from
4684 unsigned int to unsigned HOST_WIDE_INT, initialize to
4685 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4686 (ix86_function_specific_restore): Likewise.
4688 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4691 * postreload.c (reload_combine_recognize_pattern): Perform
4692 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4693 truncate_int_for_mode the result for the destination's mode.
4696 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4697 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4699 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4702 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4705 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4706 regno_allocno_class_array and sorted_pseudos.
4707 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4708 insns where regno is used.
4710 2018-03-16 Martin Liska <mliska@suse.cz>
4713 * multiple_target.c (create_dispatcher_calls): Redirect
4714 reference in the symbol table.
4716 2018-03-16 Martin Liska <mliska@suse.cz>
4719 * multiple_target.c (create_dispatcher_calls): Redirect also
4722 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4726 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4727 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4728 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4730 2018-03-16 Julia Koval <julia.koval@intel.com>
4732 * doc/invoke.texi (Skylake Server): Add CLWB.
4733 Cannonlake): Remove CLWB.
4735 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4737 PR tree-optimization/84841
4738 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4740 (FLOAT_ONE_CONST_TYPE): Define.
4741 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4742 (sort_by_operand_rank): Put entries with higher constant_type last
4743 rather than first to match comments.
4745 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4747 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4750 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4753 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4754 check and use address of target always.
4756 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4759 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4760 (indirect_thunk_bnd_needed): Likewise.
4761 (indirect_thunks_used): Likewise.
4762 (indirect_thunks_bnd_used): Likewise.
4763 (indirect_return_needed): New.
4764 (indirect_return_bnd_needed): Likewise.
4765 (output_indirect_thunk_function): Add a bool argument for
4767 (output_indirect_thunk_function): Don't generate alias for
4768 function return thunk.
4769 (ix86_code_end): Call output_indirect_thunk_function to generate
4770 function return thunks.
4771 (ix86_output_function_return): Set indirect_return_bnd_needed
4772 and indirect_return_needed instead of indirect_thunk_bnd_needed
4773 and indirect_thunk_needed.
4775 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4777 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4778 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4779 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4781 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4782 Paul Hua <paul.hua.gm@gmail.com>
4785 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4787 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4789 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4790 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4793 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4796 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4797 instead of GET_MODE_SIZE when comparing Units.
4799 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4802 * varasm.c (hash_section): Return an unchangeble hash value
4803 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4804 Return !aarch64_can_use_per_function_literal_pools_p ().
4806 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4809 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4810 rather than address of cmode as last argument to prepare_cmp_insn.
4812 2018-03-15 Julia Koval <julia.koval@intel.com>
4814 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4815 F_AVX512VNNI, F_AVX512BITALG): New.
4817 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4820 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4821 insn for floating-point loads and stores.
4823 2018-03-14 Carl Love <cel@us.ibm.com>
4825 * config/rs6000/rs6000-c.c: Add macro definitions for
4826 ALTIVEC_BUILTIN_VEC_PERMXOR.
4827 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4828 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4829 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4830 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4832 * config/doc/extend.texi: Add prototypes for vec_permxor.
4834 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4837 * diagnostic-show-locus.c (class layout_point): Convert m_line
4838 from int to linenum_type.
4839 (line_span::comparator): Use linenum "compare" function when
4840 comparing line numbers.
4841 (test_line_span): New function.
4842 (layout_range::contains_point): Convert param "row" from int to
4844 (layout_range::intersects_line_p): Likewise.
4845 (layout::will_show_line_p): Likewise.
4846 (layout::print_source_line): Likewise.
4847 (layout::should_print_annotation_line_p): Likewise.
4848 (layout::print_annotation_line): Likewise.
4849 (layout::print_leading_fixits): Likewise.
4850 (layout::annotation_line_showed_range_p): Likewise.
4851 (struct line_corrections): Likewise for field m_row.
4852 (line_corrections::line_corrections): Likewise for param "row".
4853 (layout::print_trailing_fixits): Likewise.
4854 (layout::get_state_at_point): Likewise.
4855 (layout::get_x_bound_for_row): Likewise.
4856 (layout::print_line): Likewise.
4857 (diagnostic_show_locus): Likewise for locals "last_line" and
4859 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4860 * input.c (selftest::test_linenum_comparisons): New function.
4861 (selftest::input_c_tests): Call it.
4862 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4863 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4864 * selftest.h (ASSERT_GT): New macro.
4865 (ASSERT_GT_AT): New macro.
4866 (ASSERT_LT): New macro.
4867 (ASSERT_LT_AT): New macro.
4869 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4871 PR rtl-optimization/84780
4872 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4874 2018-03-14 Martin Liska <mliska@suse.cz>
4876 * tree.c (record_node_allocation_statistics): Use
4877 get_stats_node_kind.
4878 (get_stats_node_kind): New function extracted from
4879 record_node_allocation_statistics.
4880 (free_node): Use get_stats_node_kind.
4882 2018-03-14 Richard Biener <rguenther@suse.de>
4884 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4885 that the value-set of ANTIC_IN doesn't grow.
4888 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4890 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4891 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4893 2018-03-14 Julia Koval <julia.koval@intel.com>
4895 * config.gcc (icelake-client, icelake-server): New.
4897 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4898 (initial_ix86_arch_features): Ditto.
4899 (PTA_SKYLAKE): Add SGX.
4900 (PTA_ICELAKE): Remove.
4901 (PTA_ICELAKE_CLIENT): New.
4902 (PTA_ICELAKE_SERVER): New.
4903 (ix86_option_override_internal): Split up icelake on icelake client and
4905 (get_builtin_code_for_version): Ditto.
4906 (fold_builtin_cpu): Ditto.
4907 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4908 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4909 * config/i386/i386.h (processor_type): Ditto.
4910 * doc/invoke.texi: Ditto.
4912 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4915 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4916 INTEGER_CST offset, add it together with bitpos / 8 and
4917 sign extend based on POINTER_SIZE.
4921 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4924 * config/i386/constraints.md (Yc): New register constraint.
4925 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4926 Use Yc constraint for alternative 2 of operand 0. Remove
4927 preferred_for_speed attribute.
4929 2018-03-14 Richard Biener <rguenther@suse.de>
4931 PR tree-optimization/84830
4932 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4933 with the old one to avoid oscillations.
4935 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4938 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4940 (assign_by_spills): Return a flag of reload assignment failure.
4941 Do not process the reload assignment failures. Do not spill other
4942 reload pseudos if they has the same reg class. Update n if
4944 (lra_assign): Add a return arg. Set up from the result of
4945 assign_by_spills call.
4946 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4947 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4948 usage_insns if it is not NULL.
4949 (spill_hard_reg_in_range): New function.
4950 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4951 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4952 function prototypes.
4953 (lra_assign): Change prototype.
4954 * lra.c (lra): Add code to deal with fails by splitting hard reg
4957 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4959 * config/riscv/riscv.opt (mrelax): New option.
4960 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4961 "norelax" when riscv_mrelax is disabled.
4962 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4964 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4967 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4968 reassociation for int modes.
4970 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4972 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4973 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4975 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4976 * config/aarch64/aarch64-sve.md
4977 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4978 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4979 (*extend<mode><Vwide>2): Rename to...
4980 (aarch64_sve_extend<mode><Vwide>2): ...this.
4981 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4982 renaming the old pattern to...
4983 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4985 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4986 define_expand, renaming the old pattern to...
4987 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4988 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4989 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4990 account when deciding which SVE instruction the optab should use.
4991 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4993 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4995 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4996 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4997 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4998 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4999 (tlsdesc_small_<mode>): Turn a define_expand and use
5000 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5001 (tlsdesc_small_advsimd_<mode>): ...this.
5002 (tlsdesc_small_sve_<mode>): New pattern.
5004 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5006 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5007 (UNSPEC_UMUL_HIGHPART): New constants.
5008 (MUL_HIGHPART): New int iteraor.
5009 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5010 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5012 (*<su>mul<mode>3_highpart): New define_insn.
5014 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5017 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5020 2018-03-13 Martin Liska <mliska@suse.cz>
5023 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5025 (sem_item_optimizer::~sem_item_optimizer): Release it.
5026 (sem_item_optimizer::merge_classes): Register variable aliases.
5027 (sem_item_optimizer::fixup_pt_set): New function.
5028 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5029 * ipa-icf.h: Declare new variables and functions.
5031 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5034 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5035 integer_pow2p@2 and test integer_pow2p in condition.
5036 (A < 0 ? C : 0): Similarly for @1.
5039 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5040 characters starting at p contain '\0' character, don't look beyond
5044 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5045 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5048 * reg-stack.c (change_stack): Change update_end var from int to
5049 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5050 also call set_block_for_insn on the newly added insns and rescan.
5053 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5054 on the last operand.
5057 * tree.c (stabilize_reference_1): Return save_expr (e) for
5058 STATEMENT_LIST even if it doesn't have side-effects.
5060 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5062 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5064 2018-03-12 Renlin Li <renlin.li@arm.com>
5066 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5067 aarch64_output_scalar_simd_mov_immediate.
5069 2018-03-12 Martin Sebor <msebor@redhat.com>
5071 PR tree-optimization/83456
5072 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5073 for perfectly overlapping calls to memcpy.
5074 (gimple_fold_builtin_memory_chk): Same.
5075 (gimple_fold_builtin_strcpy): Handle no-warning.
5076 (gimple_fold_builtin_stxcpy_chk): Same.
5077 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5079 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5081 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5082 parameter. Use it for SFmode.
5083 (rs6000_function_arg_advance_1): Adjust.
5084 (rs6000_function_arg): Adjust.
5085 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5087 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5089 PR rtl-optimization/84169
5090 PR rtl-optimization/84780
5091 * combine.c (can_combine_p): Check for a 2-insn combination whether
5092 the destination register is used between the two insns, too.
5094 2018-03-12 Richard Biener <rguenther@suse.de>
5096 PR tree-optimization/84803
5097 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5098 for refs DR analysis didn't process.
5100 2018-03-12 Richard Biener <rguenther@suse.de>
5102 PR tree-optimization/84777
5103 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5104 force-vectorize loops ignore whether we are optimizing for size.
5106 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5108 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5109 (TARGET_MD_ASM_ADJUST): Define.
5111 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5112 Kito Cheng <kito.cheng@gmail.com>
5113 Chung-Ju Wu <jasonwucj@gmail.com>
5115 * config/nds32/nds32.c (nds32_compute_stack_frame,
5116 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5117 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5118 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5119 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5120 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5121 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5122 * config/nds32/nds32.md (prologue, epilogue): Use macro
5123 NDS32_V3PUSH_AVAILABLE_P to do checking.
5125 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5128 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5129 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5130 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5131 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5132 addition of most attributes on !orig_type_die or the attribute not
5133 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5135 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5136 Chung-Ju Wu <jasonwucj@gmail.com>
5138 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5140 * config/nds32/nds32.opt (mvh): New option.
5142 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5143 Chung-Ju Wu <jasonwucj@gmail.com>
5145 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5147 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5148 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5151 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5152 Chung-Ju Wu <jasonwucj@gmail.com>
5154 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5156 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5157 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5159 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5160 Kito Cheng <kito.cheng@gmail.com>
5161 Chung-Ju Wu <jasonwucj@gmail.com>
5163 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5164 UNSPEC_FFMISM and UNSPEC_FLMISM.
5165 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5166 for ffb, ffmism and flmism.
5167 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5168 (unspec_ffmism): Ditto.
5169 (unspec_flmism): Ditto.
5170 (nds32_expand_builtin_impl): Check if string extension is available.
5171 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5172 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5174 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5177 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5180 * lra-assigns.c (assign_by_spills): Return a flag of reload
5181 assignment failure. Do not process the reload assignment
5182 failures. Do not spill other reload pseudos if they has the same
5184 (lra_assign): Add a return arg. Set up from the result of
5185 assign_by_spills call.
5186 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5187 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5188 usage_insns if it is not NULL.
5189 (spill_hard_reg_in_range): New function.
5190 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5191 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5192 function prototypes.
5193 (lra_assign): Change prototype.
5194 * lra.c (lra): Add code to deal with fails by splitting hard reg
5197 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5200 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5202 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5205 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5206 (dw_val_node): Add val_symbolic_view.
5207 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5208 (symview_upper_bound): New.
5209 (new_line_info_table): Initialize symviews_since_reset.
5210 (dwarf2out_source_line): Count symviews_since_reset and set
5211 symview_upper_bound.
5212 (dw_val_equal_p): Handle symview.
5213 (add_AT_symview): New.
5214 (print_dw_val): Handle symview.
5215 (attr_checksum, attr_checksum_ordered): Likewise.
5216 (same_dw_val_p, size_of_die): Likewise.
5217 (value_format, output_die): Likewise.
5218 (add_high_low_attributes): Use add_AT_symview for entry_view.
5219 (dwarf2out_finish): Reset symview_upper_bound, clear
5222 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5225 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5226 Add strict argument and use it.
5227 (rs6000_split_multireg_move): Update for new strict argument.
5228 (mem_operand_gpr): Disallow all non-offsettable addresses.
5229 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5231 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5234 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5235 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5236 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5239 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5240 decl, use remap_type if we want to use the type.
5242 2018-03-09 Martin Sebor <msebor@redhat.com>
5244 PR tree-optimization/84526
5245 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5247 (builtin_access::generic_overlap): Be prepared to handle non-array
5250 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5252 PR rtl-optimization/84682
5253 * lra-constraints.c (process_address_1): Check is_address flag
5254 for address constraints.
5255 (process_alt_operands): Likewise.
5256 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5257 preprocess_constraints.
5258 * recog.h (preprocess_constraints): Add oploc parameter.
5260 * recog.c (preprocess_constraints): Test address_operand for
5261 CT_ADDRESS constraints.
5263 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5266 * lra-assigns.c (assign_by_spills): Return a flag of reload
5267 assignment failure. Do not process the reload assignment
5268 failures. Do not spill other reload pseudos if they has the same
5270 (lra_assign): Add a return arg. Set up from the result of
5271 assign_by_spills call.
5272 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5273 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5274 usage_insns if it is not NULL.
5275 (spill_hard_reg_in_range): New function.
5276 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5277 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5278 function prototypes.
5279 (lra_assign): Change prototype.
5280 * lra.c (lra): Add code to deal with fails by splitting hard reg
5283 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5286 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5287 Accept complain bool parameter. Only emit errors if it is true.
5288 (arm_parse_cpu_option_name): Likewise.
5289 (arm_target_thumb_only): Adjust callers of the above.
5290 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5291 prototype to take a default true bool parameter.
5292 (arm_parse_arch_option_name): Likewise.
5294 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5295 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5299 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5300 * configure: Regenerate.
5301 * configure.ac ("linker --version-script option"): New.
5302 ("linker soname option"): New.
5304 2018-03-09 Richard Biener <rguenther@suse.de>
5306 PR tree-optimization/84775
5307 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5308 immediate uses of predicate stmts and mark them modified.
5311 PR tree-optimization/84178
5312 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5314 (version_loop_for_if_conversion): Delay update_ssa call.
5315 (tree_if_conversion): Delay update_ssa until after predicate
5318 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5321 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5322 when the function accesses prior frames.
5324 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5327 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5328 gen_llsym, otherwise call maybe_gen_llsym.
5331 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5332 has ',' character inside of it.
5334 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5337 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5338 as clobbering CC_REGNUM.
5340 2018-03-08 Richard Biener <rguenther@suse.de>
5343 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5344 (follow_copies_to_constant): Do not follow SSA names registered
5347 2018-03-08 Richard Biener <rguenther@suse.de>
5349 PR tree-optimization/84178
5350 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5352 (version_loop_for_if_conversion): Delay update_ssa call.
5353 (tree_if_conversion): Delay update_ssa until after predicate
5356 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5358 PR tree-optimization/84178
5359 * tree-if-conv.c (release_bb_predicate): Remove the
5360 the assertion that the stmts have NULL use_ops.
5361 Discard the statements, asserting that they haven't
5362 yet been added to a BB.
5364 2018-03-08 Richard Biener <rguenther@suse.de>
5366 PR tree-optimization/84746
5367 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5368 (phi_translate): Pass in destination ANTIC_OUT set.
5369 (phi_translate_1): Likewise. For a simplified result lookup
5370 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5371 (phi_translate_set): Adjust.
5372 (do_pre_regular_insertion): Likewise.
5373 (do_pre_partial_partial_insertion): Likewise.
5375 2018-03-08 Martin Liska <mliska@suse.cz>
5377 PR gcov-profile/84735
5378 * doc/gcov.texi: Document usage of profile files.
5379 * gcov-io.h: Document changes in the format.
5381 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5385 * dwarf2out.c (struct dw_line_info_table): Update comments for
5387 (FORCE_RESET_NEXT_VIEW): New.
5388 (FORCE_RESETTING_VIEW_P): New.
5389 (RESETTING_VIEW_P): Check for -1 too.
5390 (ZERO_VIEW_P): Likewise.
5391 (new_line_info_table): Force-reset next view.
5392 (dwarf2out_begin_function): Likewise.
5393 (dwarf2out_source_line): Simplify zero_view_p initialization.
5394 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5395 view directly. Omit view when omitting .loc at line 0.
5397 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5399 PR tree-optimization/84740
5400 * tree-switch-conversion.c (process_switch): Call build_constructors
5401 only if info.phi_count is non-zero.
5403 PR tree-optimization/84739
5404 * tree-tailcall.c (find_tail_calls): Check call arguments against
5405 DECL_ARGUMENTS (current_function_decl) rather than
5406 DECL_ARGUMENTS (func) when checking for tail recursion.
5408 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5410 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5411 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5412 Volker Reichelt's entry and add entries for people that perform
5413 GCC fuzzy testing and report numerous bugs.
5415 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5418 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5419 readonly data in sdata, if that is disabled.
5420 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5421 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5422 -mreadonly-in-sdata option.
5424 2018-03-07 Martin Sebor <msebor@redhat.com>
5426 PR tree-optimization/84468
5427 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5428 basic block when looking for nul assignment.
5430 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5433 * except.h (output_function_exception_table): Adjust prototype.
5434 * except.c (output_function_exception_table): Remove FNNAME parameter
5435 and add SECTION parameter. Ouput one part of the table at a time.
5436 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5437 the first part of the exception table and emit unwind directives.
5438 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5439 (i386_pe_seh_cold_init): Likewise.
5440 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5441 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5442 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5443 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5444 * config/i386/winnt.c: Include except.h.
5445 (struct seh_frame_state): Add reg_offset, after_prologue and
5446 in_cold_section fields.
5447 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5448 (i386_pe_seh_cold_init): New function.
5449 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5450 to seh->in_cold_section.
5451 (seh_emit_push): Record the offset of the push.
5452 (seh_emit_save): Record the offet of the save.
5453 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5454 Test seh->after_prologue to disregard the epilogue.
5455 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5456 (i386_pe_end_cold_function): New function.
5458 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5461 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5462 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5465 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5467 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5468 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5471 * multiple_target.c: Include tree-inline.h and intl.h.
5472 (expand_target_clones): Diagnose and fail if node->definition and
5473 !tree_versionable_function_p (node->decl).
5475 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5477 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5479 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5480 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5481 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5483 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5486 * combine.c (try_combine): Use reg_or_subregno instead of handling
5487 just paradoxical SUBREGs and REGs.
5489 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5491 * config/arc/arc.c (arc_finalize_pic): Remove function.
5492 (arc_must_save_register): We use single base PIC register, remove
5493 checks to save/restore the PIC register.
5494 (arc_expand_prologue): Likewise.
5495 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5497 (arc_verify_short): Likewise.
5498 (arc_attr_type): Likewise.
5499 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5500 (walk_stores): Likewise.
5501 (arc_address_cost): Make it static.
5502 (arc_verify_short): Likewise.
5503 (branch_dest): Likewise.
5504 (arc_attr_type): Likewise.
5505 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5506 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5507 (arc_final_prescan_insn): Remove inserting the nops due to
5508 hardware hazards. It is done in reorg step.
5509 (insn_length_variant_t): Remove.
5510 (insn_length_parameters_t): Likewise.
5511 (arc_insn_length_parameters): Likewise.
5512 (arc_get_insn_variants): Likewise.
5513 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5515 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5518 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5521 PR tree-optimization/84687
5522 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5524 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5526 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5528 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5529 Rename to ppc_speculation_barrier.
5530 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5531 __builtin_ppc_speculation_barrier.
5533 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5536 * combine.c (combine_simplify_rtx): Don't try to simplify if
5537 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5540 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5542 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5543 to 32 bytes when compiling for POWER9.
5545 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5548 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5549 regparm >= 3 with no arg reg available also for calls with
5550 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5553 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5555 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5557 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5560 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5562 2018-03-05 Richard Biener <rguenther@suse.de>
5564 PR tree-optimization/84486
5565 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5566 When inserting a __builtin_assume_aligned call set the LHS
5567 SSA name alignment info accordingly.
5569 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5571 PR tree-optimization/84114
5572 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5573 Avoid reassociation of FLOAT_MODE addition.
5575 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5577 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5578 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5579 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5580 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5581 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5582 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5583 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5585 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5586 __builtin_ia32_wbinvd): New builtins.
5587 (SPECIAL_ARGS2): New.
5588 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5589 (SPECIAL_ARGS2): New.
5590 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5591 (ix86_valid_target_attribute_inner_p): Ditto.
5592 (ix86_init_mmx_sse_builtins): Add special_args2.
5593 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5594 TARGET_WBNOINVD_P): New.
5595 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5596 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5597 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5598 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5599 * config/i386/pconfigintrin.h: New file.
5600 * config/i386/wbnoinvdintrin.h: Ditto.
5601 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5602 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5604 2018-03-05 Richard Biener <rguenther@suse.de>
5606 PR tree-optimization/84670
5607 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5609 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5610 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5611 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5612 doesn't grow if all successors (recursively) were visited at least
5615 2018-03-05 Richard Biener <rguenther@suse.de>
5617 PR tree-optimization/84650
5618 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5619 if executed in the loop pipeline.
5621 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5623 * doc/configfiles.texi (Configuration Files): Move info about
5624 conditionalizing $target-protos.h to...
5625 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5626 differs from $target-protos.h.
5628 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5629 Chung-Ju Wu <jasonwucj@gmail.com>
5631 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5632 * config/nds32/nds32-multiple.md (setmemsi): Define.
5633 * config/nds32/nds32-memory-manipulation.c
5634 (nds32_gen_dup_4_byte_to_word_value): New.
5635 (emit_setmem_word_loop): New.
5636 (emit_setmem_byte_loop): New.
5637 (nds32_expand_setmem_loop): New.
5638 (nds32_expand_setmem_loop_v3m): New.
5639 (nds32_expand_setmem_unroll): New.
5640 (nds32_expand_setmem): New.
5642 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5643 Chung-Ju Wu <jasonwucj@gmail.com>
5645 * config/nds32/nds32-memory-manipulation.c
5646 (nds32_emit_load_store): New.
5647 (nds32_emit_post_inc_load_store): New.
5648 (nds32_emit_mem_move): New.
5649 (nds32_emit_mem_move_block): New.
5650 (nds32_expand_movmemsi_loop_unknown_size): New.
5651 (nds32_expand_movmemsi_loop_known_size): New.
5652 (nds32_expand_movmemsi_loop): New.
5653 (nds32_expand_movmemsi_unroll): New.
5654 (nds32_expand_movmemqi): Rename ...
5655 (nds32_expand_movmemsi): ... to this.
5656 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5657 (movmemsi): ... to this.
5658 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5659 (nds32_expand_movmemsi): ... to this.
5661 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5662 Monk Chiang <sh.chiang04@gmail.com>
5663 Chung-Ju Wu <jasonwucj@gmail.com>
5665 * config/nds32/nds32-protos.h
5666 (nds32_expand_load_multiple): New arguments.
5667 (nds32_expand_store_multiple): Ditto.
5668 (nds32_valid_multiple_load_store): Rename ...
5669 (nds32_valid_multiple_load_store_p): ... to this.
5670 * config/nds32/nds32-memory-manipulation.c
5671 (nds32_expand_load_multiple): Refine implementation.
5672 (nds32_expand_store_multiple): Ditto.
5673 * config/nds32/nds32-multiple.md
5674 (load_multiple): Update nds32_expand_load_multiple interface.
5675 (store_multiple): Update nds32_expand_store_multiple interface.
5676 * config/nds32/nds32-predicates.c
5677 (nds32_valid_multiple_load_store): Rename ...
5678 (nds32_valid_multiple_load_store_p): ... to this and refine
5680 * config/nds32/predicates.md
5681 (nds32_load_multiple_and_update_address_operation): New predicate.
5682 (nds32_store_multiple_and_update_address_operation): New predicate.
5684 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5685 Chung-Ju Wu <jasonwucj@gmail.com>
5687 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5688 (combo): New attribute.
5689 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5691 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5693 * config/nds32/nds32.opt: Change -mcmodel= default value.
5695 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5696 Monk Chiang <sh.chiang04@gmail.com>
5697 Chung-Ju Wu <jasonwucj@gmail.com>
5699 * config/nds32/constants.md (unspec_element): New enum.
5700 * config/nds32/constraints.md (Umw): New constraint.
5701 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5702 * config/nds32/nds32-intrinsic.md: Likewise.
5703 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5704 (nds32_valid_smw_lwm_base_p): New.
5705 (nds32_output_smw_single_word): New.
5706 (nds32_output_lmw_single_word): New.
5707 (nds32_expand_unaligned_load): New.
5708 (nds32_expand_unaligned_store): New.
5709 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5710 (nds32_output_smw_single_word): Declare.
5711 (nds32_output_lmw_single_word): Declare.
5712 (nds32_expand_unaligned_load): Declare.
5713 (nds32_expand_unaligned_store): Declare.
5714 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5715 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5716 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5717 NDS32_BUILTIN_UASTORE_DW.
5718 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5721 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5722 Kito Cheng <kito.cheng@gmail.com>
5723 Chung-Ju Wu <jasonwucj@gmail.com>
5725 * config/nds32/nds32-intrinsic.c
5726 (nds32_expand_builtin_null_ftype_reg): Delete.
5727 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5728 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5729 (nds32_read_argument): New.
5730 (nds32_legitimize_target): Ditto.
5731 (nds32_legitimize_argument): Ditto.
5732 (nds32_check_constant_argument): Ditto.
5733 (nds32_expand_unop_builtin): Ditto.
5734 (nds32_expand_unopimm_builtin): Ditto.
5735 (nds32_expand_binop_builtin): Ditto.
5736 (nds32_builtin_decl_impl): Ditto.
5737 (builtin_description): Ditto.
5738 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5739 (nds32_init_builtins_impl): Ditto.
5740 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5741 (nds32_builtin_decl): New.
5742 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5743 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5745 2018-03-02 Jeff Law <law@redhat.com>
5747 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5748 (redundant_insn, fill_simple_delay_slots): Likewise.
5749 (fill_slots_from_thread): Likewise.
5750 * resource.c (mark_referenced_resources): Likewise.
5751 (mark_set_resources, find_dead_or_set_registers): Likewise.
5753 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5755 * substring-locations.h (format_warning_va): Formatting fix for
5757 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5759 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5760 * substring-locations.c: Include intl.h.
5761 (format_warning_va): Turned into small wrapper around
5762 format_warning_n_va, renamed to ...
5763 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5764 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5766 (format_warning_at_substring_n): New function.
5767 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5768 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5769 format_warning_at_substring with just a shorter name instead of
5770 const function pointer.
5771 (fmtwarn_n): New function.
5772 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5773 appropriate, get rid of all the fmtstr temporaries, move conditionals
5774 with G_() wrapped string literals directly into fmtwarn arguments,
5775 cast dir.len to (int), formatting fixes.
5777 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5779 * doc/invoke.texi: Remove "Cilk Plus" references.
5781 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5782 Richard Biener <rguenther@suse.de>
5785 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5786 for error or warning attributes if CALL_FROM_THUNK_P is set.
5789 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5792 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5793 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5796 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5797 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5799 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5800 instead of -1U in last predictors element's probability member.
5802 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5805 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5806 arguments if they are comparable.
5808 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5810 PR tree-optimization/84634
5811 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5812 masks and masked_loop_p with a single loop_masks, making sure it's
5813 null for bb vectorization.
5815 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5817 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5818 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5819 loop->force_vectorize to check whether there is no alias.
5821 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5824 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5826 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5828 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5829 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5830 instead of a loop around prev_real_insn.
5831 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5835 * config/i386/i386.c (ix86_print_operand): Use conditional
5836 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5839 2018-03-02 Richard Biener <rguenther@suse.de>
5841 PR tree-optimization/84427
5842 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5843 (bitmap_set_subtract_values): Rewrite to handle multiple
5846 (prune_clobbered_mems): Likewise.
5847 (phi_translate): Take edge instead of pred/phiblock.
5848 (phi_translate_1): Likewise.
5849 (phi_translate_set): Likewise. Insert all translated
5850 exprs for a value into the set, keeping possibly multiple
5851 expressions per value.
5852 (compute_antic_aux): Adjust for phi_translate changes.
5853 When intersecting union the expressions and prune those
5854 not in the final value set, keeping possibly multiple
5855 expressions per value. Do not use value-insertion
5856 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5857 all expressions. Add verification that the value-sets
5858 only shrink during iteration.
5859 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5860 (do_pre_regular_insertion): Likewise.
5861 (do_pre_partial_partial_insertion): Likewise.
5863 2018-03-02 Richard Biener <rguenther@suse.de>
5866 * config/darwin.c (saved_debug_info_level): New static global.
5867 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5868 (darwin_asm_lto_end): Restore debug info generation settings.
5870 2018-03-01 Martin Liska <mliska@suse.cz>
5873 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5876 2018-03-01 Richard Biener <rguenther@suse.de>
5879 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5880 type annotation with LTO when debug was disabled at compile-time.
5882 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5884 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5886 (mips_final_postscan_insn): Likewise.
5888 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5890 PR rtl-optimization/84528
5891 * alias.c (init_alias_target): Add commentary.
5892 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5893 a unique base value if the frame pointer is not eliminated
5894 to the stack pointer.
5896 2018-03-01 Tom de Vries <tom@codesourcery.com>
5898 PR rtl-optimization/83327
5899 * lra-int.h (hard_regs_spilled_into): Declare.
5900 * lra.c (hard_regs_spilled_into): Define.
5901 (init_reg_info): Init hard_regs_spilled_into.
5902 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5903 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5904 (process_bb_lives): Handle hard_regs_spilled_into.
5905 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5906 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5908 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5910 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5911 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5912 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5913 * config/rs6000/aix72.h: New file.
5915 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5917 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5918 instead of warning_at with conditional singular and plural messages
5922 * stor-layout.c (update_alignment_for_field): For
5923 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5924 && !DECL_PACKED (field), do the alignment update, just use
5925 only desired_align instead of MAX (type_align, desired_align)
5927 (place_field): Don't do known_align < desired_align handling
5928 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5929 is non-NULL, instead do it after rli->prev_field handling and
5930 only if not within a bitfield word. For DECL_PACKED (field)
5931 use type_align of BITS_PER_UNIT.
5933 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5935 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5936 superfluous parentheses and trailing spaces.
5938 2018-02-28 Richard Biener <rguenther@suse.de>
5940 PR tree-optimization/84584
5941 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5942 SCoPs with fake exit edge.
5944 2018-02-28 Martin Liska <mliska@suse.cz>
5947 * timevar.c (timer::print): Fix format to properly print 100%
5950 2018-02-28 Richard Biener <rguenther@suse.de>
5953 * genmatch.c (capture_info::walk_match): Do not mark
5954 captured expressions without operands as expr_p given
5955 they act more like predicates and should be subject to
5956 "lost tail" side-effect preserving.
5958 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5960 PR rtl-optimization/81611
5961 * auto-inc-dec.c (attempt_change): Move dead note from
5962 mem_insn if it's the next use of regno
5963 (find_address): Take address use of reg holding
5964 non-incremented value. Add parm to limit search to the named
5966 (merge_in_block): Attempt to use a mem insn that is the next
5967 use of the original regno.
5969 2018-02-27 Martin Sebor <msebor@redhat.com>
5972 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5973 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5975 2018-02-27 Martin Sebor <msebor@redhat.com>
5977 PR translation/84207
5978 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5979 n argument to unsigned HOST_WIDE_INT.
5980 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5981 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5982 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5983 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5985 2018-02-27 Richard Biener <rguenther@suse.de>
5987 PR tree-optimization/84512
5988 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5989 Do not use the estimate returned from record_stmt_cost for
5990 the scalar iteration cost but sum properly using add_stmt_cost.
5992 2018-02-27 Richard Biener <rguenther@suse.de>
5994 PR tree-optimization/84466
5995 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5996 Adjust last change to less strictly validate use operands.
5998 2018-02-27 Martin Liska <mliska@suse.cz>
6000 PR gcov-profile/84548
6001 * gcov.c (process_file): Allow partial overlap and consider it
6002 also as group functions.
6003 (output_lines): Properly calculate range of lines for a group.
6005 2018-02-27 Martin Liska <mliska@suse.cz>
6007 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6008 'ggc' suffixes. Change first column width.
6009 (timer::print): Fix formatting of the column.
6011 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6013 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6014 preserve inline entry blocks for the sake of debug inline
6015 entry point markers alone.
6016 (remove_unused_locals): Suggest in comments a better place to
6017 force the preservation of inline entry blocks that are
6018 otherwise unused, but do not preserve them.
6020 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6022 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6024 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6027 * config/i386/constraints.md (Bs): Replace
6028 ix86_indirect_branch_register with
6029 TARGET_INDIRECT_BRANCH_REGISTER.
6031 * config/i386/i386.md (indirect_jump): Likewise.
6032 (tablejump): Likewise.
6033 (*sibcall_memory): Likewise.
6034 (*sibcall_value_memory): Likewise.
6035 Peepholes of indirect call and jump via memory: Likewise.
6036 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6037 (*sibcall_value_GOT_32): Likewise.
6038 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6039 (GOT_memory_operand): Likewise.
6040 (call_insn_operand): Likewise.
6041 (sibcall_insn_operand): Likewise.
6042 (GOT32_symbol_operand): Likewise.
6043 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6045 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6047 PR rtl-optimization/83496
6048 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6049 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6050 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6051 redundant insn, if any.
6052 (relax_delay_slots): Likewise.
6053 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6055 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6057 PR tree-optimization/83965
6058 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6059 that grouped statements are part of a reduction chain. Return
6060 true if the statement is not marked as a reduction itself but
6062 (vect_recog_dot_prod_pattern): Don't check whether the statement
6063 is part of a group here.
6064 (vect_recog_sad_pattern): Likewise.
6065 (vect_recog_widen_sum_pattern): Likewise.
6067 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6070 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6072 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6075 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6077 (ix86_output_indirect_function_return): New prototype.
6078 (ix86_split_simple_return_pop_internal): Likewise.
6079 * config/i386/i386.c (indirect_return_via_cx): New.
6080 (indirect_return_via_cx_bnd): Likewise.
6081 (indirect_thunk_name): Handle return va CX_REG.
6082 (output_indirect_thunk_function): Create alias for
6083 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6084 (ix86_output_indirect_jmp): Remove the bool argument.
6085 (ix86_output_indirect_function_return): New function.
6086 (ix86_split_simple_return_pop_internal): Likewise.
6087 * config/i386/i386.md (*indirect_jump): Don't pass false
6088 to ix86_output_indirect_jmp.
6089 (*tablejump_1): Likewise.
6090 (simple_return_pop_internal): Change it to define_insn_and_split.
6091 Call ix86_split_simple_return_pop_internal to split it for
6093 (simple_return_indirect_internal): Call
6094 ix86_output_indirect_function_return instead of
6095 ix86_output_indirect_jmp.
6097 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6100 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6101 memset and value initialization afterwards.
6103 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6105 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6107 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6110 * common/config/aarch64/aarch64-common.c
6111 (aarch_option_optimization_table[]): Switch
6112 off fomit-frame-pointer
6114 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6115 Chung-Ju Wu <jasonwucj@gmail.com>
6117 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6119 (store_multiple): Ditto.
6121 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6123 * config.gcc: Add --with-cpu support for nds32 target.
6124 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6125 * config/nds32/nds32.opt: Add -mcpu= option.
6127 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6129 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6130 isel=yes): Warn for these deprecated options.
6132 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6134 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6135 ISA_2_5_MASKS_EMBEDDED.
6137 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6139 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6140 p->max as pointers rather than using iterative_hash_expr.
6142 2018-02-23 Carl Love <cel@us.ibm.com>
6144 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6145 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6147 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6148 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6149 P8V_BUILTIN_VEC_VUNSIGNED2.
6151 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6154 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6155 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6157 * lra-constraints.c (curr_insn_transform): Set up
6158 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6160 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6161 * lra-eliminations.c (spill_pseudos): Ditto.
6162 (process_insn_for_elimination): Ditto.
6163 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6164 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6165 LRA_NON_CLOBBERED_ALT.
6167 2018-02-22 Martin Sebor <msebor@redhat.com>
6169 PR tree-optimization/84480
6170 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6171 to maybe_diag_stxncpy_trunc. Call it.
6172 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6173 from gimple_fold_builtin_strcpy. Print inlining stack.
6174 (handle_builtin_stxncpy): Print inlining stack.
6175 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6177 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6180 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6181 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6182 and -fcheck-pointer-bounds are used together.
6183 (indirect_thunk_prefix): New enum.
6184 (indirect_thunk_need_prefix): New function.
6185 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6186 "_nt" instead of "_bnd" for NOTRACK prefix.
6187 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6188 (output_indirect_thunk_function): Likewise.
6190 (ix86_code_end): Update output_indirect_thunk_function calls.
6191 (ix86_output_indirect_branch_via_reg): Replace
6192 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6193 (ix86_output_indirect_branch_via_push): Likewise.
6194 (ix86_output_function_return): Likewise.
6195 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6196 incompatible with -fcf-protection=branch and
6197 -fcheck-pointer-bounds.
6199 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6202 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6203 Change gcc_assert call to output_operand_lossage.
6205 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6207 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6209 2018-02-22 DJ Delorie <dj@redhat.com>
6210 Sebastian Perta <sebastian.perta@renesas.com>
6211 Oleg Endo <olegendo@gcc.gnu.org>
6213 * config/rx/rx.c (rx_rtx_costs): New function.
6214 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6216 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6218 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6220 2018-02-22 Martin Liska <mliska@suse.cz>
6223 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6224 Add "native" as a possible value.
6226 2018-02-22 Martin Liska <mliska@suse.cz>
6229 * config/i386/i386.c (ix86_option_override_internal):
6230 Add "native" as a possible value for -march and -mtune.
6232 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6235 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6236 to all type variants.
6238 PR tree-optimization/84503
6239 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6240 width as info->bitpos + info->bitsize - start.
6241 (merged_store_group::merge_overlapping): Simplify width computation.
6242 (check_no_overlap): New function.
6243 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6244 start + width and last_order of the group, fail if check_no_overlap
6246 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6247 to group if check_no_overlap fails.
6249 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6251 * config/rs6000/altivec.md: Delete contraint arguments to
6252 define_expand, define_split, and define_peephole2, and in
6253 define_insn_and_split if always unused.
6254 * config/rs6000/darwin.md: Ditto.
6255 * config/rs6000/dfp.md: Ditto.
6256 * config/rs6000/rs6000.md: Ditto.
6257 * config/rs6000/sync.md: Ditto.
6258 * config/rs6000/vector.md: Ditto.
6259 * config/rs6000/vsx.md: Ditto.
6261 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6263 * config/rs6000/altivec.md: Write output control strings as braced
6264 blocks instead of double-quoted strings.
6265 * config/rs6000/darwin.md: Ditto.
6266 * config/rs6000/rs6000.md: Ditto.
6267 * config/rs6000/vector.md: Ditto.
6268 * config/rs6000/vsx.md: Ditto.
6270 2018-02-21 Jason Merrill <jason@redhat.com>
6272 PR c++/84314 - ICE with templates and fastcall attribute.
6273 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6275 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6277 * ipa-cp.c (determine_versionability): Fix comment typos.
6279 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6282 * ipa-cp.c (determine_versionability): Do not version functions caling
6285 2018-02-21 Martin Liska <mliska@suse.cz>
6288 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6289 Add "native" as a possible value.
6290 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6291 the macro when native cpu detection is available.
6293 2018-02-21 Martin Liska <mliska@suse.cz>
6296 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6297 Add "native" as a possible value.
6298 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6299 when native cpu detection is available.
6301 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6302 Martin Sebor <msebor@redhat.com>
6304 PR tree-optimization/84478
6305 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6307 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6308 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6309 support which is conservatively correct, for 2 only stay conservative
6310 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6311 argument to the 2 argument get_range_strlen, adjust 6 arg
6312 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6314 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6315 (gimple_fold_builtin_strlen): Pass true as last argument to
6318 2018-02-20 Martin Sebor <msebor@redhat.com>
6321 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6322 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6323 (builtin_memref::builtin_memref): Factor out parts into
6324 set_base_and_offset and call it.
6326 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6329 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6330 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6331 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6332 search at the associated MODE_INT.
6334 2018-02-20 Jeff Law <law@redhat.com>
6337 PR tree-optimization/81592
6339 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6340 for range data rather than using global data.
6341 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6342 range data rather than using global data.
6343 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6344 pass it to children as needed.
6345 (struct directive::fmtresult): Similarly.
6346 (struct directive::set_width): Similarly.
6347 (struct directive::set_precision): Similarly.
6348 (format_integer, format_directive, parse_directive): Similarly.
6349 (format_none): Accept unnamed vr_values parameter.
6350 (format_percent, format_floating, format_character): Similarly.
6351 (format_string, format_plain): Similarly.
6352 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6353 the EVRP range analyzer for range data rather than using global data.
6354 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6355 gimple-ssa-evrp-analyze.h
6356 (class sprintf_dom_walker): Add after_dom_children member function.
6357 Add evrp_range_analyzer member.
6358 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6359 range analyzer as needed.
6360 (sprintf_dom_walker::after_dom_children): New member function.
6361 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6363 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6364 (evrp_range_analyzer::pop_to_marker): Likewise.
6366 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6368 PR tree-optimization/84419
6369 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6370 with the required type if its current type is compatible but
6373 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6376 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6377 after vectorization.
6379 2018-02-20 Martin Liska <mliska@suse.cz>
6382 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6383 possible values if we don't have a hint.
6385 2018-02-20 Martin Liska <mliska@suse.cz>
6389 * final.c (shorten_branches): Build align_tab array with one
6391 * opts.c (finish_options): Add alignment option limit check.
6392 (MAX_CODE_ALIGN): Likewise.
6393 (MAX_CODE_ALIGN_VALUE): Likewise.
6394 * doc/invoke.texi: Document maximum allowed option value for
6395 all -falign-* options.
6397 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6400 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6401 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6402 * var-tracking.c (emit_note_insn_var_location): Remove all references
6403 to NOTE_INSN_CALL_ARG_LOCATION.
6404 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6405 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6406 Use copy_rtx_if_shared.
6407 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6408 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6409 (dwarf2out_var_location): Remove handling of
6410 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6412 * final.c (final_scan_insn): Remove all references to
6413 NOTE_INSN_CALL_ARG_LOCATION.
6414 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6415 before dumping final insns.
6416 * except.c (emit_note_eh_region_end): Remove all references to
6417 NOTE_INSN_CALL_ARG_LOCATION.
6418 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6419 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6420 * config/arc/arc.c (hwloop_optimize): Likewise.
6421 * config/arm/arm.c (create_fix_barrier): Likewise.
6422 * config/s390/s390.c (s390_chunkify_start): Likewise.
6423 * config/sh/sh.c (find_barrier): Likewise.
6424 * config/i386/i386.c (rest_of_insert_endbranch,
6425 ix86_seh_fixup_eh_fallthru): Likewise.
6426 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6427 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6428 * config/frv/frv.c (frv_function_prologue): Likewise.
6429 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6431 (note_outside_basic_block_p): Remove all references to
6432 NOTE_INSN_CALL_ARG_LOCATION.
6433 * gengtype.c (adjust_field_rtx_def): Likewise.
6434 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6436 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6437 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6440 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6443 PR tree-optimization/84452
6444 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6445 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6448 2018-02-19 Martin Liska <mliska@suse.cz>
6451 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6453 2018-02-19 Martin Liska <mliska@suse.cz>
6454 Richard Sandiford <richard.sandiford@linaro.org>
6456 PR tree-optimization/82491
6457 * gimple-fold.c (get_base_constructor): Make earlier bail out
6460 2018-02-19 Carl Love <cel@us.ibm.com>
6462 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6463 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6465 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6466 P8V_BUILTIN_VEC_NEG.
6468 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6470 * config/rl78/rl78.md (movdf): New define expand.
6472 2018-02-19 Martin Liska <mliska@suse.cz>
6475 * doc/invoke.texi: Fix typo.
6476 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6478 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6480 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6481 handle rs6000_single_float and rs6000_double_float specially for
6484 2018-02-16 Jeff Law <law@redhat.com>
6486 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6489 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6492 * ipa-inline.c (inline_small_functions): Fix a typo.
6494 2018-02-16 Nathan Sidwell <nathan@acm.org>
6496 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6498 2018-02-16 Carl Love <cel@us.ibm.com>
6500 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6501 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6502 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6503 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6504 expansion to P8V_BUILTIN_VEC_FLOAT2.
6506 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6508 PR rtl-optimization/70023
6509 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6510 src_regno into account.
6512 2018-02-16 Carl Love <cel@us.ibm.com>
6514 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6515 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6516 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6517 * config/rs6000/rs6000.c: Remove case statements for
6518 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6519 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6520 and P9V_BUILTIN_VEC_VINSERT4B.
6521 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6522 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6523 * config/rs6000/vsx.md:
6524 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6527 2018-02-16 Carl Love <cel@us.ibm.com>
6529 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6531 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6533 * config/rs6000/rs6000-c.c: Add the definitions for
6534 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6535 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6536 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6537 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6538 definition for insert4b and define insn *insert3b_internal.
6539 * doc/extend.texi: Add documentation for vec_extract4b.
6541 2018-02-16 Nathan Sidwell <nathan@acm.org>
6543 * doc/extend.texi (Backwards Compatibility): Mention friend
6544 injection. Note for-scope is deprecated.
6545 * doc/invoke.texi (-ffriend-injection): Deprecate.
6547 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6549 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6550 that moved to I2, also allow destinations that are a paradoxical
6551 subreg (instead of a normal reg).
6553 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6556 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6559 2018-02-16 Richard Biener <rguenther@suse.de>
6561 PR tree-optimization/84037
6562 PR tree-optimization/84016
6564 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6565 Adjust vec_construct for the fact we need additional higher latency
6566 128bit inserts for AVX256 and AVX512 vector builds.
6567 (ix86_add_stmt_cost): Scale vector construction cost for
6570 2018-02-16 Richard Biener <rguenther@suse.de>
6572 PR tree-optimization/84417
6573 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6574 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6575 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6577 2018-02-16 Martin Liska <mliska@suse.cz>
6580 * internal-fn.def (ASAN_CHECK): Set proper flags.
6581 (ASAN_MARK): Likewise.
6583 2018-02-16 Julia Koval <julia.koval@intel.com>
6585 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6586 from PTA_CANNONLAKE.
6588 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6591 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6592 Use ++iter rather than iter++ for std::list iterators.
6593 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6594 defer deleting them until all nodes in the forest are processed. Do
6595 free even leaf nodes. Change to_process into auto_vec.
6598 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6599 * vec.h (vec_default_construct): Use memset instead of placement new
6600 if BROKEN_VALUE_INITIALIZATION is defined.
6601 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6602 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6605 PR rtl-optimization/83723
6606 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6607 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6608 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6610 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6612 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6614 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6616 PR rtl-optimization/81443
6617 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6618 from inner REGs to paradoxical SUBREGs.
6620 2018-02-16 Richard Biener <rguenther@suse.de>
6622 PR tree-optimization/84399
6623 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6624 For operands we can analyze at their definition make sure we can
6625 analyze them at each use as well.
6627 2018-02-16 Richard Biener <rguenther@suse.de>
6629 PR tree-optimization/84190
6630 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6631 volatile accesses if the decl isn't volatile.
6633 2018-02-15 Jason Merrill <jason@redhat.com>
6635 PR c++/84314 - ICE with templates and fastcall attribute.
6636 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6637 TYPE_CANONICAL on an existing type.
6639 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6641 PR tree-optimization/84383
6642 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6643 dstoff nor call operand_equal_p if dstbase is NULL.
6645 PR tree-optimization/84334
6646 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6647 also a CONSTANT_CLASS_P, punt.
6649 2018-02-14 Jim Wilson <jimw@sifive.com>
6651 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6652 first SMALL_OPERAND check. New local min_second_step. Move assert
6653 to where locals are set. Add TARGET_RVC support.
6654 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6656 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6658 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6660 2018-02-14 Martin Sebor <msebor@redhat.com>
6662 PR tree-optimization/83698
6663 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6664 arrays constrain the offset range to their bounds.
6665 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6666 (builtin_access::overlap): Avoid setting the size of overlap if it's
6668 (maybe_diag_overlap): Also consider arrays when deciding what values
6669 of offsets to include in diagnostics.
6671 2018-02-14 Martin Sebor <msebor@redhat.com>
6674 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6675 that correspond to the kind of a declaration.
6677 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6680 * config/pa/pa.md: Load address of PIC label using the linkage table
6681 if the label is nonlocal.
6683 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6685 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6686 warning message if user requests -maltivec=be.
6687 * doc/invoke.texi: Document deprecation of -maltivec=be.
6689 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6692 * config/rs6000/rs6000-c.c: Update definitions for
6693 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6694 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6696 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6699 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6700 add _get_ssp intrinsics. Remove argument from
6701 __builtin_ia32_rdssp[d|q].
6702 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6703 * config/i386/i386-builtin.def: Remove argument from
6704 __builtin_ia32_rdssp[d|q].
6705 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6706 ix86_expand_special_args_builtin for _rdssp[d|q].
6707 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6708 Clear register before usage.
6709 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6710 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6712 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6714 PR tree-optimization/84357
6715 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6716 operand 1 of an ARRAY_REF too.
6718 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6721 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6722 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6724 (set_of_reg): New struct.
6725 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6726 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6727 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6729 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6730 Split into bitclr, bitset, bitinvert patterns if appropriate.
6731 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6732 use rx_fuse_in_memory_bitop.
6733 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6734 to named insn, correct maximum insn length.
6736 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6739 * machmode.def: Define a complex mode for PARTIAL_INT.
6740 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6742 * doc/rtl.texi: Document CSPImode.
6743 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6745 (msp430_hard_regno_nregs_with_padding): Likewise.
6747 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6750 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6752 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6754 PR rtl-optimization/84169
6755 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6756 we generated a parallel as new i3 and we split that to new i2 and i3
6757 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6758 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6759 those to i2, not i1. Partially rewrite this scan code.
6761 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6764 * stor-layout.c (place_field): For variable length fields, adjust
6765 offset_align afterwards not just based on the field's alignment,
6766 but also on the size.
6769 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6770 of exps and logs in the use_exp2 case.
6772 2018-02-13 Jeff Law <law@redhat.com>
6774 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6777 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6780 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6784 * common.opt (gas-loc-support, gas-locview-support): New.
6785 (ginline-points, ginternal-reset-location-views): New.
6786 * doc/invoke.texi: Document them. Use @itemx where intended.
6787 (gvariable-location-views): Adjust.
6788 * target.def (reset_location_view): New.
6789 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6790 (TARGET_RESET_LOCATION_VIEW): New.
6791 * doc/tm.texi: Rebuilt.
6792 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6793 (dwarf2out_default_as_locview_support): New.
6794 (output_asm_line_debug_info): Use option variables.
6795 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6796 (output_loc_list): Likewise.
6797 (add_high_low_attributes): Check option variables.
6798 Don't output entry view attribute in strict mode.
6799 (gen_inlined_subroutine_die): Check option variables.
6800 (dwarf2out_inline_entry): Likewise.
6801 (init_sections_and_labels): Likewise.
6802 (dwarf2out_early_finish): Likewise.
6803 (maybe_reset_location_view): New, from...
6804 (dwarf2out_var_location): ... here. Call it.
6805 * debug.h (dwarf2out_default_as_loc_support): Declare.
6806 (dwarf2out_default_as_locview_support): Declare.
6807 * hooks.c (hook_int_rtx_insn_0): New.
6808 * hooks.h (hook_int_rtx_insn_0): Declare.
6809 * toplev.c (process_options): Take -gas-loc-support and
6810 -gas-locview-support from dwarf2out. Enable
6811 -gvariable-location-views by default only with locview
6812 assembler support. Enable -ginternal-reset-location-views by
6813 default only if the target defines the corresponding hook.
6814 Enable -ginline-points by default if location views are
6815 enabled; force it disabled if statement frontiers are
6817 * tree-inline.c (expand_call_inline): Check option variables.
6818 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6820 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6822 PR tree-optimization/84321
6823 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6824 handling. Also check whether the anti-range contains any values
6825 that satisfy the mask; switch to a VR_RANGE if not.
6827 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6830 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6832 2018-02-13 Martin Jambor <mjambor@suse.cz>
6835 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6836 of call statements, also set location of a load to a temporary.
6838 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6840 * config/rl78/rl78.c (add_vector_labels): New function.
6841 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6842 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6843 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6844 which checks that no arguments are passed.
6845 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6846 * doc/extend.texi: Documentation for the new attribute.
6848 2018-02-13 Andreas Schwab <schwab@suse.de>
6850 * config/riscv/linux.h (CPP_SPEC): Define.
6852 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6855 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6856 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6857 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6858 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6859 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6860 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6861 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6862 temporarily for AES and PCLMUL builtins.
6864 PR tree-optimization/84339
6865 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6866 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6870 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6871 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6872 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6874 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6876 * omp-simd-clone.h: New file.
6877 * omp-simd-clone.c: Include omp-simd-clone.h.
6878 (expand_simd_clones): No longer static.
6879 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6880 cgraph.h and omp-simd-clone.h.
6881 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6882 (vect_recog_widen_shift_pattern): Formatting fix.
6883 (vect_pattern_recog_1): Don't check optab for calls.
6886 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6887 operands[2] into a REG before using gen_lowpart on it.
6889 2018-02-12 Jeff Law <law@redhat.com>
6892 * config/sh/sh.c (find_barrier): Consider a sibling call
6895 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6896 successfully back substituting a reg.
6898 2018-02-12 Richard Biener <rguenther@suse.de>
6900 PR tree-optimization/84037
6901 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6902 parameter, move visited init to caller.
6903 (vect_slp_analyze_operations): Separate cost from validity
6904 check, initialize visited once for all instances.
6905 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6907 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6909 (vect_model_promotion_demotion_cost): Likewise.
6910 (vectorizable_bswap): Guard cost modeling with !slp_node
6911 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6913 (vectorizable_call): Likewise.
6914 (vectorizable_conversion): Likewise.
6915 (vectorizable_assignment): Likewise.
6916 (vectorizable_shift): Likewise.
6917 (vectorizable_operation): Likewise.
6918 (vectorizable_store): Likewise.
6919 (vectorizable_load): Likewise.
6920 (vectorizable_condition): Likewise.
6921 (vectorizable_comparison): Likewise.
6923 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6926 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6927 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6928 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6930 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6933 * params.def (inline-min-speedup): Increase from 8 to 15.
6934 (max-inline-insns-auto): Decrease from 40 to 30.
6935 * ipa-split.c (consider_split): Add some buffer for function to
6936 be considered inlining candidate.
6937 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6940 2018-02-12 Richard Biener <rguenther@suse.de>
6942 PR tree-optimization/84037
6943 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6944 matched stmts if we cannot swap the non-matched ones.
6946 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6948 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6949 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6950 _mm_maskz_scalef_round_ss): New intrinsics.
6951 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6952 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6953 __builtin_ia32_scalefss_round): Remove.
6954 (__builtin_ia32_scalefsd_mask_round,
6955 __builtin_ia32_scalefss_mask_round): New intrinsics.
6956 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6957 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6958 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6959 "<round_constraint>")): Changed to ...
6960 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6961 "<round_scalar_constraint>")): ... this.
6962 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6963 %0, %1, %2<round_op3>}"): Changed to ...
6964 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6965 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6966 %2<round_scalar_mask_op3>}"): ... this.
6967 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6969 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6971 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6972 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6973 (_mm_maskz_sqrt_round_ss): New intrinsics.
6974 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6975 (__builtin_ia32_sqrtsd_mask_round)
6976 (__builtin_ia32_sqrtss_mask_round): New builtins.
6977 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6978 (__builtin_ia32_sqrtss_round): Remove.
6979 (__builtin_ia32_sqrtsd_mask_round)
6980 (__builtin_ia32_sqrtss_mask_round): New builtins.
6981 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6982 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6983 ((match_operand:VF_128 1 "vector_operand"
6984 "xBm,<round_constraint>")): Changed to ...
6985 ((match_operand:VF_128 1 "vector_operand"
6986 "xBm,<round_scalar_constraint>")): ... this.
6987 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6988 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6989 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6990 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6991 %<iptr>1<round_scalar_mask_op3>}): ... this.
6992 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6993 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6995 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6998 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6999 Cast vec_cmpeq result to correct type.
7000 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7001 Cast vec_cmpgt result to correct type.
7003 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7005 * final.c (final_scan_insn_1): Renamed from...
7006 (final_scan_insn): ... this. New wrapper, to recover
7007 seen from the outermost call in recursive ones.
7008 * config/sparc/sparc.c (output_return): Drop seen from call.
7009 (output_sibcall): Likewise.
7010 * config/visium/visium.c (output_branch): Likewise.
7012 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7014 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7017 2018-02-10 Alan Modra <amodra@gmail.com>
7020 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7021 Specify LR as an input.
7023 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7026 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7027 remove_member_access_dummy_vars): New functions.
7028 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7029 lower_omp_1, execute_lower_omp): Use them.
7031 PR rtl-optimization/84308
7032 * shrink-wrap.c (spread_components): Release todo vector.
7034 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7036 PR rtl-optimization/57193
7037 * ira-color.c (struct allocno_color_data): Add member
7038 conflict_allocno_hard_prefs.
7039 (update_conflict_allocno_hard_prefs): New.
7040 (bucket_allocno_compare_func): Add a preference based on
7041 conflict_allocno_hard_prefs.
7042 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7043 (color_allocnos): Remove a dead code. Initiate
7044 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7046 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7049 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7050 constraint from =wa to wa. Avoid a subreg on the output operand,
7051 instead use a pseudo and subreg it in a move.
7052 (p9_xxbrd_<mode>): Changed to ...
7053 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7054 (p9_xxbrd_v2df): New expander.
7055 (p9_xxbrw_<mode>): Changed to ...
7056 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7057 (p9_xxbrw_v4sf): New expander.
7059 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7061 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7063 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7066 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7067 multiply in 32-bit mode.
7068 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7069 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7072 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7074 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7075 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7076 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7077 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7079 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7082 * dwarf2out.c (is_trivial_indirect_ref): New function.
7083 (dwarf2out_late_global_decl): Do not generate a location
7084 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7085 and that are not defined in the current unit.
7087 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7089 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7090 instead of a libcall for UNORDERED.
7092 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7095 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7096 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7098 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7101 * config/s390/s390.c (s390_set_current_function): Invoke
7102 s390_indirect_branch_settings also if fndecl didn't change.
7104 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7106 * config/rs6000/rs6000.md (blockage): Set length to zero.
7108 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7110 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7112 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7115 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7116 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7120 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7121 PARALLEL incoming that failed vt_get_decl_and_offset check.
7124 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7125 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7127 (get_variable_section): For decls in named .bss* sections pass true as
7128 second argument to bss_initializer_p.
7130 2018-02-09 Marek Polacek <polacek@redhat.com>
7131 Jakub Jelinek <jakub@redhat.com>
7134 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7135 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7136 Sync some changes from cxx_fold_indirect_ref.
7138 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7140 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7142 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7143 (BLOCK_INLINE_ENTRY_LABEL): New.
7144 (dwarf2out_var_location): Disregard inline entry markers.
7145 (inline_entry_data): New struct.
7146 (inline_entry_data_hasher): New hashtable type.
7147 (inline_entry_data_hasher::hash): New.
7148 (inline_entry_data_hasher::equal): New.
7149 (inline_entry_data_table): New variable.
7150 (add_high_low_attributes): Add DW_AT_entry_pc and
7151 DW_AT_GNU_entry_view attributes if a pending entry is found
7152 in inline_entry_data_table. Add old entry_pc attribute only
7153 if debug nonbinding markers are disabled.
7154 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7155 markers are enabled.
7156 (block_within_block_p, dwarf2out_inline_entry): New.
7157 (dwarf2out_finish): Check that no entries remained in
7158 inline_entry_data_table.
7159 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7160 (final_scan_insn, notice_source_line): Likewise.
7161 (rest_of_clean_state): Skip inline entry markers.
7162 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7164 * gimple.c (gimple_build_debug_inline_entry): New.
7165 * gimple.h (enum gimple_debug_subcode): Add
7166 GIMPLE_DEBUG_INLINE_ENTRY.
7167 (gimple_build_debug_inline_entry): Declare.
7168 (gimple_debug_inline_entry_p): New.
7169 (gimple_debug_nonbind_marker_p): Adjust.
7170 * insn-notes.def (INLINE_ENTRY): New.
7171 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7172 inline entry marker notes.
7173 (print_insn): Likewise.
7174 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7175 (INSN_DEBUG_MARKER_KIND): Likewise.
7176 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7177 * tree-inline.c (expand_call_inline): Build and insert
7178 debug_inline_entry stmt.
7179 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7180 inline entry blocks early, if nonbind markers are enabled.
7181 (dump_scope_block): Dump fragment info.
7182 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7183 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7184 (gimple_build_debug_inline_entry): New.
7185 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7186 Enable/disable inline entry points too.
7187 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7188 (DEBUG_INSN): Describe inline entry markers.
7190 * common.opt (gvariable-location-views): New.
7191 (gvariable-location-views=incompat5): New.
7192 * config.in: Rebuilt.
7193 * configure: Rebuilt.
7194 * configure.ac: Test assembler for view support.
7195 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7196 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7197 * dwarf2out.c (var_loc_view): New typedef.
7198 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7199 (dwarf2out_locviews_in_attribute): New.
7200 (dwarf2out_locviews_in_loclist): New.
7201 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7202 (enum dw_line_info_opcode): Add LI_adv_address.
7203 (struct dw_line_info_table): Add view.
7204 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7205 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7206 (zero_view_p): New variable.
7207 (ZERO_VIEW_P): New macro.
7208 (output_asm_line_debug_info): New.
7209 (struct var_loc_node): Add view.
7210 (add_AT_view_list, AT_loc_list): New.
7211 (add_var_loc_to_decl): Add view param. Test it against last.
7212 (new_loc_list): Add view params. Record them.
7213 (AT_loc_list_ptr): Handle loc and view lists.
7214 (view_list_to_loc_list_val_node): New.
7215 (print_dw_val): Handle dw_val_class_view_list.
7216 (size_of_die): Likewise.
7217 (value_format): Likewise.
7218 (loc_list_has_views): New.
7219 (gen_llsym): Set vl_symbol too.
7220 (maybe_gen_llsym, skip_loc_list_entry): New.
7221 (dwarf2out_maybe_output_loclist_view_pair): New.
7222 (output_loc_list): Output view list or entries too.
7223 (output_view_list_offset): New.
7224 (output_die): Handle dw_val_class_view_list.
7225 (output_dwarf_version): New.
7226 (output_compilation_unit_header): Use it.
7227 (output_skeleton_debug_sections): Likewise.
7228 (output_rnglists, output_line_info): Likewise.
7229 (output_pubnames, output_aranges): Update version comments.
7230 (output_one_line_info_table): Output view numbers in asm comments.
7231 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7232 Call maybe_gen_llsym.
7233 (loc_list_from_tree_1): Adjust.
7234 (add_AT_location_description): Create view list attribute if
7235 needed, check it's absent otherwise.
7236 (convert_cfa_to_fb_loc_list): Adjust.
7237 (maybe_emit_file): Call output_asm_line_debug_info for test.
7238 (dwarf2out_var_location): Reset views as needed. Precompute
7239 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7240 attribute. Set view.
7241 (new_line_info_table): Reset next view.
7242 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7243 (dwarf2out_source_line): Likewise. Output view resets and labels to
7244 the assembler, or select appropriate line info opcodes.
7245 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7246 (optimize_string_length): Catch it. Adjust.
7247 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7248 dw_val_class_view_list, and remove it if no longer needed.
7249 (hash_loc_list): Hash view numbers.
7250 (loc_list_hasher::equal): Compare them.
7251 (optimize_location_lists): Check whether a view list symbol is
7252 needed, and whether the locview attribute is present, and
7253 whether they match. Remove the locview attribute if no longer
7255 (index_location_lists): Call skip_loc_list_entry for test.
7256 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7257 Use output_dwarf_version.
7258 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7259 (struct dw_val_node): Add val_view_list.
7260 * final.c (SEEN_NEXT_VIEW): New.
7261 (set_next_view_needed): New.
7262 (clear_next_view_needed): New.
7263 (maybe_output_next_view): New.
7264 (final_start_function): Rename to...
7265 (final_start_function_1): ... this. Take pointer to FIRST,
7266 add SEEN parameter. Emit param bindings in the initial view.
7267 (final_start_function): Reintroduce SEEN-less interface.
7268 (final): Rename to...
7269 (final_1): ... this. Take SEEN parameter. Output final pending
7270 next view at the end.
7271 (final): Reintroduce seen-less interface.
7272 (final_scan_insn): Output pending next view before switching
7273 sections or ending a block. Mark the next view as needed when
7274 outputting variable locations. Notify debug backend of section
7275 changes, and of location view changes.
7276 (rest_of_handle_final): Adjust.
7277 * toplev.c (process_options): Autodetect value for debug variable
7278 location views option. Warn on incompat5 without -gdwarf-5.
7279 * doc/invoke.texi (gvariable-location-views): New.
7280 (gvariable-location-views=incompat5): New.
7281 (gno-variable-location-views): New.
7283 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7285 PR tree-optimization/84136
7286 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7287 that the result of find_edge is non-NULL.
7289 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7292 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7293 storing integer register in SImode. Fix cost of 256 and 512
7294 byte aligned SSE register store.
7296 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7298 * config/i386/i386.c (ix86_multiplication_cost): Fix
7299 multiplication cost for TARGET_AVX512DQ.
7301 2018-02-08 Marek Polacek <polacek@redhat.com>
7303 PR tree-optimization/84238
7304 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7307 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7309 PR tree-optimization/84265
7310 * tree-vect-stmts.c (vectorizable_store): Don't treat
7311 VMAT_CONTIGUOUS accesses as grouped.
7312 (vectorizable_load): Likewise.
7314 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7316 PR tree-optimization/81635
7317 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7318 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7319 (test_round_for_mask): New functions.
7320 (wide_int_cc_tests): Call test_round_for_mask.
7321 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7322 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7323 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7324 range returned by get_range_info.
7326 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7329 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7330 * symtab.c: Include builtins.h
7331 (symtab_node::output_to_lto_symbol_table_p): Move here
7332 from lto-streamer-out.c:output_symbol_p.
7333 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7334 (output_symbol_p): Move all logic to symtab.c
7335 (produce_symtab): Update.
7337 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7339 * config/s390/s390-opts.h (enum indirect_branch): Define.
7340 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7341 (s390_indirect_branch_via_thunk)
7342 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7343 (enum s390_indirect_branch_type): Define.
7344 * config/s390/s390.c (struct s390_frame_layout, struct
7345 machine_function): Remove.
7346 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7347 (indirect_branch_table_label_no, indirect_branch_table_name):
7349 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7350 (enum s390_indirect_branch_option): Define.
7351 (s390_return_addr_from_memory): New function.
7352 (s390_handle_string_attribute): New function.
7353 (s390_attribute_table): Add new attribute handler.
7354 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7355 (s390_indirect_branch_via_thunk): New function.
7356 (s390_indirect_branch_via_inline_thunk): New function.
7357 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7358 sibling call optimization for non z10 compiles.
7359 (s390_emit_call): Force indirect branch target to be a single
7360 register. Add r1 clobber for non-z10 compiles.
7361 (s390_emit_epilogue): Emit return jump via return_use expander.
7362 (s390_reorg): Handle JUMP_INSNs as execute targets.
7363 (s390_option_override_internal): Perform validity checks for the
7364 new command line options.
7365 (s390_indirect_branch_attrvalue): New function.
7366 (s390_indirect_branch_settings): New function.
7367 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7368 (s390_output_indirect_thunk_function): New function.
7369 (s390_code_end): Implement target hook.
7370 (s390_case_values_threshold): Implement target hook.
7371 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7373 * config/s390/s390.h (struct s390_frame_layout)
7374 (struct machine_function): Move here from s390.c.
7375 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7376 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7377 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7378 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7379 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7380 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7381 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7382 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7383 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7384 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7385 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7386 (mnemonic attribute): Add values which aren't recognized
7388 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7389 pattern for branch conversion. Fix mnemonic attribute.
7390 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7391 indirect branch via thunk if requested.
7392 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7393 ("*indirect_jump"): Disable for branch conversion using out of
7395 ("indirect_jump_via_thunk<mode>_z10")
7396 ("indirect_jump_via_thunk<mode>")
7397 ("indirect_jump_via_inlinethunk<mode>_z10")
7398 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7399 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7400 ("casesi_jump_via_inlinethunk<mode>_z10")
7401 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7402 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7403 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7404 ("*indirect2_jump"): Disable for branch conversion.
7405 ("casesi_jump"): Turn into expander and expand patterns for branch
7407 ("return_use"): New expander.
7408 ("*return"): Emit return via thunk and rename it to ...
7409 ("*return<mode>"): ... this one.
7410 * config/s390/s390.opt: Add new options and and enum for the
7413 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7415 * lra-constraints.c (match_reload): Unconditionally use
7416 gen_lowpart_SUBREG, rather than selecting between that
7417 and equivalent gen_rtx_SUBREG code.
7419 2018-02-08 Richard Biener <rguenther@suse.de>
7421 PR tree-optimization/84233
7422 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7423 changed flag instead of boguously re-using phi_inserted.
7425 2018-02-08 Martin Jambor <mjambor@suse.cz>
7427 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7428 static local variables.
7430 2018-02-08 Richard Biener <rguenther@suse.de>
7432 PR tree-optimization/84278
7433 * tree-vect-stmts.c (vectorizable_store): When looking for
7434 smaller vector types to perform grouped strided loads/stores
7435 make sure the mode is supported by the target.
7436 (vectorizable_load): Likewise.
7438 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7440 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7441 Increase LDP/STP opportunities by adding adjacent callee-saves.
7443 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7445 PR rtl-optimization/84068
7446 PR rtl-optimization/83459
7447 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7449 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7451 PR tree-optimization/84224
7452 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7453 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7456 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7459 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7460 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7461 count, remove one USE.
7463 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7465 * doc/install.texi (Configuration): Document the
7466 --with-long-double-format={ibm,ieee} PowerPC configuration
7470 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7471 Convert from define_expand to be define_insn_and_split. Rework
7472 float/double/_Float128 conversions to QI/HI/SImode to work with
7473 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7474 conversions to QI/HImode types did a store and then a load to
7475 truncate the value. For conversions to VSX registers, don't split
7476 the insn, instead emit the code directly. Use the code iterator
7477 any_fix to combine signed and unsigned conversions.
7478 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7479 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7480 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7481 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7482 (fix_<mode>di2_hw): Likewise.
7483 (fixuns_<mode>di2_hw): Likewise.
7484 (fix_<mode>si2_hw): Likewise.
7485 (fixuns_<mode>si2_hw): Likewise.
7486 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7487 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7488 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7489 fix<uns>_trunc<SFDF:mode>si2_p8.
7490 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7492 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7493 (fix<uns>_<mode>_mem): Likewise.
7494 (fctiw<u>z_<mode>_mem): Likewise.
7495 (fix<uns>_<mode>_mem): Likewise.
7496 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7497 the register allocator from doing a direct move to the GPRs to do
7498 a store, and instead use the ISA 3.0 store byte/half-word from
7499 vector register instruction. For IEEE 128-bit floating point,
7500 also optimize stores of 32-bit ints.
7501 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7503 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7505 * genextract.c (push_pathstr_operand): New function to support
7507 (walk_rtx): Call push_pathstr_operand.
7508 (print_path): Support [a-zA-Z].
7510 2018-02-07 Richard Biener <rguenther@suse.de>
7512 PR tree-optimization/84037
7513 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7514 (cse_and_gimplify_to_preheader): Declare.
7515 (vect_get_place_in_interleaving_chain): Likewise.
7516 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7518 (_loop_vec_info::~_loop_vec_info): Delete it.
7519 (cse_and_gimplify_to_preheader): New function.
7520 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7521 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7522 (vectorizable_load): Likewise. For grouped stores always base
7523 the IV on the first element.
7524 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7525 condition before gimplifying.
7527 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7529 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7530 *DIV_EXPR and *MOD_EXPR.
7532 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7535 * config/i386/i386.c (ix86_option_override_internal): Mask out
7536 the CF_SET bit when checking -fcf-protection.
7538 2018-02-07 Tom de Vries <tom@codesourcery.com>
7541 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7544 2018-02-07 Richard Biener <rguenther@suse.de>
7546 PR tree-optimization/84204
7547 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7550 PR tree-optimization/84205
7551 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7552 special-case isl_ast_op_zdiv_r.
7554 PR tree-optimization/84223
7555 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7556 Only add conditions from within the region.
7557 (gather_bbs::after_dom_children): Adjust.
7559 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7562 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7563 * config/avr/avr.md: Only post-reload split REG-REG moves if
7564 either register is GENERAL_REG_P.
7566 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7568 PR tree-optimization/84235
7569 * tree-ssa-scopedtables.c
7570 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7571 if the subtraction is performed in floating point type where NaNs are
7572 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7573 build 1. Formatting fix.
7575 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7578 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7579 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7580 and skip it regardless of bb boundaries. Use CALL_P macro,
7581 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7582 unnecessarily, formatting fix.
7584 2018-02-06 Michael Collison <michael.collison@arm.com>
7586 * config/arm/thumb2.md:
7587 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7588 (*thumb_mov_notscc): Ditto.
7590 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7593 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7594 unsigned_fix, not "s".
7596 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7598 * configure.ac (gcc_fn_eh_frame_ro): New function.
7599 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7600 correct .eh_frame permissions.
7601 * configure: Regenerate.
7603 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7605 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7608 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7610 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7611 Display warning message for -mno-speculate-indirect-jumps.
7613 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7615 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7617 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7619 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7621 PR tree-optimization/84225
7622 * tree-eh.c (find_trapping_overflow): Only call
7623 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7625 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7628 * config/i386/i386.c: Reimplement the check of possible options
7629 -mibt/-mshstk conbination. Change error messages.
7630 * doc/invoke.texi: Fix a typo: remove extra '='.
7632 2018-02-06 Marek Polacek <polacek@redhat.com>
7634 PR tree-optimization/84228
7635 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7637 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7640 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7641 emitted arch directives.
7642 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7643 __ARM_FEATURE_COPROC before changing architectures.
7645 2018-02-06 Richard Biener <rguenther@suse.de>
7647 * config/i386/i386.c (print_reg): Fix typo.
7648 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7650 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7652 * configure: Regenerate.
7654 2018-02-05 Martin Sebor <msebor@redhat.com>
7656 PR tree-optimization/83369
7657 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7660 2018-02-05 Martin Liska <mliska@suse.cz>
7662 * doc/invoke.texi: Cherry-pick upstream r323995.
7664 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7666 * ira.c (ira_init_register_move_cost): Adjust comment.
7668 2018-02-05 Martin Liska <mliska@suse.cz>
7670 PR gcov-profile/84137
7671 * doc/gcov.texi: Fix typo in documentation.
7673 2018-02-05 Martin Liska <mliska@suse.cz>
7675 PR gcov-profile/83879
7676 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7677 using dlopen functionality.
7679 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7681 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7682 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7683 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7684 _mm_maskz_range_round_ss): New intrinsics.
7685 (__builtin_ia32_rangesd128_round)
7686 (__builtin_ia32_rangess128_round): Remove.
7687 (__builtin_ia32_rangesd128_mask_round,
7688 __builtin_ia32_rangess128_mask_round): New builtins.
7689 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7690 __builtin_ia32_rangess128_round): Remove.
7691 (__builtin_ia32_rangesd128_mask_round,
7692 __builtin_ia32_rangess128_mask_round): New builtins.
7693 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7694 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7695 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7696 "<round_saeonly_constraint>")): Changed to ...
7697 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7698 "<round_saeonly_scalar_constraint>")): ... this.
7699 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7700 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7701 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7702 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7703 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7705 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7707 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7709 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7710 Remove all values except native, 8540 and 8548.
7712 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7714 * config/i386/i386.c (ix86_output_function_return): Pass
7715 INVALID_REGNUM, instead of -1, as invalid register number to
7716 indirect_thunk_name and output_indirect_thunk.
7718 2018-02-02 Julia Koval <julia.koval@intel.com>
7720 * config.gcc: Add -march=icelake.
7721 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7722 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7723 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7724 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7725 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7726 (processor_target_table): Add icelake.
7727 (ix86_option_override_internal): Handle new PTAs.
7728 (get_builtin_code_for_version): Handle icelake.
7729 (M_INTEL_COREI7_ICELAKE): New.
7730 (fold_builtin_cpu): Handle icelake.
7731 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7732 * doc/invoke.texi: Add -march=icelake.
7734 2018-02-02 Julia Koval <julia.koval@intel.com>
7736 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7737 to wide_int_bitmask.
7738 * wide-int-bitmask.h: New.
7740 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7743 * config/i386/i386.md: Replace Pmode with word_mode in
7744 builtin_setjmp_setup and builtin_longjmp to support x32.
7746 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7750 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7752 (rs6000_supported_cpu_names): New static variable.
7753 (linux_cpu_translation_table): Likewise.
7754 (elf_platform) <cpu>: Define new static variable and use it.
7755 Translate kernel AT_PLATFORM name to canonical name if needed.
7756 Error if platform name is unknown.
7758 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7761 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7763 2018-02-01 Jeff Law <law@redhat.com>
7766 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7767 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7768 the scratch if RELEASE_VIA_POP is false.
7769 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7770 If we have to save a temporary register, decrement SIZE appropriately.
7771 Pass new arguments to release_scratch_register_on_entry.
7772 (ix86_adjust_stack_and_probe): Likewise.
7773 (ix86_emit_probe_stack_range): Pass new arguments to
7774 release_scratch_register_on_entry.
7776 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7778 PR rtl-optimization/84157
7779 * combine.c (change_zero_ext): Use REG_P predicate in
7780 front of HARD_REGISTER_P predicate.
7782 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7784 * config/avr/avr.c (avr_option_override): Move disabling of
7785 -fdelete-null-pointer-checks to...
7786 * common/config/avr/avr-common.c (avr_option_optimization_table):
7789 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7791 PR tree-optimization/81635
7792 * tree-data-ref.c (split_constant_offset_1): For types that
7793 wrap on overflow, try to use range info to prove that wrapping
7796 2018-02-01 Renlin Li <renlin.li@arm.com>
7799 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7801 (aarch64_register_move_cost): Likewise.
7802 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7804 (REG_CLASS_NAMES): Likewise.
7805 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7806 TAILCALL_ADDR_REGS. Remove IP registers.
7807 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7809 2018-02-01 Richard Biener <rguenther@suse.de>
7811 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7812 for specifying RPO order and allow NULL for that.
7813 * domwalk.c (dom_walker::dom_walker): Likewise.
7814 (dom_walker::walk): Handle NULL RPO order.
7815 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7817 (rewrite_update_dom_walker): Likewise.
7818 (mark_def_dom_walker): Likewise.
7820 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7822 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7823 (aarch64_maybe_expand_sve_subreg_move): Declare.
7824 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7825 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7827 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7828 that are semantically a reverse operation.
7829 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7830 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7831 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7833 (aarch64_can_change_mode_class): For big-endian, forbid changes
7834 between two SVE modes if they have different element sizes.
7836 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7838 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7839 the TImode handling for big-endian targets.
7841 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7843 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7844 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7846 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7847 Remove BSWAP handing for big-endian targets and use the form of
7848 LD1RQ appropariate for the mode.
7850 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7852 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7853 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7856 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7859 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7860 check for operands that need to go through aarch64_sve_reload_be.
7862 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7864 PR tree-optimization/81661
7865 PR tree-optimization/84117
7866 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7867 * tree-eh.c: Include gimplify.h.
7868 (find_trapping_overflow, replace_trapping_overflow,
7869 rewrite_to_non_trapping_overflow): New functions.
7870 * tree-vect-loop.c: Include tree-eh.h.
7871 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7872 * tree-data-ref.c: Include tree-eh.h.
7873 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7875 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7877 PR rtl-optimization/84123
7878 * combine.c (change_zero_ext): Check if hard register satisfies
7879 can_change_dest_mode before calling gen_lowpart_SUBREG.
7881 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7884 * ira.c (ira_init_register_move_cost): Remove assert.
7886 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7888 PR rtl-optimization/84071
7889 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7890 * doc/tm.texi: Regenerate.
7892 2018-01-31 Richard Biener <rguenther@suse.de>
7894 PR tree-optimization/84132
7895 * tree-data-ref.c (analyze_miv_subscript): Properly
7896 check whether evolution_function_is_affine_multivariate_p
7897 before calling gcd_of_steps_may_divide_p.
7899 2018-01-31 Julia Koval <julia.koval@intel.com>
7902 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7903 * config/i386/i386.md (rdpid_rex64) New.
7904 (rdpid): Make 32bit only.
7906 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7909 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7910 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7912 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7915 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7917 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7919 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7921 PR rtl-optimization/84071
7922 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7923 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7925 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7927 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7928 (arc_attribute_table): Add 'aux' attribute.
7929 (arc_in_small_data_p): Consider aux like variables.
7930 (arc_is_aux_reg_p): New function.
7931 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7932 (arc_get_aux_arg): New function.
7933 (prepare_move_operands): Handle aux-register access.
7934 (arc_handle_aux_attribute): New function.
7935 * doc/extend.texi (ARC Variable attributes): Add subsection.
7937 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7939 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7940 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7941 (arc_attribute_table): Add 'uncached' attribute.
7942 (arc_print_operand): Print '.di' flag for uncached memory
7944 (arc_in_small_data_p): Do not consider for small data the uncached
7946 (arc_is_uncached_mem_p): New function.
7947 * config/arc/predicates.md (compact_store_memory_operand): Check
7948 for uncached memory accesses.
7949 (nonvol_nonimm_operand): Likewise.
7950 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7952 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7955 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7956 falign-loops=): Add Optimization flag.
7958 2018-01-30 Jeff Law <law@redhat.com>
7961 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7962 INT_REGISTERS_SAVED. Check it prior to calling
7963 get_scratch_register_on_entry.
7964 (ix86_adjust_stack_and_probe): Similarly.
7965 (ix86_emit_probe_stack_range): Similarly.
7966 (ix86_expand_prologue): Corresponding changes.
7968 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7971 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7972 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7974 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7977 * lra-constraints.c (curr_insn_transform): Process AND in the
7980 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7982 PR rtl-optimization/83986
7983 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7984 dependence against last_pending_memory_flush in addition to
7987 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7989 PR tree-optimization/81611
7990 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7993 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7996 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7999 2018-01-30 Richard Biener <rguenther@suse.de>
8000 Jakub Jelinek <jakub@redhat.com>
8002 PR tree-optimization/84111
8003 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8004 inner loops added during recursion, as they don't have up-to-date
8007 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8010 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8011 (can_inline_edge_by_limits_p): ... here.
8012 (can_early_inline_edge_p, check_callers,
8013 update_caller_keys, update_callee_keys, recursive_inlining,
8014 add_new_edges_to_heap, speculation_useful_p,
8015 inline_small_functions,
8016 inline_small_functions, flatten_function,
8017 inline_to_all_callers_1): Update.
8019 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8021 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8024 2018-01-30 Richard Biener <rguenther@suse.de>
8026 PR tree-optimization/83008
8027 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8028 invariant and constant vector uses in stmts when they need
8031 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8034 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8035 * configure: Regenerate.
8037 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8039 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8041 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8042 Use gen_rtx_REG rather than gen_lowpart.
8044 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8046 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8047 rather than 0 when creating partial subregs.
8049 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8051 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8054 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8057 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8058 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8059 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8060 flags. This restores the settings used before the 2017-07-24.
8061 Turning off pre increment/decrement/modify allows IVOPTS to
8062 optimize DF/SF loops where the index is an int.
8064 2018-01-29 Richard Biener <rguenther@suse.de>
8065 Kelvin Nilsen <kelvin@gcc.gnu.org>
8068 * tree-vect-stmts.c (vectorizable_call): Don't call
8069 targetm.vectorize_builtin_md_vectorized_function if callee is
8072 2018-01-22 Carl Love <cel@us.ibm.com>
8074 * doc/extend.tex: Fix typo in second arg in
8075 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8077 2018-01-29 Richard Biener <rguenther@suse.de>
8079 PR tree-optimization/84086
8080 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8081 (flush_ssaname_freelist): When SSA names were released reset
8082 the SCEV hash table.
8084 2018-01-29 Richard Biener <rguenther@suse.de>
8086 PR tree-optimization/84057
8087 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8088 removed paths when removing edges.
8090 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8092 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8093 -mfunction-return=@var{choice}.
8095 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8098 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8100 (layout::print_source_line): Likewise.
8101 (test_get_line_width_without_trailing_whitespace): Add test cases.
8103 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8106 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8109 2018-01-26 Jim Wilson <jimw@sifive.com>
8111 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8113 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8116 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8118 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8119 and CMP + SUB-immediate -> SUBS.
8121 2018-01-26 Martin Sebor <msebor@redhat.com>
8123 PR tree-optimization/83896
8124 * tree-ssa-strlen.c (get_string_len): Rename...
8125 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8126 Avoid assuming length is constant.
8127 (handle_char_store): Use HOST_WIDE_INT for string length.
8129 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8132 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8133 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8135 2018-01-26 Richard Biener <rguenther@suse.de>
8137 PR rtl-optimization/84003
8138 * dse.c (record_store): Only record redundant stores when
8139 the earlier store aliases at least all accesses the later one does.
8141 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8143 PR rtl-optimization/83985
8144 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8145 REG_CFA_RESTORE insns.
8146 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8147 REG_CFA_RESTORE insns here.
8150 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8151 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8153 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8155 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8156 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8157 (arc_init): Likewise.
8158 (arc_override_options): Likewise.
8159 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8161 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8163 * config/arc/arc.h (TARGET_DBNZ): Define.
8164 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8165 properly set the tune attribute.
8166 (dbnz): Use TARGET_DBNZ guard.
8167 * config/arc/arc.opt (mtune): Add core3 option.
8169 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8171 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8172 recognize new pic like addresses.
8173 (arc_delegitimize_address): Clean up.
8175 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8177 * config/arc/arc-arches.def: Option mrf16 valid for all
8179 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8180 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8181 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8182 * config/arc/arc-tables.opt: Regenerate.
8183 * config/arc/arc.c (arc_conditional_register_usage): Handle
8184 reduced register file case.
8185 (arc_file_start): Set must have build attributes.
8186 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8188 * config/arc/arc.opt (mrf16): Add new option.
8189 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8190 * config/arc/genmultilib.awk: Handle new mrf16 option.
8191 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8192 * config/arc/t-multilib: Regenerate.
8193 * doc/invoke.texi (ARC Options): Document mrf16 option.
8195 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8197 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8198 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8199 (arc_attribute_table): Add 'secure_call' attribute.
8200 (arc_print_operand): Print secure call operand.
8201 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8203 (arc_is_secure_call_p): New function. * config/arc/arc.md
8204 (call_i): Add support for sjli instruction.
8205 (call_value_i): Likewise.
8206 * config/arc/constraints.md (Csc): New constraint.
8208 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8209 John Eric Martin <John.Martin@emmicro-us.com>
8211 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8212 * config/arc/arc.c (_arc_jli_section): New struct.
8213 (arc_jli_section): New type.
8214 (rc_jli_sections): New static variable.
8215 (arc_handle_jli_attribute): New function.
8216 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8217 (arc_file_end): New function.
8218 (TARGET_ASM_FILE_END): Define.
8219 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8220 (arc_add_jli_section): New function.
8221 (jli_call_scan): Likewise.
8222 (arc_reorg): Call jli_call_scan.
8223 (arc_output_addsi): Remove 'S' from printing asm operand.
8224 (arc_is_jli_call_p): New function.
8225 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8227 (movhi_insn): Likewise.
8228 (movsi_insn): Likewise.
8229 (movsi_set_cc_insn): Likewise.
8230 (loadqi_update): Likewise.
8231 (load_zeroextendqisi_update): Likewise.
8232 (load_signextendqisi_update): Likewise.
8233 (loadhi_update): Likewise.
8234 (load_zeroextendhisi_update): Likewise.
8235 (load_signextendhisi_update): Likewise.
8236 (loadsi_update): Likewise.
8237 (loadsf_update): Likewise.
8238 (movsicc_insn): Likewise.
8239 (bset_insn): Likewise.
8240 (bxor_insn): Likewise.
8241 (bclr_insn): Likewise.
8242 (bmsk_insn): Likewise.
8243 (bicsi3_insn): Likewise.
8244 (cmpsi_cc_c_insn): Likewise.
8245 (movsi_ne): Likewise.
8246 (movsi_cond_exec): Likewise.
8247 (clrsbsi2): Likewise.
8258 (call_i): Remove 'S' asm letter, add jli instruction.
8259 (call_value_i): Likewise.
8260 * config/arc/arc.op (mjli-always): New option.
8261 * config/arc/constraints.md (Cji): New constraint.
8262 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8264 (subsf3_fpx): Likewise.
8265 (mulsf3_fpx): Likewise.
8266 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8268 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8270 * doc/invoke.texi (ARC): Document mjli-always option.
8272 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8274 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8275 avoid addition with 0 and use incw and decw where possible.
8277 2018-01-26 Richard Biener <rguenther@suse.de>
8279 PR tree-optimization/81082
8280 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8281 association if it requires casting to unsigned.
8282 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8283 from fold_plusminus_mult_expr to catch important cases late when
8284 range info is available.
8286 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8288 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8289 * configure.ac (hidden_linkonce): New test.
8290 * configure: Regenerate.
8291 * config.in: Regenerate.
8293 2018-01-26 Julia Koval <julia.koval@intel.com>
8295 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8296 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8297 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8298 _mm_mask_bitshuffle_epi64_mask): Fix type.
8299 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8300 USI_FTYPE_V4DI_V4DI_USI): Remove.
8301 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8302 __builtin_ia32_vpshufbitqmb256_mask,
8303 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8304 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8305 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8307 2018-01-26 Alan Modra <amodra@gmail.com>
8310 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8311 UNSPEC_VBPERMQ. Sort other unspecs.
8313 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8315 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8317 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8320 * predict.c (drop_profile): Do not push/pop cfun; update also
8322 (handle_missing_profiles): Fix logic looking for zero profiles.
8324 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8327 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8328 on functions with #pragma omp declare simd or functions with simd
8330 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8331 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8332 Remove trailing \n from warning_at calls.
8334 2018-01-25 Tom de Vries <tom@codesourcery.com>
8337 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8338 for neutered workers.
8340 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8343 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8344 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8346 2018-01-24 Jeff Law <law@redhat.com>
8349 * i386.c (get_probe_interval): Move to earlier point.
8350 (ix86_compute_frame_layout): If -fstack-clash-protection and
8351 the frame is larger than the probe interval, then use pushes
8352 to save registers rather than reg->mem moves.
8353 (ix86_expand_prologue): Remove conditional for int_registers_saved
8356 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8359 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8360 min/max for never referenced object.
8362 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8365 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8367 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8368 attributes from DECL_ATTRIBUTES (decl) without affecting
8369 DECL_ATTRIBUTES (current_function_decl).
8370 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8371 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8373 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8375 PR tree-optimization/83979
8376 * fold-const.c (fold_comparison): Use constant_boolean_node
8377 instead of boolean_{true,false}_node.
8379 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8381 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8384 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8386 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8387 Simplify the clause that sets the length attribute.
8388 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8389 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8390 clause that sets the length attribute.
8391 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8393 2018-01-24 Tom de Vries <tom@codesourcery.com>
8396 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8397 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8398 Add strict parameter.
8399 (prevent_branch_around_nothing): Insert dummy insn between branch to
8400 label and label with no ptx insn inbetween.
8401 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8403 2018-01-24 Tom de Vries <tom@codesourcery.com>
8406 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8407 for neutered threads in warp.
8408 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8410 2018-01-24 Richard Biener <rguenther@suse.de>
8412 PR tree-optimization/83176
8413 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8416 2018-01-24 Richard Biener <rguenther@suse.de>
8418 PR tree-optimization/82819
8419 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8420 code generating pluses that are no-ops in the target precision.
8422 2018-01-24 Richard Biener <rguenther@suse.de>
8425 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8427 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8429 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8430 to merge probabilities.
8431 * predict.c (probably_never_executed): Also mark as cold functions
8432 with global 0 profile and guessed local profile.
8433 * profile-count.c (profile_probability::combine_with_count): New
8435 * profile-count.h (profile_probability::operator*,
8436 profile_probability::operator*=, profile_probability::operator/,
8437 profile_probability::operator/=): Reduce precision to adjusted
8438 and set value to guessed on contradictory divisions.
8439 (profile_probability::combine_with_freq): Remove.
8440 (profile_probability::combine_wiht_count): Declare.
8441 (profile_count::force_nonzero):: Set to adjusted.
8442 (profile_count::probability_in):: Set quality to adjusted.
8443 * tree-ssa-tail-merge.c (replace_block_by): Use
8446 2018-01-23 Andrew Waterman <andrew@sifive.com>
8447 Jim Wilson <jimw@sifive.com>
8449 * config/riscv/riscv.c (riscv_stack_boundary): New.
8450 (riscv_option_override): Set riscv_stack_boundary. Handle
8451 riscv_preferred_stack_boundary_arg.
8452 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8453 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8454 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8455 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8456 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8457 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8459 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8462 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8463 of struct ix86_frame.
8464 (ix86_expand_epilogue): Likewise. Add a local variable for
8465 the reg_save_offset field in struct ix86_frame.
8467 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8469 PR tree-optimization/82604
8470 * tree-loop-distribution.c (enum partition_kind): New enum item
8471 PKIND_PARTIAL_MEMSET.
8472 (partition_builtin_p): Support above new enum item.
8473 (generate_code_for_partition): Ditto.
8474 (compute_access_range): Differentiate cases that equality can be
8475 proven at all loops, the innermost loops or no loops.
8476 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8477 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8478 (finalize_partitions, distribute_loop): Don't fuse partition of
8479 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8480 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8483 2018-01-23 Martin Liska <mliska@suse.cz>
8485 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8486 order to ignore the predictor.
8487 (PRED_POLYMORPHIC_CALL): Likewise.
8488 (PRED_RECURSIVE_CALL): Likewise.
8490 2018-01-23 Martin Liska <mliska@suse.cz>
8492 * tree-profile.c (tree_profiling): Print function header to
8493 aware reader which function we are working on.
8494 * value-prof.c (gimple_find_values_to_profile): Do not print
8495 not interesting value histograms.
8497 2018-01-23 Martin Liska <mliska@suse.cz>
8499 * profile-count.h (enum profile_quality): Add
8500 profile_uninitialized as the first value. Do not number values
8501 as they are zero based.
8502 (profile_count::verify): Update sanity check.
8503 (profile_probability::verify): Likewise.
8505 2018-01-23 Nathan Sidwell <nathan@acm.org>
8507 * doc/invoke.texi (ffor-scope): Deprecate.
8509 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8511 PR tree-optimization/83510
8512 * domwalk.c (set_all_edges_as_executable): New function.
8513 (dom_walker::dom_walker): Convert bool param
8514 "skip_unreachable_blocks" to enum reachability. Move setup of
8515 edge flags to set_all_edges_as_executable and only do it when
8516 reachability is REACHABLE_BLOCKS.
8517 * domwalk.h (enum dom_walker::reachability): New enum.
8518 (dom_walker::dom_walker): Convert bool param
8519 "skip_unreachable_blocks" to enum reachability.
8520 (set_all_edges_as_executable): New decl.
8521 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8522 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8524 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8525 but converting true to REACHABLE_BLOCKS.
8526 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8528 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8529 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8530 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8532 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8533 if check_all_array_refs will be called.
8535 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8537 * tree.c (selftest::test_location_wrappers): Add more test
8540 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8542 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8543 (selftest::test_bit_in_range): Likewise.
8545 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8548 * doc/sourcebuild.texi (vect_float): Say that the selector
8549 only describes the situation when -funsafe-math-optimizations is on.
8550 (vect_float_strict): Document.
8552 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8554 PR tree-optimization/83965
8555 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8556 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8557 instead of checking only for a reduction.
8558 (vect_recog_widen_sum_pattern): Likewise.
8560 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8562 * predict.c (probably_never_executed): Only use precise profile info.
8563 (compute_function_frequency): Skip after inlining hack since we now
8564 have quality checking.
8566 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8568 * profile-count.h (profile_probability::very_unlikely,
8569 profile_probability::unlikely, profile_probability::even): Set
8570 precision to guessed.
8572 2018-01-23 Richard Biener <rguenther@suse.de>
8574 PR tree-optimization/83963
8575 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8576 Properly terminate dominator walk when crossing the exit edge not
8577 when visiting its source block.
8579 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8582 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8583 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8585 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8587 PR tree-optimization/83957
8588 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8589 semicolon after for body surrounded by braces.
8591 PR tree-optimization/83081
8592 * profile-count.h (profile_probability::split): New method.
8593 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8594 Use profile_probability::split.
8595 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8596 when splitting a single conditional jump into 2.
8598 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8600 PR tree-optimization/69452
8601 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8604 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8606 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8607 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8608 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8610 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8612 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8613 * config/rl78/rl78.md (movdi): New define_expand.
8614 * config/rl78/rl78.c (rl78_split_movdi): New function.
8616 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8619 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8621 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8622 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8623 128-bit to produce an UNSPEC move to get the double word with the
8624 signbit and then a shift directly to do signbit.
8625 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8626 implementation with a new version that just does either a direct
8627 move or a regular move. Move memory interface to separate insns.
8628 Move insns so they are next to the expander.
8629 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8630 with signbit move. Split big and little endian case.
8631 (signbit<mode>2_dm_mem_le): Likewise.
8632 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8633 (signbit<mode>2_dm2): Likewise.
8635 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8637 * config/rl78/rl78.md (anddi3): New define_expand.
8639 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8641 * config/rl78/rl78.md (umindi3): New define_expand.
8643 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8645 * config/rl78/rl78.md (smindi3): New define_expand.
8647 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8649 * config/rl78/rl78.md (smaxdi3): New define_expand.
8651 2018-01-22 Carl Love <cel@us.ibm.com>
8653 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8654 LVX_V1TI): Add macro expansion.
8655 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8656 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8657 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8658 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8659 Change check to determine if the instruction is a byte reversing
8660 entry. Fix typo in comment.
8661 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8662 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8663 Add def_builtin calls for new builtins.
8664 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8665 Add define_insn expansion.
8667 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8669 * config/rl78/rl78.md (umaxdi3): New define_expand.
8671 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8673 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8674 for non-QImode registers.
8676 2018-01-22 Richard Biener <rguenther@suse.de>
8678 PR tree-optimization/83963
8679 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8680 including the loop exit block.
8681 (scop_detection::merge_sese): Likewise.
8682 (scop_detection::add_scop): Do it here instead.
8684 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8686 * doc/sourcebuild.texi (arm_softfloat): Document.
8688 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8691 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8692 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8693 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8695 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8696 David Edelsohn <dje.gcc@gmail.com>
8699 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8700 Change "crset eq" to "crset 2".
8701 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8702 (*call_indirect_aix<mode>_nospec): Likewise.
8703 (*call_value_indirect_aix<mode>_nospec): Likewise.
8704 (*call_indirect_elfv2<mode>_nospec): Likewise.
8705 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8706 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8707 change assembly output from . to $.
8708 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8709 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8710 (*tablejump<mode>_internal1_nospec): Likewise.
8712 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8715 * config/sh/sh_optimize_sett_clrt.cc:
8716 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8718 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8720 PR tree-optimization/83940
8721 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8722 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8723 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8724 use it to pass back the definition type.
8725 (vect_check_store_rhs): Likewise rhs_dt_out.
8726 (vect_build_gather_load_calls): Add a mask_dt argument and use
8727 it instead of a call to vect_is_simple_use.
8728 (vectorizable_store): Update calls to vect_check_load_store_mask
8729 and vect_check_store_rhs. Use the dt returned by the latter instead
8730 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8731 instead of calls to vect_is_simple_use. Pass the scalar rather
8732 than the vector operand to vect_is_simple_use when handling
8733 second and subsequent copies of an rhs value.
8734 (vectorizable_load): Update calls to vect_check_load_store_mask
8735 and vect_build_gather_load_calls. Use the cached mask_dt and
8736 gs_info.offset_dt instead of calls to vect_is_simple_use.
8738 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8741 * tree-emutls.c: Include gimplify.h.
8742 (lower_emutls_2): New function.
8743 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8744 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8745 it before further processing.
8748 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8749 UINTVAL (trueop1) instead of INTVAL (op1).
8751 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8755 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8756 INCOMING_FRAME_SP_OFFSET if not defined.
8757 (scan_trace): Add ENTRY argument. If true and
8758 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8759 emit a note to adjust the CFA offset.
8760 (create_cfi_notes): Adjust scan_trace callers.
8761 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8762 INCOMING_FRAME_SP_OFFSET in the CIE.
8763 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8764 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8766 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8767 * doc/tm.texi: Regenerated.
8769 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8771 PR rtl-optimization/83147
8772 * lra-constraints.c (remove_inheritance_pseudos): Use
8773 lra_substitute_pseudo_within_insn.
8775 2018-01-19 Tom de Vries <tom@codesourcery.com>
8776 Cesar Philippidis <cesar@codesourcery.com>
8779 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8781 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8784 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8785 spaces for function labels.
8787 2018-01-19 Martin Liska <mliska@suse.cz>
8789 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8790 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8791 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8792 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8793 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8794 (PRED_CONST_RETURN): Change from 69 to 65.
8795 (PRED_NULL_RETURN): Change from 91 to 71.
8796 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8797 (PRED_LOOP_GUARD): Change from 66 to 73.
8799 2018-01-19 Martin Liska <mliska@suse.cz>
8801 * predict.c (predict_insn_def): Add new assert.
8802 (struct branch_predictor): Change type to signed integer.
8803 (test_prediction_value_range): Amend test to cover
8805 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8806 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8807 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8808 (PRED_LOOP_IV_COMPARE): Likewise.
8809 * predict.h (PROB_UNINITIALIZED): Define new constant.
8811 2018-01-19 Martin Liska <mliska@suse.cz>
8813 * predict.c (dump_prediction): Add new format for
8814 analyze_brprob.py script which is enabled with -details
8816 * profile-count.h (precise_p): New function.
8818 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8820 PR tree-optimization/83922
8821 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8822 there are no statements that need masking.
8823 (vect_active_double_reduction_p): New function.
8824 (vect_analyze_loop_operations): Use it when handling phis that
8825 are not in the loop header.
8827 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8829 PR tree-optimization/83914
8830 * tree-vect-loop.c (vectorizable_induction): Don't convert
8831 init_expr or apply the peeling adjustment for inductions
8832 that are nested within the vectorized loop.
8834 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8836 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8839 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8843 * function.h (gimplify_parameters): Add gimple_seq * argument.
8844 * function.c: Include gimple.h and options.h.
8845 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8846 for the added local temporaries if needed.
8847 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8848 if there are any parameter cleanups, wrap whole body into a
8849 try/finally with the cleanups.
8851 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8854 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8855 Use GET_MODE_CLASS for scalar floating point.
8857 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8861 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8862 Fix call of call_cgraph_insertion_hooks.
8864 2018-01-18 Martin Sebor <msebor@redhat.com>
8866 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8868 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8871 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8874 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8877 * common.opt: (-ffile-prefix-map): New option.
8878 * opts.c (common_handle_option): Defer it.
8879 * opts-global.c (handle_common_deferred_options): Handle it.
8880 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8881 * file-prefix-map.h: New file.
8882 (remap_debug_filename, add_debug_prefix_map): ...here.
8883 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8884 * final.c (debug_prefix_map, add_debug_prefix_map
8885 remap_debug_filename): Move to...
8886 * file-prefix-map.c: New file.
8887 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8888 generalize, get rid of alloca(), use strrchr() instead of strchr().
8889 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8890 Implement in terms of add_prefix_map().
8891 (remap_macro_filename, remap_debug_filename): Implement in term of
8893 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8894 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8895 * dbxout.c: Include file-prefix-map.h.
8896 * varasm.c: Likewise.
8897 * vmsdbgout.c: Likewise.
8898 * xcoffout.c: Likewise.
8899 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8900 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8901 * doc/invoke.texi (-ffile-prefix-map): Document.
8902 (-fdebug-prefix-map): Update description.
8904 2018-01-18 Martin Liska <mliska@suse.cz>
8906 * config/i386/i386.c (indirect_thunk_name): Document that also
8908 (output_indirect_thunk): Document why both instructions
8909 (pause and lfence) are generated.
8911 2018-01-18 Richard Biener <rguenther@suse.de>
8913 PR tree-optimization/83887
8914 * graphite-scop-detection.c
8915 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8916 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8917 (scop_detection::merge_sese): Re-implement with a flood-fill
8918 algorithm that properly finds a SESE region if it exists.
8920 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8923 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8924 pointer_diff optimizations use view_convert instead of convert.
8926 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8928 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8929 Generate different code for -mno-speculate-indirect-jumps.
8930 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8931 (*call_indirect_aix<mode>): Disable for
8932 -mno-speculate-indirect-jumps.
8933 (*call_indirect_aix<mode>_nospec): New define_insn.
8934 (*call_value_indirect_aix<mode>): Disable for
8935 -mno-speculate-indirect-jumps.
8936 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8937 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8938 -mno-speculate-indirect-jumps.
8939 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8941 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8943 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8944 long double type, set the flags for noting the default long double
8945 type, even if we don't pass or return a long double type.
8947 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8950 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8953 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8955 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8956 support for merge[hl].
8957 (fold_mergehl_helper): New helper function.
8958 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8959 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8960 (altivec_vmrglw_direct): Add xxmrglw insn.
8962 2018-01-17 Andrew Waterman <andrew@sifive.com>
8964 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8965 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8967 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8970 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8971 call the lto_location_cache before reading the
8972 DECL_SOURCE_LOCATION of the types.
8974 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8975 Richard Sandiford <richard.sandiford@linaro.org>
8977 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8978 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8979 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8980 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8981 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8983 * config/aarch64/constraints.md (aarch64_movti_operand):
8985 * config/aarch64/predicates.md (Uti): Add new constraint.
8987 2018-01-17 Carl Love <cel@us.ibm.com>
8989 * config/rs6000/vsx.md (define_expand xl_len_r,
8990 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8991 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8993 (define_expand, define_insn): Move the shift left from the
8994 define_insn to the define_expand for lxvl and stxvl instructions.
8995 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8996 and XL_LEN_R definitions to PURE.
8998 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9000 * config/i386/i386.c (indirect_thunk_name): Declare regno
9001 as unsigned int. Compare regno with INVALID_REGNUM.
9002 (output_indirect_thunk): Ditto.
9003 (output_indirect_thunk_function): Ditto.
9004 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9005 in the call to output_indirect_thunk_function.
9007 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9010 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9011 rather than the size of inner_type to determine the stack slot size
9012 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9014 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9017 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9020 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9022 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9023 endian Linux systems to optionally enable multilibs for selecting
9024 the long double type if the user configured an explicit type.
9025 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9026 have no long double multilibs if not defined.
9027 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9028 warn if the user used -mabi={ieee,ibm}longdouble and we built
9029 multilibs for long double.
9030 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9031 appropriate multilib option.
9032 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9034 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9035 for building long double multilibs.
9036 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9038 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9040 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9043 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9045 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9048 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9051 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9054 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9056 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9057 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9059 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9061 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9062 different rtl trees depending on TARGET_64BIT.
9063 (rs6000_gen_lvx): Likewise.
9065 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9067 * config/visium/visium.md (nop): Tweak comment.
9068 (hazard_nop): Likewise.
9070 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9072 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9073 -mspeculate-indirect-jumps.
9074 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9075 for -mno-speculate-indirect-jumps.
9076 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9077 (*call_value_indirect_elfv2<mode>): Disable for
9078 -mno-speculate-indirect-jumps.
9079 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9080 (indirect_jump): Emit different RTL for
9081 -mno-speculate-indirect-jumps.
9082 (*indirect_jump<mode>): Disable for
9083 -mno-speculate-indirect-jumps.
9084 (*indirect_jump<mode>_nospec): New define_insn.
9085 (tablejump): Emit different RTL for
9086 -mno-speculate-indirect-jumps.
9087 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9088 (tablejumpsi_nospec): New define_expand.
9089 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9090 (tablejumpdi_nospec): New define_expand.
9091 (*tablejump<mode>_internal1): Disable for
9092 -mno-speculate-indirect-jumps.
9093 (*tablejump<mode>_internal1_nospec): New define_insn.
9094 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9097 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9099 * caller-save.c (insert_save): Drop unnecessary parameter. All
9102 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9103 Richard Biener <rguenth@suse.de>
9106 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9107 return early, inline manually is_gimple_sizepos. Make sure if we
9108 call gimplify_expr we don't end up with a gimple constant.
9109 * tree.c (variably_modified_type_p): Don't return true for
9110 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9111 * gimplify.h (is_gimple_sizepos): Remove.
9113 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9115 PR tree-optimization/83857
9116 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9117 vectorizable_live_operation for pure SLP statements.
9118 (vectorizable_live_operation): Handle PHIs.
9120 2018-01-16 Richard Biener <rguenther@suse.de>
9122 PR tree-optimization/83867
9123 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9124 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9126 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9129 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9130 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9131 If off is not INTEGER_CST, issue a may not be aligned warning
9132 rather than isn't aligned. Use isn%'t rather than isn't.
9133 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9135 <case MULT_EXPR>: Improve the case when bottom and one of the
9136 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9137 operand, in that case check if the other operand is multiple of
9138 bottom divided by the INTEGER_CST operand.
9140 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9143 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9144 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9145 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9146 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9147 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9148 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9149 (pa_function_arg_size): New function.
9151 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9153 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9154 in a separate statement.
9156 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9158 PR tree-optimization/83847
9159 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9160 group gathers and scatters.
9162 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9164 PR rtl-optimization/86620
9165 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9167 PR rtl-optimization/83213
9168 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9169 to last if both are JUMP_INSNs.
9171 PR tree-optimization/83843
9172 * gimple-ssa-store-merging.c
9173 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9174 store_immediate_info for bswap/nop orig_stores.
9176 2018-01-15 Andrew Waterman <andrew@sifive.com>
9178 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9180 <UDIV>: Increase cost if !TARGET_DIV.
9182 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9184 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9185 (define_attr "cr_logical_3op"): New.
9186 (cceq_ior_compare): Adjust.
9187 (cceq_ior_compare_complement): Adjust.
9188 (*cceq_rev_compare): Adjust.
9189 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9190 (is_cracked_insn): Adjust.
9191 (insn_must_be_first_in_group): Adjust.
9192 * config/rs6000/40x.md: Adjust.
9193 * config/rs6000/440.md: Adjust.
9194 * config/rs6000/476.md: Adjust.
9195 * config/rs6000/601.md: Adjust.
9196 * config/rs6000/603.md: Adjust.
9197 * config/rs6000/6xx.md: Adjust.
9198 * config/rs6000/7450.md: Adjust.
9199 * config/rs6000/7xx.md: Adjust.
9200 * config/rs6000/8540.md: Adjust.
9201 * config/rs6000/cell.md: Adjust.
9202 * config/rs6000/e300c2c3.md: Adjust.
9203 * config/rs6000/e500mc.md: Adjust.
9204 * config/rs6000/e500mc64.md: Adjust.
9205 * config/rs6000/e5500.md: Adjust.
9206 * config/rs6000/e6500.md: Adjust.
9207 * config/rs6000/mpc.md: Adjust.
9208 * config/rs6000/power4.md: Adjust.
9209 * config/rs6000/power5.md: Adjust.
9210 * config/rs6000/power6.md: Adjust.
9211 * config/rs6000/power7.md: Adjust.
9212 * config/rs6000/power8.md: Adjust.
9213 * config/rs6000/power9.md: Adjust.
9214 * config/rs6000/rs64.md: Adjust.
9215 * config/rs6000/titan.md: Adjust.
9217 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9219 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9220 ix86_indirect_branch_register logic.
9222 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9224 * config/i386/constraints.md (Bs): Update
9225 ix86_indirect_branch_register check. Don't check
9226 ix86_indirect_branch_register with GOT_memory_operand.
9228 * config/i386/predicates.md (GOT_memory_operand): Don't check
9229 ix86_indirect_branch_register here.
9230 (GOT32_symbol_operand): Likewise.
9232 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9234 * config/i386/predicates.md (constant_call_address_operand):
9235 Rewrite ix86_indirect_branch_register logic.
9236 (sibcall_insn_operand): Likewise.
9238 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9240 * config/i386/constraints.md (Bs): Replace
9241 ix86_indirect_branch_thunk_register with
9242 ix86_indirect_branch_register.
9244 * config/i386/i386.md (indirect_jump): Likewise.
9245 (tablejump): Likewise.
9246 (*sibcall_memory): Likewise.
9247 (*sibcall_value_memory): Likewise.
9248 Peepholes of indirect call and jump via memory: Likewise.
9249 * config/i386/i386.opt: Likewise.
9250 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9251 (GOT_memory_operand): Likewise.
9252 (call_insn_operand): Likewise.
9253 (sibcall_insn_operand): Likewise.
9254 (GOT32_symbol_operand): Likewise.
9256 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9259 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9260 type rather than type addr's type points to.
9261 (expand_omp_atomic_mutex): Likewise.
9262 (expand_omp_atomic): Likewise.
9264 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9267 * config/i386/i386.c (output_indirect_thunk_function): Use
9268 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9269 for __x86_return_thunk.
9271 2018-01-15 Richard Biener <rguenther@suse.de>
9274 * expmed.c (extract_bit_field_1): Fix typo.
9276 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9279 * config/arm/iterators.md (VF): New mode iterator.
9280 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9281 Remove integer-related logic from pattern.
9282 (neon_vabd<mode>_3): Likewise.
9284 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9287 * common.opt (fstrict-overflow): No longer an alias.
9288 (fwrapv-pointer): New option.
9289 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9290 also for pointer types based on flag_wrapv_pointer.
9291 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9292 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9293 opts->x_flag_wrapv got set.
9294 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9295 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9296 POINTER_TYPE_OVERFLOW_UNDEFINED.
9297 * match.pd: Likewise in address comparison pattern.
9298 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9300 2018-01-15 Richard Biener <rguenther@suse.de>
9303 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9304 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9305 Reset type names to their identifier if their TYPE_DECL doesn't
9306 have linkage (and thus is used for ODR and devirt).
9307 (save_debug_info_for_decl): Remove.
9308 (save_debug_info_for_type): Likewise.
9309 (add_tree_to_fld_list): Adjust.
9310 * tree-pretty-print.c (dump_generic_node): Make dumping of
9311 type names more robust.
9313 2018-01-15 Richard Biener <rguenther@suse.de>
9315 * BASE-VER: Bump to 8.0.1.
9317 2018-01-14 Martin Sebor <msebor@redhat.com>
9320 * builtins.c (check_access): Avoid warning when the no-warning bit
9323 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9325 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9326 * ira-color (allocno_hard_regs_compare): Likewise.
9328 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9331 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9332 Use .pushsection/.popsection.
9334 2018-01-14 Martin Sebor <msebor@redhat.com>
9337 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9339 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9341 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9342 entry from extra_headers.
9343 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9344 extra_headers, make the list bitwise identical to the i?86-*-* one.
9346 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9348 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9349 -mcmodel=large with -mindirect-branch=thunk,
9350 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9351 -mfunction-return=thunk-extern.
9352 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9353 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9354 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9356 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9358 * config/i386/i386.c (print_reg): Print the name of the full
9359 integer register without '%'.
9360 (ix86_print_operand): Handle 'V'.
9361 * doc/extend.texi: Document 'V' modifier.
9363 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9365 * config/i386/constraints.md (Bs): Disallow memory operand for
9366 -mindirect-branch-register.
9368 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9369 (GOT_memory_operand): Likewise.
9370 (call_insn_operand): Likewise.
9371 (sibcall_insn_operand): Likewise.
9372 (GOT32_symbol_operand): Likewise.
9373 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9374 for -mindirect-branch-register.
9375 (tablejump): Likewise.
9376 (*sibcall_memory): Likewise.
9377 (*sibcall_value_memory): Likewise.
9378 Disallow peepholes of indirect call and jump via memory for
9379 -mindirect-branch-register.
9380 (*call_pop): Replace m with Bw.
9381 (*call_value_pop): Likewise.
9382 (*sibcall_pop_memory): Replace m with Bs.
9383 * config/i386/i386.opt (mindirect-branch-register): New option.
9384 * doc/invoke.texi: Document -mindirect-branch-register option.
9386 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9388 * config/i386/i386-protos.h (ix86_output_function_return): New.
9389 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9390 set function_return_type.
9391 (indirect_thunk_name): Add ret_p to indicate thunk for function
9393 (output_indirect_thunk_function): Pass false to
9394 indirect_thunk_name.
9395 (ix86_output_indirect_branch_via_reg): Likewise.
9396 (ix86_output_indirect_branch_via_push): Likewise.
9397 (output_indirect_thunk_function): Create alias for function
9398 return thunk if regno < 0.
9399 (ix86_output_function_return): New function.
9400 (ix86_handle_fndecl_attribute): Handle function_return.
9401 (ix86_attribute_table): Add function_return.
9402 * config/i386/i386.h (machine_function): Add
9403 function_return_type.
9404 * config/i386/i386.md (simple_return_internal): Use
9405 ix86_output_function_return.
9406 (simple_return_internal_long): Likewise.
9407 * config/i386/i386.opt (mfunction-return=): New option.
9408 (indirect_branch): Mention -mfunction-return=.
9409 * doc/extend.texi: Document function_return function attribute.
9410 * doc/invoke.texi: Document -mfunction-return= option.
9412 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9414 * config/i386/i386-opts.h (indirect_branch): New.
9415 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9416 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9417 with local indirect jump when converting indirect call and jump.
9418 (ix86_set_indirect_branch_type): New.
9419 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9420 (indirectlabelno): New.
9421 (indirect_thunk_needed): Likewise.
9422 (indirect_thunk_bnd_needed): Likewise.
9423 (indirect_thunks_used): Likewise.
9424 (indirect_thunks_bnd_used): Likewise.
9425 (INDIRECT_LABEL): Likewise.
9426 (indirect_thunk_name): Likewise.
9427 (output_indirect_thunk): Likewise.
9428 (output_indirect_thunk_function): Likewise.
9429 (ix86_output_indirect_branch_via_reg): Likewise.
9430 (ix86_output_indirect_branch_via_push): Likewise.
9431 (ix86_output_indirect_branch): Likewise.
9432 (ix86_output_indirect_jmp): Likewise.
9433 (ix86_code_end): Call output_indirect_thunk_function if needed.
9434 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9436 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9437 (ix86_attribute_table): Add indirect_branch.
9438 * config/i386/i386.h (machine_function): Add indirect_branch_type
9439 and has_local_indirect_jump.
9440 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9442 (tablejump): Likewise.
9443 (*indirect_jump): Use ix86_output_indirect_jmp.
9444 (*tablejump_1): Likewise.
9445 (simple_return_indirect_internal): Likewise.
9446 * config/i386/i386.opt (mindirect-branch=): New option.
9447 (indirect_branch): New.
9450 (thunk-inline): Likewise.
9451 (thunk-extern): Likewise.
9452 * doc/extend.texi: Document indirect_branch function attribute.
9453 * doc/invoke.texi: Document -mindirect-branch= option.
9455 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9458 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9460 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9462 * ipa-inline.c (want_inline_small_function_p): Return false if
9463 inlining has already failed with CIF_FINAL_ERROR.
9464 (update_caller_keys): Call want_inline_small_function_p before
9466 (update_callee_keys): Likewise.
9468 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9470 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9472 (rs6000_quadword_masked_address_p): Likewise.
9473 (quad_aligned_load_p): Likewise.
9474 (quad_aligned_store_p): Likewise.
9475 (const_load_sequence_p): Add comment to describe the outer-most loop.
9476 (mimic_memory_attributes_and_flags): New function.
9477 (rs6000_gen_stvx): Likewise.
9478 (replace_swapped_aligned_store): Likewise.
9479 (rs6000_gen_lvx): Likewise.
9480 (replace_swapped_aligned_load): Likewise.
9481 (replace_swapped_load_constant): Capitalize argument name in
9482 comment describing this function.
9483 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9484 and stores that access quad-word aligned addresses and replace
9485 with stvx or lvx instructions when appropriate.
9486 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9487 New function prototype.
9488 (rs6000_quadword_masked_address_p): Likewise.
9489 (rs6000_gen_lvx): Likewise.
9490 (rs6000_gen_stvx): Likewise.
9491 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9492 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9493 when memory address is aligned.
9494 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9495 this split to select lvx instruction when memory address is aligned.
9496 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9497 instruction when memory address is aligned.
9498 (*vsx_le_perm_load_v16qi): Likewise.
9499 (four unnamed splitters): Modify to select the stvx instruction
9500 when memory is aligned.
9502 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9504 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9505 which appears in the queue multiple times.
9507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9508 Alan Hayward <alan.hayward@arm.com>
9509 David Sherwood <david.sherwood@arm.com>
9511 * tree-vectorizer.h (vec_lower_bound): New structure.
9512 (_loop_vec_info): Add check_nonzero and lower_bounds.
9513 (LOOP_VINFO_CHECK_NONZERO): New macro.
9514 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9515 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9516 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9517 fields. Make seg_len the distance travelled, not including the
9519 (dr_direction_indicator): Declare.
9520 (dr_zero_step_indicator): Likewise.
9521 (dr_known_forward_stride_p): Likewise.
9522 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9524 (runtime_alias_check_p): Allow runtime alias checks with
9526 (operator ==): Compare access_size and align.
9527 (prune_runtime_alias_test_list): Rework for new distinction between
9528 the access_size and seg_len.
9529 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9531 (get_segment_min_max): New function.
9532 (create_intersect_range_checks): Use it.
9533 (dr_step_indicator): New function.
9534 (dr_direction_indicator): Likewise.
9535 (dr_zero_step_indicator): Likewise.
9536 (dr_known_forward_stride_p): Likewise.
9537 * tree-loop-distribution.c (data_ref_segment_size): Return
9538 DR_STEP * (niters - 1).
9539 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9541 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9542 (vect_preserves_scalar_order_p): New function, split out from...
9543 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9544 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9545 (vect_vfa_access_size): New function.
9546 (vect_vfa_align): Likewise.
9547 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9548 (dump_lower_bound): New function.
9549 (vect_check_lower_bound): Likewise.
9550 (vect_small_gap_p): Likewise.
9551 (vectorizable_with_step_bound_p): Likewise.
9552 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9553 depencies if the vectorization factor is 1. Convert the checks
9554 for nonzero steps into checks on the bounds of DR_STEP. Try using
9555 a bunds check for variable steps if the minimum required step is
9556 relatively small. Update calls to the dr_with_seg_len
9557 constructor and to vect_compile_time_alias.
9558 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9560 (vect_loop_versioning): Call it.
9561 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9563 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9566 Alan Hayward <alan.hayward@arm.com>
9567 David Sherwood <david.sherwood@arm.com>
9569 * doc/sourcebuild.texi (vect_scatter_store): Document.
9570 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9572 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9574 * genopinit.c (main): Add supports_vec_scatter_store and
9575 supports_vec_scatter_store_cached to target_optabs.
9576 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9577 IFN_MASK_SCATTER_STORE.
9578 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9580 * internal-fn.h (internal_store_fn_p): Declare.
9581 (internal_fn_stored_value_index): Likewise.
9582 * internal-fn.c (scatter_store_direct): New macro.
9583 (expand_scatter_store_optab_fn): New function.
9584 (direct_scatter_store_optab_supported_p): New macro.
9585 (internal_store_fn_p): New function.
9586 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9587 IFN_MASK_SCATTER_STORE.
9588 (internal_fn_mask_index): Likewise.
9589 (internal_fn_stored_value_index): New function.
9590 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9592 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9593 * optabs-query.c (supports_vec_scatter_store_p): New function.
9594 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9595 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9596 true for scatter stores.
9597 (vect_gather_scatter_fn_p): Handle scatter stores too.
9598 (vect_check_gather_scatter): Consider using scatter stores if
9599 supports_vec_scatter_store_p.
9600 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9602 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9603 internal_fn_stored_value_index.
9604 (check_load_store_masking): Handle scatter stores too.
9605 (vect_get_store_rhs): Make public.
9606 (vectorizable_call): Use internal_store_fn_p.
9607 (vectorizable_store): Handle scatter store internal functions.
9608 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9609 when deciding whether the end of the group has been reached.
9610 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9611 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9612 (mask_scatter_store<mode>): New insns.
9614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9615 Alan Hayward <alan.hayward@arm.com>
9616 David Sherwood <david.sherwood@arm.com>
9618 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9619 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9620 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9622 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9623 Use vect_truncate_gather_scatter_offset if we can't treat the
9624 operation as a normal gather load or scatter store.
9625 (get_group_load_store_type): Take the gather_scatter_info
9626 as argument. Try using a gather load or scatter store for
9627 single-element groups.
9628 (get_load_store_type): Update calls to get_group_load_store_type
9629 and vect_use_strided_gather_scatters_p.
9631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9632 Alan Hayward <alan.hayward@arm.com>
9633 David Sherwood <david.sherwood@arm.com>
9635 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9636 optional tree argument.
9637 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9639 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9640 but continue to use the current value as a fallback.
9641 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9642 to compare the updates.
9643 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9644 (get_load_store_type): Use it when handling a strided access.
9645 (vect_get_strided_load_store_ops): New function.
9646 (vect_get_data_ptr_increment): Likewise.
9647 (vectorizable_load): Handle strided gather loads. Always pass
9648 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9650 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9651 Alan Hayward <alan.hayward@arm.com>
9652 David Sherwood <david.sherwood@arm.com>
9654 * doc/md.texi (gather_load@var{m}): Document.
9655 (mask_gather_load@var{m}): Likewise.
9656 * genopinit.c (main): Add supports_vec_gather_load and
9657 supports_vec_gather_load_cached to target_optabs.
9658 * optabs-tree.c (init_tree_optimization_optabs): Use
9659 ggc_cleared_alloc to allocate target_optabs.
9660 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9661 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9663 * internal-fn.h (internal_load_fn_p): Declare.
9664 (internal_gather_scatter_fn_p): Likewise.
9665 (internal_fn_mask_index): Likewise.
9666 (internal_gather_scatter_fn_supported_p): Likewise.
9667 * internal-fn.c (gather_load_direct): New macro.
9668 (expand_gather_load_optab_fn): New function.
9669 (direct_gather_load_optab_supported_p): New macro.
9670 (direct_internal_fn_optab): New function.
9671 (internal_load_fn_p): Likewise.
9672 (internal_gather_scatter_fn_p): Likewise.
9673 (internal_fn_mask_index): Likewise.
9674 (internal_gather_scatter_fn_supported_p): Likewise.
9675 * optabs-query.c (supports_at_least_one_mode_p): New function.
9676 (supports_vec_gather_load_p): Likewise.
9677 * optabs-query.h (supports_vec_gather_load_p): Declare.
9678 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9679 and memory_type field.
9680 (NUM_PATTERNS): Bump to 15.
9681 * tree-vect-data-refs.c: Include internal-fn.h.
9682 (vect_gather_scatter_fn_p): New function.
9683 (vect_describe_gather_scatter_call): Likewise.
9684 (vect_check_gather_scatter): Try using internal functions for
9685 gather loads. Recognize existing calls to a gather load function.
9686 (vect_analyze_data_refs): Consider using gather loads if
9687 supports_vec_gather_load_p.
9688 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9689 (vect_get_gather_scatter_offset_type): Likewise.
9690 (vect_convert_mask_for_vectype): Likewise.
9691 (vect_add_conversion_to_patterm): Likewise.
9692 (vect_try_gather_scatter_pattern): Likewise.
9693 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9694 (vect_vect_recog_func_ptrs): Add it.
9695 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9696 internal_fn_mask_index and internal_gather_scatter_fn_p.
9697 (check_load_store_masking): Take the gather_scatter_info as an
9698 argument and handle gather loads.
9699 (vect_get_gather_scatter_ops): New function.
9700 (vectorizable_call): Check internal_load_fn_p.
9701 (vectorizable_load): Likewise. Handle gather load internal
9703 (vectorizable_store): Update call to check_load_store_masking.
9704 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9705 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9706 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9707 (aarch64_gather_scale_operand_d): New predicates.
9708 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9709 (mask_gather_load<mode>): New insns.
9711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9712 Alan Hayward <alan.hayward@arm.com>
9713 David Sherwood <david.sherwood@arm.com>
9715 * optabs.def (fold_left_plus_optab): New optab.
9716 * doc/md.texi (fold_left_plus_@var{m}): Document.
9717 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9718 * internal-fn.c (fold_left_direct): Define.
9719 (expand_fold_left_optab_fn): Likewise.
9720 (direct_fold_left_optab_supported_p): Likewise.
9721 * fold-const-call.c (fold_const_fold_left): New function.
9722 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9723 * tree-parloops.c (valid_reduction_p): New function.
9724 (gather_scalar_reductions): Use it.
9725 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9726 (vect_finish_replace_stmt): Declare.
9727 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9728 (needs_fold_left_reduction_p): New function, split out from...
9729 (vect_is_simple_reduction): ...here. Accept reductions that
9730 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9731 (vect_force_simple_reduction): Also store the reduction type in
9732 the assignment's STMT_VINFO_REDUC_TYPE.
9733 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9734 (merge_with_identity): New function.
9735 (vect_expand_fold_left): Likewise.
9736 (vectorize_fold_left_reduction): Likewise.
9737 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9738 scalar phi in place for it. Check for target support and reject
9739 cases that would reassociate the operation. Defer the transform
9740 phase to vectorize_fold_left_reduction.
9741 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9742 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9743 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9745 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9747 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9748 call to ifc_temp_var.
9750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9751 Alan Hayward <alan.hayward@arm.com>
9752 David Sherwood <david.sherwood@arm.com>
9754 * target.def (legitimize_address_displacement): Take the original
9755 offset as a poly_int.
9756 * targhooks.h (default_legitimize_address_displacement): Update
9758 * targhooks.c (default_legitimize_address_displacement): Likewise.
9759 * doc/tm.texi: Regenerate.
9760 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9761 as an argument, moving assert of ad->disp == ad->disp_term to...
9762 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9763 Try calling targetm.legitimize_address_displacement before expanding
9764 the address rather than afterwards, and adjust for the new interface.
9765 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9766 Match the new hook interface. Handle SVE addresses.
9767 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9770 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9772 * Makefile.in (OBJS): Add early-remat.o.
9773 * target.def (select_early_remat_modes): New hook.
9774 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9775 * doc/tm.texi: Regenerate.
9776 * targhooks.h (default_select_early_remat_modes): Declare.
9777 * targhooks.c (default_select_early_remat_modes): New function.
9778 * timevar.def (TV_EARLY_REMAT): New timevar.
9779 * passes.def (pass_early_remat): New pass.
9780 * tree-pass.h (make_pass_early_remat): Declare.
9781 * early-remat.c: New file.
9782 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9784 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9787 Alan Hayward <alan.hayward@arm.com>
9788 David Sherwood <david.sherwood@arm.com>
9790 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9791 vfm1 with a bound_epilog parameter.
9792 (vect_do_peeling): Update calls accordingly, and move the prologue
9793 call earlier in the function. Treat the base bound_epilog as 0 for
9794 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9795 this base when peeling for gaps.
9796 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9797 with fully-masked loops.
9798 (vect_estimate_min_profitable_iters): Handle the single peeled
9799 iteration in that case.
9801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9802 Alan Hayward <alan.hayward@arm.com>
9803 David Sherwood <david.sherwood@arm.com>
9805 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9806 single-element interleaving even if the size is not a power of 2.
9807 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9808 accesses for single-element interleaving if the group size is
9811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9812 Alan Hayward <alan.hayward@arm.com>
9813 David Sherwood <david.sherwood@arm.com>
9815 * doc/md.texi (fold_extract_last_@var{m}): Document.
9816 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9817 * optabs.def (fold_extract_last_optab): New optab.
9818 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9819 * internal-fn.c (fold_extract_direct): New macro.
9820 (expand_fold_extract_optab_fn): Likewise.
9821 (direct_fold_extract_optab_supported_p): Likewise.
9822 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9823 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9824 EXTRACT_LAST_REDUCTION.
9825 (get_initial_def_for_reduction): Do not create an initial vector
9826 for EXTRACT_LAST_REDUCTION reductions.
9827 (vectorizable_reduction): Leave the scalar phi in place for
9828 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9829 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9830 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9831 transform phase to vectorizable_condition.
9832 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9834 (vect_finish_stmt_generation): ...here.
9835 (vect_finish_replace_stmt): New function.
9836 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9837 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9839 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9841 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9842 Alan Hayward <alan.hayward@arm.com>
9843 David Sherwood <david.sherwood@arm.com>
9845 * doc/md.texi (extract_last_@var{m}): Document.
9846 * optabs.def (extract_last_optab): New optab.
9847 * internal-fn.def (EXTRACT_LAST): New internal function.
9848 * internal-fn.c (cond_unary_direct): New macro.
9849 (expand_cond_unary_optab_fn): Likewise.
9850 (direct_cond_unary_optab_supported_p): Likewise.
9851 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9852 loops using EXTRACT_LAST.
9853 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9854 (extract_last_<mode>): ...this optab.
9855 (vec_extract<mode><Vel>): Update accordingly.
9857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9858 Alan Hayward <alan.hayward@arm.com>
9859 David Sherwood <david.sherwood@arm.com>
9861 * target.def (empty_mask_is_expensive): New hook.
9862 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9863 * doc/tm.texi: Regenerate.
9864 * targhooks.h (default_empty_mask_is_expensive): Declare.
9865 * targhooks.c (default_empty_mask_is_expensive): New function.
9866 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9867 if the target says that empty masks are expensive.
9868 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9870 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9873 Alan Hayward <alan.hayward@arm.com>
9874 David Sherwood <david.sherwood@arm.com>
9876 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9877 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9878 (vect_use_loop_mask_for_alignment_p): New function.
9879 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9880 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9881 niters_skip argument. Make sure that the first niters_skip elements
9882 of the first iteration are inactive.
9883 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9884 Update call to vect_set_loop_masks_directly.
9885 (get_misalign_in_elems): New function, split out from...
9886 (vect_gen_prolog_loop_niters): ...here.
9887 (vect_update_init_of_dr): Take a code argument that specifies whether
9888 the adjustment should be added or subtracted.
9889 (vect_update_init_of_drs): Likewise.
9890 (vect_prepare_for_masked_peels): New function.
9891 (vect_do_peeling): Skip prologue peeling if we're using a mask
9892 instead. Update call to vect_update_inits_of_drs.
9893 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9895 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9896 alignment. Do not include the number of peeled iterations in
9897 the minimum threshold in that case.
9898 (vectorizable_induction): Adjust the start value down by
9899 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9900 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9901 Take the number of skipped iterations into account when calculating
9903 * tree-vect-stmts.c (vect_gen_while_not): New function.
9905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9906 Alan Hayward <alan.hayward@arm.com>
9907 David Sherwood <david.sherwood@arm.com>
9909 * doc/sourcebuild.texi (vect_fully_masked): Document.
9910 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9912 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9914 (vect_analyze_loop_2): ...here. Don't check the vectorization
9915 factor against the number of loop iterations if the loop is
9918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9919 Alan Hayward <alan.hayward@arm.com>
9920 David Sherwood <david.sherwood@arm.com>
9922 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9923 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9924 (dump_groups): Update accordingly.
9925 (iv_use::mem_type): New member variable.
9926 (address_p): New function.
9927 (record_use): Add a mem_type argument and initialize the new
9929 (record_group_use): Add a mem_type argument. Use address_p.
9930 Remove obsolete null checks of base_object. Update call to record_use.
9931 (find_interesting_uses_op): Update call to record_group_use.
9932 (find_interesting_uses_cond): Likewise.
9933 (find_interesting_uses_address): Likewise.
9934 (get_mem_type_for_internal_fn): New function.
9935 (find_address_like_use): Likewise.
9936 (find_interesting_uses_stmt): Try find_address_like_use before
9937 calling find_interesting_uses_op.
9938 (addr_offset_valid_p): Use the iv mem_type field as the type
9939 of the addressed memory.
9940 (add_autoinc_candidates): Likewise.
9941 (get_address_cost): Likewise.
9942 (split_small_address_groups_p): Use address_p.
9943 (split_address_groups): Likewise.
9944 (add_iv_candidate_for_use): Likewise.
9945 (autoinc_possible_for_pair): Likewise.
9946 (rewrite_groups): Likewise.
9947 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9948 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9949 (get_alias_ptr_type_for_ptr_address): New function.
9950 (rewrite_use_address): Rewrite address uses in calls that were
9951 identified by find_address_like_use.
9953 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9954 Alan Hayward <alan.hayward@arm.com>
9955 David Sherwood <david.sherwood@arm.com>
9957 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9959 * gimple-expr.h (is_gimple_addressable: Likewise.
9960 * gimple-expr.c (is_gimple_address): Likewise.
9961 * internal-fn.c (expand_call_mem_ref): New function.
9962 (expand_mask_load_optab_fn): Use it.
9963 (expand_mask_store_optab_fn): Likewise.
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/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9970 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9971 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9972 (cond_umax@var{mode}): Document.
9973 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9974 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9975 (cond_umin_optab, cond_umax_optab): New optabs.
9976 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9977 (COND_IOR, COND_XOR): New internal functions.
9978 * internal-fn.h (get_conditional_internal_fn): Declare.
9979 * internal-fn.c (cond_binary_direct): New macro.
9980 (expand_cond_binary_optab_fn): Likewise.
9981 (direct_cond_binary_optab_supported_p): Likewise.
9982 (get_conditional_internal_fn): New function.
9983 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9984 Cope with reduction statements that are vectorized as calls rather
9986 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9987 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9988 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9989 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9990 (UNSPEC_COND_EOR): New unspecs.
9991 (optab): Add mappings for them.
9992 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9993 (sve_int_op, sve_fp_op): New int attributes.
9995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9996 Alan Hayward <alan.hayward@arm.com>
9997 David Sherwood <david.sherwood@arm.com>
9999 * optabs.def (while_ult_optab): New optab.
10000 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10001 * internal-fn.def (WHILE_ULT): New internal function.
10002 * internal-fn.h (direct_internal_fn_supported_p): New override
10003 that takes two types as argument.
10004 * internal-fn.c (while_direct): New macro.
10005 (expand_while_optab_fn): New function.
10006 (convert_optab_supported_p): Likewise.
10007 (direct_while_optab_supported_p): New macro.
10008 * wide-int.h (wi::udiv_ceil): New function.
10009 * tree-vectorizer.h (rgroup_masks): New structure.
10010 (vec_loop_masks): New typedef.
10011 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10012 and fully_masked_p.
10013 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10014 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10015 (vect_max_vf): New function.
10016 (slpeel_make_loop_iterate_ntimes): Delete.
10017 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10018 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10019 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10020 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10021 internal-fn.h, stor-layout.h and optabs-query.h.
10022 (vect_set_loop_mask): New function.
10023 (add_preheader_seq): Likewise.
10024 (add_header_seq): Likewise.
10025 (interleave_supported_p): Likewise.
10026 (vect_maybe_permute_loop_masks): Likewise.
10027 (vect_set_loop_masks_directly): Likewise.
10028 (vect_set_loop_condition_masked): Likewise.
10029 (vect_set_loop_condition_unmasked): New function, split out from
10030 slpeel_make_loop_iterate_ntimes.
10031 (slpeel_make_loop_iterate_ntimes): Rename to..
10032 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10033 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10034 (vect_do_peeling): Update call accordingly.
10035 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10037 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10038 mask_compare_type, can_fully_mask_p and fully_masked_p.
10039 (release_vec_loop_masks): New function.
10040 (_loop_vec_info): Use it to free the loop masks.
10041 (can_produce_all_loop_masks_p): New function.
10042 (vect_get_max_nscalars_per_iter): Likewise.
10043 (vect_verify_full_masking): Likewise.
10044 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10045 retries, and free the mask rgroups before retrying. Check loop-wide
10046 reasons for disallowing fully-masked loops. Make the final decision
10047 about whether use a fully-masked loop or not.
10048 (vect_estimate_min_profitable_iters): Do not assume that peeling
10049 for the number of iterations will be needed for fully-masked loops.
10050 (vectorizable_reduction): Disable fully-masked loops.
10051 (vectorizable_live_operation): Likewise.
10052 (vect_halve_mask_nunits): New function.
10053 (vect_double_mask_nunits): Likewise.
10054 (vect_record_loop_mask): Likewise.
10055 (vect_get_loop_mask): Likewise.
10056 (vect_transform_loop): Handle the case in which the final loop
10057 iteration might handle a partial vector. Call vect_set_loop_condition
10058 instead of slpeel_make_loop_iterate_ntimes.
10059 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10060 (check_load_store_masking): New function.
10061 (prepare_load_store_mask): Likewise.
10062 (vectorizable_store): Handle fully-masked loops.
10063 (vectorizable_load): Likewise.
10064 (supportable_widening_operation): Use vect_halve_mask_nunits for
10066 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10067 (vect_gen_while): New function.
10068 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10069 (aarch64_uqdec<mode>): New insn.
10071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10072 Alan Hayward <alan.hayward@arm.com>
10073 David Sherwood <david.sherwood@arm.com>
10075 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10076 (reduc_xor_scal_optab): New optabs.
10077 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10078 (reduc_xor_scal_@var{m}): Document.
10079 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10080 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10081 internal functions.
10082 * fold-const-call.c (fold_const_call): Handle them.
10083 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10084 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10085 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10086 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10087 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10088 (UNSPEC_XORV): New unspecs.
10089 (optab): Add entries for them.
10090 (BITWISEV): New int iterator.
10091 (bit_reduc_op): New int attributes.
10093 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10094 Alan Hayward <alan.hayward@arm.com>
10095 David Sherwood <david.sherwood@arm.com>
10097 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10098 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10099 * optabs.def (vec_shl_insert_optab): New optab.
10100 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10101 (duplicate_and_interleave): Likewise.
10102 * tree-vect-loop.c: Include internal-fn.h.
10103 (neutral_op_for_slp_reduction): New function, split out from
10104 get_initial_defs_for_reduction.
10105 (get_initial_def_for_reduction): Handle option 2 for variable-length
10106 vectors by loading the neutral value into a vector and then shifting
10107 the initial value into element 0.
10108 (get_initial_defs_for_reduction): Replace the code argument with
10109 the neutral value calculated by neutral_op_for_slp_reduction.
10110 Use gimple_build_vector for constant-length vectors.
10111 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10112 but the first group_size elements have a neutral value.
10113 Use duplicate_and_interleave otherwise.
10114 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10115 Update call to get_initial_defs_for_reduction. Handle SLP
10116 reductions for variable-length vectors by creating one vector
10117 result for each scalar result, with the elements associated
10118 with other scalar results stubbed out with the neutral value.
10119 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10120 Require IFN_VEC_SHL_INSERT for double reductions on
10121 variable-length vectors, or SLP reductions that have
10122 a neutral value. Require can_duplicate_and_interleave_p
10123 support for variable-length unchained SLP reductions if there
10124 is no neutral value, such as for MIN/MAX reductions. Also require
10125 the number of vector elements to be a multiple of the number of
10126 SLP statements when doing variable-length unchained SLP reductions.
10127 Update call to vect_create_epilog_for_reduction.
10128 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10129 and remove initial values.
10130 (duplicate_and_interleave): Make public.
10131 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10132 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10135 Alan Hayward <alan.hayward@arm.com>
10136 David Sherwood <david.sherwood@arm.com>
10138 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10139 (can_duplicate_and_interleave_p): New function.
10140 (vect_get_and_check_slp_defs): Take the vector of statements
10141 rather than just the current one. Remove excess parentheses.
10142 Restriction rejectinon of vect_constant_def and vect_external_def
10143 for variable-length vectors to boolean types, or types for which
10144 can_duplicate_and_interleave_p is false.
10145 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10146 (duplicate_and_interleave): New function.
10147 (vect_get_constant_vectors): Use gimple_build_vector for
10148 constant-length vectors and suitable variable-length constant
10149 vectors. Use duplicate_and_interleave for other variable-length
10150 vectors. Don't defer the update when inserting new statements.
10152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10153 Alan Hayward <alan.hayward@arm.com>
10154 David Sherwood <david.sherwood@arm.com>
10156 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10157 min_profitable_iters doesn't go negative.
10159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10160 Alan Hayward <alan.hayward@arm.com>
10161 David Sherwood <david.sherwood@arm.com>
10163 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10164 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10165 * optabs.def (vec_mask_load_lanes_optab): New optab.
10166 (vec_mask_store_lanes_optab): Likewise.
10167 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10168 (MASK_STORE_LANES): Likewise.
10169 * internal-fn.c (mask_load_lanes_direct): New macro.
10170 (mask_store_lanes_direct): Likewise.
10171 (expand_mask_load_optab_fn): Handle masked operations.
10172 (expand_mask_load_lanes_optab_fn): New macro.
10173 (expand_mask_store_optab_fn): Handle masked operations.
10174 (expand_mask_store_lanes_optab_fn): New macro.
10175 (direct_mask_load_lanes_optab_supported_p): Likewise.
10176 (direct_mask_store_lanes_optab_supported_p): Likewise.
10177 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10179 (vect_load_lanes_supported): Likewise.
10180 * tree-vect-data-refs.c (strip_conversion): New function.
10181 (can_group_stmts_p): Likewise.
10182 (vect_analyze_data_ref_accesses): Use it instead of checking
10183 for a pair of assignments.
10184 (vect_store_lanes_supported): Take a masked_p parameter.
10185 (vect_load_lanes_supported): Likewise.
10186 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10187 vect_store_lanes_supported and vect_load_lanes_supported.
10188 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10189 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10190 parameter. Don't allow gaps for masked accesses.
10191 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10192 and vect_load_lanes_supported.
10193 (get_load_store_type): Take a masked_p parameter and update
10194 call to get_group_load_store_type.
10195 (vectorizable_store): Update call to get_load_store_type.
10196 Handle IFN_MASK_STORE_LANES.
10197 (vectorizable_load): Update call to get_load_store_type.
10198 Handle IFN_MASK_LOAD_LANES.
10200 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10201 Alan Hayward <alan.hayward@arm.com>
10202 David Sherwood <david.sherwood@arm.com>
10204 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10206 * config/aarch64/aarch64-protos.h
10207 (aarch64_sve_struct_memory_operand_p): Declare.
10208 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10209 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10210 (VPRED, vpred): Handle SVE structure modes.
10211 * config/aarch64/constraints.md (Utx): New constraint.
10212 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10213 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10214 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10215 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10216 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10217 structure modes. Split into pieces after RA.
10218 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10219 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10221 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10222 SVE structure modes.
10223 (aarch64_classify_address): Likewise.
10224 (sizetochar): Move earlier in file.
10225 (aarch64_print_operand): Handle SVE register lists.
10226 (aarch64_array_mode): New function.
10227 (aarch64_sve_struct_memory_operand_p): Likewise.
10228 (TARGET_ARRAY_MODE): Redefine.
10230 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10231 Alan Hayward <alan.hayward@arm.com>
10232 David Sherwood <david.sherwood@arm.com>
10234 * target.def (array_mode): New target hook.
10235 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10236 * doc/tm.texi: Regenerate.
10237 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10238 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10239 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10240 targetm.array_mode.
10241 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10244 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10245 Alan Hayward <alan.hayward@arm.com>
10246 David Sherwood <david.sherwood@arm.com>
10248 * fold-const.c (fold_binary_loc): Check the argument types
10249 rather than the result type when testing for a vector operation.
10251 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10253 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10254 * doc/tm.texi: Regenerate.
10256 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10257 Alan Hayward <alan.hayward@arm.com>
10258 David Sherwood <david.sherwood@arm.com>
10260 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10261 (sve): Document new AArch64 extension.
10262 * doc/md.texi (w): Extend the description of the AArch64
10263 constraint to include SVE vectors.
10264 (Upl, Upa): Document new AArch64 predicate constraints.
10265 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10267 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10268 (msve-vector-bits=): New option.
10269 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10270 SVE when these are disabled.
10271 (sve): New extension.
10272 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10273 modes. Adjust their number of units based on aarch64_sve_vg.
10274 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10275 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10276 aarch64_addr_query_type.
10277 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10278 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10279 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10280 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10281 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10282 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10283 (aarch64_simd_imm_zero_p): Delete.
10284 (aarch64_check_zero_based_sve_index_immediate): Declare.
10285 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10286 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10287 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10288 (aarch64_sve_float_mul_immediate_p): Likewise.
10289 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10290 rather than an rtx.
10291 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10292 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10293 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10294 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10295 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10296 (aarch64_regmode_natural_size): Likewise.
10297 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10298 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10300 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10301 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10302 for VG and the SVE predicate registers.
10303 (V_ALIASES): Add a "z"-prefixed alias.
10304 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10305 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10306 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10307 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10308 (REG_CLASS_NAMES): Add entries for them.
10309 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10310 and the predicate registers.
10311 (aarch64_sve_vg): Declare.
10312 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10313 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10314 (REGMODE_NATURAL_SIZE): Define.
10315 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10317 * config/aarch64/aarch64.c: Include cfgrtl.h.
10318 (simd_immediate_info): Add a constructor for series vectors,
10319 and an associated step field.
10320 (aarch64_sve_vg): New variable.
10321 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10322 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10323 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10324 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10325 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10326 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10327 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10328 (aarch64_get_mask_mode): New functions.
10329 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10330 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10331 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10332 predicate modes and predicate registers. Explicitly restrict
10333 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10334 to store a vector mode if it is recognized by
10335 aarch64_classify_vector_mode.
10336 (aarch64_regmode_natural_size): New function.
10337 (aarch64_hard_regno_caller_save_mode): Return the original mode
10339 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10340 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10341 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10342 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10344 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10345 does not overlap dest if the function is frame-related. Handle
10347 (aarch64_split_add_offset): New function.
10348 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10349 them aarch64_add_offset.
10350 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10351 and update call to aarch64_sub_sp.
10352 (aarch64_add_cfa_expression): New function.
10353 (aarch64_expand_prologue): Pass extra temporary registers to the
10354 functions above. Handle the case in which we need to emit new
10355 DW_CFA_expressions for registers that were originally saved
10356 relative to the stack pointer, but now have to be expressed
10357 relative to the frame pointer.
10358 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10360 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10361 IP0 and IP1 values for SVE frames.
10362 (aarch64_expand_vec_series): New function.
10363 (aarch64_expand_sve_widened_duplicate): Likewise.
10364 (aarch64_expand_sve_const_vector): Likewise.
10365 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10366 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10367 into the register, rather than emitting a SET directly.
10368 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10369 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10370 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10371 (offset_9bit_signed_scaled_p): New functions.
10372 (aarch64_replicate_bitmask_imm): New function.
10373 (aarch64_bitmask_imm): Use it.
10374 (aarch64_cannot_force_const_mem): Reject expressions involving
10375 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10376 (aarch64_classify_index): Handle SVE indices, by requiring
10377 a plain register index with a scale that matches the element size.
10378 (aarch64_classify_address): Handle SVE addresses. Assert that
10379 the mode of the address is VOIDmode or an integer mode.
10380 Update call to aarch64_classify_symbol.
10381 (aarch64_classify_symbolic_expression): Update call to
10382 aarch64_classify_symbol.
10383 (aarch64_const_vec_all_in_range_p): New function.
10384 (aarch64_print_vector_float_operand): Likewise.
10385 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10386 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10387 and the FP immediates 1.0 and 0.5.
10388 (aarch64_print_address_internal): Handle SVE addresses.
10389 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10390 (aarch64_regno_regclass): Handle predicate registers.
10391 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10393 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10394 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10395 (aarch64_convert_sve_vector_bits): New function.
10396 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10397 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10398 rather than an rtx.
10399 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10400 Handle SVE vector and predicate modes. Accept VL-based constants
10401 that need only one temporary register, and VL offsets that require
10402 no temporary registers.
10403 (aarch64_conditional_register_usage): Mark the predicate registers
10404 as fixed if SVE isn't available.
10405 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10406 Return true for SVE vector and predicate modes.
10407 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10408 rather than an unsigned int. Handle SVE modes.
10409 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10411 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10413 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10414 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10415 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10416 (aarch64_sve_float_mul_immediate_p): New functions.
10417 (aarch64_sve_valid_immediate): New function.
10418 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10419 Explicitly reject structure modes. Check for INDEX constants.
10420 Handle PTRUE and PFALSE constants.
10421 (aarch64_check_zero_based_sve_index_immediate): New function.
10422 (aarch64_simd_imm_zero_p): Delete.
10423 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10424 vector modes. Accept constants in the range of CNT[BHWD].
10425 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10426 ask for an Advanced SIMD mode.
10427 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10428 (aarch64_simd_vector_alignment): Handle SVE predicates.
10429 (aarch64_vectorize_preferred_vector_alignment): New function.
10430 (aarch64_simd_vector_alignment_reachable): Use it instead of
10432 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10433 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10435 (MAX_VECT_LEN): Delete.
10436 (expand_vec_perm_d): Add a vec_flags field.
10437 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10438 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10439 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10441 (aarch64_evpc_rev): Rename to...
10442 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10443 (aarch64_evpc_rev_global): New function.
10444 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10445 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10447 (aarch64_evpc_sve_tbl): New function.
10448 (aarch64_expand_vec_perm_const_1): Update after rename of
10449 aarch64_evpc_rev. Handle SVE permutes too, trying
10450 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10451 than aarch64_evpc_tbl.
10452 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10453 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10454 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10455 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10456 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10457 (aarch64_expand_sve_vcond): New functions.
10458 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10459 of aarch64_vector_mode_p.
10460 (aarch64_dwarf_poly_indeterminate_value): New function.
10461 (aarch64_compute_pressure_classes): Likewise.
10462 (aarch64_can_change_mode_class): Likewise.
10463 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10464 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10465 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10466 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10467 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10468 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10469 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10470 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10472 (Dn, Dl, Dr): Accept const as well as const_vector.
10473 (Dz): Likewise. Compare against CONST0_RTX.
10474 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10475 of "vector" where appropriate.
10476 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10477 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10478 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10479 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10480 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10481 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10482 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10483 (v_int_equiv): Extend to SVE modes.
10484 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10486 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10487 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10488 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10489 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10490 (SVE_COND_FP_CMP): New int iterators.
10491 (perm_hilo): Handle the new unpack unspecs.
10492 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10494 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10495 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10496 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10497 (aarch64_equality_operator, aarch64_constant_vector_operand)
10498 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10499 (aarch64_sve_nonimmediate_operand): Likewise.
10500 (aarch64_sve_general_operand): Likewise.
10501 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10502 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10503 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10504 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10505 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10506 (aarch64_sve_float_arith_immediate): Likewise.
10507 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10508 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10509 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10510 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10511 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10512 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10513 (aarch64_sve_float_arith_operand): Likewise.
10514 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10515 (aarch64_sve_float_mul_operand): Likewise.
10516 (aarch64_sve_vec_perm_operand): Likewise.
10517 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10518 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10519 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10520 as well as const_vector.
10521 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10522 in file. Use CONST0_RTX and CONSTM1_RTX.
10523 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10524 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10525 Use aarch64_simd_imm_zero.
10526 * config/aarch64/aarch64-sve.md: New file.
10527 * config/aarch64/aarch64.md: Include it.
10528 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10529 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10530 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10531 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10532 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10533 (sve): New attribute.
10534 (enabled): Disable instructions with the sve attribute unless
10536 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10537 aarch64_expand_mov_immediate.
10538 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10539 CNT[BHSD] immediates.
10540 (movti): Split CONST_POLY_INT moves into two halves.
10541 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10542 Split additions that need a temporary here if the destination
10543 is the stack pointer.
10544 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10545 (*add<mode>3_poly_1): New instruction.
10546 (set_clobber_cc): New expander.
10548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10550 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10551 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10552 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10553 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10554 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10555 Change innermode from fixed_mode_size to machine_mode.
10556 (simplify_subreg): Update call accordingly. Handle a constant-sized
10557 subreg of a variable-length CONST_VECTOR.
10559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10560 Alan Hayward <alan.hayward@arm.com>
10561 David Sherwood <david.sherwood@arm.com>
10563 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10564 (add_offset_to_base): New function, split out from...
10565 (create_mem_ref): ...here. When handling a scale other than 1,
10566 check first whether the address is valid without the offset.
10567 Add it into the base if so, leaving the index and scale as-is.
10569 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10572 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10573 fold_for_warn before checking if arg2 is INTEGER_CST.
10575 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10577 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10578 (store_multiple_operation): Delete.
10579 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10580 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10581 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10582 guarded by TARGET_STRING.
10583 (rs6000_output_load_multiple): Delete.
10584 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10585 OPTION_MASK_STRING / TARGET_STRING handling.
10586 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10587 (const rs6000_opt_masks) <"string">: Change mask to 0.
10588 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10589 (MASK_STRING): Delete.
10590 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10592 (load_multiple): Delete.
10599 (store_multiple): Delete.
10606 (movmemsi_8reg): Delete.
10607 (corresponding unnamed define_insn): Delete.
10608 (movmemsi_6reg): Delete.
10609 (corresponding unnamed define_insn): Delete.
10610 (movmemsi_4reg): Delete.
10611 (corresponding unnamed define_insn): Delete.
10612 (movmemsi_2reg): Delete.
10613 (corresponding unnamed define_insn): Delete.
10614 (movmemsi_1reg): Delete.
10615 (corresponding unnamed define_insn): Delete.
10616 * config/rs6000/rs6000.opt (mno-string): New.
10617 (mstring): Replace by deprecation warning stub.
10618 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10620 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10622 * regrename.c (regrename_do_replace): If replacing the same
10623 reg multiple times, try to reuse last created gen_raw_REG.
10626 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10627 main to workaround a bug in GDB.
10629 2018-01-12 Tom de Vries <tom@codesourcery.com>
10632 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10634 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10636 PR rtl-optimization/80481
10637 * ira-color.c (get_cap_member): New function.
10638 (allocnos_conflict_by_live_ranges_p): Use it.
10639 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10640 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10642 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10645 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10646 (*saddl_se_1): Ditto.
10647 (*ssubsi_1): Ditto.
10648 (*ssubl_se_1): Ditto.
10650 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10652 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10653 rather than wi::to_widest for DR_INITs.
10654 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10655 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10656 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10658 (vect_analyze_group_access_1): Note that here.
10660 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10662 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10663 polynomial type sizes.
10665 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10667 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10668 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10669 (gimple_add_tmp_var): Likewise.
10671 2018-01-12 Martin Liska <mliska@suse.cz>
10673 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10674 (gimple_alloc_sizes): Likewise.
10675 (dump_gimple_statistics): Use PRIu64 in printf format.
10676 * gimple.h: Change uint64_t to int.
10678 2018-01-12 Martin Liska <mliska@suse.cz>
10680 * tree-core.h: Use uint64_t instead of int.
10681 * tree.c (tree_node_counts): Likewise.
10682 (tree_node_sizes): Likewise.
10683 (dump_tree_statistics): Use PRIu64 in printf format.
10685 2018-01-12 Martin Liska <mliska@suse.cz>
10687 * Makefile.in: As qsort_chk is implemented in vec.c, add
10688 vec.o to linkage of gencfn-macros.
10689 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10690 passing the info to record_node_allocation_statistics.
10691 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10693 * ggc-common.c (struct ggc_usage): Add operator== and use
10694 it in operator< and compare function.
10695 * mem-stats.h (struct mem_usage): Likewise.
10696 * vec.c (struct vec_usage): Remove operator< and compare
10697 function. Can be simply inherited.
10699 2018-01-12 Martin Jambor <mjambor@suse.cz>
10702 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10703 * tree-ssa-math-opts.c: Include domwalk.h.
10704 (convert_mult_to_fma_1): New function.
10705 (fma_transformation_info): New type.
10706 (fma_deferring_state): Likewise.
10707 (cancel_fma_deferring): New function.
10708 (result_of_phi): Likewise.
10709 (last_fma_candidate_feeds_initial_phi): Likewise.
10710 (convert_mult_to_fma): Added deferring logic, split actual
10711 transformation to convert_mult_to_fma_1.
10712 (math_opts_dom_walker): New type.
10713 (math_opts_dom_walker::after_dom_children): New method, body moved
10714 here from pass_optimize_widening_mul::execute, added deferring logic
10716 (pass_optimize_widening_mul::execute): Moved most of code to
10717 math_opts_dom_walker::after_dom_children.
10718 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10719 * config/i386/i386.c (ix86_option_override_internal): Added
10720 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10722 2018-01-12 Richard Biener <rguenther@suse.de>
10725 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10726 inline instance vars.
10728 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10731 * config/rx/rx.c (rx_is_restricted_memory_address):
10732 Handle SUBREG case.
10734 2018-01-12 Richard Biener <rguenther@suse.de>
10736 PR tree-optimization/80846
10737 * target.def (split_reduction): New target hook.
10738 * targhooks.c (default_split_reduction): New function.
10739 * targhooks.h (default_split_reduction): Declare.
10740 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10741 target requests first reduce vectors by combining low and high
10743 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10744 (get_vectype_for_scalar_type_and_size): Export.
10745 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10746 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10747 * doc/tm.texi: Regenerate.
10748 * config/i386/i386.c (ix86_split_reduction): Implement
10749 TARGET_VECTORIZE_SPLIT_REDUCTION.
10751 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10754 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10755 in PIC mode except for TARGET_VXWORKS_RTP.
10756 * config/sparc/sparc.c: Include cfgrtl.h.
10757 (TARGET_INIT_PIC_REG): Define.
10758 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10759 (sparc_pic_register_p): New predicate.
10760 (sparc_legitimate_address_p): Use it.
10761 (sparc_legitimize_pic_address): Likewise.
10762 (sparc_delegitimize_address): Likewise.
10763 (sparc_mode_dependent_address_p): Likewise.
10764 (gen_load_pcrel_sym): Remove 4th parameter.
10765 (load_got_register): Adjust call to above. Remove obsolete stuff.
10766 (sparc_expand_prologue): Do not call load_got_register here.
10767 (sparc_flat_expand_prologue): Likewise.
10768 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10769 (sparc_use_pseudo_pic_reg): New function.
10770 (sparc_init_pic_reg): Likewise.
10771 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10772 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10774 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10776 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10777 Add item for branch_cost.
10779 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10781 PR rtl-optimization/83565
10782 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10783 not extend the result to a larger mode for rotate operations.
10784 (num_sign_bit_copies1): Likewise.
10786 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10789 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10791 Use values-Xc.o for -pedantic.
10792 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10794 2018-01-12 Martin Liska <mliska@suse.cz>
10797 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10799 (possible_polymorphic_call_targets): Use it.
10800 (ipa_devirt): Likewise.
10802 2018-01-12 Martin Liska <mliska@suse.cz>
10804 * profile-count.h (enum profile_quality): Use 0 as invalid
10805 enum value of profile_quality.
10807 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10809 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10810 -mext-string options.
10812 2018-01-12 Richard Biener <rguenther@suse.de>
10814 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10815 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10816 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10818 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10820 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10822 * configure.ac (--with-long-double-format): Add support for the
10823 configuration option to change the default long double format on
10825 * config.gcc (powerpc*-linux*-*): Likewise.
10826 * configure: Regenerate.
10827 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10828 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10829 used without modification.
10831 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10833 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10834 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10835 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10836 MISC_BUILTIN_SPEC_BARRIER.
10837 (rs6000_init_builtins): Likewise.
10838 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10840 (speculation_barrier): New define_insn.
10841 * doc/extend.texi: Document __builtin_speculation_barrier.
10843 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10846 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10847 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10848 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10850 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10851 integral modes instead of "ss" and "sd".
10852 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10853 vectors with 32-bit and 64-bit elements.
10854 (vecdupssescalarmodesuffix): New mode attribute.
10855 (vec_dup<mode>): Use it.
10857 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10860 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10861 frame if argument is passed on stack.
10863 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10866 * ree.c (combine_reaching_defs): Optimize also
10867 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10868 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10870 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10872 PR middle-end/83189
10873 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10875 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10877 PR middle-end/83718
10878 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10879 after they are computed.
10881 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10883 PR tree-optimization/83695
10884 * gimple-loop-linterchange.cc
10885 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10886 reset cached scev information after interchange.
10887 (pass_linterchange::execute): Remove call to scev_reset_htab.
10889 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10891 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10892 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10893 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10894 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10895 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10896 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10897 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10898 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10899 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10900 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10901 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10902 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10903 (V_lane_reg): Likewise.
10904 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10906 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10907 (vfmal_lane_low<mode>_intrinsic,
10908 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10909 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10910 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10911 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10912 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10913 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10915 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10917 * config/arm/arm-cpus.in (fp16fml): New feature.
10918 (ALL_SIMD): Add fp16fml.
10919 (armv8.2-a): Add fp16fml as an option.
10920 (armv8.3-a): Likewise.
10921 (armv8.4-a): Add fp16fml as part of fp16.
10922 * config/arm/arm.h (TARGET_FP16FML): Define.
10923 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10925 * config/arm/arm-modes.def (V2HF): Define.
10926 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10927 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10928 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10929 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10930 vfmsl_low, vfmsl_high): New set of builtins.
10931 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10932 (vfml_op): New code attribute.
10933 (VFMLHALVES): New int iterator.
10934 (VFML, VFMLSEL): New mode attributes.
10935 (V_reg): Define mapping for V2HF.
10936 (V_hi, V_lo): New mode attributes.
10937 (VF_constraint): Likewise.
10938 (vfml_half, vfml_half_selector): New int attributes.
10939 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10941 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10942 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10944 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10945 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10946 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10947 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10949 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10950 Document new effective target and option set.
10952 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10954 * config/arm/arm-cpus.in (armv8_4): New feature.
10955 (ARMv8_4a): New fgroup.
10956 (armv8.4-a): New arch.
10957 * config/arm/arm-tables.opt: Regenerate.
10958 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10959 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10960 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10961 Add matching rules for -march=armv8.4-a and extensions.
10962 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10964 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10967 * config/rx/rx.md (BW): New mode attribute.
10968 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10970 2018-01-11 Richard Biener <rguenther@suse.de>
10972 PR tree-optimization/83435
10973 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10974 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10975 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10977 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10978 Alan Hayward <alan.hayward@arm.com>
10979 David Sherwood <david.sherwood@arm.com>
10981 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10983 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10984 (aarch64_print_address_internal): Use it to check for a zero offset.
10986 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10987 Alan Hayward <alan.hayward@arm.com>
10988 David Sherwood <david.sherwood@arm.com>
10990 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10991 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10992 Return a poly_int64 rather than a HOST_WIDE_INT.
10993 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10994 rather than a HOST_WIDE_INT.
10995 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10996 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10997 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10998 final_offset from HOST_WIDE_INT to poly_int64.
10999 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11000 to_constant when getting the number of units in an Advanced SIMD
11002 (aarch64_builtin_vectorized_function): Check for a constant number
11004 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11006 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11007 attribute instead of GET_MODE_NUNITS.
11008 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11009 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11010 GET_MODE_SIZE for fixed-size registers.
11011 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11012 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11013 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11014 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11015 (aarch64_print_operand, aarch64_print_address_internal)
11016 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11017 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11018 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11019 Handle polynomial GET_MODE_SIZE.
11020 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11021 wider than SImode without modification.
11022 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11023 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11024 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11025 passing and returning SVE modes.
11026 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11027 rather than GEN_INT.
11028 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11029 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11030 (aarch64_allocate_and_probe_stack_space): Likewise.
11031 (aarch64_layout_frame): Cope with polynomial offsets.
11032 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11033 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11034 polynomial offsets.
11035 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11036 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11037 poly_int64 rather than a HOST_WIDE_INT.
11038 (aarch64_get_separate_components, aarch64_process_components)
11039 (aarch64_expand_prologue, aarch64_expand_epilogue)
11040 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11041 (aarch64_anchor_offset): New function, split out from...
11042 (aarch64_legitimize_address): ...here.
11043 (aarch64_builtin_vectorization_cost): Handle polynomial
11044 TYPE_VECTOR_SUBPARTS.
11045 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11047 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11048 number of elements from the PARALLEL rather than the mode.
11049 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11050 rather than GET_MODE_BITSIZE.
11051 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11052 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11053 (aarch64_expand_vec_perm_const_1): Handle polynomial
11054 d->perm.length () and d->perm elements.
11055 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11056 Apply to_constant to d->perm elements.
11057 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11058 polynomial CONST_VECTOR_NUNITS.
11059 (aarch64_move_pointer): Take amount as a poly_int64 rather
11061 (aarch64_progress_pointer): Avoid temporary variable.
11062 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11063 the mode attribute instead of GET_MODE.
11065 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11066 Alan Hayward <alan.hayward@arm.com>
11067 David Sherwood <david.sherwood@arm.com>
11069 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11070 x exists before using it.
11071 (aarch64_add_constant_internal): Rename to...
11072 (aarch64_add_offset_1): ...this. Replace regnum with separate
11073 src and dest rtxes. Handle the case in which they're different,
11074 including when the offset is zero. Replace scratchreg with an rtx.
11075 Use 2 additions if there is no spare register into which we can
11076 move a 16-bit constant.
11077 (aarch64_add_constant): Delete.
11078 (aarch64_add_offset): Replace reg with separate src and dest
11079 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11080 Use aarch64_add_offset_1.
11081 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11082 an rtx rather than an int. Take the delta as a poly_int64
11083 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11084 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11085 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11086 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11087 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11088 and aarch64_add_sp.
11089 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11090 aarch64_add_constant.
11092 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11094 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11095 Use scalar_float_mode.
11097 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11099 * config/aarch64/aarch64-simd.md
11100 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11101 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11102 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11103 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11104 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11105 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11106 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11107 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11108 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11109 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11111 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11114 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11115 targ_options->x_arm_arch_string is non NULL.
11117 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11119 * config/aarch64/aarch64.h
11120 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11122 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11125 * expmed.c (emit_store_flag_force): Swap if const op0
11126 and change VOIDmode to mode of op0.
11128 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11130 PR rtl-optimization/83761
11131 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11132 than bytes to mode_for_size.
11134 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11136 PR middle-end/83189
11137 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11138 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11141 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11143 PR middle-end/83575
11144 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11145 when in layout mode.
11146 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11147 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11150 2018-01-10 Michael Collison <michael.collison@arm.com>
11152 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11153 * config/aarch64/aarch64-option-extension.def: Add
11154 AARCH64_OPT_EXTENSION of 'fp16fml'.
11155 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11156 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11157 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11158 * config/aarch64/constraints.md (Ui7): New constraint.
11159 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11160 (VFMLA_SEL_W): Ditto.
11163 (VFMLA16_LOW): New int iterator.
11164 (VFMLA16_HIGH): Ditto.
11165 (UNSPEC_FMLAL): New unspec.
11166 (UNSPEC_FMLSL): Ditto.
11167 (UNSPEC_FMLAL2): Ditto.
11168 (UNSPEC_FMLSL2): Ditto.
11169 (f16mac): New code attribute.
11170 * config/aarch64/aarch64-simd-builtins.def
11171 (aarch64_fmlal_lowv2sf): Ditto.
11172 (aarch64_fmlsl_lowv2sf): Ditto.
11173 (aarch64_fmlalq_lowv4sf): Ditto.
11174 (aarch64_fmlslq_lowv4sf): Ditto.
11175 (aarch64_fmlal_highv2sf): Ditto.
11176 (aarch64_fmlsl_highv2sf): Ditto.
11177 (aarch64_fmlalq_highv4sf): Ditto.
11178 (aarch64_fmlslq_highv4sf): Ditto.
11179 (aarch64_fmlal_lane_lowv2sf): Ditto.
11180 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11181 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11182 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11183 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11184 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11185 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11186 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11187 (aarch64_fmlal_lane_highv2sf): Ditto.
11188 (aarch64_fmlsl_lane_highv2sf): Ditto.
11189 (aarch64_fmlal_laneq_highv2sf): Ditto.
11190 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11191 (aarch64_fmlalq_lane_highv4sf): Ditto.
11192 (aarch64_fmlsl_lane_highv4sf): Ditto.
11193 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11194 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11195 * config/aarch64/aarch64-simd.md:
11196 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11197 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11198 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11199 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11200 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11201 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11202 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11203 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11204 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11205 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11206 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11207 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11208 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11209 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11210 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11211 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11212 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11213 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11214 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11215 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11216 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11217 (vfmlsl_low_u32): Ditto.
11218 (vfmlalq_low_u32): Ditto.
11219 (vfmlslq_low_u32): Ditto.
11220 (vfmlal_high_u32): Ditto.
11221 (vfmlsl_high_u32): Ditto.
11222 (vfmlalq_high_u32): Ditto.
11223 (vfmlslq_high_u32): Ditto.
11224 (vfmlal_lane_low_u32): Ditto.
11225 (vfmlsl_lane_low_u32): Ditto.
11226 (vfmlal_laneq_low_u32): Ditto.
11227 (vfmlsl_laneq_low_u32): Ditto.
11228 (vfmlalq_lane_low_u32): Ditto.
11229 (vfmlslq_lane_low_u32): Ditto.
11230 (vfmlalq_laneq_low_u32): Ditto.
11231 (vfmlslq_laneq_low_u32): Ditto.
11232 (vfmlal_lane_high_u32): Ditto.
11233 (vfmlsl_lane_high_u32): Ditto.
11234 (vfmlal_laneq_high_u32): Ditto.
11235 (vfmlsl_laneq_high_u32): Ditto.
11236 (vfmlalq_lane_high_u32): Ditto.
11237 (vfmlslq_lane_high_u32): Ditto.
11238 (vfmlalq_laneq_high_u32): Ditto.
11239 (vfmlslq_laneq_high_u32): Ditto.
11240 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11241 (AARCH64_FL_FOR_ARCH8_4): New.
11242 (AARCH64_ISA_F16FML): New ISA flag.
11243 (TARGET_F16FML): New feature flag for fp16fml.
11244 (doc/invoke.texi): Document new fp16fml option.
11246 2018-01-10 Michael Collison <michael.collison@arm.com>
11248 * config/aarch64/aarch64-builtins.c:
11249 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11250 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11251 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11252 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11253 (AARCH64_ISA_SHA3): New ISA flag.
11254 (TARGET_SHA3): New feature flag for sha3.
11255 * config/aarch64/iterators.md (sha512_op): New int attribute.
11256 (CRYPTO_SHA512): New int iterator.
11257 (UNSPEC_SHA512H): New unspec.
11258 (UNSPEC_SHA512H2): Ditto.
11259 (UNSPEC_SHA512SU0): Ditto.
11260 (UNSPEC_SHA512SU1): Ditto.
11261 * config/aarch64/aarch64-simd-builtins.def
11262 (aarch64_crypto_sha512hqv2di): New builtin.
11263 (aarch64_crypto_sha512h2qv2di): Ditto.
11264 (aarch64_crypto_sha512su0qv2di): Ditto.
11265 (aarch64_crypto_sha512su1qv2di): Ditto.
11266 (aarch64_eor3qv8hi): Ditto.
11267 (aarch64_rax1qv2di): Ditto.
11268 (aarch64_xarqv2di): Ditto.
11269 (aarch64_bcaxqv8hi): Ditto.
11270 * config/aarch64/aarch64-simd.md:
11271 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11272 (aarch64_crypto_sha512su0qv2di): Ditto.
11273 (aarch64_crypto_sha512su1qv2di): Ditto.
11274 (aarch64_eor3qv8hi): Ditto.
11275 (aarch64_rax1qv2di): Ditto.
11276 (aarch64_xarqv2di): Ditto.
11277 (aarch64_bcaxqv8hi): Ditto.
11278 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11279 (vsha512h2q_u64): Ditto.
11280 (vsha512su0q_u64): Ditto.
11281 (vsha512su1q_u64): Ditto.
11282 (veor3q_u16): Ditto.
11283 (vrax1q_u64): Ditto.
11284 (vxarq_u64): Ditto.
11285 (vbcaxq_u16): Ditto.
11286 * config/arm/types.md (crypto_sha512): New type attribute.
11287 (crypto_sha3): Ditto.
11288 (doc/invoke.texi): Document new sha3 option.
11290 2018-01-10 Michael Collison <michael.collison@arm.com>
11292 * config/aarch64/aarch64-builtins.c:
11293 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11294 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11295 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11296 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11297 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11298 (AARCH64_ISA_SM4): New ISA flag.
11299 (TARGET_SM4): New feature flag for sm4.
11300 * config/aarch64/aarch64-simd-builtins.def
11301 (aarch64_sm3ss1qv4si): Ditto.
11302 (aarch64_sm3tt1aq4si): Ditto.
11303 (aarch64_sm3tt1bq4si): Ditto.
11304 (aarch64_sm3tt2aq4si): Ditto.
11305 (aarch64_sm3tt2bq4si): Ditto.
11306 (aarch64_sm3partw1qv4si): Ditto.
11307 (aarch64_sm3partw2qv4si): Ditto.
11308 (aarch64_sm4eqv4si): Ditto.
11309 (aarch64_sm4ekeyqv4si): Ditto.
11310 * config/aarch64/aarch64-simd.md:
11311 (aarch64_sm3ss1qv4si): Ditto.
11312 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11313 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11314 (aarch64_sm4eqv4si): Ditto.
11315 (aarch64_sm4ekeyqv4si): Ditto.
11316 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11317 (sm3part_op): Ditto.
11318 (CRYPTO_SM3TT): Ditto.
11319 (CRYPTO_SM3PART): Ditto.
11320 (UNSPEC_SM3SS1): New unspec.
11321 (UNSPEC_SM3TT1A): Ditto.
11322 (UNSPEC_SM3TT1B): Ditto.
11323 (UNSPEC_SM3TT2A): Ditto.
11324 (UNSPEC_SM3TT2B): Ditto.
11325 (UNSPEC_SM3PARTW1): Ditto.
11326 (UNSPEC_SM3PARTW2): Ditto.
11327 (UNSPEC_SM4E): Ditto.
11328 (UNSPEC_SM4EKEY): Ditto.
11329 * config/aarch64/constraints.md (Ui2): New constraint.
11330 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11331 * config/arm/types.md (crypto_sm3): New type attribute.
11332 (crypto_sm4): Ditto.
11333 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11334 (vsm3tt1aq_u32): Ditto.
11335 (vsm3tt1bq_u32): Ditto.
11336 (vsm3tt2aq_u32): Ditto.
11337 (vsm3tt2bq_u32): Ditto.
11338 (vsm3partw1q_u32): Ditto.
11339 (vsm3partw2q_u32): Ditto.
11340 (vsm4eq_u32): Ditto.
11341 (vsm4ekeyq_u32): Ditto.
11342 (doc/invoke.texi): Document new sm4 option.
11344 2018-01-10 Michael Collison <michael.collison@arm.com>
11346 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11347 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11348 (AARCH64_FL_FOR_ARCH8_4): New.
11349 (AARCH64_FL_V8_4): New flag.
11350 (doc/invoke.texi): Document new armv8.4-a option.
11352 2018-01-10 Michael Collison <michael.collison@arm.com>
11354 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11355 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11356 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11357 * config/aarch64/aarch64-option-extension.def: Add
11358 AARCH64_OPT_EXTENSION of 'sha2'.
11359 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11360 (crypto): Disable sha2 and aes if crypto disabled.
11361 (crypto): Enable aes and sha2 if enabled.
11362 (simd): Disable sha2 and aes if simd disabled.
11363 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11365 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11366 (TARGET_SHA2): New feature flag for sha2.
11367 (TARGET_AES): New feature flag for aes.
11368 * config/aarch64/aarch64-simd.md:
11369 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11370 conditional on TARGET_AES.
11371 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11372 (aarch64_crypto_sha1hsi): Make pattern conditional
11374 (aarch64_crypto_sha1hv4si): Ditto.
11375 (aarch64_be_crypto_sha1hv4si): Ditto.
11376 (aarch64_crypto_sha1su1v4si): Ditto.
11377 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11378 (aarch64_crypto_sha1su0v4si): Ditto.
11379 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11380 (aarch64_crypto_sha256su0v4si): Ditto.
11381 (aarch64_crypto_sha256su1v4si): Ditto.
11382 (doc/invoke.texi): Document new aes and sha2 options.
11384 2018-01-10 Martin Sebor <msebor@redhat.com>
11386 PR tree-optimization/83781
11387 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11390 2018-01-11 Martin Sebor <msebor@gmail.com>
11391 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11393 PR tree-optimization/83501
11394 PR tree-optimization/81703
11396 * tree-ssa-strlen.c (get_string_cst): Rename...
11397 (get_string_len): ...to this. Handle global constants.
11398 (handle_char_store): Adjust.
11400 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11401 Jim Wilson <jimw@sifive.com>
11403 * config/riscv/riscv-protos.h (riscv_output_return): New.
11404 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11405 (riscv_attribute_table, riscv_output_return),
11406 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11407 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11408 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11410 (riscv_expand_prologue): Add early return for naked function.
11411 (riscv_expand_epilogue): Likewise.
11412 (riscv_function_ok_for_sibcall): Return false for naked function.
11413 (riscv_set_current_function): New.
11414 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11415 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11416 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11417 * doc/extend.texi (RISC-V Function Attributes): New.
11419 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11421 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11422 check for 128-bit long double before checking TCmode.
11423 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11424 128-bit long doubles before checking TFmode or TCmode.
11425 (FLOAT128_IBM_P): Likewise.
11427 2018-01-10 Martin Sebor <msebor@redhat.com>
11429 PR tree-optimization/83671
11430 * builtins.c (c_strlen): Unconditionally return zero for the empty
11432 Use -Warray-bounds for warnings.
11433 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11434 for non-constant array indices with COMPONENT_REF, arrays of
11435 arrays, and pointers to arrays.
11436 (gimple_fold_builtin_strlen): Determine and set length range for
11437 non-constant character arrays.
11439 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11441 PR middle-end/81897
11442 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11445 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11447 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11449 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11452 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11453 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11454 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11455 indexed_or_indirect_operand predicate.
11456 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11457 (*vsx_le_perm_load_v8hi): Likewise.
11458 (*vsx_le_perm_load_v16qi): Likewise.
11459 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11460 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11461 (*vsx_le_perm_store_v8hi): Likewise.
11462 (*vsx_le_perm_store_v16qi): Likewise.
11463 (eight unnamed splitters): Likewise.
11465 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11467 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11468 * config/rs6000/emmintrin.h: Likewise.
11469 * config/rs6000/mmintrin.h: Likewise.
11470 * config/rs6000/xmmintrin.h: Likewise.
11472 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11475 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11477 * tree.c (tree_nop_conversion): Return true for location wrapper
11479 (maybe_wrap_with_location): New function.
11480 (selftest::check_strip_nops): New function.
11481 (selftest::test_location_wrappers): New function.
11482 (selftest::tree_c_tests): Call it.
11483 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11484 (maybe_wrap_with_location): New decl.
11485 (EXPR_LOCATION_WRAPPER_P): New macro.
11486 (location_wrapper_p): New inline function.
11487 (tree_strip_any_location_wrapper): New inline function.
11489 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11492 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11493 stack_realign_offset for the largest alignment of stack slot
11495 (ix86_find_max_used_stack_alignment): New function.
11496 (ix86_finalize_stack_frame_flags): Use it. Set
11497 max_used_stack_alignment if we don't realign stack.
11498 * config/i386/i386.h (machine_function): Add
11499 max_used_stack_alignment.
11501 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11503 * config/arm/arm.opt (-mbranch-cost): New option.
11504 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11507 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11510 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11511 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11513 2018-01-10 Richard Biener <rguenther@suse.de>
11516 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11517 early out so it also covers the case where we have a non-NULL
11520 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11522 PR tree-optimization/83753
11523 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11524 for non-strided grouped accesses if the number of elements is 1.
11526 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11529 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11530 * i386.h (TARGET_USE_GATHER): Define.
11531 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11533 2018-01-10 Martin Liska <mliska@suse.cz>
11536 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11537 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11539 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11540 CLEANUP_NO_PARTITIONING is not set.
11542 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11544 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11545 for vectors, as a partial revert of r254296.
11546 * rtl.h (const_vec_p): Delete.
11547 (const_vec_duplicate_p): Don't test for vector CONSTs.
11548 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11549 * expmed.c (make_tree): Likewise.
11552 * common.md (E, F): Use CONSTANT_P instead of checking for
11554 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11555 checking for CONST_VECTOR.
11557 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11559 PR middle-end/83575
11560 * predict.c (force_edge_cold): Handle in more sane way edges
11561 with no prediction.
11563 2018-01-09 Carl Love <cel@us.ibm.com>
11565 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11567 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11568 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11569 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11570 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11571 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11572 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11573 * config/rs6000/rs6000-protos.h: Add extern defition for
11574 rs6000_generate_float2_double_code.
11575 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11577 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11578 (float2_v2df): Add define_expand.
11580 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11583 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11584 op_mode in the force_to_mode call.
11586 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11588 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11589 instead of checking each element individually.
11590 (aarch64_evpc_uzp): Likewise.
11591 (aarch64_evpc_zip): Likewise.
11592 (aarch64_evpc_ext): Likewise.
11593 (aarch64_evpc_rev): Likewise.
11594 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11595 instead of checking each element individually. Return true without
11597 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11598 whether all selected elements come from the same input, instead of
11599 checking each element individually. Remove calls to gen_rtx_REG,
11600 start_sequence and end_sequence and instead assert that no rtl is
11603 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11605 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11606 order of HIGH and CONST checks.
11608 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11610 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11611 if the destination isn't an SSA_NAME.
11613 2018-01-09 Richard Biener <rguenther@suse.de>
11615 PR tree-optimization/83668
11616 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11618 (canonicalize_loop_form): ... here, renamed from ...
11619 (canonicalize_loop_closed_ssa_form): ... this and amended to
11620 swap successor edges for loop exit blocks to make us use
11621 the RPO order we need for initial schedule generation.
11623 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11625 PR tree-optimization/64811
11626 * match.pd: When optimizing comparisons with Inf, avoid
11627 introducing or losing exceptions from comparisons with NaN.
11629 2018-01-09 Martin Liska <mliska@suse.cz>
11632 * asan.c (shadow_mem_size): Add gcc_assert.
11634 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11636 Don't save registers in main().
11639 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11640 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11641 * config/avr/avr.c (avr_set_current_function): Don't error if
11642 naked, OS_task or OS_main are specified at the same time.
11643 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11645 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11647 * common/config/avr/avr-common.c (avr_option_optimization_table):
11648 Switch on -mmain-is-OS_task for optimizing compilations.
11650 2018-01-09 Richard Biener <rguenther@suse.de>
11652 PR tree-optimization/83572
11653 * graphite.c: Include cfganal.h.
11654 (graphite_transform_loops): Connect infinite loops to exit
11655 and remove fake edges at the end.
11657 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11659 * ipa-inline.c (edge_badness): Revert accidental checkin.
11661 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11664 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11665 symbols; not inline clones.
11667 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11670 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11671 hard registers. Formatting fixes.
11673 PR preprocessor/83722
11674 * gcc.c (try_generate_repro): Pass
11675 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11676 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11679 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11680 Kito Cheng <kito.cheng@gmail.com>
11682 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11683 (riscv_leaf_function_p): Delete.
11684 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11686 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11688 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11690 (do_ifelse): New function.
11691 (do_isel): New function.
11692 (do_sub3): New function.
11693 (do_add3): New function.
11694 (do_load_mask_compare): New function.
11695 (do_overlap_load_compare): New function.
11696 (expand_compare_loop): New function.
11697 (expand_block_compare): Call expand_compare_loop() when appropriate.
11698 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11699 option description.
11700 (-mblock-compare-inline-loop-limit): New option.
11702 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11705 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11706 Reverse order of second and third operands in first alternative.
11707 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11708 of first and second elements in UNSPEC_VPERMR vector.
11709 (altivec_expand_vec_perm_le): Likewise.
11711 2018-01-08 Jeff Law <law@redhat.com>
11713 PR rtl-optimizatin/81308
11714 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11715 (process_switch): If group_case_labels makes a change, then set
11717 (pass_convert_switch::execute): If a switch is converted, then
11718 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11720 PR rtl-optimization/81308
11721 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11724 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11726 PR target/83663 - Revert r255946
11727 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11728 generation for cases where splatting a value is not useful.
11729 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11730 across a vec_duplicate and a paradoxical subreg forming a vector
11731 mode to a vec_concat.
11733 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11735 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11736 -march=armv8.3-a variants.
11737 * config/arm/t-multilib: Likewise.
11738 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11740 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11742 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11744 (cceq_ior_compare_complement): Give it a name so I can use it, and
11745 change boolean_or_operator predicate to boolean_operator so it can
11746 be used to generate a crand.
11747 (eqne): New code iterator.
11748 (bd/bd_neg): New code_attrs.
11749 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11750 a single define_insn.
11751 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11752 decrement (bdnzt/bdnzf/bdzt/bdzf).
11753 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11754 with the new names of the branch decrement patterns, and added the
11755 names of the branch decrement conditional patterns.
11757 2018-01-08 Richard Biener <rguenther@suse.de>
11759 PR tree-optimization/83563
11760 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11763 2018-01-08 Richard Biener <rguenther@suse.de>
11765 PR middle-end/83713
11766 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11768 2018-01-08 Richard Biener <rguenther@suse.de>
11770 PR tree-optimization/83685
11771 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11772 references to abnormals.
11774 2018-01-08 Richard Biener <rguenther@suse.de>
11777 * dwarf2out.c (output_indirect_strings): Handle empty
11778 skeleton_debug_str_hash.
11779 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11781 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11783 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11784 (emit_store_direct): Likewise.
11785 (arc_trampoline_adjust_address): Likewise.
11786 (arc_asm_trampoline_template): New function.
11787 (arc_initialize_trampoline): Use asm_trampoline_template.
11788 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11789 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11790 * config/arc/arc.md (flush_icache): Delete pattern.
11792 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11794 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11795 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11798 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11801 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11802 by not USED_FOR_TARGET.
11803 (make_pass_resolve_sw_modes): Likewise.
11805 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11807 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11810 2018-01-08 Richard Biener <rguenther@suse.de>
11812 PR middle-end/83580
11813 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11815 2018-01-08 Richard Biener <rguenther@suse.de>
11817 PR middle-end/83517
11818 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11820 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11822 PR middle-end/81897
11823 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11824 basic blocks with a small number of successors.
11825 (convert_control_dep_chain_into_preds): Improve handling of
11827 (dump_predicates): Split apart into...
11828 (dump_pred_chain): ...here...
11829 (dump_pred_info): ...and here.
11830 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11831 (can_chain_union_be_invalidated_p): Improve check for invalidation
11833 (uninit_uses_cannot_happen): Avoid unnecessary if
11834 convert_control_dep_chain_into_preds yielded nothing.
11836 2018-01-06 Martin Sebor <msebor@redhat.com>
11838 PR tree-optimization/83640
11839 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11840 subtracting negative offset from size.
11841 (builtin_access::overlap): Adjust offset bounds of the access to fall
11842 within the size of the object if possible.
11844 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11846 PR rtl-optimization/83699
11847 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11848 extract_bit_field_as_subreg to cases in which the extracted
11849 value is also a vector.
11851 * lra-constraints.c (process_alt_operands): Test for the equivalence
11852 substitutions when detecting a possible reload cycle.
11854 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11857 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11858 by default if flag_selective_schedling{,2}. Formatting fixes.
11860 PR rtl-optimization/83682
11861 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11862 if it has non-VECTOR_MODE element mode.
11863 (vec_duplicate_p): Likewise.
11865 PR middle-end/83694
11866 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11867 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11869 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11872 * config/i386/i386-builtin.def
11873 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11874 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11875 Require also OPTION_MASK_ISA_AVX512F in addition to
11876 OPTION_MASK_ISA_GFNI.
11877 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11878 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11879 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11880 to OPTION_MASK_ISA_GFNI.
11881 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11882 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11883 OPTION_MASK_ISA_AVX512BW.
11884 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11885 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11886 addition to OPTION_MASK_ISA_GFNI.
11887 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11888 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11889 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11890 to OPTION_MASK_ISA_GFNI.
11891 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11892 a requirement for all ISAs rather than any of them with a few
11894 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11896 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11897 bitmasks to be enabled with 3 exceptions, instead of requiring any
11898 enabled ISA with lots of exceptions.
11899 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11900 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11901 Change avx512bw in isa attribute to avx512f.
11902 * config/i386/sgxintrin.h: Add license boilerplate.
11903 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11904 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11905 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11906 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11908 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11909 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11910 temporarily sse2 rather than sse if not enabled already.
11913 * config/i386/sse.md (VI248_VLBW): Rename to ...
11914 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11915 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11916 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11917 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11918 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11919 mode iterator instead of VI248_VLBW.
11921 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11923 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11924 (record_modified): Skip clobbers; add debug output.
11925 (param_change_prob): Use sreal frequencies.
11927 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11929 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11930 punt for user-aligned variables.
11932 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11934 * tree-chrec.c (chrec_contains_symbols): Return true for
11937 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11940 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11941 of (x|y) == x for BICS pattern.
11943 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11945 PR tree-optimization/83605
11946 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11947 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11950 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11952 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11953 * config/epiphany/rtems.h: New file.
11955 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11956 Uros Bizjak <ubizjak@gmail.com>
11959 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11960 QIreg_operand instead of register_operand predicate.
11961 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11962 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11963 comments instead of -fmitigate[-_]rop.
11965 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11968 * cgraphunit.c (symbol_table::compile): Switch to text_section
11969 before calling assembly_start debug hook.
11970 * run-rtl-passes.c (run_rtl_passes): Likewise.
11973 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11975 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11976 range_int_cst_p rather than !symbolic_range_p before calling
11977 extract_range_from_multiplicative_op_1.
11979 2018-01-04 Jeff Law <law@redhat.com>
11981 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11982 redundant test in assertion.
11984 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11986 * doc/rtl.texi: Document machine_mode wrapper classes.
11988 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11990 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11991 using tree_to_uhwi.
11993 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11995 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11996 the VEC_PERM_EXPR fold to fail.
11998 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12001 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12002 to switched_sections.
12004 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12007 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12008 test for d.testing.
12010 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12013 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12014 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12016 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12019 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12020 is BLKmode and bitpos not zero or mode change is needed.
12022 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12025 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12028 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12031 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12032 instead of MULT rtx. Update all corresponding splitters.
12033 (*saddl_se): Ditto.
12034 (*ssub<modesuffix>): Ditto.
12035 (*ssubl_se): Ditto.
12036 (*cmp_sadd_di): Update split patterns.
12037 (*cmp_sadd_si): Ditto.
12038 (*cmp_sadd_sidi): Ditto.
12039 (*cmp_ssub_di): Ditto.
12040 (*cmp_ssub_si): Ditto.
12041 (*cmp_ssub_sidi): Ditto.
12042 * config/alpha/predicates.md (const23_operand): New predicate.
12043 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12044 Look for ASHIFT, not MULT inner operand.
12045 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12047 2018-01-04 Martin Liska <mliska@suse.cz>
12049 PR gcov-profile/83669
12050 * gcov.c (output_intermediate_file): Add version to intermediate
12052 * doc/gcov.texi: Document new field 'version' in intermediate
12053 file format. Fix location of '-k' option of gcov command.
12055 2018-01-04 Martin Liska <mliska@suse.cz>
12058 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12060 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12062 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12064 2018-01-03 Martin Sebor <msebor@redhat.com>
12066 PR tree-optimization/83655
12067 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12068 checking calls with invalid arguments.
12070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12072 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12073 (vectorizable_mask_load_store): Delete.
12074 (vectorizable_call): Return false for masked loads and stores.
12075 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12076 instead of gimple_assign_rhs1.
12077 (vectorizable_load): Handle IFN_MASK_LOAD.
12078 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12082 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12084 (vectorizable_mask_load_store): ...here.
12085 (vectorizable_load): ...and here.
12087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12089 * tree-vect-stmts.c (vect_build_all_ones_mask)
12090 (vect_build_zero_merge_argument): New functions, split out from...
12091 (vectorizable_load): ...here.
12093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12095 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12097 (vectorizable_mask_load_store): ...here.
12098 (vectorizable_store): ...and here.
12100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12102 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12104 (vectorizable_mask_load_store): ...here.
12106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12108 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12109 (vect_model_store_cost): Take a vec_load_store_type instead of a
12111 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12112 (vect_model_store_cost): Take a vec_load_store_type instead of a
12114 (vectorizable_mask_load_store): Update accordingly.
12115 (vectorizable_store): Likewise.
12116 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12120 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12121 IFN_MASK_LOAD calls here rather than...
12122 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12125 Alan Hayward <alan.hayward@arm.com>
12126 David Sherwood <david.sherwood@arm.com>
12128 * expmed.c (extract_bit_field_1): For vector extracts,
12129 fall back to extract_bit_field_as_subreg if vec_extract
12132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12133 Alan Hayward <alan.hayward@arm.com>
12134 David Sherwood <david.sherwood@arm.com>
12136 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12137 they are variable or constant sized.
12138 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12139 slots for constant-sized data.
12141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12142 Alan Hayward <alan.hayward@arm.com>
12143 David Sherwood <david.sherwood@arm.com>
12145 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12146 handling COND_EXPRs with boolean comparisons, try to find a better
12147 basis for the mask type than the boolean itself.
12149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12151 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12152 is calculated and how it can be overridden.
12153 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12154 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12156 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12160 Alan Hayward <alan.hayward@arm.com>
12161 David Sherwood <david.sherwood@arm.com>
12163 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12164 Remove the mode argument.
12165 (aarch64_simd_valid_immediate): Remove the mode and inverse
12167 * config/aarch64/iterators.md (bitsize): New iterator.
12168 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12169 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12170 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12171 aarch64_simd_valid_immediate.
12172 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12173 (aarch64_reg_or_bic_imm): Likewise.
12174 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12175 with an insn_type enum and msl with a modifier_type enum.
12176 Replace element_width with a scalar_mode. Change the shift
12177 to unsigned int. Add constructors for scalar_float_mode and
12178 scalar_int_mode elements.
12179 (aarch64_vect_float_const_representable_p): Delete.
12180 (aarch64_can_const_movi_rtx_p)
12181 (aarch64_simd_scalar_immediate_valid_for_move)
12182 (aarch64_simd_make_constant): Update call to
12183 aarch64_simd_valid_immediate.
12184 (aarch64_advsimd_valid_immediate_hs): New function.
12185 (aarch64_advsimd_valid_immediate): Likewise.
12186 (aarch64_simd_valid_immediate): Remove mode and inverse
12187 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12188 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12189 and aarch64_float_const_representable_p on the result.
12190 (aarch64_output_simd_mov_immediate): Remove mode argument.
12191 Update call to aarch64_simd_valid_immediate and use of
12192 simd_immediate_info.
12193 (aarch64_output_scalar_simd_mov_immediate): Update call
12196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12197 Alan Hayward <alan.hayward@arm.com>
12198 David Sherwood <david.sherwood@arm.com>
12200 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12201 (mode_nunits): Likewise CONST_MODE_NUNITS.
12202 * machmode.def (ADJUST_NUNITS): Document.
12203 * genmodes.c (mode_data::need_nunits_adj): New field.
12204 (blank_mode): Update accordingly.
12205 (adj_nunits): New variable.
12206 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12208 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12209 listed in adj_nunits.
12210 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12211 listed in adj_nunits. Don't emit case statements for such modes.
12212 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12213 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12214 nothing if adj_nunits is nonnull.
12215 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12216 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12217 (emit_mode_fbit): Update use of print_maybe_const_decl.
12218 (emit_move_size): Likewise. Treat the array as non-const
12220 (emit_mode_adjustments): Handle adj_nunits.
12222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12224 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12225 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12226 (VECTOR_MODES): Use it.
12227 (make_vector_modes): Take the prefix as an argument.
12229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12230 Alan Hayward <alan.hayward@arm.com>
12231 David Sherwood <david.sherwood@arm.com>
12233 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12234 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12235 for MODE_VECTOR_BOOL.
12236 * machmode.def (VECTOR_BOOL_MODE): Document.
12237 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12238 (make_vector_bool_mode): New function.
12239 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12241 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12242 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12244 * stor-layout.c (int_mode_for_mode): Likewise.
12245 * tree.c (build_vector_type_for_mode): Likewise.
12246 * varasm.c (output_constant_pool_2): Likewise.
12247 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12248 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12249 for MODE_VECTOR_BOOL.
12250 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12251 of mode class checks.
12252 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12253 instead of a list of mode class checks.
12254 (expand_vector_scalar_condition): Likewise.
12255 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12258 Alan Hayward <alan.hayward@arm.com>
12259 David Sherwood <david.sherwood@arm.com>
12261 * machmode.h (mode_size): Change from unsigned short to
12263 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12264 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12265 or if measurement_type is not polynomial.
12266 (fixed_size_mode::includes_p): Check for constant-sized modes.
12267 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12268 return a poly_uint16 rather than an unsigned short.
12269 (emit_mode_size): Change the type of mode_size from unsigned short
12270 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12271 (emit_mode_adjustments): Cope with polynomial vector sizes.
12272 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12274 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12276 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12277 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12278 * caller-save.c (setup_save_areas): Likewise.
12279 (replace_reg_with_saved_mem): Likewise.
12280 * calls.c (emit_library_call_value_1): Likewise.
12281 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12282 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12283 (gen_lowpart_for_combine): Likewise.
12284 * convert.c (convert_to_integer_1): Likewise.
12285 * cse.c (equiv_constant, cse_insn): Likewise.
12286 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12287 (cselib_subst_to_values): Likewise.
12288 * dce.c (word_dce_process_block): Likewise.
12289 * df-problems.c (df_word_lr_mark_ref): Likewise.
12290 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12291 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12292 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12293 (rtl_for_decl_location): Likewise.
12294 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12295 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12296 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12297 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12298 (expand_expr_real_1): Likewise.
12299 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12300 (pad_below): Likewise.
12301 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12302 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12303 * ira.c (get_subreg_tracking_sizes): Likewise.
12304 * ira-build.c (ira_create_allocno_objects): Likewise.
12305 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12306 (ira_sort_regnos_for_alter_reg): Likewise.
12307 * ira-costs.c (record_operand_costs): Likewise.
12308 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12309 (resolve_simple_move): Likewise.
12310 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12311 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12312 (lra_constraints): Likewise.
12313 (CONST_POOL_OK_P): Reject variable-sized modes.
12314 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12315 (add_pseudo_to_slot, lra_spill): Likewise.
12316 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12317 * optabs-query.c (get_best_extraction_insn): Likewise.
12318 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12319 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12320 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12321 * recog.c (offsettable_address_addr_space_p): Likewise.
12322 * regcprop.c (maybe_mode_change): Likewise.
12323 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12324 * regrename.c (build_def_use): Likewise.
12325 * regstat.c (dump_reg_info): Likewise.
12326 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12327 (find_reloads, find_reloads_subreg_address): Likewise.
12328 * reload1.c (eliminate_regs_1): Likewise.
12329 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12330 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12331 (simplify_binary_operation_1, simplify_subreg): Likewise.
12332 * targhooks.c (default_function_arg_padding): Likewise.
12333 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12334 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12335 (verify_gimple_assign_ternary): Likewise.
12336 * tree-inline.c (estimate_move_cost): Likewise.
12337 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12338 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12339 (get_address_cost_ainc): Likewise.
12340 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12341 (vect_supportable_dr_alignment): Likewise.
12342 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12343 (vectorizable_reduction): Likewise.
12344 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12345 (vectorizable_operation, vectorizable_load): Likewise.
12346 * tree.c (build_same_sized_truth_vector_type): Likewise.
12347 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12348 * var-tracking.c (emit_note_insn_var_location): Likewise.
12349 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12350 (ADDR_VEC_ALIGN): Likewise.
12352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12353 Alan Hayward <alan.hayward@arm.com>
12354 David Sherwood <david.sherwood@arm.com>
12356 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12358 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12359 or if measurement_type is polynomial.
12360 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12361 * combine.c (make_extraction): Likewise.
12362 * dse.c (find_shift_sequence): Likewise.
12363 * dwarf2out.c (mem_loc_descriptor): Likewise.
12364 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12365 (extract_bit_field, extract_low_bits): Likewise.
12366 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12367 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12368 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12369 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12370 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12371 * reload.c (find_reloads): Likewise.
12372 * reload1.c (alter_reg): Likewise.
12373 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12374 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12375 * tree-if-conv.c (predicate_mem_writes): Likewise.
12376 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12377 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12378 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12379 * valtrack.c (dead_debug_insert_temp): Likewise.
12380 * varasm.c (mergeable_constant_section): Likewise.
12381 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12384 Alan Hayward <alan.hayward@arm.com>
12385 David Sherwood <david.sherwood@arm.com>
12387 * expr.c (expand_assignment): Cope with polynomial mode sizes
12388 when assigning to a CONCAT.
12390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12391 Alan Hayward <alan.hayward@arm.com>
12392 David Sherwood <david.sherwood@arm.com>
12394 * machmode.h (mode_precision): Change from unsigned short to
12396 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12398 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12399 or if measurement_type is not polynomial.
12400 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12401 in which the mode is already known to be a scalar_int_mode.
12402 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12403 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12405 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12406 for GET_MODE_PRECISION.
12407 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12408 for GET_MODE_PRECISION.
12409 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12411 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12412 (expand_field_assignment, make_extraction): Likewise.
12413 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12414 (get_last_value): Likewise.
12415 * convert.c (convert_to_integer_1): Likewise.
12416 * cse.c (cse_insn): Likewise.
12417 * expr.c (expand_expr_real_1): Likewise.
12418 * lra-constraints.c (simplify_operand_subreg): Likewise.
12419 * optabs-query.c (can_atomic_load_p): Likewise.
12420 * optabs.c (expand_atomic_load): Likewise.
12421 (expand_atomic_store): Likewise.
12422 * ree.c (combine_reaching_defs): Likewise.
12423 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12424 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12425 * tree.h (type_has_mode_precision_p): Likewise.
12426 * ubsan.c (instrument_si_overflow): Likewise.
12428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12429 Alan Hayward <alan.hayward@arm.com>
12430 David Sherwood <david.sherwood@arm.com>
12432 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12433 polynomial numbers of units.
12434 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12435 (valid_vector_subparts_p): New function.
12436 (build_vector_type): Remove temporary shim and take the number
12437 of units as a poly_uint64 rather than an int.
12438 (build_opaque_vector_type): Take the number of units as a
12439 poly_uint64 rather than an int.
12440 * tree.c (build_vector_from_ctor): Handle polynomial
12441 TYPE_VECTOR_SUBPARTS.
12442 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12443 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12444 (build_vector_from_val): If the number of units is variable,
12445 use build_vec_duplicate_cst for constant operands and
12446 VEC_DUPLICATE_EXPR otherwise.
12447 (make_vector_type): Remove temporary is_constant ().
12448 (build_vector_type, build_opaque_vector_type): Take the number of
12449 units as a poly_uint64 rather than an int.
12450 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12452 * cfgexpand.c (expand_debug_expr): Likewise.
12453 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12454 (store_constructor, expand_expr_real_1): Likewise.
12455 (const_scalar_mask_from_tree): Likewise.
12456 * fold-const-call.c (fold_const_reduction): Likewise.
12457 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12458 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12459 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12460 (fold_relational_const): Likewise.
12461 (native_interpret_vector): Likewise. Change the size from an
12462 int to an unsigned int.
12463 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12464 TYPE_VECTOR_SUBPARTS.
12465 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12466 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12467 duplicating a non-constant operand into a variable-length vector.
12468 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12469 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12470 * ipa-icf.c (sem_variable::equals): Likewise.
12471 * match.pd: Likewise.
12472 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12473 * print-tree.c (print_node): Likewise.
12474 * stor-layout.c (layout_type): Likewise.
12475 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12476 * tree-cfg.c (verify_gimple_comparison): Likewise.
12477 (verify_gimple_assign_binary): Likewise.
12478 (verify_gimple_assign_ternary): Likewise.
12479 (verify_gimple_assign_single): Likewise.
12480 * tree-pretty-print.c (dump_generic_node): Likewise.
12481 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12482 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12483 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12484 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12485 (vect_shift_permute_load_chain): Likewise.
12486 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12487 (expand_vector_condition, optimize_vector_constructor): Likewise.
12488 (lower_vec_perm, get_compute_type): Likewise.
12489 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12490 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12491 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12492 (vect_recog_mask_conversion_pattern): Likewise.
12493 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12494 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12495 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12496 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12497 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12498 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12499 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12500 (supportable_widening_operation): Likewise.
12501 (supportable_narrowing_operation): Likewise.
12502 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12504 * varasm.c (output_constant): Likewise.
12506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12507 Alan Hayward <alan.hayward@arm.com>
12508 David Sherwood <david.sherwood@arm.com>
12510 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12511 so that both the length == 3 and length != 3 cases set up their
12512 own permute vectors. Add comments explaining why we know the
12513 number of elements is constant.
12514 (vect_permute_load_chain): Likewise.
12516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12517 Alan Hayward <alan.hayward@arm.com>
12518 David Sherwood <david.sherwood@arm.com>
12520 * machmode.h (mode_nunits): Change from unsigned char to
12522 (ONLY_FIXED_SIZE_MODES): New macro.
12523 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12524 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12525 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12527 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12528 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12529 or if measurement_type is not polynomial.
12530 * genmodes.c (ZERO_COEFFS): New macro.
12531 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12533 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12534 Use ZERO_COEFFS when emitting initializers.
12535 * data-streamer.h (bp_pack_poly_value): New function.
12536 (bp_unpack_poly_value): Likewise.
12537 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12538 for GET_MODE_NUNITS.
12539 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12540 for GET_MODE_NUNITS.
12541 * tree.c (make_vector_type): Remove temporary shim and make
12542 the real function take the number of units as a poly_uint64
12543 rather than an int.
12544 (build_vector_type_for_mode): Handle polynomial nunits.
12545 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12546 * emit-rtl.c (const_vec_series_p_1): Likewise.
12547 (gen_rtx_CONST_VECTOR): Likewise.
12548 * fold-const.c (test_vec_duplicate_folding): Likewise.
12549 * genrecog.c (validate_pattern): Likewise.
12550 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12551 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12552 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12553 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12554 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12555 * rtlanal.c (subreg_get_info): Likewise.
12556 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12557 (vect_grouped_load_supported): Likewise.
12558 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12559 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12560 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12561 (simplify_const_unary_operation, simplify_binary_operation_1)
12562 (simplify_const_binary_operation, simplify_ternary_operation)
12563 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12564 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12565 instead of CONST_VECTOR_NUNITS.
12566 * varasm.c (output_constant_pool_2): Likewise.
12567 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12568 explicit-encoded elements in the XVEC for variable-length vectors.
12570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12572 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12575 Alan Hayward <alan.hayward@arm.com>
12576 David Sherwood <david.sherwood@arm.com>
12578 * coretypes.h (fixed_size_mode): Declare.
12579 (fixed_size_mode_pod): New typedef.
12580 * builtins.h (target_builtins::x_apply_args_mode)
12581 (target_builtins::x_apply_result_mode): Change type to
12582 fixed_size_mode_pod.
12583 * builtins.c (apply_args_size, apply_result_size, result_vector)
12584 (expand_builtin_apply_args_1, expand_builtin_apply)
12585 (expand_builtin_return): Update accordingly.
12587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12589 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12590 * cselib.c (cselib_hash_rtx): Likewise.
12591 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12592 CONST_VECTOR encoding.
12594 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12595 Jeff Law <law@redhat.com>
12598 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12599 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12600 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12601 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12604 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12605 explicitly probe *sp in a noreturn function if there were any callee
12606 register saves or frame pointer is needed.
12608 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12611 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12612 BLKmode for ternary, binary or unary expressions.
12615 * var-tracking.c (delete_vta_debug_insn): New inline function.
12616 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12617 insns from get_insns () to NULL instead of each bb separately.
12618 Use delete_vta_debug_insn. No longer static.
12619 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12620 delete_vta_debug_insns callers.
12621 * rtl.h (delete_vta_debug_insns): Declare.
12622 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12623 instead of variable_tracking_main.
12625 2018-01-03 Martin Sebor <msebor@redhat.com>
12627 PR tree-optimization/83603
12628 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12629 arguments past the endof the argument list in functions declared
12630 without a prototype.
12631 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12632 Avoid checking when arguments are null.
12634 2018-01-03 Martin Sebor <msebor@redhat.com>
12637 * doc/extend.texi (attribute const): Fix a typo.
12638 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12639 issuing -Wsuggest-attribute for void functions.
12641 2018-01-03 Martin Sebor <msebor@redhat.com>
12643 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12644 offset_int::from instead of wide_int::to_shwi.
12645 (maybe_diag_overlap): Remove assertion.
12646 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12647 * gimple-ssa-sprintf.c (format_directive): Same.
12648 (parse_directive): Same.
12649 (sprintf_dom_walker::compute_format_length): Same.
12650 (try_substitute_return_value): Same.
12652 2018-01-03 Jeff Law <law@redhat.com>
12654 PR middle-end/83654
12655 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12656 non-constant residual for zero at runtime and avoid probing in
12657 that case. Reorganize code for trailing problem to mirror handling
12660 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12662 PR tree-optimization/83501
12663 * tree-ssa-strlen.c (get_string_cst): New.
12664 (handle_char_store): Call get_string_cst.
12666 2018-01-03 Martin Liska <mliska@suse.cz>
12668 PR tree-optimization/83593
12669 * tree-ssa-strlen.c: Include tree-cfg.h.
12670 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12671 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12672 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12674 (strlen_dom_walker::before_dom_children): Call
12675 gimple_purge_dead_eh_edges. Dump tranformation with details
12677 (strlen_dom_walker::before_dom_children): Update call by adding
12678 new argument cleanup_eh.
12679 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12681 2018-01-03 Martin Liska <mliska@suse.cz>
12684 * cif-code.def (VARIADIC_THUNK): New enum value.
12685 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12688 2018-01-03 Jan Beulich <jbeulich@suse.com>
12690 * sse.md (mov<mode>_internal): Tighten condition for when to use
12691 vmovdqu<ssescalarsize> for TI and OI modes.
12693 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12695 Update copyright years.
12697 2018-01-03 Martin Liska <mliska@suse.cz>
12700 * ipa-visibility.c (function_and_variable_visibility): Skip
12701 functions with noipa attribure.
12703 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12705 * gcc.c (process_command): Update copyright notice dates.
12706 * gcov-dump.c (print_version): Ditto.
12707 * gcov.c (print_version): Ditto.
12708 * gcov-tool.c (print_version): Ditto.
12709 * gengtype.c (create_file): Ditto.
12710 * doc/cpp.texi: Bump @copying's copyright year.
12711 * doc/cppinternals.texi: Ditto.
12712 * doc/gcc.texi: Ditto.
12713 * doc/gccint.texi: Ditto.
12714 * doc/gcov.texi: Ditto.
12715 * doc/install.texi: Ditto.
12716 * doc/invoke.texi: Ditto.
12718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12720 * vector-builder.h (vector_builder::m_full_nelts): Change from
12721 unsigned int to poly_uint64.
12722 (vector_builder::full_nelts): Update prototype accordingly.
12723 (vector_builder::new_vector): Likewise.
12724 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12725 (vector_builder::operator ==): Likewise.
12726 (vector_builder::finalize): Likewise.
12727 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12728 Take the number of elements as a poly_uint64 rather than an
12730 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12731 from unsigned int to poly_uint64.
12732 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12733 (vec_perm_indices::new_vector): Likewise.
12734 (vec_perm_indices::length): Likewise.
12735 (vec_perm_indices::nelts_per_input): Likewise.
12736 (vec_perm_indices::input_nelts): Likewise.
12737 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12738 number of elements per input as a poly_uint64 rather than an
12739 unsigned int. Use the original encoding for variable-length
12740 vectors, rather than clamping each individual element.
12741 For the second and subsequent elements in each pattern,
12742 clamp the step and base before clamping their sum.
12743 (vec_perm_indices::series_p): Handle polynomial element counts.
12744 (vec_perm_indices::all_in_range_p): Likewise.
12745 (vec_perm_indices_to_tree): Likewise.
12746 (vec_perm_indices_to_rtx): Likewise.
12747 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12748 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12749 (tree_vector_builder::new_binary_operation): Handle polynomial
12750 element counts. Return false if we need to know the number
12751 of elements at compile time.
12752 * fold-const.c (fold_vec_perm): Punt if the number of elements
12753 isn't known at compile time.
12755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12757 * vec-perm-indices.h (vec_perm_builder): Change element type
12758 from HOST_WIDE_INT to poly_int64.
12759 (vec_perm_indices::element_type): Update accordingly.
12760 (vec_perm_indices::clamp): Handle polynomial element_types.
12761 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12762 (vec_perm_indices::all_in_range_p): Likewise.
12763 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12765 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12766 polynomial vec_perm_indices element types.
12767 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12768 * fold-const.c (fold_vec_perm): Likewise.
12769 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12770 * tree-vect-generic.c (lower_vec_perm): Likewise.
12771 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12772 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12773 element type to HOST_WIDE_INT.
12775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12776 Alan Hayward <alan.hayward@arm.com>
12777 David Sherwood <david.sherwood@arm.com>
12779 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12780 rather than an int. Use plus_constant.
12781 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12782 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12785 Alan Hayward <alan.hayward@arm.com>
12786 David Sherwood <david.sherwood@arm.com>
12788 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12789 a HOST_WIDE_INT to a poly_int64.
12791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12792 Alan Hayward <alan.hayward@arm.com>
12793 David Sherwood <david.sherwood@arm.com>
12795 * calls.c (load_register_parameters): Cope with polynomial
12796 mode sizes. Require a constant size for BLKmode parameters
12797 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12798 forces a parameter to be padded at the lsb end in order to
12799 fill a complete number of words, require the parameter size
12800 to be ordered wrt UNITS_PER_WORD.
12802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12803 Alan Hayward <alan.hayward@arm.com>
12804 David Sherwood <david.sherwood@arm.com>
12806 * reload1.c (spill_stack_slot_width): Change element type
12807 from unsigned int to poly_uint64_pod.
12808 (alter_reg): Treat mode sizes as polynomial.
12810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12811 Alan Hayward <alan.hayward@arm.com>
12812 David Sherwood <david.sherwood@arm.com>
12814 * reload.c (complex_word_subreg_p): New function.
12815 (reload_inner_reg_of_subreg, push_reload): Use it.
12817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12818 Alan Hayward <alan.hayward@arm.com>
12819 David Sherwood <david.sherwood@arm.com>
12821 * lra-constraints.c (process_alt_operands): Reject matched
12822 operands whose sizes aren't ordered.
12823 (match_reload): Refer to this check here.
12825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12826 Alan Hayward <alan.hayward@arm.com>
12827 David Sherwood <david.sherwood@arm.com>
12829 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12830 that the mode size is in the set {1, 2, 4, 8, 16}.
12832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12833 Alan Hayward <alan.hayward@arm.com>
12834 David Sherwood <david.sherwood@arm.com>
12836 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12837 Use plus_constant instead of gen_rtx_PLUS.
12839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12840 Alan Hayward <alan.hayward@arm.com>
12841 David Sherwood <david.sherwood@arm.com>
12843 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12844 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12845 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12846 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12847 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12848 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12849 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12850 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12851 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12852 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12854 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12855 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12856 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12857 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12858 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12859 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12860 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12861 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12862 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12863 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12865 * expr.c (emit_move_resolve_push): Treat the input and result
12866 of PUSH_ROUNDING as a poly_int64.
12867 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12868 (emit_push_insn): Likewise.
12869 * lra-eliminations.c (mark_not_eliminable): Likewise.
12870 * recog.c (push_operand): Likewise.
12871 * reload1.c (elimination_effects): Likewise.
12872 * rtlanal.c (nonzero_bits1): Likewise.
12873 * calls.c (store_one_arg): Likewise. Require the padding to be
12874 known at compile time.
12876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12877 Alan Hayward <alan.hayward@arm.com>
12878 David Sherwood <david.sherwood@arm.com>
12880 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12881 Use plus_constant instead of gen_rtx_PLUS.
12883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12884 Alan Hayward <alan.hayward@arm.com>
12885 David Sherwood <david.sherwood@arm.com>
12887 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12888 rather than an int.
12890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12891 Alan Hayward <alan.hayward@arm.com>
12892 David Sherwood <david.sherwood@arm.com>
12894 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12895 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12896 via stack temporaries. Treat the mode size as polynomial too.
12898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12899 Alan Hayward <alan.hayward@arm.com>
12900 David Sherwood <david.sherwood@arm.com>
12902 * expr.c (expand_expr_real_2): When handling conversions involving
12903 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12904 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12905 as a poly_uint64 too.
12907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12908 Alan Hayward <alan.hayward@arm.com>
12909 David Sherwood <david.sherwood@arm.com>
12911 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12914 Alan Hayward <alan.hayward@arm.com>
12915 David Sherwood <david.sherwood@arm.com>
12917 * combine.c (can_change_dest_mode): Handle polynomial
12918 REGMODE_NATURAL_SIZE.
12919 * expmed.c (store_bit_field_1): Likewise.
12920 * expr.c (store_constructor): Likewise.
12921 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12922 and polynomial REGMODE_NATURAL_SIZE.
12923 (gen_lowpart_common): Likewise.
12924 * reginfo.c (record_subregs_of_mode): Likewise.
12925 * rtlanal.c (read_modify_subreg_p): Likewise.
12927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12928 Alan Hayward <alan.hayward@arm.com>
12929 David Sherwood <david.sherwood@arm.com>
12931 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12932 numbers of elements.
12934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12935 Alan Hayward <alan.hayward@arm.com>
12936 David Sherwood <david.sherwood@arm.com>
12938 * match.pd: Cope with polynomial numbers of vector elements.
12940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12941 Alan Hayward <alan.hayward@arm.com>
12942 David Sherwood <david.sherwood@arm.com>
12944 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12945 in a POINTER_PLUS_EXPR.
12947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12948 Alan Hayward <alan.hayward@arm.com>
12949 David Sherwood <david.sherwood@arm.com>
12951 * omp-simd-clone.c (simd_clone_subparts): New function.
12952 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12953 (ipa_simd_modify_function_body): Likewise.
12955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12956 Alan Hayward <alan.hayward@arm.com>
12957 David Sherwood <david.sherwood@arm.com>
12959 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12960 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12961 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12962 (expand_vector_condition, vector_element): Likewise.
12963 (subparts_gt): New function.
12964 (get_compute_type): Use subparts_gt.
12965 (count_type_subparts): Delete.
12966 (expand_vector_operations_1): Use subparts_gt instead of
12967 count_type_subparts.
12969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12970 Alan Hayward <alan.hayward@arm.com>
12971 David Sherwood <david.sherwood@arm.com>
12973 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12974 (vect_compile_time_alias): ...this new function. Do the calculation
12975 on poly_ints rather than trees.
12976 (vect_prune_runtime_alias_test_list): Update call accordingly.
12978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12979 Alan Hayward <alan.hayward@arm.com>
12980 David Sherwood <david.sherwood@arm.com>
12982 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12984 (vect_schedule_slp_instance): Likewise.
12986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12987 Alan Hayward <alan.hayward@arm.com>
12988 David Sherwood <david.sherwood@arm.com>
12990 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12991 constant and extern definitions for variable-length vectors.
12992 (vect_get_constant_vectors): Note that the number of units
12993 is known to be constant.
12995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12996 Alan Hayward <alan.hayward@arm.com>
12997 David Sherwood <david.sherwood@arm.com>
12999 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13000 of units as polynomial. Choose between WIDE and NARROW based
13003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13004 Alan Hayward <alan.hayward@arm.com>
13005 David Sherwood <david.sherwood@arm.com>
13007 * tree-vect-stmts.c (simd_clone_subparts): New function.
13008 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13011 Alan Hayward <alan.hayward@arm.com>
13012 David Sherwood <david.sherwood@arm.com>
13014 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13015 vectors as polynomial. Use build_index_vector for
13016 IFN_GOMP_SIMD_LANE.
13018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13019 Alan Hayward <alan.hayward@arm.com>
13020 David Sherwood <david.sherwood@arm.com>
13022 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13023 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13024 for variable-length vectors.
13025 (vectorizable_mask_load_store): Treat the number of units as
13026 polynomial, asserting that it is constant if the condition has
13027 already been enforced.
13028 (vectorizable_store, vectorizable_load): Likewise.
13030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13031 Alan Hayward <alan.hayward@arm.com>
13032 David Sherwood <david.sherwood@arm.com>
13034 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13035 of units as polynomial. Punt if we can't tell at compile time
13036 which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
13043 of units as polynomial. Punt on SLP inductions. Use an integer
13044 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13045 cast of such a series for variable-length floating-point
13048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13049 Alan Hayward <alan.hayward@arm.com>
13050 David Sherwood <david.sherwood@arm.com>
13052 * tree.h (build_index_vector): Declare.
13053 * tree.c (build_index_vector): New function.
13054 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13055 of units as polynomial, forcibly converting it to a constant if
13056 vectorizable_reduction has already enforced the condition.
13057 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13058 to create a {1,2,3,...} vector.
13059 (vectorizable_reduction): Treat the number of units as polynomial.
13060 Choose vectype_in based on the largest scalar element size rather
13061 than the smallest number of units. Enforce the restrictions
13064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13065 Alan Hayward <alan.hayward@arm.com>
13066 David Sherwood <david.sherwood@arm.com>
13068 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13069 number of units as polynomial.
13071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13072 Alan Hayward <alan.hayward@arm.com>
13073 David Sherwood <david.sherwood@arm.com>
13075 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13076 * target.def (autovectorize_vector_sizes): Return the vector sizes
13077 by pointer, using vector_sizes rather than a bitmask.
13078 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13079 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13080 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13082 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13083 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13084 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13085 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13086 * omp-general.c (omp_max_vf): Likewise.
13087 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13088 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13089 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13090 * tree-vect-slp.c (vect_slp_bb): Likewise.
13091 * doc/tm.texi: Regenerate.
13092 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13094 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13095 the vector size as a poly_uint64 rather than an unsigned int.
13096 (current_vector_size): Change from an unsigned int to a poly_uint64.
13097 (get_vectype_for_scalar_type): Update accordingly.
13098 * tree.h (build_truth_vector_type): Take the size and number of
13099 units as a poly_uint64 rather than an unsigned int.
13100 (build_vector_type): Add a temporary overload that takes
13101 the number of units as a poly_uint64 rather than an unsigned int.
13102 * tree.c (make_vector_type): Likewise.
13103 (build_truth_vector_type): Take the number of units as a poly_uint64
13104 rather than an unsigned int.
13106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13107 Alan Hayward <alan.hayward@arm.com>
13108 David Sherwood <david.sherwood@arm.com>
13110 * target.def (get_mask_mode): Take the number of units and length
13111 as poly_uint64s rather than unsigned ints.
13112 * targhooks.h (default_get_mask_mode): Update accordingly.
13113 * targhooks.c (default_get_mask_mode): Likewise.
13114 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13115 * doc/tm.texi: Regenerate.
13117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13118 Alan Hayward <alan.hayward@arm.com>
13119 David Sherwood <david.sherwood@arm.com>
13121 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13122 * omp-general.c (omp_max_vf): Likewise.
13123 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13124 (expand_omp_simd): Handle polynomial safelen.
13125 * omp-low.c (omplow_simd_context): Add a default constructor.
13126 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13127 (lower_rec_simd_input_clauses): Update accordingly.
13128 (lower_rec_input_clauses): Likewise.
13130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13131 Alan Hayward <alan.hayward@arm.com>
13132 David Sherwood <david.sherwood@arm.com>
13134 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13135 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13136 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13137 (vect_analyze_slp_cost): Likewise.
13138 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13139 (vect_model_load_cost): Likewise.
13141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13142 Alan Hayward <alan.hayward@arm.com>
13143 David Sherwood <david.sherwood@arm.com>
13145 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13146 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13147 from an unsigned int * to a poly_uint64_pod *.
13148 (calculate_unrolling_factor): New function.
13149 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13152 Alan Hayward <alan.hayward@arm.com>
13153 David Sherwood <david.sherwood@arm.com>
13155 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13156 from an unsigned int to a poly_uint64.
13157 (_loop_vec_info::slp_unrolling_factor): Likewise.
13158 (_loop_vec_info::vectorization_factor): Change from an int
13160 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13161 (vect_get_num_vectors): New function.
13162 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13163 (vect_get_num_copies): Use vect_get_num_vectors.
13164 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13165 to an unsigned int *.
13166 (vect_analyze_data_refs): Change min_vf from an int * to a
13168 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13169 than an unsigned HOST_WIDE_INT.
13170 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13171 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13172 to an unsigned int *.
13173 (vect_analyze_data_ref_dependences): Likewise.
13174 (vect_compute_data_ref_alignment): Handle polynomial vf.
13175 (vect_enhance_data_refs_alignment): Likewise.
13176 (vect_prune_runtime_alias_test_list): Likewise.
13177 (vect_shift_permute_load_chain): Likewise.
13178 (vect_supportable_dr_alignment): Likewise.
13179 (dependence_distance_ge_vf): Take the vectorization factor as a
13180 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13181 (vect_analyze_data_refs): Change min_vf from an int * to a
13183 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13184 vfm1 as a poly_uint64 rather than an int. Make the same change
13185 for the returned bound_scalar.
13186 (vect_gen_vector_loop_niters): Handle polynomial vf.
13187 (vect_do_peeling): Likewise. Update call to
13188 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13189 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13191 * tree-vect-loop.c (vect_determine_vectorization_factor)
13192 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13193 (vect_get_known_peeling_cost): Likewise.
13194 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13195 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13196 (vect_transform_loop): Likewise. Use the lowest possible VF when
13197 updating the upper bounds of the loop.
13198 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13199 rather than an int.
13200 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13201 polynomial unroll factors.
13202 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13203 (vect_make_slp_decision): Likewise.
13204 (vect_supported_load_permutation_p): Likewise, and polynomial
13206 (vect_analyze_slp_cost): Handle polynomial vf.
13207 (vect_slp_analyze_node_operations): Likewise.
13208 (vect_slp_analyze_bb_1): Likewise.
13209 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13210 than an unsigned HOST_WIDE_INT.
13211 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13212 (vectorizable_load): Handle polynomial vf.
13213 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13215 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13218 Alan Hayward <alan.hayward@arm.com>
13219 David Sherwood <david.sherwood@arm.com>
13221 * match.pd: Handle bit operations involving three constants
13222 and try to fold one pair.
13224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13226 * tree-vect-loop-manip.c: Include gimple-fold.h.
13227 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13228 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13229 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13230 Add a path that uses a step of VF instead of 1, but disable it
13232 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13233 and niters_no_overflow parameters. Update calls to
13234 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13235 Create a new SSA name if the latter choses to use a ste other
13236 than zero, and return it via niters_vector_mult_vf_var.
13237 * tree-vect-loop.c (vect_transform_loop): Update calls to
13238 vect_do_peeling, vect_gen_vector_loop_niters and
13239 slpeel_make_loop_iterate_ntimes.
13240 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13241 (vect_gen_vector_loop_niters): Update declarations after above changes.
13243 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13245 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13246 128-bit round to integer instructions.
13247 (ceil<mode>2): Likewise.
13248 (btrunc<mode>2): Likewise.
13249 (round<mode>2): Likewise.
13251 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13253 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13254 unaligned VSX load/store on P8/P9.
13255 (expand_block_clear): Allow the use of unaligned VSX
13256 load/store on P8/P9.
13258 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13260 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13262 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13263 swap associated with both a load and a store.
13265 2018-01-02 Andrew Waterman <andrew@sifive.com>
13267 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13268 * config/riscv/riscv.md (clear_cache): Use it.
13270 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13272 * web.c: Remove out-of-date comment.
13274 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13276 * expr.c (fixup_args_size_notes): Check that any existing
13277 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13278 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13279 (emit_single_push_insn): ...here.
13281 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13283 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13284 (const_vector_encoded_nelts): New function.
13285 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13286 (const_vector_int_elt, const_vector_elt): Declare.
13287 * emit-rtl.c (const_vector_int_elt_1): New function.
13288 (const_vector_elt): Likewise.
13289 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13290 of CONST_VECTOR_ELT.
13292 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13294 * expr.c: Include rtx-vector-builder.h.
13295 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13296 directly on the tree encoding.
13297 (const_vector_from_tree): Likewise.
13298 * optabs.c: Include rtx-vector-builder.h.
13299 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13300 sequence of "u" values.
13301 * vec-perm-indices.c: Include rtx-vector-builder.h.
13302 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13303 directly on the vec_perm_indices encoding.
13305 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13307 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13308 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13309 * rtx-vector-builder.h: New file.
13310 * rtx-vector-builder.c: Likewise.
13311 * rtl.h (rtx_def::u2): Add a const_vector field.
13312 (CONST_VECTOR_NPATTERNS): New macro.
13313 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13314 (CONST_VECTOR_DUPLICATE_P): Likewise.
13315 (CONST_VECTOR_STEPPED_P): Likewise.
13316 (CONST_VECTOR_ENCODED_ELT): Likewise.
13317 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13318 (unwrap_const_vec_duplicate): Likewise.
13319 (const_vec_series_p): Check for a non-duplicated vector encoding.
13320 Say that the function only returns true for integer vectors.
13321 * emit-rtl.c: Include rtx-vector-builder.h.
13322 (gen_const_vec_duplicate_1): Delete.
13323 (gen_const_vector): Call gen_const_vec_duplicate instead of
13324 gen_const_vec_duplicate_1.
13325 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13326 (gen_const_vec_duplicate): Use rtx_vector_builder.
13327 (gen_const_vec_series): Likewise.
13328 (gen_rtx_CONST_VECTOR): Likewise.
13329 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13330 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13331 Build a new vector rather than modifying a CONST_VECTOR in-place.
13332 (handle_special_swappables): Update call accordingly.
13333 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13334 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13335 Build a new vector rather than modifying a CONST_VECTOR in-place.
13336 (handle_special_swappables): Update call accordingly.
13338 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13340 * simplify-rtx.c (simplify_const_binary_operation): Use
13341 CONST_VECTOR_ELT instead of XVECEXP.
13343 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13345 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13346 the selector elements to be different from the data elements
13347 if the selector is a VECTOR_CST.
13348 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13349 ssizetype for the selector.
13351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13353 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13354 before testing each element individually.
13355 * tree-vect-generic.c (lower_vec_perm): Likewise.
13357 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13359 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13360 * selftest-run-tests.c (selftest::run_tests): Call it.
13361 * vector-builder.h (vector_builder::operator ==): New function.
13362 (vector_builder::operator !=): Likewise.
13363 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13364 (vec_perm_indices::all_from_input_p): New function.
13365 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13366 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13367 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13368 instead of reading the VECTOR_CST directly. Detect whether both
13369 vector inputs are the same before constructing the vec_perm_indices,
13370 and update the number of inputs argument accordingly. Use the
13371 utility functions added above. Only construct sel2 if we need to.
13373 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13375 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13376 the broadcast of the low byte.
13377 (expand_mult_highpart): Use an explicit encoding for the permutes.
13378 * optabs-query.c (can_mult_highpart_p): Likewise.
13379 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13380 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13381 (vectorizable_bswap): Likewise.
13382 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13383 explicit encoding for the power-of-2 permutes.
13384 (vect_permute_store_chain): Likewise.
13385 (vect_grouped_load_supported): Likewise.
13386 (vect_permute_load_chain): Likewise.
13388 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13390 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13391 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13392 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13393 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13394 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13395 (vect_gen_perm_mask_any): Likewise.
13397 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13399 * int-vector-builder.h: New file.
13400 * vec-perm-indices.h: Include int-vector-builder.h.
13401 (vec_perm_indices): Redefine as an int_vector_builder.
13402 (auto_vec_perm_indices): Delete.
13403 (vec_perm_builder): Redefine as a stand-alone class.
13404 (vec_perm_indices::vec_perm_indices): New function.
13405 (vec_perm_indices::clamp): Likewise.
13406 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13407 (vec_perm_indices::new_vector): New function.
13408 (vec_perm_indices::new_expanded_vector): Update for new
13409 vec_perm_indices class.
13410 (vec_perm_indices::rotate_inputs): New function.
13411 (vec_perm_indices::all_in_range_p): Operate directly on the
13412 encoded form, without computing elided elements.
13413 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13414 encoding. Update for new vec_perm_indices class.
13415 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13416 the given vec_perm_builder.
13417 (expand_vec_perm_var): Update vec_perm_builder constructor.
13418 (expand_mult_highpart): Use vec_perm_builder instead of
13419 auto_vec_perm_indices.
13420 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13421 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13422 or double series encoding as appropriate.
13423 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13424 vec_perm_indices instead of auto_vec_perm_indices.
13425 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13426 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13427 (vect_permute_store_chain): Likewise.
13428 (vect_grouped_load_supported): Likewise.
13429 (vect_permute_load_chain): Likewise.
13430 (vect_shift_permute_load_chain): Likewise.
13431 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13432 (vect_transform_slp_perm_load): Likewise.
13433 (vect_schedule_slp_instance): Likewise.
13434 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13435 (vectorizable_mask_load_store): Likewise.
13436 (vectorizable_bswap): Likewise.
13437 (vectorizable_store): Likewise.
13438 (vectorizable_load): Likewise.
13439 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13440 vec_perm_indices instead of auto_vec_perm_indices. Use
13441 tree_to_vec_perm_builder to read the vector from a tree.
13442 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13443 vec_perm_builder instead of a vec_perm_indices.
13444 (have_whole_vector_shift): Use vec_perm_builder and
13445 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13446 truncation to calc_vec_perm_mask_for_shift.
13447 (vect_create_epilog_for_reduction): Likewise.
13448 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13449 from auto_vec_perm_indices to vec_perm_indices.
13450 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13451 instead of changing individual elements.
13452 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13453 the vector in d.perm.
13454 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13455 from auto_vec_perm_indices to vec_perm_indices.
13456 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13457 instead of changing individual elements.
13458 (arm_vectorize_vec_perm_const): Use new_vector to install
13459 the vector in d.perm.
13460 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13461 Update vec_perm_builder constructor.
13462 (rs6000_expand_interleave): Likewise.
13463 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13464 (rs6000_expand_interleave): Likewise.
13466 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13468 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13469 to qimode could truncate the indices.
13470 * optabs.c (expand_vec_perm_var): Likewise.
13472 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13474 * Makefile.in (OBJS): Add vec-perm-indices.o.
13475 * vec-perm-indices.h: New file.
13476 * vec-perm-indices.c: Likewise.
13477 * target.h (vec_perm_indices): Replace with a forward class
13479 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13480 * optabs.h: Include vec-perm-indices.h.
13481 (expand_vec_perm): Delete.
13482 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13483 (expand_vec_perm_const): Declare.
13484 * target.def (vec_perm_const_ok): Replace with...
13485 (vec_perm_const): ...this new hook.
13486 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13487 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13488 * doc/tm.texi: Regenerate.
13489 * optabs.def (vec_perm_const): Delete.
13490 * doc/md.texi (vec_perm_const): Likewise.
13491 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13492 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13493 expand_vec_perm for constant permutation vectors. Assert that
13494 the mode of variable permutation vectors is the integer equivalent
13495 of the mode that is being permuted.
13496 * optabs-query.h (selector_fits_mode_p): Declare.
13497 * optabs-query.c: Include vec-perm-indices.h.
13498 (selector_fits_mode_p): New function.
13499 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13500 is defined, instead of checking whether the vec_perm_const_optab
13501 exists. Use targetm.vectorize.vec_perm_const instead of
13502 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13503 fit in the vector mode before using a variable permute.
13504 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13505 vec_perm_indices instead of an rtx.
13506 (expand_vec_perm): Replace with...
13507 (expand_vec_perm_const): ...this new function. Take the selector
13508 as a vec_perm_indices rather than an rtx. Also take the mode of
13509 the selector. Update call to shift_amt_for_vec_perm_mask.
13510 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13511 Use vec_perm_indices::new_expanded_vector to expand the original
13512 selector into bytes. Check whether the indices fit in the vector
13513 mode before using a variable permute.
13514 (expand_vec_perm_var): Make global.
13515 (expand_mult_highpart): Use expand_vec_perm_const.
13516 * fold-const.c: Includes vec-perm-indices.h.
13517 * tree-ssa-forwprop.c: Likewise.
13518 * tree-vect-data-refs.c: Likewise.
13519 * tree-vect-generic.c: Likewise.
13520 * tree-vect-loop.c: Likewise.
13521 * tree-vect-slp.c: Likewise.
13522 * tree-vect-stmts.c: Likewise.
13523 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13525 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13526 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13527 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13528 (aarch64_vectorize_vec_perm_const): ...this new function.
13529 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13530 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13531 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13532 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13533 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13534 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13535 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13537 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13538 check for NEON modes.
13539 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13540 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13541 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13542 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13544 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13545 the old VEC_PERM_CONST conditions.
13546 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13547 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13548 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13549 (ia64_vectorize_vec_perm_const_ok): Merge into...
13550 (ia64_vectorize_vec_perm_const): ...this new function.
13551 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13552 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13553 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13554 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13555 * config/mips/mips.c (mips_expand_vec_perm_const)
13556 (mips_vectorize_vec_perm_const_ok): Merge into...
13557 (mips_vectorize_vec_perm_const): ...this new function.
13558 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13559 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13560 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13561 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13562 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13563 (rs6000_expand_vec_perm_const): Delete.
13564 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13566 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13567 (altivec_expand_vec_perm_const_le): Take each operand individually.
13568 Operate on constant selectors rather than rtxes.
13569 (altivec_expand_vec_perm_const): Likewise. Update call to
13570 altivec_expand_vec_perm_const_le.
13571 (rs6000_expand_vec_perm_const): Delete.
13572 (rs6000_vectorize_vec_perm_const_ok): Delete.
13573 (rs6000_vectorize_vec_perm_const): New function.
13574 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13575 an element count and rtx array.
13576 (rs6000_expand_extract_even): Update call accordingly.
13577 (rs6000_expand_interleave): Likewise.
13578 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13579 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13580 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13581 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13582 (rs6000_expand_vec_perm_const): Delete.
13583 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13584 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13585 (altivec_expand_vec_perm_const_le): Take each operand individually.
13586 Operate on constant selectors rather than rtxes.
13587 (altivec_expand_vec_perm_const): Likewise. Update call to
13588 altivec_expand_vec_perm_const_le.
13589 (rs6000_expand_vec_perm_const): Delete.
13590 (rs6000_vectorize_vec_perm_const_ok): Delete.
13591 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13592 reference to the SPE evmerge intructions.
13593 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13594 an element count and rtx array.
13595 (rs6000_expand_extract_even): Update call accordingly.
13596 (rs6000_expand_interleave): Likewise.
13597 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13598 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13600 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13602 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13604 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13605 vector mode and that that mode matches the mode of the data
13607 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13608 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13609 directly using expand_vec_perm_1 when forcing selectors into
13611 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13613 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13615 * optabs-query.h (can_vec_perm_p): Delete.
13616 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13617 * optabs-query.c (can_vec_perm_p): Split into...
13618 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13619 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13620 particular selector is valid.
13621 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13622 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13623 (vect_grouped_load_supported): Likewise.
13624 (vect_shift_permute_load_chain): Likewise.
13625 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13626 (vect_transform_slp_perm_load): Likewise.
13627 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13628 (vectorizable_bswap): Likewise.
13629 (vect_gen_perm_mask_checked): Likewise.
13630 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13631 implementations of variable permutation vectors into account
13632 when deciding which selector to use.
13633 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13634 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13635 with a false third argument.
13636 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13637 to test whether the constant selector is valid and can_vec_perm_var_p
13638 to test whether a variable selector is valid.
13640 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13642 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13643 * optabs-query.c (can_vec_perm_p): Likewise.
13644 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13645 instead of vec_perm_indices.
13646 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13647 (vect_gen_perm_mask_checked): Likewise,
13648 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13649 (vect_gen_perm_mask_checked): Likewise,
13651 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13653 * optabs-query.h (qimode_for_vec_perm): Declare.
13654 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13655 (qimode_for_vec_perm): ...this new function.
13656 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13658 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13660 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13661 does not have a conditional at the top.
13663 2018-01-02 Richard Biener <rguenther@suse.de>
13665 * ipa-inline.c (big_speedup_p): Fix expression.
13667 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13670 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13673 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13677 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13678 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13679 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13680 cond_taken_branch_cost 3->4.
13682 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13684 PR tree-optimization/83581
13685 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13686 TODO_cleanup_cfg if any changes have been made.
13688 PR middle-end/83608
13689 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13690 convert_modes if target mode has the right side, but different mode
13693 PR middle-end/83609
13694 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13695 last argument when extracting from CONCAT. If either from_real or
13696 from_imag is NULL, use expansion through memory. If result is not
13697 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13698 the parts directly to inner mode, if even that fails, use expansion
13701 PR middle-end/83623
13702 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13703 check for bswap in mode rather than HImode and use that in expand_unop
13706 Copyright (C) 2018 Free Software Foundation, Inc.
13708 Copying and distribution of this file, with or without modification,
13709 are permitted in any medium without royalty provided the copyright
13710 notice and this notice are preserved.