1 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
5 define __ibm128 as long double.
6 * config/rs6000/rs6000.c (rs6000_init_builtins): Always create
7 __ibm128 as a distinct type.
8 (init_float128_ieee): Fix up conversions between IFmode and IEEE
9 128-bit types to use the correct functions.
10 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
11 convert between 128-bit floating point types that have different
12 modes but the same representation, instead of using gen_lowpart to
14 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
16 (IFKF_reg): New attributes to give the register constraints for
18 (extend<mode>tf2_internal): New insns to mark an explicit
19 conversion between 128-bit floating point types that have a
20 different mode but share the same representation.
22 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
24 PR tree-optimization/85814
25 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
26 a null return from get_strinfo when unsharing the next
29 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
32 * varasm.c (weak_finish): Clean up weak_decls.
34 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
36 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
37 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
39 * config/aarch64/iterators.md (ABAL): New int iterator.
42 (sur): Add mappings for the above.
43 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
45 (aarch64_<sur>abal<mode>_4): Likewise.
46 (aarch64_<sur>adalp<mode>_3): Likewise.
47 (<sur>sadv16qi): New define_expand.
49 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
51 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
52 (*movdf_internal): Ditto.
54 (*rsqrtsf2_sse): Ditto.
55 (*sqrt<mode>2_sse): Ditto.
57 2018-05-21 Tamar Christina <tamar.christina@arm.com>
59 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
61 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
62 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
63 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
64 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
66 * config/aarch64/arm_neon.h: Likewise.
67 * config/aarch64/iterators.md (VQ_I): New.
69 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
71 * config.gcc: Add arc/t-multilib-linux to tmake_file for
73 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
76 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
78 * config/nds32/constraints.md (S): New constraint.
79 * config/nds32/nds32.md (call_internal): Use constraint S.
80 (call_value_internal): Likewise.
81 (sibcall_internal): Likewise.
82 (sibcall_value_internal): Likewise.
84 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
85 Chung-Ju Wu <jasonwucj@gmail.com>
87 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
90 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
91 Chung-Ju Wu <jasonwucj@gmail.com>
93 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
94 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
95 (nds32_rtx_costs_impl): Simplify.
96 (nds32_address_cost_impl): Simplify.
97 (nds32_init_rtx_costs): New function.
98 (nds32_rtx_costs_speed_prefer): Likewise.
99 (nds32_rtx_costs_size_prefer): Likewise.
100 (nds32_address_cost_speed_prefer): Likewise.
101 (nds32_address_cost_speed_fwprop): Likewise.
102 (nds32_address_cost_size_prefer): Likewise.
103 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
104 * config/nds32/nds32.c (nds32_option_override): Use
105 nds32_init_rtx_costs function.
107 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
109 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
110 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
111 (TARGET_PIPELINE_N8): Likewise.
112 (TARGET_PIPELINE_N10): Likewise.
113 (TARGET_PIPELINE_N13): Likewise.
114 (TARGET_PIPELINE_GRAYWOLF): Likewise.
116 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
118 * config/nds32/nds32-fpu.md: Update copyright year.
120 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
122 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
124 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
126 * config/nds32/nds32.c
127 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
128 * config/nds32/nds32.opt (minline-asm-r15): New option.
130 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
132 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
134 * config/nds32/nds32.md (abssi2): New pattern.
136 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
138 * config/i386/i386.md (rex64namesuffix): New mode attribute.
139 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
140 Merge insn pattern from sse_cvtsi2ss<round_name> and
141 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
142 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
143 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
144 using SWI48 mode iterator.
145 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
146 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
147 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
148 pattern from sse_cvttss2si<round_saeonly_name>
149 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
150 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
151 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
152 using SWI48 mode iterator.
153 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
154 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
155 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
156 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
157 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
158 using SWI48 mode iterator.
159 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
160 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
161 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
162 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
163 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
165 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
166 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
167 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
168 pattern from sse_cvttsd2si<round_saeonly_name>
169 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
171 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
173 * config/nds32/nds32-md-auxiliary.c
174 (nds32_valid_smw_lwm_base_p): Refine.
175 (nds32_output_smw_single_word): Refine.
176 (nds32_output_smw_double_word): New.
177 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
179 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
181 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
182 (nds32_output_stack_pop): Refine.
183 (nds32_expand_unaligned_load): Refine.
184 (nds32_expand_unaligned_store): Refine.
186 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
187 Chung-Ju Wu <jasonwucj@gmail.com>
189 * config/nds32/constants.md: Add TP_REGNUM constant.
190 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
191 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
193 * config/nds32/nds32-doubleword.md: Consider flag_pic.
194 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
195 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
196 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
197 and PIC code generation.
198 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
200 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
202 * config/nds32/nds32.md: Support TLS and PIC.
203 * config/nds32/nds32.c: Support TLS and PIC.
204 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
205 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
208 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
210 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
213 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
214 Chung-Ju Wu <jasonwucj@gmail.com>
216 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
217 * config/nds32/nds32-md-auxiliary.c
218 (symbolic_reference_mentioned_p): New.
219 (nds32_legitimize_ict_address): New.
220 (nds32_expand_ict_move): New.
221 (nds32_indirect_call_referenced_p): New.
222 (nds32_symbol_binds_local_p): Delete.
223 (nds32_long_call_p): Modify.
224 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
225 * config/nds32/nds32-protos.h
226 (symbolic_reference_mentioned_p): Declare.
227 (nds32_legitimize_ict_address): Declare.
228 (nds32_expand_ict_move): Declare.
229 (nds32_indirect_call_referenced_p): Declare.
230 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
231 (nds32_relax_group): Use nds32_ict_const_p as condition.
232 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
233 (nds32_asm_file_start): Output ict_model directive in asm code.
234 (nds32_legitimate_address_p): Consider indirect call.
235 (nds32_print_operand): Consider indirect call.
236 (nds32_print_operand_address): Consider indirect call.
237 (nds32_insert_attributes): Handle "indirect_call" attribute.
238 (TARGET_LEGITIMATE_ADDRESS_P): Define.
239 (TARGET_LEGITIMATE_CONSTANT_P): Define.
240 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
241 (TARGET_DELEGITIMIZE_ADDRESS): Define.
242 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
243 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
244 (TARGET_ICT_MODEL_SMALL): Define.
245 (TARGET_ICT_MODEL_LARGE): Define.
246 * config/nds32/nds32.md (movsi): Consider ict model.
247 (call, call_value): Consider ict model.
248 (sibcall, sibcall_value): Consider ict model.
249 * config/nds32/nds32.opt (mict-model): New option.
250 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
253 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
254 Monk Chiang <sh.chiang04@gmail.com>
255 Jim Wilson <jimw@sifive.com>
257 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
258 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
259 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
260 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
261 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
262 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
263 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
264 compute save_libcall_adjustment properly.
265 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
266 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
267 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
268 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
269 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
270 (ABI_SPEC): Handle mabi=ilp32e.
271 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
273 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
274 <-march>: Add rv32e as an example.
276 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
279 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
280 (intra_create_variable_infos): Handle C++ constructors.
282 2018-05-18 Martin Liska <mliska@suse.cz>
284 * passes.def: Remove a redundant pass.
286 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
289 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
291 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
293 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
295 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
297 (armv2, armv2a, armv3, armv3m): Delete architectures.
298 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
299 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
300 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
302 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
303 (*mulsidi3adddi): Likewise.
304 (mulsidi3): Likewise.
305 (*mulsidi3_nov6): Likewise.
306 (umulsidi3): Likewise.
307 (umulsidi3_nov6): Likewise.
308 (umaddsidi4): Likewise.
309 (*umulsidi3adddi): Likewise.
310 (smulsi3_highpart): Likewise.
311 (*smulsi3_highpart_nov6): Likewise.
312 (umulsi3_highpart): Likewise.
313 (*umulsi3_highpart_nov6): Likewise.
314 * config/arm/arm.h (arm_arch3m): Delete.
315 * config/arm/arm.c (arm_arch3m): Delete.
316 (arm_option_override_internal): Update armv3-related comment.
317 (arm_configure_build_target): Delete use of isa_bit_mode32.
318 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
319 (arm_rtx_costs_internal): Delete check of arm_arch3m.
320 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
323 * config/arm/arm-protos.h (arm_arch3m): Delete.
324 * config/arm/arm-tables.opt: Regenerate.
325 * config/arm/arm-tune.md: Likewise.
326 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
327 deleted architectures.
329 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
331 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
332 (armv5t, armv5te): New features.
333 (ARMv5, ARMv5e): Delete fgroups.
334 (ARMv5t, ARMv5te): Adjust for above changes.
336 (armv5, armv5e): Delete arches.
337 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
339 (*call_reg_arm): Likewise.
340 (*call_value_reg_armv5): Likewise.
341 (*call_value_reg_arm): Likewise.
342 (*call_symbol): Likewise.
343 (*call_value_symbol): Likewise.
344 (*sibcall_insn): Likewise.
345 (*sibcall_value_insn): Likewise.
347 (prefetch): Likewise.
348 (define_split and define_peephole2 dependent on arm_arch5):
350 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
352 (TARGET_ARM_QBIT): Likewise.
353 (TARGET_DSP_MULTIPLY): Likewise.
354 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
355 (arm_arch5, arm_arch5e): Delete.
356 (arm_arch5t, arm_arch5te): Declare.
357 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
358 (arm_arch5t): Declare.
359 (arm_option_reconfigure_globals): Update for the above.
360 (arm_options_perform_arch_sanity_checks): Update comment, replace
361 use of arm_arch5 with arm_arch5t.
362 (use_return_insn): Likewise.
363 (arm_emit_call_insn): Likewise.
364 (output_return_instruction): Likewise.
365 (arm_final_prescan_insn): Likewise.
366 (arm_coproc_builtin_available): Likewise.
367 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
368 arm_arch5e with arm_arch5t and arm_arch5te.
369 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
370 (arm_arch5t, arm_arch5te): Declare.
371 * config/arm/arm-tables.opt: Regenerate.
372 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
373 * config/arm/t-multilib: Likewise.
374 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
375 instead of arm_arch5.
376 (*call_reg_thumb1): Likewise.
377 (*call_value_reg_thumb1_v5): Likewise.
378 (*call_value_reg_thumb1): Likewise.
379 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
381 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
383 2018-05-18 Martin Liska <mliska@suse.cz>
385 PR gcov-profile/84846
386 * doc/gcov.texi: Document -t option of gcov tool.
388 2018-05-18 Martin Liska <mliska@suse.cz>
390 PR gcov-profile/84846
391 * gcov.c (print_usage): Add new -t option.
392 (process_args): Handle the option.
393 (generate_results): Use stdout as output when requested by
396 2018-05-18 Martin Liska <mliska@suse.cz>
398 PR gcov-profile/84846
399 * coverage.c (coverage_init): Write PWD to .gcno file.
400 * doc/gcov.texi: Document how working directory is printed.
401 * gcov-dump.c (dump_gcov_file): Print PWD.
402 * gcov.c (output_intermediate_file): Likewise.
403 (read_graph_file): Read PWD string.
404 (output_lines): Print PWD.
406 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
409 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
410 for retval and return false if all args to phi are zero.
412 2018-05-18 Richard Biener <rguenther@suse.de>
414 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
416 (evrp_dom_walker::before_dom_children): Call it.
418 2018-05-18 Richard Biener <rguenther@suse.de>
420 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
421 results when processing array refs with variable index.
423 2018-05-18 Toon Moene <toon@moene.org>
425 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
426 directly after that of -floop-interchange. Indicate that both
427 options are enabled by default when specifying -O3.
429 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
431 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
432 iterator. Delete separate integer-mode vec_set<mode> expander.
433 (aarch64_simd_vec_setv2di): Delete.
434 (vec_setv2di): Delete.
435 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
436 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
437 the "w, r" alternative.
439 2018-05-18 Martin Liska <mliska@suse.cz>
441 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
442 * tree-pass.h (make_pass_lower_switch_O0): New function.
443 * tree-switch-conversion.c (node_has_low_bound): Remove.
444 (node_has_high_bound): Likewise.
445 (node_is_bounded): Likewise.
446 (class pass_lower_switch): Make it a template type and create
448 (pass_lower_switch::execute): Add template argument.
449 (make_pass_lower_switch): New function.
450 (make_pass_lower_switch_O0): New function.
451 (do_jump_if_equal): Remove.
452 (emit_case_nodes): Simplify to just handle all 3 cases and leave
453 all the hard work to tree optimization passes.
455 2018-05-18 Martin Liska <mliska@suse.cz>
457 * dbgcnt.c (limit_low): Renamed from limit.
458 (limit_high): New variable.
459 (dbg_cnt_is_enabled): Check for upper limit.
460 (dbg_cnt): Adjust dumping.
461 (dbg_cnt_set_limit_by_index): Add new argument for high
463 (dbg_cnt_set_limit_by_name): Likewise.
464 (dbg_cnt_process_single_pair): Parse new format.
465 (dbg_cnt_process_opt): Use strtok.
466 (dbg_cnt_list_all_counters): Remove 'value' and add
468 * doc/invoke.texi: Document changes.
470 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
472 * doc/sourcebuild.texi (scalar_all_fma): Document.
473 * tree.def (FMA_EXPR): Delete.
474 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
475 * internal-fn.c (ternary_direct): New macro.
476 (expand_ternary_optab_fn): Likewise.
477 (direct_ternary_optab_supported_p): Likewise.
478 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
479 * builtins.c (fold_builtin_fma): Delete.
480 (fold_builtin_3): Don't call it.
481 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
482 * expr.c (expand_expr_real_2): Likewise.
483 * fold-const.c (operand_equal_p): Likewise.
484 (fold_ternary_loc): Likewise.
485 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
486 * gimple.c (DEFTREECODE): Likewise.
487 * gimplify.c (gimplify_expr): Likewise.
488 * optabs-tree.c (optab_for_tree_code): Likewise.
489 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
490 * tree-eh.c (operation_could_trap_p): Likewise.
491 (stmt_could_throw_1_p): Likewise.
492 * tree-inline.c (estimate_operator_cost): Likewise.
493 * tree-pretty-print.c (dump_generic_node): Likewise.
494 (op_code_prio): Likewise.
495 * tree-ssa-loop-im.c (stmt_cost): Likewise.
496 * tree-ssa-operands.c (get_expr_operands): Likewise.
497 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
498 * fold-const-call.h (fold_fma): Delete.
499 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
500 CFN_FNMA and CFN_FNMS.
502 * genmatch.c (combined_fn): New enum.
503 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
504 (commutative_op): New function.
505 (commutate): Use it. Handle more than 2 operands.
506 (dt_operand::gen_gimple_expr): Use commutative_op.
507 (parser::parse_expr): Allow :c to be used with non-binary
508 operators if the commutative operand is known.
509 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
510 CFN_FMS, CFN_FNMA and CFN_FNMS.
511 (backprop::process_assign_use): Remove FMA_EXPR handling.
512 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
513 (gen_hsa_fma): New function.
514 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
515 IFN_FNMA and IFN_FNMS.
516 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
517 * gimple-fold.h (follow_all_ssa_edges): Declare.
518 * gimple-fold.c (follow_all_ssa_edges): New function.
519 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
520 gimple_build interface and use follow_all_ssa_edges to fold the result.
521 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
522 instead of checking for optabs directly.
523 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
524 rather than FMA_EXPRs.
525 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
526 call to IFN_FMA instead of an FMA_EXPR.
528 2018-05-17 Jim Wilson <jimw@sifive.com>
530 * expr.c (do_tablejump): When converting index to Pmode, if we have a
531 sign extended promoted subreg, and the range does not have the sign bit
532 set, then do a sign extend.
534 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
535 test, check for sign extended subreg and/or constant operands, and
536 do a sign extend in that case.
538 2018-05-17 Steve Ellcey <sellcey@cavium.com>
540 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
541 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
543 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
544 Change logics_shift_reg to logics_shift_imm.
545 (thunderx2t99_fp_loadpair_basic): Delete.
546 (thunderx2t99_fp_storepair_basic): Delete.
547 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
548 (thunderx2t99_asimd_polynomial): Delete.
549 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
550 and neon_fp_mul_d_scalar_q.
551 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
552 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
553 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
554 (thunderx2t99_asimd_lut): Add missing tbl types.
555 (thunderx2t99_asimd_ext): Delete.
556 (thunderx2t99_asimd_load1_1_mult): Delete.
557 (thunderx2t99_asimd_load1_2_mult): Delete.
558 (thunderx2t99_asimd_load1_ldp): New.
559 (thunderx2t99_asimd_load1): New.
560 (thunderx2t99_asimd_load2): Add missing *load2* types.
561 (thunderx2t99_asimd_load3): New.
562 (thunderx2t99_asimd_load4): New.
563 (thunderx2t99_asimd_store1_1_mult): Delete.
564 (thunderx2t99_asimd_store1_2_mult): Delete.
565 (thunderx2t99_asimd_store2_mult): Delete.
566 (thunderx2t99_asimd_store2_onelane): Delete.
567 (thunderx2t99_asimd_store_stp): New.
568 (thunderx2t99_asimd_store1): New.
569 (thunderx2t99_asimd_store2): New.
570 (thunderx2t99_asimd_store3): New.
571 (thunderx2t99_asimd_store4): New.
573 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
575 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
576 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
578 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
579 Segher Boessenkool <segher@kernel.crashing.org>
582 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
585 2018-05-17 Richard Biener <rguenther@suse.de>
587 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
588 for pruning loop and prune defs feeding only already visited PHIs.
590 2018-05-17 Richard Biener <rguenther@suse.de>
592 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
594 2018-05-17 Bin Cheng <bin.cheng@arm.com>
595 Richard Biener <rguenther@suse.de>
597 PR tree-optimization/85793
598 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
599 for VMAT_ELEMENTWISE.
601 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
603 * internal-fn.h (lookup_internal_fn): Declare
604 * internal-fn.c (lookup_internal_fn): New function.
605 * gimple.c (gimple_build_call_from_tree): Handle calls to
607 * gimple-pretty-print.c (dump_gimple_call): Print "." before
608 internal function names.
609 * tree-pretty-print.c (dump_generic_node): Likewise.
610 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
612 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
614 * gimple-fold.h (gimple_build): Make the function forms take
615 combined_fn rather than built_in_function.
616 (gimple_simplify): Likewise.
617 * gimple-match-head.c (gimple_simplify): Likewise.
618 * gimple-fold.c (gimple_build): Likewise.
619 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
620 rather than gimple_build_call_internal.
621 (get_initial_defs_for_reduction): Likewise.
622 (vect_create_epilog_for_reduction): Likewise.
623 (vectorizable_live_operation): Likewise.
625 2018-05-17 Martin Liska <mliska@suse.cz>
627 * gimple-ssa-sprintf.c (format_directive): Do not use
628 space in between 'G_' and '('.
630 2018-05-17 Jakub Jelinek <jakub@redhat.com>
633 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
634 even if the mask is not all ones.
637 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
639 (ix86_gimple_fold_builtin): Likewise.
642 * config/i386/i386.c: Include tree-vector-builder.h.
643 (ix86_vector_shift_count): New function.
644 (ix86_fold_builtin): Fold shift builtins by scalar count.
645 (ix86_gimple_fold_builtin): Likewise.
647 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
648 _mm512_setzero): New intrinsics.
650 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
651 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
653 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
654 code generation for cases where splatting a value is not useful.
655 * simplify-rtx.c (simplify_ternary_operation): Simplify
656 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
657 mode to a vec_concat.
659 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
661 * config.gcc: Support "goldmont-plus".
662 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
664 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
665 PROCESSOR_GOLDMONT_PLUS.
666 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
667 (processor_target_table): Add "goldmont-plus".
668 (PTA_GOLDMONT_PLUS): Define.
669 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
670 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
671 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
672 (fold_builtin_cpu): Add "goldmont-plus".
673 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
674 (ix86_option_override_internal): Add "goldmont-plus".
675 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
676 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
677 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
678 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
680 2018-05-17 Richard Biener <rguenther@suse.de>
682 PR tree-optimization/85757
683 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
684 remove defs that only feed that PHI from further processing.
686 2018-05-16 Jim Wilson <jimw@sifive.com>
688 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
690 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
692 2018-05-16 Mark Wielaard <mark@klomp.org>
694 * dwarf2out.c (count_index_strings): New function.
695 (output_indirect_strings): Call count_index_strings and generate
696 header for dwarf_version >= 5.
698 2018-05-16 Mark Wielaard <mark@klomp.org>
700 * dwarf2out.c (dwarf_FORM): New function.
701 (set_indirect_string): Use dwarf_FORM.
702 (reset_indirect_string): Likewise.
703 (size_of_die): Likewise.
704 (value_format): Likewise.
705 (output_die): Likewise.
706 (add_skeleton_AT_string): Likewise.
707 (output_macinfo_op): Likewise.
708 (index_string): Likewise.
709 (output_index_string_offset): Likewise.
710 (output_index_string): Likewise.
711 (count_index_strings): Likewise.
713 2018-05-16 Carl Love <cel@us.ibm.com>
715 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
716 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
718 2018-05-16 Martin Jambor <mjambor@suse.cz>
720 * ipa-prop.c (ipa_free_all_edge_args): Remove.
721 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
723 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
725 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
726 (fnma<mode>4): Likewise.
727 (fms<mode>4): Likewise.
728 (fnms<mode>4): Likewise.
729 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
730 (aarch64_fnma<mode>4): Likewise.
731 (aarch64_fms<mode>4): Likewise.
732 (aarch64_fnms<mode>4): Likewise.
733 (aarch64_fnmadd<mode>4): Likewise.
735 2018-05-16 Jason Merrill <jason@redhat.com>
737 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
739 2018-05-16 Richard Biener <rguenther@suse.de>
741 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
742 (dump_stmt_cost): Declare.
743 (add_stmt_cost): Dump cost we add.
744 (add_stmt_costs): New function.
745 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
747 (vect_analyze_stmt): Adjust prototype.
748 (vectorizable_condition): Likewise.
749 (vectorizable_live_operation): Likewise.
750 (vectorizable_reduction): Likewise.
751 (vectorizable_induction): Likewise.
752 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
753 cost vector to pass to vectorizable_ and record afterwards.
754 (vect_model_reduction_cost): Take cost vector argument and adjust.
755 (vect_model_induction_cost): Likewise.
756 (vectorizable_reduction): Likewise.
757 (vectorizable_induction): Likewise.
758 (vectorizable_live_operation): Likewise.
759 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
760 SLP_TREE_NUMBER_OF_VEC_STMTS.
761 (vect_analyze_slp_cost_1): Remove.
762 (vect_analyze_slp_cost): Likewise.
763 (vect_slp_analyze_node_operations): Take visited args and
764 a target cost vector. Avoid processing already visited stmt sets.
765 (vect_slp_analyze_operations): Use a local cost vector to gather
766 costs and register those of non-discarded instances.
767 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
768 (vect_schedule_slp_instance): Remove copying of
769 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
771 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
772 adding cost. Record cost entry location.
773 (vect_prologue_cost_for_slp_op): Function to compute cost of
774 a constant or invariant generated for SLP vect in the prologue,
775 split out from vect_analyze_slp_cost_1.
776 (vect_model_simple_cost): Make static. Adjust for SLP costing.
777 (vect_model_promotion_demotion_cost): Likewise.
778 (vect_model_store_cost): Likewise, make static.
779 (vect_model_load_cost): Likewise.
780 (vectorizable_bswap): Add cost vector arg and adjust.
781 (vectorizable_call): Likewise.
782 (vectorizable_simd_clone_call): Likewise.
783 (vectorizable_conversion): Likewise.
784 (vectorizable_assignment): Likewise.
785 (vectorizable_shift): Likewise.
786 (vectorizable_operation): Likewise.
787 (vectorizable_store): Likewise.
788 (vectorizable_load): Likewise.
789 (vectorizable_condition): Likewise.
790 (vectorizable_comparison): Likewise.
791 (can_vectorize_live_stmts): Likewise.
792 (vect_analyze_stmt): Likewise.
793 (vect_transform_stmt): Adjust calls to vectorizable_*.
794 * tree-vectorizer.c: Include gimple-pretty-print.h.
795 (dump_stmt_cost): New function.
797 2018-05-16 Richard Biener <rguenther@suse.de>
799 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
800 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
801 * tree-ssa-dse.c: Include tree-ssa-loop.h.
802 (check_name): New callback.
803 (dse_classify_store): Track cycles via a visited bitmap of PHI
804 defs and simplify handling of in-loop and across loop dead stores
805 and properly fail for loop-variant refs. Handle byte-tracking with
806 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
809 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
811 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
812 (vect_get_mask_type_for_stmt): Likewise.
813 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
815 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
816 to determine the statement's vector type and the vector type that
817 should be used for calculating nunits. Deal with cases in which
818 the type has to be deferred.
819 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
820 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
821 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
822 (vect_determine_vf_for_stmt): New functions, split out from...
823 (vect_determine_vectorization_factor): ...here.
824 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
825 (vect_get_mask_type_for_stmt): New functions, split out from
826 vect_determine_vectorization_factor.
828 2018-05-16 Richard Biener <rguenther@suse.de>
830 * tree-cfg.c (verify_gimple_assign_ternary): Properly
831 verify the [VEC_]COND_EXPR embedded comparison.
833 2018-05-15 Martin Sebor <msebor@redhat.com>
835 PR tree-optimization/85753
836 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
837 RECORD_TYPE in addition to ARRAY_TYPE.
839 2018-05-15 Martin Sebor <msebor@redhat.com>
842 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
844 2018-05-15 Richard Biener <rguenther@suse.de>
846 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
847 add by_clobber_p one. Change algorithm to collect all defs
848 representing uses we need to walk and try reducing them to
849 a single one before failing.
850 (dse_dom_walker::dse_optimize_stmt): Adjust.
852 2018-05-13 Mark Wielaard <mark@klomp.org>
854 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
855 (size_of_loc_descr): Likewise.
856 (output_loc_operands): Likewise.
857 (output_loc_operands_raw): Likewise.
858 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
859 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
860 (hash_loc_operands): Likewise.
861 (compare_loc_operands): Likewise.
863 2018-05-14 Mark Wielaard <mark@klomp.org>
865 * dwarf2out.c (count_index_addrs): New function.
866 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
868 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
870 PR tree-optimization/83648
871 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
872 return value as malloc candidate.
874 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
877 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
878 as true in call to suggest_attribute.
880 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
882 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
885 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
887 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
889 (aarch64_crypto_aesd_fused): Likewise.
891 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
893 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
894 (movsi_aarch64): Likewise.
895 (load_pairsi): Likewise.
896 (load_pairdi): Likewise.
897 (store_pairsi): Likewise.
898 (store_pairdi): Likewise.
899 (load_pairsf): Likewise.
900 (load_pairdf): Likewise.
901 (store_pairsf): Likewise.
902 (store_pairdf): Likewise.
903 (zero_extend): Likewise.
904 (trunc): Swap alternatives.
905 (fcvt_target): Add '?' to prefer w over r.
907 2018-05-14 Jakub Jelinek <jakub@redhat.com>
910 * config/i386/i386.md: Disallow non-commutative arithmetics in
911 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
912 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
913 in the peephole2 before it.
915 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
917 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
918 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
919 (ix86_handle_option): Handle -mcldemote.
920 * config.gcc: New header.
921 * config/i386/cldemoteintrin.h: New file.
922 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
923 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
925 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
926 OPTION_MASK_ISA_CLDEMOTE.
927 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
928 (ix86_valid_target_attribute_inner_p): Ditto.
929 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
930 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
931 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
932 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
933 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
935 * config/i386/i386.opt: Add -mcldemote.
936 * config/i386/x86intrin.h: New header.
937 * doc/invoke.texi: Add -mcldemote.
939 2018-05-14 Richard Biener <rguenther@suse.de>
941 * doc/match-and-simplify.texi: Adjust :s documentation.
943 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
945 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
946 intended memcpy size.
947 (REORDER_45): Likewise.
949 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
952 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
953 * vec.c (qsort_chk): Use gcc_qsort.
954 * Makefile.in (OBJS-libcommon): Add sort.o.
955 (build/sort.o): New target. Use it...
956 (BUILD_RTL): ... here, and...
957 (build/gencfn-macros): ... here, and...
958 (build/genmatch): ... here.
960 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
961 Chung-Ju Wu <jasonwucj@gmail.com>
963 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
964 * config/nds32/nds32-graywolf.md: New file.
965 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
966 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
968 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
969 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
970 * config/nds32/nds32.md (pipeline_model): Add graywolf.
971 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
972 * config/nds32/pipelines.md: Include n15 settings.
974 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
975 Chung-Ju Wu <jasonwucj@gmail.com>
977 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
978 * config/nds32/nds32-n13.md: New file.
979 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
980 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
982 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
983 * config/nds32/nds32.md (pipeline_model): Add n13.
984 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
985 * config/nds32/pipelines.md: Include n13 settings.
987 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
988 Chung-Ju Wu <jasonwucj@gmail.com>
990 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
991 * config/nds32/nds32-n10.md: New file.
992 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
993 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
995 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
996 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
997 * config/nds32/nds32.md (pipeline_model): Add n10.
998 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
999 * config/nds32/pipelines.md: Include n10 settings.
1001 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1002 Kito Cheng <kito.cheng@gmail.com>
1003 Chung-Ju Wu <jasonwucj@gmail.com>
1005 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1006 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1007 Add enum values for DSP extension instructions.
1008 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1010 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1011 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1013 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1014 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1015 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1016 * config/nds32/nds32-intrinsic.md: Likewise.
1017 * config/nds32/nds32_intrinsic.h: Likewise.
1018 * config/nds32/nds32-md-auxiliary.c: Likewise.
1019 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1020 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1021 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1022 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1023 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1024 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1025 TYPE_DMAC in switch statement.
1026 * config/nds32/nds32.c: New checking and implementation for DSP
1027 extension instructions.
1028 * config/nds32/nds32.h: Likewise.
1029 * config/nds32/nds32.md: Likewise.
1030 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1031 * config/nds32/predicates.md: Implement new predicates for DSP
1034 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1036 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1037 Reformat alternatives and attributes so it is easier to identify
1038 which constraints/attributes go with which instruction.
1039 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1040 (mov<mode>_softfloat32, FMOVE64): Likewise.
1041 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1042 (mov<mode>_softfloat64, FMOVE64): Likewise.
1044 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1046 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1048 (Basic PowerPC Built-in Functions): The new name of the
1049 subsection previously known as "PowerPC Built-in Functions".
1050 (Basic PowerPC Built-in Functions Available on all Configurations):
1052 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1053 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1054 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1055 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1057 2018-05-11 Martin Jambor <mjambor@suse.cz>
1060 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1063 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1066 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1068 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1070 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1071 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1072 (ix86_handle_option): Handle -mwaitpkg.
1073 * config.gcc: New header.
1074 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1075 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1076 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1078 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1079 OPTION_MASK_ISA_WAITPKG.
1080 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1081 (ix86_option_override_internal): Add PTA_WAITPKG.
1082 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1083 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1084 IX86_BUILTIN_TPAUSE.
1085 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1086 __builtin_ia32_umwait and __builtin_ia32_tpause.
1087 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1088 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1089 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1090 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1091 UNSPECV_TPAUSE): New.
1092 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1093 * config/i386/i386.opt: Add -mwaitpkg.
1094 * config/i386/waitpkgintrin.h: New file.
1095 * config/i386/x86intrin.h: New header.
1096 * doc/invoke.texi: Add -mwaitpkg.
1098 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1101 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1103 (cortex-m0): Use armv6s-m isa.
1104 (cortex-m0plus): Likewise.
1105 (cortex-m1): Likewise.
1106 (cortex-m0.small-multiply): Likewise.
1107 (cortex-m0plus.small-multiply): Likewise.
1108 (cortex-m1.small-multiply): Likewise.
1110 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1111 Jakub Jelinek <jakub@redhat.com>
1113 PR tree-optimization/85692
1114 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1115 source permute as well.
1117 2018-05-11 Martin Liska <mliska@suse.cz>
1120 * doc/extend.texi: Document LLVM style format for no_sanitize
1123 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1125 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1126 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1127 (mode_supports_vsx_dform_quad): Likewise.
1128 (mode_supports_vmx_dform): Move these functions to be next to the
1129 other mode_supports functions.
1130 (mode_supports_dq_form): Likewise.
1131 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1132 mode_supports_dq_form.
1133 (reg_offset_addressing_ok_p): Likewise.
1134 (offsettable_ok_by_alignment): Likewise.
1135 (rs6000_legitimate_offset_address_p): Likewise.
1136 (legitimate_lo_sum_address_p): Likewise.
1137 (rs6000_legitimize_address): Likewise.
1138 (rs6000_legitimize_reload_address): Likewise.
1139 (rs6000_secondary_reload_inner): Likewise.
1140 (rs6000_preferred_reload_class): Likewise.
1141 (rs6000_output_move_128bit): Likewise.
1143 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1145 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1146 Generate SImode target register for null target.
1147 <case IX86_BUILTIN_XGETBV>: Ditto.
1148 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1149 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1151 2018-05-10 Carl Love <cel@us.ibm.com>
1153 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1154 dcbtt and dcbtstt if operands[2] is 0.
1156 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1159 * config/i386/sse.md (usadv64qi): New expander.
1161 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1163 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1164 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1165 -maltivec=be support.
1166 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1167 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1168 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1169 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1170 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1171 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1172 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1173 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1174 altivec_vsumsws): Adjust.
1175 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1176 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1177 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1179 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1180 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1181 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1182 (altivec_lve<VI_char>x): Delete expand.
1183 (*altivec_lve<VI_char>x_internal): Rename to...
1184 (altivec_lve<VI_char>x): ... this.
1185 (altivec_lvxl_<mode>): Delete expand.
1186 (*altivec_lvxl_<mode>_internal): Rename to ...
1187 (altivec_lvxl_<mode>): ... this.
1188 (altivec_stvxl_<mode>): Delete expand.
1189 (*altivec_stvxl_<mode>_internal): Rename to ...
1190 (altivec_stvxl_<mode>): ... this.
1191 (altivec_stve<VI_char>x): Delete expand.
1192 (*altivec_stve<VI_char>x_internal): Rename to ...
1193 (altivec_stve<VI_char>x): ... this.
1194 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1195 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1196 reduc_plus_scal_<mode>): Adjust.
1197 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1199 (rs6000_cpu_cpp_builtins): Adjust.
1200 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1201 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1202 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1203 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1204 -maltivec=be support.
1205 (rs6000_split_vec_extract_var): Adjust.
1206 (rs6000_split_v4si_init): Adjust.
1207 (swap_selector_for_mode): Delete.
1208 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1209 altivec_expand_stvex_be): Delete.
1210 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1211 -maltivec=be support.
1212 (rs6000_gimple_fold_builtin): Ditto.
1213 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1215 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1216 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1217 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1218 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1219 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1220 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1221 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1222 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1223 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1224 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1225 anonymous split): Adjust.
1226 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1227 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1229 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1231 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1232 when --with-gxx-include-dir is also specified.
1233 * configure: Regenerate.
1235 2018-05-09 Jim Wilson <jimw@sifive.com>
1238 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1239 * config/riscv/t-withmultilib: New.
1240 * config/riscv/withmultilib.h: New.
1241 * doc/install.texi: Document RISC-V --with-multilib-list support.
1243 2018-05-09 Richard Biener <rguenther@suse.de>
1245 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1247 (vect_bb_vectorization_profitable_p): Adjust. Compute
1248 actual scalar cost using the cost vector and the add_stmt_cost
1251 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1253 PR rtl-optimization/85645
1254 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1255 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1257 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1259 PR rtl-optimization/85645
1260 * shrink-wrap.c (spread_components): Return a boolean saying if
1261 anything was changed.
1262 (try_shrink_wrapping_separate): Iterate spread_components until
1263 nothing changes anymore.
1265 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1267 PR rtl-optimization/85645
1268 * regrename.c (build_def_use): Also kill the chains that include the
1269 destination of a REG_CFA_REGISTER note.
1271 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1273 PR rtl-optimization/85645
1274 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1275 insn that has a REG_CFA_REGISTER note.
1277 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1279 * cfgexpand.c (expand_clobber): New function.
1280 (expand_gimple_stmt_1): Use it.
1281 * tree-vect-stmts.c (vect_clobber_variable): New function,
1283 (vectorizable_simd_clone_call): ...here.
1284 (vectorizable_store): Emit a clobber either side of an
1285 IFN_STORE_LANES sequence.
1286 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1288 2018-05-09 Tom de Vries <tom@codesourcery.com>
1291 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1292 (define_insn "trap_if_false"): Add exit after trap.
1294 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1296 PR rtl-optimization/85638
1297 * bb-reorder.c: Include common/common-target.h.
1298 (create_forwarder_block): New function extracted from...
1299 (fix_up_crossing_landing_pad): ...here. Rename into...
1300 (dw2_fix_up_crossing_landing_pad): ...this.
1301 (sjlj_fix_up_crossing_landing_pad): New function.
1302 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1303 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1304 from both partitions and exit the loop after one iteration.
1306 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1309 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1311 (Basic PowerPC Built-in Functions): The new name of the
1312 subsection previously known as "PowerPC Built-in Functions".
1313 (Basic PowerPC Built-in Functions Available on all Configurations):
1315 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1317 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1319 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1321 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1324 2018-05-08 Jim Wilson <jimw@sifive.com>
1326 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1327 (LD_EMUL_SUFFIX): New.
1328 (LINK_SPEC): Use it.
1330 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1332 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1334 (Basic PowerPC Built-in Functions): The new name of the
1335 subsection previously known as "PowerPC Built-in Functions".
1336 (Basic PowerPC Built-in Functions Available on all Configurations):
1338 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1340 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1342 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1344 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1347 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1350 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1351 after cmpelim optimization.
1353 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1355 * config.gcc: Support "goldmont".
1356 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1357 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1359 * config/i386/i386.c (m_GOLDMONT): Define.
1360 (processor_target_table): Add "goldmont".
1361 (PTA_GOLDMONT): Define.
1362 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1363 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1364 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1365 (fold_builtin_cpu): Add "goldmont".
1366 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1367 (ix86_option_override_internal): Add "goldmont".
1368 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1369 (processor_type): Add PROCESSOR_GOLDMONT.
1370 * config/i386/i386.md: Add CPU "glm".
1371 * config/i386/glm.md: New file.
1372 * config/i386/x86-tune.def: Add m_GOLDMONT.
1373 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1375 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1378 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1380 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1381 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1382 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1385 * config/i386/i386.c (ix86_fold_builtin): Handle
1386 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1389 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1390 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1392 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1395 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1396 (check_arch): Likewise.
1397 (check_fpu): Return the result rather than printing it.
1398 (end arch): Fix operator precedence.
1399 (end cpu): Likewise.
1400 (END): Print the result from check_fpu.
1402 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1403 Alan Hayward <alan.hayward@arm.com>
1404 David Sherwood <david.sherwood@arm.com>
1406 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1407 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1408 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1409 (*fcmuo<mode>_and): New patterns.
1411 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1413 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1414 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1415 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1416 (cmp_op, sve_imm_con): New code attributes.
1417 (SVE_COND_INT_CMP, imm_con): Delete.
1418 (cmp_op): Remove above unspecs from int attribute.
1419 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1421 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1422 comparison-specific unspecs.
1423 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1424 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1425 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1426 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1427 (*vec_fcm<cmp_op><mode>): Rename to...
1428 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1429 (*vec_fcmuo<mode>): Rename to...
1430 (*fcmuo<mode>): ...this and adjust likewise.
1431 (*pred_fcm<cmp_op><mode>): New pattern.
1432 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1433 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1435 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1437 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1438 (aarch64_emit_sve_predicated_cond): New function.
1439 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1440 (aarch64_emit_unspec_cond_or): Replace with...
1441 (aarch64_emit_sve_or_conds): ...this new function. Use
1442 aarch64_emit_sve_ptrue_op for the individual comparisons and
1443 aarch64_emit_binop to OR them together.
1444 (aarch64_emit_inverted_unspec_cond): Replace with...
1445 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1446 aarch64_emit_sve_ptrue_op for the comparison and
1447 aarch64_emit_unop to invert the result.
1448 (aarch64_expand_sve_vec_cmp_float): Update after the above
1449 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1451 2018-05-07 Nathan Sidwell <nathan@acm.org>
1453 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1454 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1455 (Backwards Compatibility): Likewise.
1457 2018-05-07 Luis Machado <luis.machado@linaro.org>
1461 2018-05-07 Luis Machado <luis.machado@linaro.org>
1463 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1464 <prefetch_dynamic_strides>: New const bool field.
1465 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1466 prefetch_dynamic_strides.
1467 (exynosm1_prefetch_tune): Likewise.
1468 (thunderxt88_prefetch_tune): Likewise.
1469 (thunderx_prefetch_tune): Likewise.
1470 (thunderx2t99_prefetch_tune): Likewise.
1471 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1472 (aarch64_override_options_internal): Update to set
1473 PARAM_PREFETCH_DYNAMIC_STRIDES.
1474 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1475 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1476 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1477 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1478 prefetch-dynamic-strides setting.
1480 2018-05-07 Luis Machado <luis.machado@linaro.org>
1482 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1483 <minimum_stride>: New const int field.
1484 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1485 minimum_stride field.
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 minimum_stride to 2048.
1491 (aarch64_override_options_internal): Update to set
1492 PARAM_PREFETCH_MINIMUM_STRIDE.
1493 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1494 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1495 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1496 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1497 stride is constant and is below the minimum stride threshold.
1499 2018-05-07 Luis Machado <luis.machado@linaro.org>
1501 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1504 2018-05-07 Luis Machado <luis.machado@linaro.org>
1506 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1507 <prefetch_dynamic_strides>: New const bool field.
1508 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1509 prefetch_dynamic_strides.
1510 (exynosm1_prefetch_tune): Likewise.
1511 (thunderxt88_prefetch_tune): Likewise.
1512 (thunderx_prefetch_tune): Likewise.
1513 (thunderx2t99_prefetch_tune): Likewise.
1514 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1515 (aarch64_override_options_internal): Update to set
1516 PARAM_PREFETCH_DYNAMIC_STRIDES.
1517 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1518 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1519 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1520 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1521 prefetch-dynamic-strides setting.
1523 2018-05-07 Luis Machado <luis.machado@linaro.org>
1525 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1526 <minimum_stride>: New const int field.
1527 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1528 minimum_stride field.
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 minimum_stride to 2048.
1534 (aarch64_override_options_internal): Update to set
1535 PARAM_PREFETCH_MINIMUM_STRIDE.
1536 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1537 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1538 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1539 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1540 stride is constant and is below the minimum stride threshold.
1542 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1545 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1546 the type is addressable. Don't force op into register if it has
1549 2018-05-05 Roland McGrath <mcgrathr@google.com>
1552 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1553 any section for which we don't know a specific type it should have,
1554 regardless of name. Previously this was done only for the exact
1555 names ".init_array", ".fini_array", and ".preinit_array".
1556 (default_elf_asm_named_section): Add comment about
1557 relationship with default_section_type_flags and SECTION_NOTYPE.
1558 (get_section): Don't consider it a type conflict if one side has
1559 SECTION_NOTYPE and the other doesn't, as long as neither has the
1560 SECTION_BSS et al used in the default_section_type_flags logic.
1562 2018-05-05 Tom de Vries <tom@codesourcery.com>
1565 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1566 (workaround_barsyncs): New function.
1567 (nvptx_reorg): Use workaround_barsyncs.
1568 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1569 (define_expand "nvptx_membar_cta"): New define_expand.
1570 (define_insn "*nvptx_membar_cta"): New insn.
1572 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1574 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1575 To improve optimization opportunities.
1576 * builtin-types.def: The new needed builtin types for the above.
1578 2018-05-04 Richard Biener <rguenther@suse.de>
1580 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1581 * gimple-ssa-store-merging.c
1582 (imm_store_chain_info::output_merged_store): Remove redundant create,
1583 release split_store vector contents on failure.
1584 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1585 scalar stmt vector on cache hit.
1587 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1589 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1591 * config.gcc (powerpc-xilinx-eabi*): Remove.
1592 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1594 (fusion_addis_mem_combo_load): Ditto.
1595 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1597 (rs6000_cpu_cpp_builtins): Ditto.
1598 * config/rs6000/rs6000-linux.c
1599 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1600 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1601 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1603 (rs6000_setup_reg_addr_masks): Ditto.
1604 (rs6000_init_hard_regno_mode_ok): Ditto.
1605 (rs6000_option_override_internal): Ditto.
1606 (legitimate_lo_sum_address_p): Ditto.
1607 (rs6000_legitimize_address): Ditto.
1608 (rs6000_legitimize_reload_address): Ditto.
1609 (rs6000_legitimate_address_p): Ditto.
1610 (abi_v4_pass_in_fpr): Ditto.
1611 (setup_incoming_varargs): Ditto.
1612 (rs6000_gimplify_va_arg): Ditto.
1613 (rs6000_split_multireg_move): Ditto.
1614 (rs6000_savres_strategy): Ditto.
1615 (rs6000_emit_prologue_components): Ditto.
1616 (rs6000_emit_epilogue_components): Ditto.
1617 (rs6000_emit_prologue): Ditto.
1618 (rs6000_emit_epilogue): Ditto.
1619 (rs6000_elf_file_end): Ditto.
1620 (rs6000_function_value): Ditto.
1621 (rs6000_libcall_value): Ditto.
1622 * config/rs6000/rs6000.h: Ditto.
1623 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1624 (TARGET_MINMAX): ... this. New.
1625 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1626 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1627 (*movsi_internal1_single): Delete.
1628 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1629 mfpu=, mxilinx-fpu): Delete.
1630 * config/rs6000/singlefp.h: Delete.
1631 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1632 * config/rs6000/t-rs6000: Ditto.
1633 * config/rs6000/t-xilinx: Delete.
1634 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1635 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1636 (VStype_simple): Delete.
1637 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1638 * config/rs6000/xfpu.h: Delete.
1639 * config/rs6000/xfpu.md: Delete.
1640 * config/rs6000/xilinx.h: Delete.
1641 * config/rs6000/xilinx.opt: Delete.
1642 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1643 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1645 2018-05-04 Tom de Vries <tom@codesourcery.com>
1648 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1651 2018-05-04 Richard Biener <rguenther@suse.de>
1654 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1655 (expand_complex_div_wide): Likewise.
1656 (expand_complex_operations_1): Likewise.
1657 (expand_complex_libcall): Preserve EH info of the original stmt.
1658 (tree_lower_complex): Handle removed blocks.
1659 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1660 on complex multiplication and division libcall builtins.
1662 2018-05-04 Richard Biener <rguenther@suse.de>
1665 * fold-const.c (negate_expr_p): Restrict negation of operand
1666 zero of a division to when we know that can happen without
1668 (fold_negate_expr_1): Likewise.
1670 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1673 * real.h (real_nextafter): Declare.
1674 * real.c (real_nextafter): New function.
1675 * fold-const-call.c (fold_const_nextafter): New function.
1676 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1677 CASE_CFN_NEXTTOWARD.
1678 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1679 even when arg1_mode is different from arg0_mode.
1681 2018-05-03 Nathan Sidwell <nathan@acm.org>
1683 * doc/extend.texi (Deprecated Features): Remove
1685 (Backwards Compatibility): Likewise.
1686 * doc/invoke.texi (C++ Language Options): Likewise.
1687 (C++ Dialect Options): Likewise.
1689 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1692 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1693 _mm512_mask_mullox_epi64): New intrinsics.
1695 2018-05-03 Tom de Vries <tom@codesourcery.com>
1698 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1699 dump files): Add offload-tree.
1701 2018-05-03 Richard Biener <rguenther@suse.de>
1703 PR tree-optimization/85615
1704 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1705 to loops not nested in BBs loop father to avoid creating multi-entry
1708 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1710 PR tree-optimization/70291
1711 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1712 arguments. Change return type to tree. Emit libcall as a new
1713 statement rather than replacing existing one when inplace_p is true.
1714 (expand_complex_multiplication_components): New function.
1715 (expand_complex_multiplication): Expand floating-point complex
1716 multiplication using the above.
1717 (expand_complex_division): Rename inner_type parameter to type.
1718 Update expand_complex_libcall call-site.
1719 (expand_complex_operations_1): Update expand_complex_multiplication
1720 and expand_complex_division call-sites.
1722 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1725 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1726 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1727 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1728 the highest significant bit of the shift count mask is clear. In
1729 check whether and[sq]i3 is needed verify that all significant bits
1730 of the shift count other than the highest are set.
1732 2018-05-02 Tom de Vries <tom@codesourcery.com>
1735 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1736 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1737 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1738 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1739 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1740 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1741 __builtin_goacc_parlevel_size.
1743 2018-05-02 Richard Biener <rguenther@suse.de>
1745 PR tree-optimization/85597
1746 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1747 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1750 2018-05-02 Tom de Vries <tom@codesourcery.com>
1753 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1754 dump files): Add ltrans-tree.
1756 2018-05-02 Tom de Vries <tom@codesourcery.com>
1759 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1760 dump files): Add wpa-ipa.
1762 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1764 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1765 powerpc*-*-linux*paired* target.
1766 * config/rs6000/750cl.h: Delete.
1767 * config/rs6000/paired.h: Delete.
1768 * config/rs6000/paired.md: Delete.
1769 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1771 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1772 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1773 comment. Remove paired float support.
1774 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1775 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1777 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1778 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1780 * config/rs6000/rs6000.c: Remove paired float support.
1781 (paired_expand_vector_init, paired_expand_vector_move,
1782 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1783 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1784 paired_expand_builtin, paired_expand_predicate_builtin,
1785 paired_init_builtins): Delete.
1786 * config/rs6000/rs6000.h: Remove paired float support.
1787 * config/rs6000/rs6000.md: Remove paired float support.
1788 (move_from_CR_ov_bit): Delete.
1789 * config/rs6000/rs6000.opt (mpaired): Delete.
1790 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1791 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1793 2018-05-02 Richard Biener <rguenther@suse.de>
1796 * gimplify.c (gimplify_save_expr): When in SSA form allow
1797 SAVE_EXPRs to compute to SSA vars.
1799 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1802 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1803 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1804 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1805 clobber operands[2], instead use a new pseudo. Formatting fixes.
1807 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1809 PR tree-optimization/85586
1810 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1811 exit early for statements in the same group if the accesses are
1814 2018-05-02 Tom de Vries <tom@codesourcery.com>
1817 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1820 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1822 PR tree-optimization/85143
1823 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1825 2018-05-01 Tom de Vries <tom@codesourcery.com>
1828 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1829 not found" error message.
1831 2018-05-01 Tom de Vries <tom@codesourcery.com>
1834 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1835 * vec.c (test_ordered_remove_if): New function.
1836 (vec_c_tests): Call test_ordered_remove_if.
1837 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1838 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1839 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1840 VEC_ORDERED_REMOVE_IF.
1842 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1844 PR tree-optimization/82665
1845 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1846 pointer subtraction where arguments come from a memchr call.
1848 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1850 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1851 --push-state --as-needed and --pop-state instead of --as-needed and
1852 --no-as-needed if ld supports it.
1853 * configure: Regenerated.
1856 * doc/install.texi2html: Replace _002d with - and _002a with * in
1857 generated html files using sed.
1859 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1862 * gcc-rich-location.c (blank_line_before_p): New function.
1863 (use_new_line): New function.
1864 (gcc_rich_location::add_fixit_insert_formatted): New function.
1865 * gcc-rich-location.h
1866 (gcc_rich_location::add_fixit_insert_formatted): New function.
1868 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1870 * selftest.c (assert_streq): Rename "expected" and "actual" to
1871 "val1" and "val2". Extend NULL-handling to cover both inputs
1872 symmetrically, while still requiring both to be non-NULL for a pass.
1873 * selftest.h (assert_streq): Rename "expected" and "actual" to
1875 (ASSERT_EQ): Likewise.
1876 (ASSERT_EQ_AT): Likewise.
1877 (ASSERT_KNOWN_EQ): Likewise.
1878 (ASSERT_KNOWN_EQ_AT): Likewise.
1879 (ASSERT_NE): Likewise.
1880 (ASSERT_MAYBE_NE): Likewise.
1881 (ASSERT_MAYBE_NE_AT): Likewise.
1882 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1883 the assertion to pass.
1884 (ASSERT_STREQ_AT): Likewise.
1886 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1888 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1889 interaction with -pie.
1891 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1893 * selftest.h: Fix alphabetization of per-source-file selftest
1896 2018-04-30 Jason Merrill <jason@redhat.com>
1898 PR c++/61982 - dead stores to destroyed objects.
1899 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1902 2018-04-30 Jason Merrill <jason@redhat.com>
1904 * tree.c (build_clobber): New.
1905 * tree.h: Declare it.
1906 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1908 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1910 * diagnostic-show-locus.c (layout::layout): Update for
1911 location_get_source_line returning a char_span.
1912 (struct char_span): Move to input.h.
1913 (struct correction): Update for fields in char_span becoming
1915 (struct source_line): Update for location_get_source_line
1916 returning a char_span.
1917 (layout::print_line): Likewise.
1918 * edit-context.c (edited_file::print_content): Likewise.
1919 (edited_file::print_diff_hunk): Likewise.
1920 (edited_file::print_run_of_changed_lines): Likewise.
1921 (edited_file::get_num_lines): Likewise.
1922 (edited_line::edited_line): Likewise.
1923 * final.c (asm_show_source): Likewise.
1924 * input.c (location_get_source_line): Convert return type
1925 from const char * to char_span, losing the final "line_len"
1927 (dump_location_info): Update for the above.
1928 (get_substring_ranges_for_loc): Likewise. Use a char_span
1929 when handling the literal within the line.
1930 (test_reading_source_line): Update for location_get_source_line
1931 returning a char_span.
1932 * input.h (class char_span): Move here from
1933 diagnostic-show-locus.c, converting from a struct to a class.
1934 Make data members private.
1935 (char_span::operator bool): New.
1936 (char_span::length): New.
1937 (char_span::get_buffer): New.
1938 (char_span::operator[]): New.
1939 (char_span::subspan): Make const.
1940 (char_span::xstrdup): New.
1941 (location_get_source_line): Convert return type from const char *
1942 to char_span, losing the final "line_size" param.
1944 2018-04-30 Jan Hubicka <jh@suse.cz>
1946 * lto-wrapper.c (ltrans_priorities): New static var.
1947 (cmp_priority): New.
1948 (run_gcc): Read priorities and if doing parallel build order
1949 the Makefile by them.
1951 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1953 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1955 2018-04-30 Richard Biener <rguenther@suse.de>
1957 * tree-cfg.c (verify_address): Remove base argument, add
1958 flag whether to check TREE_ADDRESSABLE and do that.
1959 (verify_expr): Remove.
1960 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1961 (verify_gimple_assign_single): Likewise.
1962 (verify_gimple_switch): Likewise.
1963 (verify_expr_location_1): Dereference tp once. Add (disabled)
1964 piece from verify_expr.
1965 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1967 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1969 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1971 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1973 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1974 (small_data_pattern): Likewise.
1975 (arc_rewrite_small_data): Likewise.
1976 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1977 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1978 (get_symbol_alignment): New function.
1979 (legitimate_small_data_address_p): Likewise.
1980 (legitimate_scaled_address): Update, call
1981 legitimate_small_data_address_p.
1982 (output_sdata): New static variable.
1983 (arc_print_operand): Update how we handle small data operands.
1984 (arc_print_operand_address): Likewise.
1985 (arc_legitimate_address_p): Update, use
1986 legitimate_small_data_address_p.
1987 (arc_rewrite_small_data_p): Remove.
1988 (arc_rewrite_small_data_1): Likewise.
1989 (arc_rewrite_small_data): Likewise.
1990 (small_data_pattern): Likewise.
1991 (compact_sda_memory_operand): Update to use
1992 legitimate_small_data_address_p and get_symbol_alignment.
1993 (prepare_move_operands): Don't rewite sdata pattern.
1994 (prepare_extend_operands): Remove.
1995 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1997 (zero_extendqisi2): Likewise.
1998 (zero_extendhisi2): Likewise.
1999 (extendqihi2): Likewise.
2000 (extendqisi2): Likewise.
2001 (extendhisi2): Likewise.
2005 * config/arc/constraints.md (Usd): Change it to memory constraint.
2007 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2009 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2010 as source of std instructions.
2011 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2012 allow 6-bit constants as source for store instructions.
2013 (movdi_insn): Update instruction pattern to allow 6-bit constants
2014 as source for store instructions.
2016 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2018 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2020 2018-04-30 Nathan Sidwell <nathan@acm.org>
2021 Sandra Loosemore <sandra@codesourcery.com>
2023 * dumpfile.c (dump_open): Allow '-' for stdout.
2024 * doc/invoke.texi (Developer Options): Document dump filename
2025 determination early. Document stdin/stdout selection.
2027 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2029 Microblaze Target: PIC data text relative
2031 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2032 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2034 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2035 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2036 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2037 New addressing mode for data-text relative position indepenedent code.
2038 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2039 'ADDRESS_SYMBOLIC_TXT_REL'.
2040 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2041 (microblaze_legitimate_pic_operand): Exclude function calls from
2042 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2043 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2045 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2046 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2047 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2048 for 'address + offset'.
2049 (microblaze_expand_prologue): Add new function prologue call for
2051 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2052 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2053 table in case of TARGET_PIC_DATA_TEXT_REL.
2054 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2055 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2056 Add new macros 'UNSPEC_TEXT',
2057 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2058 + exclude function calls from 'UNSPEC_PLT' in case of data text
2060 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2061 new target hook for generating address diff vector tables in case of
2063 * doc/tm.texi : Regenerate.
2064 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2065 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2066 of addr diff vector generation.
2067 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2068 target hook definition.
2069 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2070 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2071 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2072 option for data text relative.
2074 2018-04-30 Richard Biener <rguenther@suse.de>
2076 * tree-chrec.h (evolution_function_is_constant_p): Remove
2078 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2080 2018-04-30 Richard Biener <rguenther@suse.de>
2083 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2085 2018-04-30 Richard Biener <rguenther@suse.de>
2087 PR tree-optimization/28364
2088 PR tree-optimization/85275
2089 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2090 copying first exit test.
2092 2018-04-28 Mark Wielaard <mark@klomp.org>
2094 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2096 (dwarf_AT): Handle DW_AT_addr_base.
2097 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2099 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2102 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2103 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2104 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2105 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2107 2018-04-28 Richard Biener <rguenther@suse.de>
2109 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2110 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2111 to reflect use. Only add interesting stmts.
2113 2018-04-27 Martin Jambor <mjambor@suse.cz>
2116 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2117 the jump function allows for passing through aggregate values.
2119 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2121 * input.h (in_system_header_at): Convert from macro to inline
2123 (from_macro_expansion_at): Likewise.
2124 (from_macro_definition_at): Likewise.
2126 2018-04-27 Jeff Law <law@redhat.com>
2128 * config.gcc: Mark tile* targets as deprecated/obsolete.
2130 2018-04-27 Richard Biener <rguenther@suse.de>
2132 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2135 2018-04-27 Richard Biener <rguenther@suse.de>
2137 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2139 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2141 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2142 with Yd constraint. Set "preferred_for_speed" attribute from
2143 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2145 (*movdi_internal): Ditto.
2146 (movti_interunit splitters): Remove
2147 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2148 (movdi_interunit splitters): Ditto.
2149 * config/i386/constraints.md (Ye): Remove.
2150 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2152 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2155 * config/aarch64/constraints.md (Usg): Limit to 31.
2158 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2160 PR tree-optimization/85529
2161 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2162 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2163 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2164 zero extension or masking of the MSB bit.
2165 (optimize_range_tests): Add FIRST_BB argument, pass it through
2166 to optimize_range_tests_var_bound.
2167 (maybe_optimize_range_tests, reassociate_bb): Adjust
2168 optimize_range_tests callers.
2170 2018-04-26 Richard Biener <rguenther@suse.de>
2171 Jakub Jelinek <jakub@redhat.com>
2173 * cgraph.h (symbol_table): Just declare debug method here.
2174 * symtab.c (symbol_table::debug): Define.
2176 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2178 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2180 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2182 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2183 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2184 (*movdi_internal): Substitute Yi and Yj constraint with x
2185 and Ym and Yn constraint with y constraint. Update "isa"
2186 attribute and set "preferred_for_speed" attribute from
2187 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2188 (*movsi_internal): Ditto.
2189 (*movdf_internal): Ditto.
2190 (*movsf_internal): Ditto.
2191 (*zero_extendsidi2): Ditto.
2192 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2193 (sse2_loadld): Ditto.
2194 (*vec_extract<ssevecmodelower>_0): Ditto.
2195 (*vec_extractv4si_0_zext_sse4): Ditto.
2196 (vec_concatv2di): Ditto.
2197 (*vec_dup<mode>): Ditto.
2198 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2199 * config/i386/constraints.md (Yi): Remove.
2204 2018-04-26 Nathan Sidwell <nathan@acm.org>
2206 * dumpfile.c (dump_open): New.
2207 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2208 (dump_finish): Detect stdio/stderr by value not name.
2210 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2212 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2214 2018-04-26 Tom de Vries <tom@codesourcery.com>
2217 * config/nvptx/nvptx.c (verify_neutering_jumps)
2218 (verify_neutering_labels): New function
2219 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2221 2018-04-26 Tom de Vries <tom@codesourcery.com>
2224 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2225 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2228 2018-04-26 Richard Biener <rguenther@suse.de>
2229 Tom de Vries <tom@codesourcery.com>
2232 * lto-streamer-out.c (output_function): Fixup loops if required to match
2233 discovery done in the reader.
2235 2018-04-26 Richard Biener <rguenther@suse.de>
2237 PR tree-optimization/85116
2238 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2239 have a loop exit from the single latch predecessor. Remove
2240 case of header with just condition.
2241 (ch_base::copy_headers): Exclude infinite loops from any
2243 (pass_ch::execute): Record exits.
2245 2018-04-26 Richard Biener <rguenther@suse.de>
2247 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2248 prologue cost vector and pass it to vect_get_load_cost.
2249 (vect_get_peeling_costs_all_drs): Likewise.
2250 (vect_peeling_hash_get_lowest_cost): Likewise.
2251 (vect_enhance_data_refs_alignment): Likewise.
2253 2018-04-26 Richard Biener <rguenther@suse.de>
2256 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2257 checking of integer<->pointer conversions.
2258 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2259 sign-/zero-extending pointer types.
2260 (expand_omp_for_static_chunk): Likewise.
2262 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2263 Jean Lee <xiaoyur347@gmail.com>
2265 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2266 (TARGET_ASAN_SHADOW_OFFSET): Define.
2267 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2268 true for -fsanitize=address.
2270 2018-04-25 Mark Wielaard <mark@klomp.org>
2272 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2275 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2277 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2278 than "alu", remove explicit "memory" and "imm_disp" attributes.
2279 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2282 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2283 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2284 gen_lowpart_no_emit.
2286 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2289 * config/i386/i386.c (ix86_expand_builtin): Change memory
2290 operand to XI, extend p0 to Pmode.
2291 * config/i386/i386.md: Change unspec volatile and operand
2292 1 mode to XI, change operand 0 mode to P.
2294 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2296 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2297 GET_MODE_MASK before any checking.
2298 (nds32_can_use_bset_p): Likewise.
2299 (nds32_can_use_btgl_p): Likewise.
2301 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2303 * config/nds32/nds32-doubleword.md: New define_split pattern for
2304 illegal register number.
2306 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2308 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2310 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2312 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2314 2018-04-25 Richard Biener <rguenther@suse.de>
2316 * lto-streamer.h (LTO_major_version): Bump to 8.
2318 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2320 * BASE-VER: Set to 9.0.0.
2322 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2324 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2325 in __abskf2 and __powikf2.
2327 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2330 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2331 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2332 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2333 Use the above on operand 2. Reindent.
2334 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2336 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2339 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2341 * config/i386/i386.opt (mcet): Removed.
2342 * doc/install.texi: Remove -mcet documentation.
2343 * doc/invoke.texi: Likewise.
2345 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2348 * doc/install.texi: Remove -mcet from bootstrap-cet.
2350 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2353 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2354 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2358 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2359 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2360 containing a CONST_VECTOR.
2362 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2364 * doc/install.texi: Update newlib dependency for nvptx.
2366 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2369 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2370 instead of INTVAL when shifting x left.
2372 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2374 PR tree-optimization/85478
2375 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2376 vect_grouped_store_supported for single element vectors.
2378 2018-04-24 Richard Biener <rguenther@suse.de>
2381 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2382 load cost increase to the case of non-constant step.
2384 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2387 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2388 destination if any_malformed_asm.
2390 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2393 * expr.c (store_field): In the bitfield case, if the value comes from
2394 a function call and is returned in registers by means of a PARALLEL,
2395 do not change the mode of the temporary unless BLKmode and VOIDmode.
2397 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2399 PR rtl-optimization/85423
2400 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2401 dependencies to debug insns when the previous insn is non-debug.
2403 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2405 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2406 enums into a single definition.
2407 (fls): Fix predicates and printing.
2410 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2412 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2413 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2414 and short u6 immediate.
2415 (check_if_valid_sleep_operand): Remove.
2416 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2418 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2420 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2421 flag_always_save_lp condition.
2422 * config/nds32/nds32.opt (malways-save-lp): New option.
2424 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2426 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2427 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2428 * config/nds32/nds32.h
2429 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2430 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2432 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2434 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2435 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2437 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2438 Chung-Ju Wu <jasonwucj@gmail.com>
2440 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2442 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2443 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2445 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2447 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2449 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2451 * config/nds32/nds32-protos.h (nds32_data_alignment,
2452 nds32_local_alignment): Declare.
2453 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2454 nds32_local_alignment): New functions.
2455 (TARGET_CONSTANT_ALIGNMENT): Define.
2456 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2458 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2460 * config/nds32/nds32.c
2461 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2462 (TARGET_MODES_TIEABLE_P): Likewise.
2464 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2466 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2469 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2470 Chung-Ju Wu <jasonwucj@gmail.com>
2472 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2473 for unaligned access.
2474 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2476 * config/nds32/nds32-intrinsic.md: Likewise.
2477 * config/nds32/nds32_intrinsic.h: Likewise.
2478 * config/nds32/nds32.h (nds32_builtins): Likewise.
2479 * config/nds32/nds32.opt (munaligned-access): New option.
2480 * config/nds32/nds32.c (nds32_asm_file_start): Display
2481 flag_unaligned_access status.
2483 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2485 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2486 -mno-relax is present.
2487 * config/riscv/linux.h (LINK_SPEC): Ditto.
2489 2018-04-20 Martin Sebor <msebor@redhat.com>
2492 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2494 (gimple_fold_builtin_stxcpy_chk): Same.
2495 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2497 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2500 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2501 __powikf2 when long double is IEEE 128-bit.
2503 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2505 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2506 step to make sure stack always aligned.
2508 2018-04-20 Carl Love <cel@us.ibm.com>
2511 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2512 size check for arg0.
2514 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2515 Tom de Vries <tom@codesourcery.com>
2518 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2519 Emit insns for calls too.
2520 (nvptx_find_par): Always look for worker-level predecessor insn.
2521 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2523 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2524 (nvptx_process_pars): Propagate frames for calls.
2526 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2529 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2531 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2532 (ix86_handle_option): Don't handle OPT_mibt.
2533 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2535 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2537 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2538 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2539 (ix86_target_macros): Define __CET__ with flag_cf_protection
2540 for -fcf-protection.
2541 * config/i386/i386.c (isa2_opts): Remove -mibt.
2542 * config/i386/i386.h (TARGET_IBT): Removed.
2543 (TARGET_IBT_P): Likewise.
2544 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2545 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2546 * config/i386/i386.opt (mcet): Update help message.
2549 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2550 -mcet as an alias for -mshstk.
2552 2018-04-20 Richard Biener <rguenther@suse.de>
2555 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2556 complexity by forcing a single use of the multiply operand.
2558 2018-04-20 Martin Jambor <mjambor@suse.cz>
2561 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2562 recursion dependency to only apply to non-clones.
2564 2018-04-20 Martin Jambor <mjambor@suse.cz>
2567 * ipa-cp.c (create_specialized_node): Check that clones of
2568 self-recursive edges exist during IPA-CP.
2570 2018-04-19 Toon Moene <toon@moene.org>
2572 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2575 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2577 PR tree-optimization/85467
2578 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2579 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2580 VECTOR_CST element to type.
2582 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2585 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2586 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2587 (builtin_longjmp): Likewise.
2588 (save_stack_nonlocal): New pattern.
2589 (restore_stack_nonlocal): Likewise.
2591 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2594 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2595 Replace ASM_OUTPUT_LABEL with fprintf.
2597 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2600 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2601 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2602 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2603 define __IBT__ and __SHSTK__ for -fcf-protection.
2604 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2606 (ix86_trampoline_init): Likewise.
2607 (x86_output_mi_thunk): Likewise.
2608 (ix86_notrack_prefixed_insn_p): Likewise.
2609 (ix86_option_override_internal): Don't disallow -fcf-protection.
2610 * config/i386/i386.md (rdssp<mode>): Also enable for
2612 (incssp<mode>): Likewise.
2613 (nop_endbr): Likewise.
2614 * config/i386/i386.opt (mcet): Change help message to built-in
2618 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2619 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2620 enable CET built-in functions.
2622 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2624 * common/config/i386/i386-common.c
2625 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2626 OPTION_MASK_ISA_MOVDIRI_UNSET,
2627 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2628 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2629 * config.gcc (movdirintrin.h): New header.
2630 * config/i386/cpuid.h (bit_MOVDIRI,
2631 bit_MOVDIR64B): New bits.
2632 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2634 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2635 (VOID, PVOID, PCVOID)): New function types.
2636 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2637 __builtin_ia32_directstoreu_u64,
2638 __builtin_ia32_movdir64b): New builtins.
2639 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2640 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2642 (ix86_valid_target_attribute_inner_p): Ditto.
2643 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2644 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2645 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2646 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2647 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2648 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2649 (movdiri<mode>, movdir64b_<mode>): New.
2650 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2651 * config/i386/immintrin.h: Include movdirintrin.h.
2652 * config/i386/movdirintrin.h: New file.
2653 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2655 2018-04-19 Richard Biener <rguenther@suse.de>
2658 * cfg.c (clear_bb_flags): When loop state says we have
2659 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2661 2018-04-19 Richard Biener <rguenther@suse.de>
2663 PR tree-optimization/84737
2664 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2665 copying restrict info.
2666 (vect_setup_realignment): Use it.
2667 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2668 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2669 the first DR to all generated stores.
2670 (vectorizable_load): Likewise for loads.
2672 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2674 PR tree-optimization/85446
2675 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2676 the integral and pointer types to have the same precision.
2678 * doc/install.texi: Document --disable-cet being the default and
2681 2018-04-18 Martin Liska <mliska@suse.cz>
2683 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2686 2018-04-18 Martin Liska <mliska@suse.cz>
2689 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2692 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2693 arguments if they are comparable.
2695 2018-04-18 Martin Liska <mliska@suse.cz>
2698 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2701 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2704 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2707 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2708 ENDBR after calling __morestack.
2710 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2713 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2714 by using gcc_base_ver to generate a gcc_driver_version, and use
2715 it when generating GCC_DRIVER_NAME.
2716 * configure: Regenerate.
2718 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2721 * config.gcc: Obsolete powerpc*-*-*spe*.
2723 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2726 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2727 (stabstr_D): Change type of unum from unsigned int to
2728 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2731 2018-04-17 Jim Wilson <jimw@sifive.com>
2734 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2735 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2736 Set arg_pointer_offset after using pretend_args_size.
2738 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2740 PR rtl-optimization/85431
2741 * dse.c (record_store): Ignore zero width stores.
2744 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2745 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2746 __builtin_stack_restore rather than after it.
2747 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2748 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2749 argument instead of virtual_dynamic_stack_rtx.
2751 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2753 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2755 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2756 Add note to error message to explain internal mapping of overloaded
2757 built-in function name to non-overloaded built-in function name.
2758 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2761 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2764 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2765 where the inputs overlap with the output.
2767 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2769 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2770 (=v, v) alternative and explicit "memory" attribute.
2771 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2772 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2774 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2775 "sselog1" type instead of "sselog".
2776 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2777 "sselog". Remove explicit "memory" attribute.
2778 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2779 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2781 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2782 "sselog1" type instead of "sselog". Remove explicit "memory"
2784 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2785 use "sselog1" type instead of "sselog". Remove explicit "memory"
2787 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2788 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2790 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2791 "sselog1" type instead of "sselog". Remove explicit "memory"
2793 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2794 use "sselog1" type instead of "sselog". Remove explicit "memory"
2798 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2801 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2804 2018-04-17 Martin Jambor <mjambor@suse.cz>
2807 * ipa-cp.c (create_specialized_node): Call
2808 expand_all_artificial_thunks if necessary.
2810 2018-04-17 Martin Liska <mliska@suse.cz>
2813 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2814 in message, remote space in between '_G' and '('.
2816 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2819 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2820 avx512f_vmcmp<mode>3<round_saeonly_name>,
2821 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2822 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2823 avx512f_rndscale<mode><round_saeonly_name>,
2824 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2825 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2826 Use %<iptr>2 instead of %2 for -masm=intel.
2827 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2828 avx512f_vcvttss2usi<round_saeonly_name>,
2829 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2831 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2832 avx512f_vcvttsd2usi<round_saeonly_name>,
2833 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2834 Use %q1 instead of %1 for -masm=intel.
2835 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2836 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2837 of %3 for -masm=intel.
2838 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2840 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2842 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2844 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2845 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2847 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2849 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2850 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2851 %g1 and one with %0 and %1.
2852 (avx512er_vmrcp28<mode><round_saeonly_name>,
2853 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2855 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2856 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2857 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2858 of %0 and %{%4%} for -masm=intel.
2859 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2860 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2861 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2862 order of %0 and %{%5%}%{z%} for -masm=intel.
2864 2018-04-17 Jan Hubicka <jh@suse.cz>
2867 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2869 2018-04-17 Martin Liska <mliska@suse.cz>
2872 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2873 for target_clone error message. Make default implementation
2874 clone to be a local declaration.
2875 (separate_attrs): Add new argument and check for an empty
2877 (expand_target_clones): Handle it.
2878 (ipa_target_clone): Make redirection just for target_clones
2881 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2882 Tom de Vries <tom@codesourcery.com>
2885 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2886 tiled basic blocks without omp continue statements.
2888 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2891 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2892 vec_extract expression as having side effects to make sure it gets
2895 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2898 * config/i386/i386.c (get_builtin_code_for_version): Check
2901 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2904 * gcc/config.gcc: Support "skylake".
2905 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2907 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2908 (processor_target_table): Add "skylake".
2909 (ix86_option_override_internal): Add "skylake".
2910 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2911 PROCESSOR_CANNONLAKE.
2912 (get_builtin_code_for_version): Fix priority for
2913 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2914 PROCESSOR_SKYLAKE-AVX512.
2915 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2916 (processor_type): Add PROCESSOR_SKYLAKE.
2918 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2919 Jason Merrill <jason@redhat.com>
2922 * convert.c (convert_to_integer_1): Use direct recursion for
2923 enumeral types and types with a precision less than the number
2924 of bits in their mode.
2926 2018-04-16 Julia Koval <julia.koval@intel.com>
2929 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2930 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2932 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2935 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2936 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2937 and -mno-direct-move.
2939 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2942 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2943 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2944 Ensure negative shifts result in {0}.
2946 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2948 PR rtl-optimization/79916
2949 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2950 regs (if any) to define how to gnerate SD moves when LRA is in
2953 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2955 PR rtl-optimization/85393
2956 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2957 * except.c (expand_dw2_landing_pad_for_region): Make static.
2958 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2959 a label and unconditional jump to old_bb, rather than
2960 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2963 PR rtl-optimization/85376
2964 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2965 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2966 instead of a specific value.
2968 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2969 Bin Cheng <bin.cheng@arm.com>
2971 PR tree-optimization/82965
2972 PR tree-optimization/83991
2973 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2974 by_profile_only parameter.
2975 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2976 information if the loop was predicted to iterate too many times.
2977 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2979 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2982 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2985 2018-04-13 Martin Liska <mliska@suse.cz>
2986 Jakub Jelinek <jakub@redhat.com>
2989 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2990 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2991 * builtins.c (expand_builtin_memory_copy_args): Use
2992 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2993 handle dest_addr == pc_rtx.
2995 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2998 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2999 asked to not generate direct moves.
3000 (fix_trunc<mode>si2_stfiwx): Similar.
3001 (fix_trunc<mode>si2_internal): Similar.
3003 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3006 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3007 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3008 lookup if dest in some wider mode is known to be const0_rtx and
3009 if so, record permanent equivalence for it to be ZERO_EXTEND of
3010 the narrower mode destination.
3012 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3014 * lto-streamer-out.c (output_function): Revert 259346.
3015 * omp-expand.c (expand_oacc_for): Likewise.
3017 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3019 PR rtl-optimization/85354
3020 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3021 * sel-sched.c (sel_global_init): ... here.
3023 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3026 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3027 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3028 mode for PE-COFF targets.
3029 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3030 (i386_pe_asm_lto_end): Likewise.
3031 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3032 (TARGET_ASM_LTO_END): Likewise.
3033 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3034 (i386_pe_asm_lto_start): New function.
3035 (i386_pe_asm_lto_end): Likewise.
3037 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3038 Richard Biener <rguenther@suse.de>
3041 * lto-streamer-out.c (output_function): Fix CFG loop state before
3043 * omp-expand.c (expand_oacc_for): Handle calls to internal
3044 functions like regular functions.
3046 2018-04-12 Richard Biener <rguenther@suse.de>
3049 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3050 for the early LTO debug to properly generate references to it
3051 during DIE emission. Do not re-use that for the skeleton for
3053 (dwarf2out_early_finish): Likewise.
3055 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3058 * config/i386/sse.md
3059 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3060 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3061 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3062 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3063 and output is a reg, avoid creating invalid lowpart subreg, but
3064 instead split into a 512-bit move. Don't split if not AVX512VL,
3065 input is xmm16+ reg and output is a mem.
3066 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3067 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3068 xmm16+ reg and output is a mem.
3070 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3072 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3073 also for flag_dwarf2_cfi_asm.
3075 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3077 PR rtl-optimization/85342
3078 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3079 a bool scalar var inside of the loop instead. Don't try to update
3080 recog_data.operand after failed apply_change_group.
3082 2018-04-12 Tom de Vries <tom@codesourcery.com>
3085 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3086 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3087 array with flexible array member as array without given dimension.
3088 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3089 argument for undefined param to true.
3091 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3094 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3095 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3096 from PowerPC section.
3097 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3098 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3099 help text that is too long.
3100 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3101 help text that is too long.
3102 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3103 help text that is too long.
3105 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3107 * config/alpha/alpha.md (stack_probe_internal): Rename
3108 from "probe_stack". Update all callers.
3110 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3112 PR rtl-optimization/84566
3113 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3114 sched_macro_fuse_insns.
3116 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3119 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3120 (compute_block_dependences): ... from here.
3122 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3124 PR tree-optimization/85331
3125 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3126 from int to HOST_WIDE_INT.
3128 2018-04-11 Martin Jambor <mjambor@suse.cz>
3131 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3132 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3133 not the same as the source val.
3134 (cgraph_edge_brings_value_p): New parameter.
3135 (gather_edges_for_value): Pass destination value to
3136 cgraph_edge_brings_value_p.
3137 (perhaps_add_new_callers): Likewise.
3138 (get_info_about_necessary_edges): Likewise and exclude values brought
3139 only by self-recursive edges.
3140 (create_specialized_node): Redirect only clones of self-calling edges.
3141 (+self_recursive_pass_through_p): New function.
3142 (find_more_scalar_values_for_callers_subset): Use it.
3143 (find_aggregate_values_for_callers_subset): Likewise.
3144 (known_aggs_to_agg_replacement_list): Removed.
3145 (decide_whether_version_node): Re-calculate known constants for all
3146 remaining context clones.
3148 2018-04-11 Richard Biener <rguenther@suse.de>
3151 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3152 from early DWARF output.
3153 (dwarf2out_early_finish): Output line info unconditionally into
3154 early DWARF and add reference to it.
3156 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3159 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3160 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3161 other than V2DFmode using iptr mode attribute.
3162 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3164 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3166 PR rtl-optimization/84659
3167 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3169 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3172 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3174 (output_loc_list): Pass address of a dummy size variable even in the
3175 locview handling loop.
3176 (index_location_lists): Add comment on why skip_loc_list_entry can't
3179 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3182 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3185 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3188 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3189 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3190 and -mstring-compare-inline-limit.
3192 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3195 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3196 for stack clash protection in a register whenever we need it to be in
3199 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3201 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3202 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3204 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3207 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3209 (mlong-double-): Ditto.
3210 * config/rs6000/sysv4.opt (msdata=): Ditto.
3211 (mtls-size=): Ditto.
3213 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3215 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3216 erroneous entries for
3217 "vector int vec_ldl (int, long int *)", and
3218 "vector unsigned int vec_ldl (int, unsigned long int *)".
3219 Add comments and entries for
3220 "vector bool char vec_ldl (int, bool char *)",
3221 "vector bool short vec_ldl (int, bool short *)",
3222 "vector bool int vec_ldl (int, bool int *)",
3223 "vector bool long long vec_ldl (int, bool long long *)",
3224 "vector pixel vec_ldl (int, pixel *)",
3225 "vector long long vec_ldl (int, long long *)",
3226 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3227 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3228 type tree bool_long_long_type_node and correct definition of
3229 bool_V2DI_type_node to make reference to this new type tree.
3230 (rs6000_mangle_type): Replace erroneous reference to
3231 bool_long_type_node with bool_long_long_type_node.
3232 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3233 comments to emphasize sign distinctions for char and int types and
3234 replace RS6000_BTI_bool_long constant with
3235 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3236 use of RS6000_BTI_pixel.
3237 (bool_long_type_node): Remove this macro definition.
3238 (bool_long_long_type_node): New macro definition
3240 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3242 PR rtl-optimization/85300
3243 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3244 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3245 simplify_unary_operation fails.
3247 2018-04-10 Martin Liska <mliska@suse.cz>
3249 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3250 cgraph_edge and ipa_ref.
3252 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3256 * config/i386/sse.md
3257 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3258 computation of the VEC_MERGE selector from mask.
3259 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3260 Fix decoding of the VEC_MERGE selector into mask.
3262 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3264 PR tree-optimization/85286
3265 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3267 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3269 * final.c (final_1): Set insn_last_address as well as
3270 insn_current_address.
3272 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3275 * explow.c (emit_stack_probe): Call validize_mem on memory location
3276 before passing it to gen_probe_stack. Create address operand and
3277 legitimize it for the probe_stack_address case.
3279 2018-04-09 Jan Hubicka <jh@suse.cz>
3282 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3283 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3284 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3286 (free_lang_data): Rebuild type inheritance graph.
3288 2018-04-09 Martin Sebor <msebor@redhat.com>
3290 * invoke.texi (-finline-small-functions): Mention other optimization
3292 (-findirect-inlining, -fpartial-inlining): Same.
3293 (-finline-functions-called-once): Same.
3294 (-freorder-blocks-and-partition): Same.
3296 2018-04-09 Jan Hubicka <jh@suse.cz>
3299 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3300 jumps; choose last target that matches the criteria (i.e.
3301 no partition changes for non-crossing jumps).
3302 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3303 support for redirecting crossing jumps to non-crossing.
3305 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3307 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3308 also for naked functions.
3310 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3312 * config/arc/arc.md (add_shift): New pattern.
3313 (add_shift2): Likewise.
3314 (sub_shift): Likewise.
3315 (sub_shift_cmp0_noout): Likewise.
3316 (compare_si_ashiftsi): Likewise.
3317 (xbfu_cmp0_noout): New combine pattern.
3318 (xbfu_cmp0"): Likewise.
3319 (movsi_set_cc_insn): Place the predicable variant first.
3320 (commutative_binary_cmp0_noout): Remove clobber.
3321 (commutative_binary_cmp0): New pattern.
3322 (noncommutative_binary_cmp0): Likewise.
3323 (noncommutative_binary_cmp0_noout): Likewise.
3324 (noncommutative_binary_comparison_result_used): Removed.
3325 (rsub_cmp0): New pattern.
3326 (rsub_cmp0_noout): Likewise.
3327 (extzvsi): Changed, keep only meaningful variants.
3328 (SQH, SEZ): New iterators.
3329 (SQH_postfix): New mode attribute.
3330 (SEZ_prefix): New code attribute.
3331 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3332 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3333 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3335 (noncommutative_operator): Check the availability of barrel
3338 2018-04-09 Richard Biener <rguenther@suse.de>
3340 PR tree-optimization/85284
3341 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3342 Only use the niter constraining form of simple_iv when the exit
3345 2018-04-09 Tom de Vries <tom@codesourcery.com>
3348 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3349 (define_expand "*memory_barrier"): New define_expand.
3350 (define_insn "memory_barrier"): New insn.
3352 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3354 PR rtl-optimization/80463
3355 PR rtl-optimization/83972
3356 PR rtl-optimization/83480
3358 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3359 correct producer for the insn.
3360 (tidy_control_flow): Fixup seqnos in case of debug insns.
3362 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3364 PR rtl-optimization/83913
3366 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3367 different sched-times when merging exprs.
3369 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3371 PR rtl-optimization/83962
3373 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3374 tidy_fallthru_edge and tidy_control_flow.
3376 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3378 PR rtl-optimization/83530
3380 * sel-sched.c (force_next_insn): New global variable.
3381 (remove_insn_for_debug): When force_next_insn is true, also leave only
3382 next insn in the ready list.
3383 (sel_sched_region): When the region wasn't scheduled, make another pass
3384 over it with force_next_insn set to 1.
3386 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3388 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3390 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3391 for interrupt control.
3392 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3393 functions for interrupt control.
3394 * config/nds32/nds32-intrinsic.md: Likewise.
3395 * config/nds32/nds32_intrinsic.h: Likewise.
3396 * config/nds32/nds32.h (nds32_builtins): Likewise.
3398 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3400 * config/nds32/nds32.c (nds32_init_machine_status,
3401 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3402 strict_aligned_p field.
3403 (nds32_expand_to_rtl_hook): New function.
3404 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3405 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3407 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3408 Chung-Ju Wu <jasonwucj@gmail.com>
3410 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3411 * config/nds32/nds32-n7.md: New file.
3412 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3413 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3415 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3416 * config/nds32/nds32.md (pipeline_model): Add n7.
3417 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3418 * config/nds32/pipelines.md: Include n7 settings.
3420 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3421 Chung-Ju Wu <jasonwucj@gmail.com>
3423 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3424 * config/nds32/nds32-e8.md: New file.
3425 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3426 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3428 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3429 * config/nds32/nds32.md (pipeline_model): Add e8.
3430 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3431 * config/nds32/pipelines.md: Include e8 settings.
3433 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3434 Chung-Ju Wu <jasonwucj@gmail.com>
3436 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3437 * config/nds32/nds32-n8.md: New file.
3438 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3439 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3441 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3442 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3443 * config/nds32/nds32.md (pipeline_model): Add n8.
3444 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3445 * config/nds32/pipelines.md: Include n8 settings.
3447 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3448 Chung-Ju Wu <jasonwucj@gmail.com>
3450 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3451 * config/nds32/nds32-n9-2r1w.md: New file.
3452 * config/nds32/nds32-n9-3r2w.md: New file.
3453 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3454 nds32_register_ports): New or modify for cpu n9.
3455 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3457 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3458 * config/nds32/nds32-utils.c: New file.
3459 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3460 TARGET_MUL_SLOW): Define.
3461 * config/nds32/nds32.md (pipeline_model): New attribute.
3462 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3463 New options that support cpu n9.
3464 * config/nds32/pipelines.md: Include n9 settings.
3465 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3467 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3469 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3470 information if necessary.
3471 (output_cond_branch_compare_zero): Likewise.
3472 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3473 (nds32_target_alignment): Refine for alignment.
3474 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3475 (FUNCTION_BOUNDARY): Modify.
3476 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3478 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3480 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3482 * config/nds32/constants.md (unspec_volatile_element): Add values for
3483 TLB operation and data prefetch.
3484 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3485 functions for TLB operation and data prefetch.
3486 * config/nds32/nds32-intrinsic.md: Likewise.
3487 * config/nds32/nds32_intrinsic.h: Likewise.
3488 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3489 (nds32_print_operand): Likewise.
3490 * config/nds32/nds32.h (nds32_builtins): Likewise.
3492 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3493 Andrew Pinski <pinsika@gcc.gnu.org>
3496 * match.pd: Use constant_boolean_node of correct type instead of
3497 boolean_true_node or boolean_false_node for simplifying
3498 pointer comparisons to zero.
3500 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3502 PR tree-optimization/80021
3503 * tree.c (verify_type_variant): Make error call in verify_variant_match
3504 translatable and remove final full stop.
3506 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3508 * config/nds32/constants.md (unspec_volatile_element): Add
3509 UNSPEC_VOLATILE_EH_RETURN.
3510 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3511 nds32_output_stack_pop): Support dwarf exception handling process.
3512 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3513 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3514 exception handling process.
3515 (nds32_compute_stack_frame): Likewise.
3516 (nds32_return_addr_rtx): Likewise.
3517 (nds32_initial_elimination_offset): Likewise.
3518 (nds32_expand_prologue): Likewise.
3519 (nds32_expand_epilogue): Likewise.
3520 (nds32_dynamic_chain_address): New function.
3521 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3523 (DYNAMIC_CHAIN_ADDRESS): Define.
3524 (EH_RETURN_DATA_REGNO): Define.
3525 (EH_RETURN_STACKADJ_RTX): Define.
3526 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3527 patterns for dwarf exception handling.
3529 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3531 * config/nds32/nds32.h: Clean up obsolete macros.
3533 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3535 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3536 Add enum values for particular instructions.
3537 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3538 particular intrinsic functions.
3539 * config/nds32/nds32-intrinsic.md: Likewise.
3540 * config/nds32/nds32_intrinsic.h: Likewise.
3541 * config/nds32/nds32.h (nds32_builtins): Likewise.
3542 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3543 (btst, ave): New patterns for particular instructions.
3545 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3547 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3548 Add enum values for atomic load/store and memory sync.
3549 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3551 * config/nds32/nds32-intrinsic.md: Likewise.
3552 * config/nds32/nds32_intrinsic.h: Likewise.
3553 * config/nds32/nds32.h (nds32_builtins): Likewise.
3555 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3557 PR tree-optimization/85257
3558 * fold-const.c (native_encode_vector): If not all elts could fit
3559 and off is -1, return 0 rather than offset.
3560 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3561 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3562 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3563 adjust buffer in native_interpret_expr call.
3565 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3567 * config/nds32/constants.md (unspec_volatile_element): Add cache
3568 control enum values.
3569 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3570 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3571 * config/nds32/nds32.c (nds32_cctl_names): New.
3572 (nds32_print_operand): Handle cache control register names.
3573 * config/nds32/nds32.h (nds32_builtins): New enum values.
3574 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3576 * config/nds32/nds32.md (type): Add mmu.
3577 * config/nds32/pipelines.md (simple_insn): Add mmu.
3579 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3581 * config/nds32/nds32.md (type): Remove call.
3582 * config/nds32/pipelines.md (simple_insn): Likewise.
3584 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3586 * config/nds32/constants.md (unspec_volatile_element): Add
3587 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3588 UNSPEC_VOLATILE_FMFCFG.
3589 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3590 description for fmfcfg and fmfcsr.
3591 (bdesc_1arg): Add fmtcsr.
3592 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3593 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3594 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3595 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3596 unspec_fmfcfg): New patterns.
3597 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3598 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3599 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3600 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3601 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3602 __nds32__fmfcfg): Define.
3604 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3606 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3607 intrinsic register names.
3608 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3609 intrinsic register enum values and macros.
3611 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3613 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3614 for load/store addressing form.
3615 (nds32_print_operand_address): Likewise.
3617 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3620 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3621 based on LABEL_REF. Remove useless assertion.
3622 (pic_address_needs_scratch): Fix formatting.
3623 (sparc_legitimize_pic_address): Minor tweaks.
3624 (sparc_delegitimize_address): Adjust assertion accordingly.
3625 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3626 into symbolic_operand.
3627 (movsi_high_pic_label_ref): Likewise.
3628 (movsi_lo_sum_pic_label_ref): Likewise.
3629 (movdi_pic_label_ref): Likewise.
3630 (movdi_high_pic_label_ref): Likewise.
3631 (movdi_lo_sum_pic_label_ref): Likewise.
3633 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3635 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3636 custom LIB_SPEC setup.
3638 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3639 Kito Cheng <kito.cheng@gmail.com>
3641 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3642 * config/riscv/freebsd.h: New.
3644 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3646 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3647 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3650 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3651 Kito Cheng <kito.cheng@gmail.com>
3653 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3654 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3655 * config/nds32/nds32-protos.h (nds32_output_call,
3656 nds32_output_return): Declare.
3657 * config/nds32/nds32.md: Refine all the call and return patterns.
3659 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3662 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3663 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3665 PR rtl-optimization/84872
3666 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3667 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3670 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3672 * expr.c (copy_blkmode_to_reg): Revert 254862.
3673 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3675 2018-04-06 Richard Biener <rguenther@suse.de>
3678 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3679 after seeing a component reference with an adjacent field. Treat
3680 refs to arrays at struct end of external decls similar to
3681 refs to unconstrained commons.
3683 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3686 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3687 look through SAVE_EXPRs with non-side-effects argument. Adjust
3689 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3692 2018-04-06 Richard Biener <rguenther@suse.de>
3695 * alias.c (find_base_term): New wrapper around find_base_term
3696 unwinding CSELIB_VAL_PTR changes.
3697 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3700 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3702 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3704 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3705 constant definitions.
3706 ("nop"): lr 0,0 -> nopr r0
3707 ("nop_lr0", "nop_lr1"): New insn definitions.
3709 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3711 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3712 NDS32_V3PUSH_AVAILABLE_P macro.
3714 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3715 Chung-Ju Wu <jasonwucj@gmail.com>
3717 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3718 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3719 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3720 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3721 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3722 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3723 * config/nds32/constraints.md: New constraints and checking for hard
3724 float configuration.
3725 * config/nds32/iterators.md: New mode iterator and attribute for hard
3726 float configuration.
3727 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3729 * config/nds32/nds32-fpu.md: New file.
3730 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3731 deal with hard float code generation.
3732 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3734 (abi_type, float_reg_number): New enum type.
3735 * config/nds32/nds32-predicates.c: New predicates for hard float.
3736 * config/nds32/nds32-protos.h: Declare functions for hard float.
3737 * config/nds32/nds32.c: Implementation for hard float configuration.
3738 * config/nds32/nds32.h: Definitions for hard float configuration.
3739 * config/nds32/nds32.md: Include hard float machine description and
3740 modify patterns for hard float configuration.
3741 * config/nds32/nds32.opt: New options for hard float configuration.
3742 * config/nds32/predicates.md: New predicates for hard float
3745 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3747 * common/config/nds32/nds32-common.c
3748 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3750 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3753 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3754 CONSTRUCTOR_ELT (ctor, ...)->value.
3756 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3759 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3761 2018-04-05 Tom de Vries <tom@codesourcery.com>
3764 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3767 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3768 Kito Cheng <kito.cheng@gmail.com>
3770 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3771 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3772 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3773 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3775 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3776 Kito Cheng <kito.cheng@gmail.com>
3778 * config/nds32/constraints.md (Ufe): New memory constraint.
3779 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3780 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3781 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3783 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3784 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3786 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3788 * config/nds32/nds32.md: Use optimize_size in the condition for
3789 alu-shift instructions.
3791 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3793 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3795 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3797 * config/nds32/nds32.md (negsi2): Refine pattern.
3799 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3800 Chung-Ju Wu <jasonwucj@gmail.com>
3802 * config/nds32/iterators.md (shift_rotate): New code iterator.
3803 (shift): New code attribute.
3804 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3805 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3806 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3807 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3808 bit-wise operations.
3809 (andsi3, *andsi3): Ditto.
3810 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3811 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3812 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3813 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3814 nds32_ior_operand, nds32_xor_operand): New predicates.
3816 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3818 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3819 (addsi3, subsi3): ... this.
3821 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3823 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3825 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3827 * config/nds32/nds32.md: Adjust indention.
3829 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3831 * config/nds32/nds32.md (feature): New attribute.
3833 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3835 * config/nds32/nds32.md (subtype): New attribute.
3837 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3840 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3841 expansion to perform a bitwise AND of the argument followed by a
3842 boolean negation of the result.
3844 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3846 PR rtl-optimization/84878
3847 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3848 the basic block. Assert the use reference is not artificial and that
3849 it has an associated insn.
3851 2018-04-04 Michael Matz <matz@suse.de>
3853 * builtins.c (compute_objsize): Pass correct operand
3854 to array_at_struct_end_p.
3856 2018-04-04 Richard Biener <rguenther@suse.de>
3859 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3860 from contexts for DINFO_LEVEL_TERSE and below.
3862 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3864 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3865 resiter_operand condition.
3866 * config/nds32/nds32.md (*move<mode>): Ditto.
3868 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3869 Monk Chiang <sh.chiang04@gmail.com>
3871 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3873 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3875 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3877 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3878 Kito Cheng <kito.cheng@gmail.com>
3880 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3881 nds32_cond_code_str, output_cond_branch,
3882 output_cond_branch_compare_zero, nds32_expand_cbranch,
3883 nds32_expand_cstore, nds32_expand_movcc,
3884 nds32_output_cbranchsi4_equality_zero,
3885 nds32_output_cbranchsi4_equality_reg,
3886 nds32_output_cbranchsi4_equality_reg_or_const_int,
3887 nds32_output_cbranchsi4_greater_less_zero: New functions.
3888 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3889 nds32_expand_cstore, nds32_expand_movcc,
3890 nds32_output_cbranchsi4_equality_zero,
3891 nds32_output_cbranchsi4_equality_reg,
3892 nds32_output_cbranchsi4_equality_reg_or_const_int,
3893 nds32_output_cbranchsi4_greater_less_zero): Declare.
3894 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3895 nds32_rimm11s_operand): New predicates.
3896 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3897 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3900 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3902 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3903 * config/nds32/nds32.md: Ditto.
3904 * config/nds32/pipelines.md: Ditto.
3906 2018-04-04 Richard Biener <rguenther@suse.de>
3908 PR tree-optimization/85168
3909 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3910 propagating abnormals.
3912 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3914 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3916 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3917 Kito Cheng <kito.cheng@gmail.com>
3919 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3920 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3921 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3922 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3923 * config/nds32/nds32.md (sibcall_internal): New.
3924 (sibcall_register): Remove.
3925 (sibcall_immediate): Remove.
3926 (sibcall_value_internal): New.
3927 (sibcall_value_register): Remove.
3928 (sibcall_value_immediate): Remove.
3929 * config/nds32/predicates.md (nds32_general_register_operand): New.
3930 (nds32_call_address_operand): New.
3932 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3934 PR rtl-optimization/85167
3935 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3936 bb_defs if *split_p, instead preinitialize it to NULL.
3938 PR tree-optimization/85156
3939 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3940 evaluating the argument multiple times.
3942 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3944 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3946 (_mm_cvtpd_ps): Likewise.
3947 (_mm_cvttpd_epi32): Likewise.
3948 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3949 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3950 vector, pixel, and bool following altivec.h include.
3952 2018-04-03 Martin Sebor <msebor@redhat.com>
3954 * doc/extend.texi (Common Function Attributes): Clarify.
3955 (const attribute): Likewise.
3956 (pure attribute): Likewise.
3958 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3961 * config/i386/i386.c (ix86_expand_vector_set): Use
3962 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3964 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3966 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3967 instructions when changing rounding bits to preserve precision bits
3968 in the x87 control word.
3970 2018-04-03 Martin Liska <mliska@suse.cz>
3972 PR tree-optimization/82491
3973 * rtl.h (strip_offset_and_add): Replace += suboffset with
3974 poly_uint64 () + suboffset.
3976 2018-03-29 Martin Liska <mliska@suse.cz>
3977 Martin Jambor <mjambor@suse.cz>
3980 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3981 param_type is not an integral or pointer type.
3983 2018-04-03 Richard Biener <rguenther@suse.de>
3985 * sese.h (recompute_all_dominators): Remove.
3987 2018-04-02 Martin Sebor <msebor@redhat.com>
3989 * doc/invoke.texi (-Wrestrict): Fix typos.
3991 2018-04-02 Jim Wilson <jimw@sifive.com>
3993 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3994 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3995 (<optab>di3, <optab>si3_extend): Likewise.
3996 (<optab>si3_mask, <optab>si3_mask_1): New.
3997 (<optab>di3_mask, <optab>di3_mask_1): New.
3998 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3999 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4000 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4002 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4004 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4007 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4009 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4010 (nds32_canonicalize_comparison): New function.
4012 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4013 Kito Cheng <kito.cheng@gmail.com>
4014 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4016 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4017 * config/nds32/constants.md (unspec_volatile_element): Add
4018 UNSPEC_VOLATILE_RELAX_GROUP.
4019 * config/nds32/nds32-relax-opt.c: New file.
4020 * config/nds32/nds32-predicates.c
4021 (nds32_symbol_load_store_p): New function.
4022 * config/nds32/nds32-protos.h
4023 (nds32_symbol_load_store_p): Declare function.
4024 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4025 * config/nds32/nds32.c
4026 (nds32_register_pass): New function to register pass.
4027 (nds32_register_passes): New function to register passes.
4028 * config/nds32/nds32.md (relax_group): New pattern.
4029 * config/nds32/nds32.opt (mrelax-hint): New option.
4030 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4032 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4034 * config/nds32/t-nds32: Modify files dependency.
4036 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4038 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4039 (PROFILE_HOOK): Define its implementation.
4041 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4043 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4044 type and 32-bit size.
4046 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4049 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4050 (V_128_256): New mode iterator.
4051 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4052 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4053 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4055 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4056 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4058 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4061 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4062 NaN inputs correctly.
4064 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4067 * config/rs6000/vsx.md (??r): New mode attribute.
4068 (*vsx_mov<mode>_64bit): Use it.
4069 (*vsx_mov<mode>_32bit): Likewise.
4071 2018-03-30 Martin Sebor <msebor@redhat.com>
4073 PR tree-optimization/84818
4074 * builtins.c (check_access): Use warning_n.
4076 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4079 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4081 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4084 2018-03-30 Julia Koval <julia.koval@intel.com>
4087 * x86-tune.def (movx, partial_reg_dependency): Enable for
4090 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4093 * lra-constraints.c (process_alt_operands): Move setting
4094 this_alternative_matches below.
4096 2018-03-29 Martin Liska <mliska@suse.cz>
4099 * doc/invoke.texi: Document how LTO works with debug info.
4100 Describe auto-load support of binutils. Mention 'x86-64'
4101 as valid option value of -march option.
4103 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4105 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4108 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4109 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4110 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4113 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4116 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4117 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4118 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4119 for RS6000_BTM_POWERPC64.
4120 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4121 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4122 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4127 2018-03-28 Carl Love <cel@us.ibm.com>
4130 2017-09-27 Carl Love <cel@us.ibm.com>
4132 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4133 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4134 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4137 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4139 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4140 instead of __vector bool.
4141 (_mm_max_pu8): Likewise.
4142 (_mm_min_pi16): Likewise.
4144 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4147 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4148 (DIVWEUO): Likewise.
4150 (DIVDEUO): Likewise.
4151 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4152 DIVWEUO and DIVDEUO.
4153 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4154 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4155 (div_extend): Likewise.
4156 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4158 (__builtin_divweuo): Likewise.
4159 (__builtin_divdeo): Likewise.
4160 (__builtin_divdeuo): Likewise.
4162 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4165 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4166 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4168 PR tree-optimization/82004
4169 * gimple-match-head.c (optimize_pow_to_exp): New function.
4170 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4171 Don't fold to exp if optimize_pow_to_exp is false.
4173 2018-03-28 Martin Liska <mliska@suse.cz>
4176 * calls.c (initialize_argument_information): Fix trailing space.
4177 * common.opt: Fix typo and provide better explanation for
4178 -fsanitize-coverage option.
4179 * config/i386/i386.opt: Fix typo.
4181 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4182 Martin Liska <mliska@suse.cz>
4185 * gimplify.c (asan_poison_variable): Don't do the check for
4186 gimplify_omp_ctxp here.
4187 (gimplify_decl_expr): Do it here.
4188 (gimplify_target_expr): Likewise.
4190 2018-03-28 Martin Liska <mliska@suse.cz>
4193 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4194 chkp_type_bounds_count if MPX is not enabled.
4196 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4198 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4200 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4203 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4204 function to create the function decl for complex long double
4205 multiply and divide for -mabi=ieeelongdouble.
4206 (init_float128_ieee): Call it.
4208 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4211 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4212 -fcf-protection=branch -mibt.
4213 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4215 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4218 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4220 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4223 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4224 extern array declarations.
4226 2018-03-27 Richard Biener <rguenther@suse.de>
4229 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4230 explicit single_use checks.
4232 2018-03-27 Richard Biener <rguenther@suse.de>
4234 PR tree-optimization/85082
4235 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4238 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4240 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4241 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4242 Turn on fasynchronous-unwind-tables and funwind-tables.
4244 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4247 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4248 (*bmi_blsr_<mode>_ccz): Ditto.
4250 2018-03-26 Tom de Vries <tom@codesourcery.com>
4252 PR tree-optimization/85063
4253 * omp-general.c (offloading_function_p): New function. Factor out
4255 * omp-offload.c (pass_omp_target_link::gate): ... here.
4256 * omp-general.h (offloading_function_p): Declare.
4257 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4258 with attribute omp declare target for offloading functions.
4260 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4262 PR tree-optimization/84005
4263 * tree-data-ref.h (get_base_for_alignment): Declare.
4264 * tree-data-ref.c (get_base_for_alignment_1): New function.
4265 (get_base_for_alignment): Likewise.
4266 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4267 get_base_for_alignment to find a suitable base object, instead
4268 of always using drb->base_address.
4270 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4273 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4274 known size by default.
4276 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4279 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4280 and non BLKmode operands.
4282 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4285 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4286 Clean up attributes.
4288 2018-03-23 Richard Biener <rguenther@suse.de>
4291 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4292 we are going to emit early debug for LTO.
4294 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4297 * function.c (match_asm_constraints_1): Don't optimize if input
4298 doesn't satisfy general_operand predicate for output's mode.
4301 * alias.c (write_dependence_p): Don't require for x_canonicalized
4302 non-VOIDmode if x has VOIDmode.
4305 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4306 just don't try to optimize it rather than assert it never happens.
4308 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4310 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4311 macro expansions for definition of ST_INTERNAL_<mode> and
4312 LD_INTERNAL_<mode> builtins.
4313 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4315 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4317 (altivec_expand_st_builtin): Likewise.
4318 (altivec_expand_builtin): Remove calls to deleted functions.
4319 (rs6000_address_for_altivec): Delete this function.
4320 * config/rs6000/vector.md: Remove expands for
4321 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4323 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4326 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4327 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4328 re-computing once computed.
4329 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4330 (arm_init_machine_status): Initialize
4331 machine->static_chain_stack_bytes.
4333 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4336 * doc/extend.texi: Add four new prototypes for vec_ld.
4337 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4338 definitions for more logical presentation.
4339 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4340 entries for V1TI variants of __builtin_altivec_ld builtin.
4341 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4342 handling of V1TI variant of LVX icode pattern.
4343 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4344 (rs6000_gimple_fold_builtin): Likewise.
4345 (altivec_init_builtins): Add code to define
4346 __builtin_altivec_lvx_v1ti function.
4348 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4351 * function.c (match_asm_constraints_1): Don't do the optimization
4352 if input isn't a REG, SUBREG, MEM or constant.
4354 2018-03-22 Tom de Vries <tom@codesourcery.com>
4356 PR tree-optimization/84956
4357 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4358 bb_has_abnormal_pred.
4360 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4363 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4364 DECL_INITIAL (decl) to decl at the end.
4365 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4368 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4370 * doc/extend.texi (__builtin_tgmath): Document when complex
4371 integer types are treated as _Complex _Float64.
4373 2018-03-21 Tom de Vries <tom@codesourcery.com>
4375 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4377 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4379 PR tree-optimization/84960
4380 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4381 if it is ENTRY block, move them into single succ of ENTRY in that case.
4383 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4385 PR tree-optimization/84811
4386 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4387 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4388 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4389 (known_subrange_p): Update accordingly. Cast each value involved
4390 in the size comparison, rather than casting the result of the
4393 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4395 PR tree-optimization/84982
4396 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4397 by flipping the least significant bit rather than all bits from
4398 bitpos to bitpos + bitsize - 1.
4400 2018-03-21 Nathan Sidwell <nathan@acm.org>
4402 * doc/extend.texi (Deprecated Features): Remove mention of
4403 long-deleted deprecations.
4405 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4408 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4409 * configure: Regenerate.
4411 2018-03-21 Tom de Vries <tom@codesourcery.com>
4413 PR tree-optimization/83126
4414 * tree-parloops.c (num_phis): New function.
4415 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4417 2018-03-21 Nathan Sidwell <nathan@acm.org>
4419 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4420 mention anon-struct/union members and trailing attributes.
4422 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4424 PR tree-optimization/84969
4425 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4426 builtin memset partitions if they set different rhs values.
4428 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4430 PR rtl-optimization/84989
4431 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4432 VEC_DUPLICATE with scalar result mode.
4434 2018-03-21 Martin Liska <mliska@suse.cz>
4437 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4438 not intended return statement.
4440 2018-03-21 Martin Liska <mliska@suse.cz>
4443 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4444 (chkp_find_bound_slots_1): Limit number of iterations.
4446 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4449 * Minor grammar fixes for x86 options.
4451 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4454 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4455 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4457 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4460 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4461 (altivec_lvx_<mode>_1op): Likewise.
4462 (altivec_stvx_<mode>_2op): Likewise.
4463 (altivec_stvx_<mode>_1op): Likewise.
4464 (altivec_lvx_<VM2:mode>): New define_expand.
4465 (altivec_stvx_<VM2:mode>): Likewise.
4466 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4467 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4468 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4469 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4470 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4471 (rs6000_gen_lvx): Likewise.
4472 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4473 (altivec_expand_stv_builtin): Likewise.
4474 (altivec_expand_builtin): Likewise.
4475 * config/rs6000/vector.md: Likewise.
4477 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4480 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4483 2018-03-20 Richard Biener <rguenther@suse.de>
4486 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4487 sign-conversions as zero, fall back to standard scalar_stmt
4490 2018-03-20 Martin Liska <mliska@suse.cz>
4493 * predict.c (rebuild_frequencies): Handle case when we have
4494 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4496 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4499 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4500 flag_section_anchors.
4501 * varasm.c (use_blocks_for_decl_p): Remove hack for
4502 dw2_force_const_mem.
4505 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4507 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4508 be created, use lowpart_subreg of operands[0] rather than operands[0]
4510 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4511 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4512 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4513 and n constraint instead of aarch64_shift_imm_di and Usd.
4514 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4515 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4517 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4520 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4521 to favor GPR over NEON registers.
4522 (<shift>di3_neon): Likewise.
4524 2018-03-20 Tom de Vries <tom@codesourcery.com>
4527 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4528 (nvptx_process_pars): Emit bar.sync asap and alap.
4530 2018-03-20 Tom de Vries <tom@codesourcery.com>
4533 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4534 seen_label if seen_label is already set.
4536 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4539 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4540 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4541 Use 1U instead of 1. Formatting fixes.
4544 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4545 instead of TREE_TYPE (s1) for the return value.
4547 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4549 PR tree-optimization/84946
4550 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4551 bitsize + bitsize in poly_uint64 rather than poly_int64.
4554 * dwarf2asm.c: Include fold-const.c.
4555 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4556 of decl rather than decl itself.
4558 PR rtl-optimization/84643
4559 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4561 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4564 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4565 calling assemble_variable.
4567 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4570 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4571 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4573 2018-03-19 Jim Wilson <jimw@sifive.com>
4576 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4577 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4578 (riscv_first_stack_step): Likewise.
4579 (riscv_option_override): Use STACK_BOUNDARY instead of
4581 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4583 (BIGGEST_ALIGNMENT): Set to 128.
4584 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4585 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4588 2018-03-19 Richard Biener <rguenther@suse.de>
4590 PR tree-optimization/84933
4591 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4592 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4594 2018-03-19 Richard Biener <rguenther@suse.de>
4596 PR tree-optimization/84859
4597 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4598 (cond_if_else_store_replacement): Perform sinking operation on
4599 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4600 Generalize what a BB with a single eligible store is.
4602 2018-03-19 Richard Biener <rguenther@suse.de>
4604 PR tree-optimization/84929
4605 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4606 chrec_is_positive against non-chrec arg.
4608 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4611 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4613 2018-03-18 Martin Liska <mliska@suse.cz>
4615 PR rtl-optimization/84635
4616 * regrename.c (build_def_use): Use matches_mode only when
4619 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4621 PR tree-optimization/84913
4622 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4623 vectorize chains of COND_EXPRs.
4625 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4627 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4629 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4631 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4633 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4635 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4637 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4638 Kito Cheng <kito.cheng@gmail.com>
4640 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4641 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4642 (nds32_adjust_reg_alloc_order): New function.
4643 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4645 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4647 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4648 nds32_print_operand, nds32_print_operand_address): Use
4649 HOST_WIDE_INT_PRINT_DEC instead.
4651 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4653 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4655 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4658 * config/i386/i386.c (initial_ix86_tune_features,
4659 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4661 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4662 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4663 rather than 1u << ix86_tune. Formatting fix.
4664 (ix86_option_override_internal): Change ix86_arch_mask from
4665 unsigned int to unsigned HOST_WIDE_INT, initialize to
4666 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4667 (ix86_function_specific_restore): Likewise.
4669 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4672 * postreload.c (reload_combine_recognize_pattern): Perform
4673 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4674 truncate_int_for_mode the result for the destination's mode.
4677 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4678 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4680 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4683 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4686 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4687 regno_allocno_class_array and sorted_pseudos.
4688 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4689 insns where regno is used.
4691 2018-03-16 Martin Liska <mliska@suse.cz>
4694 * multiple_target.c (create_dispatcher_calls): Redirect
4695 reference in the symbol table.
4697 2018-03-16 Martin Liska <mliska@suse.cz>
4700 * multiple_target.c (create_dispatcher_calls): Redirect also
4703 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4707 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4708 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4709 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4711 2018-03-16 Julia Koval <julia.koval@intel.com>
4713 * doc/invoke.texi (Skylake Server): Add CLWB.
4714 Cannonlake): Remove CLWB.
4716 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4718 PR tree-optimization/84841
4719 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4721 (FLOAT_ONE_CONST_TYPE): Define.
4722 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4723 (sort_by_operand_rank): Put entries with higher constant_type last
4724 rather than first to match comments.
4726 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4728 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4731 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4734 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4735 check and use address of target always.
4737 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4740 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4741 (indirect_thunk_bnd_needed): Likewise.
4742 (indirect_thunks_used): Likewise.
4743 (indirect_thunks_bnd_used): Likewise.
4744 (indirect_return_needed): New.
4745 (indirect_return_bnd_needed): Likewise.
4746 (output_indirect_thunk_function): Add a bool argument for
4748 (output_indirect_thunk_function): Don't generate alias for
4749 function return thunk.
4750 (ix86_code_end): Call output_indirect_thunk_function to generate
4751 function return thunks.
4752 (ix86_output_function_return): Set indirect_return_bnd_needed
4753 and indirect_return_needed instead of indirect_thunk_bnd_needed
4754 and indirect_thunk_needed.
4756 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4758 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4759 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4760 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4762 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4763 Paul Hua <paul.hua.gm@gmail.com>
4766 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4768 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4770 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4771 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4774 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4777 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4778 instead of GET_MODE_SIZE when comparing Units.
4780 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4783 * varasm.c (hash_section): Return an unchangeble hash value
4784 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4785 Return !aarch64_can_use_per_function_literal_pools_p ().
4787 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4790 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4791 rather than address of cmode as last argument to prepare_cmp_insn.
4793 2018-03-15 Julia Koval <julia.koval@intel.com>
4795 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4796 F_AVX512VNNI, F_AVX512BITALG): New.
4798 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4801 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4802 insn for floating-point loads and stores.
4804 2018-03-14 Carl Love <cel@us.ibm.com>
4806 * config/rs6000/rs6000-c.c: Add macro definitions for
4807 ALTIVEC_BUILTIN_VEC_PERMXOR.
4808 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4809 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4810 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4811 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4813 * config/doc/extend.texi: Add prototypes for vec_permxor.
4815 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4818 * diagnostic-show-locus.c (class layout_point): Convert m_line
4819 from int to linenum_type.
4820 (line_span::comparator): Use linenum "compare" function when
4821 comparing line numbers.
4822 (test_line_span): New function.
4823 (layout_range::contains_point): Convert param "row" from int to
4825 (layout_range::intersects_line_p): Likewise.
4826 (layout::will_show_line_p): Likewise.
4827 (layout::print_source_line): Likewise.
4828 (layout::should_print_annotation_line_p): Likewise.
4829 (layout::print_annotation_line): Likewise.
4830 (layout::print_leading_fixits): Likewise.
4831 (layout::annotation_line_showed_range_p): Likewise.
4832 (struct line_corrections): Likewise for field m_row.
4833 (line_corrections::line_corrections): Likewise for param "row".
4834 (layout::print_trailing_fixits): Likewise.
4835 (layout::get_state_at_point): Likewise.
4836 (layout::get_x_bound_for_row): Likewise.
4837 (layout::print_line): Likewise.
4838 (diagnostic_show_locus): Likewise for locals "last_line" and
4840 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4841 * input.c (selftest::test_linenum_comparisons): New function.
4842 (selftest::input_c_tests): Call it.
4843 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4844 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4845 * selftest.h (ASSERT_GT): New macro.
4846 (ASSERT_GT_AT): New macro.
4847 (ASSERT_LT): New macro.
4848 (ASSERT_LT_AT): New macro.
4850 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4852 PR rtl-optimization/84780
4853 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4855 2018-03-14 Martin Liska <mliska@suse.cz>
4857 * tree.c (record_node_allocation_statistics): Use
4858 get_stats_node_kind.
4859 (get_stats_node_kind): New function extracted from
4860 record_node_allocation_statistics.
4861 (free_node): Use get_stats_node_kind.
4863 2018-03-14 Richard Biener <rguenther@suse.de>
4865 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4866 that the value-set of ANTIC_IN doesn't grow.
4869 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4871 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4872 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4874 2018-03-14 Julia Koval <julia.koval@intel.com>
4876 * config.gcc (icelake-client, icelake-server): New.
4878 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4879 (initial_ix86_arch_features): Ditto.
4880 (PTA_SKYLAKE): Add SGX.
4881 (PTA_ICELAKE): Remove.
4882 (PTA_ICELAKE_CLIENT): New.
4883 (PTA_ICELAKE_SERVER): New.
4884 (ix86_option_override_internal): Split up icelake on icelake client and
4886 (get_builtin_code_for_version): Ditto.
4887 (fold_builtin_cpu): Ditto.
4888 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4889 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4890 * config/i386/i386.h (processor_type): Ditto.
4891 * doc/invoke.texi: Ditto.
4893 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4896 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4897 INTEGER_CST offset, add it together with bitpos / 8 and
4898 sign extend based on POINTER_SIZE.
4902 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4905 * config/i386/constraints.md (Yc): New register constraint.
4906 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4907 Use Yc constraint for alternative 2 of operand 0. Remove
4908 preferred_for_speed attribute.
4910 2018-03-14 Richard Biener <rguenther@suse.de>
4912 PR tree-optimization/84830
4913 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4914 with the old one to avoid oscillations.
4916 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4919 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4921 (assign_by_spills): Return a flag of reload assignment failure.
4922 Do not process the reload assignment failures. Do not spill other
4923 reload pseudos if they has the same reg class. Update n if
4925 (lra_assign): Add a return arg. Set up from the result of
4926 assign_by_spills call.
4927 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4928 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4929 usage_insns if it is not NULL.
4930 (spill_hard_reg_in_range): New function.
4931 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4932 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4933 function prototypes.
4934 (lra_assign): Change prototype.
4935 * lra.c (lra): Add code to deal with fails by splitting hard reg
4938 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4940 * config/riscv/riscv.opt (mrelax): New option.
4941 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4942 "norelax" when riscv_mrelax is disabled.
4943 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4945 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4948 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4949 reassociation for int modes.
4951 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4953 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4954 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4956 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4957 * config/aarch64/aarch64-sve.md
4958 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4959 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4960 (*extend<mode><Vwide>2): Rename to...
4961 (aarch64_sve_extend<mode><Vwide>2): ...this.
4962 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4963 renaming the old pattern to...
4964 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4966 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4967 define_expand, renaming the old pattern to...
4968 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4969 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4970 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4971 account when deciding which SVE instruction the optab should use.
4972 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4974 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4976 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4977 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4978 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4979 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4980 (tlsdesc_small_<mode>): Turn a define_expand and use
4981 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4982 (tlsdesc_small_advsimd_<mode>): ...this.
4983 (tlsdesc_small_sve_<mode>): New pattern.
4985 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4987 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4988 (UNSPEC_UMUL_HIGHPART): New constants.
4989 (MUL_HIGHPART): New int iteraor.
4990 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4991 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4993 (*<su>mul<mode>3_highpart): New define_insn.
4995 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4998 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5001 2018-03-13 Martin Liska <mliska@suse.cz>
5004 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5006 (sem_item_optimizer::~sem_item_optimizer): Release it.
5007 (sem_item_optimizer::merge_classes): Register variable aliases.
5008 (sem_item_optimizer::fixup_pt_set): New function.
5009 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5010 * ipa-icf.h: Declare new variables and functions.
5012 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5015 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5016 integer_pow2p@2 and test integer_pow2p in condition.
5017 (A < 0 ? C : 0): Similarly for @1.
5020 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5021 characters starting at p contain '\0' character, don't look beyond
5025 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5026 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5029 * reg-stack.c (change_stack): Change update_end var from int to
5030 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5031 also call set_block_for_insn on the newly added insns and rescan.
5034 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5035 on the last operand.
5038 * tree.c (stabilize_reference_1): Return save_expr (e) for
5039 STATEMENT_LIST even if it doesn't have side-effects.
5041 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5043 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5045 2018-03-12 Renlin Li <renlin.li@arm.com>
5047 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5048 aarch64_output_scalar_simd_mov_immediate.
5050 2018-03-12 Martin Sebor <msebor@redhat.com>
5052 PR tree-optimization/83456
5053 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5054 for perfectly overlapping calls to memcpy.
5055 (gimple_fold_builtin_memory_chk): Same.
5056 (gimple_fold_builtin_strcpy): Handle no-warning.
5057 (gimple_fold_builtin_stxcpy_chk): Same.
5058 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5060 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5062 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5063 parameter. Use it for SFmode.
5064 (rs6000_function_arg_advance_1): Adjust.
5065 (rs6000_function_arg): Adjust.
5066 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5068 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5070 PR rtl-optimization/84169
5071 PR rtl-optimization/84780
5072 * combine.c (can_combine_p): Check for a 2-insn combination whether
5073 the destination register is used between the two insns, too.
5075 2018-03-12 Richard Biener <rguenther@suse.de>
5077 PR tree-optimization/84803
5078 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5079 for refs DR analysis didn't process.
5081 2018-03-12 Richard Biener <rguenther@suse.de>
5083 PR tree-optimization/84777
5084 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5085 force-vectorize loops ignore whether we are optimizing for size.
5087 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5089 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5090 (TARGET_MD_ASM_ADJUST): Define.
5092 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5093 Kito Cheng <kito.cheng@gmail.com>
5094 Chung-Ju Wu <jasonwucj@gmail.com>
5096 * config/nds32/nds32.c (nds32_compute_stack_frame,
5097 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5098 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5099 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5100 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5101 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5102 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5103 * config/nds32/nds32.md (prologue, epilogue): Use macro
5104 NDS32_V3PUSH_AVAILABLE_P to do checking.
5106 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5109 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5110 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5111 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5112 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5113 addition of most attributes on !orig_type_die or the attribute not
5114 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5116 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5117 Chung-Ju Wu <jasonwucj@gmail.com>
5119 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5121 * config/nds32/nds32.opt (mvh): New option.
5123 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5124 Chung-Ju Wu <jasonwucj@gmail.com>
5126 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5128 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5129 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5132 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5133 Chung-Ju Wu <jasonwucj@gmail.com>
5135 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5137 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5138 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5140 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5141 Kito Cheng <kito.cheng@gmail.com>
5142 Chung-Ju Wu <jasonwucj@gmail.com>
5144 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5145 UNSPEC_FFMISM and UNSPEC_FLMISM.
5146 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5147 for ffb, ffmism and flmism.
5148 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5149 (unspec_ffmism): Ditto.
5150 (unspec_flmism): Ditto.
5151 (nds32_expand_builtin_impl): Check if string extension is available.
5152 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5153 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5155 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5158 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5161 * lra-assigns.c (assign_by_spills): Return a flag of reload
5162 assignment failure. Do not process the reload assignment
5163 failures. Do not spill other reload pseudos if they has the same
5165 (lra_assign): Add a return arg. Set up from the result of
5166 assign_by_spills call.
5167 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5168 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5169 usage_insns if it is not NULL.
5170 (spill_hard_reg_in_range): New function.
5171 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5172 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5173 function prototypes.
5174 (lra_assign): Change prototype.
5175 * lra.c (lra): Add code to deal with fails by splitting hard reg
5178 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5181 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5183 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5186 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5187 (dw_val_node): Add val_symbolic_view.
5188 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5189 (symview_upper_bound): New.
5190 (new_line_info_table): Initialize symviews_since_reset.
5191 (dwarf2out_source_line): Count symviews_since_reset and set
5192 symview_upper_bound.
5193 (dw_val_equal_p): Handle symview.
5194 (add_AT_symview): New.
5195 (print_dw_val): Handle symview.
5196 (attr_checksum, attr_checksum_ordered): Likewise.
5197 (same_dw_val_p, size_of_die): Likewise.
5198 (value_format, output_die): Likewise.
5199 (add_high_low_attributes): Use add_AT_symview for entry_view.
5200 (dwarf2out_finish): Reset symview_upper_bound, clear
5203 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5206 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5207 Add strict argument and use it.
5208 (rs6000_split_multireg_move): Update for new strict argument.
5209 (mem_operand_gpr): Disallow all non-offsettable addresses.
5210 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5212 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5215 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5216 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5217 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5220 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5221 decl, use remap_type if we want to use the type.
5223 2018-03-09 Martin Sebor <msebor@redhat.com>
5225 PR tree-optimization/84526
5226 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5228 (builtin_access::generic_overlap): Be prepared to handle non-array
5231 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5233 PR rtl-optimization/84682
5234 * lra-constraints.c (process_address_1): Check is_address flag
5235 for address constraints.
5236 (process_alt_operands): Likewise.
5237 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5238 preprocess_constraints.
5239 * recog.h (preprocess_constraints): Add oploc parameter.
5241 * recog.c (preprocess_constraints): Test address_operand for
5242 CT_ADDRESS constraints.
5244 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5247 * lra-assigns.c (assign_by_spills): Return a flag of reload
5248 assignment failure. Do not process the reload assignment
5249 failures. Do not spill other reload pseudos if they has the same
5251 (lra_assign): Add a return arg. Set up from the result of
5252 assign_by_spills call.
5253 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5254 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5255 usage_insns if it is not NULL.
5256 (spill_hard_reg_in_range): New function.
5257 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5258 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5259 function prototypes.
5260 (lra_assign): Change prototype.
5261 * lra.c (lra): Add code to deal with fails by splitting hard reg
5264 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5267 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5268 Accept complain bool parameter. Only emit errors if it is true.
5269 (arm_parse_cpu_option_name): Likewise.
5270 (arm_target_thumb_only): Adjust callers of the above.
5271 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5272 prototype to take a default true bool parameter.
5273 (arm_parse_arch_option_name): Likewise.
5275 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5276 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5280 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5281 * configure: Regenerate.
5282 * configure.ac ("linker --version-script option"): New.
5283 ("linker soname option"): New.
5285 2018-03-09 Richard Biener <rguenther@suse.de>
5287 PR tree-optimization/84775
5288 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5289 immediate uses of predicate stmts and mark them modified.
5292 PR tree-optimization/84178
5293 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5295 (version_loop_for_if_conversion): Delay update_ssa call.
5296 (tree_if_conversion): Delay update_ssa until after predicate
5299 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5302 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5303 when the function accesses prior frames.
5305 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5308 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5309 gen_llsym, otherwise call maybe_gen_llsym.
5312 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5313 has ',' character inside of it.
5315 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5318 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5319 as clobbering CC_REGNUM.
5321 2018-03-08 Richard Biener <rguenther@suse.de>
5324 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5325 (follow_copies_to_constant): Do not follow SSA names registered
5328 2018-03-08 Richard Biener <rguenther@suse.de>
5330 PR tree-optimization/84178
5331 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5333 (version_loop_for_if_conversion): Delay update_ssa call.
5334 (tree_if_conversion): Delay update_ssa until after predicate
5337 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5339 PR tree-optimization/84178
5340 * tree-if-conv.c (release_bb_predicate): Remove the
5341 the assertion that the stmts have NULL use_ops.
5342 Discard the statements, asserting that they haven't
5343 yet been added to a BB.
5345 2018-03-08 Richard Biener <rguenther@suse.de>
5347 PR tree-optimization/84746
5348 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5349 (phi_translate): Pass in destination ANTIC_OUT set.
5350 (phi_translate_1): Likewise. For a simplified result lookup
5351 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5352 (phi_translate_set): Adjust.
5353 (do_pre_regular_insertion): Likewise.
5354 (do_pre_partial_partial_insertion): Likewise.
5356 2018-03-08 Martin Liska <mliska@suse.cz>
5358 PR gcov-profile/84735
5359 * doc/gcov.texi: Document usage of profile files.
5360 * gcov-io.h: Document changes in the format.
5362 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5366 * dwarf2out.c (struct dw_line_info_table): Update comments for
5368 (FORCE_RESET_NEXT_VIEW): New.
5369 (FORCE_RESETTING_VIEW_P): New.
5370 (RESETTING_VIEW_P): Check for -1 too.
5371 (ZERO_VIEW_P): Likewise.
5372 (new_line_info_table): Force-reset next view.
5373 (dwarf2out_begin_function): Likewise.
5374 (dwarf2out_source_line): Simplify zero_view_p initialization.
5375 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5376 view directly. Omit view when omitting .loc at line 0.
5378 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5380 PR tree-optimization/84740
5381 * tree-switch-conversion.c (process_switch): Call build_constructors
5382 only if info.phi_count is non-zero.
5384 PR tree-optimization/84739
5385 * tree-tailcall.c (find_tail_calls): Check call arguments against
5386 DECL_ARGUMENTS (current_function_decl) rather than
5387 DECL_ARGUMENTS (func) when checking for tail recursion.
5389 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5391 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5392 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5393 Volker Reichelt's entry and add entries for people that perform
5394 GCC fuzzy testing and report numerous bugs.
5396 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5399 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5400 readonly data in sdata, if that is disabled.
5401 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5402 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5403 -mreadonly-in-sdata option.
5405 2018-03-07 Martin Sebor <msebor@redhat.com>
5407 PR tree-optimization/84468
5408 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5409 basic block when looking for nul assignment.
5411 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5414 * except.h (output_function_exception_table): Adjust prototype.
5415 * except.c (output_function_exception_table): Remove FNNAME parameter
5416 and add SECTION parameter. Ouput one part of the table at a time.
5417 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5418 the first part of the exception table and emit unwind directives.
5419 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5420 (i386_pe_seh_cold_init): Likewise.
5421 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5422 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5423 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5424 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5425 * config/i386/winnt.c: Include except.h.
5426 (struct seh_frame_state): Add reg_offset, after_prologue and
5427 in_cold_section fields.
5428 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5429 (i386_pe_seh_cold_init): New function.
5430 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5431 to seh->in_cold_section.
5432 (seh_emit_push): Record the offset of the push.
5433 (seh_emit_save): Record the offet of the save.
5434 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5435 Test seh->after_prologue to disregard the epilogue.
5436 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5437 (i386_pe_end_cold_function): New function.
5439 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5442 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5443 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5446 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5448 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5449 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5452 * multiple_target.c: Include tree-inline.h and intl.h.
5453 (expand_target_clones): Diagnose and fail if node->definition and
5454 !tree_versionable_function_p (node->decl).
5456 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5458 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5460 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5461 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5462 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5464 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5467 * combine.c (try_combine): Use reg_or_subregno instead of handling
5468 just paradoxical SUBREGs and REGs.
5470 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5472 * config/arc/arc.c (arc_finalize_pic): Remove function.
5473 (arc_must_save_register): We use single base PIC register, remove
5474 checks to save/restore the PIC register.
5475 (arc_expand_prologue): Likewise.
5476 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5478 (arc_verify_short): Likewise.
5479 (arc_attr_type): Likewise.
5480 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5481 (walk_stores): Likewise.
5482 (arc_address_cost): Make it static.
5483 (arc_verify_short): Likewise.
5484 (branch_dest): Likewise.
5485 (arc_attr_type): Likewise.
5486 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5487 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5488 (arc_final_prescan_insn): Remove inserting the nops due to
5489 hardware hazards. It is done in reorg step.
5490 (insn_length_variant_t): Remove.
5491 (insn_length_parameters_t): Likewise.
5492 (arc_insn_length_parameters): Likewise.
5493 (arc_get_insn_variants): Likewise.
5494 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5496 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5499 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5502 PR tree-optimization/84687
5503 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5505 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5507 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5509 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5510 Rename to ppc_speculation_barrier.
5511 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5512 __builtin_ppc_speculation_barrier.
5514 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5517 * combine.c (combine_simplify_rtx): Don't try to simplify if
5518 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5521 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5523 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5524 to 32 bytes when compiling for POWER9.
5526 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5529 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5530 regparm >= 3 with no arg reg available also for calls with
5531 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5534 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5536 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5538 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5541 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5543 2018-03-05 Richard Biener <rguenther@suse.de>
5545 PR tree-optimization/84486
5546 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5547 When inserting a __builtin_assume_aligned call set the LHS
5548 SSA name alignment info accordingly.
5550 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5552 PR tree-optimization/84114
5553 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5554 Avoid reassociation of FLOAT_MODE addition.
5556 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5558 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5559 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5560 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5561 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5562 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5563 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5564 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5566 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5567 __builtin_ia32_wbinvd): New builtins.
5568 (SPECIAL_ARGS2): New.
5569 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5570 (SPECIAL_ARGS2): New.
5571 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5572 (ix86_valid_target_attribute_inner_p): Ditto.
5573 (ix86_init_mmx_sse_builtins): Add special_args2.
5574 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5575 TARGET_WBNOINVD_P): New.
5576 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5577 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5578 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5579 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5580 * config/i386/pconfigintrin.h: New file.
5581 * config/i386/wbnoinvdintrin.h: Ditto.
5582 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5583 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5585 2018-03-05 Richard Biener <rguenther@suse.de>
5587 PR tree-optimization/84670
5588 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5590 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5591 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5592 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5593 doesn't grow if all successors (recursively) were visited at least
5596 2018-03-05 Richard Biener <rguenther@suse.de>
5598 PR tree-optimization/84650
5599 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5600 if executed in the loop pipeline.
5602 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5604 * doc/configfiles.texi (Configuration Files): Move info about
5605 conditionalizing $target-protos.h to...
5606 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5607 differs from $target-protos.h.
5609 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5610 Chung-Ju Wu <jasonwucj@gmail.com>
5612 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5613 * config/nds32/nds32-multiple.md (setmemsi): Define.
5614 * config/nds32/nds32-memory-manipulation.c
5615 (nds32_gen_dup_4_byte_to_word_value): New.
5616 (emit_setmem_word_loop): New.
5617 (emit_setmem_byte_loop): New.
5618 (nds32_expand_setmem_loop): New.
5619 (nds32_expand_setmem_loop_v3m): New.
5620 (nds32_expand_setmem_unroll): New.
5621 (nds32_expand_setmem): New.
5623 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5624 Chung-Ju Wu <jasonwucj@gmail.com>
5626 * config/nds32/nds32-memory-manipulation.c
5627 (nds32_emit_load_store): New.
5628 (nds32_emit_post_inc_load_store): New.
5629 (nds32_emit_mem_move): New.
5630 (nds32_emit_mem_move_block): New.
5631 (nds32_expand_movmemsi_loop_unknown_size): New.
5632 (nds32_expand_movmemsi_loop_known_size): New.
5633 (nds32_expand_movmemsi_loop): New.
5634 (nds32_expand_movmemsi_unroll): New.
5635 (nds32_expand_movmemqi): Rename ...
5636 (nds32_expand_movmemsi): ... to this.
5637 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5638 (movmemsi): ... to this.
5639 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5640 (nds32_expand_movmemsi): ... to this.
5642 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5643 Monk Chiang <sh.chiang04@gmail.com>
5644 Chung-Ju Wu <jasonwucj@gmail.com>
5646 * config/nds32/nds32-protos.h
5647 (nds32_expand_load_multiple): New arguments.
5648 (nds32_expand_store_multiple): Ditto.
5649 (nds32_valid_multiple_load_store): Rename ...
5650 (nds32_valid_multiple_load_store_p): ... to this.
5651 * config/nds32/nds32-memory-manipulation.c
5652 (nds32_expand_load_multiple): Refine implementation.
5653 (nds32_expand_store_multiple): Ditto.
5654 * config/nds32/nds32-multiple.md
5655 (load_multiple): Update nds32_expand_load_multiple interface.
5656 (store_multiple): Update nds32_expand_store_multiple interface.
5657 * config/nds32/nds32-predicates.c
5658 (nds32_valid_multiple_load_store): Rename ...
5659 (nds32_valid_multiple_load_store_p): ... to this and refine
5661 * config/nds32/predicates.md
5662 (nds32_load_multiple_and_update_address_operation): New predicate.
5663 (nds32_store_multiple_and_update_address_operation): New predicate.
5665 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5666 Chung-Ju Wu <jasonwucj@gmail.com>
5668 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5669 (combo): New attribute.
5670 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5672 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5674 * config/nds32/nds32.opt: Change -mcmodel= default value.
5676 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5677 Monk Chiang <sh.chiang04@gmail.com>
5678 Chung-Ju Wu <jasonwucj@gmail.com>
5680 * config/nds32/constants.md (unspec_element): New enum.
5681 * config/nds32/constraints.md (Umw): New constraint.
5682 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5683 * config/nds32/nds32-intrinsic.md: Likewise.
5684 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5685 (nds32_valid_smw_lwm_base_p): New.
5686 (nds32_output_smw_single_word): New.
5687 (nds32_output_lmw_single_word): New.
5688 (nds32_expand_unaligned_load): New.
5689 (nds32_expand_unaligned_store): New.
5690 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5691 (nds32_output_smw_single_word): Declare.
5692 (nds32_output_lmw_single_word): Declare.
5693 (nds32_expand_unaligned_load): Declare.
5694 (nds32_expand_unaligned_store): Declare.
5695 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5696 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5697 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5698 NDS32_BUILTIN_UASTORE_DW.
5699 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5702 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5703 Kito Cheng <kito.cheng@gmail.com>
5704 Chung-Ju Wu <jasonwucj@gmail.com>
5706 * config/nds32/nds32-intrinsic.c
5707 (nds32_expand_builtin_null_ftype_reg): Delete.
5708 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5709 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5710 (nds32_read_argument): New.
5711 (nds32_legitimize_target): Ditto.
5712 (nds32_legitimize_argument): Ditto.
5713 (nds32_check_constant_argument): Ditto.
5714 (nds32_expand_unop_builtin): Ditto.
5715 (nds32_expand_unopimm_builtin): Ditto.
5716 (nds32_expand_binop_builtin): Ditto.
5717 (nds32_builtin_decl_impl): Ditto.
5718 (builtin_description): Ditto.
5719 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5720 (nds32_init_builtins_impl): Ditto.
5721 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5722 (nds32_builtin_decl): New.
5723 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5724 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5726 2018-03-02 Jeff Law <law@redhat.com>
5728 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5729 (redundant_insn, fill_simple_delay_slots): Likewise.
5730 (fill_slots_from_thread): Likewise.
5731 * resource.c (mark_referenced_resources): Likewise.
5732 (mark_set_resources, find_dead_or_set_registers): Likewise.
5734 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5736 * substring-locations.h (format_warning_va): Formatting fix for
5738 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5740 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5741 * substring-locations.c: Include intl.h.
5742 (format_warning_va): Turned into small wrapper around
5743 format_warning_n_va, renamed to ...
5744 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5745 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5747 (format_warning_at_substring_n): New function.
5748 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5749 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5750 format_warning_at_substring with just a shorter name instead of
5751 const function pointer.
5752 (fmtwarn_n): New function.
5753 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5754 appropriate, get rid of all the fmtstr temporaries, move conditionals
5755 with G_() wrapped string literals directly into fmtwarn arguments,
5756 cast dir.len to (int), formatting fixes.
5758 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5760 * doc/invoke.texi: Remove "Cilk Plus" references.
5762 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5763 Richard Biener <rguenther@suse.de>
5766 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5767 for error or warning attributes if CALL_FROM_THUNK_P is set.
5770 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5773 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5774 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5777 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5778 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5780 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5781 instead of -1U in last predictors element's probability member.
5783 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5786 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5787 arguments if they are comparable.
5789 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5791 PR tree-optimization/84634
5792 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5793 masks and masked_loop_p with a single loop_masks, making sure it's
5794 null for bb vectorization.
5796 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5798 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5799 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5800 loop->force_vectorize to check whether there is no alias.
5802 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5805 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5807 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5809 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5810 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5811 instead of a loop around prev_real_insn.
5812 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5816 * config/i386/i386.c (ix86_print_operand): Use conditional
5817 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5820 2018-03-02 Richard Biener <rguenther@suse.de>
5822 PR tree-optimization/84427
5823 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5824 (bitmap_set_subtract_values): Rewrite to handle multiple
5827 (prune_clobbered_mems): Likewise.
5828 (phi_translate): Take edge instead of pred/phiblock.
5829 (phi_translate_1): Likewise.
5830 (phi_translate_set): Likewise. Insert all translated
5831 exprs for a value into the set, keeping possibly multiple
5832 expressions per value.
5833 (compute_antic_aux): Adjust for phi_translate changes.
5834 When intersecting union the expressions and prune those
5835 not in the final value set, keeping possibly multiple
5836 expressions per value. Do not use value-insertion
5837 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5838 all expressions. Add verification that the value-sets
5839 only shrink during iteration.
5840 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5841 (do_pre_regular_insertion): Likewise.
5842 (do_pre_partial_partial_insertion): Likewise.
5844 2018-03-02 Richard Biener <rguenther@suse.de>
5847 * config/darwin.c (saved_debug_info_level): New static global.
5848 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5849 (darwin_asm_lto_end): Restore debug info generation settings.
5851 2018-03-01 Martin Liska <mliska@suse.cz>
5854 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5857 2018-03-01 Richard Biener <rguenther@suse.de>
5860 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5861 type annotation with LTO when debug was disabled at compile-time.
5863 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5865 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5867 (mips_final_postscan_insn): Likewise.
5869 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5871 PR rtl-optimization/84528
5872 * alias.c (init_alias_target): Add commentary.
5873 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5874 a unique base value if the frame pointer is not eliminated
5875 to the stack pointer.
5877 2018-03-01 Tom de Vries <tom@codesourcery.com>
5879 PR rtl-optimization/83327
5880 * lra-int.h (hard_regs_spilled_into): Declare.
5881 * lra.c (hard_regs_spilled_into): Define.
5882 (init_reg_info): Init hard_regs_spilled_into.
5883 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5884 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5885 (process_bb_lives): Handle hard_regs_spilled_into.
5886 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5887 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5889 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5891 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5892 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5893 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5894 * config/rs6000/aix72.h: New file.
5896 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5898 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5899 instead of warning_at with conditional singular and plural messages
5903 * stor-layout.c (update_alignment_for_field): For
5904 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5905 && !DECL_PACKED (field), do the alignment update, just use
5906 only desired_align instead of MAX (type_align, desired_align)
5908 (place_field): Don't do known_align < desired_align handling
5909 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5910 is non-NULL, instead do it after rli->prev_field handling and
5911 only if not within a bitfield word. For DECL_PACKED (field)
5912 use type_align of BITS_PER_UNIT.
5914 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5916 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5917 superfluous parentheses and trailing spaces.
5919 2018-02-28 Richard Biener <rguenther@suse.de>
5921 PR tree-optimization/84584
5922 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5923 SCoPs with fake exit edge.
5925 2018-02-28 Martin Liska <mliska@suse.cz>
5928 * timevar.c (timer::print): Fix format to properly print 100%
5931 2018-02-28 Richard Biener <rguenther@suse.de>
5934 * genmatch.c (capture_info::walk_match): Do not mark
5935 captured expressions without operands as expr_p given
5936 they act more like predicates and should be subject to
5937 "lost tail" side-effect preserving.
5939 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5941 PR rtl-optimization/81611
5942 * auto-inc-dec.c (attempt_change): Move dead note from
5943 mem_insn if it's the next use of regno
5944 (find_address): Take address use of reg holding
5945 non-incremented value. Add parm to limit search to the named
5947 (merge_in_block): Attempt to use a mem insn that is the next
5948 use of the original regno.
5950 2018-02-27 Martin Sebor <msebor@redhat.com>
5953 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5954 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5956 2018-02-27 Martin Sebor <msebor@redhat.com>
5958 PR translation/84207
5959 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5960 n argument to unsigned HOST_WIDE_INT.
5961 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5962 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5963 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5964 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5966 2018-02-27 Richard Biener <rguenther@suse.de>
5968 PR tree-optimization/84512
5969 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5970 Do not use the estimate returned from record_stmt_cost for
5971 the scalar iteration cost but sum properly using add_stmt_cost.
5973 2018-02-27 Richard Biener <rguenther@suse.de>
5975 PR tree-optimization/84466
5976 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5977 Adjust last change to less strictly validate use operands.
5979 2018-02-27 Martin Liska <mliska@suse.cz>
5981 PR gcov-profile/84548
5982 * gcov.c (process_file): Allow partial overlap and consider it
5983 also as group functions.
5984 (output_lines): Properly calculate range of lines for a group.
5986 2018-02-27 Martin Liska <mliska@suse.cz>
5988 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5989 'ggc' suffixes. Change first column width.
5990 (timer::print): Fix formatting of the column.
5992 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5994 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5995 preserve inline entry blocks for the sake of debug inline
5996 entry point markers alone.
5997 (remove_unused_locals): Suggest in comments a better place to
5998 force the preservation of inline entry blocks that are
5999 otherwise unused, but do not preserve them.
6001 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6003 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6005 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6008 * config/i386/constraints.md (Bs): Replace
6009 ix86_indirect_branch_register with
6010 TARGET_INDIRECT_BRANCH_REGISTER.
6012 * config/i386/i386.md (indirect_jump): Likewise.
6013 (tablejump): Likewise.
6014 (*sibcall_memory): Likewise.
6015 (*sibcall_value_memory): Likewise.
6016 Peepholes of indirect call and jump via memory: Likewise.
6017 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6018 (*sibcall_value_GOT_32): Likewise.
6019 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6020 (GOT_memory_operand): Likewise.
6021 (call_insn_operand): Likewise.
6022 (sibcall_insn_operand): Likewise.
6023 (GOT32_symbol_operand): Likewise.
6024 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6026 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6028 PR rtl-optimization/83496
6029 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6030 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6031 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6032 redundant insn, if any.
6033 (relax_delay_slots): Likewise.
6034 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6036 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6038 PR tree-optimization/83965
6039 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6040 that grouped statements are part of a reduction chain. Return
6041 true if the statement is not marked as a reduction itself but
6043 (vect_recog_dot_prod_pattern): Don't check whether the statement
6044 is part of a group here.
6045 (vect_recog_sad_pattern): Likewise.
6046 (vect_recog_widen_sum_pattern): Likewise.
6048 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6051 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6053 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6056 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6058 (ix86_output_indirect_function_return): New prototype.
6059 (ix86_split_simple_return_pop_internal): Likewise.
6060 * config/i386/i386.c (indirect_return_via_cx): New.
6061 (indirect_return_via_cx_bnd): Likewise.
6062 (indirect_thunk_name): Handle return va CX_REG.
6063 (output_indirect_thunk_function): Create alias for
6064 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6065 (ix86_output_indirect_jmp): Remove the bool argument.
6066 (ix86_output_indirect_function_return): New function.
6067 (ix86_split_simple_return_pop_internal): Likewise.
6068 * config/i386/i386.md (*indirect_jump): Don't pass false
6069 to ix86_output_indirect_jmp.
6070 (*tablejump_1): Likewise.
6071 (simple_return_pop_internal): Change it to define_insn_and_split.
6072 Call ix86_split_simple_return_pop_internal to split it for
6074 (simple_return_indirect_internal): Call
6075 ix86_output_indirect_function_return instead of
6076 ix86_output_indirect_jmp.
6078 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6081 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6082 memset and value initialization afterwards.
6084 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6086 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6088 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6091 * common/config/aarch64/aarch64-common.c
6092 (aarch_option_optimization_table[]): Switch
6093 off fomit-frame-pointer
6095 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6096 Chung-Ju Wu <jasonwucj@gmail.com>
6098 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6100 (store_multiple): Ditto.
6102 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6104 * config.gcc: Add --with-cpu support for nds32 target.
6105 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6106 * config/nds32/nds32.opt: Add -mcpu= option.
6108 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6110 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6111 isel=yes): Warn for these deprecated options.
6113 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6115 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6116 ISA_2_5_MASKS_EMBEDDED.
6118 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6120 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6121 p->max as pointers rather than using iterative_hash_expr.
6123 2018-02-23 Carl Love <cel@us.ibm.com>
6125 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6126 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6128 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6129 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6130 P8V_BUILTIN_VEC_VUNSIGNED2.
6132 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6135 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6136 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6138 * lra-constraints.c (curr_insn_transform): Set up
6139 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6141 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6142 * lra-eliminations.c (spill_pseudos): Ditto.
6143 (process_insn_for_elimination): Ditto.
6144 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6145 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6146 LRA_NON_CLOBBERED_ALT.
6148 2018-02-22 Martin Sebor <msebor@redhat.com>
6150 PR tree-optimization/84480
6151 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6152 to maybe_diag_stxncpy_trunc. Call it.
6153 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6154 from gimple_fold_builtin_strcpy. Print inlining stack.
6155 (handle_builtin_stxncpy): Print inlining stack.
6156 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6158 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6161 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6162 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6163 and -fcheck-pointer-bounds are used together.
6164 (indirect_thunk_prefix): New enum.
6165 (indirect_thunk_need_prefix): New function.
6166 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6167 "_nt" instead of "_bnd" for NOTRACK prefix.
6168 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6169 (output_indirect_thunk_function): Likewise.
6171 (ix86_code_end): Update output_indirect_thunk_function calls.
6172 (ix86_output_indirect_branch_via_reg): Replace
6173 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6174 (ix86_output_indirect_branch_via_push): Likewise.
6175 (ix86_output_function_return): Likewise.
6176 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6177 incompatible with -fcf-protection=branch and
6178 -fcheck-pointer-bounds.
6180 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6183 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6184 Change gcc_assert call to output_operand_lossage.
6186 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6188 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6190 2018-02-22 DJ Delorie <dj@redhat.com>
6191 Sebastian Perta <sebastian.perta@renesas.com>
6192 Oleg Endo <olegendo@gcc.gnu.org>
6194 * config/rx/rx.c (rx_rtx_costs): New function.
6195 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6197 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6199 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6201 2018-02-22 Martin Liska <mliska@suse.cz>
6204 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6205 Add "native" as a possible value.
6207 2018-02-22 Martin Liska <mliska@suse.cz>
6210 * config/i386/i386.c (ix86_option_override_internal):
6211 Add "native" as a possible value for -march and -mtune.
6213 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6216 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6217 to all type variants.
6219 PR tree-optimization/84503
6220 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6221 width as info->bitpos + info->bitsize - start.
6222 (merged_store_group::merge_overlapping): Simplify width computation.
6223 (check_no_overlap): New function.
6224 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6225 start + width and last_order of the group, fail if check_no_overlap
6227 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6228 to group if check_no_overlap fails.
6230 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6232 * config/rs6000/altivec.md: Delete contraint arguments to
6233 define_expand, define_split, and define_peephole2, and in
6234 define_insn_and_split if always unused.
6235 * config/rs6000/darwin.md: Ditto.
6236 * config/rs6000/dfp.md: Ditto.
6237 * config/rs6000/rs6000.md: Ditto.
6238 * config/rs6000/sync.md: Ditto.
6239 * config/rs6000/vector.md: Ditto.
6240 * config/rs6000/vsx.md: Ditto.
6242 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6244 * config/rs6000/altivec.md: Write output control strings as braced
6245 blocks instead of double-quoted strings.
6246 * config/rs6000/darwin.md: Ditto.
6247 * config/rs6000/rs6000.md: Ditto.
6248 * config/rs6000/vector.md: Ditto.
6249 * config/rs6000/vsx.md: Ditto.
6251 2018-02-21 Jason Merrill <jason@redhat.com>
6253 PR c++/84314 - ICE with templates and fastcall attribute.
6254 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6256 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6258 * ipa-cp.c (determine_versionability): Fix comment typos.
6260 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6263 * ipa-cp.c (determine_versionability): Do not version functions caling
6266 2018-02-21 Martin Liska <mliska@suse.cz>
6269 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6270 Add "native" as a possible value.
6271 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6272 the macro when native cpu detection is available.
6274 2018-02-21 Martin Liska <mliska@suse.cz>
6277 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6278 Add "native" as a possible value.
6279 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6280 when native cpu detection is available.
6282 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6283 Martin Sebor <msebor@redhat.com>
6285 PR tree-optimization/84478
6286 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6288 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6289 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6290 support which is conservatively correct, for 2 only stay conservative
6291 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6292 argument to the 2 argument get_range_strlen, adjust 6 arg
6293 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6295 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6296 (gimple_fold_builtin_strlen): Pass true as last argument to
6299 2018-02-20 Martin Sebor <msebor@redhat.com>
6302 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6303 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6304 (builtin_memref::builtin_memref): Factor out parts into
6305 set_base_and_offset and call it.
6307 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6310 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6311 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6312 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6313 search at the associated MODE_INT.
6315 2018-02-20 Jeff Law <law@redhat.com>
6318 PR tree-optimization/81592
6320 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6321 for range data rather than using global data.
6322 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6323 range data rather than using global data.
6324 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6325 pass it to children as needed.
6326 (struct directive::fmtresult): Similarly.
6327 (struct directive::set_width): Similarly.
6328 (struct directive::set_precision): Similarly.
6329 (format_integer, format_directive, parse_directive): Similarly.
6330 (format_none): Accept unnamed vr_values parameter.
6331 (format_percent, format_floating, format_character): Similarly.
6332 (format_string, format_plain): Similarly.
6333 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6334 the EVRP range analyzer for range data rather than using global data.
6335 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6336 gimple-ssa-evrp-analyze.h
6337 (class sprintf_dom_walker): Add after_dom_children member function.
6338 Add evrp_range_analyzer member.
6339 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6340 range analyzer as needed.
6341 (sprintf_dom_walker::after_dom_children): New member function.
6342 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6344 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6345 (evrp_range_analyzer::pop_to_marker): Likewise.
6347 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6349 PR tree-optimization/84419
6350 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6351 with the required type if its current type is compatible but
6354 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6357 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6358 after vectorization.
6360 2018-02-20 Martin Liska <mliska@suse.cz>
6363 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6364 possible values if we don't have a hint.
6366 2018-02-20 Martin Liska <mliska@suse.cz>
6370 * final.c (shorten_branches): Build align_tab array with one
6372 * opts.c (finish_options): Add alignment option limit check.
6373 (MAX_CODE_ALIGN): Likewise.
6374 (MAX_CODE_ALIGN_VALUE): Likewise.
6375 * doc/invoke.texi: Document maximum allowed option value for
6376 all -falign-* options.
6378 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6381 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6382 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6383 * var-tracking.c (emit_note_insn_var_location): Remove all references
6384 to NOTE_INSN_CALL_ARG_LOCATION.
6385 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6386 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6387 Use copy_rtx_if_shared.
6388 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6389 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6390 (dwarf2out_var_location): Remove handling of
6391 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6393 * final.c (final_scan_insn): Remove all references to
6394 NOTE_INSN_CALL_ARG_LOCATION.
6395 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6396 before dumping final insns.
6397 * except.c (emit_note_eh_region_end): Remove all references to
6398 NOTE_INSN_CALL_ARG_LOCATION.
6399 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6400 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6401 * config/arc/arc.c (hwloop_optimize): Likewise.
6402 * config/arm/arm.c (create_fix_barrier): Likewise.
6403 * config/s390/s390.c (s390_chunkify_start): Likewise.
6404 * config/sh/sh.c (find_barrier): Likewise.
6405 * config/i386/i386.c (rest_of_insert_endbranch,
6406 ix86_seh_fixup_eh_fallthru): Likewise.
6407 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6408 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6409 * config/frv/frv.c (frv_function_prologue): Likewise.
6410 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6412 (note_outside_basic_block_p): Remove all references to
6413 NOTE_INSN_CALL_ARG_LOCATION.
6414 * gengtype.c (adjust_field_rtx_def): Likewise.
6415 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6417 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6418 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6421 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6424 PR tree-optimization/84452
6425 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6426 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6429 2018-02-19 Martin Liska <mliska@suse.cz>
6432 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6434 2018-02-19 Martin Liska <mliska@suse.cz>
6435 Richard Sandiford <richard.sandiford@linaro.org>
6437 PR tree-optimization/82491
6438 * gimple-fold.c (get_base_constructor): Make earlier bail out
6441 2018-02-19 Carl Love <cel@us.ibm.com>
6443 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6444 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6446 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6447 P8V_BUILTIN_VEC_NEG.
6449 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6451 * config/rl78/rl78.md (movdf): New define expand.
6453 2018-02-19 Martin Liska <mliska@suse.cz>
6456 * doc/invoke.texi: Fix typo.
6457 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6459 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6461 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6462 handle rs6000_single_float and rs6000_double_float specially for
6465 2018-02-16 Jeff Law <law@redhat.com>
6467 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6470 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6473 * ipa-inline.c (inline_small_functions): Fix a typo.
6475 2018-02-16 Nathan Sidwell <nathan@acm.org>
6477 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6479 2018-02-16 Carl Love <cel@us.ibm.com>
6481 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6482 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6483 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6484 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6485 expansion to P8V_BUILTIN_VEC_FLOAT2.
6487 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6489 PR rtl-optimization/70023
6490 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6491 src_regno into account.
6493 2018-02-16 Carl Love <cel@us.ibm.com>
6495 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6496 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6497 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6498 * config/rs6000/rs6000.c: Remove case statements for
6499 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6500 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6501 and P9V_BUILTIN_VEC_VINSERT4B.
6502 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6503 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6504 * config/rs6000/vsx.md:
6505 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6508 2018-02-16 Carl Love <cel@us.ibm.com>
6510 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6512 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6514 * config/rs6000/rs6000-c.c: Add the definitions for
6515 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6516 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6517 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6518 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6519 definition for insert4b and define insn *insert3b_internal.
6520 * doc/extend.texi: Add documentation for vec_extract4b.
6522 2018-02-16 Nathan Sidwell <nathan@acm.org>
6524 * doc/extend.texi (Backwards Compatibility): Mention friend
6525 injection. Note for-scope is deprecated.
6526 * doc/invoke.texi (-ffriend-injection): Deprecate.
6528 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6530 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6531 that moved to I2, also allow destinations that are a paradoxical
6532 subreg (instead of a normal reg).
6534 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6537 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6540 2018-02-16 Richard Biener <rguenther@suse.de>
6542 PR tree-optimization/84037
6543 PR tree-optimization/84016
6545 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6546 Adjust vec_construct for the fact we need additional higher latency
6547 128bit inserts for AVX256 and AVX512 vector builds.
6548 (ix86_add_stmt_cost): Scale vector construction cost for
6551 2018-02-16 Richard Biener <rguenther@suse.de>
6553 PR tree-optimization/84417
6554 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6555 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6556 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6558 2018-02-16 Martin Liska <mliska@suse.cz>
6561 * internal-fn.def (ASAN_CHECK): Set proper flags.
6562 (ASAN_MARK): Likewise.
6564 2018-02-16 Julia Koval <julia.koval@intel.com>
6566 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6567 from PTA_CANNONLAKE.
6569 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6572 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6573 Use ++iter rather than iter++ for std::list iterators.
6574 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6575 defer deleting them until all nodes in the forest are processed. Do
6576 free even leaf nodes. Change to_process into auto_vec.
6579 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6580 * vec.h (vec_default_construct): Use memset instead of placement new
6581 if BROKEN_VALUE_INITIALIZATION is defined.
6582 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6583 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6586 PR rtl-optimization/83723
6587 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6588 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6589 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6591 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6593 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6595 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6597 PR rtl-optimization/81443
6598 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6599 from inner REGs to paradoxical SUBREGs.
6601 2018-02-16 Richard Biener <rguenther@suse.de>
6603 PR tree-optimization/84399
6604 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6605 For operands we can analyze at their definition make sure we can
6606 analyze them at each use as well.
6608 2018-02-16 Richard Biener <rguenther@suse.de>
6610 PR tree-optimization/84190
6611 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6612 volatile accesses if the decl isn't volatile.
6614 2018-02-15 Jason Merrill <jason@redhat.com>
6616 PR c++/84314 - ICE with templates and fastcall attribute.
6617 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6618 TYPE_CANONICAL on an existing type.
6620 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6622 PR tree-optimization/84383
6623 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6624 dstoff nor call operand_equal_p if dstbase is NULL.
6626 PR tree-optimization/84334
6627 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6628 also a CONSTANT_CLASS_P, punt.
6630 2018-02-14 Jim Wilson <jimw@sifive.com>
6632 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6633 first SMALL_OPERAND check. New local min_second_step. Move assert
6634 to where locals are set. Add TARGET_RVC support.
6635 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6637 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6639 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6641 2018-02-14 Martin Sebor <msebor@redhat.com>
6643 PR tree-optimization/83698
6644 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6645 arrays constrain the offset range to their bounds.
6646 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6647 (builtin_access::overlap): Avoid setting the size of overlap if it's
6649 (maybe_diag_overlap): Also consider arrays when deciding what values
6650 of offsets to include in diagnostics.
6652 2018-02-14 Martin Sebor <msebor@redhat.com>
6655 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6656 that correspond to the kind of a declaration.
6658 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6661 * config/pa/pa.md: Load address of PIC label using the linkage table
6662 if the label is nonlocal.
6664 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6666 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6667 warning message if user requests -maltivec=be.
6668 * doc/invoke.texi: Document deprecation of -maltivec=be.
6670 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6673 * config/rs6000/rs6000-c.c: Update definitions for
6674 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6675 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6677 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6680 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6681 add _get_ssp intrinsics. Remove argument from
6682 __builtin_ia32_rdssp[d|q].
6683 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6684 * config/i386/i386-builtin.def: Remove argument from
6685 __builtin_ia32_rdssp[d|q].
6686 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6687 ix86_expand_special_args_builtin for _rdssp[d|q].
6688 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6689 Clear register before usage.
6690 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6691 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6693 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6695 PR tree-optimization/84357
6696 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6697 operand 1 of an ARRAY_REF too.
6699 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6702 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6703 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6705 (set_of_reg): New struct.
6706 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6707 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6708 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6710 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6711 Split into bitclr, bitset, bitinvert patterns if appropriate.
6712 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6713 use rx_fuse_in_memory_bitop.
6714 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6715 to named insn, correct maximum insn length.
6717 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6720 * machmode.def: Define a complex mode for PARTIAL_INT.
6721 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6723 * doc/rtl.texi: Document CSPImode.
6724 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6726 (msp430_hard_regno_nregs_with_padding): Likewise.
6728 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6731 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6733 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6735 PR rtl-optimization/84169
6736 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6737 we generated a parallel as new i3 and we split that to new i2 and i3
6738 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6739 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6740 those to i2, not i1. Partially rewrite this scan code.
6742 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6745 * stor-layout.c (place_field): For variable length fields, adjust
6746 offset_align afterwards not just based on the field's alignment,
6747 but also on the size.
6750 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6751 of exps and logs in the use_exp2 case.
6753 2018-02-13 Jeff Law <law@redhat.com>
6755 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6758 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6761 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6765 * common.opt (gas-loc-support, gas-locview-support): New.
6766 (ginline-points, ginternal-reset-location-views): New.
6767 * doc/invoke.texi: Document them. Use @itemx where intended.
6768 (gvariable-location-views): Adjust.
6769 * target.def (reset_location_view): New.
6770 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6771 (TARGET_RESET_LOCATION_VIEW): New.
6772 * doc/tm.texi: Rebuilt.
6773 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6774 (dwarf2out_default_as_locview_support): New.
6775 (output_asm_line_debug_info): Use option variables.
6776 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6777 (output_loc_list): Likewise.
6778 (add_high_low_attributes): Check option variables.
6779 Don't output entry view attribute in strict mode.
6780 (gen_inlined_subroutine_die): Check option variables.
6781 (dwarf2out_inline_entry): Likewise.
6782 (init_sections_and_labels): Likewise.
6783 (dwarf2out_early_finish): Likewise.
6784 (maybe_reset_location_view): New, from...
6785 (dwarf2out_var_location): ... here. Call it.
6786 * debug.h (dwarf2out_default_as_loc_support): Declare.
6787 (dwarf2out_default_as_locview_support): Declare.
6788 * hooks.c (hook_int_rtx_insn_0): New.
6789 * hooks.h (hook_int_rtx_insn_0): Declare.
6790 * toplev.c (process_options): Take -gas-loc-support and
6791 -gas-locview-support from dwarf2out. Enable
6792 -gvariable-location-views by default only with locview
6793 assembler support. Enable -ginternal-reset-location-views by
6794 default only if the target defines the corresponding hook.
6795 Enable -ginline-points by default if location views are
6796 enabled; force it disabled if statement frontiers are
6798 * tree-inline.c (expand_call_inline): Check option variables.
6799 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6801 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6803 PR tree-optimization/84321
6804 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6805 handling. Also check whether the anti-range contains any values
6806 that satisfy the mask; switch to a VR_RANGE if not.
6808 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6811 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6813 2018-02-13 Martin Jambor <mjambor@suse.cz>
6816 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6817 of call statements, also set location of a load to a temporary.
6819 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6821 * config/rl78/rl78.c (add_vector_labels): New function.
6822 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6823 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6824 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6825 which checks that no arguments are passed.
6826 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6827 * doc/extend.texi: Documentation for the new attribute.
6829 2018-02-13 Andreas Schwab <schwab@suse.de>
6831 * config/riscv/linux.h (CPP_SPEC): Define.
6833 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6836 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6837 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6838 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6839 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6840 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6841 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6842 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6843 temporarily for AES and PCLMUL builtins.
6845 PR tree-optimization/84339
6846 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6847 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6851 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6852 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6853 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6855 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6857 * omp-simd-clone.h: New file.
6858 * omp-simd-clone.c: Include omp-simd-clone.h.
6859 (expand_simd_clones): No longer static.
6860 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6861 cgraph.h and omp-simd-clone.h.
6862 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6863 (vect_recog_widen_shift_pattern): Formatting fix.
6864 (vect_pattern_recog_1): Don't check optab for calls.
6867 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6868 operands[2] into a REG before using gen_lowpart on it.
6870 2018-02-12 Jeff Law <law@redhat.com>
6873 * config/sh/sh.c (find_barrier): Consider a sibling call
6876 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6877 successfully back substituting a reg.
6879 2018-02-12 Richard Biener <rguenther@suse.de>
6881 PR tree-optimization/84037
6882 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6883 parameter, move visited init to caller.
6884 (vect_slp_analyze_operations): Separate cost from validity
6885 check, initialize visited once for all instances.
6886 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6888 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6890 (vect_model_promotion_demotion_cost): Likewise.
6891 (vectorizable_bswap): Guard cost modeling with !slp_node
6892 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6894 (vectorizable_call): Likewise.
6895 (vectorizable_conversion): Likewise.
6896 (vectorizable_assignment): Likewise.
6897 (vectorizable_shift): Likewise.
6898 (vectorizable_operation): Likewise.
6899 (vectorizable_store): Likewise.
6900 (vectorizable_load): Likewise.
6901 (vectorizable_condition): Likewise.
6902 (vectorizable_comparison): Likewise.
6904 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6907 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6908 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6909 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6911 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6914 * params.def (inline-min-speedup): Increase from 8 to 15.
6915 (max-inline-insns-auto): Decrease from 40 to 30.
6916 * ipa-split.c (consider_split): Add some buffer for function to
6917 be considered inlining candidate.
6918 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6921 2018-02-12 Richard Biener <rguenther@suse.de>
6923 PR tree-optimization/84037
6924 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6925 matched stmts if we cannot swap the non-matched ones.
6927 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6929 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6930 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6931 _mm_maskz_scalef_round_ss): New intrinsics.
6932 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6933 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6934 __builtin_ia32_scalefss_round): Remove.
6935 (__builtin_ia32_scalefsd_mask_round,
6936 __builtin_ia32_scalefss_mask_round): New intrinsics.
6937 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6938 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6939 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6940 "<round_constraint>")): Changed to ...
6941 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6942 "<round_scalar_constraint>")): ... this.
6943 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6944 %0, %1, %2<round_op3>}"): Changed to ...
6945 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6946 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6947 %2<round_scalar_mask_op3>}"): ... this.
6948 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6950 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6952 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6953 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6954 (_mm_maskz_sqrt_round_ss): New intrinsics.
6955 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6956 (__builtin_ia32_sqrtsd_mask_round)
6957 (__builtin_ia32_sqrtss_mask_round): New builtins.
6958 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6959 (__builtin_ia32_sqrtss_round): Remove.
6960 (__builtin_ia32_sqrtsd_mask_round)
6961 (__builtin_ia32_sqrtss_mask_round): New builtins.
6962 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6963 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6964 ((match_operand:VF_128 1 "vector_operand"
6965 "xBm,<round_constraint>")): Changed to ...
6966 ((match_operand:VF_128 1 "vector_operand"
6967 "xBm,<round_scalar_constraint>")): ... this.
6968 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6969 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6970 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6971 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6972 %<iptr>1<round_scalar_mask_op3>}): ... this.
6973 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6974 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6976 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6979 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6980 Cast vec_cmpeq result to correct type.
6981 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6982 Cast vec_cmpgt result to correct type.
6984 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6986 * final.c (final_scan_insn_1): Renamed from...
6987 (final_scan_insn): ... this. New wrapper, to recover
6988 seen from the outermost call in recursive ones.
6989 * config/sparc/sparc.c (output_return): Drop seen from call.
6990 (output_sibcall): Likewise.
6991 * config/visium/visium.c (output_branch): Likewise.
6993 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6995 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6998 2018-02-10 Alan Modra <amodra@gmail.com>
7001 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7002 Specify LR as an input.
7004 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7007 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7008 remove_member_access_dummy_vars): New functions.
7009 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7010 lower_omp_1, execute_lower_omp): Use them.
7012 PR rtl-optimization/84308
7013 * shrink-wrap.c (spread_components): Release todo vector.
7015 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7017 PR rtl-optimization/57193
7018 * ira-color.c (struct allocno_color_data): Add member
7019 conflict_allocno_hard_prefs.
7020 (update_conflict_allocno_hard_prefs): New.
7021 (bucket_allocno_compare_func): Add a preference based on
7022 conflict_allocno_hard_prefs.
7023 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7024 (color_allocnos): Remove a dead code. Initiate
7025 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7027 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7030 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7031 constraint from =wa to wa. Avoid a subreg on the output operand,
7032 instead use a pseudo and subreg it in a move.
7033 (p9_xxbrd_<mode>): Changed to ...
7034 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7035 (p9_xxbrd_v2df): New expander.
7036 (p9_xxbrw_<mode>): Changed to ...
7037 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7038 (p9_xxbrw_v4sf): New expander.
7040 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7042 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7044 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7047 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7048 multiply in 32-bit mode.
7049 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7050 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7053 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7055 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7056 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7057 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7058 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7060 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7063 * dwarf2out.c (is_trivial_indirect_ref): New function.
7064 (dwarf2out_late_global_decl): Do not generate a location
7065 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7066 and that are not defined in the current unit.
7068 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7070 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7071 instead of a libcall for UNORDERED.
7073 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7076 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7077 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7079 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7082 * config/s390/s390.c (s390_set_current_function): Invoke
7083 s390_indirect_branch_settings also if fndecl didn't change.
7085 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7087 * config/rs6000/rs6000.md (blockage): Set length to zero.
7089 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7091 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7093 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7096 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7097 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7101 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7102 PARALLEL incoming that failed vt_get_decl_and_offset check.
7105 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7106 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7108 (get_variable_section): For decls in named .bss* sections pass true as
7109 second argument to bss_initializer_p.
7111 2018-02-09 Marek Polacek <polacek@redhat.com>
7112 Jakub Jelinek <jakub@redhat.com>
7115 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7116 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7117 Sync some changes from cxx_fold_indirect_ref.
7119 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7121 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7123 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7124 (BLOCK_INLINE_ENTRY_LABEL): New.
7125 (dwarf2out_var_location): Disregard inline entry markers.
7126 (inline_entry_data): New struct.
7127 (inline_entry_data_hasher): New hashtable type.
7128 (inline_entry_data_hasher::hash): New.
7129 (inline_entry_data_hasher::equal): New.
7130 (inline_entry_data_table): New variable.
7131 (add_high_low_attributes): Add DW_AT_entry_pc and
7132 DW_AT_GNU_entry_view attributes if a pending entry is found
7133 in inline_entry_data_table. Add old entry_pc attribute only
7134 if debug nonbinding markers are disabled.
7135 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7136 markers are enabled.
7137 (block_within_block_p, dwarf2out_inline_entry): New.
7138 (dwarf2out_finish): Check that no entries remained in
7139 inline_entry_data_table.
7140 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7141 (final_scan_insn, notice_source_line): Likewise.
7142 (rest_of_clean_state): Skip inline entry markers.
7143 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7145 * gimple.c (gimple_build_debug_inline_entry): New.
7146 * gimple.h (enum gimple_debug_subcode): Add
7147 GIMPLE_DEBUG_INLINE_ENTRY.
7148 (gimple_build_debug_inline_entry): Declare.
7149 (gimple_debug_inline_entry_p): New.
7150 (gimple_debug_nonbind_marker_p): Adjust.
7151 * insn-notes.def (INLINE_ENTRY): New.
7152 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7153 inline entry marker notes.
7154 (print_insn): Likewise.
7155 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7156 (INSN_DEBUG_MARKER_KIND): Likewise.
7157 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7158 * tree-inline.c (expand_call_inline): Build and insert
7159 debug_inline_entry stmt.
7160 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7161 inline entry blocks early, if nonbind markers are enabled.
7162 (dump_scope_block): Dump fragment info.
7163 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7164 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7165 (gimple_build_debug_inline_entry): New.
7166 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7167 Enable/disable inline entry points too.
7168 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7169 (DEBUG_INSN): Describe inline entry markers.
7171 * common.opt (gvariable-location-views): New.
7172 (gvariable-location-views=incompat5): New.
7173 * config.in: Rebuilt.
7174 * configure: Rebuilt.
7175 * configure.ac: Test assembler for view support.
7176 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7177 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7178 * dwarf2out.c (var_loc_view): New typedef.
7179 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7180 (dwarf2out_locviews_in_attribute): New.
7181 (dwarf2out_locviews_in_loclist): New.
7182 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7183 (enum dw_line_info_opcode): Add LI_adv_address.
7184 (struct dw_line_info_table): Add view.
7185 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7186 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7187 (zero_view_p): New variable.
7188 (ZERO_VIEW_P): New macro.
7189 (output_asm_line_debug_info): New.
7190 (struct var_loc_node): Add view.
7191 (add_AT_view_list, AT_loc_list): New.
7192 (add_var_loc_to_decl): Add view param. Test it against last.
7193 (new_loc_list): Add view params. Record them.
7194 (AT_loc_list_ptr): Handle loc and view lists.
7195 (view_list_to_loc_list_val_node): New.
7196 (print_dw_val): Handle dw_val_class_view_list.
7197 (size_of_die): Likewise.
7198 (value_format): Likewise.
7199 (loc_list_has_views): New.
7200 (gen_llsym): Set vl_symbol too.
7201 (maybe_gen_llsym, skip_loc_list_entry): New.
7202 (dwarf2out_maybe_output_loclist_view_pair): New.
7203 (output_loc_list): Output view list or entries too.
7204 (output_view_list_offset): New.
7205 (output_die): Handle dw_val_class_view_list.
7206 (output_dwarf_version): New.
7207 (output_compilation_unit_header): Use it.
7208 (output_skeleton_debug_sections): Likewise.
7209 (output_rnglists, output_line_info): Likewise.
7210 (output_pubnames, output_aranges): Update version comments.
7211 (output_one_line_info_table): Output view numbers in asm comments.
7212 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7213 Call maybe_gen_llsym.
7214 (loc_list_from_tree_1): Adjust.
7215 (add_AT_location_description): Create view list attribute if
7216 needed, check it's absent otherwise.
7217 (convert_cfa_to_fb_loc_list): Adjust.
7218 (maybe_emit_file): Call output_asm_line_debug_info for test.
7219 (dwarf2out_var_location): Reset views as needed. Precompute
7220 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7221 attribute. Set view.
7222 (new_line_info_table): Reset next view.
7223 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7224 (dwarf2out_source_line): Likewise. Output view resets and labels to
7225 the assembler, or select appropriate line info opcodes.
7226 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7227 (optimize_string_length): Catch it. Adjust.
7228 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7229 dw_val_class_view_list, and remove it if no longer needed.
7230 (hash_loc_list): Hash view numbers.
7231 (loc_list_hasher::equal): Compare them.
7232 (optimize_location_lists): Check whether a view list symbol is
7233 needed, and whether the locview attribute is present, and
7234 whether they match. Remove the locview attribute if no longer
7236 (index_location_lists): Call skip_loc_list_entry for test.
7237 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7238 Use output_dwarf_version.
7239 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7240 (struct dw_val_node): Add val_view_list.
7241 * final.c (SEEN_NEXT_VIEW): New.
7242 (set_next_view_needed): New.
7243 (clear_next_view_needed): New.
7244 (maybe_output_next_view): New.
7245 (final_start_function): Rename to...
7246 (final_start_function_1): ... this. Take pointer to FIRST,
7247 add SEEN parameter. Emit param bindings in the initial view.
7248 (final_start_function): Reintroduce SEEN-less interface.
7249 (final): Rename to...
7250 (final_1): ... this. Take SEEN parameter. Output final pending
7251 next view at the end.
7252 (final): Reintroduce seen-less interface.
7253 (final_scan_insn): Output pending next view before switching
7254 sections or ending a block. Mark the next view as needed when
7255 outputting variable locations. Notify debug backend of section
7256 changes, and of location view changes.
7257 (rest_of_handle_final): Adjust.
7258 * toplev.c (process_options): Autodetect value for debug variable
7259 location views option. Warn on incompat5 without -gdwarf-5.
7260 * doc/invoke.texi (gvariable-location-views): New.
7261 (gvariable-location-views=incompat5): New.
7262 (gno-variable-location-views): New.
7264 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7266 PR tree-optimization/84136
7267 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7268 that the result of find_edge is non-NULL.
7270 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7273 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7274 storing integer register in SImode. Fix cost of 256 and 512
7275 byte aligned SSE register store.
7277 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7279 * config/i386/i386.c (ix86_multiplication_cost): Fix
7280 multiplication cost for TARGET_AVX512DQ.
7282 2018-02-08 Marek Polacek <polacek@redhat.com>
7284 PR tree-optimization/84238
7285 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7288 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7290 PR tree-optimization/84265
7291 * tree-vect-stmts.c (vectorizable_store): Don't treat
7292 VMAT_CONTIGUOUS accesses as grouped.
7293 (vectorizable_load): Likewise.
7295 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7297 PR tree-optimization/81635
7298 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7299 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7300 (test_round_for_mask): New functions.
7301 (wide_int_cc_tests): Call test_round_for_mask.
7302 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7303 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7304 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7305 range returned by get_range_info.
7307 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7310 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7311 * symtab.c: Include builtins.h
7312 (symtab_node::output_to_lto_symbol_table_p): Move here
7313 from lto-streamer-out.c:output_symbol_p.
7314 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7315 (output_symbol_p): Move all logic to symtab.c
7316 (produce_symtab): Update.
7318 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7320 * config/s390/s390-opts.h (enum indirect_branch): Define.
7321 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7322 (s390_indirect_branch_via_thunk)
7323 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7324 (enum s390_indirect_branch_type): Define.
7325 * config/s390/s390.c (struct s390_frame_layout, struct
7326 machine_function): Remove.
7327 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7328 (indirect_branch_table_label_no, indirect_branch_table_name):
7330 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7331 (enum s390_indirect_branch_option): Define.
7332 (s390_return_addr_from_memory): New function.
7333 (s390_handle_string_attribute): New function.
7334 (s390_attribute_table): Add new attribute handler.
7335 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7336 (s390_indirect_branch_via_thunk): New function.
7337 (s390_indirect_branch_via_inline_thunk): New function.
7338 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7339 sibling call optimization for non z10 compiles.
7340 (s390_emit_call): Force indirect branch target to be a single
7341 register. Add r1 clobber for non-z10 compiles.
7342 (s390_emit_epilogue): Emit return jump via return_use expander.
7343 (s390_reorg): Handle JUMP_INSNs as execute targets.
7344 (s390_option_override_internal): Perform validity checks for the
7345 new command line options.
7346 (s390_indirect_branch_attrvalue): New function.
7347 (s390_indirect_branch_settings): New function.
7348 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7349 (s390_output_indirect_thunk_function): New function.
7350 (s390_code_end): Implement target hook.
7351 (s390_case_values_threshold): Implement target hook.
7352 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7354 * config/s390/s390.h (struct s390_frame_layout)
7355 (struct machine_function): Move here from s390.c.
7356 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7357 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7358 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7359 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7360 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7361 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7362 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7363 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7364 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7365 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7366 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7367 (mnemonic attribute): Add values which aren't recognized
7369 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7370 pattern for branch conversion. Fix mnemonic attribute.
7371 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7372 indirect branch via thunk if requested.
7373 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7374 ("*indirect_jump"): Disable for branch conversion using out of
7376 ("indirect_jump_via_thunk<mode>_z10")
7377 ("indirect_jump_via_thunk<mode>")
7378 ("indirect_jump_via_inlinethunk<mode>_z10")
7379 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7380 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7381 ("casesi_jump_via_inlinethunk<mode>_z10")
7382 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7383 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7384 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7385 ("*indirect2_jump"): Disable for branch conversion.
7386 ("casesi_jump"): Turn into expander and expand patterns for branch
7388 ("return_use"): New expander.
7389 ("*return"): Emit return via thunk and rename it to ...
7390 ("*return<mode>"): ... this one.
7391 * config/s390/s390.opt: Add new options and and enum for the
7394 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7396 * lra-constraints.c (match_reload): Unconditionally use
7397 gen_lowpart_SUBREG, rather than selecting between that
7398 and equivalent gen_rtx_SUBREG code.
7400 2018-02-08 Richard Biener <rguenther@suse.de>
7402 PR tree-optimization/84233
7403 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7404 changed flag instead of boguously re-using phi_inserted.
7406 2018-02-08 Martin Jambor <mjambor@suse.cz>
7408 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7409 static local variables.
7411 2018-02-08 Richard Biener <rguenther@suse.de>
7413 PR tree-optimization/84278
7414 * tree-vect-stmts.c (vectorizable_store): When looking for
7415 smaller vector types to perform grouped strided loads/stores
7416 make sure the mode is supported by the target.
7417 (vectorizable_load): Likewise.
7419 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7421 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7422 Increase LDP/STP opportunities by adding adjacent callee-saves.
7424 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7426 PR rtl-optimization/84068
7427 PR rtl-optimization/83459
7428 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7430 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7432 PR tree-optimization/84224
7433 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7434 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7437 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7440 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7441 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7442 count, remove one USE.
7444 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7446 * doc/install.texi (Configuration): Document the
7447 --with-long-double-format={ibm,ieee} PowerPC configuration
7451 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7452 Convert from define_expand to be define_insn_and_split. Rework
7453 float/double/_Float128 conversions to QI/HI/SImode to work with
7454 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7455 conversions to QI/HImode types did a store and then a load to
7456 truncate the value. For conversions to VSX registers, don't split
7457 the insn, instead emit the code directly. Use the code iterator
7458 any_fix to combine signed and unsigned conversions.
7459 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7460 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7461 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7462 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7463 (fix_<mode>di2_hw): Likewise.
7464 (fixuns_<mode>di2_hw): Likewise.
7465 (fix_<mode>si2_hw): Likewise.
7466 (fixuns_<mode>si2_hw): Likewise.
7467 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7468 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7469 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7470 fix<uns>_trunc<SFDF:mode>si2_p8.
7471 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7473 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7474 (fix<uns>_<mode>_mem): Likewise.
7475 (fctiw<u>z_<mode>_mem): Likewise.
7476 (fix<uns>_<mode>_mem): Likewise.
7477 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7478 the register allocator from doing a direct move to the GPRs to do
7479 a store, and instead use the ISA 3.0 store byte/half-word from
7480 vector register instruction. For IEEE 128-bit floating point,
7481 also optimize stores of 32-bit ints.
7482 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7484 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7486 * genextract.c (push_pathstr_operand): New function to support
7488 (walk_rtx): Call push_pathstr_operand.
7489 (print_path): Support [a-zA-Z].
7491 2018-02-07 Richard Biener <rguenther@suse.de>
7493 PR tree-optimization/84037
7494 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7495 (cse_and_gimplify_to_preheader): Declare.
7496 (vect_get_place_in_interleaving_chain): Likewise.
7497 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7499 (_loop_vec_info::~_loop_vec_info): Delete it.
7500 (cse_and_gimplify_to_preheader): New function.
7501 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7502 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7503 (vectorizable_load): Likewise. For grouped stores always base
7504 the IV on the first element.
7505 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7506 condition before gimplifying.
7508 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7510 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7511 *DIV_EXPR and *MOD_EXPR.
7513 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7516 * config/i386/i386.c (ix86_option_override_internal): Mask out
7517 the CF_SET bit when checking -fcf-protection.
7519 2018-02-07 Tom de Vries <tom@codesourcery.com>
7522 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7525 2018-02-07 Richard Biener <rguenther@suse.de>
7527 PR tree-optimization/84204
7528 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7531 PR tree-optimization/84205
7532 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7533 special-case isl_ast_op_zdiv_r.
7535 PR tree-optimization/84223
7536 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7537 Only add conditions from within the region.
7538 (gather_bbs::after_dom_children): Adjust.
7540 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7543 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7544 * config/avr/avr.md: Only post-reload split REG-REG moves if
7545 either register is GENERAL_REG_P.
7547 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7549 PR tree-optimization/84235
7550 * tree-ssa-scopedtables.c
7551 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7552 if the subtraction is performed in floating point type where NaNs are
7553 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7554 build 1. Formatting fix.
7556 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7559 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7560 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7561 and skip it regardless of bb boundaries. Use CALL_P macro,
7562 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7563 unnecessarily, formatting fix.
7565 2018-02-06 Michael Collison <michael.collison@arm.com>
7567 * config/arm/thumb2.md:
7568 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7569 (*thumb_mov_notscc): Ditto.
7571 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7574 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7575 unsigned_fix, not "s".
7577 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7579 * configure.ac (gcc_fn_eh_frame_ro): New function.
7580 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7581 correct .eh_frame permissions.
7582 * configure: Regenerate.
7584 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7586 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7589 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7591 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7592 Display warning message for -mno-speculate-indirect-jumps.
7594 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7596 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7598 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7600 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7602 PR tree-optimization/84225
7603 * tree-eh.c (find_trapping_overflow): Only call
7604 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7606 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7609 * config/i386/i386.c: Reimplement the check of possible options
7610 -mibt/-mshstk conbination. Change error messages.
7611 * doc/invoke.texi: Fix a typo: remove extra '='.
7613 2018-02-06 Marek Polacek <polacek@redhat.com>
7615 PR tree-optimization/84228
7616 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7618 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7621 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7622 emitted arch directives.
7623 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7624 __ARM_FEATURE_COPROC before changing architectures.
7626 2018-02-06 Richard Biener <rguenther@suse.de>
7628 * config/i386/i386.c (print_reg): Fix typo.
7629 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7631 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7633 * configure: Regenerate.
7635 2018-02-05 Martin Sebor <msebor@redhat.com>
7637 PR tree-optimization/83369
7638 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7641 2018-02-05 Martin Liska <mliska@suse.cz>
7643 * doc/invoke.texi: Cherry-pick upstream r323995.
7645 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7647 * ira.c (ira_init_register_move_cost): Adjust comment.
7649 2018-02-05 Martin Liska <mliska@suse.cz>
7651 PR gcov-profile/84137
7652 * doc/gcov.texi: Fix typo in documentation.
7654 2018-02-05 Martin Liska <mliska@suse.cz>
7656 PR gcov-profile/83879
7657 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7658 using dlopen functionality.
7660 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7662 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7663 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7664 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7665 _mm_maskz_range_round_ss): New intrinsics.
7666 (__builtin_ia32_rangesd128_round)
7667 (__builtin_ia32_rangess128_round): Remove.
7668 (__builtin_ia32_rangesd128_mask_round,
7669 __builtin_ia32_rangess128_mask_round): New builtins.
7670 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7671 __builtin_ia32_rangess128_round): Remove.
7672 (__builtin_ia32_rangesd128_mask_round,
7673 __builtin_ia32_rangess128_mask_round): New builtins.
7674 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7675 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7676 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7677 "<round_saeonly_constraint>")): Changed to ...
7678 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7679 "<round_saeonly_scalar_constraint>")): ... this.
7680 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7681 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7682 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7683 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7684 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7686 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7688 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7690 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7691 Remove all values except native, 8540 and 8548.
7693 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7695 * config/i386/i386.c (ix86_output_function_return): Pass
7696 INVALID_REGNUM, instead of -1, as invalid register number to
7697 indirect_thunk_name and output_indirect_thunk.
7699 2018-02-02 Julia Koval <julia.koval@intel.com>
7701 * config.gcc: Add -march=icelake.
7702 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7703 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7704 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7705 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7706 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7707 (processor_target_table): Add icelake.
7708 (ix86_option_override_internal): Handle new PTAs.
7709 (get_builtin_code_for_version): Handle icelake.
7710 (M_INTEL_COREI7_ICELAKE): New.
7711 (fold_builtin_cpu): Handle icelake.
7712 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7713 * doc/invoke.texi: Add -march=icelake.
7715 2018-02-02 Julia Koval <julia.koval@intel.com>
7717 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7718 to wide_int_bitmask.
7719 * wide-int-bitmask.h: New.
7721 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7724 * config/i386/i386.md: Replace Pmode with word_mode in
7725 builtin_setjmp_setup and builtin_longjmp to support x32.
7727 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7731 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7733 (rs6000_supported_cpu_names): New static variable.
7734 (linux_cpu_translation_table): Likewise.
7735 (elf_platform) <cpu>: Define new static variable and use it.
7736 Translate kernel AT_PLATFORM name to canonical name if needed.
7737 Error if platform name is unknown.
7739 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7742 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7744 2018-02-01 Jeff Law <law@redhat.com>
7747 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7748 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7749 the scratch if RELEASE_VIA_POP is false.
7750 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7751 If we have to save a temporary register, decrement SIZE appropriately.
7752 Pass new arguments to release_scratch_register_on_entry.
7753 (ix86_adjust_stack_and_probe): Likewise.
7754 (ix86_emit_probe_stack_range): Pass new arguments to
7755 release_scratch_register_on_entry.
7757 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7759 PR rtl-optimization/84157
7760 * combine.c (change_zero_ext): Use REG_P predicate in
7761 front of HARD_REGISTER_P predicate.
7763 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7765 * config/avr/avr.c (avr_option_override): Move disabling of
7766 -fdelete-null-pointer-checks to...
7767 * common/config/avr/avr-common.c (avr_option_optimization_table):
7770 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7772 PR tree-optimization/81635
7773 * tree-data-ref.c (split_constant_offset_1): For types that
7774 wrap on overflow, try to use range info to prove that wrapping
7777 2018-02-01 Renlin Li <renlin.li@arm.com>
7780 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7782 (aarch64_register_move_cost): Likewise.
7783 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7785 (REG_CLASS_NAMES): Likewise.
7786 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7787 TAILCALL_ADDR_REGS. Remove IP registers.
7788 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7790 2018-02-01 Richard Biener <rguenther@suse.de>
7792 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7793 for specifying RPO order and allow NULL for that.
7794 * domwalk.c (dom_walker::dom_walker): Likewise.
7795 (dom_walker::walk): Handle NULL RPO order.
7796 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7798 (rewrite_update_dom_walker): Likewise.
7799 (mark_def_dom_walker): Likewise.
7801 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7803 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7804 (aarch64_maybe_expand_sve_subreg_move): Declare.
7805 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7806 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7808 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7809 that are semantically a reverse operation.
7810 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7811 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7812 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7814 (aarch64_can_change_mode_class): For big-endian, forbid changes
7815 between two SVE modes if they have different element sizes.
7817 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7819 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7820 the TImode handling for big-endian targets.
7822 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7824 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7825 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7827 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7828 Remove BSWAP handing for big-endian targets and use the form of
7829 LD1RQ appropariate for the mode.
7831 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7833 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7834 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7837 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7840 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7841 check for operands that need to go through aarch64_sve_reload_be.
7843 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7845 PR tree-optimization/81661
7846 PR tree-optimization/84117
7847 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7848 * tree-eh.c: Include gimplify.h.
7849 (find_trapping_overflow, replace_trapping_overflow,
7850 rewrite_to_non_trapping_overflow): New functions.
7851 * tree-vect-loop.c: Include tree-eh.h.
7852 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7853 * tree-data-ref.c: Include tree-eh.h.
7854 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7856 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7858 PR rtl-optimization/84123
7859 * combine.c (change_zero_ext): Check if hard register satisfies
7860 can_change_dest_mode before calling gen_lowpart_SUBREG.
7862 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7865 * ira.c (ira_init_register_move_cost): Remove assert.
7867 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7869 PR rtl-optimization/84071
7870 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7871 * doc/tm.texi: Regenerate.
7873 2018-01-31 Richard Biener <rguenther@suse.de>
7875 PR tree-optimization/84132
7876 * tree-data-ref.c (analyze_miv_subscript): Properly
7877 check whether evolution_function_is_affine_multivariate_p
7878 before calling gcd_of_steps_may_divide_p.
7880 2018-01-31 Julia Koval <julia.koval@intel.com>
7883 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7884 * config/i386/i386.md (rdpid_rex64) New.
7885 (rdpid): Make 32bit only.
7887 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7890 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7891 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7893 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7896 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7898 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7900 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7902 PR rtl-optimization/84071
7903 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7904 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7906 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7908 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7909 (arc_attribute_table): Add 'aux' attribute.
7910 (arc_in_small_data_p): Consider aux like variables.
7911 (arc_is_aux_reg_p): New function.
7912 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7913 (arc_get_aux_arg): New function.
7914 (prepare_move_operands): Handle aux-register access.
7915 (arc_handle_aux_attribute): New function.
7916 * doc/extend.texi (ARC Variable attributes): Add subsection.
7918 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7920 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7921 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7922 (arc_attribute_table): Add 'uncached' attribute.
7923 (arc_print_operand): Print '.di' flag for uncached memory
7925 (arc_in_small_data_p): Do not consider for small data the uncached
7927 (arc_is_uncached_mem_p): New function.
7928 * config/arc/predicates.md (compact_store_memory_operand): Check
7929 for uncached memory accesses.
7930 (nonvol_nonimm_operand): Likewise.
7931 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7933 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7936 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7937 falign-loops=): Add Optimization flag.
7939 2018-01-30 Jeff Law <law@redhat.com>
7942 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7943 INT_REGISTERS_SAVED. Check it prior to calling
7944 get_scratch_register_on_entry.
7945 (ix86_adjust_stack_and_probe): Similarly.
7946 (ix86_emit_probe_stack_range): Similarly.
7947 (ix86_expand_prologue): Corresponding changes.
7949 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7952 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7953 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7955 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7958 * lra-constraints.c (curr_insn_transform): Process AND in the
7961 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7963 PR rtl-optimization/83986
7964 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7965 dependence against last_pending_memory_flush in addition to
7968 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7970 PR tree-optimization/81611
7971 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7974 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7977 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7980 2018-01-30 Richard Biener <rguenther@suse.de>
7981 Jakub Jelinek <jakub@redhat.com>
7983 PR tree-optimization/84111
7984 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7985 inner loops added during recursion, as they don't have up-to-date
7988 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7991 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7992 (can_inline_edge_by_limits_p): ... here.
7993 (can_early_inline_edge_p, check_callers,
7994 update_caller_keys, update_callee_keys, recursive_inlining,
7995 add_new_edges_to_heap, speculation_useful_p,
7996 inline_small_functions,
7997 inline_small_functions, flatten_function,
7998 inline_to_all_callers_1): Update.
8000 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8002 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8005 2018-01-30 Richard Biener <rguenther@suse.de>
8007 PR tree-optimization/83008
8008 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8009 invariant and constant vector uses in stmts when they need
8012 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8015 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8016 * configure: Regenerate.
8018 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8020 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8022 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8023 Use gen_rtx_REG rather than gen_lowpart.
8025 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8027 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8028 rather than 0 when creating partial subregs.
8030 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8032 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8035 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8038 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8039 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8040 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8041 flags. This restores the settings used before the 2017-07-24.
8042 Turning off pre increment/decrement/modify allows IVOPTS to
8043 optimize DF/SF loops where the index is an int.
8045 2018-01-29 Richard Biener <rguenther@suse.de>
8046 Kelvin Nilsen <kelvin@gcc.gnu.org>
8049 * tree-vect-stmts.c (vectorizable_call): Don't call
8050 targetm.vectorize_builtin_md_vectorized_function if callee is
8053 2018-01-22 Carl Love <cel@us.ibm.com>
8055 * doc/extend.tex: Fix typo in second arg in
8056 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8058 2018-01-29 Richard Biener <rguenther@suse.de>
8060 PR tree-optimization/84086
8061 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8062 (flush_ssaname_freelist): When SSA names were released reset
8063 the SCEV hash table.
8065 2018-01-29 Richard Biener <rguenther@suse.de>
8067 PR tree-optimization/84057
8068 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8069 removed paths when removing edges.
8071 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8073 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8074 -mfunction-return=@var{choice}.
8076 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8079 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8081 (layout::print_source_line): Likewise.
8082 (test_get_line_width_without_trailing_whitespace): Add test cases.
8084 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8087 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8090 2018-01-26 Jim Wilson <jimw@sifive.com>
8092 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8094 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8097 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8099 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8100 and CMP + SUB-immediate -> SUBS.
8102 2018-01-26 Martin Sebor <msebor@redhat.com>
8104 PR tree-optimization/83896
8105 * tree-ssa-strlen.c (get_string_len): Rename...
8106 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8107 Avoid assuming length is constant.
8108 (handle_char_store): Use HOST_WIDE_INT for string length.
8110 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8113 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8114 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8116 2018-01-26 Richard Biener <rguenther@suse.de>
8118 PR rtl-optimization/84003
8119 * dse.c (record_store): Only record redundant stores when
8120 the earlier store aliases at least all accesses the later one does.
8122 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8124 PR rtl-optimization/83985
8125 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8126 REG_CFA_RESTORE insns.
8127 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8128 REG_CFA_RESTORE insns here.
8131 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8132 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8134 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8136 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8137 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8138 (arc_init): Likewise.
8139 (arc_override_options): Likewise.
8140 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8142 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8144 * config/arc/arc.h (TARGET_DBNZ): Define.
8145 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8146 properly set the tune attribute.
8147 (dbnz): Use TARGET_DBNZ guard.
8148 * config/arc/arc.opt (mtune): Add core3 option.
8150 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8152 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8153 recognize new pic like addresses.
8154 (arc_delegitimize_address): Clean up.
8156 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8158 * config/arc/arc-arches.def: Option mrf16 valid for all
8160 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8161 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8162 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8163 * config/arc/arc-tables.opt: Regenerate.
8164 * config/arc/arc.c (arc_conditional_register_usage): Handle
8165 reduced register file case.
8166 (arc_file_start): Set must have build attributes.
8167 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8169 * config/arc/arc.opt (mrf16): Add new option.
8170 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8171 * config/arc/genmultilib.awk: Handle new mrf16 option.
8172 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8173 * config/arc/t-multilib: Regenerate.
8174 * doc/invoke.texi (ARC Options): Document mrf16 option.
8176 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8178 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8179 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8180 (arc_attribute_table): Add 'secure_call' attribute.
8181 (arc_print_operand): Print secure call operand.
8182 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8184 (arc_is_secure_call_p): New function. * config/arc/arc.md
8185 (call_i): Add support for sjli instruction.
8186 (call_value_i): Likewise.
8187 * config/arc/constraints.md (Csc): New constraint.
8189 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8190 John Eric Martin <John.Martin@emmicro-us.com>
8192 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8193 * config/arc/arc.c (_arc_jli_section): New struct.
8194 (arc_jli_section): New type.
8195 (rc_jli_sections): New static variable.
8196 (arc_handle_jli_attribute): New function.
8197 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8198 (arc_file_end): New function.
8199 (TARGET_ASM_FILE_END): Define.
8200 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8201 (arc_add_jli_section): New function.
8202 (jli_call_scan): Likewise.
8203 (arc_reorg): Call jli_call_scan.
8204 (arc_output_addsi): Remove 'S' from printing asm operand.
8205 (arc_is_jli_call_p): New function.
8206 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8208 (movhi_insn): Likewise.
8209 (movsi_insn): Likewise.
8210 (movsi_set_cc_insn): Likewise.
8211 (loadqi_update): Likewise.
8212 (load_zeroextendqisi_update): Likewise.
8213 (load_signextendqisi_update): Likewise.
8214 (loadhi_update): Likewise.
8215 (load_zeroextendhisi_update): Likewise.
8216 (load_signextendhisi_update): Likewise.
8217 (loadsi_update): Likewise.
8218 (loadsf_update): Likewise.
8219 (movsicc_insn): Likewise.
8220 (bset_insn): Likewise.
8221 (bxor_insn): Likewise.
8222 (bclr_insn): Likewise.
8223 (bmsk_insn): Likewise.
8224 (bicsi3_insn): Likewise.
8225 (cmpsi_cc_c_insn): Likewise.
8226 (movsi_ne): Likewise.
8227 (movsi_cond_exec): Likewise.
8228 (clrsbsi2): Likewise.
8239 (call_i): Remove 'S' asm letter, add jli instruction.
8240 (call_value_i): Likewise.
8241 * config/arc/arc.op (mjli-always): New option.
8242 * config/arc/constraints.md (Cji): New constraint.
8243 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8245 (subsf3_fpx): Likewise.
8246 (mulsf3_fpx): Likewise.
8247 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8249 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8251 * doc/invoke.texi (ARC): Document mjli-always option.
8253 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8255 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8256 avoid addition with 0 and use incw and decw where possible.
8258 2018-01-26 Richard Biener <rguenther@suse.de>
8260 PR tree-optimization/81082
8261 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8262 association if it requires casting to unsigned.
8263 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8264 from fold_plusminus_mult_expr to catch important cases late when
8265 range info is available.
8267 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8269 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8270 * configure.ac (hidden_linkonce): New test.
8271 * configure: Regenerate.
8272 * config.in: Regenerate.
8274 2018-01-26 Julia Koval <julia.koval@intel.com>
8276 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8277 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8278 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8279 _mm_mask_bitshuffle_epi64_mask): Fix type.
8280 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8281 USI_FTYPE_V4DI_V4DI_USI): Remove.
8282 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8283 __builtin_ia32_vpshufbitqmb256_mask,
8284 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8285 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8286 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8288 2018-01-26 Alan Modra <amodra@gmail.com>
8291 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8292 UNSPEC_VBPERMQ. Sort other unspecs.
8294 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8296 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8298 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8301 * predict.c (drop_profile): Do not push/pop cfun; update also
8303 (handle_missing_profiles): Fix logic looking for zero profiles.
8305 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8308 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8309 on functions with #pragma omp declare simd or functions with simd
8311 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8312 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8313 Remove trailing \n from warning_at calls.
8315 2018-01-25 Tom de Vries <tom@codesourcery.com>
8318 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8319 for neutered workers.
8321 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8324 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8325 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8327 2018-01-24 Jeff Law <law@redhat.com>
8330 * i386.c (get_probe_interval): Move to earlier point.
8331 (ix86_compute_frame_layout): If -fstack-clash-protection and
8332 the frame is larger than the probe interval, then use pushes
8333 to save registers rather than reg->mem moves.
8334 (ix86_expand_prologue): Remove conditional for int_registers_saved
8337 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8340 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8341 min/max for never referenced object.
8343 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8346 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8348 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8349 attributes from DECL_ATTRIBUTES (decl) without affecting
8350 DECL_ATTRIBUTES (current_function_decl).
8351 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8352 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8354 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8356 PR tree-optimization/83979
8357 * fold-const.c (fold_comparison): Use constant_boolean_node
8358 instead of boolean_{true,false}_node.
8360 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8362 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8365 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8367 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8368 Simplify the clause that sets the length attribute.
8369 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8370 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8371 clause that sets the length attribute.
8372 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8374 2018-01-24 Tom de Vries <tom@codesourcery.com>
8377 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8378 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8379 Add strict parameter.
8380 (prevent_branch_around_nothing): Insert dummy insn between branch to
8381 label and label with no ptx insn inbetween.
8382 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8384 2018-01-24 Tom de Vries <tom@codesourcery.com>
8387 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8388 for neutered threads in warp.
8389 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8391 2018-01-24 Richard Biener <rguenther@suse.de>
8393 PR tree-optimization/83176
8394 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8397 2018-01-24 Richard Biener <rguenther@suse.de>
8399 PR tree-optimization/82819
8400 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8401 code generating pluses that are no-ops in the target precision.
8403 2018-01-24 Richard Biener <rguenther@suse.de>
8406 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8408 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8410 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8411 to merge probabilities.
8412 * predict.c (probably_never_executed): Also mark as cold functions
8413 with global 0 profile and guessed local profile.
8414 * profile-count.c (profile_probability::combine_with_count): New
8416 * profile-count.h (profile_probability::operator*,
8417 profile_probability::operator*=, profile_probability::operator/,
8418 profile_probability::operator/=): Reduce precision to adjusted
8419 and set value to guessed on contradictory divisions.
8420 (profile_probability::combine_with_freq): Remove.
8421 (profile_probability::combine_wiht_count): Declare.
8422 (profile_count::force_nonzero):: Set to adjusted.
8423 (profile_count::probability_in):: Set quality to adjusted.
8424 * tree-ssa-tail-merge.c (replace_block_by): Use
8427 2018-01-23 Andrew Waterman <andrew@sifive.com>
8428 Jim Wilson <jimw@sifive.com>
8430 * config/riscv/riscv.c (riscv_stack_boundary): New.
8431 (riscv_option_override): Set riscv_stack_boundary. Handle
8432 riscv_preferred_stack_boundary_arg.
8433 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8434 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8435 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8436 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8437 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8438 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8440 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8443 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8444 of struct ix86_frame.
8445 (ix86_expand_epilogue): Likewise. Add a local variable for
8446 the reg_save_offset field in struct ix86_frame.
8448 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8450 PR tree-optimization/82604
8451 * tree-loop-distribution.c (enum partition_kind): New enum item
8452 PKIND_PARTIAL_MEMSET.
8453 (partition_builtin_p): Support above new enum item.
8454 (generate_code_for_partition): Ditto.
8455 (compute_access_range): Differentiate cases that equality can be
8456 proven at all loops, the innermost loops or no loops.
8457 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8458 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8459 (finalize_partitions, distribute_loop): Don't fuse partition of
8460 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8461 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8464 2018-01-23 Martin Liska <mliska@suse.cz>
8466 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8467 order to ignore the predictor.
8468 (PRED_POLYMORPHIC_CALL): Likewise.
8469 (PRED_RECURSIVE_CALL): Likewise.
8471 2018-01-23 Martin Liska <mliska@suse.cz>
8473 * tree-profile.c (tree_profiling): Print function header to
8474 aware reader which function we are working on.
8475 * value-prof.c (gimple_find_values_to_profile): Do not print
8476 not interesting value histograms.
8478 2018-01-23 Martin Liska <mliska@suse.cz>
8480 * profile-count.h (enum profile_quality): Add
8481 profile_uninitialized as the first value. Do not number values
8482 as they are zero based.
8483 (profile_count::verify): Update sanity check.
8484 (profile_probability::verify): Likewise.
8486 2018-01-23 Nathan Sidwell <nathan@acm.org>
8488 * doc/invoke.texi (ffor-scope): Deprecate.
8490 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8492 PR tree-optimization/83510
8493 * domwalk.c (set_all_edges_as_executable): New function.
8494 (dom_walker::dom_walker): Convert bool param
8495 "skip_unreachable_blocks" to enum reachability. Move setup of
8496 edge flags to set_all_edges_as_executable and only do it when
8497 reachability is REACHABLE_BLOCKS.
8498 * domwalk.h (enum dom_walker::reachability): New enum.
8499 (dom_walker::dom_walker): Convert bool param
8500 "skip_unreachable_blocks" to enum reachability.
8501 (set_all_edges_as_executable): New decl.
8502 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8503 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8505 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8506 but converting true to REACHABLE_BLOCKS.
8507 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8509 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8510 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8511 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8513 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8514 if check_all_array_refs will be called.
8516 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8518 * tree.c (selftest::test_location_wrappers): Add more test
8521 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8523 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8524 (selftest::test_bit_in_range): Likewise.
8526 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8529 * doc/sourcebuild.texi (vect_float): Say that the selector
8530 only describes the situation when -funsafe-math-optimizations is on.
8531 (vect_float_strict): Document.
8533 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8535 PR tree-optimization/83965
8536 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8537 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8538 instead of checking only for a reduction.
8539 (vect_recog_widen_sum_pattern): Likewise.
8541 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8543 * predict.c (probably_never_executed): Only use precise profile info.
8544 (compute_function_frequency): Skip after inlining hack since we now
8545 have quality checking.
8547 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8549 * profile-count.h (profile_probability::very_unlikely,
8550 profile_probability::unlikely, profile_probability::even): Set
8551 precision to guessed.
8553 2018-01-23 Richard Biener <rguenther@suse.de>
8555 PR tree-optimization/83963
8556 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8557 Properly terminate dominator walk when crossing the exit edge not
8558 when visiting its source block.
8560 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8563 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8564 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8566 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8568 PR tree-optimization/83957
8569 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8570 semicolon after for body surrounded by braces.
8572 PR tree-optimization/83081
8573 * profile-count.h (profile_probability::split): New method.
8574 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8575 Use profile_probability::split.
8576 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8577 when splitting a single conditional jump into 2.
8579 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8581 PR tree-optimization/69452
8582 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8585 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8587 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8588 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8589 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8591 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8593 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8594 * config/rl78/rl78.md (movdi): New define_expand.
8595 * config/rl78/rl78.c (rl78_split_movdi): New function.
8597 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8600 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8602 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8603 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8604 128-bit to produce an UNSPEC move to get the double word with the
8605 signbit and then a shift directly to do signbit.
8606 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8607 implementation with a new version that just does either a direct
8608 move or a regular move. Move memory interface to separate insns.
8609 Move insns so they are next to the expander.
8610 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8611 with signbit move. Split big and little endian case.
8612 (signbit<mode>2_dm_mem_le): Likewise.
8613 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8614 (signbit<mode>2_dm2): Likewise.
8616 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8618 * config/rl78/rl78.md (anddi3): New define_expand.
8620 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8622 * config/rl78/rl78.md (umindi3): New define_expand.
8624 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8626 * config/rl78/rl78.md (smindi3): New define_expand.
8628 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8630 * config/rl78/rl78.md (smaxdi3): New define_expand.
8632 2018-01-22 Carl Love <cel@us.ibm.com>
8634 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8635 LVX_V1TI): Add macro expansion.
8636 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8637 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8638 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8639 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8640 Change check to determine if the instruction is a byte reversing
8641 entry. Fix typo in comment.
8642 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8643 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8644 Add def_builtin calls for new builtins.
8645 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8646 Add define_insn expansion.
8648 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8650 * config/rl78/rl78.md (umaxdi3): New define_expand.
8652 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8654 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8655 for non-QImode registers.
8657 2018-01-22 Richard Biener <rguenther@suse.de>
8659 PR tree-optimization/83963
8660 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8661 including the loop exit block.
8662 (scop_detection::merge_sese): Likewise.
8663 (scop_detection::add_scop): Do it here instead.
8665 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8667 * doc/sourcebuild.texi (arm_softfloat): Document.
8669 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8672 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8673 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8674 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8676 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8677 David Edelsohn <dje.gcc@gmail.com>
8680 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8681 Change "crset eq" to "crset 2".
8682 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8683 (*call_indirect_aix<mode>_nospec): Likewise.
8684 (*call_value_indirect_aix<mode>_nospec): Likewise.
8685 (*call_indirect_elfv2<mode>_nospec): Likewise.
8686 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8687 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8688 change assembly output from . to $.
8689 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8690 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8691 (*tablejump<mode>_internal1_nospec): Likewise.
8693 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8696 * config/sh/sh_optimize_sett_clrt.cc:
8697 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8699 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8701 PR tree-optimization/83940
8702 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8703 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8704 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8705 use it to pass back the definition type.
8706 (vect_check_store_rhs): Likewise rhs_dt_out.
8707 (vect_build_gather_load_calls): Add a mask_dt argument and use
8708 it instead of a call to vect_is_simple_use.
8709 (vectorizable_store): Update calls to vect_check_load_store_mask
8710 and vect_check_store_rhs. Use the dt returned by the latter instead
8711 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8712 instead of calls to vect_is_simple_use. Pass the scalar rather
8713 than the vector operand to vect_is_simple_use when handling
8714 second and subsequent copies of an rhs value.
8715 (vectorizable_load): Update calls to vect_check_load_store_mask
8716 and vect_build_gather_load_calls. Use the cached mask_dt and
8717 gs_info.offset_dt instead of calls to vect_is_simple_use.
8719 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8722 * tree-emutls.c: Include gimplify.h.
8723 (lower_emutls_2): New function.
8724 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8725 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8726 it before further processing.
8729 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8730 UINTVAL (trueop1) instead of INTVAL (op1).
8732 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8736 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8737 INCOMING_FRAME_SP_OFFSET if not defined.
8738 (scan_trace): Add ENTRY argument. If true and
8739 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8740 emit a note to adjust the CFA offset.
8741 (create_cfi_notes): Adjust scan_trace callers.
8742 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8743 INCOMING_FRAME_SP_OFFSET in the CIE.
8744 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8745 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8747 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8748 * doc/tm.texi: Regenerated.
8750 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8752 PR rtl-optimization/83147
8753 * lra-constraints.c (remove_inheritance_pseudos): Use
8754 lra_substitute_pseudo_within_insn.
8756 2018-01-19 Tom de Vries <tom@codesourcery.com>
8757 Cesar Philippidis <cesar@codesourcery.com>
8760 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8762 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8765 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8766 spaces for function labels.
8768 2018-01-19 Martin Liska <mliska@suse.cz>
8770 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8771 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8772 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8773 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8774 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8775 (PRED_CONST_RETURN): Change from 69 to 65.
8776 (PRED_NULL_RETURN): Change from 91 to 71.
8777 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8778 (PRED_LOOP_GUARD): Change from 66 to 73.
8780 2018-01-19 Martin Liska <mliska@suse.cz>
8782 * predict.c (predict_insn_def): Add new assert.
8783 (struct branch_predictor): Change type to signed integer.
8784 (test_prediction_value_range): Amend test to cover
8786 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8787 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8788 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8789 (PRED_LOOP_IV_COMPARE): Likewise.
8790 * predict.h (PROB_UNINITIALIZED): Define new constant.
8792 2018-01-19 Martin Liska <mliska@suse.cz>
8794 * predict.c (dump_prediction): Add new format for
8795 analyze_brprob.py script which is enabled with -details
8797 * profile-count.h (precise_p): New function.
8799 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8801 PR tree-optimization/83922
8802 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8803 there are no statements that need masking.
8804 (vect_active_double_reduction_p): New function.
8805 (vect_analyze_loop_operations): Use it when handling phis that
8806 are not in the loop header.
8808 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8810 PR tree-optimization/83914
8811 * tree-vect-loop.c (vectorizable_induction): Don't convert
8812 init_expr or apply the peeling adjustment for inductions
8813 that are nested within the vectorized loop.
8815 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8817 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8820 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8824 * function.h (gimplify_parameters): Add gimple_seq * argument.
8825 * function.c: Include gimple.h and options.h.
8826 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8827 for the added local temporaries if needed.
8828 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8829 if there are any parameter cleanups, wrap whole body into a
8830 try/finally with the cleanups.
8832 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8835 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8836 Use GET_MODE_CLASS for scalar floating point.
8838 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8842 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8843 Fix call of call_cgraph_insertion_hooks.
8845 2018-01-18 Martin Sebor <msebor@redhat.com>
8847 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8849 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8852 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8855 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8858 * common.opt: (-ffile-prefix-map): New option.
8859 * opts.c (common_handle_option): Defer it.
8860 * opts-global.c (handle_common_deferred_options): Handle it.
8861 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8862 * file-prefix-map.h: New file.
8863 (remap_debug_filename, add_debug_prefix_map): ...here.
8864 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8865 * final.c (debug_prefix_map, add_debug_prefix_map
8866 remap_debug_filename): Move to...
8867 * file-prefix-map.c: New file.
8868 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8869 generalize, get rid of alloca(), use strrchr() instead of strchr().
8870 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8871 Implement in terms of add_prefix_map().
8872 (remap_macro_filename, remap_debug_filename): Implement in term of
8874 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8875 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8876 * dbxout.c: Include file-prefix-map.h.
8877 * varasm.c: Likewise.
8878 * vmsdbgout.c: Likewise.
8879 * xcoffout.c: Likewise.
8880 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8881 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8882 * doc/invoke.texi (-ffile-prefix-map): Document.
8883 (-fdebug-prefix-map): Update description.
8885 2018-01-18 Martin Liska <mliska@suse.cz>
8887 * config/i386/i386.c (indirect_thunk_name): Document that also
8889 (output_indirect_thunk): Document why both instructions
8890 (pause and lfence) are generated.
8892 2018-01-18 Richard Biener <rguenther@suse.de>
8894 PR tree-optimization/83887
8895 * graphite-scop-detection.c
8896 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8897 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8898 (scop_detection::merge_sese): Re-implement with a flood-fill
8899 algorithm that properly finds a SESE region if it exists.
8901 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8904 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8905 pointer_diff optimizations use view_convert instead of convert.
8907 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8909 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8910 Generate different code for -mno-speculate-indirect-jumps.
8911 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8912 (*call_indirect_aix<mode>): Disable for
8913 -mno-speculate-indirect-jumps.
8914 (*call_indirect_aix<mode>_nospec): New define_insn.
8915 (*call_value_indirect_aix<mode>): Disable for
8916 -mno-speculate-indirect-jumps.
8917 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8918 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8919 -mno-speculate-indirect-jumps.
8920 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8922 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8924 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8925 long double type, set the flags for noting the default long double
8926 type, even if we don't pass or return a long double type.
8928 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8931 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8934 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8936 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8937 support for merge[hl].
8938 (fold_mergehl_helper): New helper function.
8939 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8940 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8941 (altivec_vmrglw_direct): Add xxmrglw insn.
8943 2018-01-17 Andrew Waterman <andrew@sifive.com>
8945 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8946 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8948 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8951 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8952 call the lto_location_cache before reading the
8953 DECL_SOURCE_LOCATION of the types.
8955 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8956 Richard Sandiford <richard.sandiford@linaro.org>
8958 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8959 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8960 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8961 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8962 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8964 * config/aarch64/constraints.md (aarch64_movti_operand):
8966 * config/aarch64/predicates.md (Uti): Add new constraint.
8968 2018-01-17 Carl Love <cel@us.ibm.com>
8970 * config/rs6000/vsx.md (define_expand xl_len_r,
8971 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8972 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8974 (define_expand, define_insn): Move the shift left from the
8975 define_insn to the define_expand for lxvl and stxvl instructions.
8976 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8977 and XL_LEN_R definitions to PURE.
8979 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8981 * config/i386/i386.c (indirect_thunk_name): Declare regno
8982 as unsigned int. Compare regno with INVALID_REGNUM.
8983 (output_indirect_thunk): Ditto.
8984 (output_indirect_thunk_function): Ditto.
8985 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8986 in the call to output_indirect_thunk_function.
8988 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8991 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8992 rather than the size of inner_type to determine the stack slot size
8993 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8995 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8998 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9001 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9003 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9004 endian Linux systems to optionally enable multilibs for selecting
9005 the long double type if the user configured an explicit type.
9006 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9007 have no long double multilibs if not defined.
9008 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9009 warn if the user used -mabi={ieee,ibm}longdouble and we built
9010 multilibs for long double.
9011 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9012 appropriate multilib option.
9013 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9015 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9016 for building long double multilibs.
9017 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9019 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9021 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9024 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9026 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9029 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9032 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9035 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9037 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9038 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9040 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9042 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9043 different rtl trees depending on TARGET_64BIT.
9044 (rs6000_gen_lvx): Likewise.
9046 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9048 * config/visium/visium.md (nop): Tweak comment.
9049 (hazard_nop): Likewise.
9051 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9053 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9054 -mspeculate-indirect-jumps.
9055 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9056 for -mno-speculate-indirect-jumps.
9057 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9058 (*call_value_indirect_elfv2<mode>): Disable for
9059 -mno-speculate-indirect-jumps.
9060 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9061 (indirect_jump): Emit different RTL for
9062 -mno-speculate-indirect-jumps.
9063 (*indirect_jump<mode>): Disable for
9064 -mno-speculate-indirect-jumps.
9065 (*indirect_jump<mode>_nospec): New define_insn.
9066 (tablejump): Emit different RTL for
9067 -mno-speculate-indirect-jumps.
9068 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9069 (tablejumpsi_nospec): New define_expand.
9070 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9071 (tablejumpdi_nospec): New define_expand.
9072 (*tablejump<mode>_internal1): Disable for
9073 -mno-speculate-indirect-jumps.
9074 (*tablejump<mode>_internal1_nospec): New define_insn.
9075 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9078 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9080 * caller-save.c (insert_save): Drop unnecessary parameter. All
9083 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9084 Richard Biener <rguenth@suse.de>
9087 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9088 return early, inline manually is_gimple_sizepos. Make sure if we
9089 call gimplify_expr we don't end up with a gimple constant.
9090 * tree.c (variably_modified_type_p): Don't return true for
9091 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9092 * gimplify.h (is_gimple_sizepos): Remove.
9094 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9096 PR tree-optimization/83857
9097 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9098 vectorizable_live_operation for pure SLP statements.
9099 (vectorizable_live_operation): Handle PHIs.
9101 2018-01-16 Richard Biener <rguenther@suse.de>
9103 PR tree-optimization/83867
9104 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9105 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9107 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9110 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9111 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9112 If off is not INTEGER_CST, issue a may not be aligned warning
9113 rather than isn't aligned. Use isn%'t rather than isn't.
9114 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9116 <case MULT_EXPR>: Improve the case when bottom and one of the
9117 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9118 operand, in that case check if the other operand is multiple of
9119 bottom divided by the INTEGER_CST operand.
9121 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9124 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9125 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9126 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9127 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9128 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9129 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9130 (pa_function_arg_size): New function.
9132 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9134 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9135 in a separate statement.
9137 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9139 PR tree-optimization/83847
9140 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9141 group gathers and scatters.
9143 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9145 PR rtl-optimization/86620
9146 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9148 PR rtl-optimization/83213
9149 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9150 to last if both are JUMP_INSNs.
9152 PR tree-optimization/83843
9153 * gimple-ssa-store-merging.c
9154 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9155 store_immediate_info for bswap/nop orig_stores.
9157 2018-01-15 Andrew Waterman <andrew@sifive.com>
9159 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9161 <UDIV>: Increase cost if !TARGET_DIV.
9163 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9165 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9166 (define_attr "cr_logical_3op"): New.
9167 (cceq_ior_compare): Adjust.
9168 (cceq_ior_compare_complement): Adjust.
9169 (*cceq_rev_compare): Adjust.
9170 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9171 (is_cracked_insn): Adjust.
9172 (insn_must_be_first_in_group): Adjust.
9173 * config/rs6000/40x.md: Adjust.
9174 * config/rs6000/440.md: Adjust.
9175 * config/rs6000/476.md: Adjust.
9176 * config/rs6000/601.md: Adjust.
9177 * config/rs6000/603.md: Adjust.
9178 * config/rs6000/6xx.md: Adjust.
9179 * config/rs6000/7450.md: Adjust.
9180 * config/rs6000/7xx.md: Adjust.
9181 * config/rs6000/8540.md: Adjust.
9182 * config/rs6000/cell.md: Adjust.
9183 * config/rs6000/e300c2c3.md: Adjust.
9184 * config/rs6000/e500mc.md: Adjust.
9185 * config/rs6000/e500mc64.md: Adjust.
9186 * config/rs6000/e5500.md: Adjust.
9187 * config/rs6000/e6500.md: Adjust.
9188 * config/rs6000/mpc.md: Adjust.
9189 * config/rs6000/power4.md: Adjust.
9190 * config/rs6000/power5.md: Adjust.
9191 * config/rs6000/power6.md: Adjust.
9192 * config/rs6000/power7.md: Adjust.
9193 * config/rs6000/power8.md: Adjust.
9194 * config/rs6000/power9.md: Adjust.
9195 * config/rs6000/rs64.md: Adjust.
9196 * config/rs6000/titan.md: Adjust.
9198 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9200 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9201 ix86_indirect_branch_register logic.
9203 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9205 * config/i386/constraints.md (Bs): Update
9206 ix86_indirect_branch_register check. Don't check
9207 ix86_indirect_branch_register with GOT_memory_operand.
9209 * config/i386/predicates.md (GOT_memory_operand): Don't check
9210 ix86_indirect_branch_register here.
9211 (GOT32_symbol_operand): Likewise.
9213 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9215 * config/i386/predicates.md (constant_call_address_operand):
9216 Rewrite ix86_indirect_branch_register logic.
9217 (sibcall_insn_operand): Likewise.
9219 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9221 * config/i386/constraints.md (Bs): Replace
9222 ix86_indirect_branch_thunk_register with
9223 ix86_indirect_branch_register.
9225 * config/i386/i386.md (indirect_jump): Likewise.
9226 (tablejump): Likewise.
9227 (*sibcall_memory): Likewise.
9228 (*sibcall_value_memory): Likewise.
9229 Peepholes of indirect call and jump via memory: Likewise.
9230 * config/i386/i386.opt: Likewise.
9231 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9232 (GOT_memory_operand): Likewise.
9233 (call_insn_operand): Likewise.
9234 (sibcall_insn_operand): Likewise.
9235 (GOT32_symbol_operand): Likewise.
9237 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9240 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9241 type rather than type addr's type points to.
9242 (expand_omp_atomic_mutex): Likewise.
9243 (expand_omp_atomic): Likewise.
9245 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9248 * config/i386/i386.c (output_indirect_thunk_function): Use
9249 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9250 for __x86_return_thunk.
9252 2018-01-15 Richard Biener <rguenther@suse.de>
9255 * expmed.c (extract_bit_field_1): Fix typo.
9257 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9260 * config/arm/iterators.md (VF): New mode iterator.
9261 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9262 Remove integer-related logic from pattern.
9263 (neon_vabd<mode>_3): Likewise.
9265 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9268 * common.opt (fstrict-overflow): No longer an alias.
9269 (fwrapv-pointer): New option.
9270 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9271 also for pointer types based on flag_wrapv_pointer.
9272 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9273 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9274 opts->x_flag_wrapv got set.
9275 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9276 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9277 POINTER_TYPE_OVERFLOW_UNDEFINED.
9278 * match.pd: Likewise in address comparison pattern.
9279 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9281 2018-01-15 Richard Biener <rguenther@suse.de>
9284 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9285 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9286 Reset type names to their identifier if their TYPE_DECL doesn't
9287 have linkage (and thus is used for ODR and devirt).
9288 (save_debug_info_for_decl): Remove.
9289 (save_debug_info_for_type): Likewise.
9290 (add_tree_to_fld_list): Adjust.
9291 * tree-pretty-print.c (dump_generic_node): Make dumping of
9292 type names more robust.
9294 2018-01-15 Richard Biener <rguenther@suse.de>
9296 * BASE-VER: Bump to 8.0.1.
9298 2018-01-14 Martin Sebor <msebor@redhat.com>
9301 * builtins.c (check_access): Avoid warning when the no-warning bit
9304 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9306 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9307 * ira-color (allocno_hard_regs_compare): Likewise.
9309 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9312 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9313 Use .pushsection/.popsection.
9315 2018-01-14 Martin Sebor <msebor@redhat.com>
9318 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9320 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9322 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9323 entry from extra_headers.
9324 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9325 extra_headers, make the list bitwise identical to the i?86-*-* one.
9327 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9329 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9330 -mcmodel=large with -mindirect-branch=thunk,
9331 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9332 -mfunction-return=thunk-extern.
9333 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9334 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9335 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9337 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9339 * config/i386/i386.c (print_reg): Print the name of the full
9340 integer register without '%'.
9341 (ix86_print_operand): Handle 'V'.
9342 * doc/extend.texi: Document 'V' modifier.
9344 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9346 * config/i386/constraints.md (Bs): Disallow memory operand for
9347 -mindirect-branch-register.
9349 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9350 (GOT_memory_operand): Likewise.
9351 (call_insn_operand): Likewise.
9352 (sibcall_insn_operand): Likewise.
9353 (GOT32_symbol_operand): Likewise.
9354 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9355 for -mindirect-branch-register.
9356 (tablejump): Likewise.
9357 (*sibcall_memory): Likewise.
9358 (*sibcall_value_memory): Likewise.
9359 Disallow peepholes of indirect call and jump via memory for
9360 -mindirect-branch-register.
9361 (*call_pop): Replace m with Bw.
9362 (*call_value_pop): Likewise.
9363 (*sibcall_pop_memory): Replace m with Bs.
9364 * config/i386/i386.opt (mindirect-branch-register): New option.
9365 * doc/invoke.texi: Document -mindirect-branch-register option.
9367 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9369 * config/i386/i386-protos.h (ix86_output_function_return): New.
9370 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9371 set function_return_type.
9372 (indirect_thunk_name): Add ret_p to indicate thunk for function
9374 (output_indirect_thunk_function): Pass false to
9375 indirect_thunk_name.
9376 (ix86_output_indirect_branch_via_reg): Likewise.
9377 (ix86_output_indirect_branch_via_push): Likewise.
9378 (output_indirect_thunk_function): Create alias for function
9379 return thunk if regno < 0.
9380 (ix86_output_function_return): New function.
9381 (ix86_handle_fndecl_attribute): Handle function_return.
9382 (ix86_attribute_table): Add function_return.
9383 * config/i386/i386.h (machine_function): Add
9384 function_return_type.
9385 * config/i386/i386.md (simple_return_internal): Use
9386 ix86_output_function_return.
9387 (simple_return_internal_long): Likewise.
9388 * config/i386/i386.opt (mfunction-return=): New option.
9389 (indirect_branch): Mention -mfunction-return=.
9390 * doc/extend.texi: Document function_return function attribute.
9391 * doc/invoke.texi: Document -mfunction-return= option.
9393 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9395 * config/i386/i386-opts.h (indirect_branch): New.
9396 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9397 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9398 with local indirect jump when converting indirect call and jump.
9399 (ix86_set_indirect_branch_type): New.
9400 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9401 (indirectlabelno): New.
9402 (indirect_thunk_needed): Likewise.
9403 (indirect_thunk_bnd_needed): Likewise.
9404 (indirect_thunks_used): Likewise.
9405 (indirect_thunks_bnd_used): Likewise.
9406 (INDIRECT_LABEL): Likewise.
9407 (indirect_thunk_name): Likewise.
9408 (output_indirect_thunk): Likewise.
9409 (output_indirect_thunk_function): Likewise.
9410 (ix86_output_indirect_branch_via_reg): Likewise.
9411 (ix86_output_indirect_branch_via_push): Likewise.
9412 (ix86_output_indirect_branch): Likewise.
9413 (ix86_output_indirect_jmp): Likewise.
9414 (ix86_code_end): Call output_indirect_thunk_function if needed.
9415 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9417 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9418 (ix86_attribute_table): Add indirect_branch.
9419 * config/i386/i386.h (machine_function): Add indirect_branch_type
9420 and has_local_indirect_jump.
9421 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9423 (tablejump): Likewise.
9424 (*indirect_jump): Use ix86_output_indirect_jmp.
9425 (*tablejump_1): Likewise.
9426 (simple_return_indirect_internal): Likewise.
9427 * config/i386/i386.opt (mindirect-branch=): New option.
9428 (indirect_branch): New.
9431 (thunk-inline): Likewise.
9432 (thunk-extern): Likewise.
9433 * doc/extend.texi: Document indirect_branch function attribute.
9434 * doc/invoke.texi: Document -mindirect-branch= option.
9436 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9439 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9441 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9443 * ipa-inline.c (want_inline_small_function_p): Return false if
9444 inlining has already failed with CIF_FINAL_ERROR.
9445 (update_caller_keys): Call want_inline_small_function_p before
9447 (update_callee_keys): Likewise.
9449 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9451 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9453 (rs6000_quadword_masked_address_p): Likewise.
9454 (quad_aligned_load_p): Likewise.
9455 (quad_aligned_store_p): Likewise.
9456 (const_load_sequence_p): Add comment to describe the outer-most loop.
9457 (mimic_memory_attributes_and_flags): New function.
9458 (rs6000_gen_stvx): Likewise.
9459 (replace_swapped_aligned_store): Likewise.
9460 (rs6000_gen_lvx): Likewise.
9461 (replace_swapped_aligned_load): Likewise.
9462 (replace_swapped_load_constant): Capitalize argument name in
9463 comment describing this function.
9464 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9465 and stores that access quad-word aligned addresses and replace
9466 with stvx or lvx instructions when appropriate.
9467 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9468 New function prototype.
9469 (rs6000_quadword_masked_address_p): Likewise.
9470 (rs6000_gen_lvx): Likewise.
9471 (rs6000_gen_stvx): Likewise.
9472 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9473 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9474 when memory address is aligned.
9475 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9476 this split to select lvx instruction when memory address is aligned.
9477 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9478 instruction when memory address is aligned.
9479 (*vsx_le_perm_load_v16qi): Likewise.
9480 (four unnamed splitters): Modify to select the stvx instruction
9481 when memory is aligned.
9483 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9485 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9486 which appears in the queue multiple times.
9488 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9489 Alan Hayward <alan.hayward@arm.com>
9490 David Sherwood <david.sherwood@arm.com>
9492 * tree-vectorizer.h (vec_lower_bound): New structure.
9493 (_loop_vec_info): Add check_nonzero and lower_bounds.
9494 (LOOP_VINFO_CHECK_NONZERO): New macro.
9495 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9496 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9497 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9498 fields. Make seg_len the distance travelled, not including the
9500 (dr_direction_indicator): Declare.
9501 (dr_zero_step_indicator): Likewise.
9502 (dr_known_forward_stride_p): Likewise.
9503 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9505 (runtime_alias_check_p): Allow runtime alias checks with
9507 (operator ==): Compare access_size and align.
9508 (prune_runtime_alias_test_list): Rework for new distinction between
9509 the access_size and seg_len.
9510 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9512 (get_segment_min_max): New function.
9513 (create_intersect_range_checks): Use it.
9514 (dr_step_indicator): New function.
9515 (dr_direction_indicator): Likewise.
9516 (dr_zero_step_indicator): Likewise.
9517 (dr_known_forward_stride_p): Likewise.
9518 * tree-loop-distribution.c (data_ref_segment_size): Return
9519 DR_STEP * (niters - 1).
9520 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9522 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9523 (vect_preserves_scalar_order_p): New function, split out from...
9524 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9525 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9526 (vect_vfa_access_size): New function.
9527 (vect_vfa_align): Likewise.
9528 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9529 (dump_lower_bound): New function.
9530 (vect_check_lower_bound): Likewise.
9531 (vect_small_gap_p): Likewise.
9532 (vectorizable_with_step_bound_p): Likewise.
9533 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9534 depencies if the vectorization factor is 1. Convert the checks
9535 for nonzero steps into checks on the bounds of DR_STEP. Try using
9536 a bunds check for variable steps if the minimum required step is
9537 relatively small. Update calls to the dr_with_seg_len
9538 constructor and to vect_compile_time_alias.
9539 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9541 (vect_loop_versioning): Call it.
9542 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9544 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9546 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9547 Alan Hayward <alan.hayward@arm.com>
9548 David Sherwood <david.sherwood@arm.com>
9550 * doc/sourcebuild.texi (vect_scatter_store): Document.
9551 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9553 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9555 * genopinit.c (main): Add supports_vec_scatter_store and
9556 supports_vec_scatter_store_cached to target_optabs.
9557 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9558 IFN_MASK_SCATTER_STORE.
9559 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9561 * internal-fn.h (internal_store_fn_p): Declare.
9562 (internal_fn_stored_value_index): Likewise.
9563 * internal-fn.c (scatter_store_direct): New macro.
9564 (expand_scatter_store_optab_fn): New function.
9565 (direct_scatter_store_optab_supported_p): New macro.
9566 (internal_store_fn_p): New function.
9567 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9568 IFN_MASK_SCATTER_STORE.
9569 (internal_fn_mask_index): Likewise.
9570 (internal_fn_stored_value_index): New function.
9571 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9573 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9574 * optabs-query.c (supports_vec_scatter_store_p): New function.
9575 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9576 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9577 true for scatter stores.
9578 (vect_gather_scatter_fn_p): Handle scatter stores too.
9579 (vect_check_gather_scatter): Consider using scatter stores if
9580 supports_vec_scatter_store_p.
9581 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9583 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9584 internal_fn_stored_value_index.
9585 (check_load_store_masking): Handle scatter stores too.
9586 (vect_get_store_rhs): Make public.
9587 (vectorizable_call): Use internal_store_fn_p.
9588 (vectorizable_store): Handle scatter store internal functions.
9589 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9590 when deciding whether the end of the group has been reached.
9591 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9592 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9593 (mask_scatter_store<mode>): New insns.
9595 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9596 Alan Hayward <alan.hayward@arm.com>
9597 David Sherwood <david.sherwood@arm.com>
9599 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9600 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9601 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9603 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9604 Use vect_truncate_gather_scatter_offset if we can't treat the
9605 operation as a normal gather load or scatter store.
9606 (get_group_load_store_type): Take the gather_scatter_info
9607 as argument. Try using a gather load or scatter store for
9608 single-element groups.
9609 (get_load_store_type): Update calls to get_group_load_store_type
9610 and vect_use_strided_gather_scatters_p.
9612 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9613 Alan Hayward <alan.hayward@arm.com>
9614 David Sherwood <david.sherwood@arm.com>
9616 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9617 optional tree argument.
9618 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9620 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9621 but continue to use the current value as a fallback.
9622 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9623 to compare the updates.
9624 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9625 (get_load_store_type): Use it when handling a strided access.
9626 (vect_get_strided_load_store_ops): New function.
9627 (vect_get_data_ptr_increment): Likewise.
9628 (vectorizable_load): Handle strided gather loads. Always pass
9629 a step to vect_create_data_ref_ptr and bump_vector_ptr.
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 * doc/md.texi (gather_load@var{m}): Document.
9636 (mask_gather_load@var{m}): Likewise.
9637 * genopinit.c (main): Add supports_vec_gather_load and
9638 supports_vec_gather_load_cached to target_optabs.
9639 * optabs-tree.c (init_tree_optimization_optabs): Use
9640 ggc_cleared_alloc to allocate target_optabs.
9641 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9642 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9644 * internal-fn.h (internal_load_fn_p): Declare.
9645 (internal_gather_scatter_fn_p): Likewise.
9646 (internal_fn_mask_index): Likewise.
9647 (internal_gather_scatter_fn_supported_p): Likewise.
9648 * internal-fn.c (gather_load_direct): New macro.
9649 (expand_gather_load_optab_fn): New function.
9650 (direct_gather_load_optab_supported_p): New macro.
9651 (direct_internal_fn_optab): New function.
9652 (internal_load_fn_p): Likewise.
9653 (internal_gather_scatter_fn_p): Likewise.
9654 (internal_fn_mask_index): Likewise.
9655 (internal_gather_scatter_fn_supported_p): Likewise.
9656 * optabs-query.c (supports_at_least_one_mode_p): New function.
9657 (supports_vec_gather_load_p): Likewise.
9658 * optabs-query.h (supports_vec_gather_load_p): Declare.
9659 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9660 and memory_type field.
9661 (NUM_PATTERNS): Bump to 15.
9662 * tree-vect-data-refs.c: Include internal-fn.h.
9663 (vect_gather_scatter_fn_p): New function.
9664 (vect_describe_gather_scatter_call): Likewise.
9665 (vect_check_gather_scatter): Try using internal functions for
9666 gather loads. Recognize existing calls to a gather load function.
9667 (vect_analyze_data_refs): Consider using gather loads if
9668 supports_vec_gather_load_p.
9669 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9670 (vect_get_gather_scatter_offset_type): Likewise.
9671 (vect_convert_mask_for_vectype): Likewise.
9672 (vect_add_conversion_to_patterm): Likewise.
9673 (vect_try_gather_scatter_pattern): Likewise.
9674 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9675 (vect_vect_recog_func_ptrs): Add it.
9676 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9677 internal_fn_mask_index and internal_gather_scatter_fn_p.
9678 (check_load_store_masking): Take the gather_scatter_info as an
9679 argument and handle gather loads.
9680 (vect_get_gather_scatter_ops): New function.
9681 (vectorizable_call): Check internal_load_fn_p.
9682 (vectorizable_load): Likewise. Handle gather load internal
9684 (vectorizable_store): Update call to check_load_store_masking.
9685 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9686 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9687 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9688 (aarch64_gather_scale_operand_d): New predicates.
9689 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9690 (mask_gather_load<mode>): New insns.
9692 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9693 Alan Hayward <alan.hayward@arm.com>
9694 David Sherwood <david.sherwood@arm.com>
9696 * optabs.def (fold_left_plus_optab): New optab.
9697 * doc/md.texi (fold_left_plus_@var{m}): Document.
9698 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9699 * internal-fn.c (fold_left_direct): Define.
9700 (expand_fold_left_optab_fn): Likewise.
9701 (direct_fold_left_optab_supported_p): Likewise.
9702 * fold-const-call.c (fold_const_fold_left): New function.
9703 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9704 * tree-parloops.c (valid_reduction_p): New function.
9705 (gather_scalar_reductions): Use it.
9706 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9707 (vect_finish_replace_stmt): Declare.
9708 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9709 (needs_fold_left_reduction_p): New function, split out from...
9710 (vect_is_simple_reduction): ...here. Accept reductions that
9711 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9712 (vect_force_simple_reduction): Also store the reduction type in
9713 the assignment's STMT_VINFO_REDUC_TYPE.
9714 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9715 (merge_with_identity): New function.
9716 (vect_expand_fold_left): Likewise.
9717 (vectorize_fold_left_reduction): Likewise.
9718 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9719 scalar phi in place for it. Check for target support and reject
9720 cases that would reassociate the operation. Defer the transform
9721 phase to vectorize_fold_left_reduction.
9722 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9723 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9724 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9726 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9728 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9729 call to ifc_temp_var.
9731 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9732 Alan Hayward <alan.hayward@arm.com>
9733 David Sherwood <david.sherwood@arm.com>
9735 * target.def (legitimize_address_displacement): Take the original
9736 offset as a poly_int.
9737 * targhooks.h (default_legitimize_address_displacement): Update
9739 * targhooks.c (default_legitimize_address_displacement): Likewise.
9740 * doc/tm.texi: Regenerate.
9741 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9742 as an argument, moving assert of ad->disp == ad->disp_term to...
9743 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9744 Try calling targetm.legitimize_address_displacement before expanding
9745 the address rather than afterwards, and adjust for the new interface.
9746 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9747 Match the new hook interface. Handle SVE addresses.
9748 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9753 * Makefile.in (OBJS): Add early-remat.o.
9754 * target.def (select_early_remat_modes): New hook.
9755 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9756 * doc/tm.texi: Regenerate.
9757 * targhooks.h (default_select_early_remat_modes): Declare.
9758 * targhooks.c (default_select_early_remat_modes): New function.
9759 * timevar.def (TV_EARLY_REMAT): New timevar.
9760 * passes.def (pass_early_remat): New pass.
9761 * tree-pass.h (make_pass_early_remat): Declare.
9762 * early-remat.c: New file.
9763 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9765 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9768 Alan Hayward <alan.hayward@arm.com>
9769 David Sherwood <david.sherwood@arm.com>
9771 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9772 vfm1 with a bound_epilog parameter.
9773 (vect_do_peeling): Update calls accordingly, and move the prologue
9774 call earlier in the function. Treat the base bound_epilog as 0 for
9775 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9776 this base when peeling for gaps.
9777 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9778 with fully-masked loops.
9779 (vect_estimate_min_profitable_iters): Handle the single peeled
9780 iteration in that case.
9782 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9783 Alan Hayward <alan.hayward@arm.com>
9784 David Sherwood <david.sherwood@arm.com>
9786 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9787 single-element interleaving even if the size is not a power of 2.
9788 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9789 accesses for single-element interleaving if the group size is
9792 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9793 Alan Hayward <alan.hayward@arm.com>
9794 David Sherwood <david.sherwood@arm.com>
9796 * doc/md.texi (fold_extract_last_@var{m}): Document.
9797 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9798 * optabs.def (fold_extract_last_optab): New optab.
9799 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9800 * internal-fn.c (fold_extract_direct): New macro.
9801 (expand_fold_extract_optab_fn): Likewise.
9802 (direct_fold_extract_optab_supported_p): Likewise.
9803 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9804 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9805 EXTRACT_LAST_REDUCTION.
9806 (get_initial_def_for_reduction): Do not create an initial vector
9807 for EXTRACT_LAST_REDUCTION reductions.
9808 (vectorizable_reduction): Leave the scalar phi in place for
9809 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9810 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9811 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9812 transform phase to vectorizable_condition.
9813 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9815 (vect_finish_stmt_generation): ...here.
9816 (vect_finish_replace_stmt): New function.
9817 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9818 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9820 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9822 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9823 Alan Hayward <alan.hayward@arm.com>
9824 David Sherwood <david.sherwood@arm.com>
9826 * doc/md.texi (extract_last_@var{m}): Document.
9827 * optabs.def (extract_last_optab): New optab.
9828 * internal-fn.def (EXTRACT_LAST): New internal function.
9829 * internal-fn.c (cond_unary_direct): New macro.
9830 (expand_cond_unary_optab_fn): Likewise.
9831 (direct_cond_unary_optab_supported_p): Likewise.
9832 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9833 loops using EXTRACT_LAST.
9834 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9835 (extract_last_<mode>): ...this optab.
9836 (vec_extract<mode><Vel>): Update accordingly.
9838 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9839 Alan Hayward <alan.hayward@arm.com>
9840 David Sherwood <david.sherwood@arm.com>
9842 * target.def (empty_mask_is_expensive): New hook.
9843 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9844 * doc/tm.texi: Regenerate.
9845 * targhooks.h (default_empty_mask_is_expensive): Declare.
9846 * targhooks.c (default_empty_mask_is_expensive): New function.
9847 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9848 if the target says that empty masks are expensive.
9849 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9851 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9853 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9854 Alan Hayward <alan.hayward@arm.com>
9855 David Sherwood <david.sherwood@arm.com>
9857 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9858 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9859 (vect_use_loop_mask_for_alignment_p): New function.
9860 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9861 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9862 niters_skip argument. Make sure that the first niters_skip elements
9863 of the first iteration are inactive.
9864 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9865 Update call to vect_set_loop_masks_directly.
9866 (get_misalign_in_elems): New function, split out from...
9867 (vect_gen_prolog_loop_niters): ...here.
9868 (vect_update_init_of_dr): Take a code argument that specifies whether
9869 the adjustment should be added or subtracted.
9870 (vect_update_init_of_drs): Likewise.
9871 (vect_prepare_for_masked_peels): New function.
9872 (vect_do_peeling): Skip prologue peeling if we're using a mask
9873 instead. Update call to vect_update_inits_of_drs.
9874 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9876 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9877 alignment. Do not include the number of peeled iterations in
9878 the minimum threshold in that case.
9879 (vectorizable_induction): Adjust the start value down by
9880 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9881 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9882 Take the number of skipped iterations into account when calculating
9884 * tree-vect-stmts.c (vect_gen_while_not): New function.
9886 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9887 Alan Hayward <alan.hayward@arm.com>
9888 David Sherwood <david.sherwood@arm.com>
9890 * doc/sourcebuild.texi (vect_fully_masked): Document.
9891 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9893 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9895 (vect_analyze_loop_2): ...here. Don't check the vectorization
9896 factor against the number of loop iterations if the loop is
9899 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9900 Alan Hayward <alan.hayward@arm.com>
9901 David Sherwood <david.sherwood@arm.com>
9903 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9904 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9905 (dump_groups): Update accordingly.
9906 (iv_use::mem_type): New member variable.
9907 (address_p): New function.
9908 (record_use): Add a mem_type argument and initialize the new
9910 (record_group_use): Add a mem_type argument. Use address_p.
9911 Remove obsolete null checks of base_object. Update call to record_use.
9912 (find_interesting_uses_op): Update call to record_group_use.
9913 (find_interesting_uses_cond): Likewise.
9914 (find_interesting_uses_address): Likewise.
9915 (get_mem_type_for_internal_fn): New function.
9916 (find_address_like_use): Likewise.
9917 (find_interesting_uses_stmt): Try find_address_like_use before
9918 calling find_interesting_uses_op.
9919 (addr_offset_valid_p): Use the iv mem_type field as the type
9920 of the addressed memory.
9921 (add_autoinc_candidates): Likewise.
9922 (get_address_cost): Likewise.
9923 (split_small_address_groups_p): Use address_p.
9924 (split_address_groups): Likewise.
9925 (add_iv_candidate_for_use): Likewise.
9926 (autoinc_possible_for_pair): Likewise.
9927 (rewrite_groups): Likewise.
9928 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9929 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9930 (get_alias_ptr_type_for_ptr_address): New function.
9931 (rewrite_use_address): Rewrite address uses in calls that were
9932 identified by find_address_like_use.
9934 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9935 Alan Hayward <alan.hayward@arm.com>
9936 David Sherwood <david.sherwood@arm.com>
9938 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9940 * gimple-expr.h (is_gimple_addressable: Likewise.
9941 * gimple-expr.c (is_gimple_address): Likewise.
9942 * internal-fn.c (expand_call_mem_ref): New function.
9943 (expand_mask_load_optab_fn): Use it.
9944 (expand_mask_store_optab_fn): Likewise.
9946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9947 Alan Hayward <alan.hayward@arm.com>
9948 David Sherwood <david.sherwood@arm.com>
9950 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9951 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9952 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9953 (cond_umax@var{mode}): Document.
9954 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9955 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9956 (cond_umin_optab, cond_umax_optab): New optabs.
9957 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9958 (COND_IOR, COND_XOR): New internal functions.
9959 * internal-fn.h (get_conditional_internal_fn): Declare.
9960 * internal-fn.c (cond_binary_direct): New macro.
9961 (expand_cond_binary_optab_fn): Likewise.
9962 (direct_cond_binary_optab_supported_p): Likewise.
9963 (get_conditional_internal_fn): New function.
9964 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9965 Cope with reduction statements that are vectorized as calls rather
9967 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9968 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9969 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9970 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9971 (UNSPEC_COND_EOR): New unspecs.
9972 (optab): Add mappings for them.
9973 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9974 (sve_int_op, sve_fp_op): New int attributes.
9976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9977 Alan Hayward <alan.hayward@arm.com>
9978 David Sherwood <david.sherwood@arm.com>
9980 * optabs.def (while_ult_optab): New optab.
9981 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9982 * internal-fn.def (WHILE_ULT): New internal function.
9983 * internal-fn.h (direct_internal_fn_supported_p): New override
9984 that takes two types as argument.
9985 * internal-fn.c (while_direct): New macro.
9986 (expand_while_optab_fn): New function.
9987 (convert_optab_supported_p): Likewise.
9988 (direct_while_optab_supported_p): New macro.
9989 * wide-int.h (wi::udiv_ceil): New function.
9990 * tree-vectorizer.h (rgroup_masks): New structure.
9991 (vec_loop_masks): New typedef.
9992 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9994 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9995 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9996 (vect_max_vf): New function.
9997 (slpeel_make_loop_iterate_ntimes): Delete.
9998 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9999 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10000 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10001 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10002 internal-fn.h, stor-layout.h and optabs-query.h.
10003 (vect_set_loop_mask): New function.
10004 (add_preheader_seq): Likewise.
10005 (add_header_seq): Likewise.
10006 (interleave_supported_p): Likewise.
10007 (vect_maybe_permute_loop_masks): Likewise.
10008 (vect_set_loop_masks_directly): Likewise.
10009 (vect_set_loop_condition_masked): Likewise.
10010 (vect_set_loop_condition_unmasked): New function, split out from
10011 slpeel_make_loop_iterate_ntimes.
10012 (slpeel_make_loop_iterate_ntimes): Rename to..
10013 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10014 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10015 (vect_do_peeling): Update call accordingly.
10016 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10018 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10019 mask_compare_type, can_fully_mask_p and fully_masked_p.
10020 (release_vec_loop_masks): New function.
10021 (_loop_vec_info): Use it to free the loop masks.
10022 (can_produce_all_loop_masks_p): New function.
10023 (vect_get_max_nscalars_per_iter): Likewise.
10024 (vect_verify_full_masking): Likewise.
10025 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10026 retries, and free the mask rgroups before retrying. Check loop-wide
10027 reasons for disallowing fully-masked loops. Make the final decision
10028 about whether use a fully-masked loop or not.
10029 (vect_estimate_min_profitable_iters): Do not assume that peeling
10030 for the number of iterations will be needed for fully-masked loops.
10031 (vectorizable_reduction): Disable fully-masked loops.
10032 (vectorizable_live_operation): Likewise.
10033 (vect_halve_mask_nunits): New function.
10034 (vect_double_mask_nunits): Likewise.
10035 (vect_record_loop_mask): Likewise.
10036 (vect_get_loop_mask): Likewise.
10037 (vect_transform_loop): Handle the case in which the final loop
10038 iteration might handle a partial vector. Call vect_set_loop_condition
10039 instead of slpeel_make_loop_iterate_ntimes.
10040 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10041 (check_load_store_masking): New function.
10042 (prepare_load_store_mask): Likewise.
10043 (vectorizable_store): Handle fully-masked loops.
10044 (vectorizable_load): Likewise.
10045 (supportable_widening_operation): Use vect_halve_mask_nunits for
10047 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10048 (vect_gen_while): New function.
10049 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10050 (aarch64_uqdec<mode>): New insn.
10052 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10053 Alan Hayward <alan.hayward@arm.com>
10054 David Sherwood <david.sherwood@arm.com>
10056 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10057 (reduc_xor_scal_optab): New optabs.
10058 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10059 (reduc_xor_scal_@var{m}): Document.
10060 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10061 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10062 internal functions.
10063 * fold-const-call.c (fold_const_call): Handle them.
10064 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10065 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10066 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10067 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10068 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10069 (UNSPEC_XORV): New unspecs.
10070 (optab): Add entries for them.
10071 (BITWISEV): New int iterator.
10072 (bit_reduc_op): New int attributes.
10074 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10075 Alan Hayward <alan.hayward@arm.com>
10076 David Sherwood <david.sherwood@arm.com>
10078 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10079 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10080 * optabs.def (vec_shl_insert_optab): New optab.
10081 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10082 (duplicate_and_interleave): Likewise.
10083 * tree-vect-loop.c: Include internal-fn.h.
10084 (neutral_op_for_slp_reduction): New function, split out from
10085 get_initial_defs_for_reduction.
10086 (get_initial_def_for_reduction): Handle option 2 for variable-length
10087 vectors by loading the neutral value into a vector and then shifting
10088 the initial value into element 0.
10089 (get_initial_defs_for_reduction): Replace the code argument with
10090 the neutral value calculated by neutral_op_for_slp_reduction.
10091 Use gimple_build_vector for constant-length vectors.
10092 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10093 but the first group_size elements have a neutral value.
10094 Use duplicate_and_interleave otherwise.
10095 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10096 Update call to get_initial_defs_for_reduction. Handle SLP
10097 reductions for variable-length vectors by creating one vector
10098 result for each scalar result, with the elements associated
10099 with other scalar results stubbed out with the neutral value.
10100 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10101 Require IFN_VEC_SHL_INSERT for double reductions on
10102 variable-length vectors, or SLP reductions that have
10103 a neutral value. Require can_duplicate_and_interleave_p
10104 support for variable-length unchained SLP reductions if there
10105 is no neutral value, such as for MIN/MAX reductions. Also require
10106 the number of vector elements to be a multiple of the number of
10107 SLP statements when doing variable-length unchained SLP reductions.
10108 Update call to vect_create_epilog_for_reduction.
10109 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10110 and remove initial values.
10111 (duplicate_and_interleave): Make public.
10112 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10113 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10115 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10116 Alan Hayward <alan.hayward@arm.com>
10117 David Sherwood <david.sherwood@arm.com>
10119 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10120 (can_duplicate_and_interleave_p): New function.
10121 (vect_get_and_check_slp_defs): Take the vector of statements
10122 rather than just the current one. Remove excess parentheses.
10123 Restriction rejectinon of vect_constant_def and vect_external_def
10124 for variable-length vectors to boolean types, or types for which
10125 can_duplicate_and_interleave_p is false.
10126 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10127 (duplicate_and_interleave): New function.
10128 (vect_get_constant_vectors): Use gimple_build_vector for
10129 constant-length vectors and suitable variable-length constant
10130 vectors. Use duplicate_and_interleave for other variable-length
10131 vectors. Don't defer the update when inserting new statements.
10133 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10134 Alan Hayward <alan.hayward@arm.com>
10135 David Sherwood <david.sherwood@arm.com>
10137 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10138 min_profitable_iters doesn't go negative.
10140 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10141 Alan Hayward <alan.hayward@arm.com>
10142 David Sherwood <david.sherwood@arm.com>
10144 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10145 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10146 * optabs.def (vec_mask_load_lanes_optab): New optab.
10147 (vec_mask_store_lanes_optab): Likewise.
10148 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10149 (MASK_STORE_LANES): Likewise.
10150 * internal-fn.c (mask_load_lanes_direct): New macro.
10151 (mask_store_lanes_direct): Likewise.
10152 (expand_mask_load_optab_fn): Handle masked operations.
10153 (expand_mask_load_lanes_optab_fn): New macro.
10154 (expand_mask_store_optab_fn): Handle masked operations.
10155 (expand_mask_store_lanes_optab_fn): New macro.
10156 (direct_mask_load_lanes_optab_supported_p): Likewise.
10157 (direct_mask_store_lanes_optab_supported_p): Likewise.
10158 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10160 (vect_load_lanes_supported): Likewise.
10161 * tree-vect-data-refs.c (strip_conversion): New function.
10162 (can_group_stmts_p): Likewise.
10163 (vect_analyze_data_ref_accesses): Use it instead of checking
10164 for a pair of assignments.
10165 (vect_store_lanes_supported): Take a masked_p parameter.
10166 (vect_load_lanes_supported): Likewise.
10167 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10168 vect_store_lanes_supported and vect_load_lanes_supported.
10169 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10170 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10171 parameter. Don't allow gaps for masked accesses.
10172 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10173 and vect_load_lanes_supported.
10174 (get_load_store_type): Take a masked_p parameter and update
10175 call to get_group_load_store_type.
10176 (vectorizable_store): Update call to get_load_store_type.
10177 Handle IFN_MASK_STORE_LANES.
10178 (vectorizable_load): Update call to get_load_store_type.
10179 Handle IFN_MASK_LOAD_LANES.
10181 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10182 Alan Hayward <alan.hayward@arm.com>
10183 David Sherwood <david.sherwood@arm.com>
10185 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10187 * config/aarch64/aarch64-protos.h
10188 (aarch64_sve_struct_memory_operand_p): Declare.
10189 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10190 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10191 (VPRED, vpred): Handle SVE structure modes.
10192 * config/aarch64/constraints.md (Utx): New constraint.
10193 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10194 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10195 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10196 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10197 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10198 structure modes. Split into pieces after RA.
10199 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10200 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10202 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10203 SVE structure modes.
10204 (aarch64_classify_address): Likewise.
10205 (sizetochar): Move earlier in file.
10206 (aarch64_print_operand): Handle SVE register lists.
10207 (aarch64_array_mode): New function.
10208 (aarch64_sve_struct_memory_operand_p): Likewise.
10209 (TARGET_ARRAY_MODE): Redefine.
10211 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10212 Alan Hayward <alan.hayward@arm.com>
10213 David Sherwood <david.sherwood@arm.com>
10215 * target.def (array_mode): New target hook.
10216 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10217 * doc/tm.texi: Regenerate.
10218 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10219 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10220 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10221 targetm.array_mode.
10222 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10226 Alan Hayward <alan.hayward@arm.com>
10227 David Sherwood <david.sherwood@arm.com>
10229 * fold-const.c (fold_binary_loc): Check the argument types
10230 rather than the result type when testing for a vector operation.
10232 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10234 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10235 * doc/tm.texi: Regenerate.
10237 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10238 Alan Hayward <alan.hayward@arm.com>
10239 David Sherwood <david.sherwood@arm.com>
10241 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10242 (sve): Document new AArch64 extension.
10243 * doc/md.texi (w): Extend the description of the AArch64
10244 constraint to include SVE vectors.
10245 (Upl, Upa): Document new AArch64 predicate constraints.
10246 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10248 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10249 (msve-vector-bits=): New option.
10250 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10251 SVE when these are disabled.
10252 (sve): New extension.
10253 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10254 modes. Adjust their number of units based on aarch64_sve_vg.
10255 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10256 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10257 aarch64_addr_query_type.
10258 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10259 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10260 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10261 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10262 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10263 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10264 (aarch64_simd_imm_zero_p): Delete.
10265 (aarch64_check_zero_based_sve_index_immediate): Declare.
10266 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10267 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10268 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10269 (aarch64_sve_float_mul_immediate_p): Likewise.
10270 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10271 rather than an rtx.
10272 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10273 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10274 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10275 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10276 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10277 (aarch64_regmode_natural_size): Likewise.
10278 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10279 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10281 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10282 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10283 for VG and the SVE predicate registers.
10284 (V_ALIASES): Add a "z"-prefixed alias.
10285 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10286 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10287 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10288 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10289 (REG_CLASS_NAMES): Add entries for them.
10290 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10291 and the predicate registers.
10292 (aarch64_sve_vg): Declare.
10293 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10294 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10295 (REGMODE_NATURAL_SIZE): Define.
10296 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10298 * config/aarch64/aarch64.c: Include cfgrtl.h.
10299 (simd_immediate_info): Add a constructor for series vectors,
10300 and an associated step field.
10301 (aarch64_sve_vg): New variable.
10302 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10303 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10304 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10305 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10306 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10307 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10308 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10309 (aarch64_get_mask_mode): New functions.
10310 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10311 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10312 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10313 predicate modes and predicate registers. Explicitly restrict
10314 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10315 to store a vector mode if it is recognized by
10316 aarch64_classify_vector_mode.
10317 (aarch64_regmode_natural_size): New function.
10318 (aarch64_hard_regno_caller_save_mode): Return the original mode
10320 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10321 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10322 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10323 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10325 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10326 does not overlap dest if the function is frame-related. Handle
10328 (aarch64_split_add_offset): New function.
10329 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10330 them aarch64_add_offset.
10331 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10332 and update call to aarch64_sub_sp.
10333 (aarch64_add_cfa_expression): New function.
10334 (aarch64_expand_prologue): Pass extra temporary registers to the
10335 functions above. Handle the case in which we need to emit new
10336 DW_CFA_expressions for registers that were originally saved
10337 relative to the stack pointer, but now have to be expressed
10338 relative to the frame pointer.
10339 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10341 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10342 IP0 and IP1 values for SVE frames.
10343 (aarch64_expand_vec_series): New function.
10344 (aarch64_expand_sve_widened_duplicate): Likewise.
10345 (aarch64_expand_sve_const_vector): Likewise.
10346 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10347 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10348 into the register, rather than emitting a SET directly.
10349 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10350 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10351 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10352 (offset_9bit_signed_scaled_p): New functions.
10353 (aarch64_replicate_bitmask_imm): New function.
10354 (aarch64_bitmask_imm): Use it.
10355 (aarch64_cannot_force_const_mem): Reject expressions involving
10356 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10357 (aarch64_classify_index): Handle SVE indices, by requiring
10358 a plain register index with a scale that matches the element size.
10359 (aarch64_classify_address): Handle SVE addresses. Assert that
10360 the mode of the address is VOIDmode or an integer mode.
10361 Update call to aarch64_classify_symbol.
10362 (aarch64_classify_symbolic_expression): Update call to
10363 aarch64_classify_symbol.
10364 (aarch64_const_vec_all_in_range_p): New function.
10365 (aarch64_print_vector_float_operand): Likewise.
10366 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10367 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10368 and the FP immediates 1.0 and 0.5.
10369 (aarch64_print_address_internal): Handle SVE addresses.
10370 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10371 (aarch64_regno_regclass): Handle predicate registers.
10372 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10374 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10375 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10376 (aarch64_convert_sve_vector_bits): New function.
10377 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10378 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10379 rather than an rtx.
10380 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10381 Handle SVE vector and predicate modes. Accept VL-based constants
10382 that need only one temporary register, and VL offsets that require
10383 no temporary registers.
10384 (aarch64_conditional_register_usage): Mark the predicate registers
10385 as fixed if SVE isn't available.
10386 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10387 Return true for SVE vector and predicate modes.
10388 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10389 rather than an unsigned int. Handle SVE modes.
10390 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10392 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10394 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10395 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10396 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10397 (aarch64_sve_float_mul_immediate_p): New functions.
10398 (aarch64_sve_valid_immediate): New function.
10399 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10400 Explicitly reject structure modes. Check for INDEX constants.
10401 Handle PTRUE and PFALSE constants.
10402 (aarch64_check_zero_based_sve_index_immediate): New function.
10403 (aarch64_simd_imm_zero_p): Delete.
10404 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10405 vector modes. Accept constants in the range of CNT[BHWD].
10406 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10407 ask for an Advanced SIMD mode.
10408 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10409 (aarch64_simd_vector_alignment): Handle SVE predicates.
10410 (aarch64_vectorize_preferred_vector_alignment): New function.
10411 (aarch64_simd_vector_alignment_reachable): Use it instead of
10413 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10414 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10416 (MAX_VECT_LEN): Delete.
10417 (expand_vec_perm_d): Add a vec_flags field.
10418 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10419 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10420 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10422 (aarch64_evpc_rev): Rename to...
10423 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10424 (aarch64_evpc_rev_global): New function.
10425 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10426 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10428 (aarch64_evpc_sve_tbl): New function.
10429 (aarch64_expand_vec_perm_const_1): Update after rename of
10430 aarch64_evpc_rev. Handle SVE permutes too, trying
10431 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10432 than aarch64_evpc_tbl.
10433 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10434 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10435 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10436 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10437 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10438 (aarch64_expand_sve_vcond): New functions.
10439 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10440 of aarch64_vector_mode_p.
10441 (aarch64_dwarf_poly_indeterminate_value): New function.
10442 (aarch64_compute_pressure_classes): Likewise.
10443 (aarch64_can_change_mode_class): Likewise.
10444 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10445 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10446 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10447 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10448 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10449 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10450 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10451 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10453 (Dn, Dl, Dr): Accept const as well as const_vector.
10454 (Dz): Likewise. Compare against CONST0_RTX.
10455 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10456 of "vector" where appropriate.
10457 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10458 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10459 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10460 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10461 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10462 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10463 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10464 (v_int_equiv): Extend to SVE modes.
10465 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10467 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10468 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10469 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10470 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10471 (SVE_COND_FP_CMP): New int iterators.
10472 (perm_hilo): Handle the new unpack unspecs.
10473 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10475 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10476 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10477 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10478 (aarch64_equality_operator, aarch64_constant_vector_operand)
10479 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10480 (aarch64_sve_nonimmediate_operand): Likewise.
10481 (aarch64_sve_general_operand): Likewise.
10482 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10483 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10484 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10485 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10486 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10487 (aarch64_sve_float_arith_immediate): Likewise.
10488 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10489 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10490 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10491 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10492 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10493 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10494 (aarch64_sve_float_arith_operand): Likewise.
10495 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10496 (aarch64_sve_float_mul_operand): Likewise.
10497 (aarch64_sve_vec_perm_operand): Likewise.
10498 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10499 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10500 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10501 as well as const_vector.
10502 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10503 in file. Use CONST0_RTX and CONSTM1_RTX.
10504 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10505 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10506 Use aarch64_simd_imm_zero.
10507 * config/aarch64/aarch64-sve.md: New file.
10508 * config/aarch64/aarch64.md: Include it.
10509 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10510 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10511 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10512 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10513 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10514 (sve): New attribute.
10515 (enabled): Disable instructions with the sve attribute unless
10517 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10518 aarch64_expand_mov_immediate.
10519 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10520 CNT[BHSD] immediates.
10521 (movti): Split CONST_POLY_INT moves into two halves.
10522 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10523 Split additions that need a temporary here if the destination
10524 is the stack pointer.
10525 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10526 (*add<mode>3_poly_1): New instruction.
10527 (set_clobber_cc): New expander.
10529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10531 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10532 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10533 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10534 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10535 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10536 Change innermode from fixed_mode_size to machine_mode.
10537 (simplify_subreg): Update call accordingly. Handle a constant-sized
10538 subreg of a variable-length CONST_VECTOR.
10540 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10541 Alan Hayward <alan.hayward@arm.com>
10542 David Sherwood <david.sherwood@arm.com>
10544 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10545 (add_offset_to_base): New function, split out from...
10546 (create_mem_ref): ...here. When handling a scale other than 1,
10547 check first whether the address is valid without the offset.
10548 Add it into the base if so, leaving the index and scale as-is.
10550 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10553 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10554 fold_for_warn before checking if arg2 is INTEGER_CST.
10556 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10558 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10559 (store_multiple_operation): Delete.
10560 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10561 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10562 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10563 guarded by TARGET_STRING.
10564 (rs6000_output_load_multiple): Delete.
10565 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10566 OPTION_MASK_STRING / TARGET_STRING handling.
10567 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10568 (const rs6000_opt_masks) <"string">: Change mask to 0.
10569 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10570 (MASK_STRING): Delete.
10571 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10573 (load_multiple): Delete.
10580 (store_multiple): Delete.
10587 (movmemsi_8reg): Delete.
10588 (corresponding unnamed define_insn): Delete.
10589 (movmemsi_6reg): Delete.
10590 (corresponding unnamed define_insn): Delete.
10591 (movmemsi_4reg): Delete.
10592 (corresponding unnamed define_insn): Delete.
10593 (movmemsi_2reg): Delete.
10594 (corresponding unnamed define_insn): Delete.
10595 (movmemsi_1reg): Delete.
10596 (corresponding unnamed define_insn): Delete.
10597 * config/rs6000/rs6000.opt (mno-string): New.
10598 (mstring): Replace by deprecation warning stub.
10599 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10601 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10603 * regrename.c (regrename_do_replace): If replacing the same
10604 reg multiple times, try to reuse last created gen_raw_REG.
10607 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10608 main to workaround a bug in GDB.
10610 2018-01-12 Tom de Vries <tom@codesourcery.com>
10613 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10615 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10617 PR rtl-optimization/80481
10618 * ira-color.c (get_cap_member): New function.
10619 (allocnos_conflict_by_live_ranges_p): Use it.
10620 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10621 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10623 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10626 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10627 (*saddl_se_1): Ditto.
10628 (*ssubsi_1): Ditto.
10629 (*ssubl_se_1): Ditto.
10631 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10633 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10634 rather than wi::to_widest for DR_INITs.
10635 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10636 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10637 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10639 (vect_analyze_group_access_1): Note that here.
10641 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10643 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10644 polynomial type sizes.
10646 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10648 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10649 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10650 (gimple_add_tmp_var): Likewise.
10652 2018-01-12 Martin Liska <mliska@suse.cz>
10654 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10655 (gimple_alloc_sizes): Likewise.
10656 (dump_gimple_statistics): Use PRIu64 in printf format.
10657 * gimple.h: Change uint64_t to int.
10659 2018-01-12 Martin Liska <mliska@suse.cz>
10661 * tree-core.h: Use uint64_t instead of int.
10662 * tree.c (tree_node_counts): Likewise.
10663 (tree_node_sizes): Likewise.
10664 (dump_tree_statistics): Use PRIu64 in printf format.
10666 2018-01-12 Martin Liska <mliska@suse.cz>
10668 * Makefile.in: As qsort_chk is implemented in vec.c, add
10669 vec.o to linkage of gencfn-macros.
10670 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10671 passing the info to record_node_allocation_statistics.
10672 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10674 * ggc-common.c (struct ggc_usage): Add operator== and use
10675 it in operator< and compare function.
10676 * mem-stats.h (struct mem_usage): Likewise.
10677 * vec.c (struct vec_usage): Remove operator< and compare
10678 function. Can be simply inherited.
10680 2018-01-12 Martin Jambor <mjambor@suse.cz>
10683 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10684 * tree-ssa-math-opts.c: Include domwalk.h.
10685 (convert_mult_to_fma_1): New function.
10686 (fma_transformation_info): New type.
10687 (fma_deferring_state): Likewise.
10688 (cancel_fma_deferring): New function.
10689 (result_of_phi): Likewise.
10690 (last_fma_candidate_feeds_initial_phi): Likewise.
10691 (convert_mult_to_fma): Added deferring logic, split actual
10692 transformation to convert_mult_to_fma_1.
10693 (math_opts_dom_walker): New type.
10694 (math_opts_dom_walker::after_dom_children): New method, body moved
10695 here from pass_optimize_widening_mul::execute, added deferring logic
10697 (pass_optimize_widening_mul::execute): Moved most of code to
10698 math_opts_dom_walker::after_dom_children.
10699 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10700 * config/i386/i386.c (ix86_option_override_internal): Added
10701 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10703 2018-01-12 Richard Biener <rguenther@suse.de>
10706 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10707 inline instance vars.
10709 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10712 * config/rx/rx.c (rx_is_restricted_memory_address):
10713 Handle SUBREG case.
10715 2018-01-12 Richard Biener <rguenther@suse.de>
10717 PR tree-optimization/80846
10718 * target.def (split_reduction): New target hook.
10719 * targhooks.c (default_split_reduction): New function.
10720 * targhooks.h (default_split_reduction): Declare.
10721 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10722 target requests first reduce vectors by combining low and high
10724 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10725 (get_vectype_for_scalar_type_and_size): Export.
10726 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10727 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10728 * doc/tm.texi: Regenerate.
10729 * config/i386/i386.c (ix86_split_reduction): Implement
10730 TARGET_VECTORIZE_SPLIT_REDUCTION.
10732 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10735 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10736 in PIC mode except for TARGET_VXWORKS_RTP.
10737 * config/sparc/sparc.c: Include cfgrtl.h.
10738 (TARGET_INIT_PIC_REG): Define.
10739 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10740 (sparc_pic_register_p): New predicate.
10741 (sparc_legitimate_address_p): Use it.
10742 (sparc_legitimize_pic_address): Likewise.
10743 (sparc_delegitimize_address): Likewise.
10744 (sparc_mode_dependent_address_p): Likewise.
10745 (gen_load_pcrel_sym): Remove 4th parameter.
10746 (load_got_register): Adjust call to above. Remove obsolete stuff.
10747 (sparc_expand_prologue): Do not call load_got_register here.
10748 (sparc_flat_expand_prologue): Likewise.
10749 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10750 (sparc_use_pseudo_pic_reg): New function.
10751 (sparc_init_pic_reg): Likewise.
10752 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10753 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10755 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10757 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10758 Add item for branch_cost.
10760 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10762 PR rtl-optimization/83565
10763 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10764 not extend the result to a larger mode for rotate operations.
10765 (num_sign_bit_copies1): Likewise.
10767 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10770 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10772 Use values-Xc.o for -pedantic.
10773 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10775 2018-01-12 Martin Liska <mliska@suse.cz>
10778 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10780 (possible_polymorphic_call_targets): Use it.
10781 (ipa_devirt): Likewise.
10783 2018-01-12 Martin Liska <mliska@suse.cz>
10785 * profile-count.h (enum profile_quality): Use 0 as invalid
10786 enum value of profile_quality.
10788 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10790 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10791 -mext-string options.
10793 2018-01-12 Richard Biener <rguenther@suse.de>
10795 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10796 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10797 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10799 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10801 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10803 * configure.ac (--with-long-double-format): Add support for the
10804 configuration option to change the default long double format on
10806 * config.gcc (powerpc*-linux*-*): Likewise.
10807 * configure: Regenerate.
10808 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10809 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10810 used without modification.
10812 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10814 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10815 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10816 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10817 MISC_BUILTIN_SPEC_BARRIER.
10818 (rs6000_init_builtins): Likewise.
10819 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10821 (speculation_barrier): New define_insn.
10822 * doc/extend.texi: Document __builtin_speculation_barrier.
10824 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10827 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10828 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10829 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10831 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10832 integral modes instead of "ss" and "sd".
10833 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10834 vectors with 32-bit and 64-bit elements.
10835 (vecdupssescalarmodesuffix): New mode attribute.
10836 (vec_dup<mode>): Use it.
10838 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10841 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10842 frame if argument is passed on stack.
10844 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10847 * ree.c (combine_reaching_defs): Optimize also
10848 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10849 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10851 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10853 PR middle-end/83189
10854 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10856 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10858 PR middle-end/83718
10859 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10860 after they are computed.
10862 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10864 PR tree-optimization/83695
10865 * gimple-loop-linterchange.cc
10866 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10867 reset cached scev information after interchange.
10868 (pass_linterchange::execute): Remove call to scev_reset_htab.
10870 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10872 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10873 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10874 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10875 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10876 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10877 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10878 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10879 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10880 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10881 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10882 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10883 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10884 (V_lane_reg): Likewise.
10885 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10887 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10888 (vfmal_lane_low<mode>_intrinsic,
10889 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10890 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10891 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10892 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10893 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10894 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10896 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10898 * config/arm/arm-cpus.in (fp16fml): New feature.
10899 (ALL_SIMD): Add fp16fml.
10900 (armv8.2-a): Add fp16fml as an option.
10901 (armv8.3-a): Likewise.
10902 (armv8.4-a): Add fp16fml as part of fp16.
10903 * config/arm/arm.h (TARGET_FP16FML): Define.
10904 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10906 * config/arm/arm-modes.def (V2HF): Define.
10907 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10908 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10909 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10910 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10911 vfmsl_low, vfmsl_high): New set of builtins.
10912 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10913 (vfml_op): New code attribute.
10914 (VFMLHALVES): New int iterator.
10915 (VFML, VFMLSEL): New mode attributes.
10916 (V_reg): Define mapping for V2HF.
10917 (V_hi, V_lo): New mode attributes.
10918 (VF_constraint): Likewise.
10919 (vfml_half, vfml_half_selector): New int attributes.
10920 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10922 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10923 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10925 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10926 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10927 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10928 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10930 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10931 Document new effective target and option set.
10933 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10935 * config/arm/arm-cpus.in (armv8_4): New feature.
10936 (ARMv8_4a): New fgroup.
10937 (armv8.4-a): New arch.
10938 * config/arm/arm-tables.opt: Regenerate.
10939 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10940 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10941 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10942 Add matching rules for -march=armv8.4-a and extensions.
10943 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10945 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10948 * config/rx/rx.md (BW): New mode attribute.
10949 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10951 2018-01-11 Richard Biener <rguenther@suse.de>
10953 PR tree-optimization/83435
10954 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10955 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10956 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10958 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10959 Alan Hayward <alan.hayward@arm.com>
10960 David Sherwood <david.sherwood@arm.com>
10962 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10964 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10965 (aarch64_print_address_internal): Use it to check for a zero offset.
10967 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10968 Alan Hayward <alan.hayward@arm.com>
10969 David Sherwood <david.sherwood@arm.com>
10971 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10972 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10973 Return a poly_int64 rather than a HOST_WIDE_INT.
10974 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10975 rather than a HOST_WIDE_INT.
10976 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10977 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10978 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10979 final_offset from HOST_WIDE_INT to poly_int64.
10980 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10981 to_constant when getting the number of units in an Advanced SIMD
10983 (aarch64_builtin_vectorized_function): Check for a constant number
10985 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10987 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10988 attribute instead of GET_MODE_NUNITS.
10989 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10990 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10991 GET_MODE_SIZE for fixed-size registers.
10992 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10993 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10994 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10995 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10996 (aarch64_print_operand, aarch64_print_address_internal)
10997 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10998 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10999 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11000 Handle polynomial GET_MODE_SIZE.
11001 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11002 wider than SImode without modification.
11003 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11004 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11005 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11006 passing and returning SVE modes.
11007 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11008 rather than GEN_INT.
11009 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11010 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11011 (aarch64_allocate_and_probe_stack_space): Likewise.
11012 (aarch64_layout_frame): Cope with polynomial offsets.
11013 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11014 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11015 polynomial offsets.
11016 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11017 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11018 poly_int64 rather than a HOST_WIDE_INT.
11019 (aarch64_get_separate_components, aarch64_process_components)
11020 (aarch64_expand_prologue, aarch64_expand_epilogue)
11021 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11022 (aarch64_anchor_offset): New function, split out from...
11023 (aarch64_legitimize_address): ...here.
11024 (aarch64_builtin_vectorization_cost): Handle polynomial
11025 TYPE_VECTOR_SUBPARTS.
11026 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11028 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11029 number of elements from the PARALLEL rather than the mode.
11030 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11031 rather than GET_MODE_BITSIZE.
11032 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11033 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11034 (aarch64_expand_vec_perm_const_1): Handle polynomial
11035 d->perm.length () and d->perm elements.
11036 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11037 Apply to_constant to d->perm elements.
11038 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11039 polynomial CONST_VECTOR_NUNITS.
11040 (aarch64_move_pointer): Take amount as a poly_int64 rather
11042 (aarch64_progress_pointer): Avoid temporary variable.
11043 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11044 the mode attribute instead of GET_MODE.
11046 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11047 Alan Hayward <alan.hayward@arm.com>
11048 David Sherwood <david.sherwood@arm.com>
11050 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11051 x exists before using it.
11052 (aarch64_add_constant_internal): Rename to...
11053 (aarch64_add_offset_1): ...this. Replace regnum with separate
11054 src and dest rtxes. Handle the case in which they're different,
11055 including when the offset is zero. Replace scratchreg with an rtx.
11056 Use 2 additions if there is no spare register into which we can
11057 move a 16-bit constant.
11058 (aarch64_add_constant): Delete.
11059 (aarch64_add_offset): Replace reg with separate src and dest
11060 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11061 Use aarch64_add_offset_1.
11062 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11063 an rtx rather than an int. Take the delta as a poly_int64
11064 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11065 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11066 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11067 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11068 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11069 and aarch64_add_sp.
11070 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11071 aarch64_add_constant.
11073 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11075 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11076 Use scalar_float_mode.
11078 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11080 * config/aarch64/aarch64-simd.md
11081 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11082 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11083 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11084 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11085 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11086 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11087 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11088 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11089 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11090 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11092 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11095 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11096 targ_options->x_arm_arch_string is non NULL.
11098 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11100 * config/aarch64/aarch64.h
11101 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11103 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11106 * expmed.c (emit_store_flag_force): Swap if const op0
11107 and change VOIDmode to mode of op0.
11109 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11111 PR rtl-optimization/83761
11112 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11113 than bytes to mode_for_size.
11115 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11117 PR middle-end/83189
11118 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11119 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11122 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11124 PR middle-end/83575
11125 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11126 when in layout mode.
11127 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11128 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11131 2018-01-10 Michael Collison <michael.collison@arm.com>
11133 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11134 * config/aarch64/aarch64-option-extension.def: Add
11135 AARCH64_OPT_EXTENSION of 'fp16fml'.
11136 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11137 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11138 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11139 * config/aarch64/constraints.md (Ui7): New constraint.
11140 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11141 (VFMLA_SEL_W): Ditto.
11144 (VFMLA16_LOW): New int iterator.
11145 (VFMLA16_HIGH): Ditto.
11146 (UNSPEC_FMLAL): New unspec.
11147 (UNSPEC_FMLSL): Ditto.
11148 (UNSPEC_FMLAL2): Ditto.
11149 (UNSPEC_FMLSL2): Ditto.
11150 (f16mac): New code attribute.
11151 * config/aarch64/aarch64-simd-builtins.def
11152 (aarch64_fmlal_lowv2sf): Ditto.
11153 (aarch64_fmlsl_lowv2sf): Ditto.
11154 (aarch64_fmlalq_lowv4sf): Ditto.
11155 (aarch64_fmlslq_lowv4sf): Ditto.
11156 (aarch64_fmlal_highv2sf): Ditto.
11157 (aarch64_fmlsl_highv2sf): Ditto.
11158 (aarch64_fmlalq_highv4sf): Ditto.
11159 (aarch64_fmlslq_highv4sf): Ditto.
11160 (aarch64_fmlal_lane_lowv2sf): Ditto.
11161 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11162 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11163 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11164 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11165 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11166 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11167 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11168 (aarch64_fmlal_lane_highv2sf): Ditto.
11169 (aarch64_fmlsl_lane_highv2sf): Ditto.
11170 (aarch64_fmlal_laneq_highv2sf): Ditto.
11171 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11172 (aarch64_fmlalq_lane_highv4sf): Ditto.
11173 (aarch64_fmlsl_lane_highv4sf): Ditto.
11174 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11175 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11176 * config/aarch64/aarch64-simd.md:
11177 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11178 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11179 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11180 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11181 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11182 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11183 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11184 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11185 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11186 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11187 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11188 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11189 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11190 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11191 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11192 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11193 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11194 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11195 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11196 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11197 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11198 (vfmlsl_low_u32): Ditto.
11199 (vfmlalq_low_u32): Ditto.
11200 (vfmlslq_low_u32): Ditto.
11201 (vfmlal_high_u32): Ditto.
11202 (vfmlsl_high_u32): Ditto.
11203 (vfmlalq_high_u32): Ditto.
11204 (vfmlslq_high_u32): Ditto.
11205 (vfmlal_lane_low_u32): Ditto.
11206 (vfmlsl_lane_low_u32): Ditto.
11207 (vfmlal_laneq_low_u32): Ditto.
11208 (vfmlsl_laneq_low_u32): Ditto.
11209 (vfmlalq_lane_low_u32): Ditto.
11210 (vfmlslq_lane_low_u32): Ditto.
11211 (vfmlalq_laneq_low_u32): Ditto.
11212 (vfmlslq_laneq_low_u32): Ditto.
11213 (vfmlal_lane_high_u32): Ditto.
11214 (vfmlsl_lane_high_u32): Ditto.
11215 (vfmlal_laneq_high_u32): Ditto.
11216 (vfmlsl_laneq_high_u32): Ditto.
11217 (vfmlalq_lane_high_u32): Ditto.
11218 (vfmlslq_lane_high_u32): Ditto.
11219 (vfmlalq_laneq_high_u32): Ditto.
11220 (vfmlslq_laneq_high_u32): Ditto.
11221 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11222 (AARCH64_FL_FOR_ARCH8_4): New.
11223 (AARCH64_ISA_F16FML): New ISA flag.
11224 (TARGET_F16FML): New feature flag for fp16fml.
11225 (doc/invoke.texi): Document new fp16fml option.
11227 2018-01-10 Michael Collison <michael.collison@arm.com>
11229 * config/aarch64/aarch64-builtins.c:
11230 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11231 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11232 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11233 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11234 (AARCH64_ISA_SHA3): New ISA flag.
11235 (TARGET_SHA3): New feature flag for sha3.
11236 * config/aarch64/iterators.md (sha512_op): New int attribute.
11237 (CRYPTO_SHA512): New int iterator.
11238 (UNSPEC_SHA512H): New unspec.
11239 (UNSPEC_SHA512H2): Ditto.
11240 (UNSPEC_SHA512SU0): Ditto.
11241 (UNSPEC_SHA512SU1): Ditto.
11242 * config/aarch64/aarch64-simd-builtins.def
11243 (aarch64_crypto_sha512hqv2di): New builtin.
11244 (aarch64_crypto_sha512h2qv2di): Ditto.
11245 (aarch64_crypto_sha512su0qv2di): Ditto.
11246 (aarch64_crypto_sha512su1qv2di): Ditto.
11247 (aarch64_eor3qv8hi): Ditto.
11248 (aarch64_rax1qv2di): Ditto.
11249 (aarch64_xarqv2di): Ditto.
11250 (aarch64_bcaxqv8hi): Ditto.
11251 * config/aarch64/aarch64-simd.md:
11252 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11253 (aarch64_crypto_sha512su0qv2di): Ditto.
11254 (aarch64_crypto_sha512su1qv2di): Ditto.
11255 (aarch64_eor3qv8hi): Ditto.
11256 (aarch64_rax1qv2di): Ditto.
11257 (aarch64_xarqv2di): Ditto.
11258 (aarch64_bcaxqv8hi): Ditto.
11259 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11260 (vsha512h2q_u64): Ditto.
11261 (vsha512su0q_u64): Ditto.
11262 (vsha512su1q_u64): Ditto.
11263 (veor3q_u16): Ditto.
11264 (vrax1q_u64): Ditto.
11265 (vxarq_u64): Ditto.
11266 (vbcaxq_u16): Ditto.
11267 * config/arm/types.md (crypto_sha512): New type attribute.
11268 (crypto_sha3): Ditto.
11269 (doc/invoke.texi): Document new sha3 option.
11271 2018-01-10 Michael Collison <michael.collison@arm.com>
11273 * config/aarch64/aarch64-builtins.c:
11274 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11275 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11276 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11277 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11278 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11279 (AARCH64_ISA_SM4): New ISA flag.
11280 (TARGET_SM4): New feature flag for sm4.
11281 * config/aarch64/aarch64-simd-builtins.def
11282 (aarch64_sm3ss1qv4si): Ditto.
11283 (aarch64_sm3tt1aq4si): Ditto.
11284 (aarch64_sm3tt1bq4si): Ditto.
11285 (aarch64_sm3tt2aq4si): Ditto.
11286 (aarch64_sm3tt2bq4si): Ditto.
11287 (aarch64_sm3partw1qv4si): Ditto.
11288 (aarch64_sm3partw2qv4si): Ditto.
11289 (aarch64_sm4eqv4si): Ditto.
11290 (aarch64_sm4ekeyqv4si): Ditto.
11291 * config/aarch64/aarch64-simd.md:
11292 (aarch64_sm3ss1qv4si): Ditto.
11293 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11294 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11295 (aarch64_sm4eqv4si): Ditto.
11296 (aarch64_sm4ekeyqv4si): Ditto.
11297 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11298 (sm3part_op): Ditto.
11299 (CRYPTO_SM3TT): Ditto.
11300 (CRYPTO_SM3PART): Ditto.
11301 (UNSPEC_SM3SS1): New unspec.
11302 (UNSPEC_SM3TT1A): Ditto.
11303 (UNSPEC_SM3TT1B): Ditto.
11304 (UNSPEC_SM3TT2A): Ditto.
11305 (UNSPEC_SM3TT2B): Ditto.
11306 (UNSPEC_SM3PARTW1): Ditto.
11307 (UNSPEC_SM3PARTW2): Ditto.
11308 (UNSPEC_SM4E): Ditto.
11309 (UNSPEC_SM4EKEY): Ditto.
11310 * config/aarch64/constraints.md (Ui2): New constraint.
11311 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11312 * config/arm/types.md (crypto_sm3): New type attribute.
11313 (crypto_sm4): Ditto.
11314 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11315 (vsm3tt1aq_u32): Ditto.
11316 (vsm3tt1bq_u32): Ditto.
11317 (vsm3tt2aq_u32): Ditto.
11318 (vsm3tt2bq_u32): Ditto.
11319 (vsm3partw1q_u32): Ditto.
11320 (vsm3partw2q_u32): Ditto.
11321 (vsm4eq_u32): Ditto.
11322 (vsm4ekeyq_u32): Ditto.
11323 (doc/invoke.texi): Document new sm4 option.
11325 2018-01-10 Michael Collison <michael.collison@arm.com>
11327 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11328 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11329 (AARCH64_FL_FOR_ARCH8_4): New.
11330 (AARCH64_FL_V8_4): New flag.
11331 (doc/invoke.texi): Document new armv8.4-a option.
11333 2018-01-10 Michael Collison <michael.collison@arm.com>
11335 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11336 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11337 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11338 * config/aarch64/aarch64-option-extension.def: Add
11339 AARCH64_OPT_EXTENSION of 'sha2'.
11340 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11341 (crypto): Disable sha2 and aes if crypto disabled.
11342 (crypto): Enable aes and sha2 if enabled.
11343 (simd): Disable sha2 and aes if simd disabled.
11344 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11346 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11347 (TARGET_SHA2): New feature flag for sha2.
11348 (TARGET_AES): New feature flag for aes.
11349 * config/aarch64/aarch64-simd.md:
11350 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11351 conditional on TARGET_AES.
11352 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11353 (aarch64_crypto_sha1hsi): Make pattern conditional
11355 (aarch64_crypto_sha1hv4si): Ditto.
11356 (aarch64_be_crypto_sha1hv4si): Ditto.
11357 (aarch64_crypto_sha1su1v4si): Ditto.
11358 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11359 (aarch64_crypto_sha1su0v4si): Ditto.
11360 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11361 (aarch64_crypto_sha256su0v4si): Ditto.
11362 (aarch64_crypto_sha256su1v4si): Ditto.
11363 (doc/invoke.texi): Document new aes and sha2 options.
11365 2018-01-10 Martin Sebor <msebor@redhat.com>
11367 PR tree-optimization/83781
11368 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11371 2018-01-11 Martin Sebor <msebor@gmail.com>
11372 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11374 PR tree-optimization/83501
11375 PR tree-optimization/81703
11377 * tree-ssa-strlen.c (get_string_cst): Rename...
11378 (get_string_len): ...to this. Handle global constants.
11379 (handle_char_store): Adjust.
11381 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11382 Jim Wilson <jimw@sifive.com>
11384 * config/riscv/riscv-protos.h (riscv_output_return): New.
11385 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11386 (riscv_attribute_table, riscv_output_return),
11387 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11388 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11389 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11391 (riscv_expand_prologue): Add early return for naked function.
11392 (riscv_expand_epilogue): Likewise.
11393 (riscv_function_ok_for_sibcall): Return false for naked function.
11394 (riscv_set_current_function): New.
11395 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11396 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11397 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11398 * doc/extend.texi (RISC-V Function Attributes): New.
11400 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11402 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11403 check for 128-bit long double before checking TCmode.
11404 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11405 128-bit long doubles before checking TFmode or TCmode.
11406 (FLOAT128_IBM_P): Likewise.
11408 2018-01-10 Martin Sebor <msebor@redhat.com>
11410 PR tree-optimization/83671
11411 * builtins.c (c_strlen): Unconditionally return zero for the empty
11413 Use -Warray-bounds for warnings.
11414 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11415 for non-constant array indices with COMPONENT_REF, arrays of
11416 arrays, and pointers to arrays.
11417 (gimple_fold_builtin_strlen): Determine and set length range for
11418 non-constant character arrays.
11420 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11422 PR middle-end/81897
11423 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11426 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11428 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11430 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11433 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11434 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11435 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11436 indexed_or_indirect_operand predicate.
11437 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11438 (*vsx_le_perm_load_v8hi): Likewise.
11439 (*vsx_le_perm_load_v16qi): Likewise.
11440 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11441 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11442 (*vsx_le_perm_store_v8hi): Likewise.
11443 (*vsx_le_perm_store_v16qi): Likewise.
11444 (eight unnamed splitters): Likewise.
11446 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11448 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11449 * config/rs6000/emmintrin.h: Likewise.
11450 * config/rs6000/mmintrin.h: Likewise.
11451 * config/rs6000/xmmintrin.h: Likewise.
11453 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11456 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11458 * tree.c (tree_nop_conversion): Return true for location wrapper
11460 (maybe_wrap_with_location): New function.
11461 (selftest::check_strip_nops): New function.
11462 (selftest::test_location_wrappers): New function.
11463 (selftest::tree_c_tests): Call it.
11464 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11465 (maybe_wrap_with_location): New decl.
11466 (EXPR_LOCATION_WRAPPER_P): New macro.
11467 (location_wrapper_p): New inline function.
11468 (tree_strip_any_location_wrapper): New inline function.
11470 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11473 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11474 stack_realign_offset for the largest alignment of stack slot
11476 (ix86_find_max_used_stack_alignment): New function.
11477 (ix86_finalize_stack_frame_flags): Use it. Set
11478 max_used_stack_alignment if we don't realign stack.
11479 * config/i386/i386.h (machine_function): Add
11480 max_used_stack_alignment.
11482 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11484 * config/arm/arm.opt (-mbranch-cost): New option.
11485 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11488 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11491 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11492 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11494 2018-01-10 Richard Biener <rguenther@suse.de>
11497 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11498 early out so it also covers the case where we have a non-NULL
11501 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11503 PR tree-optimization/83753
11504 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11505 for non-strided grouped accesses if the number of elements is 1.
11507 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11510 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11511 * i386.h (TARGET_USE_GATHER): Define.
11512 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11514 2018-01-10 Martin Liska <mliska@suse.cz>
11517 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11518 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11520 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11521 CLEANUP_NO_PARTITIONING is not set.
11523 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11525 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11526 for vectors, as a partial revert of r254296.
11527 * rtl.h (const_vec_p): Delete.
11528 (const_vec_duplicate_p): Don't test for vector CONSTs.
11529 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11530 * expmed.c (make_tree): Likewise.
11533 * common.md (E, F): Use CONSTANT_P instead of checking for
11535 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11536 checking for CONST_VECTOR.
11538 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11540 PR middle-end/83575
11541 * predict.c (force_edge_cold): Handle in more sane way edges
11542 with no prediction.
11544 2018-01-09 Carl Love <cel@us.ibm.com>
11546 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11548 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11549 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11550 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11551 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11552 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11553 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11554 * config/rs6000/rs6000-protos.h: Add extern defition for
11555 rs6000_generate_float2_double_code.
11556 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11558 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11559 (float2_v2df): Add define_expand.
11561 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11564 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11565 op_mode in the force_to_mode call.
11567 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11569 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11570 instead of checking each element individually.
11571 (aarch64_evpc_uzp): Likewise.
11572 (aarch64_evpc_zip): Likewise.
11573 (aarch64_evpc_ext): Likewise.
11574 (aarch64_evpc_rev): Likewise.
11575 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11576 instead of checking each element individually. Return true without
11578 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11579 whether all selected elements come from the same input, instead of
11580 checking each element individually. Remove calls to gen_rtx_REG,
11581 start_sequence and end_sequence and instead assert that no rtl is
11584 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11586 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11587 order of HIGH and CONST checks.
11589 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11591 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11592 if the destination isn't an SSA_NAME.
11594 2018-01-09 Richard Biener <rguenther@suse.de>
11596 PR tree-optimization/83668
11597 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11599 (canonicalize_loop_form): ... here, renamed from ...
11600 (canonicalize_loop_closed_ssa_form): ... this and amended to
11601 swap successor edges for loop exit blocks to make us use
11602 the RPO order we need for initial schedule generation.
11604 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11606 PR tree-optimization/64811
11607 * match.pd: When optimizing comparisons with Inf, avoid
11608 introducing or losing exceptions from comparisons with NaN.
11610 2018-01-09 Martin Liska <mliska@suse.cz>
11613 * asan.c (shadow_mem_size): Add gcc_assert.
11615 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11617 Don't save registers in main().
11620 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11621 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11622 * config/avr/avr.c (avr_set_current_function): Don't error if
11623 naked, OS_task or OS_main are specified at the same time.
11624 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11626 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11628 * common/config/avr/avr-common.c (avr_option_optimization_table):
11629 Switch on -mmain-is-OS_task for optimizing compilations.
11631 2018-01-09 Richard Biener <rguenther@suse.de>
11633 PR tree-optimization/83572
11634 * graphite.c: Include cfganal.h.
11635 (graphite_transform_loops): Connect infinite loops to exit
11636 and remove fake edges at the end.
11638 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11640 * ipa-inline.c (edge_badness): Revert accidental checkin.
11642 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11645 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11646 symbols; not inline clones.
11648 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11651 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11652 hard registers. Formatting fixes.
11654 PR preprocessor/83722
11655 * gcc.c (try_generate_repro): Pass
11656 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11657 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11660 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11661 Kito Cheng <kito.cheng@gmail.com>
11663 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11664 (riscv_leaf_function_p): Delete.
11665 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11667 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11669 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11671 (do_ifelse): New function.
11672 (do_isel): New function.
11673 (do_sub3): New function.
11674 (do_add3): New function.
11675 (do_load_mask_compare): New function.
11676 (do_overlap_load_compare): New function.
11677 (expand_compare_loop): New function.
11678 (expand_block_compare): Call expand_compare_loop() when appropriate.
11679 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11680 option description.
11681 (-mblock-compare-inline-loop-limit): New option.
11683 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11686 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11687 Reverse order of second and third operands in first alternative.
11688 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11689 of first and second elements in UNSPEC_VPERMR vector.
11690 (altivec_expand_vec_perm_le): Likewise.
11692 2018-01-08 Jeff Law <law@redhat.com>
11694 PR rtl-optimizatin/81308
11695 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11696 (process_switch): If group_case_labels makes a change, then set
11698 (pass_convert_switch::execute): If a switch is converted, then
11699 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11701 PR rtl-optimization/81308
11702 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11705 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11707 PR target/83663 - Revert r255946
11708 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11709 generation for cases where splatting a value is not useful.
11710 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11711 across a vec_duplicate and a paradoxical subreg forming a vector
11712 mode to a vec_concat.
11714 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11716 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11717 -march=armv8.3-a variants.
11718 * config/arm/t-multilib: Likewise.
11719 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11721 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11723 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11725 (cceq_ior_compare_complement): Give it a name so I can use it, and
11726 change boolean_or_operator predicate to boolean_operator so it can
11727 be used to generate a crand.
11728 (eqne): New code iterator.
11729 (bd/bd_neg): New code_attrs.
11730 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11731 a single define_insn.
11732 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11733 decrement (bdnzt/bdnzf/bdzt/bdzf).
11734 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11735 with the new names of the branch decrement patterns, and added the
11736 names of the branch decrement conditional patterns.
11738 2018-01-08 Richard Biener <rguenther@suse.de>
11740 PR tree-optimization/83563
11741 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11744 2018-01-08 Richard Biener <rguenther@suse.de>
11746 PR middle-end/83713
11747 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11749 2018-01-08 Richard Biener <rguenther@suse.de>
11751 PR tree-optimization/83685
11752 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11753 references to abnormals.
11755 2018-01-08 Richard Biener <rguenther@suse.de>
11758 * dwarf2out.c (output_indirect_strings): Handle empty
11759 skeleton_debug_str_hash.
11760 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11762 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11764 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11765 (emit_store_direct): Likewise.
11766 (arc_trampoline_adjust_address): Likewise.
11767 (arc_asm_trampoline_template): New function.
11768 (arc_initialize_trampoline): Use asm_trampoline_template.
11769 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11770 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11771 * config/arc/arc.md (flush_icache): Delete pattern.
11773 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11775 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11776 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11779 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11782 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11783 by not USED_FOR_TARGET.
11784 (make_pass_resolve_sw_modes): Likewise.
11786 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11788 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11791 2018-01-08 Richard Biener <rguenther@suse.de>
11793 PR middle-end/83580
11794 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11796 2018-01-08 Richard Biener <rguenther@suse.de>
11798 PR middle-end/83517
11799 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11801 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11803 PR middle-end/81897
11804 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11805 basic blocks with a small number of successors.
11806 (convert_control_dep_chain_into_preds): Improve handling of
11808 (dump_predicates): Split apart into...
11809 (dump_pred_chain): ...here...
11810 (dump_pred_info): ...and here.
11811 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11812 (can_chain_union_be_invalidated_p): Improve check for invalidation
11814 (uninit_uses_cannot_happen): Avoid unnecessary if
11815 convert_control_dep_chain_into_preds yielded nothing.
11817 2018-01-06 Martin Sebor <msebor@redhat.com>
11819 PR tree-optimization/83640
11820 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11821 subtracting negative offset from size.
11822 (builtin_access::overlap): Adjust offset bounds of the access to fall
11823 within the size of the object if possible.
11825 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11827 PR rtl-optimization/83699
11828 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11829 extract_bit_field_as_subreg to cases in which the extracted
11830 value is also a vector.
11832 * lra-constraints.c (process_alt_operands): Test for the equivalence
11833 substitutions when detecting a possible reload cycle.
11835 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11838 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11839 by default if flag_selective_schedling{,2}. Formatting fixes.
11841 PR rtl-optimization/83682
11842 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11843 if it has non-VECTOR_MODE element mode.
11844 (vec_duplicate_p): Likewise.
11846 PR middle-end/83694
11847 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11848 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11850 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11853 * config/i386/i386-builtin.def
11854 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11855 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11856 Require also OPTION_MASK_ISA_AVX512F in addition to
11857 OPTION_MASK_ISA_GFNI.
11858 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11859 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11860 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11861 to OPTION_MASK_ISA_GFNI.
11862 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11863 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11864 OPTION_MASK_ISA_AVX512BW.
11865 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11866 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11867 addition to OPTION_MASK_ISA_GFNI.
11868 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11869 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11870 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11871 to OPTION_MASK_ISA_GFNI.
11872 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11873 a requirement for all ISAs rather than any of them with a few
11875 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11877 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11878 bitmasks to be enabled with 3 exceptions, instead of requiring any
11879 enabled ISA with lots of exceptions.
11880 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11881 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11882 Change avx512bw in isa attribute to avx512f.
11883 * config/i386/sgxintrin.h: Add license boilerplate.
11884 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11885 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11886 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11887 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11889 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11890 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11891 temporarily sse2 rather than sse if not enabled already.
11894 * config/i386/sse.md (VI248_VLBW): Rename to ...
11895 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11896 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11897 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11898 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11899 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11900 mode iterator instead of VI248_VLBW.
11902 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11904 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11905 (record_modified): Skip clobbers; add debug output.
11906 (param_change_prob): Use sreal frequencies.
11908 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11910 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11911 punt for user-aligned variables.
11913 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11915 * tree-chrec.c (chrec_contains_symbols): Return true for
11918 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11921 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11922 of (x|y) == x for BICS pattern.
11924 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11926 PR tree-optimization/83605
11927 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11928 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11931 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11933 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11934 * config/epiphany/rtems.h: New file.
11936 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11937 Uros Bizjak <ubizjak@gmail.com>
11940 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11941 QIreg_operand instead of register_operand predicate.
11942 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11943 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11944 comments instead of -fmitigate[-_]rop.
11946 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11949 * cgraphunit.c (symbol_table::compile): Switch to text_section
11950 before calling assembly_start debug hook.
11951 * run-rtl-passes.c (run_rtl_passes): Likewise.
11954 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11956 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11957 range_int_cst_p rather than !symbolic_range_p before calling
11958 extract_range_from_multiplicative_op_1.
11960 2018-01-04 Jeff Law <law@redhat.com>
11962 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11963 redundant test in assertion.
11965 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11967 * doc/rtl.texi: Document machine_mode wrapper classes.
11969 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11971 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11972 using tree_to_uhwi.
11974 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11976 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11977 the VEC_PERM_EXPR fold to fail.
11979 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11982 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11983 to switched_sections.
11985 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11988 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11989 test for d.testing.
11991 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11994 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11995 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11997 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12000 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12001 is BLKmode and bitpos not zero or mode change is needed.
12003 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12006 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12009 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12012 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12013 instead of MULT rtx. Update all corresponding splitters.
12014 (*saddl_se): Ditto.
12015 (*ssub<modesuffix>): Ditto.
12016 (*ssubl_se): Ditto.
12017 (*cmp_sadd_di): Update split patterns.
12018 (*cmp_sadd_si): Ditto.
12019 (*cmp_sadd_sidi): Ditto.
12020 (*cmp_ssub_di): Ditto.
12021 (*cmp_ssub_si): Ditto.
12022 (*cmp_ssub_sidi): Ditto.
12023 * config/alpha/predicates.md (const23_operand): New predicate.
12024 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12025 Look for ASHIFT, not MULT inner operand.
12026 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12028 2018-01-04 Martin Liska <mliska@suse.cz>
12030 PR gcov-profile/83669
12031 * gcov.c (output_intermediate_file): Add version to intermediate
12033 * doc/gcov.texi: Document new field 'version' in intermediate
12034 file format. Fix location of '-k' option of gcov command.
12036 2018-01-04 Martin Liska <mliska@suse.cz>
12039 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12041 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12043 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12045 2018-01-03 Martin Sebor <msebor@redhat.com>
12047 PR tree-optimization/83655
12048 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12049 checking calls with invalid arguments.
12051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12053 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12054 (vectorizable_mask_load_store): Delete.
12055 (vectorizable_call): Return false for masked loads and stores.
12056 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12057 instead of gimple_assign_rhs1.
12058 (vectorizable_load): Handle IFN_MASK_LOAD.
12059 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12063 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12065 (vectorizable_mask_load_store): ...here.
12066 (vectorizable_load): ...and here.
12068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12070 * tree-vect-stmts.c (vect_build_all_ones_mask)
12071 (vect_build_zero_merge_argument): New functions, split out from...
12072 (vectorizable_load): ...here.
12074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12076 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12078 (vectorizable_mask_load_store): ...here.
12079 (vectorizable_store): ...and here.
12081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12083 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12085 (vectorizable_mask_load_store): ...here.
12087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12089 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12090 (vect_model_store_cost): Take a vec_load_store_type instead of a
12092 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12093 (vect_model_store_cost): Take a vec_load_store_type instead of a
12095 (vectorizable_mask_load_store): Update accordingly.
12096 (vectorizable_store): Likewise.
12097 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12101 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12102 IFN_MASK_LOAD calls here rather than...
12103 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12106 Alan Hayward <alan.hayward@arm.com>
12107 David Sherwood <david.sherwood@arm.com>
12109 * expmed.c (extract_bit_field_1): For vector extracts,
12110 fall back to extract_bit_field_as_subreg if vec_extract
12113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12114 Alan Hayward <alan.hayward@arm.com>
12115 David Sherwood <david.sherwood@arm.com>
12117 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12118 they are variable or constant sized.
12119 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12120 slots for constant-sized data.
12122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12123 Alan Hayward <alan.hayward@arm.com>
12124 David Sherwood <david.sherwood@arm.com>
12126 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12127 handling COND_EXPRs with boolean comparisons, try to find a better
12128 basis for the mask type than the boolean itself.
12130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12132 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12133 is calculated and how it can be overridden.
12134 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12135 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12137 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12141 Alan Hayward <alan.hayward@arm.com>
12142 David Sherwood <david.sherwood@arm.com>
12144 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12145 Remove the mode argument.
12146 (aarch64_simd_valid_immediate): Remove the mode and inverse
12148 * config/aarch64/iterators.md (bitsize): New iterator.
12149 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12150 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12151 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12152 aarch64_simd_valid_immediate.
12153 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12154 (aarch64_reg_or_bic_imm): Likewise.
12155 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12156 with an insn_type enum and msl with a modifier_type enum.
12157 Replace element_width with a scalar_mode. Change the shift
12158 to unsigned int. Add constructors for scalar_float_mode and
12159 scalar_int_mode elements.
12160 (aarch64_vect_float_const_representable_p): Delete.
12161 (aarch64_can_const_movi_rtx_p)
12162 (aarch64_simd_scalar_immediate_valid_for_move)
12163 (aarch64_simd_make_constant): Update call to
12164 aarch64_simd_valid_immediate.
12165 (aarch64_advsimd_valid_immediate_hs): New function.
12166 (aarch64_advsimd_valid_immediate): Likewise.
12167 (aarch64_simd_valid_immediate): Remove mode and inverse
12168 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12169 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12170 and aarch64_float_const_representable_p on the result.
12171 (aarch64_output_simd_mov_immediate): Remove mode argument.
12172 Update call to aarch64_simd_valid_immediate and use of
12173 simd_immediate_info.
12174 (aarch64_output_scalar_simd_mov_immediate): Update call
12177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12178 Alan Hayward <alan.hayward@arm.com>
12179 David Sherwood <david.sherwood@arm.com>
12181 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12182 (mode_nunits): Likewise CONST_MODE_NUNITS.
12183 * machmode.def (ADJUST_NUNITS): Document.
12184 * genmodes.c (mode_data::need_nunits_adj): New field.
12185 (blank_mode): Update accordingly.
12186 (adj_nunits): New variable.
12187 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12189 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12190 listed in adj_nunits.
12191 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12192 listed in adj_nunits. Don't emit case statements for such modes.
12193 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12194 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12195 nothing if adj_nunits is nonnull.
12196 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12197 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12198 (emit_mode_fbit): Update use of print_maybe_const_decl.
12199 (emit_move_size): Likewise. Treat the array as non-const
12201 (emit_mode_adjustments): Handle adj_nunits.
12203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12205 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12206 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12207 (VECTOR_MODES): Use it.
12208 (make_vector_modes): Take the prefix as an argument.
12210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12211 Alan Hayward <alan.hayward@arm.com>
12212 David Sherwood <david.sherwood@arm.com>
12214 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12215 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12216 for MODE_VECTOR_BOOL.
12217 * machmode.def (VECTOR_BOOL_MODE): Document.
12218 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12219 (make_vector_bool_mode): New function.
12220 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12222 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12223 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12225 * stor-layout.c (int_mode_for_mode): Likewise.
12226 * tree.c (build_vector_type_for_mode): Likewise.
12227 * varasm.c (output_constant_pool_2): Likewise.
12228 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12229 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12230 for MODE_VECTOR_BOOL.
12231 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12232 of mode class checks.
12233 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12234 instead of a list of mode class checks.
12235 (expand_vector_scalar_condition): Likewise.
12236 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12239 Alan Hayward <alan.hayward@arm.com>
12240 David Sherwood <david.sherwood@arm.com>
12242 * machmode.h (mode_size): Change from unsigned short to
12244 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12245 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12246 or if measurement_type is not polynomial.
12247 (fixed_size_mode::includes_p): Check for constant-sized modes.
12248 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12249 return a poly_uint16 rather than an unsigned short.
12250 (emit_mode_size): Change the type of mode_size from unsigned short
12251 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12252 (emit_mode_adjustments): Cope with polynomial vector sizes.
12253 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12255 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12257 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12258 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12259 * caller-save.c (setup_save_areas): Likewise.
12260 (replace_reg_with_saved_mem): Likewise.
12261 * calls.c (emit_library_call_value_1): Likewise.
12262 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12263 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12264 (gen_lowpart_for_combine): Likewise.
12265 * convert.c (convert_to_integer_1): Likewise.
12266 * cse.c (equiv_constant, cse_insn): Likewise.
12267 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12268 (cselib_subst_to_values): Likewise.
12269 * dce.c (word_dce_process_block): Likewise.
12270 * df-problems.c (df_word_lr_mark_ref): Likewise.
12271 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12272 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12273 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12274 (rtl_for_decl_location): Likewise.
12275 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12276 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12277 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12278 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12279 (expand_expr_real_1): Likewise.
12280 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12281 (pad_below): Likewise.
12282 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12283 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12284 * ira.c (get_subreg_tracking_sizes): Likewise.
12285 * ira-build.c (ira_create_allocno_objects): Likewise.
12286 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12287 (ira_sort_regnos_for_alter_reg): Likewise.
12288 * ira-costs.c (record_operand_costs): Likewise.
12289 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12290 (resolve_simple_move): Likewise.
12291 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12292 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12293 (lra_constraints): Likewise.
12294 (CONST_POOL_OK_P): Reject variable-sized modes.
12295 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12296 (add_pseudo_to_slot, lra_spill): Likewise.
12297 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12298 * optabs-query.c (get_best_extraction_insn): Likewise.
12299 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12300 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12301 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12302 * recog.c (offsettable_address_addr_space_p): Likewise.
12303 * regcprop.c (maybe_mode_change): Likewise.
12304 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12305 * regrename.c (build_def_use): Likewise.
12306 * regstat.c (dump_reg_info): Likewise.
12307 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12308 (find_reloads, find_reloads_subreg_address): Likewise.
12309 * reload1.c (eliminate_regs_1): Likewise.
12310 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12311 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12312 (simplify_binary_operation_1, simplify_subreg): Likewise.
12313 * targhooks.c (default_function_arg_padding): Likewise.
12314 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12315 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12316 (verify_gimple_assign_ternary): Likewise.
12317 * tree-inline.c (estimate_move_cost): Likewise.
12318 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12319 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12320 (get_address_cost_ainc): Likewise.
12321 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12322 (vect_supportable_dr_alignment): Likewise.
12323 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12324 (vectorizable_reduction): Likewise.
12325 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12326 (vectorizable_operation, vectorizable_load): Likewise.
12327 * tree.c (build_same_sized_truth_vector_type): Likewise.
12328 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12329 * var-tracking.c (emit_note_insn_var_location): Likewise.
12330 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12331 (ADDR_VEC_ALIGN): Likewise.
12333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12334 Alan Hayward <alan.hayward@arm.com>
12335 David Sherwood <david.sherwood@arm.com>
12337 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12339 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12340 or if measurement_type is polynomial.
12341 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12342 * combine.c (make_extraction): Likewise.
12343 * dse.c (find_shift_sequence): Likewise.
12344 * dwarf2out.c (mem_loc_descriptor): Likewise.
12345 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12346 (extract_bit_field, extract_low_bits): Likewise.
12347 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12348 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12349 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12350 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12351 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12352 * reload.c (find_reloads): Likewise.
12353 * reload1.c (alter_reg): Likewise.
12354 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12355 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12356 * tree-if-conv.c (predicate_mem_writes): Likewise.
12357 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12358 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12359 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12360 * valtrack.c (dead_debug_insert_temp): Likewise.
12361 * varasm.c (mergeable_constant_section): Likewise.
12362 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12365 Alan Hayward <alan.hayward@arm.com>
12366 David Sherwood <david.sherwood@arm.com>
12368 * expr.c (expand_assignment): Cope with polynomial mode sizes
12369 when assigning to a CONCAT.
12371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12372 Alan Hayward <alan.hayward@arm.com>
12373 David Sherwood <david.sherwood@arm.com>
12375 * machmode.h (mode_precision): Change from unsigned short to
12377 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12379 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12380 or if measurement_type is not polynomial.
12381 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12382 in which the mode is already known to be a scalar_int_mode.
12383 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12384 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12386 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12387 for GET_MODE_PRECISION.
12388 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12389 for GET_MODE_PRECISION.
12390 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12392 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12393 (expand_field_assignment, make_extraction): Likewise.
12394 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12395 (get_last_value): Likewise.
12396 * convert.c (convert_to_integer_1): Likewise.
12397 * cse.c (cse_insn): Likewise.
12398 * expr.c (expand_expr_real_1): Likewise.
12399 * lra-constraints.c (simplify_operand_subreg): Likewise.
12400 * optabs-query.c (can_atomic_load_p): Likewise.
12401 * optabs.c (expand_atomic_load): Likewise.
12402 (expand_atomic_store): Likewise.
12403 * ree.c (combine_reaching_defs): Likewise.
12404 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12405 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12406 * tree.h (type_has_mode_precision_p): Likewise.
12407 * ubsan.c (instrument_si_overflow): Likewise.
12409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12410 Alan Hayward <alan.hayward@arm.com>
12411 David Sherwood <david.sherwood@arm.com>
12413 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12414 polynomial numbers of units.
12415 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12416 (valid_vector_subparts_p): New function.
12417 (build_vector_type): Remove temporary shim and take the number
12418 of units as a poly_uint64 rather than an int.
12419 (build_opaque_vector_type): Take the number of units as a
12420 poly_uint64 rather than an int.
12421 * tree.c (build_vector_from_ctor): Handle polynomial
12422 TYPE_VECTOR_SUBPARTS.
12423 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12424 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12425 (build_vector_from_val): If the number of units is variable,
12426 use build_vec_duplicate_cst for constant operands and
12427 VEC_DUPLICATE_EXPR otherwise.
12428 (make_vector_type): Remove temporary is_constant ().
12429 (build_vector_type, build_opaque_vector_type): Take the number of
12430 units as a poly_uint64 rather than an int.
12431 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12433 * cfgexpand.c (expand_debug_expr): Likewise.
12434 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12435 (store_constructor, expand_expr_real_1): Likewise.
12436 (const_scalar_mask_from_tree): Likewise.
12437 * fold-const-call.c (fold_const_reduction): Likewise.
12438 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12439 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12440 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12441 (fold_relational_const): Likewise.
12442 (native_interpret_vector): Likewise. Change the size from an
12443 int to an unsigned int.
12444 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12445 TYPE_VECTOR_SUBPARTS.
12446 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12447 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12448 duplicating a non-constant operand into a variable-length vector.
12449 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12450 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12451 * ipa-icf.c (sem_variable::equals): Likewise.
12452 * match.pd: Likewise.
12453 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12454 * print-tree.c (print_node): Likewise.
12455 * stor-layout.c (layout_type): Likewise.
12456 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12457 * tree-cfg.c (verify_gimple_comparison): Likewise.
12458 (verify_gimple_assign_binary): Likewise.
12459 (verify_gimple_assign_ternary): Likewise.
12460 (verify_gimple_assign_single): Likewise.
12461 * tree-pretty-print.c (dump_generic_node): Likewise.
12462 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12463 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12464 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12465 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12466 (vect_shift_permute_load_chain): Likewise.
12467 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12468 (expand_vector_condition, optimize_vector_constructor): Likewise.
12469 (lower_vec_perm, get_compute_type): Likewise.
12470 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12471 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12472 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12473 (vect_recog_mask_conversion_pattern): Likewise.
12474 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12475 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12476 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12477 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12478 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12479 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12480 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12481 (supportable_widening_operation): Likewise.
12482 (supportable_narrowing_operation): Likewise.
12483 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12485 * varasm.c (output_constant): Likewise.
12487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12488 Alan Hayward <alan.hayward@arm.com>
12489 David Sherwood <david.sherwood@arm.com>
12491 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12492 so that both the length == 3 and length != 3 cases set up their
12493 own permute vectors. Add comments explaining why we know the
12494 number of elements is constant.
12495 (vect_permute_load_chain): Likewise.
12497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12498 Alan Hayward <alan.hayward@arm.com>
12499 David Sherwood <david.sherwood@arm.com>
12501 * machmode.h (mode_nunits): Change from unsigned char to
12503 (ONLY_FIXED_SIZE_MODES): New macro.
12504 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12505 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12506 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12508 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12509 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12510 or if measurement_type is not polynomial.
12511 * genmodes.c (ZERO_COEFFS): New macro.
12512 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12514 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12515 Use ZERO_COEFFS when emitting initializers.
12516 * data-streamer.h (bp_pack_poly_value): New function.
12517 (bp_unpack_poly_value): Likewise.
12518 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12519 for GET_MODE_NUNITS.
12520 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12521 for GET_MODE_NUNITS.
12522 * tree.c (make_vector_type): Remove temporary shim and make
12523 the real function take the number of units as a poly_uint64
12524 rather than an int.
12525 (build_vector_type_for_mode): Handle polynomial nunits.
12526 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12527 * emit-rtl.c (const_vec_series_p_1): Likewise.
12528 (gen_rtx_CONST_VECTOR): Likewise.
12529 * fold-const.c (test_vec_duplicate_folding): Likewise.
12530 * genrecog.c (validate_pattern): Likewise.
12531 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12532 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12533 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12534 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12535 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12536 * rtlanal.c (subreg_get_info): Likewise.
12537 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12538 (vect_grouped_load_supported): Likewise.
12539 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12540 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12541 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12542 (simplify_const_unary_operation, simplify_binary_operation_1)
12543 (simplify_const_binary_operation, simplify_ternary_operation)
12544 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12545 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12546 instead of CONST_VECTOR_NUNITS.
12547 * varasm.c (output_constant_pool_2): Likewise.
12548 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12549 explicit-encoded elements in the XVEC for variable-length vectors.
12551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12553 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12556 Alan Hayward <alan.hayward@arm.com>
12557 David Sherwood <david.sherwood@arm.com>
12559 * coretypes.h (fixed_size_mode): Declare.
12560 (fixed_size_mode_pod): New typedef.
12561 * builtins.h (target_builtins::x_apply_args_mode)
12562 (target_builtins::x_apply_result_mode): Change type to
12563 fixed_size_mode_pod.
12564 * builtins.c (apply_args_size, apply_result_size, result_vector)
12565 (expand_builtin_apply_args_1, expand_builtin_apply)
12566 (expand_builtin_return): Update accordingly.
12568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12570 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12571 * cselib.c (cselib_hash_rtx): Likewise.
12572 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12573 CONST_VECTOR encoding.
12575 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12576 Jeff Law <law@redhat.com>
12579 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12580 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12581 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12582 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12585 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12586 explicitly probe *sp in a noreturn function if there were any callee
12587 register saves or frame pointer is needed.
12589 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12592 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12593 BLKmode for ternary, binary or unary expressions.
12596 * var-tracking.c (delete_vta_debug_insn): New inline function.
12597 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12598 insns from get_insns () to NULL instead of each bb separately.
12599 Use delete_vta_debug_insn. No longer static.
12600 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12601 delete_vta_debug_insns callers.
12602 * rtl.h (delete_vta_debug_insns): Declare.
12603 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12604 instead of variable_tracking_main.
12606 2018-01-03 Martin Sebor <msebor@redhat.com>
12608 PR tree-optimization/83603
12609 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12610 arguments past the endof the argument list in functions declared
12611 without a prototype.
12612 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12613 Avoid checking when arguments are null.
12615 2018-01-03 Martin Sebor <msebor@redhat.com>
12618 * doc/extend.texi (attribute const): Fix a typo.
12619 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12620 issuing -Wsuggest-attribute for void functions.
12622 2018-01-03 Martin Sebor <msebor@redhat.com>
12624 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12625 offset_int::from instead of wide_int::to_shwi.
12626 (maybe_diag_overlap): Remove assertion.
12627 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12628 * gimple-ssa-sprintf.c (format_directive): Same.
12629 (parse_directive): Same.
12630 (sprintf_dom_walker::compute_format_length): Same.
12631 (try_substitute_return_value): Same.
12633 2018-01-03 Jeff Law <law@redhat.com>
12635 PR middle-end/83654
12636 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12637 non-constant residual for zero at runtime and avoid probing in
12638 that case. Reorganize code for trailing problem to mirror handling
12641 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12643 PR tree-optimization/83501
12644 * tree-ssa-strlen.c (get_string_cst): New.
12645 (handle_char_store): Call get_string_cst.
12647 2018-01-03 Martin Liska <mliska@suse.cz>
12649 PR tree-optimization/83593
12650 * tree-ssa-strlen.c: Include tree-cfg.h.
12651 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12652 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12653 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12655 (strlen_dom_walker::before_dom_children): Call
12656 gimple_purge_dead_eh_edges. Dump tranformation with details
12658 (strlen_dom_walker::before_dom_children): Update call by adding
12659 new argument cleanup_eh.
12660 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12662 2018-01-03 Martin Liska <mliska@suse.cz>
12665 * cif-code.def (VARIADIC_THUNK): New enum value.
12666 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12669 2018-01-03 Jan Beulich <jbeulich@suse.com>
12671 * sse.md (mov<mode>_internal): Tighten condition for when to use
12672 vmovdqu<ssescalarsize> for TI and OI modes.
12674 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12676 Update copyright years.
12678 2018-01-03 Martin Liska <mliska@suse.cz>
12681 * ipa-visibility.c (function_and_variable_visibility): Skip
12682 functions with noipa attribure.
12684 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12686 * gcc.c (process_command): Update copyright notice dates.
12687 * gcov-dump.c (print_version): Ditto.
12688 * gcov.c (print_version): Ditto.
12689 * gcov-tool.c (print_version): Ditto.
12690 * gengtype.c (create_file): Ditto.
12691 * doc/cpp.texi: Bump @copying's copyright year.
12692 * doc/cppinternals.texi: Ditto.
12693 * doc/gcc.texi: Ditto.
12694 * doc/gccint.texi: Ditto.
12695 * doc/gcov.texi: Ditto.
12696 * doc/install.texi: Ditto.
12697 * doc/invoke.texi: Ditto.
12699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12701 * vector-builder.h (vector_builder::m_full_nelts): Change from
12702 unsigned int to poly_uint64.
12703 (vector_builder::full_nelts): Update prototype accordingly.
12704 (vector_builder::new_vector): Likewise.
12705 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12706 (vector_builder::operator ==): Likewise.
12707 (vector_builder::finalize): Likewise.
12708 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12709 Take the number of elements as a poly_uint64 rather than an
12711 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12712 from unsigned int to poly_uint64.
12713 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12714 (vec_perm_indices::new_vector): Likewise.
12715 (vec_perm_indices::length): Likewise.
12716 (vec_perm_indices::nelts_per_input): Likewise.
12717 (vec_perm_indices::input_nelts): Likewise.
12718 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12719 number of elements per input as a poly_uint64 rather than an
12720 unsigned int. Use the original encoding for variable-length
12721 vectors, rather than clamping each individual element.
12722 For the second and subsequent elements in each pattern,
12723 clamp the step and base before clamping their sum.
12724 (vec_perm_indices::series_p): Handle polynomial element counts.
12725 (vec_perm_indices::all_in_range_p): Likewise.
12726 (vec_perm_indices_to_tree): Likewise.
12727 (vec_perm_indices_to_rtx): Likewise.
12728 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12729 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12730 (tree_vector_builder::new_binary_operation): Handle polynomial
12731 element counts. Return false if we need to know the number
12732 of elements at compile time.
12733 * fold-const.c (fold_vec_perm): Punt if the number of elements
12734 isn't known at compile time.
12736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12738 * vec-perm-indices.h (vec_perm_builder): Change element type
12739 from HOST_WIDE_INT to poly_int64.
12740 (vec_perm_indices::element_type): Update accordingly.
12741 (vec_perm_indices::clamp): Handle polynomial element_types.
12742 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12743 (vec_perm_indices::all_in_range_p): Likewise.
12744 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12746 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12747 polynomial vec_perm_indices element types.
12748 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12749 * fold-const.c (fold_vec_perm): Likewise.
12750 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12751 * tree-vect-generic.c (lower_vec_perm): Likewise.
12752 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12753 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12754 element type to HOST_WIDE_INT.
12756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12757 Alan Hayward <alan.hayward@arm.com>
12758 David Sherwood <david.sherwood@arm.com>
12760 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12761 rather than an int. Use plus_constant.
12762 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12763 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12766 Alan Hayward <alan.hayward@arm.com>
12767 David Sherwood <david.sherwood@arm.com>
12769 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12770 a HOST_WIDE_INT to a poly_int64.
12772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12773 Alan Hayward <alan.hayward@arm.com>
12774 David Sherwood <david.sherwood@arm.com>
12776 * calls.c (load_register_parameters): Cope with polynomial
12777 mode sizes. Require a constant size for BLKmode parameters
12778 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12779 forces a parameter to be padded at the lsb end in order to
12780 fill a complete number of words, require the parameter size
12781 to be ordered wrt UNITS_PER_WORD.
12783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12784 Alan Hayward <alan.hayward@arm.com>
12785 David Sherwood <david.sherwood@arm.com>
12787 * reload1.c (spill_stack_slot_width): Change element type
12788 from unsigned int to poly_uint64_pod.
12789 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
12796 (reload_inner_reg_of_subreg, push_reload): Use it.
12798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12799 Alan Hayward <alan.hayward@arm.com>
12800 David Sherwood <david.sherwood@arm.com>
12802 * lra-constraints.c (process_alt_operands): Reject matched
12803 operands whose sizes aren't ordered.
12804 (match_reload): Refer to this check here.
12806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12807 Alan Hayward <alan.hayward@arm.com>
12808 David Sherwood <david.sherwood@arm.com>
12810 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12811 that the mode size is in the set {1, 2, 4, 8, 16}.
12813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12814 Alan Hayward <alan.hayward@arm.com>
12815 David Sherwood <david.sherwood@arm.com>
12817 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12818 Use plus_constant instead of gen_rtx_PLUS.
12820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12821 Alan Hayward <alan.hayward@arm.com>
12822 David Sherwood <david.sherwood@arm.com>
12824 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12825 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12826 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12827 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12828 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12829 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12830 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12831 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12832 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12833 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12835 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12836 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12837 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12838 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12839 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12840 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12841 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12842 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12843 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12844 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12846 * expr.c (emit_move_resolve_push): Treat the input and result
12847 of PUSH_ROUNDING as a poly_int64.
12848 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12849 (emit_push_insn): Likewise.
12850 * lra-eliminations.c (mark_not_eliminable): Likewise.
12851 * recog.c (push_operand): Likewise.
12852 * reload1.c (elimination_effects): Likewise.
12853 * rtlanal.c (nonzero_bits1): Likewise.
12854 * calls.c (store_one_arg): Likewise. Require the padding to be
12855 known at compile time.
12857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12858 Alan Hayward <alan.hayward@arm.com>
12859 David Sherwood <david.sherwood@arm.com>
12861 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12862 Use plus_constant instead of gen_rtx_PLUS.
12864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12865 Alan Hayward <alan.hayward@arm.com>
12866 David Sherwood <david.sherwood@arm.com>
12868 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12869 rather than an int.
12871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12872 Alan Hayward <alan.hayward@arm.com>
12873 David Sherwood <david.sherwood@arm.com>
12875 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12876 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12877 via stack temporaries. Treat the mode size as polynomial too.
12879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12880 Alan Hayward <alan.hayward@arm.com>
12881 David Sherwood <david.sherwood@arm.com>
12883 * expr.c (expand_expr_real_2): When handling conversions involving
12884 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12885 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12886 as a poly_uint64 too.
12888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12889 Alan Hayward <alan.hayward@arm.com>
12890 David Sherwood <david.sherwood@arm.com>
12892 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12895 Alan Hayward <alan.hayward@arm.com>
12896 David Sherwood <david.sherwood@arm.com>
12898 * combine.c (can_change_dest_mode): Handle polynomial
12899 REGMODE_NATURAL_SIZE.
12900 * expmed.c (store_bit_field_1): Likewise.
12901 * expr.c (store_constructor): Likewise.
12902 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12903 and polynomial REGMODE_NATURAL_SIZE.
12904 (gen_lowpart_common): Likewise.
12905 * reginfo.c (record_subregs_of_mode): Likewise.
12906 * rtlanal.c (read_modify_subreg_p): Likewise.
12908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12909 Alan Hayward <alan.hayward@arm.com>
12910 David Sherwood <david.sherwood@arm.com>
12912 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12913 numbers of elements.
12915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12916 Alan Hayward <alan.hayward@arm.com>
12917 David Sherwood <david.sherwood@arm.com>
12919 * match.pd: Cope with polynomial numbers of vector elements.
12921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12922 Alan Hayward <alan.hayward@arm.com>
12923 David Sherwood <david.sherwood@arm.com>
12925 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12926 in a POINTER_PLUS_EXPR.
12928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12929 Alan Hayward <alan.hayward@arm.com>
12930 David Sherwood <david.sherwood@arm.com>
12932 * omp-simd-clone.c (simd_clone_subparts): New function.
12933 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12934 (ipa_simd_modify_function_body): Likewise.
12936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12937 Alan Hayward <alan.hayward@arm.com>
12938 David Sherwood <david.sherwood@arm.com>
12940 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12941 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12942 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12943 (expand_vector_condition, vector_element): Likewise.
12944 (subparts_gt): New function.
12945 (get_compute_type): Use subparts_gt.
12946 (count_type_subparts): Delete.
12947 (expand_vector_operations_1): Use subparts_gt instead of
12948 count_type_subparts.
12950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12951 Alan Hayward <alan.hayward@arm.com>
12952 David Sherwood <david.sherwood@arm.com>
12954 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12955 (vect_compile_time_alias): ...this new function. Do the calculation
12956 on poly_ints rather than trees.
12957 (vect_prune_runtime_alias_test_list): Update call accordingly.
12959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12960 Alan Hayward <alan.hayward@arm.com>
12961 David Sherwood <david.sherwood@arm.com>
12963 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12965 (vect_schedule_slp_instance): Likewise.
12967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12968 Alan Hayward <alan.hayward@arm.com>
12969 David Sherwood <david.sherwood@arm.com>
12971 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12972 constant and extern definitions for variable-length vectors.
12973 (vect_get_constant_vectors): Note that the number of units
12974 is known to be constant.
12976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12977 Alan Hayward <alan.hayward@arm.com>
12978 David Sherwood <david.sherwood@arm.com>
12980 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12981 of units as polynomial. Choose between WIDE and NARROW based
12984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12985 Alan Hayward <alan.hayward@arm.com>
12986 David Sherwood <david.sherwood@arm.com>
12988 * tree-vect-stmts.c (simd_clone_subparts): New function.
12989 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12992 Alan Hayward <alan.hayward@arm.com>
12993 David Sherwood <david.sherwood@arm.com>
12995 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12996 vectors as polynomial. Use build_index_vector for
12997 IFN_GOMP_SIMD_LANE.
12999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13000 Alan Hayward <alan.hayward@arm.com>
13001 David Sherwood <david.sherwood@arm.com>
13003 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13004 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13005 for variable-length vectors.
13006 (vectorizable_mask_load_store): Treat the number of units as
13007 polynomial, asserting that it is constant if the condition has
13008 already been enforced.
13009 (vectorizable_store, vectorizable_load): Likewise.
13011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13012 Alan Hayward <alan.hayward@arm.com>
13013 David Sherwood <david.sherwood@arm.com>
13015 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13016 of units as polynomial. Punt if we can't tell at compile time
13017 which vector contains the final result.
13019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13020 Alan Hayward <alan.hayward@arm.com>
13021 David Sherwood <david.sherwood@arm.com>
13023 * tree-vect-loop.c (vectorizable_induction): Treat the number
13024 of units as polynomial. Punt on SLP inductions. Use an integer
13025 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13026 cast of such a series for variable-length floating-point
13029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13030 Alan Hayward <alan.hayward@arm.com>
13031 David Sherwood <david.sherwood@arm.com>
13033 * tree.h (build_index_vector): Declare.
13034 * tree.c (build_index_vector): New function.
13035 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13036 of units as polynomial, forcibly converting it to a constant if
13037 vectorizable_reduction has already enforced the condition.
13038 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13039 to create a {1,2,3,...} vector.
13040 (vectorizable_reduction): Treat the number of units as polynomial.
13041 Choose vectype_in based on the largest scalar element size rather
13042 than the smallest number of units. Enforce the restrictions
13045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13046 Alan Hayward <alan.hayward@arm.com>
13047 David Sherwood <david.sherwood@arm.com>
13049 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13050 number of units as polynomial.
13052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13053 Alan Hayward <alan.hayward@arm.com>
13054 David Sherwood <david.sherwood@arm.com>
13056 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13057 * target.def (autovectorize_vector_sizes): Return the vector sizes
13058 by pointer, using vector_sizes rather than a bitmask.
13059 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13060 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13061 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13063 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13064 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13065 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13066 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13067 * omp-general.c (omp_max_vf): Likewise.
13068 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13069 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13070 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13071 * tree-vect-slp.c (vect_slp_bb): Likewise.
13072 * doc/tm.texi: Regenerate.
13073 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13075 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13076 the vector size as a poly_uint64 rather than an unsigned int.
13077 (current_vector_size): Change from an unsigned int to a poly_uint64.
13078 (get_vectype_for_scalar_type): Update accordingly.
13079 * tree.h (build_truth_vector_type): Take the size and number of
13080 units as a poly_uint64 rather than an unsigned int.
13081 (build_vector_type): Add a temporary overload that takes
13082 the number of units as a poly_uint64 rather than an unsigned int.
13083 * tree.c (make_vector_type): Likewise.
13084 (build_truth_vector_type): Take the number of units as a poly_uint64
13085 rather than an unsigned int.
13087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13088 Alan Hayward <alan.hayward@arm.com>
13089 David Sherwood <david.sherwood@arm.com>
13091 * target.def (get_mask_mode): Take the number of units and length
13092 as poly_uint64s rather than unsigned ints.
13093 * targhooks.h (default_get_mask_mode): Update accordingly.
13094 * targhooks.c (default_get_mask_mode): Likewise.
13095 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13096 * doc/tm.texi: Regenerate.
13098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13099 Alan Hayward <alan.hayward@arm.com>
13100 David Sherwood <david.sherwood@arm.com>
13102 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13103 * omp-general.c (omp_max_vf): Likewise.
13104 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13105 (expand_omp_simd): Handle polynomial safelen.
13106 * omp-low.c (omplow_simd_context): Add a default constructor.
13107 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13108 (lower_rec_simd_input_clauses): Update accordingly.
13109 (lower_rec_input_clauses): Likewise.
13111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13112 Alan Hayward <alan.hayward@arm.com>
13113 David Sherwood <david.sherwood@arm.com>
13115 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13116 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13117 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13118 (vect_analyze_slp_cost): Likewise.
13119 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13120 (vect_model_load_cost): Likewise.
13122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13123 Alan Hayward <alan.hayward@arm.com>
13124 David Sherwood <david.sherwood@arm.com>
13126 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13127 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13128 from an unsigned int * to a poly_uint64_pod *.
13129 (calculate_unrolling_factor): New function.
13130 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13133 Alan Hayward <alan.hayward@arm.com>
13134 David Sherwood <david.sherwood@arm.com>
13136 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13137 from an unsigned int to a poly_uint64.
13138 (_loop_vec_info::slp_unrolling_factor): Likewise.
13139 (_loop_vec_info::vectorization_factor): Change from an int
13141 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13142 (vect_get_num_vectors): New function.
13143 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13144 (vect_get_num_copies): Use vect_get_num_vectors.
13145 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13146 to an unsigned int *.
13147 (vect_analyze_data_refs): Change min_vf from an int * to a
13149 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13150 than an unsigned HOST_WIDE_INT.
13151 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13152 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13153 to an unsigned int *.
13154 (vect_analyze_data_ref_dependences): Likewise.
13155 (vect_compute_data_ref_alignment): Handle polynomial vf.
13156 (vect_enhance_data_refs_alignment): Likewise.
13157 (vect_prune_runtime_alias_test_list): Likewise.
13158 (vect_shift_permute_load_chain): Likewise.
13159 (vect_supportable_dr_alignment): Likewise.
13160 (dependence_distance_ge_vf): Take the vectorization factor as a
13161 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13162 (vect_analyze_data_refs): Change min_vf from an int * to a
13164 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13165 vfm1 as a poly_uint64 rather than an int. Make the same change
13166 for the returned bound_scalar.
13167 (vect_gen_vector_loop_niters): Handle polynomial vf.
13168 (vect_do_peeling): Likewise. Update call to
13169 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13170 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13172 * tree-vect-loop.c (vect_determine_vectorization_factor)
13173 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13174 (vect_get_known_peeling_cost): Likewise.
13175 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13176 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13177 (vect_transform_loop): Likewise. Use the lowest possible VF when
13178 updating the upper bounds of the loop.
13179 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13180 rather than an int.
13181 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13182 polynomial unroll factors.
13183 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13184 (vect_make_slp_decision): Likewise.
13185 (vect_supported_load_permutation_p): Likewise, and polynomial
13187 (vect_analyze_slp_cost): Handle polynomial vf.
13188 (vect_slp_analyze_node_operations): Likewise.
13189 (vect_slp_analyze_bb_1): Likewise.
13190 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13191 than an unsigned HOST_WIDE_INT.
13192 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13193 (vectorizable_load): Handle polynomial vf.
13194 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13196 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13199 Alan Hayward <alan.hayward@arm.com>
13200 David Sherwood <david.sherwood@arm.com>
13202 * match.pd: Handle bit operations involving three constants
13203 and try to fold one pair.
13205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13207 * tree-vect-loop-manip.c: Include gimple-fold.h.
13208 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13209 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13210 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13211 Add a path that uses a step of VF instead of 1, but disable it
13213 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13214 and niters_no_overflow parameters. Update calls to
13215 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13216 Create a new SSA name if the latter choses to use a ste other
13217 than zero, and return it via niters_vector_mult_vf_var.
13218 * tree-vect-loop.c (vect_transform_loop): Update calls to
13219 vect_do_peeling, vect_gen_vector_loop_niters and
13220 slpeel_make_loop_iterate_ntimes.
13221 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13222 (vect_gen_vector_loop_niters): Update declarations after above changes.
13224 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13226 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13227 128-bit round to integer instructions.
13228 (ceil<mode>2): Likewise.
13229 (btrunc<mode>2): Likewise.
13230 (round<mode>2): Likewise.
13232 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13234 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13235 unaligned VSX load/store on P8/P9.
13236 (expand_block_clear): Allow the use of unaligned VSX
13237 load/store on P8/P9.
13239 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13241 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13243 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13244 swap associated with both a load and a store.
13246 2018-01-02 Andrew Waterman <andrew@sifive.com>
13248 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13249 * config/riscv/riscv.md (clear_cache): Use it.
13251 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13253 * web.c: Remove out-of-date comment.
13255 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13257 * expr.c (fixup_args_size_notes): Check that any existing
13258 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13259 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13260 (emit_single_push_insn): ...here.
13262 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13264 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13265 (const_vector_encoded_nelts): New function.
13266 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13267 (const_vector_int_elt, const_vector_elt): Declare.
13268 * emit-rtl.c (const_vector_int_elt_1): New function.
13269 (const_vector_elt): Likewise.
13270 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13271 of CONST_VECTOR_ELT.
13273 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13275 * expr.c: Include rtx-vector-builder.h.
13276 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13277 directly on the tree encoding.
13278 (const_vector_from_tree): Likewise.
13279 * optabs.c: Include rtx-vector-builder.h.
13280 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13281 sequence of "u" values.
13282 * vec-perm-indices.c: Include rtx-vector-builder.h.
13283 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13284 directly on the vec_perm_indices encoding.
13286 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13288 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13289 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13290 * rtx-vector-builder.h: New file.
13291 * rtx-vector-builder.c: Likewise.
13292 * rtl.h (rtx_def::u2): Add a const_vector field.
13293 (CONST_VECTOR_NPATTERNS): New macro.
13294 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13295 (CONST_VECTOR_DUPLICATE_P): Likewise.
13296 (CONST_VECTOR_STEPPED_P): Likewise.
13297 (CONST_VECTOR_ENCODED_ELT): Likewise.
13298 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13299 (unwrap_const_vec_duplicate): Likewise.
13300 (const_vec_series_p): Check for a non-duplicated vector encoding.
13301 Say that the function only returns true for integer vectors.
13302 * emit-rtl.c: Include rtx-vector-builder.h.
13303 (gen_const_vec_duplicate_1): Delete.
13304 (gen_const_vector): Call gen_const_vec_duplicate instead of
13305 gen_const_vec_duplicate_1.
13306 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13307 (gen_const_vec_duplicate): Use rtx_vector_builder.
13308 (gen_const_vec_series): Likewise.
13309 (gen_rtx_CONST_VECTOR): Likewise.
13310 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13311 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13312 Build a new vector rather than modifying a CONST_VECTOR in-place.
13313 (handle_special_swappables): Update call accordingly.
13314 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13315 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13316 Build a new vector rather than modifying a CONST_VECTOR in-place.
13317 (handle_special_swappables): Update call accordingly.
13319 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13321 * simplify-rtx.c (simplify_const_binary_operation): Use
13322 CONST_VECTOR_ELT instead of XVECEXP.
13324 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13326 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13327 the selector elements to be different from the data elements
13328 if the selector is a VECTOR_CST.
13329 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13330 ssizetype for the selector.
13332 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13334 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13335 before testing each element individually.
13336 * tree-vect-generic.c (lower_vec_perm): Likewise.
13338 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13340 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13341 * selftest-run-tests.c (selftest::run_tests): Call it.
13342 * vector-builder.h (vector_builder::operator ==): New function.
13343 (vector_builder::operator !=): Likewise.
13344 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13345 (vec_perm_indices::all_from_input_p): New function.
13346 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13347 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13348 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13349 instead of reading the VECTOR_CST directly. Detect whether both
13350 vector inputs are the same before constructing the vec_perm_indices,
13351 and update the number of inputs argument accordingly. Use the
13352 utility functions added above. Only construct sel2 if we need to.
13354 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13356 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13357 the broadcast of the low byte.
13358 (expand_mult_highpart): Use an explicit encoding for the permutes.
13359 * optabs-query.c (can_mult_highpart_p): Likewise.
13360 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13361 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13362 (vectorizable_bswap): Likewise.
13363 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13364 explicit encoding for the power-of-2 permutes.
13365 (vect_permute_store_chain): Likewise.
13366 (vect_grouped_load_supported): Likewise.
13367 (vect_permute_load_chain): Likewise.
13369 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13371 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13372 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13373 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13374 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13375 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13376 (vect_gen_perm_mask_any): Likewise.
13378 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13380 * int-vector-builder.h: New file.
13381 * vec-perm-indices.h: Include int-vector-builder.h.
13382 (vec_perm_indices): Redefine as an int_vector_builder.
13383 (auto_vec_perm_indices): Delete.
13384 (vec_perm_builder): Redefine as a stand-alone class.
13385 (vec_perm_indices::vec_perm_indices): New function.
13386 (vec_perm_indices::clamp): Likewise.
13387 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13388 (vec_perm_indices::new_vector): New function.
13389 (vec_perm_indices::new_expanded_vector): Update for new
13390 vec_perm_indices class.
13391 (vec_perm_indices::rotate_inputs): New function.
13392 (vec_perm_indices::all_in_range_p): Operate directly on the
13393 encoded form, without computing elided elements.
13394 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13395 encoding. Update for new vec_perm_indices class.
13396 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13397 the given vec_perm_builder.
13398 (expand_vec_perm_var): Update vec_perm_builder constructor.
13399 (expand_mult_highpart): Use vec_perm_builder instead of
13400 auto_vec_perm_indices.
13401 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13402 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13403 or double series encoding as appropriate.
13404 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13405 vec_perm_indices instead of auto_vec_perm_indices.
13406 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13407 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13408 (vect_permute_store_chain): Likewise.
13409 (vect_grouped_load_supported): Likewise.
13410 (vect_permute_load_chain): Likewise.
13411 (vect_shift_permute_load_chain): Likewise.
13412 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13413 (vect_transform_slp_perm_load): Likewise.
13414 (vect_schedule_slp_instance): Likewise.
13415 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13416 (vectorizable_mask_load_store): Likewise.
13417 (vectorizable_bswap): Likewise.
13418 (vectorizable_store): Likewise.
13419 (vectorizable_load): Likewise.
13420 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13421 vec_perm_indices instead of auto_vec_perm_indices. Use
13422 tree_to_vec_perm_builder to read the vector from a tree.
13423 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13424 vec_perm_builder instead of a vec_perm_indices.
13425 (have_whole_vector_shift): Use vec_perm_builder and
13426 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13427 truncation to calc_vec_perm_mask_for_shift.
13428 (vect_create_epilog_for_reduction): Likewise.
13429 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13430 from auto_vec_perm_indices to vec_perm_indices.
13431 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13432 instead of changing individual elements.
13433 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13434 the vector in d.perm.
13435 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13436 from auto_vec_perm_indices to vec_perm_indices.
13437 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13438 instead of changing individual elements.
13439 (arm_vectorize_vec_perm_const): Use new_vector to install
13440 the vector in d.perm.
13441 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13442 Update vec_perm_builder constructor.
13443 (rs6000_expand_interleave): Likewise.
13444 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13445 (rs6000_expand_interleave): Likewise.
13447 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13449 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13450 to qimode could truncate the indices.
13451 * optabs.c (expand_vec_perm_var): Likewise.
13453 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13455 * Makefile.in (OBJS): Add vec-perm-indices.o.
13456 * vec-perm-indices.h: New file.
13457 * vec-perm-indices.c: Likewise.
13458 * target.h (vec_perm_indices): Replace with a forward class
13460 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13461 * optabs.h: Include vec-perm-indices.h.
13462 (expand_vec_perm): Delete.
13463 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13464 (expand_vec_perm_const): Declare.
13465 * target.def (vec_perm_const_ok): Replace with...
13466 (vec_perm_const): ...this new hook.
13467 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13468 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13469 * doc/tm.texi: Regenerate.
13470 * optabs.def (vec_perm_const): Delete.
13471 * doc/md.texi (vec_perm_const): Likewise.
13472 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13473 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13474 expand_vec_perm for constant permutation vectors. Assert that
13475 the mode of variable permutation vectors is the integer equivalent
13476 of the mode that is being permuted.
13477 * optabs-query.h (selector_fits_mode_p): Declare.
13478 * optabs-query.c: Include vec-perm-indices.h.
13479 (selector_fits_mode_p): New function.
13480 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13481 is defined, instead of checking whether the vec_perm_const_optab
13482 exists. Use targetm.vectorize.vec_perm_const instead of
13483 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13484 fit in the vector mode before using a variable permute.
13485 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13486 vec_perm_indices instead of an rtx.
13487 (expand_vec_perm): Replace with...
13488 (expand_vec_perm_const): ...this new function. Take the selector
13489 as a vec_perm_indices rather than an rtx. Also take the mode of
13490 the selector. Update call to shift_amt_for_vec_perm_mask.
13491 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13492 Use vec_perm_indices::new_expanded_vector to expand the original
13493 selector into bytes. Check whether the indices fit in the vector
13494 mode before using a variable permute.
13495 (expand_vec_perm_var): Make global.
13496 (expand_mult_highpart): Use expand_vec_perm_const.
13497 * fold-const.c: Includes vec-perm-indices.h.
13498 * tree-ssa-forwprop.c: Likewise.
13499 * tree-vect-data-refs.c: Likewise.
13500 * tree-vect-generic.c: Likewise.
13501 * tree-vect-loop.c: Likewise.
13502 * tree-vect-slp.c: Likewise.
13503 * tree-vect-stmts.c: Likewise.
13504 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13506 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13507 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13508 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13509 (aarch64_vectorize_vec_perm_const): ...this new function.
13510 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13511 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13512 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13513 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13514 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13515 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13516 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13518 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13519 check for NEON modes.
13520 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13521 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13522 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13523 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13525 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13526 the old VEC_PERM_CONST conditions.
13527 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13528 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13529 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13530 (ia64_vectorize_vec_perm_const_ok): Merge into...
13531 (ia64_vectorize_vec_perm_const): ...this new function.
13532 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13533 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13534 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13535 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13536 * config/mips/mips.c (mips_expand_vec_perm_const)
13537 (mips_vectorize_vec_perm_const_ok): Merge into...
13538 (mips_vectorize_vec_perm_const): ...this new function.
13539 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13540 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13541 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13542 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13543 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13544 (rs6000_expand_vec_perm_const): Delete.
13545 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13547 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13548 (altivec_expand_vec_perm_const_le): Take each operand individually.
13549 Operate on constant selectors rather than rtxes.
13550 (altivec_expand_vec_perm_const): Likewise. Update call to
13551 altivec_expand_vec_perm_const_le.
13552 (rs6000_expand_vec_perm_const): Delete.
13553 (rs6000_vectorize_vec_perm_const_ok): Delete.
13554 (rs6000_vectorize_vec_perm_const): New function.
13555 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13556 an element count and rtx array.
13557 (rs6000_expand_extract_even): Update call accordingly.
13558 (rs6000_expand_interleave): Likewise.
13559 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13560 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13561 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13562 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13563 (rs6000_expand_vec_perm_const): Delete.
13564 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13565 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13566 (altivec_expand_vec_perm_const_le): Take each operand individually.
13567 Operate on constant selectors rather than rtxes.
13568 (altivec_expand_vec_perm_const): Likewise. Update call to
13569 altivec_expand_vec_perm_const_le.
13570 (rs6000_expand_vec_perm_const): Delete.
13571 (rs6000_vectorize_vec_perm_const_ok): Delete.
13572 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13573 reference to the SPE evmerge intructions.
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/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13579 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13581 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13583 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13585 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13586 vector mode and that that mode matches the mode of the data
13588 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13589 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13590 directly using expand_vec_perm_1 when forcing selectors into
13592 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13594 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13596 * optabs-query.h (can_vec_perm_p): Delete.
13597 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13598 * optabs-query.c (can_vec_perm_p): Split into...
13599 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13600 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13601 particular selector is valid.
13602 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13603 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13604 (vect_grouped_load_supported): Likewise.
13605 (vect_shift_permute_load_chain): Likewise.
13606 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13607 (vect_transform_slp_perm_load): Likewise.
13608 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13609 (vectorizable_bswap): Likewise.
13610 (vect_gen_perm_mask_checked): Likewise.
13611 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13612 implementations of variable permutation vectors into account
13613 when deciding which selector to use.
13614 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13615 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13616 with a false third argument.
13617 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13618 to test whether the constant selector is valid and can_vec_perm_var_p
13619 to test whether a variable selector is valid.
13621 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13623 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13624 * optabs-query.c (can_vec_perm_p): Likewise.
13625 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13626 instead of vec_perm_indices.
13627 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13628 (vect_gen_perm_mask_checked): Likewise,
13629 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13630 (vect_gen_perm_mask_checked): Likewise,
13632 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13634 * optabs-query.h (qimode_for_vec_perm): Declare.
13635 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13636 (qimode_for_vec_perm): ...this new function.
13637 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13639 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13641 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13642 does not have a conditional at the top.
13644 2018-01-02 Richard Biener <rguenther@suse.de>
13646 * ipa-inline.c (big_speedup_p): Fix expression.
13648 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13651 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13654 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13658 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13659 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13660 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13661 cond_taken_branch_cost 3->4.
13663 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13665 PR tree-optimization/83581
13666 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13667 TODO_cleanup_cfg if any changes have been made.
13669 PR middle-end/83608
13670 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13671 convert_modes if target mode has the right side, but different mode
13674 PR middle-end/83609
13675 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13676 last argument when extracting from CONCAT. If either from_real or
13677 from_imag is NULL, use expansion through memory. If result is not
13678 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13679 the parts directly to inner mode, if even that fails, use expansion
13682 PR middle-end/83623
13683 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13684 check for bswap in mode rather than HImode and use that in expand_unop
13687 Copyright (C) 2018 Free Software Foundation, Inc.
13689 Copying and distribution of this file, with or without modification,
13690 are permitted in any medium without royalty provided the copyright
13691 notice and this notice are preserved.