1 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4 * config/aarch64/aarch64.md: New patterns to generate stp
6 (store_pair_sw, store_pair_dw): New patterns to generate stp for
7 single words and double words.
8 (load_pair_sw, load_pair_dw): Likewise.
9 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
11 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
13 * config/aarch64/aarch64-ldpstp.md: Modify peephole
14 for different mode ldpstp and add peephole for merged zero stores.
16 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18 (aarch64_gen_store_pair): Rename calls to match new patterns.
19 (aarch64_gen_load_pair): Rename calls to match new patterns.
20 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
21 (load_pair<DREG:mode><DREG2:mode>): ... This.
22 (store_pair<mode>): Rename to...
23 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
24 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
26 (V_INT_EQUIV): Handle SImode.
27 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
30 2018-05-22 Martin Sebor <msebor@redhat.com>
32 PR tree-optimization/85826
33 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
34 assuming that a DECL necesarily has a constant size.
36 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
39 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
41 2018-05-22 Richard Biener <rguenther@suse.de>
43 PR tree-optimization/85834
44 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
45 non-constant and non-zero memset arguments.
47 2018-05-22 Martin Liska <mliska@suse.cz>
50 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
52 2018-05-22 Richard Biener <rguenther@suse.de>
54 PR tree-optimization/85863
55 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
56 comparisons when vectype is specified.
57 (vectorizable_condition): Do not specify vectype for
58 vect_is_simple_cond when SLP vectorizing.
60 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
63 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
64 define __ibm128 as long double.
65 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
66 as a distinct type when IEEE 128-bit support is enabled.
67 (init_float128_ieee): Fix up conversions between IFmode and IEEE
68 128-bit types to use the correct functions.
69 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
70 convert between 128-bit floating point types that have different
71 modes but the same representation, instead of using gen_lowpart to
73 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
75 (IFKF_reg): New attributes to give the register constraints for
77 (extend<mode>tf2_internal): New insns to mark an explicit
78 conversion between 128-bit floating point types that have a
79 different mode but share the same representation.
81 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
83 PR tree-optimization/85814
84 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
85 a null return from get_strinfo when unsharing the next
88 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
91 * varasm.c (weak_finish): Clean up weak_decls.
93 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
95 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
96 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
98 * config/aarch64/iterators.md (ABAL): New int iterator.
101 (sur): Add mappings for the above.
102 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
104 (aarch64_<sur>abal<mode>_4): Likewise.
105 (aarch64_<sur>adalp<mode>_3): Likewise.
106 (<sur>sadv16qi): New define_expand.
108 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
110 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
111 (*movdf_internal): Ditto.
112 (*rcpsf2_sse): Ditto.
113 (*rsqrtsf2_sse): Ditto.
114 (*sqrt<mode>2_sse): Ditto.
116 2018-05-21 Tamar Christina <tamar.christina@arm.com>
118 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
120 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
121 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
122 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
123 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
125 * config/aarch64/arm_neon.h: Likewise.
126 * config/aarch64/iterators.md (VQ_I): New.
128 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
130 * config.gcc: Add arc/t-multilib-linux to tmake_file for
132 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
135 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
137 * config/nds32/constraints.md (S): New constraint.
138 * config/nds32/nds32.md (call_internal): Use constraint S.
139 (call_value_internal): Likewise.
140 (sibcall_internal): Likewise.
141 (sibcall_value_internal): Likewise.
143 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
144 Chung-Ju Wu <jasonwucj@gmail.com>
146 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
149 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
150 Chung-Ju Wu <jasonwucj@gmail.com>
152 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
153 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
154 (nds32_rtx_costs_impl): Simplify.
155 (nds32_address_cost_impl): Simplify.
156 (nds32_init_rtx_costs): New function.
157 (nds32_rtx_costs_speed_prefer): Likewise.
158 (nds32_rtx_costs_size_prefer): Likewise.
159 (nds32_address_cost_speed_prefer): Likewise.
160 (nds32_address_cost_speed_fwprop): Likewise.
161 (nds32_address_cost_size_prefer): Likewise.
162 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
163 * config/nds32/nds32.c (nds32_option_override): Use
164 nds32_init_rtx_costs function.
166 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
168 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
169 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
170 (TARGET_PIPELINE_N8): Likewise.
171 (TARGET_PIPELINE_N10): Likewise.
172 (TARGET_PIPELINE_N13): Likewise.
173 (TARGET_PIPELINE_GRAYWOLF): Likewise.
175 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
177 * config/nds32/nds32-fpu.md: Update copyright year.
179 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
181 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
183 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
185 * config/nds32/nds32.c
186 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
187 * config/nds32/nds32.opt (minline-asm-r15): New option.
189 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
191 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
193 * config/nds32/nds32.md (abssi2): New pattern.
195 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
197 * config/i386/i386.md (rex64namesuffix): New mode attribute.
198 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
199 Merge insn pattern from sse_cvtsi2ss<round_name> and
200 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
201 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
202 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
203 using SWI48 mode iterator.
204 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
205 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
206 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
207 pattern from sse_cvttss2si<round_saeonly_name>
208 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
209 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
210 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
211 using SWI48 mode iterator.
212 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
213 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
214 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
215 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
216 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
217 using SWI48 mode iterator.
218 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
219 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
220 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
221 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
222 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
224 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
225 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
226 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
227 pattern from sse_cvttsd2si<round_saeonly_name>
228 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
230 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
232 * config/nds32/nds32-md-auxiliary.c
233 (nds32_valid_smw_lwm_base_p): Refine.
234 (nds32_output_smw_single_word): Refine.
235 (nds32_output_smw_double_word): New.
236 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
238 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
240 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
241 (nds32_output_stack_pop): Refine.
242 (nds32_expand_unaligned_load): Refine.
243 (nds32_expand_unaligned_store): Refine.
245 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
246 Chung-Ju Wu <jasonwucj@gmail.com>
248 * config/nds32/constants.md: Add TP_REGNUM constant.
249 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
250 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
252 * config/nds32/nds32-doubleword.md: Consider flag_pic.
253 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
254 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
255 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
256 and PIC code generation.
257 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
259 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
261 * config/nds32/nds32.md: Support TLS and PIC.
262 * config/nds32/nds32.c: Support TLS and PIC.
263 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
264 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
267 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
269 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
272 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
273 Chung-Ju Wu <jasonwucj@gmail.com>
275 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
276 * config/nds32/nds32-md-auxiliary.c
277 (symbolic_reference_mentioned_p): New.
278 (nds32_legitimize_ict_address): New.
279 (nds32_expand_ict_move): New.
280 (nds32_indirect_call_referenced_p): New.
281 (nds32_symbol_binds_local_p): Delete.
282 (nds32_long_call_p): Modify.
283 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
284 * config/nds32/nds32-protos.h
285 (symbolic_reference_mentioned_p): Declare.
286 (nds32_legitimize_ict_address): Declare.
287 (nds32_expand_ict_move): Declare.
288 (nds32_indirect_call_referenced_p): Declare.
289 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
290 (nds32_relax_group): Use nds32_ict_const_p as condition.
291 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
292 (nds32_asm_file_start): Output ict_model directive in asm code.
293 (nds32_legitimate_address_p): Consider indirect call.
294 (nds32_print_operand): Consider indirect call.
295 (nds32_print_operand_address): Consider indirect call.
296 (nds32_insert_attributes): Handle "indirect_call" attribute.
297 (TARGET_LEGITIMATE_ADDRESS_P): Define.
298 (TARGET_LEGITIMATE_CONSTANT_P): Define.
299 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
300 (TARGET_DELEGITIMIZE_ADDRESS): Define.
301 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
302 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
303 (TARGET_ICT_MODEL_SMALL): Define.
304 (TARGET_ICT_MODEL_LARGE): Define.
305 * config/nds32/nds32.md (movsi): Consider ict model.
306 (call, call_value): Consider ict model.
307 (sibcall, sibcall_value): Consider ict model.
308 * config/nds32/nds32.opt (mict-model): New option.
309 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
312 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
313 Monk Chiang <sh.chiang04@gmail.com>
314 Jim Wilson <jimw@sifive.com>
316 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
317 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
318 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
319 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
320 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
321 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
322 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
323 compute save_libcall_adjustment properly.
324 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
325 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
326 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
327 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
328 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
329 (ABI_SPEC): Handle mabi=ilp32e.
330 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
332 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
333 <-march>: Add rv32e as an example.
335 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
338 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
339 (intra_create_variable_infos): Handle C++ constructors.
341 2018-05-18 Martin Liska <mliska@suse.cz>
343 * passes.def: Remove a redundant pass.
345 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
348 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
350 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
352 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
354 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
356 (armv2, armv2a, armv3, armv3m): Delete architectures.
357 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
358 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
359 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
361 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
362 (*mulsidi3adddi): Likewise.
363 (mulsidi3): Likewise.
364 (*mulsidi3_nov6): Likewise.
365 (umulsidi3): Likewise.
366 (umulsidi3_nov6): Likewise.
367 (umaddsidi4): Likewise.
368 (*umulsidi3adddi): Likewise.
369 (smulsi3_highpart): Likewise.
370 (*smulsi3_highpart_nov6): Likewise.
371 (umulsi3_highpart): Likewise.
372 (*umulsi3_highpart_nov6): Likewise.
373 * config/arm/arm.h (arm_arch3m): Delete.
374 * config/arm/arm.c (arm_arch3m): Delete.
375 (arm_option_override_internal): Update armv3-related comment.
376 (arm_configure_build_target): Delete use of isa_bit_mode32.
377 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
378 (arm_rtx_costs_internal): Delete check of arm_arch3m.
379 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
382 * config/arm/arm-protos.h (arm_arch3m): Delete.
383 * config/arm/arm-tables.opt: Regenerate.
384 * config/arm/arm-tune.md: Likewise.
385 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
386 deleted architectures.
388 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
390 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
391 (armv5t, armv5te): New features.
392 (ARMv5, ARMv5e): Delete fgroups.
393 (ARMv5t, ARMv5te): Adjust for above changes.
395 (armv5, armv5e): Delete arches.
396 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
398 (*call_reg_arm): Likewise.
399 (*call_value_reg_armv5): Likewise.
400 (*call_value_reg_arm): Likewise.
401 (*call_symbol): Likewise.
402 (*call_value_symbol): Likewise.
403 (*sibcall_insn): Likewise.
404 (*sibcall_value_insn): Likewise.
406 (prefetch): Likewise.
407 (define_split and define_peephole2 dependent on arm_arch5):
409 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
411 (TARGET_ARM_QBIT): Likewise.
412 (TARGET_DSP_MULTIPLY): Likewise.
413 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
414 (arm_arch5, arm_arch5e): Delete.
415 (arm_arch5t, arm_arch5te): Declare.
416 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
417 (arm_arch5t): Declare.
418 (arm_option_reconfigure_globals): Update for the above.
419 (arm_options_perform_arch_sanity_checks): Update comment, replace
420 use of arm_arch5 with arm_arch5t.
421 (use_return_insn): Likewise.
422 (arm_emit_call_insn): Likewise.
423 (output_return_instruction): Likewise.
424 (arm_final_prescan_insn): Likewise.
425 (arm_coproc_builtin_available): Likewise.
426 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
427 arm_arch5e with arm_arch5t and arm_arch5te.
428 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
429 (arm_arch5t, arm_arch5te): Declare.
430 * config/arm/arm-tables.opt: Regenerate.
431 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
432 * config/arm/t-multilib: Likewise.
433 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
434 instead of arm_arch5.
435 (*call_reg_thumb1): Likewise.
436 (*call_value_reg_thumb1_v5): Likewise.
437 (*call_value_reg_thumb1): Likewise.
438 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
440 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
442 2018-05-18 Martin Liska <mliska@suse.cz>
444 PR gcov-profile/84846
445 * doc/gcov.texi: Document -t option of gcov tool.
447 2018-05-18 Martin Liska <mliska@suse.cz>
449 PR gcov-profile/84846
450 * gcov.c (print_usage): Add new -t option.
451 (process_args): Handle the option.
452 (generate_results): Use stdout as output when requested by
455 2018-05-18 Martin Liska <mliska@suse.cz>
457 PR gcov-profile/84846
458 * coverage.c (coverage_init): Write PWD to .gcno file.
459 * doc/gcov.texi: Document how working directory is printed.
460 * gcov-dump.c (dump_gcov_file): Print PWD.
461 * gcov.c (output_intermediate_file): Likewise.
462 (read_graph_file): Read PWD string.
463 (output_lines): Print PWD.
465 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
468 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
469 for retval and return false if all args to phi are zero.
471 2018-05-18 Richard Biener <rguenther@suse.de>
473 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
475 (evrp_dom_walker::before_dom_children): Call it.
477 2018-05-18 Richard Biener <rguenther@suse.de>
479 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
480 results when processing array refs with variable index.
482 2018-05-18 Toon Moene <toon@moene.org>
484 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
485 directly after that of -floop-interchange. Indicate that both
486 options are enabled by default when specifying -O3.
488 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
490 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
491 iterator. Delete separate integer-mode vec_set<mode> expander.
492 (aarch64_simd_vec_setv2di): Delete.
493 (vec_setv2di): Delete.
494 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
495 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
496 the "w, r" alternative.
498 2018-05-18 Martin Liska <mliska@suse.cz>
500 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
501 * tree-pass.h (make_pass_lower_switch_O0): New function.
502 * tree-switch-conversion.c (node_has_low_bound): Remove.
503 (node_has_high_bound): Likewise.
504 (node_is_bounded): Likewise.
505 (class pass_lower_switch): Make it a template type and create
507 (pass_lower_switch::execute): Add template argument.
508 (make_pass_lower_switch): New function.
509 (make_pass_lower_switch_O0): New function.
510 (do_jump_if_equal): Remove.
511 (emit_case_nodes): Simplify to just handle all 3 cases and leave
512 all the hard work to tree optimization passes.
514 2018-05-18 Martin Liska <mliska@suse.cz>
516 * dbgcnt.c (limit_low): Renamed from limit.
517 (limit_high): New variable.
518 (dbg_cnt_is_enabled): Check for upper limit.
519 (dbg_cnt): Adjust dumping.
520 (dbg_cnt_set_limit_by_index): Add new argument for high
522 (dbg_cnt_set_limit_by_name): Likewise.
523 (dbg_cnt_process_single_pair): Parse new format.
524 (dbg_cnt_process_opt): Use strtok.
525 (dbg_cnt_list_all_counters): Remove 'value' and add
527 * doc/invoke.texi: Document changes.
529 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
531 * doc/sourcebuild.texi (scalar_all_fma): Document.
532 * tree.def (FMA_EXPR): Delete.
533 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
534 * internal-fn.c (ternary_direct): New macro.
535 (expand_ternary_optab_fn): Likewise.
536 (direct_ternary_optab_supported_p): Likewise.
537 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
538 * builtins.c (fold_builtin_fma): Delete.
539 (fold_builtin_3): Don't call it.
540 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
541 * expr.c (expand_expr_real_2): Likewise.
542 * fold-const.c (operand_equal_p): Likewise.
543 (fold_ternary_loc): Likewise.
544 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
545 * gimple.c (DEFTREECODE): Likewise.
546 * gimplify.c (gimplify_expr): Likewise.
547 * optabs-tree.c (optab_for_tree_code): Likewise.
548 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
549 * tree-eh.c (operation_could_trap_p): Likewise.
550 (stmt_could_throw_1_p): Likewise.
551 * tree-inline.c (estimate_operator_cost): Likewise.
552 * tree-pretty-print.c (dump_generic_node): Likewise.
553 (op_code_prio): Likewise.
554 * tree-ssa-loop-im.c (stmt_cost): Likewise.
555 * tree-ssa-operands.c (get_expr_operands): Likewise.
556 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
557 * fold-const-call.h (fold_fma): Delete.
558 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
559 CFN_FNMA and CFN_FNMS.
561 * genmatch.c (combined_fn): New enum.
562 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
563 (commutative_op): New function.
564 (commutate): Use it. Handle more than 2 operands.
565 (dt_operand::gen_gimple_expr): Use commutative_op.
566 (parser::parse_expr): Allow :c to be used with non-binary
567 operators if the commutative operand is known.
568 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
569 CFN_FMS, CFN_FNMA and CFN_FNMS.
570 (backprop::process_assign_use): Remove FMA_EXPR handling.
571 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
572 (gen_hsa_fma): New function.
573 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
574 IFN_FNMA and IFN_FNMS.
575 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
576 * gimple-fold.h (follow_all_ssa_edges): Declare.
577 * gimple-fold.c (follow_all_ssa_edges): New function.
578 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
579 gimple_build interface and use follow_all_ssa_edges to fold the result.
580 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
581 instead of checking for optabs directly.
582 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
583 rather than FMA_EXPRs.
584 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
585 call to IFN_FMA instead of an FMA_EXPR.
587 2018-05-17 Jim Wilson <jimw@sifive.com>
589 * expr.c (do_tablejump): When converting index to Pmode, if we have a
590 sign extended promoted subreg, and the range does not have the sign bit
591 set, then do a sign extend.
593 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
594 test, check for sign extended subreg and/or constant operands, and
595 do a sign extend in that case.
597 2018-05-17 Steve Ellcey <sellcey@cavium.com>
599 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
600 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
602 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
603 Change logics_shift_reg to logics_shift_imm.
604 (thunderx2t99_fp_loadpair_basic): Delete.
605 (thunderx2t99_fp_storepair_basic): Delete.
606 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
607 (thunderx2t99_asimd_polynomial): Delete.
608 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
609 and neon_fp_mul_d_scalar_q.
610 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
611 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
612 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
613 (thunderx2t99_asimd_lut): Add missing tbl types.
614 (thunderx2t99_asimd_ext): Delete.
615 (thunderx2t99_asimd_load1_1_mult): Delete.
616 (thunderx2t99_asimd_load1_2_mult): Delete.
617 (thunderx2t99_asimd_load1_ldp): New.
618 (thunderx2t99_asimd_load1): New.
619 (thunderx2t99_asimd_load2): Add missing *load2* types.
620 (thunderx2t99_asimd_load3): New.
621 (thunderx2t99_asimd_load4): New.
622 (thunderx2t99_asimd_store1_1_mult): Delete.
623 (thunderx2t99_asimd_store1_2_mult): Delete.
624 (thunderx2t99_asimd_store2_mult): Delete.
625 (thunderx2t99_asimd_store2_onelane): Delete.
626 (thunderx2t99_asimd_store_stp): New.
627 (thunderx2t99_asimd_store1): New.
628 (thunderx2t99_asimd_store2): New.
629 (thunderx2t99_asimd_store3): New.
630 (thunderx2t99_asimd_store4): New.
632 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
634 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
635 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
637 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
638 Segher Boessenkool <segher@kernel.crashing.org>
641 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
644 2018-05-17 Richard Biener <rguenther@suse.de>
646 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
647 for pruning loop and prune defs feeding only already visited PHIs.
649 2018-05-17 Richard Biener <rguenther@suse.de>
651 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
653 2018-05-17 Bin Cheng <bin.cheng@arm.com>
654 Richard Biener <rguenther@suse.de>
656 PR tree-optimization/85793
657 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
658 for VMAT_ELEMENTWISE.
660 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
662 * internal-fn.h (lookup_internal_fn): Declare
663 * internal-fn.c (lookup_internal_fn): New function.
664 * gimple.c (gimple_build_call_from_tree): Handle calls to
666 * gimple-pretty-print.c (dump_gimple_call): Print "." before
667 internal function names.
668 * tree-pretty-print.c (dump_generic_node): Likewise.
669 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
671 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
673 * gimple-fold.h (gimple_build): Make the function forms take
674 combined_fn rather than built_in_function.
675 (gimple_simplify): Likewise.
676 * gimple-match-head.c (gimple_simplify): Likewise.
677 * gimple-fold.c (gimple_build): Likewise.
678 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
679 rather than gimple_build_call_internal.
680 (get_initial_defs_for_reduction): Likewise.
681 (vect_create_epilog_for_reduction): Likewise.
682 (vectorizable_live_operation): Likewise.
684 2018-05-17 Martin Liska <mliska@suse.cz>
686 * gimple-ssa-sprintf.c (format_directive): Do not use
687 space in between 'G_' and '('.
689 2018-05-17 Jakub Jelinek <jakub@redhat.com>
692 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
693 even if the mask is not all ones.
696 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
698 (ix86_gimple_fold_builtin): Likewise.
701 * config/i386/i386.c: Include tree-vector-builder.h.
702 (ix86_vector_shift_count): New function.
703 (ix86_fold_builtin): Fold shift builtins by scalar count.
704 (ix86_gimple_fold_builtin): Likewise.
706 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
707 _mm512_setzero): New intrinsics.
709 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
710 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
712 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
713 code generation for cases where splatting a value is not useful.
714 * simplify-rtx.c (simplify_ternary_operation): Simplify
715 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
716 mode to a vec_concat.
718 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
720 * config.gcc: Support "goldmont-plus".
721 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
723 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
724 PROCESSOR_GOLDMONT_PLUS.
725 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
726 (processor_target_table): Add "goldmont-plus".
727 (PTA_GOLDMONT_PLUS): Define.
728 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
729 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
730 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
731 (fold_builtin_cpu): Add "goldmont-plus".
732 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
733 (ix86_option_override_internal): Add "goldmont-plus".
734 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
735 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
736 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
737 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
739 2018-05-17 Richard Biener <rguenther@suse.de>
741 PR tree-optimization/85757
742 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
743 remove defs that only feed that PHI from further processing.
745 2018-05-16 Jim Wilson <jimw@sifive.com>
747 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
749 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
751 2018-05-16 Mark Wielaard <mark@klomp.org>
753 * dwarf2out.c (count_index_strings): New function.
754 (output_indirect_strings): Call count_index_strings and generate
755 header for dwarf_version >= 5.
757 2018-05-16 Mark Wielaard <mark@klomp.org>
759 * dwarf2out.c (dwarf_FORM): New function.
760 (set_indirect_string): Use dwarf_FORM.
761 (reset_indirect_string): Likewise.
762 (size_of_die): Likewise.
763 (value_format): Likewise.
764 (output_die): Likewise.
765 (add_skeleton_AT_string): Likewise.
766 (output_macinfo_op): Likewise.
767 (index_string): Likewise.
768 (output_index_string_offset): Likewise.
769 (output_index_string): Likewise.
770 (count_index_strings): Likewise.
772 2018-05-16 Carl Love <cel@us.ibm.com>
774 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
775 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
777 2018-05-16 Martin Jambor <mjambor@suse.cz>
779 * ipa-prop.c (ipa_free_all_edge_args): Remove.
780 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
782 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
784 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
785 (fnma<mode>4): Likewise.
786 (fms<mode>4): Likewise.
787 (fnms<mode>4): Likewise.
788 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
789 (aarch64_fnma<mode>4): Likewise.
790 (aarch64_fms<mode>4): Likewise.
791 (aarch64_fnms<mode>4): Likewise.
792 (aarch64_fnmadd<mode>4): Likewise.
794 2018-05-16 Jason Merrill <jason@redhat.com>
796 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
798 2018-05-16 Richard Biener <rguenther@suse.de>
800 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
801 (dump_stmt_cost): Declare.
802 (add_stmt_cost): Dump cost we add.
803 (add_stmt_costs): New function.
804 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
806 (vect_analyze_stmt): Adjust prototype.
807 (vectorizable_condition): Likewise.
808 (vectorizable_live_operation): Likewise.
809 (vectorizable_reduction): Likewise.
810 (vectorizable_induction): Likewise.
811 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
812 cost vector to pass to vectorizable_ and record afterwards.
813 (vect_model_reduction_cost): Take cost vector argument and adjust.
814 (vect_model_induction_cost): Likewise.
815 (vectorizable_reduction): Likewise.
816 (vectorizable_induction): Likewise.
817 (vectorizable_live_operation): Likewise.
818 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
819 SLP_TREE_NUMBER_OF_VEC_STMTS.
820 (vect_analyze_slp_cost_1): Remove.
821 (vect_analyze_slp_cost): Likewise.
822 (vect_slp_analyze_node_operations): Take visited args and
823 a target cost vector. Avoid processing already visited stmt sets.
824 (vect_slp_analyze_operations): Use a local cost vector to gather
825 costs and register those of non-discarded instances.
826 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
827 (vect_schedule_slp_instance): Remove copying of
828 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
830 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
831 adding cost. Record cost entry location.
832 (vect_prologue_cost_for_slp_op): Function to compute cost of
833 a constant or invariant generated for SLP vect in the prologue,
834 split out from vect_analyze_slp_cost_1.
835 (vect_model_simple_cost): Make static. Adjust for SLP costing.
836 (vect_model_promotion_demotion_cost): Likewise.
837 (vect_model_store_cost): Likewise, make static.
838 (vect_model_load_cost): Likewise.
839 (vectorizable_bswap): Add cost vector arg and adjust.
840 (vectorizable_call): Likewise.
841 (vectorizable_simd_clone_call): Likewise.
842 (vectorizable_conversion): Likewise.
843 (vectorizable_assignment): Likewise.
844 (vectorizable_shift): Likewise.
845 (vectorizable_operation): Likewise.
846 (vectorizable_store): Likewise.
847 (vectorizable_load): Likewise.
848 (vectorizable_condition): Likewise.
849 (vectorizable_comparison): Likewise.
850 (can_vectorize_live_stmts): Likewise.
851 (vect_analyze_stmt): Likewise.
852 (vect_transform_stmt): Adjust calls to vectorizable_*.
853 * tree-vectorizer.c: Include gimple-pretty-print.h.
854 (dump_stmt_cost): New function.
856 2018-05-16 Richard Biener <rguenther@suse.de>
858 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
859 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
860 * tree-ssa-dse.c: Include tree-ssa-loop.h.
861 (check_name): New callback.
862 (dse_classify_store): Track cycles via a visited bitmap of PHI
863 defs and simplify handling of in-loop and across loop dead stores
864 and properly fail for loop-variant refs. Handle byte-tracking with
865 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
868 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
870 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
871 (vect_get_mask_type_for_stmt): Likewise.
872 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
874 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
875 to determine the statement's vector type and the vector type that
876 should be used for calculating nunits. Deal with cases in which
877 the type has to be deferred.
878 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
879 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
880 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
881 (vect_determine_vf_for_stmt): New functions, split out from...
882 (vect_determine_vectorization_factor): ...here.
883 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
884 (vect_get_mask_type_for_stmt): New functions, split out from
885 vect_determine_vectorization_factor.
887 2018-05-16 Richard Biener <rguenther@suse.de>
889 * tree-cfg.c (verify_gimple_assign_ternary): Properly
890 verify the [VEC_]COND_EXPR embedded comparison.
892 2018-05-15 Martin Sebor <msebor@redhat.com>
894 PR tree-optimization/85753
895 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
896 RECORD_TYPE in addition to ARRAY_TYPE.
898 2018-05-15 Martin Sebor <msebor@redhat.com>
901 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
903 2018-05-15 Richard Biener <rguenther@suse.de>
905 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
906 add by_clobber_p one. Change algorithm to collect all defs
907 representing uses we need to walk and try reducing them to
908 a single one before failing.
909 (dse_dom_walker::dse_optimize_stmt): Adjust.
911 2018-05-13 Mark Wielaard <mark@klomp.org>
913 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
914 (size_of_loc_descr): Likewise.
915 (output_loc_operands): Likewise.
916 (output_loc_operands_raw): Likewise.
917 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
918 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
919 (hash_loc_operands): Likewise.
920 (compare_loc_operands): Likewise.
922 2018-05-14 Mark Wielaard <mark@klomp.org>
924 * dwarf2out.c (count_index_addrs): New function.
925 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
927 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
929 PR tree-optimization/83648
930 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
931 return value as malloc candidate.
933 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
936 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
937 as true in call to suggest_attribute.
939 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
941 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
944 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
946 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
948 (aarch64_crypto_aesd_fused): Likewise.
950 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
952 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
953 (movsi_aarch64): Likewise.
954 (load_pairsi): Likewise.
955 (load_pairdi): Likewise.
956 (store_pairsi): Likewise.
957 (store_pairdi): Likewise.
958 (load_pairsf): Likewise.
959 (load_pairdf): Likewise.
960 (store_pairsf): Likewise.
961 (store_pairdf): Likewise.
962 (zero_extend): Likewise.
963 (trunc): Swap alternatives.
964 (fcvt_target): Add '?' to prefer w over r.
966 2018-05-14 Jakub Jelinek <jakub@redhat.com>
969 * config/i386/i386.md: Disallow non-commutative arithmetics in
970 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
971 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
972 in the peephole2 before it.
974 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
976 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
977 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
978 (ix86_handle_option): Handle -mcldemote.
979 * config.gcc: New header.
980 * config/i386/cldemoteintrin.h: New file.
981 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
982 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
984 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
985 OPTION_MASK_ISA_CLDEMOTE.
986 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
987 (ix86_valid_target_attribute_inner_p): Ditto.
988 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
989 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
990 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
991 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
992 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
994 * config/i386/i386.opt: Add -mcldemote.
995 * config/i386/x86intrin.h: New header.
996 * doc/invoke.texi: Add -mcldemote.
998 2018-05-14 Richard Biener <rguenther@suse.de>
1000 * doc/match-and-simplify.texi: Adjust :s documentation.
1002 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1004 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1005 intended memcpy size.
1006 (REORDER_45): Likewise.
1008 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1010 * sort.cc: New file.
1011 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1012 * vec.c (qsort_chk): Use gcc_qsort.
1013 * Makefile.in (OBJS-libcommon): Add sort.o.
1014 (build/sort.o): New target. Use it...
1015 (BUILD_RTL): ... here, and...
1016 (build/gencfn-macros): ... here, and...
1017 (build/genmatch): ... here.
1019 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1020 Chung-Ju Wu <jasonwucj@gmail.com>
1022 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1023 * config/nds32/nds32-graywolf.md: New file.
1024 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1025 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1027 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1028 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1029 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1030 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1031 * config/nds32/pipelines.md: Include n15 settings.
1033 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1034 Chung-Ju Wu <jasonwucj@gmail.com>
1036 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1037 * config/nds32/nds32-n13.md: New file.
1038 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1039 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1041 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1042 * config/nds32/nds32.md (pipeline_model): Add n13.
1043 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1044 * config/nds32/pipelines.md: Include n13 settings.
1046 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1047 Chung-Ju Wu <jasonwucj@gmail.com>
1049 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1050 * config/nds32/nds32-n10.md: New file.
1051 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1052 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1054 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1055 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1056 * config/nds32/nds32.md (pipeline_model): Add n10.
1057 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1058 * config/nds32/pipelines.md: Include n10 settings.
1060 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1061 Kito Cheng <kito.cheng@gmail.com>
1062 Chung-Ju Wu <jasonwucj@gmail.com>
1064 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1065 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1066 Add enum values for DSP extension instructions.
1067 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1069 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1070 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1072 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1073 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1074 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1075 * config/nds32/nds32-intrinsic.md: Likewise.
1076 * config/nds32/nds32_intrinsic.h: Likewise.
1077 * config/nds32/nds32-md-auxiliary.c: Likewise.
1078 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1079 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1080 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1081 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1082 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1083 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1084 TYPE_DMAC in switch statement.
1085 * config/nds32/nds32.c: New checking and implementation for DSP
1086 extension instructions.
1087 * config/nds32/nds32.h: Likewise.
1088 * config/nds32/nds32.md: Likewise.
1089 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1090 * config/nds32/predicates.md: Implement new predicates for DSP
1093 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1095 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1096 Reformat alternatives and attributes so it is easier to identify
1097 which constraints/attributes go with which instruction.
1098 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1099 (mov<mode>_softfloat32, FMOVE64): Likewise.
1100 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1101 (mov<mode>_softfloat64, FMOVE64): Likewise.
1103 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1105 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1107 (Basic PowerPC Built-in Functions): The new name of the
1108 subsection previously known as "PowerPC Built-in Functions".
1109 (Basic PowerPC Built-in Functions Available on all Configurations):
1111 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1112 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1113 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1114 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1116 2018-05-11 Martin Jambor <mjambor@suse.cz>
1119 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1122 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1125 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1127 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1129 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1130 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1131 (ix86_handle_option): Handle -mwaitpkg.
1132 * config.gcc: New header.
1133 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1134 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1135 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1137 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1138 OPTION_MASK_ISA_WAITPKG.
1139 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1140 (ix86_option_override_internal): Add PTA_WAITPKG.
1141 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1142 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1143 IX86_BUILTIN_TPAUSE.
1144 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1145 __builtin_ia32_umwait and __builtin_ia32_tpause.
1146 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1147 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1148 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1149 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1150 UNSPECV_TPAUSE): New.
1151 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1152 * config/i386/i386.opt: Add -mwaitpkg.
1153 * config/i386/waitpkgintrin.h: New file.
1154 * config/i386/x86intrin.h: New header.
1155 * doc/invoke.texi: Add -mwaitpkg.
1157 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1160 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1162 (cortex-m0): Use armv6s-m isa.
1163 (cortex-m0plus): Likewise.
1164 (cortex-m1): Likewise.
1165 (cortex-m0.small-multiply): Likewise.
1166 (cortex-m0plus.small-multiply): Likewise.
1167 (cortex-m1.small-multiply): Likewise.
1169 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1170 Jakub Jelinek <jakub@redhat.com>
1172 PR tree-optimization/85692
1173 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1174 source permute as well.
1176 2018-05-11 Martin Liska <mliska@suse.cz>
1179 * doc/extend.texi: Document LLVM style format for no_sanitize
1182 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1184 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1185 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1186 (mode_supports_vsx_dform_quad): Likewise.
1187 (mode_supports_vmx_dform): Move these functions to be next to the
1188 other mode_supports functions.
1189 (mode_supports_dq_form): Likewise.
1190 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1191 mode_supports_dq_form.
1192 (reg_offset_addressing_ok_p): Likewise.
1193 (offsettable_ok_by_alignment): Likewise.
1194 (rs6000_legitimate_offset_address_p): Likewise.
1195 (legitimate_lo_sum_address_p): Likewise.
1196 (rs6000_legitimize_address): Likewise.
1197 (rs6000_legitimize_reload_address): Likewise.
1198 (rs6000_secondary_reload_inner): Likewise.
1199 (rs6000_preferred_reload_class): Likewise.
1200 (rs6000_output_move_128bit): Likewise.
1202 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1204 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1205 Generate SImode target register for null target.
1206 <case IX86_BUILTIN_XGETBV>: Ditto.
1207 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1208 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1210 2018-05-10 Carl Love <cel@us.ibm.com>
1212 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1213 dcbtt and dcbtstt if operands[2] is 0.
1215 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1218 * config/i386/sse.md (usadv64qi): New expander.
1220 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1222 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1223 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1224 -maltivec=be support.
1225 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1226 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1227 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1228 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1229 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1230 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1231 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1232 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1233 altivec_vsumsws): Adjust.
1234 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1235 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1236 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1238 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1239 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1240 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1241 (altivec_lve<VI_char>x): Delete expand.
1242 (*altivec_lve<VI_char>x_internal): Rename to...
1243 (altivec_lve<VI_char>x): ... this.
1244 (altivec_lvxl_<mode>): Delete expand.
1245 (*altivec_lvxl_<mode>_internal): Rename to ...
1246 (altivec_lvxl_<mode>): ... this.
1247 (altivec_stvxl_<mode>): Delete expand.
1248 (*altivec_stvxl_<mode>_internal): Rename to ...
1249 (altivec_stvxl_<mode>): ... this.
1250 (altivec_stve<VI_char>x): Delete expand.
1251 (*altivec_stve<VI_char>x_internal): Rename to ...
1252 (altivec_stve<VI_char>x): ... this.
1253 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1254 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1255 reduc_plus_scal_<mode>): Adjust.
1256 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1258 (rs6000_cpu_cpp_builtins): Adjust.
1259 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1260 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1261 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1262 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1263 -maltivec=be support.
1264 (rs6000_split_vec_extract_var): Adjust.
1265 (rs6000_split_v4si_init): Adjust.
1266 (swap_selector_for_mode): Delete.
1267 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1268 altivec_expand_stvex_be): Delete.
1269 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1270 -maltivec=be support.
1271 (rs6000_gimple_fold_builtin): Ditto.
1272 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1274 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1275 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1276 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1277 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1278 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1279 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1280 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1281 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1282 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1283 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1284 anonymous split): Adjust.
1285 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1286 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1288 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1290 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1291 when --with-gxx-include-dir is also specified.
1292 * configure: Regenerate.
1294 2018-05-09 Jim Wilson <jimw@sifive.com>
1297 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1298 * config/riscv/t-withmultilib: New.
1299 * config/riscv/withmultilib.h: New.
1300 * doc/install.texi: Document RISC-V --with-multilib-list support.
1302 2018-05-09 Richard Biener <rguenther@suse.de>
1304 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1306 (vect_bb_vectorization_profitable_p): Adjust. Compute
1307 actual scalar cost using the cost vector and the add_stmt_cost
1310 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1312 PR rtl-optimization/85645
1313 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1314 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1316 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1318 PR rtl-optimization/85645
1319 * shrink-wrap.c (spread_components): Return a boolean saying if
1320 anything was changed.
1321 (try_shrink_wrapping_separate): Iterate spread_components until
1322 nothing changes anymore.
1324 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1326 PR rtl-optimization/85645
1327 * regrename.c (build_def_use): Also kill the chains that include the
1328 destination of a REG_CFA_REGISTER note.
1330 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1332 PR rtl-optimization/85645
1333 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1334 insn that has a REG_CFA_REGISTER note.
1336 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1338 * cfgexpand.c (expand_clobber): New function.
1339 (expand_gimple_stmt_1): Use it.
1340 * tree-vect-stmts.c (vect_clobber_variable): New function,
1342 (vectorizable_simd_clone_call): ...here.
1343 (vectorizable_store): Emit a clobber either side of an
1344 IFN_STORE_LANES sequence.
1345 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1347 2018-05-09 Tom de Vries <tom@codesourcery.com>
1350 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1351 (define_insn "trap_if_false"): Add exit after trap.
1353 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1355 PR rtl-optimization/85638
1356 * bb-reorder.c: Include common/common-target.h.
1357 (create_forwarder_block): New function extracted from...
1358 (fix_up_crossing_landing_pad): ...here. Rename into...
1359 (dw2_fix_up_crossing_landing_pad): ...this.
1360 (sjlj_fix_up_crossing_landing_pad): New function.
1361 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1362 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1363 from both partitions and exit the loop after one iteration.
1365 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1368 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1370 (Basic PowerPC Built-in Functions): The new name of the
1371 subsection previously known as "PowerPC Built-in Functions".
1372 (Basic PowerPC Built-in Functions Available on all Configurations):
1374 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1376 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1378 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1380 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1383 2018-05-08 Jim Wilson <jimw@sifive.com>
1385 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1386 (LD_EMUL_SUFFIX): New.
1387 (LINK_SPEC): Use it.
1389 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1391 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1393 (Basic PowerPC Built-in Functions): The new name of the
1394 subsection previously known as "PowerPC Built-in Functions".
1395 (Basic PowerPC Built-in Functions Available on all Configurations):
1397 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1399 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1401 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1403 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1406 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1409 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1410 after cmpelim optimization.
1412 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1414 * config.gcc: Support "goldmont".
1415 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1416 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1418 * config/i386/i386.c (m_GOLDMONT): Define.
1419 (processor_target_table): Add "goldmont".
1420 (PTA_GOLDMONT): Define.
1421 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1422 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1423 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1424 (fold_builtin_cpu): Add "goldmont".
1425 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1426 (ix86_option_override_internal): Add "goldmont".
1427 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1428 (processor_type): Add PROCESSOR_GOLDMONT.
1429 * config/i386/i386.md: Add CPU "glm".
1430 * config/i386/glm.md: New file.
1431 * config/i386/x86-tune.def: Add m_GOLDMONT.
1432 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1434 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1437 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1439 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1440 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1441 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1444 * config/i386/i386.c (ix86_fold_builtin): Handle
1445 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1448 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1449 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1451 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1454 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1455 (check_arch): Likewise.
1456 (check_fpu): Return the result rather than printing it.
1457 (end arch): Fix operator precedence.
1458 (end cpu): Likewise.
1459 (END): Print the result from check_fpu.
1461 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1462 Alan Hayward <alan.hayward@arm.com>
1463 David Sherwood <david.sherwood@arm.com>
1465 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1466 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1467 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1468 (*fcmuo<mode>_and): New patterns.
1470 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1472 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1473 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1474 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1475 (cmp_op, sve_imm_con): New code attributes.
1476 (SVE_COND_INT_CMP, imm_con): Delete.
1477 (cmp_op): Remove above unspecs from int attribute.
1478 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1480 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1481 comparison-specific unspecs.
1482 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1483 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1484 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1485 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1486 (*vec_fcm<cmp_op><mode>): Rename to...
1487 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1488 (*vec_fcmuo<mode>): Rename to...
1489 (*fcmuo<mode>): ...this and adjust likewise.
1490 (*pred_fcm<cmp_op><mode>): New pattern.
1491 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1492 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1494 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1496 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1497 (aarch64_emit_sve_predicated_cond): New function.
1498 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1499 (aarch64_emit_unspec_cond_or): Replace with...
1500 (aarch64_emit_sve_or_conds): ...this new function. Use
1501 aarch64_emit_sve_ptrue_op for the individual comparisons and
1502 aarch64_emit_binop to OR them together.
1503 (aarch64_emit_inverted_unspec_cond): Replace with...
1504 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1505 aarch64_emit_sve_ptrue_op for the comparison and
1506 aarch64_emit_unop to invert the result.
1507 (aarch64_expand_sve_vec_cmp_float): Update after the above
1508 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1510 2018-05-07 Nathan Sidwell <nathan@acm.org>
1512 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1513 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1514 (Backwards Compatibility): Likewise.
1516 2018-05-07 Luis Machado <luis.machado@linaro.org>
1520 2018-05-07 Luis Machado <luis.machado@linaro.org>
1522 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1523 <prefetch_dynamic_strides>: New const bool field.
1524 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1525 prefetch_dynamic_strides.
1526 (exynosm1_prefetch_tune): Likewise.
1527 (thunderxt88_prefetch_tune): Likewise.
1528 (thunderx_prefetch_tune): Likewise.
1529 (thunderx2t99_prefetch_tune): Likewise.
1530 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1531 (aarch64_override_options_internal): Update to set
1532 PARAM_PREFETCH_DYNAMIC_STRIDES.
1533 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1534 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1535 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1536 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1537 prefetch-dynamic-strides setting.
1539 2018-05-07 Luis Machado <luis.machado@linaro.org>
1541 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1542 <minimum_stride>: New const int field.
1543 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1544 minimum_stride field.
1545 (exynosm1_prefetch_tune): Likewise.
1546 (thunderxt88_prefetch_tune): Likewise.
1547 (thunderx_prefetch_tune): Likewise.
1548 (thunderx2t99_prefetch_tune): Likewise.
1549 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1550 (aarch64_override_options_internal): Update to set
1551 PARAM_PREFETCH_MINIMUM_STRIDE.
1552 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1553 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1554 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1555 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1556 stride is constant and is below the minimum stride threshold.
1558 2018-05-07 Luis Machado <luis.machado@linaro.org>
1560 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1563 2018-05-07 Luis Machado <luis.machado@linaro.org>
1565 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1566 <prefetch_dynamic_strides>: New const bool field.
1567 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1568 prefetch_dynamic_strides.
1569 (exynosm1_prefetch_tune): Likewise.
1570 (thunderxt88_prefetch_tune): Likewise.
1571 (thunderx_prefetch_tune): Likewise.
1572 (thunderx2t99_prefetch_tune): Likewise.
1573 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1574 (aarch64_override_options_internal): Update to set
1575 PARAM_PREFETCH_DYNAMIC_STRIDES.
1576 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1577 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1578 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1579 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1580 prefetch-dynamic-strides setting.
1582 2018-05-07 Luis Machado <luis.machado@linaro.org>
1584 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1585 <minimum_stride>: New const int field.
1586 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1587 minimum_stride field.
1588 (exynosm1_prefetch_tune): Likewise.
1589 (thunderxt88_prefetch_tune): Likewise.
1590 (thunderx_prefetch_tune): Likewise.
1591 (thunderx2t99_prefetch_tune): Likewise.
1592 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1593 (aarch64_override_options_internal): Update to set
1594 PARAM_PREFETCH_MINIMUM_STRIDE.
1595 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1596 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1597 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1598 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1599 stride is constant and is below the minimum stride threshold.
1601 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1604 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1605 the type is addressable. Don't force op into register if it has
1608 2018-05-05 Roland McGrath <mcgrathr@google.com>
1611 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1612 any section for which we don't know a specific type it should have,
1613 regardless of name. Previously this was done only for the exact
1614 names ".init_array", ".fini_array", and ".preinit_array".
1615 (default_elf_asm_named_section): Add comment about
1616 relationship with default_section_type_flags and SECTION_NOTYPE.
1617 (get_section): Don't consider it a type conflict if one side has
1618 SECTION_NOTYPE and the other doesn't, as long as neither has the
1619 SECTION_BSS et al used in the default_section_type_flags logic.
1621 2018-05-05 Tom de Vries <tom@codesourcery.com>
1624 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1625 (workaround_barsyncs): New function.
1626 (nvptx_reorg): Use workaround_barsyncs.
1627 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1628 (define_expand "nvptx_membar_cta"): New define_expand.
1629 (define_insn "*nvptx_membar_cta"): New insn.
1631 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1633 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1634 To improve optimization opportunities.
1635 * builtin-types.def: The new needed builtin types for the above.
1637 2018-05-04 Richard Biener <rguenther@suse.de>
1639 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1640 * gimple-ssa-store-merging.c
1641 (imm_store_chain_info::output_merged_store): Remove redundant create,
1642 release split_store vector contents on failure.
1643 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1644 scalar stmt vector on cache hit.
1646 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1648 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1650 * config.gcc (powerpc-xilinx-eabi*): Remove.
1651 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1653 (fusion_addis_mem_combo_load): Ditto.
1654 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1656 (rs6000_cpu_cpp_builtins): Ditto.
1657 * config/rs6000/rs6000-linux.c
1658 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1659 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1660 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1662 (rs6000_setup_reg_addr_masks): Ditto.
1663 (rs6000_init_hard_regno_mode_ok): Ditto.
1664 (rs6000_option_override_internal): Ditto.
1665 (legitimate_lo_sum_address_p): Ditto.
1666 (rs6000_legitimize_address): Ditto.
1667 (rs6000_legitimize_reload_address): Ditto.
1668 (rs6000_legitimate_address_p): Ditto.
1669 (abi_v4_pass_in_fpr): Ditto.
1670 (setup_incoming_varargs): Ditto.
1671 (rs6000_gimplify_va_arg): Ditto.
1672 (rs6000_split_multireg_move): Ditto.
1673 (rs6000_savres_strategy): Ditto.
1674 (rs6000_emit_prologue_components): Ditto.
1675 (rs6000_emit_epilogue_components): Ditto.
1676 (rs6000_emit_prologue): Ditto.
1677 (rs6000_emit_epilogue): Ditto.
1678 (rs6000_elf_file_end): Ditto.
1679 (rs6000_function_value): Ditto.
1680 (rs6000_libcall_value): Ditto.
1681 * config/rs6000/rs6000.h: Ditto.
1682 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1683 (TARGET_MINMAX): ... this. New.
1684 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1685 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1686 (*movsi_internal1_single): Delete.
1687 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1688 mfpu=, mxilinx-fpu): Delete.
1689 * config/rs6000/singlefp.h: Delete.
1690 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1691 * config/rs6000/t-rs6000: Ditto.
1692 * config/rs6000/t-xilinx: Delete.
1693 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1694 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1695 (VStype_simple): Delete.
1696 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1697 * config/rs6000/xfpu.h: Delete.
1698 * config/rs6000/xfpu.md: Delete.
1699 * config/rs6000/xilinx.h: Delete.
1700 * config/rs6000/xilinx.opt: Delete.
1701 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1702 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1704 2018-05-04 Tom de Vries <tom@codesourcery.com>
1707 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1710 2018-05-04 Richard Biener <rguenther@suse.de>
1713 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1714 (expand_complex_div_wide): Likewise.
1715 (expand_complex_operations_1): Likewise.
1716 (expand_complex_libcall): Preserve EH info of the original stmt.
1717 (tree_lower_complex): Handle removed blocks.
1718 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1719 on complex multiplication and division libcall builtins.
1721 2018-05-04 Richard Biener <rguenther@suse.de>
1724 * fold-const.c (negate_expr_p): Restrict negation of operand
1725 zero of a division to when we know that can happen without
1727 (fold_negate_expr_1): Likewise.
1729 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1732 * real.h (real_nextafter): Declare.
1733 * real.c (real_nextafter): New function.
1734 * fold-const-call.c (fold_const_nextafter): New function.
1735 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1736 CASE_CFN_NEXTTOWARD.
1737 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1738 even when arg1_mode is different from arg0_mode.
1740 2018-05-03 Nathan Sidwell <nathan@acm.org>
1742 * doc/extend.texi (Deprecated Features): Remove
1744 (Backwards Compatibility): Likewise.
1745 * doc/invoke.texi (C++ Language Options): Likewise.
1746 (C++ Dialect Options): Likewise.
1748 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1751 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1752 _mm512_mask_mullox_epi64): New intrinsics.
1754 2018-05-03 Tom de Vries <tom@codesourcery.com>
1757 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1758 dump files): Add offload-tree.
1760 2018-05-03 Richard Biener <rguenther@suse.de>
1762 PR tree-optimization/85615
1763 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1764 to loops not nested in BBs loop father to avoid creating multi-entry
1767 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1769 PR tree-optimization/70291
1770 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1771 arguments. Change return type to tree. Emit libcall as a new
1772 statement rather than replacing existing one when inplace_p is true.
1773 (expand_complex_multiplication_components): New function.
1774 (expand_complex_multiplication): Expand floating-point complex
1775 multiplication using the above.
1776 (expand_complex_division): Rename inner_type parameter to type.
1777 Update expand_complex_libcall call-site.
1778 (expand_complex_operations_1): Update expand_complex_multiplication
1779 and expand_complex_division call-sites.
1781 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1784 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1785 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1786 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1787 the highest significant bit of the shift count mask is clear. In
1788 check whether and[sq]i3 is needed verify that all significant bits
1789 of the shift count other than the highest are set.
1791 2018-05-02 Tom de Vries <tom@codesourcery.com>
1794 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1795 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1796 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1797 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1798 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1799 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1800 __builtin_goacc_parlevel_size.
1802 2018-05-02 Richard Biener <rguenther@suse.de>
1804 PR tree-optimization/85597
1805 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1806 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1809 2018-05-02 Tom de Vries <tom@codesourcery.com>
1812 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1813 dump files): Add ltrans-tree.
1815 2018-05-02 Tom de Vries <tom@codesourcery.com>
1818 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1819 dump files): Add wpa-ipa.
1821 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1823 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1824 powerpc*-*-linux*paired* target.
1825 * config/rs6000/750cl.h: Delete.
1826 * config/rs6000/paired.h: Delete.
1827 * config/rs6000/paired.md: Delete.
1828 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1830 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1831 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1832 comment. Remove paired float support.
1833 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1834 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1836 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1837 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1839 * config/rs6000/rs6000.c: Remove paired float support.
1840 (paired_expand_vector_init, paired_expand_vector_move,
1841 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1842 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1843 paired_expand_builtin, paired_expand_predicate_builtin,
1844 paired_init_builtins): Delete.
1845 * config/rs6000/rs6000.h: Remove paired float support.
1846 * config/rs6000/rs6000.md: Remove paired float support.
1847 (move_from_CR_ov_bit): Delete.
1848 * config/rs6000/rs6000.opt (mpaired): Delete.
1849 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1850 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1852 2018-05-02 Richard Biener <rguenther@suse.de>
1855 * gimplify.c (gimplify_save_expr): When in SSA form allow
1856 SAVE_EXPRs to compute to SSA vars.
1858 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1861 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1862 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1863 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1864 clobber operands[2], instead use a new pseudo. Formatting fixes.
1866 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1868 PR tree-optimization/85586
1869 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1870 exit early for statements in the same group if the accesses are
1873 2018-05-02 Tom de Vries <tom@codesourcery.com>
1876 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1879 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1881 PR tree-optimization/85143
1882 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1884 2018-05-01 Tom de Vries <tom@codesourcery.com>
1887 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1888 not found" error message.
1890 2018-05-01 Tom de Vries <tom@codesourcery.com>
1893 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1894 * vec.c (test_ordered_remove_if): New function.
1895 (vec_c_tests): Call test_ordered_remove_if.
1896 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1897 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1898 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1899 VEC_ORDERED_REMOVE_IF.
1901 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1903 PR tree-optimization/82665
1904 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1905 pointer subtraction where arguments come from a memchr call.
1907 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1909 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1910 --push-state --as-needed and --pop-state instead of --as-needed and
1911 --no-as-needed if ld supports it.
1912 * configure: Regenerated.
1915 * doc/install.texi2html: Replace _002d with - and _002a with * in
1916 generated html files using sed.
1918 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1921 * gcc-rich-location.c (blank_line_before_p): New function.
1922 (use_new_line): New function.
1923 (gcc_rich_location::add_fixit_insert_formatted): New function.
1924 * gcc-rich-location.h
1925 (gcc_rich_location::add_fixit_insert_formatted): New function.
1927 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1929 * selftest.c (assert_streq): Rename "expected" and "actual" to
1930 "val1" and "val2". Extend NULL-handling to cover both inputs
1931 symmetrically, while still requiring both to be non-NULL for a pass.
1932 * selftest.h (assert_streq): Rename "expected" and "actual" to
1934 (ASSERT_EQ): Likewise.
1935 (ASSERT_EQ_AT): Likewise.
1936 (ASSERT_KNOWN_EQ): Likewise.
1937 (ASSERT_KNOWN_EQ_AT): Likewise.
1938 (ASSERT_NE): Likewise.
1939 (ASSERT_MAYBE_NE): Likewise.
1940 (ASSERT_MAYBE_NE_AT): Likewise.
1941 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1942 the assertion to pass.
1943 (ASSERT_STREQ_AT): Likewise.
1945 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1947 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1948 interaction with -pie.
1950 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1952 * selftest.h: Fix alphabetization of per-source-file selftest
1955 2018-04-30 Jason Merrill <jason@redhat.com>
1957 PR c++/61982 - dead stores to destroyed objects.
1958 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1961 2018-04-30 Jason Merrill <jason@redhat.com>
1963 * tree.c (build_clobber): New.
1964 * tree.h: Declare it.
1965 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1967 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1969 * diagnostic-show-locus.c (layout::layout): Update for
1970 location_get_source_line returning a char_span.
1971 (struct char_span): Move to input.h.
1972 (struct correction): Update for fields in char_span becoming
1974 (struct source_line): Update for location_get_source_line
1975 returning a char_span.
1976 (layout::print_line): Likewise.
1977 * edit-context.c (edited_file::print_content): Likewise.
1978 (edited_file::print_diff_hunk): Likewise.
1979 (edited_file::print_run_of_changed_lines): Likewise.
1980 (edited_file::get_num_lines): Likewise.
1981 (edited_line::edited_line): Likewise.
1982 * final.c (asm_show_source): Likewise.
1983 * input.c (location_get_source_line): Convert return type
1984 from const char * to char_span, losing the final "line_len"
1986 (dump_location_info): Update for the above.
1987 (get_substring_ranges_for_loc): Likewise. Use a char_span
1988 when handling the literal within the line.
1989 (test_reading_source_line): Update for location_get_source_line
1990 returning a char_span.
1991 * input.h (class char_span): Move here from
1992 diagnostic-show-locus.c, converting from a struct to a class.
1993 Make data members private.
1994 (char_span::operator bool): New.
1995 (char_span::length): New.
1996 (char_span::get_buffer): New.
1997 (char_span::operator[]): New.
1998 (char_span::subspan): Make const.
1999 (char_span::xstrdup): New.
2000 (location_get_source_line): Convert return type from const char *
2001 to char_span, losing the final "line_size" param.
2003 2018-04-30 Jan Hubicka <jh@suse.cz>
2005 * lto-wrapper.c (ltrans_priorities): New static var.
2006 (cmp_priority): New.
2007 (run_gcc): Read priorities and if doing parallel build order
2008 the Makefile by them.
2010 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2012 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2014 2018-04-30 Richard Biener <rguenther@suse.de>
2016 * tree-cfg.c (verify_address): Remove base argument, add
2017 flag whether to check TREE_ADDRESSABLE and do that.
2018 (verify_expr): Remove.
2019 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2020 (verify_gimple_assign_single): Likewise.
2021 (verify_gimple_switch): Likewise.
2022 (verify_expr_location_1): Dereference tp once. Add (disabled)
2023 piece from verify_expr.
2024 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2026 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2028 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2030 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2032 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2033 (small_data_pattern): Likewise.
2034 (arc_rewrite_small_data): Likewise.
2035 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2036 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2037 (get_symbol_alignment): New function.
2038 (legitimate_small_data_address_p): Likewise.
2039 (legitimate_scaled_address): Update, call
2040 legitimate_small_data_address_p.
2041 (output_sdata): New static variable.
2042 (arc_print_operand): Update how we handle small data operands.
2043 (arc_print_operand_address): Likewise.
2044 (arc_legitimate_address_p): Update, use
2045 legitimate_small_data_address_p.
2046 (arc_rewrite_small_data_p): Remove.
2047 (arc_rewrite_small_data_1): Likewise.
2048 (arc_rewrite_small_data): Likewise.
2049 (small_data_pattern): Likewise.
2050 (compact_sda_memory_operand): Update to use
2051 legitimate_small_data_address_p and get_symbol_alignment.
2052 (prepare_move_operands): Don't rewite sdata pattern.
2053 (prepare_extend_operands): Remove.
2054 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2056 (zero_extendqisi2): Likewise.
2057 (zero_extendhisi2): Likewise.
2058 (extendqihi2): Likewise.
2059 (extendqisi2): Likewise.
2060 (extendhisi2): Likewise.
2064 * config/arc/constraints.md (Usd): Change it to memory constraint.
2066 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2068 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2069 as source of std instructions.
2070 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2071 allow 6-bit constants as source for store instructions.
2072 (movdi_insn): Update instruction pattern to allow 6-bit constants
2073 as source for store instructions.
2075 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2077 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2079 2018-04-30 Nathan Sidwell <nathan@acm.org>
2080 Sandra Loosemore <sandra@codesourcery.com>
2082 * dumpfile.c (dump_open): Allow '-' for stdout.
2083 * doc/invoke.texi (Developer Options): Document dump filename
2084 determination early. Document stdin/stdout selection.
2086 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2088 Microblaze Target: PIC data text relative
2090 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2091 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2093 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2094 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2095 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2096 New addressing mode for data-text relative position indepenedent code.
2097 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2098 'ADDRESS_SYMBOLIC_TXT_REL'.
2099 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2100 (microblaze_legitimate_pic_operand): Exclude function calls from
2101 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2102 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2104 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2105 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2106 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2107 for 'address + offset'.
2108 (microblaze_expand_prologue): Add new function prologue call for
2110 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2111 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2112 table in case of TARGET_PIC_DATA_TEXT_REL.
2113 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2114 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2115 Add new macros 'UNSPEC_TEXT',
2116 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2117 + exclude function calls from 'UNSPEC_PLT' in case of data text
2119 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2120 new target hook for generating address diff vector tables in case of
2122 * doc/tm.texi : Regenerate.
2123 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2124 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2125 of addr diff vector generation.
2126 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2127 target hook definition.
2128 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2129 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2130 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2131 option for data text relative.
2133 2018-04-30 Richard Biener <rguenther@suse.de>
2135 * tree-chrec.h (evolution_function_is_constant_p): Remove
2137 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2139 2018-04-30 Richard Biener <rguenther@suse.de>
2142 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2144 2018-04-30 Richard Biener <rguenther@suse.de>
2146 PR tree-optimization/28364
2147 PR tree-optimization/85275
2148 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2149 copying first exit test.
2151 2018-04-28 Mark Wielaard <mark@klomp.org>
2153 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2155 (dwarf_AT): Handle DW_AT_addr_base.
2156 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2158 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2161 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2162 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2163 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2164 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2166 2018-04-28 Richard Biener <rguenther@suse.de>
2168 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2169 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2170 to reflect use. Only add interesting stmts.
2172 2018-04-27 Martin Jambor <mjambor@suse.cz>
2175 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2176 the jump function allows for passing through aggregate values.
2178 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2180 * input.h (in_system_header_at): Convert from macro to inline
2182 (from_macro_expansion_at): Likewise.
2183 (from_macro_definition_at): Likewise.
2185 2018-04-27 Jeff Law <law@redhat.com>
2187 * config.gcc: Mark tile* targets as deprecated/obsolete.
2189 2018-04-27 Richard Biener <rguenther@suse.de>
2191 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2194 2018-04-27 Richard Biener <rguenther@suse.de>
2196 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2198 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2200 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2201 with Yd constraint. Set "preferred_for_speed" attribute from
2202 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2204 (*movdi_internal): Ditto.
2205 (movti_interunit splitters): Remove
2206 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2207 (movdi_interunit splitters): Ditto.
2208 * config/i386/constraints.md (Ye): Remove.
2209 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2211 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2214 * config/aarch64/constraints.md (Usg): Limit to 31.
2217 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2219 PR tree-optimization/85529
2220 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2221 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2222 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2223 zero extension or masking of the MSB bit.
2224 (optimize_range_tests): Add FIRST_BB argument, pass it through
2225 to optimize_range_tests_var_bound.
2226 (maybe_optimize_range_tests, reassociate_bb): Adjust
2227 optimize_range_tests callers.
2229 2018-04-26 Richard Biener <rguenther@suse.de>
2230 Jakub Jelinek <jakub@redhat.com>
2232 * cgraph.h (symbol_table): Just declare debug method here.
2233 * symtab.c (symbol_table::debug): Define.
2235 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2237 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2239 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2241 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2242 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2243 (*movdi_internal): Substitute Yi and Yj constraint with x
2244 and Ym and Yn constraint with y constraint. Update "isa"
2245 attribute and set "preferred_for_speed" attribute from
2246 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2247 (*movsi_internal): Ditto.
2248 (*movdf_internal): Ditto.
2249 (*movsf_internal): Ditto.
2250 (*zero_extendsidi2): Ditto.
2251 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2252 (sse2_loadld): Ditto.
2253 (*vec_extract<ssevecmodelower>_0): Ditto.
2254 (*vec_extractv4si_0_zext_sse4): Ditto.
2255 (vec_concatv2di): Ditto.
2256 (*vec_dup<mode>): Ditto.
2257 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2258 * config/i386/constraints.md (Yi): Remove.
2263 2018-04-26 Nathan Sidwell <nathan@acm.org>
2265 * dumpfile.c (dump_open): New.
2266 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2267 (dump_finish): Detect stdio/stderr by value not name.
2269 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2271 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2273 2018-04-26 Tom de Vries <tom@codesourcery.com>
2276 * config/nvptx/nvptx.c (verify_neutering_jumps)
2277 (verify_neutering_labels): New function
2278 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2280 2018-04-26 Tom de Vries <tom@codesourcery.com>
2283 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2284 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2287 2018-04-26 Richard Biener <rguenther@suse.de>
2288 Tom de Vries <tom@codesourcery.com>
2291 * lto-streamer-out.c (output_function): Fixup loops if required to match
2292 discovery done in the reader.
2294 2018-04-26 Richard Biener <rguenther@suse.de>
2296 PR tree-optimization/85116
2297 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2298 have a loop exit from the single latch predecessor. Remove
2299 case of header with just condition.
2300 (ch_base::copy_headers): Exclude infinite loops from any
2302 (pass_ch::execute): Record exits.
2304 2018-04-26 Richard Biener <rguenther@suse.de>
2306 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2307 prologue cost vector and pass it to vect_get_load_cost.
2308 (vect_get_peeling_costs_all_drs): Likewise.
2309 (vect_peeling_hash_get_lowest_cost): Likewise.
2310 (vect_enhance_data_refs_alignment): Likewise.
2312 2018-04-26 Richard Biener <rguenther@suse.de>
2315 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2316 checking of integer<->pointer conversions.
2317 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2318 sign-/zero-extending pointer types.
2319 (expand_omp_for_static_chunk): Likewise.
2321 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2322 Jean Lee <xiaoyur347@gmail.com>
2324 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2325 (TARGET_ASAN_SHADOW_OFFSET): Define.
2326 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2327 true for -fsanitize=address.
2329 2018-04-25 Mark Wielaard <mark@klomp.org>
2331 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2334 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2336 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2337 than "alu", remove explicit "memory" and "imm_disp" attributes.
2338 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2341 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2342 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2343 gen_lowpart_no_emit.
2345 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2348 * config/i386/i386.c (ix86_expand_builtin): Change memory
2349 operand to XI, extend p0 to Pmode.
2350 * config/i386/i386.md: Change unspec volatile and operand
2351 1 mode to XI, change operand 0 mode to P.
2353 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2355 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2356 GET_MODE_MASK before any checking.
2357 (nds32_can_use_bset_p): Likewise.
2358 (nds32_can_use_btgl_p): Likewise.
2360 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2362 * config/nds32/nds32-doubleword.md: New define_split pattern for
2363 illegal register number.
2365 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2367 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2369 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2371 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2373 2018-04-25 Richard Biener <rguenther@suse.de>
2375 * lto-streamer.h (LTO_major_version): Bump to 8.
2377 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2379 * BASE-VER: Set to 9.0.0.
2381 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2383 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2384 in __abskf2 and __powikf2.
2386 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2389 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2390 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2391 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2392 Use the above on operand 2. Reindent.
2393 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2395 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2398 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2400 * config/i386/i386.opt (mcet): Removed.
2401 * doc/install.texi: Remove -mcet documentation.
2402 * doc/invoke.texi: Likewise.
2404 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2407 * doc/install.texi: Remove -mcet from bootstrap-cet.
2409 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2412 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2413 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2417 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2418 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2419 containing a CONST_VECTOR.
2421 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2423 * doc/install.texi: Update newlib dependency for nvptx.
2425 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2428 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2429 instead of INTVAL when shifting x left.
2431 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2433 PR tree-optimization/85478
2434 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2435 vect_grouped_store_supported for single element vectors.
2437 2018-04-24 Richard Biener <rguenther@suse.de>
2440 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2441 load cost increase to the case of non-constant step.
2443 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2446 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2447 destination if any_malformed_asm.
2449 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2452 * expr.c (store_field): In the bitfield case, if the value comes from
2453 a function call and is returned in registers by means of a PARALLEL,
2454 do not change the mode of the temporary unless BLKmode and VOIDmode.
2456 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2458 PR rtl-optimization/85423
2459 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2460 dependencies to debug insns when the previous insn is non-debug.
2462 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2464 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2465 enums into a single definition.
2466 (fls): Fix predicates and printing.
2469 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2471 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2472 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2473 and short u6 immediate.
2474 (check_if_valid_sleep_operand): Remove.
2475 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2477 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2479 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2480 flag_always_save_lp condition.
2481 * config/nds32/nds32.opt (malways-save-lp): New option.
2483 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2485 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2486 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2487 * config/nds32/nds32.h
2488 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2489 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2491 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2493 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2494 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2496 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2497 Chung-Ju Wu <jasonwucj@gmail.com>
2499 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2501 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2502 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2504 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2506 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2508 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2510 * config/nds32/nds32-protos.h (nds32_data_alignment,
2511 nds32_local_alignment): Declare.
2512 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2513 nds32_local_alignment): New functions.
2514 (TARGET_CONSTANT_ALIGNMENT): Define.
2515 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2517 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2519 * config/nds32/nds32.c
2520 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2521 (TARGET_MODES_TIEABLE_P): Likewise.
2523 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2525 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2528 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2529 Chung-Ju Wu <jasonwucj@gmail.com>
2531 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2532 for unaligned access.
2533 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2535 * config/nds32/nds32-intrinsic.md: Likewise.
2536 * config/nds32/nds32_intrinsic.h: Likewise.
2537 * config/nds32/nds32.h (nds32_builtins): Likewise.
2538 * config/nds32/nds32.opt (munaligned-access): New option.
2539 * config/nds32/nds32.c (nds32_asm_file_start): Display
2540 flag_unaligned_access status.
2542 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2544 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2545 -mno-relax is present.
2546 * config/riscv/linux.h (LINK_SPEC): Ditto.
2548 2018-04-20 Martin Sebor <msebor@redhat.com>
2551 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2553 (gimple_fold_builtin_stxcpy_chk): Same.
2554 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2556 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2559 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2560 __powikf2 when long double is IEEE 128-bit.
2562 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2564 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2565 step to make sure stack always aligned.
2567 2018-04-20 Carl Love <cel@us.ibm.com>
2570 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2571 size check for arg0.
2573 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2574 Tom de Vries <tom@codesourcery.com>
2577 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2578 Emit insns for calls too.
2579 (nvptx_find_par): Always look for worker-level predecessor insn.
2580 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2582 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2583 (nvptx_process_pars): Propagate frames for calls.
2585 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2588 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2590 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2591 (ix86_handle_option): Don't handle OPT_mibt.
2592 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2594 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2596 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2597 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2598 (ix86_target_macros): Define __CET__ with flag_cf_protection
2599 for -fcf-protection.
2600 * config/i386/i386.c (isa2_opts): Remove -mibt.
2601 * config/i386/i386.h (TARGET_IBT): Removed.
2602 (TARGET_IBT_P): Likewise.
2603 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2604 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2605 * config/i386/i386.opt (mcet): Update help message.
2608 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2609 -mcet as an alias for -mshstk.
2611 2018-04-20 Richard Biener <rguenther@suse.de>
2614 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2615 complexity by forcing a single use of the multiply operand.
2617 2018-04-20 Martin Jambor <mjambor@suse.cz>
2620 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2621 recursion dependency to only apply to non-clones.
2623 2018-04-20 Martin Jambor <mjambor@suse.cz>
2626 * ipa-cp.c (create_specialized_node): Check that clones of
2627 self-recursive edges exist during IPA-CP.
2629 2018-04-19 Toon Moene <toon@moene.org>
2631 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2634 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2636 PR tree-optimization/85467
2637 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2638 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2639 VECTOR_CST element to type.
2641 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2644 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2645 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2646 (builtin_longjmp): Likewise.
2647 (save_stack_nonlocal): New pattern.
2648 (restore_stack_nonlocal): Likewise.
2650 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2653 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2654 Replace ASM_OUTPUT_LABEL with fprintf.
2656 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2659 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2660 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2661 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2662 define __IBT__ and __SHSTK__ for -fcf-protection.
2663 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2665 (ix86_trampoline_init): Likewise.
2666 (x86_output_mi_thunk): Likewise.
2667 (ix86_notrack_prefixed_insn_p): Likewise.
2668 (ix86_option_override_internal): Don't disallow -fcf-protection.
2669 * config/i386/i386.md (rdssp<mode>): Also enable for
2671 (incssp<mode>): Likewise.
2672 (nop_endbr): Likewise.
2673 * config/i386/i386.opt (mcet): Change help message to built-in
2677 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2678 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2679 enable CET built-in functions.
2681 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2683 * common/config/i386/i386-common.c
2684 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2685 OPTION_MASK_ISA_MOVDIRI_UNSET,
2686 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2687 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2688 * config.gcc (movdirintrin.h): New header.
2689 * config/i386/cpuid.h (bit_MOVDIRI,
2690 bit_MOVDIR64B): New bits.
2691 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2693 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2694 (VOID, PVOID, PCVOID)): New function types.
2695 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2696 __builtin_ia32_directstoreu_u64,
2697 __builtin_ia32_movdir64b): New builtins.
2698 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2699 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2701 (ix86_valid_target_attribute_inner_p): Ditto.
2702 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2703 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2704 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2705 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2706 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2707 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2708 (movdiri<mode>, movdir64b_<mode>): New.
2709 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2710 * config/i386/immintrin.h: Include movdirintrin.h.
2711 * config/i386/movdirintrin.h: New file.
2712 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2714 2018-04-19 Richard Biener <rguenther@suse.de>
2717 * cfg.c (clear_bb_flags): When loop state says we have
2718 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2720 2018-04-19 Richard Biener <rguenther@suse.de>
2722 PR tree-optimization/84737
2723 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2724 copying restrict info.
2725 (vect_setup_realignment): Use it.
2726 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2727 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2728 the first DR to all generated stores.
2729 (vectorizable_load): Likewise for loads.
2731 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2733 PR tree-optimization/85446
2734 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2735 the integral and pointer types to have the same precision.
2737 * doc/install.texi: Document --disable-cet being the default and
2740 2018-04-18 Martin Liska <mliska@suse.cz>
2742 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2745 2018-04-18 Martin Liska <mliska@suse.cz>
2748 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2751 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2752 arguments if they are comparable.
2754 2018-04-18 Martin Liska <mliska@suse.cz>
2757 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2760 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2763 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2766 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2767 ENDBR after calling __morestack.
2769 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2772 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2773 by using gcc_base_ver to generate a gcc_driver_version, and use
2774 it when generating GCC_DRIVER_NAME.
2775 * configure: Regenerate.
2777 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2780 * config.gcc: Obsolete powerpc*-*-*spe*.
2782 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2785 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2786 (stabstr_D): Change type of unum from unsigned int to
2787 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2790 2018-04-17 Jim Wilson <jimw@sifive.com>
2793 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2794 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2795 Set arg_pointer_offset after using pretend_args_size.
2797 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2799 PR rtl-optimization/85431
2800 * dse.c (record_store): Ignore zero width stores.
2803 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2804 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2805 __builtin_stack_restore rather than after it.
2806 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2807 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2808 argument instead of virtual_dynamic_stack_rtx.
2810 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2812 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2814 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2815 Add note to error message to explain internal mapping of overloaded
2816 built-in function name to non-overloaded built-in function name.
2817 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2820 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2823 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2824 where the inputs overlap with the output.
2826 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2828 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2829 (=v, v) alternative and explicit "memory" attribute.
2830 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2831 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2833 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2834 "sselog1" type instead of "sselog".
2835 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2836 "sselog". Remove explicit "memory" attribute.
2837 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2838 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2840 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2841 "sselog1" type instead of "sselog". Remove explicit "memory"
2843 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2844 use "sselog1" type instead of "sselog". Remove explicit "memory"
2846 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2847 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2849 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2850 "sselog1" type instead of "sselog". Remove explicit "memory"
2852 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2853 use "sselog1" type instead of "sselog". Remove explicit "memory"
2857 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2860 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2863 2018-04-17 Martin Jambor <mjambor@suse.cz>
2866 * ipa-cp.c (create_specialized_node): Call
2867 expand_all_artificial_thunks if necessary.
2869 2018-04-17 Martin Liska <mliska@suse.cz>
2872 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2873 in message, remote space in between '_G' and '('.
2875 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2878 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2879 avx512f_vmcmp<mode>3<round_saeonly_name>,
2880 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2881 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2882 avx512f_rndscale<mode><round_saeonly_name>,
2883 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2884 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2885 Use %<iptr>2 instead of %2 for -masm=intel.
2886 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2887 avx512f_vcvttss2usi<round_saeonly_name>,
2888 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2890 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2891 avx512f_vcvttsd2usi<round_saeonly_name>,
2892 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2893 Use %q1 instead of %1 for -masm=intel.
2894 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2895 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2896 of %3 for -masm=intel.
2897 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2899 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2901 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2903 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2904 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2906 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2908 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2909 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2910 %g1 and one with %0 and %1.
2911 (avx512er_vmrcp28<mode><round_saeonly_name>,
2912 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2914 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2915 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2916 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2917 of %0 and %{%4%} for -masm=intel.
2918 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2919 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2920 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2921 order of %0 and %{%5%}%{z%} for -masm=intel.
2923 2018-04-17 Jan Hubicka <jh@suse.cz>
2926 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2928 2018-04-17 Martin Liska <mliska@suse.cz>
2931 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2932 for target_clone error message. Make default implementation
2933 clone to be a local declaration.
2934 (separate_attrs): Add new argument and check for an empty
2936 (expand_target_clones): Handle it.
2937 (ipa_target_clone): Make redirection just for target_clones
2940 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2941 Tom de Vries <tom@codesourcery.com>
2944 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2945 tiled basic blocks without omp continue statements.
2947 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2950 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2951 vec_extract expression as having side effects to make sure it gets
2954 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2957 * config/i386/i386.c (get_builtin_code_for_version): Check
2960 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2963 * gcc/config.gcc: Support "skylake".
2964 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2966 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2967 (processor_target_table): Add "skylake".
2968 (ix86_option_override_internal): Add "skylake".
2969 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2970 PROCESSOR_CANNONLAKE.
2971 (get_builtin_code_for_version): Fix priority for
2972 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2973 PROCESSOR_SKYLAKE-AVX512.
2974 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2975 (processor_type): Add PROCESSOR_SKYLAKE.
2977 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2978 Jason Merrill <jason@redhat.com>
2981 * convert.c (convert_to_integer_1): Use direct recursion for
2982 enumeral types and types with a precision less than the number
2983 of bits in their mode.
2985 2018-04-16 Julia Koval <julia.koval@intel.com>
2988 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2989 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2991 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2994 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2995 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2996 and -mno-direct-move.
2998 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3001 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3002 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3003 Ensure negative shifts result in {0}.
3005 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3007 PR rtl-optimization/79916
3008 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3009 regs (if any) to define how to gnerate SD moves when LRA is in
3012 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3014 PR rtl-optimization/85393
3015 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3016 * except.c (expand_dw2_landing_pad_for_region): Make static.
3017 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3018 a label and unconditional jump to old_bb, rather than
3019 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3022 PR rtl-optimization/85376
3023 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3024 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3025 instead of a specific value.
3027 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3028 Bin Cheng <bin.cheng@arm.com>
3030 PR tree-optimization/82965
3031 PR tree-optimization/83991
3032 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3033 by_profile_only parameter.
3034 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3035 information if the loop was predicted to iterate too many times.
3036 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3038 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3041 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3044 2018-04-13 Martin Liska <mliska@suse.cz>
3045 Jakub Jelinek <jakub@redhat.com>
3048 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3049 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3050 * builtins.c (expand_builtin_memory_copy_args): Use
3051 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3052 handle dest_addr == pc_rtx.
3054 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3057 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3058 asked to not generate direct moves.
3059 (fix_trunc<mode>si2_stfiwx): Similar.
3060 (fix_trunc<mode>si2_internal): Similar.
3062 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3065 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3066 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3067 lookup if dest in some wider mode is known to be const0_rtx and
3068 if so, record permanent equivalence for it to be ZERO_EXTEND of
3069 the narrower mode destination.
3071 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3073 * lto-streamer-out.c (output_function): Revert 259346.
3074 * omp-expand.c (expand_oacc_for): Likewise.
3076 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3078 PR rtl-optimization/85354
3079 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3080 * sel-sched.c (sel_global_init): ... here.
3082 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3085 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3086 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3087 mode for PE-COFF targets.
3088 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3089 (i386_pe_asm_lto_end): Likewise.
3090 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3091 (TARGET_ASM_LTO_END): Likewise.
3092 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3093 (i386_pe_asm_lto_start): New function.
3094 (i386_pe_asm_lto_end): Likewise.
3096 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3097 Richard Biener <rguenther@suse.de>
3100 * lto-streamer-out.c (output_function): Fix CFG loop state before
3102 * omp-expand.c (expand_oacc_for): Handle calls to internal
3103 functions like regular functions.
3105 2018-04-12 Richard Biener <rguenther@suse.de>
3108 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3109 for the early LTO debug to properly generate references to it
3110 during DIE emission. Do not re-use that for the skeleton for
3112 (dwarf2out_early_finish): Likewise.
3114 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3117 * config/i386/sse.md
3118 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3119 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3120 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3121 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3122 and output is a reg, avoid creating invalid lowpart subreg, but
3123 instead split into a 512-bit move. Don't split if not AVX512VL,
3124 input is xmm16+ reg and output is a mem.
3125 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3126 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3127 xmm16+ reg and output is a mem.
3129 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3131 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3132 also for flag_dwarf2_cfi_asm.
3134 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3136 PR rtl-optimization/85342
3137 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3138 a bool scalar var inside of the loop instead. Don't try to update
3139 recog_data.operand after failed apply_change_group.
3141 2018-04-12 Tom de Vries <tom@codesourcery.com>
3144 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3145 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3146 array with flexible array member as array without given dimension.
3147 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3148 argument for undefined param to true.
3150 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3153 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3154 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3155 from PowerPC section.
3156 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3157 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3158 help text that is too long.
3159 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3160 help text that is too long.
3161 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3162 help text that is too long.
3164 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3166 * config/alpha/alpha.md (stack_probe_internal): Rename
3167 from "probe_stack". Update all callers.
3169 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3171 PR rtl-optimization/84566
3172 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3173 sched_macro_fuse_insns.
3175 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3178 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3179 (compute_block_dependences): ... from here.
3181 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3183 PR tree-optimization/85331
3184 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3185 from int to HOST_WIDE_INT.
3187 2018-04-11 Martin Jambor <mjambor@suse.cz>
3190 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3191 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3192 not the same as the source val.
3193 (cgraph_edge_brings_value_p): New parameter.
3194 (gather_edges_for_value): Pass destination value to
3195 cgraph_edge_brings_value_p.
3196 (perhaps_add_new_callers): Likewise.
3197 (get_info_about_necessary_edges): Likewise and exclude values brought
3198 only by self-recursive edges.
3199 (create_specialized_node): Redirect only clones of self-calling edges.
3200 (+self_recursive_pass_through_p): New function.
3201 (find_more_scalar_values_for_callers_subset): Use it.
3202 (find_aggregate_values_for_callers_subset): Likewise.
3203 (known_aggs_to_agg_replacement_list): Removed.
3204 (decide_whether_version_node): Re-calculate known constants for all
3205 remaining context clones.
3207 2018-04-11 Richard Biener <rguenther@suse.de>
3210 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3211 from early DWARF output.
3212 (dwarf2out_early_finish): Output line info unconditionally into
3213 early DWARF and add reference to it.
3215 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3218 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3219 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3220 other than V2DFmode using iptr mode attribute.
3221 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3223 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3225 PR rtl-optimization/84659
3226 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3228 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3231 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3233 (output_loc_list): Pass address of a dummy size variable even in the
3234 locview handling loop.
3235 (index_location_lists): Add comment on why skip_loc_list_entry can't
3238 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3241 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3244 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3247 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3248 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3249 and -mstring-compare-inline-limit.
3251 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3254 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3255 for stack clash protection in a register whenever we need it to be in
3258 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3260 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3261 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3263 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3266 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3268 (mlong-double-): Ditto.
3269 * config/rs6000/sysv4.opt (msdata=): Ditto.
3270 (mtls-size=): Ditto.
3272 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3274 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3275 erroneous entries for
3276 "vector int vec_ldl (int, long int *)", and
3277 "vector unsigned int vec_ldl (int, unsigned long int *)".
3278 Add comments and entries for
3279 "vector bool char vec_ldl (int, bool char *)",
3280 "vector bool short vec_ldl (int, bool short *)",
3281 "vector bool int vec_ldl (int, bool int *)",
3282 "vector bool long long vec_ldl (int, bool long long *)",
3283 "vector pixel vec_ldl (int, pixel *)",
3284 "vector long long vec_ldl (int, long long *)",
3285 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3286 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3287 type tree bool_long_long_type_node and correct definition of
3288 bool_V2DI_type_node to make reference to this new type tree.
3289 (rs6000_mangle_type): Replace erroneous reference to
3290 bool_long_type_node with bool_long_long_type_node.
3291 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3292 comments to emphasize sign distinctions for char and int types and
3293 replace RS6000_BTI_bool_long constant with
3294 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3295 use of RS6000_BTI_pixel.
3296 (bool_long_type_node): Remove this macro definition.
3297 (bool_long_long_type_node): New macro definition
3299 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3301 PR rtl-optimization/85300
3302 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3303 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3304 simplify_unary_operation fails.
3306 2018-04-10 Martin Liska <mliska@suse.cz>
3308 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3309 cgraph_edge and ipa_ref.
3311 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3315 * config/i386/sse.md
3316 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3317 computation of the VEC_MERGE selector from mask.
3318 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3319 Fix decoding of the VEC_MERGE selector into mask.
3321 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3323 PR tree-optimization/85286
3324 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3326 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3328 * final.c (final_1): Set insn_last_address as well as
3329 insn_current_address.
3331 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3334 * explow.c (emit_stack_probe): Call validize_mem on memory location
3335 before passing it to gen_probe_stack. Create address operand and
3336 legitimize it for the probe_stack_address case.
3338 2018-04-09 Jan Hubicka <jh@suse.cz>
3341 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3342 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3343 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3345 (free_lang_data): Rebuild type inheritance graph.
3347 2018-04-09 Martin Sebor <msebor@redhat.com>
3349 * invoke.texi (-finline-small-functions): Mention other optimization
3351 (-findirect-inlining, -fpartial-inlining): Same.
3352 (-finline-functions-called-once): Same.
3353 (-freorder-blocks-and-partition): Same.
3355 2018-04-09 Jan Hubicka <jh@suse.cz>
3358 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3359 jumps; choose last target that matches the criteria (i.e.
3360 no partition changes for non-crossing jumps).
3361 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3362 support for redirecting crossing jumps to non-crossing.
3364 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3366 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3367 also for naked functions.
3369 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3371 * config/arc/arc.md (add_shift): New pattern.
3372 (add_shift2): Likewise.
3373 (sub_shift): Likewise.
3374 (sub_shift_cmp0_noout): Likewise.
3375 (compare_si_ashiftsi): Likewise.
3376 (xbfu_cmp0_noout): New combine pattern.
3377 (xbfu_cmp0"): Likewise.
3378 (movsi_set_cc_insn): Place the predicable variant first.
3379 (commutative_binary_cmp0_noout): Remove clobber.
3380 (commutative_binary_cmp0): New pattern.
3381 (noncommutative_binary_cmp0): Likewise.
3382 (noncommutative_binary_cmp0_noout): Likewise.
3383 (noncommutative_binary_comparison_result_used): Removed.
3384 (rsub_cmp0): New pattern.
3385 (rsub_cmp0_noout): Likewise.
3386 (extzvsi): Changed, keep only meaningful variants.
3387 (SQH, SEZ): New iterators.
3388 (SQH_postfix): New mode attribute.
3389 (SEZ_prefix): New code attribute.
3390 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3391 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3392 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3394 (noncommutative_operator): Check the availability of barrel
3397 2018-04-09 Richard Biener <rguenther@suse.de>
3399 PR tree-optimization/85284
3400 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3401 Only use the niter constraining form of simple_iv when the exit
3404 2018-04-09 Tom de Vries <tom@codesourcery.com>
3407 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3408 (define_expand "*memory_barrier"): New define_expand.
3409 (define_insn "memory_barrier"): New insn.
3411 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3413 PR rtl-optimization/80463
3414 PR rtl-optimization/83972
3415 PR rtl-optimization/83480
3417 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3418 correct producer for the insn.
3419 (tidy_control_flow): Fixup seqnos in case of debug insns.
3421 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3423 PR rtl-optimization/83913
3425 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3426 different sched-times when merging exprs.
3428 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3430 PR rtl-optimization/83962
3432 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3433 tidy_fallthru_edge and tidy_control_flow.
3435 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3437 PR rtl-optimization/83530
3439 * sel-sched.c (force_next_insn): New global variable.
3440 (remove_insn_for_debug): When force_next_insn is true, also leave only
3441 next insn in the ready list.
3442 (sel_sched_region): When the region wasn't scheduled, make another pass
3443 over it with force_next_insn set to 1.
3445 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3447 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3449 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3450 for interrupt control.
3451 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3452 functions for interrupt control.
3453 * config/nds32/nds32-intrinsic.md: Likewise.
3454 * config/nds32/nds32_intrinsic.h: Likewise.
3455 * config/nds32/nds32.h (nds32_builtins): Likewise.
3457 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3459 * config/nds32/nds32.c (nds32_init_machine_status,
3460 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3461 strict_aligned_p field.
3462 (nds32_expand_to_rtl_hook): New function.
3463 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3464 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3466 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3467 Chung-Ju Wu <jasonwucj@gmail.com>
3469 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3470 * config/nds32/nds32-n7.md: New file.
3471 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3472 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3474 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3475 * config/nds32/nds32.md (pipeline_model): Add n7.
3476 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3477 * config/nds32/pipelines.md: Include n7 settings.
3479 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3480 Chung-Ju Wu <jasonwucj@gmail.com>
3482 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3483 * config/nds32/nds32-e8.md: New file.
3484 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3485 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3487 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3488 * config/nds32/nds32.md (pipeline_model): Add e8.
3489 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3490 * config/nds32/pipelines.md: Include e8 settings.
3492 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3493 Chung-Ju Wu <jasonwucj@gmail.com>
3495 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3496 * config/nds32/nds32-n8.md: New file.
3497 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3498 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3500 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3501 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3502 * config/nds32/nds32.md (pipeline_model): Add n8.
3503 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3504 * config/nds32/pipelines.md: Include n8 settings.
3506 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3507 Chung-Ju Wu <jasonwucj@gmail.com>
3509 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3510 * config/nds32/nds32-n9-2r1w.md: New file.
3511 * config/nds32/nds32-n9-3r2w.md: New file.
3512 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3513 nds32_register_ports): New or modify for cpu n9.
3514 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3516 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3517 * config/nds32/nds32-utils.c: New file.
3518 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3519 TARGET_MUL_SLOW): Define.
3520 * config/nds32/nds32.md (pipeline_model): New attribute.
3521 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3522 New options that support cpu n9.
3523 * config/nds32/pipelines.md: Include n9 settings.
3524 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3526 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3528 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3529 information if necessary.
3530 (output_cond_branch_compare_zero): Likewise.
3531 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3532 (nds32_target_alignment): Refine for alignment.
3533 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3534 (FUNCTION_BOUNDARY): Modify.
3535 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3537 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3539 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3541 * config/nds32/constants.md (unspec_volatile_element): Add values for
3542 TLB operation and data prefetch.
3543 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3544 functions for TLB operation and data prefetch.
3545 * config/nds32/nds32-intrinsic.md: Likewise.
3546 * config/nds32/nds32_intrinsic.h: Likewise.
3547 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3548 (nds32_print_operand): Likewise.
3549 * config/nds32/nds32.h (nds32_builtins): Likewise.
3551 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3552 Andrew Pinski <pinsika@gcc.gnu.org>
3555 * match.pd: Use constant_boolean_node of correct type instead of
3556 boolean_true_node or boolean_false_node for simplifying
3557 pointer comparisons to zero.
3559 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3561 PR tree-optimization/80021
3562 * tree.c (verify_type_variant): Make error call in verify_variant_match
3563 translatable and remove final full stop.
3565 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3567 * config/nds32/constants.md (unspec_volatile_element): Add
3568 UNSPEC_VOLATILE_EH_RETURN.
3569 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3570 nds32_output_stack_pop): Support dwarf exception handling process.
3571 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3572 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3573 exception handling process.
3574 (nds32_compute_stack_frame): Likewise.
3575 (nds32_return_addr_rtx): Likewise.
3576 (nds32_initial_elimination_offset): Likewise.
3577 (nds32_expand_prologue): Likewise.
3578 (nds32_expand_epilogue): Likewise.
3579 (nds32_dynamic_chain_address): New function.
3580 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3582 (DYNAMIC_CHAIN_ADDRESS): Define.
3583 (EH_RETURN_DATA_REGNO): Define.
3584 (EH_RETURN_STACKADJ_RTX): Define.
3585 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3586 patterns for dwarf exception handling.
3588 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3590 * config/nds32/nds32.h: Clean up obsolete macros.
3592 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3594 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3595 Add enum values for particular instructions.
3596 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3597 particular intrinsic functions.
3598 * config/nds32/nds32-intrinsic.md: Likewise.
3599 * config/nds32/nds32_intrinsic.h: Likewise.
3600 * config/nds32/nds32.h (nds32_builtins): Likewise.
3601 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3602 (btst, ave): New patterns for particular instructions.
3604 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3606 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3607 Add enum values for atomic load/store and memory sync.
3608 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3610 * config/nds32/nds32-intrinsic.md: Likewise.
3611 * config/nds32/nds32_intrinsic.h: Likewise.
3612 * config/nds32/nds32.h (nds32_builtins): Likewise.
3614 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3616 PR tree-optimization/85257
3617 * fold-const.c (native_encode_vector): If not all elts could fit
3618 and off is -1, return 0 rather than offset.
3619 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3620 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3621 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3622 adjust buffer in native_interpret_expr call.
3624 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3626 * config/nds32/constants.md (unspec_volatile_element): Add cache
3627 control enum values.
3628 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3629 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3630 * config/nds32/nds32.c (nds32_cctl_names): New.
3631 (nds32_print_operand): Handle cache control register names.
3632 * config/nds32/nds32.h (nds32_builtins): New enum values.
3633 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3635 * config/nds32/nds32.md (type): Add mmu.
3636 * config/nds32/pipelines.md (simple_insn): Add mmu.
3638 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3640 * config/nds32/nds32.md (type): Remove call.
3641 * config/nds32/pipelines.md (simple_insn): Likewise.
3643 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3645 * config/nds32/constants.md (unspec_volatile_element): Add
3646 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3647 UNSPEC_VOLATILE_FMFCFG.
3648 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3649 description for fmfcfg and fmfcsr.
3650 (bdesc_1arg): Add fmtcsr.
3651 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3652 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3653 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3654 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3655 unspec_fmfcfg): New patterns.
3656 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3657 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3658 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3659 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3660 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3661 __nds32__fmfcfg): Define.
3663 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3665 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3666 intrinsic register names.
3667 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3668 intrinsic register enum values and macros.
3670 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3672 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3673 for load/store addressing form.
3674 (nds32_print_operand_address): Likewise.
3676 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3679 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3680 based on LABEL_REF. Remove useless assertion.
3681 (pic_address_needs_scratch): Fix formatting.
3682 (sparc_legitimize_pic_address): Minor tweaks.
3683 (sparc_delegitimize_address): Adjust assertion accordingly.
3684 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3685 into symbolic_operand.
3686 (movsi_high_pic_label_ref): Likewise.
3687 (movsi_lo_sum_pic_label_ref): Likewise.
3688 (movdi_pic_label_ref): Likewise.
3689 (movdi_high_pic_label_ref): Likewise.
3690 (movdi_lo_sum_pic_label_ref): Likewise.
3692 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3694 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3695 custom LIB_SPEC setup.
3697 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3698 Kito Cheng <kito.cheng@gmail.com>
3700 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3701 * config/riscv/freebsd.h: New.
3703 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3705 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3706 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3709 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3710 Kito Cheng <kito.cheng@gmail.com>
3712 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3713 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3714 * config/nds32/nds32-protos.h (nds32_output_call,
3715 nds32_output_return): Declare.
3716 * config/nds32/nds32.md: Refine all the call and return patterns.
3718 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3721 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3722 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3724 PR rtl-optimization/84872
3725 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3726 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3729 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3731 * expr.c (copy_blkmode_to_reg): Revert 254862.
3732 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3734 2018-04-06 Richard Biener <rguenther@suse.de>
3737 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3738 after seeing a component reference with an adjacent field. Treat
3739 refs to arrays at struct end of external decls similar to
3740 refs to unconstrained commons.
3742 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3745 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3746 look through SAVE_EXPRs with non-side-effects argument. Adjust
3748 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3751 2018-04-06 Richard Biener <rguenther@suse.de>
3754 * alias.c (find_base_term): New wrapper around find_base_term
3755 unwinding CSELIB_VAL_PTR changes.
3756 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3759 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3761 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3763 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3764 constant definitions.
3765 ("nop"): lr 0,0 -> nopr r0
3766 ("nop_lr0", "nop_lr1"): New insn definitions.
3768 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3770 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3771 NDS32_V3PUSH_AVAILABLE_P macro.
3773 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3774 Chung-Ju Wu <jasonwucj@gmail.com>
3776 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3777 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3778 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3779 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3780 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3781 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3782 * config/nds32/constraints.md: New constraints and checking for hard
3783 float configuration.
3784 * config/nds32/iterators.md: New mode iterator and attribute for hard
3785 float configuration.
3786 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3788 * config/nds32/nds32-fpu.md: New file.
3789 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3790 deal with hard float code generation.
3791 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3793 (abi_type, float_reg_number): New enum type.
3794 * config/nds32/nds32-predicates.c: New predicates for hard float.
3795 * config/nds32/nds32-protos.h: Declare functions for hard float.
3796 * config/nds32/nds32.c: Implementation for hard float configuration.
3797 * config/nds32/nds32.h: Definitions for hard float configuration.
3798 * config/nds32/nds32.md: Include hard float machine description and
3799 modify patterns for hard float configuration.
3800 * config/nds32/nds32.opt: New options for hard float configuration.
3801 * config/nds32/predicates.md: New predicates for hard float
3804 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3806 * common/config/nds32/nds32-common.c
3807 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3809 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3812 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3813 CONSTRUCTOR_ELT (ctor, ...)->value.
3815 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3818 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3820 2018-04-05 Tom de Vries <tom@codesourcery.com>
3823 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3826 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3827 Kito Cheng <kito.cheng@gmail.com>
3829 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3830 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3831 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3832 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3834 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3835 Kito Cheng <kito.cheng@gmail.com>
3837 * config/nds32/constraints.md (Ufe): New memory constraint.
3838 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3839 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3840 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3842 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3843 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3845 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3847 * config/nds32/nds32.md: Use optimize_size in the condition for
3848 alu-shift instructions.
3850 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3852 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3854 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3856 * config/nds32/nds32.md (negsi2): Refine pattern.
3858 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3859 Chung-Ju Wu <jasonwucj@gmail.com>
3861 * config/nds32/iterators.md (shift_rotate): New code iterator.
3862 (shift): New code attribute.
3863 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3864 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3865 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3866 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3867 bit-wise operations.
3868 (andsi3, *andsi3): Ditto.
3869 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3870 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3871 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3872 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3873 nds32_ior_operand, nds32_xor_operand): New predicates.
3875 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3877 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3878 (addsi3, subsi3): ... this.
3880 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3882 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3884 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3886 * config/nds32/nds32.md: Adjust indention.
3888 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3890 * config/nds32/nds32.md (feature): New attribute.
3892 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3894 * config/nds32/nds32.md (subtype): New attribute.
3896 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3899 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3900 expansion to perform a bitwise AND of the argument followed by a
3901 boolean negation of the result.
3903 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3905 PR rtl-optimization/84878
3906 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3907 the basic block. Assert the use reference is not artificial and that
3908 it has an associated insn.
3910 2018-04-04 Michael Matz <matz@suse.de>
3912 * builtins.c (compute_objsize): Pass correct operand
3913 to array_at_struct_end_p.
3915 2018-04-04 Richard Biener <rguenther@suse.de>
3918 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3919 from contexts for DINFO_LEVEL_TERSE and below.
3921 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3923 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3924 resiter_operand condition.
3925 * config/nds32/nds32.md (*move<mode>): Ditto.
3927 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3928 Monk Chiang <sh.chiang04@gmail.com>
3930 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3932 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3934 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3936 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3937 Kito Cheng <kito.cheng@gmail.com>
3939 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3940 nds32_cond_code_str, output_cond_branch,
3941 output_cond_branch_compare_zero, nds32_expand_cbranch,
3942 nds32_expand_cstore, nds32_expand_movcc,
3943 nds32_output_cbranchsi4_equality_zero,
3944 nds32_output_cbranchsi4_equality_reg,
3945 nds32_output_cbranchsi4_equality_reg_or_const_int,
3946 nds32_output_cbranchsi4_greater_less_zero: New functions.
3947 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3948 nds32_expand_cstore, nds32_expand_movcc,
3949 nds32_output_cbranchsi4_equality_zero,
3950 nds32_output_cbranchsi4_equality_reg,
3951 nds32_output_cbranchsi4_equality_reg_or_const_int,
3952 nds32_output_cbranchsi4_greater_less_zero): Declare.
3953 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3954 nds32_rimm11s_operand): New predicates.
3955 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3956 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3959 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3961 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3962 * config/nds32/nds32.md: Ditto.
3963 * config/nds32/pipelines.md: Ditto.
3965 2018-04-04 Richard Biener <rguenther@suse.de>
3967 PR tree-optimization/85168
3968 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3969 propagating abnormals.
3971 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3973 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3975 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3976 Kito Cheng <kito.cheng@gmail.com>
3978 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3979 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3980 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3981 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3982 * config/nds32/nds32.md (sibcall_internal): New.
3983 (sibcall_register): Remove.
3984 (sibcall_immediate): Remove.
3985 (sibcall_value_internal): New.
3986 (sibcall_value_register): Remove.
3987 (sibcall_value_immediate): Remove.
3988 * config/nds32/predicates.md (nds32_general_register_operand): New.
3989 (nds32_call_address_operand): New.
3991 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3993 PR rtl-optimization/85167
3994 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3995 bb_defs if *split_p, instead preinitialize it to NULL.
3997 PR tree-optimization/85156
3998 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3999 evaluating the argument multiple times.
4001 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4003 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4005 (_mm_cvtpd_ps): Likewise.
4006 (_mm_cvttpd_epi32): Likewise.
4007 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4008 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4009 vector, pixel, and bool following altivec.h include.
4011 2018-04-03 Martin Sebor <msebor@redhat.com>
4013 * doc/extend.texi (Common Function Attributes): Clarify.
4014 (const attribute): Likewise.
4015 (pure attribute): Likewise.
4017 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4020 * config/i386/i386.c (ix86_expand_vector_set): Use
4021 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4023 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4025 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4026 instructions when changing rounding bits to preserve precision bits
4027 in the x87 control word.
4029 2018-04-03 Martin Liska <mliska@suse.cz>
4031 PR tree-optimization/82491
4032 * rtl.h (strip_offset_and_add): Replace += suboffset with
4033 poly_uint64 () + suboffset.
4035 2018-03-29 Martin Liska <mliska@suse.cz>
4036 Martin Jambor <mjambor@suse.cz>
4039 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4040 param_type is not an integral or pointer type.
4042 2018-04-03 Richard Biener <rguenther@suse.de>
4044 * sese.h (recompute_all_dominators): Remove.
4046 2018-04-02 Martin Sebor <msebor@redhat.com>
4048 * doc/invoke.texi (-Wrestrict): Fix typos.
4050 2018-04-02 Jim Wilson <jimw@sifive.com>
4052 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4053 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4054 (<optab>di3, <optab>si3_extend): Likewise.
4055 (<optab>si3_mask, <optab>si3_mask_1): New.
4056 (<optab>di3_mask, <optab>di3_mask_1): New.
4057 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4058 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4059 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4061 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4063 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4066 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4068 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4069 (nds32_canonicalize_comparison): New function.
4071 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4072 Kito Cheng <kito.cheng@gmail.com>
4073 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4075 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4076 * config/nds32/constants.md (unspec_volatile_element): Add
4077 UNSPEC_VOLATILE_RELAX_GROUP.
4078 * config/nds32/nds32-relax-opt.c: New file.
4079 * config/nds32/nds32-predicates.c
4080 (nds32_symbol_load_store_p): New function.
4081 * config/nds32/nds32-protos.h
4082 (nds32_symbol_load_store_p): Declare function.
4083 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4084 * config/nds32/nds32.c
4085 (nds32_register_pass): New function to register pass.
4086 (nds32_register_passes): New function to register passes.
4087 * config/nds32/nds32.md (relax_group): New pattern.
4088 * config/nds32/nds32.opt (mrelax-hint): New option.
4089 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4091 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4093 * config/nds32/t-nds32: Modify files dependency.
4095 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4097 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4098 (PROFILE_HOOK): Define its implementation.
4100 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4102 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4103 type and 32-bit size.
4105 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4108 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4109 (V_128_256): New mode iterator.
4110 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4111 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4112 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4114 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4115 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4117 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4120 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4121 NaN inputs correctly.
4123 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4126 * config/rs6000/vsx.md (??r): New mode attribute.
4127 (*vsx_mov<mode>_64bit): Use it.
4128 (*vsx_mov<mode>_32bit): Likewise.
4130 2018-03-30 Martin Sebor <msebor@redhat.com>
4132 PR tree-optimization/84818
4133 * builtins.c (check_access): Use warning_n.
4135 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4138 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4140 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4143 2018-03-30 Julia Koval <julia.koval@intel.com>
4146 * x86-tune.def (movx, partial_reg_dependency): Enable for
4149 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4152 * lra-constraints.c (process_alt_operands): Move setting
4153 this_alternative_matches below.
4155 2018-03-29 Martin Liska <mliska@suse.cz>
4158 * doc/invoke.texi: Document how LTO works with debug info.
4159 Describe auto-load support of binutils. Mention 'x86-64'
4160 as valid option value of -march option.
4162 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4164 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4167 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4168 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4169 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4172 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4175 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4176 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4177 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4178 for RS6000_BTM_POWERPC64.
4179 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4180 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4181 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4186 2018-03-28 Carl Love <cel@us.ibm.com>
4189 2017-09-27 Carl Love <cel@us.ibm.com>
4191 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4192 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4193 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4196 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4198 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4199 instead of __vector bool.
4200 (_mm_max_pu8): Likewise.
4201 (_mm_min_pi16): Likewise.
4203 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4206 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4207 (DIVWEUO): Likewise.
4209 (DIVDEUO): Likewise.
4210 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4211 DIVWEUO and DIVDEUO.
4212 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4213 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4214 (div_extend): Likewise.
4215 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4217 (__builtin_divweuo): Likewise.
4218 (__builtin_divdeo): Likewise.
4219 (__builtin_divdeuo): Likewise.
4221 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4224 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4225 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4227 PR tree-optimization/82004
4228 * gimple-match-head.c (optimize_pow_to_exp): New function.
4229 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4230 Don't fold to exp if optimize_pow_to_exp is false.
4232 2018-03-28 Martin Liska <mliska@suse.cz>
4235 * calls.c (initialize_argument_information): Fix trailing space.
4236 * common.opt: Fix typo and provide better explanation for
4237 -fsanitize-coverage option.
4238 * config/i386/i386.opt: Fix typo.
4240 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4241 Martin Liska <mliska@suse.cz>
4244 * gimplify.c (asan_poison_variable): Don't do the check for
4245 gimplify_omp_ctxp here.
4246 (gimplify_decl_expr): Do it here.
4247 (gimplify_target_expr): Likewise.
4249 2018-03-28 Martin Liska <mliska@suse.cz>
4252 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4253 chkp_type_bounds_count if MPX is not enabled.
4255 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4257 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4259 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4262 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4263 function to create the function decl for complex long double
4264 multiply and divide for -mabi=ieeelongdouble.
4265 (init_float128_ieee): Call it.
4267 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4270 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4271 -fcf-protection=branch -mibt.
4272 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4274 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4277 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4279 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4282 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4283 extern array declarations.
4285 2018-03-27 Richard Biener <rguenther@suse.de>
4288 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4289 explicit single_use checks.
4291 2018-03-27 Richard Biener <rguenther@suse.de>
4293 PR tree-optimization/85082
4294 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4297 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4299 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4300 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4301 Turn on fasynchronous-unwind-tables and funwind-tables.
4303 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4306 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4307 (*bmi_blsr_<mode>_ccz): Ditto.
4309 2018-03-26 Tom de Vries <tom@codesourcery.com>
4311 PR tree-optimization/85063
4312 * omp-general.c (offloading_function_p): New function. Factor out
4314 * omp-offload.c (pass_omp_target_link::gate): ... here.
4315 * omp-general.h (offloading_function_p): Declare.
4316 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4317 with attribute omp declare target for offloading functions.
4319 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4321 PR tree-optimization/84005
4322 * tree-data-ref.h (get_base_for_alignment): Declare.
4323 * tree-data-ref.c (get_base_for_alignment_1): New function.
4324 (get_base_for_alignment): Likewise.
4325 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4326 get_base_for_alignment to find a suitable base object, instead
4327 of always using drb->base_address.
4329 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4332 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4333 known size by default.
4335 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4338 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4339 and non BLKmode operands.
4341 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4344 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4345 Clean up attributes.
4347 2018-03-23 Richard Biener <rguenther@suse.de>
4350 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4351 we are going to emit early debug for LTO.
4353 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4356 * function.c (match_asm_constraints_1): Don't optimize if input
4357 doesn't satisfy general_operand predicate for output's mode.
4360 * alias.c (write_dependence_p): Don't require for x_canonicalized
4361 non-VOIDmode if x has VOIDmode.
4364 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4365 just don't try to optimize it rather than assert it never happens.
4367 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4369 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4370 macro expansions for definition of ST_INTERNAL_<mode> and
4371 LD_INTERNAL_<mode> builtins.
4372 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4374 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4376 (altivec_expand_st_builtin): Likewise.
4377 (altivec_expand_builtin): Remove calls to deleted functions.
4378 (rs6000_address_for_altivec): Delete this function.
4379 * config/rs6000/vector.md: Remove expands for
4380 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4382 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4385 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4386 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4387 re-computing once computed.
4388 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4389 (arm_init_machine_status): Initialize
4390 machine->static_chain_stack_bytes.
4392 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4395 * doc/extend.texi: Add four new prototypes for vec_ld.
4396 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4397 definitions for more logical presentation.
4398 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4399 entries for V1TI variants of __builtin_altivec_ld builtin.
4400 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4401 handling of V1TI variant of LVX icode pattern.
4402 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4403 (rs6000_gimple_fold_builtin): Likewise.
4404 (altivec_init_builtins): Add code to define
4405 __builtin_altivec_lvx_v1ti function.
4407 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4410 * function.c (match_asm_constraints_1): Don't do the optimization
4411 if input isn't a REG, SUBREG, MEM or constant.
4413 2018-03-22 Tom de Vries <tom@codesourcery.com>
4415 PR tree-optimization/84956
4416 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4417 bb_has_abnormal_pred.
4419 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4422 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4423 DECL_INITIAL (decl) to decl at the end.
4424 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4427 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4429 * doc/extend.texi (__builtin_tgmath): Document when complex
4430 integer types are treated as _Complex _Float64.
4432 2018-03-21 Tom de Vries <tom@codesourcery.com>
4434 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4436 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4438 PR tree-optimization/84960
4439 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4440 if it is ENTRY block, move them into single succ of ENTRY in that case.
4442 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4444 PR tree-optimization/84811
4445 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4446 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4447 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4448 (known_subrange_p): Update accordingly. Cast each value involved
4449 in the size comparison, rather than casting the result of the
4452 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4454 PR tree-optimization/84982
4455 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4456 by flipping the least significant bit rather than all bits from
4457 bitpos to bitpos + bitsize - 1.
4459 2018-03-21 Nathan Sidwell <nathan@acm.org>
4461 * doc/extend.texi (Deprecated Features): Remove mention of
4462 long-deleted deprecations.
4464 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4467 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4468 * configure: Regenerate.
4470 2018-03-21 Tom de Vries <tom@codesourcery.com>
4472 PR tree-optimization/83126
4473 * tree-parloops.c (num_phis): New function.
4474 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4476 2018-03-21 Nathan Sidwell <nathan@acm.org>
4478 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4479 mention anon-struct/union members and trailing attributes.
4481 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4483 PR tree-optimization/84969
4484 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4485 builtin memset partitions if they set different rhs values.
4487 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4489 PR rtl-optimization/84989
4490 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4491 VEC_DUPLICATE with scalar result mode.
4493 2018-03-21 Martin Liska <mliska@suse.cz>
4496 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4497 not intended return statement.
4499 2018-03-21 Martin Liska <mliska@suse.cz>
4502 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4503 (chkp_find_bound_slots_1): Limit number of iterations.
4505 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4508 * Minor grammar fixes for x86 options.
4510 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4513 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4514 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4516 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4519 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4520 (altivec_lvx_<mode>_1op): Likewise.
4521 (altivec_stvx_<mode>_2op): Likewise.
4522 (altivec_stvx_<mode>_1op): Likewise.
4523 (altivec_lvx_<VM2:mode>): New define_expand.
4524 (altivec_stvx_<VM2:mode>): Likewise.
4525 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4526 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4527 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4528 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4529 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4530 (rs6000_gen_lvx): Likewise.
4531 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4532 (altivec_expand_stv_builtin): Likewise.
4533 (altivec_expand_builtin): Likewise.
4534 * config/rs6000/vector.md: Likewise.
4536 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4539 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4542 2018-03-20 Richard Biener <rguenther@suse.de>
4545 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4546 sign-conversions as zero, fall back to standard scalar_stmt
4549 2018-03-20 Martin Liska <mliska@suse.cz>
4552 * predict.c (rebuild_frequencies): Handle case when we have
4553 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4555 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4558 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4559 flag_section_anchors.
4560 * varasm.c (use_blocks_for_decl_p): Remove hack for
4561 dw2_force_const_mem.
4564 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4566 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4567 be created, use lowpart_subreg of operands[0] rather than operands[0]
4569 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4570 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4571 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4572 and n constraint instead of aarch64_shift_imm_di and Usd.
4573 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4574 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4576 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4579 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4580 to favor GPR over NEON registers.
4581 (<shift>di3_neon): Likewise.
4583 2018-03-20 Tom de Vries <tom@codesourcery.com>
4586 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4587 (nvptx_process_pars): Emit bar.sync asap and alap.
4589 2018-03-20 Tom de Vries <tom@codesourcery.com>
4592 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4593 seen_label if seen_label is already set.
4595 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4598 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4599 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4600 Use 1U instead of 1. Formatting fixes.
4603 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4604 instead of TREE_TYPE (s1) for the return value.
4606 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4608 PR tree-optimization/84946
4609 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4610 bitsize + bitsize in poly_uint64 rather than poly_int64.
4613 * dwarf2asm.c: Include fold-const.c.
4614 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4615 of decl rather than decl itself.
4617 PR rtl-optimization/84643
4618 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4620 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4623 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4624 calling assemble_variable.
4626 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4629 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4630 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4632 2018-03-19 Jim Wilson <jimw@sifive.com>
4635 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4636 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4637 (riscv_first_stack_step): Likewise.
4638 (riscv_option_override): Use STACK_BOUNDARY instead of
4640 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4642 (BIGGEST_ALIGNMENT): Set to 128.
4643 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4644 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4647 2018-03-19 Richard Biener <rguenther@suse.de>
4649 PR tree-optimization/84933
4650 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4651 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4653 2018-03-19 Richard Biener <rguenther@suse.de>
4655 PR tree-optimization/84859
4656 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4657 (cond_if_else_store_replacement): Perform sinking operation on
4658 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4659 Generalize what a BB with a single eligible store is.
4661 2018-03-19 Richard Biener <rguenther@suse.de>
4663 PR tree-optimization/84929
4664 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4665 chrec_is_positive against non-chrec arg.
4667 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4670 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4672 2018-03-18 Martin Liska <mliska@suse.cz>
4674 PR rtl-optimization/84635
4675 * regrename.c (build_def_use): Use matches_mode only when
4678 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4680 PR tree-optimization/84913
4681 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4682 vectorize chains of COND_EXPRs.
4684 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4686 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4688 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4690 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4692 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4694 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4696 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4697 Kito Cheng <kito.cheng@gmail.com>
4699 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4700 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4701 (nds32_adjust_reg_alloc_order): New function.
4702 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4704 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4706 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4707 nds32_print_operand, nds32_print_operand_address): Use
4708 HOST_WIDE_INT_PRINT_DEC instead.
4710 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4712 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4714 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4717 * config/i386/i386.c (initial_ix86_tune_features,
4718 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4720 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4721 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4722 rather than 1u << ix86_tune. Formatting fix.
4723 (ix86_option_override_internal): Change ix86_arch_mask from
4724 unsigned int to unsigned HOST_WIDE_INT, initialize to
4725 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4726 (ix86_function_specific_restore): Likewise.
4728 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4731 * postreload.c (reload_combine_recognize_pattern): Perform
4732 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4733 truncate_int_for_mode the result for the destination's mode.
4736 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4737 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4739 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4742 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4745 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4746 regno_allocno_class_array and sorted_pseudos.
4747 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4748 insns where regno is used.
4750 2018-03-16 Martin Liska <mliska@suse.cz>
4753 * multiple_target.c (create_dispatcher_calls): Redirect
4754 reference in the symbol table.
4756 2018-03-16 Martin Liska <mliska@suse.cz>
4759 * multiple_target.c (create_dispatcher_calls): Redirect also
4762 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4766 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4767 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4768 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4770 2018-03-16 Julia Koval <julia.koval@intel.com>
4772 * doc/invoke.texi (Skylake Server): Add CLWB.
4773 Cannonlake): Remove CLWB.
4775 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4777 PR tree-optimization/84841
4778 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4780 (FLOAT_ONE_CONST_TYPE): Define.
4781 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4782 (sort_by_operand_rank): Put entries with higher constant_type last
4783 rather than first to match comments.
4785 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4787 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4790 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4793 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4794 check and use address of target always.
4796 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4799 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4800 (indirect_thunk_bnd_needed): Likewise.
4801 (indirect_thunks_used): Likewise.
4802 (indirect_thunks_bnd_used): Likewise.
4803 (indirect_return_needed): New.
4804 (indirect_return_bnd_needed): Likewise.
4805 (output_indirect_thunk_function): Add a bool argument for
4807 (output_indirect_thunk_function): Don't generate alias for
4808 function return thunk.
4809 (ix86_code_end): Call output_indirect_thunk_function to generate
4810 function return thunks.
4811 (ix86_output_function_return): Set indirect_return_bnd_needed
4812 and indirect_return_needed instead of indirect_thunk_bnd_needed
4813 and indirect_thunk_needed.
4815 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4817 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4818 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4819 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4821 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4822 Paul Hua <paul.hua.gm@gmail.com>
4825 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4827 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4829 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4830 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4833 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4836 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4837 instead of GET_MODE_SIZE when comparing Units.
4839 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4842 * varasm.c (hash_section): Return an unchangeble hash value
4843 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4844 Return !aarch64_can_use_per_function_literal_pools_p ().
4846 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4849 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4850 rather than address of cmode as last argument to prepare_cmp_insn.
4852 2018-03-15 Julia Koval <julia.koval@intel.com>
4854 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4855 F_AVX512VNNI, F_AVX512BITALG): New.
4857 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4860 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4861 insn for floating-point loads and stores.
4863 2018-03-14 Carl Love <cel@us.ibm.com>
4865 * config/rs6000/rs6000-c.c: Add macro definitions for
4866 ALTIVEC_BUILTIN_VEC_PERMXOR.
4867 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4868 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4869 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4870 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4872 * config/doc/extend.texi: Add prototypes for vec_permxor.
4874 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4877 * diagnostic-show-locus.c (class layout_point): Convert m_line
4878 from int to linenum_type.
4879 (line_span::comparator): Use linenum "compare" function when
4880 comparing line numbers.
4881 (test_line_span): New function.
4882 (layout_range::contains_point): Convert param "row" from int to
4884 (layout_range::intersects_line_p): Likewise.
4885 (layout::will_show_line_p): Likewise.
4886 (layout::print_source_line): Likewise.
4887 (layout::should_print_annotation_line_p): Likewise.
4888 (layout::print_annotation_line): Likewise.
4889 (layout::print_leading_fixits): Likewise.
4890 (layout::annotation_line_showed_range_p): Likewise.
4891 (struct line_corrections): Likewise for field m_row.
4892 (line_corrections::line_corrections): Likewise for param "row".
4893 (layout::print_trailing_fixits): Likewise.
4894 (layout::get_state_at_point): Likewise.
4895 (layout::get_x_bound_for_row): Likewise.
4896 (layout::print_line): Likewise.
4897 (diagnostic_show_locus): Likewise for locals "last_line" and
4899 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4900 * input.c (selftest::test_linenum_comparisons): New function.
4901 (selftest::input_c_tests): Call it.
4902 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4903 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4904 * selftest.h (ASSERT_GT): New macro.
4905 (ASSERT_GT_AT): New macro.
4906 (ASSERT_LT): New macro.
4907 (ASSERT_LT_AT): New macro.
4909 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4911 PR rtl-optimization/84780
4912 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4914 2018-03-14 Martin Liska <mliska@suse.cz>
4916 * tree.c (record_node_allocation_statistics): Use
4917 get_stats_node_kind.
4918 (get_stats_node_kind): New function extracted from
4919 record_node_allocation_statistics.
4920 (free_node): Use get_stats_node_kind.
4922 2018-03-14 Richard Biener <rguenther@suse.de>
4924 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4925 that the value-set of ANTIC_IN doesn't grow.
4928 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4930 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4931 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4933 2018-03-14 Julia Koval <julia.koval@intel.com>
4935 * config.gcc (icelake-client, icelake-server): New.
4937 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4938 (initial_ix86_arch_features): Ditto.
4939 (PTA_SKYLAKE): Add SGX.
4940 (PTA_ICELAKE): Remove.
4941 (PTA_ICELAKE_CLIENT): New.
4942 (PTA_ICELAKE_SERVER): New.
4943 (ix86_option_override_internal): Split up icelake on icelake client and
4945 (get_builtin_code_for_version): Ditto.
4946 (fold_builtin_cpu): Ditto.
4947 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4948 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4949 * config/i386/i386.h (processor_type): Ditto.
4950 * doc/invoke.texi: Ditto.
4952 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4955 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4956 INTEGER_CST offset, add it together with bitpos / 8 and
4957 sign extend based on POINTER_SIZE.
4961 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4964 * config/i386/constraints.md (Yc): New register constraint.
4965 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4966 Use Yc constraint for alternative 2 of operand 0. Remove
4967 preferred_for_speed attribute.
4969 2018-03-14 Richard Biener <rguenther@suse.de>
4971 PR tree-optimization/84830
4972 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4973 with the old one to avoid oscillations.
4975 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4978 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4980 (assign_by_spills): Return a flag of reload assignment failure.
4981 Do not process the reload assignment failures. Do not spill other
4982 reload pseudos if they has the same reg class. Update n if
4984 (lra_assign): Add a return arg. Set up from the result of
4985 assign_by_spills call.
4986 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4987 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4988 usage_insns if it is not NULL.
4989 (spill_hard_reg_in_range): New function.
4990 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4991 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4992 function prototypes.
4993 (lra_assign): Change prototype.
4994 * lra.c (lra): Add code to deal with fails by splitting hard reg
4997 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4999 * config/riscv/riscv.opt (mrelax): New option.
5000 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5001 "norelax" when riscv_mrelax is disabled.
5002 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5004 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5007 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5008 reassociation for int modes.
5010 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5012 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5013 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5015 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5016 * config/aarch64/aarch64-sve.md
5017 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5018 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5019 (*extend<mode><Vwide>2): Rename to...
5020 (aarch64_sve_extend<mode><Vwide>2): ...this.
5021 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5022 renaming the old pattern to...
5023 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5025 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5026 define_expand, renaming the old pattern to...
5027 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5028 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5029 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5030 account when deciding which SVE instruction the optab should use.
5031 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5033 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5035 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5036 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5037 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5038 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5039 (tlsdesc_small_<mode>): Turn a define_expand and use
5040 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5041 (tlsdesc_small_advsimd_<mode>): ...this.
5042 (tlsdesc_small_sve_<mode>): New pattern.
5044 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5046 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5047 (UNSPEC_UMUL_HIGHPART): New constants.
5048 (MUL_HIGHPART): New int iteraor.
5049 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5050 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5052 (*<su>mul<mode>3_highpart): New define_insn.
5054 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5057 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5060 2018-03-13 Martin Liska <mliska@suse.cz>
5063 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5065 (sem_item_optimizer::~sem_item_optimizer): Release it.
5066 (sem_item_optimizer::merge_classes): Register variable aliases.
5067 (sem_item_optimizer::fixup_pt_set): New function.
5068 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5069 * ipa-icf.h: Declare new variables and functions.
5071 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5074 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5075 integer_pow2p@2 and test integer_pow2p in condition.
5076 (A < 0 ? C : 0): Similarly for @1.
5079 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5080 characters starting at p contain '\0' character, don't look beyond
5084 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5085 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5088 * reg-stack.c (change_stack): Change update_end var from int to
5089 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5090 also call set_block_for_insn on the newly added insns and rescan.
5093 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5094 on the last operand.
5097 * tree.c (stabilize_reference_1): Return save_expr (e) for
5098 STATEMENT_LIST even if it doesn't have side-effects.
5100 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5102 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5104 2018-03-12 Renlin Li <renlin.li@arm.com>
5106 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5107 aarch64_output_scalar_simd_mov_immediate.
5109 2018-03-12 Martin Sebor <msebor@redhat.com>
5111 PR tree-optimization/83456
5112 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5113 for perfectly overlapping calls to memcpy.
5114 (gimple_fold_builtin_memory_chk): Same.
5115 (gimple_fold_builtin_strcpy): Handle no-warning.
5116 (gimple_fold_builtin_stxcpy_chk): Same.
5117 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5119 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5121 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5122 parameter. Use it for SFmode.
5123 (rs6000_function_arg_advance_1): Adjust.
5124 (rs6000_function_arg): Adjust.
5125 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5127 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5129 PR rtl-optimization/84169
5130 PR rtl-optimization/84780
5131 * combine.c (can_combine_p): Check for a 2-insn combination whether
5132 the destination register is used between the two insns, too.
5134 2018-03-12 Richard Biener <rguenther@suse.de>
5136 PR tree-optimization/84803
5137 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5138 for refs DR analysis didn't process.
5140 2018-03-12 Richard Biener <rguenther@suse.de>
5142 PR tree-optimization/84777
5143 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5144 force-vectorize loops ignore whether we are optimizing for size.
5146 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5148 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5149 (TARGET_MD_ASM_ADJUST): Define.
5151 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5152 Kito Cheng <kito.cheng@gmail.com>
5153 Chung-Ju Wu <jasonwucj@gmail.com>
5155 * config/nds32/nds32.c (nds32_compute_stack_frame,
5156 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5157 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5158 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5159 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5160 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5161 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5162 * config/nds32/nds32.md (prologue, epilogue): Use macro
5163 NDS32_V3PUSH_AVAILABLE_P to do checking.
5165 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5168 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5169 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5170 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5171 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5172 addition of most attributes on !orig_type_die or the attribute not
5173 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5175 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5176 Chung-Ju Wu <jasonwucj@gmail.com>
5178 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5180 * config/nds32/nds32.opt (mvh): New option.
5182 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5183 Chung-Ju Wu <jasonwucj@gmail.com>
5185 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5187 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5188 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5191 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5192 Chung-Ju Wu <jasonwucj@gmail.com>
5194 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5196 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5197 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5199 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5200 Kito Cheng <kito.cheng@gmail.com>
5201 Chung-Ju Wu <jasonwucj@gmail.com>
5203 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5204 UNSPEC_FFMISM and UNSPEC_FLMISM.
5205 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5206 for ffb, ffmism and flmism.
5207 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5208 (unspec_ffmism): Ditto.
5209 (unspec_flmism): Ditto.
5210 (nds32_expand_builtin_impl): Check if string extension is available.
5211 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5212 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5214 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5217 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5220 * lra-assigns.c (assign_by_spills): Return a flag of reload
5221 assignment failure. Do not process the reload assignment
5222 failures. Do not spill other reload pseudos if they has the same
5224 (lra_assign): Add a return arg. Set up from the result of
5225 assign_by_spills call.
5226 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5227 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5228 usage_insns if it is not NULL.
5229 (spill_hard_reg_in_range): New function.
5230 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5231 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5232 function prototypes.
5233 (lra_assign): Change prototype.
5234 * lra.c (lra): Add code to deal with fails by splitting hard reg
5237 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5240 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5242 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5245 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5246 (dw_val_node): Add val_symbolic_view.
5247 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5248 (symview_upper_bound): New.
5249 (new_line_info_table): Initialize symviews_since_reset.
5250 (dwarf2out_source_line): Count symviews_since_reset and set
5251 symview_upper_bound.
5252 (dw_val_equal_p): Handle symview.
5253 (add_AT_symview): New.
5254 (print_dw_val): Handle symview.
5255 (attr_checksum, attr_checksum_ordered): Likewise.
5256 (same_dw_val_p, size_of_die): Likewise.
5257 (value_format, output_die): Likewise.
5258 (add_high_low_attributes): Use add_AT_symview for entry_view.
5259 (dwarf2out_finish): Reset symview_upper_bound, clear
5262 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5265 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5266 Add strict argument and use it.
5267 (rs6000_split_multireg_move): Update for new strict argument.
5268 (mem_operand_gpr): Disallow all non-offsettable addresses.
5269 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5271 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5274 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5275 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5276 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5279 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5280 decl, use remap_type if we want to use the type.
5282 2018-03-09 Martin Sebor <msebor@redhat.com>
5284 PR tree-optimization/84526
5285 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5287 (builtin_access::generic_overlap): Be prepared to handle non-array
5290 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5292 PR rtl-optimization/84682
5293 * lra-constraints.c (process_address_1): Check is_address flag
5294 for address constraints.
5295 (process_alt_operands): Likewise.
5296 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5297 preprocess_constraints.
5298 * recog.h (preprocess_constraints): Add oploc parameter.
5300 * recog.c (preprocess_constraints): Test address_operand for
5301 CT_ADDRESS constraints.
5303 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5306 * lra-assigns.c (assign_by_spills): Return a flag of reload
5307 assignment failure. Do not process the reload assignment
5308 failures. Do not spill other reload pseudos if they has the same
5310 (lra_assign): Add a return arg. Set up from the result of
5311 assign_by_spills call.
5312 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5313 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5314 usage_insns if it is not NULL.
5315 (spill_hard_reg_in_range): New function.
5316 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5317 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5318 function prototypes.
5319 (lra_assign): Change prototype.
5320 * lra.c (lra): Add code to deal with fails by splitting hard reg
5323 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5326 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5327 Accept complain bool parameter. Only emit errors if it is true.
5328 (arm_parse_cpu_option_name): Likewise.
5329 (arm_target_thumb_only): Adjust callers of the above.
5330 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5331 prototype to take a default true bool parameter.
5332 (arm_parse_arch_option_name): Likewise.
5334 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5335 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5339 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5340 * configure: Regenerate.
5341 * configure.ac ("linker --version-script option"): New.
5342 ("linker soname option"): New.
5344 2018-03-09 Richard Biener <rguenther@suse.de>
5346 PR tree-optimization/84775
5347 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5348 immediate uses of predicate stmts and mark them modified.
5351 PR tree-optimization/84178
5352 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5354 (version_loop_for_if_conversion): Delay update_ssa call.
5355 (tree_if_conversion): Delay update_ssa until after predicate
5358 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5361 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5362 when the function accesses prior frames.
5364 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5367 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5368 gen_llsym, otherwise call maybe_gen_llsym.
5371 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5372 has ',' character inside of it.
5374 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5377 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5378 as clobbering CC_REGNUM.
5380 2018-03-08 Richard Biener <rguenther@suse.de>
5383 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5384 (follow_copies_to_constant): Do not follow SSA names registered
5387 2018-03-08 Richard Biener <rguenther@suse.de>
5389 PR tree-optimization/84178
5390 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5392 (version_loop_for_if_conversion): Delay update_ssa call.
5393 (tree_if_conversion): Delay update_ssa until after predicate
5396 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5398 PR tree-optimization/84178
5399 * tree-if-conv.c (release_bb_predicate): Remove the
5400 the assertion that the stmts have NULL use_ops.
5401 Discard the statements, asserting that they haven't
5402 yet been added to a BB.
5404 2018-03-08 Richard Biener <rguenther@suse.de>
5406 PR tree-optimization/84746
5407 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5408 (phi_translate): Pass in destination ANTIC_OUT set.
5409 (phi_translate_1): Likewise. For a simplified result lookup
5410 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5411 (phi_translate_set): Adjust.
5412 (do_pre_regular_insertion): Likewise.
5413 (do_pre_partial_partial_insertion): Likewise.
5415 2018-03-08 Martin Liska <mliska@suse.cz>
5417 PR gcov-profile/84735
5418 * doc/gcov.texi: Document usage of profile files.
5419 * gcov-io.h: Document changes in the format.
5421 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5425 * dwarf2out.c (struct dw_line_info_table): Update comments for
5427 (FORCE_RESET_NEXT_VIEW): New.
5428 (FORCE_RESETTING_VIEW_P): New.
5429 (RESETTING_VIEW_P): Check for -1 too.
5430 (ZERO_VIEW_P): Likewise.
5431 (new_line_info_table): Force-reset next view.
5432 (dwarf2out_begin_function): Likewise.
5433 (dwarf2out_source_line): Simplify zero_view_p initialization.
5434 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5435 view directly. Omit view when omitting .loc at line 0.
5437 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5439 PR tree-optimization/84740
5440 * tree-switch-conversion.c (process_switch): Call build_constructors
5441 only if info.phi_count is non-zero.
5443 PR tree-optimization/84739
5444 * tree-tailcall.c (find_tail_calls): Check call arguments against
5445 DECL_ARGUMENTS (current_function_decl) rather than
5446 DECL_ARGUMENTS (func) when checking for tail recursion.
5448 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5450 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5451 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5452 Volker Reichelt's entry and add entries for people that perform
5453 GCC fuzzy testing and report numerous bugs.
5455 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5458 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5459 readonly data in sdata, if that is disabled.
5460 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5461 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5462 -mreadonly-in-sdata option.
5464 2018-03-07 Martin Sebor <msebor@redhat.com>
5466 PR tree-optimization/84468
5467 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5468 basic block when looking for nul assignment.
5470 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5473 * except.h (output_function_exception_table): Adjust prototype.
5474 * except.c (output_function_exception_table): Remove FNNAME parameter
5475 and add SECTION parameter. Ouput one part of the table at a time.
5476 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5477 the first part of the exception table and emit unwind directives.
5478 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5479 (i386_pe_seh_cold_init): Likewise.
5480 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5481 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5482 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5483 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5484 * config/i386/winnt.c: Include except.h.
5485 (struct seh_frame_state): Add reg_offset, after_prologue and
5486 in_cold_section fields.
5487 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5488 (i386_pe_seh_cold_init): New function.
5489 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5490 to seh->in_cold_section.
5491 (seh_emit_push): Record the offset of the push.
5492 (seh_emit_save): Record the offet of the save.
5493 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5494 Test seh->after_prologue to disregard the epilogue.
5495 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5496 (i386_pe_end_cold_function): New function.
5498 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5501 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5502 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5505 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5507 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5508 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5511 * multiple_target.c: Include tree-inline.h and intl.h.
5512 (expand_target_clones): Diagnose and fail if node->definition and
5513 !tree_versionable_function_p (node->decl).
5515 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5517 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5519 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5520 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5521 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5523 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5526 * combine.c (try_combine): Use reg_or_subregno instead of handling
5527 just paradoxical SUBREGs and REGs.
5529 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5531 * config/arc/arc.c (arc_finalize_pic): Remove function.
5532 (arc_must_save_register): We use single base PIC register, remove
5533 checks to save/restore the PIC register.
5534 (arc_expand_prologue): Likewise.
5535 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5537 (arc_verify_short): Likewise.
5538 (arc_attr_type): Likewise.
5539 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5540 (walk_stores): Likewise.
5541 (arc_address_cost): Make it static.
5542 (arc_verify_short): Likewise.
5543 (branch_dest): Likewise.
5544 (arc_attr_type): Likewise.
5545 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5546 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5547 (arc_final_prescan_insn): Remove inserting the nops due to
5548 hardware hazards. It is done in reorg step.
5549 (insn_length_variant_t): Remove.
5550 (insn_length_parameters_t): Likewise.
5551 (arc_insn_length_parameters): Likewise.
5552 (arc_get_insn_variants): Likewise.
5553 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5555 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5558 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5561 PR tree-optimization/84687
5562 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5564 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5566 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5568 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5569 Rename to ppc_speculation_barrier.
5570 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5571 __builtin_ppc_speculation_barrier.
5573 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5576 * combine.c (combine_simplify_rtx): Don't try to simplify if
5577 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5580 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5582 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5583 to 32 bytes when compiling for POWER9.
5585 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5588 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5589 regparm >= 3 with no arg reg available also for calls with
5590 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5593 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5595 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5597 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5600 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5602 2018-03-05 Richard Biener <rguenther@suse.de>
5604 PR tree-optimization/84486
5605 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5606 When inserting a __builtin_assume_aligned call set the LHS
5607 SSA name alignment info accordingly.
5609 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5611 PR tree-optimization/84114
5612 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5613 Avoid reassociation of FLOAT_MODE addition.
5615 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5617 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5618 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5619 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5620 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5621 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5622 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5623 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5625 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5626 __builtin_ia32_wbinvd): New builtins.
5627 (SPECIAL_ARGS2): New.
5628 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5629 (SPECIAL_ARGS2): New.
5630 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5631 (ix86_valid_target_attribute_inner_p): Ditto.
5632 (ix86_init_mmx_sse_builtins): Add special_args2.
5633 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5634 TARGET_WBNOINVD_P): New.
5635 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5636 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5637 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5638 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5639 * config/i386/pconfigintrin.h: New file.
5640 * config/i386/wbnoinvdintrin.h: Ditto.
5641 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5642 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5644 2018-03-05 Richard Biener <rguenther@suse.de>
5646 PR tree-optimization/84670
5647 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5649 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5650 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5651 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5652 doesn't grow if all successors (recursively) were visited at least
5655 2018-03-05 Richard Biener <rguenther@suse.de>
5657 PR tree-optimization/84650
5658 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5659 if executed in the loop pipeline.
5661 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5663 * doc/configfiles.texi (Configuration Files): Move info about
5664 conditionalizing $target-protos.h to...
5665 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5666 differs from $target-protos.h.
5668 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5669 Chung-Ju Wu <jasonwucj@gmail.com>
5671 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5672 * config/nds32/nds32-multiple.md (setmemsi): Define.
5673 * config/nds32/nds32-memory-manipulation.c
5674 (nds32_gen_dup_4_byte_to_word_value): New.
5675 (emit_setmem_word_loop): New.
5676 (emit_setmem_byte_loop): New.
5677 (nds32_expand_setmem_loop): New.
5678 (nds32_expand_setmem_loop_v3m): New.
5679 (nds32_expand_setmem_unroll): New.
5680 (nds32_expand_setmem): New.
5682 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5683 Chung-Ju Wu <jasonwucj@gmail.com>
5685 * config/nds32/nds32-memory-manipulation.c
5686 (nds32_emit_load_store): New.
5687 (nds32_emit_post_inc_load_store): New.
5688 (nds32_emit_mem_move): New.
5689 (nds32_emit_mem_move_block): New.
5690 (nds32_expand_movmemsi_loop_unknown_size): New.
5691 (nds32_expand_movmemsi_loop_known_size): New.
5692 (nds32_expand_movmemsi_loop): New.
5693 (nds32_expand_movmemsi_unroll): New.
5694 (nds32_expand_movmemqi): Rename ...
5695 (nds32_expand_movmemsi): ... to this.
5696 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5697 (movmemsi): ... to this.
5698 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5699 (nds32_expand_movmemsi): ... to this.
5701 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5702 Monk Chiang <sh.chiang04@gmail.com>
5703 Chung-Ju Wu <jasonwucj@gmail.com>
5705 * config/nds32/nds32-protos.h
5706 (nds32_expand_load_multiple): New arguments.
5707 (nds32_expand_store_multiple): Ditto.
5708 (nds32_valid_multiple_load_store): Rename ...
5709 (nds32_valid_multiple_load_store_p): ... to this.
5710 * config/nds32/nds32-memory-manipulation.c
5711 (nds32_expand_load_multiple): Refine implementation.
5712 (nds32_expand_store_multiple): Ditto.
5713 * config/nds32/nds32-multiple.md
5714 (load_multiple): Update nds32_expand_load_multiple interface.
5715 (store_multiple): Update nds32_expand_store_multiple interface.
5716 * config/nds32/nds32-predicates.c
5717 (nds32_valid_multiple_load_store): Rename ...
5718 (nds32_valid_multiple_load_store_p): ... to this and refine
5720 * config/nds32/predicates.md
5721 (nds32_load_multiple_and_update_address_operation): New predicate.
5722 (nds32_store_multiple_and_update_address_operation): New predicate.
5724 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5725 Chung-Ju Wu <jasonwucj@gmail.com>
5727 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5728 (combo): New attribute.
5729 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5731 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5733 * config/nds32/nds32.opt: Change -mcmodel= default value.
5735 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5736 Monk Chiang <sh.chiang04@gmail.com>
5737 Chung-Ju Wu <jasonwucj@gmail.com>
5739 * config/nds32/constants.md (unspec_element): New enum.
5740 * config/nds32/constraints.md (Umw): New constraint.
5741 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5742 * config/nds32/nds32-intrinsic.md: Likewise.
5743 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5744 (nds32_valid_smw_lwm_base_p): New.
5745 (nds32_output_smw_single_word): New.
5746 (nds32_output_lmw_single_word): New.
5747 (nds32_expand_unaligned_load): New.
5748 (nds32_expand_unaligned_store): New.
5749 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5750 (nds32_output_smw_single_word): Declare.
5751 (nds32_output_lmw_single_word): Declare.
5752 (nds32_expand_unaligned_load): Declare.
5753 (nds32_expand_unaligned_store): Declare.
5754 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5755 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5756 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5757 NDS32_BUILTIN_UASTORE_DW.
5758 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5761 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5762 Kito Cheng <kito.cheng@gmail.com>
5763 Chung-Ju Wu <jasonwucj@gmail.com>
5765 * config/nds32/nds32-intrinsic.c
5766 (nds32_expand_builtin_null_ftype_reg): Delete.
5767 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5768 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5769 (nds32_read_argument): New.
5770 (nds32_legitimize_target): Ditto.
5771 (nds32_legitimize_argument): Ditto.
5772 (nds32_check_constant_argument): Ditto.
5773 (nds32_expand_unop_builtin): Ditto.
5774 (nds32_expand_unopimm_builtin): Ditto.
5775 (nds32_expand_binop_builtin): Ditto.
5776 (nds32_builtin_decl_impl): Ditto.
5777 (builtin_description): Ditto.
5778 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5779 (nds32_init_builtins_impl): Ditto.
5780 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5781 (nds32_builtin_decl): New.
5782 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5783 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5785 2018-03-02 Jeff Law <law@redhat.com>
5787 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5788 (redundant_insn, fill_simple_delay_slots): Likewise.
5789 (fill_slots_from_thread): Likewise.
5790 * resource.c (mark_referenced_resources): Likewise.
5791 (mark_set_resources, find_dead_or_set_registers): Likewise.
5793 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5795 * substring-locations.h (format_warning_va): Formatting fix for
5797 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5799 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5800 * substring-locations.c: Include intl.h.
5801 (format_warning_va): Turned into small wrapper around
5802 format_warning_n_va, renamed to ...
5803 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5804 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5806 (format_warning_at_substring_n): New function.
5807 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5808 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5809 format_warning_at_substring with just a shorter name instead of
5810 const function pointer.
5811 (fmtwarn_n): New function.
5812 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5813 appropriate, get rid of all the fmtstr temporaries, move conditionals
5814 with G_() wrapped string literals directly into fmtwarn arguments,
5815 cast dir.len to (int), formatting fixes.
5817 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5819 * doc/invoke.texi: Remove "Cilk Plus" references.
5821 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5822 Richard Biener <rguenther@suse.de>
5825 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5826 for error or warning attributes if CALL_FROM_THUNK_P is set.
5829 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5832 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5833 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5836 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5837 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5839 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5840 instead of -1U in last predictors element's probability member.
5842 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5845 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5846 arguments if they are comparable.
5848 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5850 PR tree-optimization/84634
5851 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5852 masks and masked_loop_p with a single loop_masks, making sure it's
5853 null for bb vectorization.
5855 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5857 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5858 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5859 loop->force_vectorize to check whether there is no alias.
5861 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5864 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5866 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5868 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5869 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5870 instead of a loop around prev_real_insn.
5871 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5875 * config/i386/i386.c (ix86_print_operand): Use conditional
5876 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5879 2018-03-02 Richard Biener <rguenther@suse.de>
5881 PR tree-optimization/84427
5882 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5883 (bitmap_set_subtract_values): Rewrite to handle multiple
5886 (prune_clobbered_mems): Likewise.
5887 (phi_translate): Take edge instead of pred/phiblock.
5888 (phi_translate_1): Likewise.
5889 (phi_translate_set): Likewise. Insert all translated
5890 exprs for a value into the set, keeping possibly multiple
5891 expressions per value.
5892 (compute_antic_aux): Adjust for phi_translate changes.
5893 When intersecting union the expressions and prune those
5894 not in the final value set, keeping possibly multiple
5895 expressions per value. Do not use value-insertion
5896 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5897 all expressions. Add verification that the value-sets
5898 only shrink during iteration.
5899 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5900 (do_pre_regular_insertion): Likewise.
5901 (do_pre_partial_partial_insertion): Likewise.
5903 2018-03-02 Richard Biener <rguenther@suse.de>
5906 * config/darwin.c (saved_debug_info_level): New static global.
5907 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5908 (darwin_asm_lto_end): Restore debug info generation settings.
5910 2018-03-01 Martin Liska <mliska@suse.cz>
5913 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5916 2018-03-01 Richard Biener <rguenther@suse.de>
5919 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5920 type annotation with LTO when debug was disabled at compile-time.
5922 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5924 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5926 (mips_final_postscan_insn): Likewise.
5928 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5930 PR rtl-optimization/84528
5931 * alias.c (init_alias_target): Add commentary.
5932 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5933 a unique base value if the frame pointer is not eliminated
5934 to the stack pointer.
5936 2018-03-01 Tom de Vries <tom@codesourcery.com>
5938 PR rtl-optimization/83327
5939 * lra-int.h (hard_regs_spilled_into): Declare.
5940 * lra.c (hard_regs_spilled_into): Define.
5941 (init_reg_info): Init hard_regs_spilled_into.
5942 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5943 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5944 (process_bb_lives): Handle hard_regs_spilled_into.
5945 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5946 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5948 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5950 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5951 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5952 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5953 * config/rs6000/aix72.h: New file.
5955 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5957 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5958 instead of warning_at with conditional singular and plural messages
5962 * stor-layout.c (update_alignment_for_field): For
5963 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5964 && !DECL_PACKED (field), do the alignment update, just use
5965 only desired_align instead of MAX (type_align, desired_align)
5967 (place_field): Don't do known_align < desired_align handling
5968 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5969 is non-NULL, instead do it after rli->prev_field handling and
5970 only if not within a bitfield word. For DECL_PACKED (field)
5971 use type_align of BITS_PER_UNIT.
5973 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5975 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5976 superfluous parentheses and trailing spaces.
5978 2018-02-28 Richard Biener <rguenther@suse.de>
5980 PR tree-optimization/84584
5981 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5982 SCoPs with fake exit edge.
5984 2018-02-28 Martin Liska <mliska@suse.cz>
5987 * timevar.c (timer::print): Fix format to properly print 100%
5990 2018-02-28 Richard Biener <rguenther@suse.de>
5993 * genmatch.c (capture_info::walk_match): Do not mark
5994 captured expressions without operands as expr_p given
5995 they act more like predicates and should be subject to
5996 "lost tail" side-effect preserving.
5998 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6000 PR rtl-optimization/81611
6001 * auto-inc-dec.c (attempt_change): Move dead note from
6002 mem_insn if it's the next use of regno
6003 (find_address): Take address use of reg holding
6004 non-incremented value. Add parm to limit search to the named
6006 (merge_in_block): Attempt to use a mem insn that is the next
6007 use of the original regno.
6009 2018-02-27 Martin Sebor <msebor@redhat.com>
6012 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6013 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6015 2018-02-27 Martin Sebor <msebor@redhat.com>
6017 PR translation/84207
6018 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6019 n argument to unsigned HOST_WIDE_INT.
6020 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6021 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6022 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6023 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6025 2018-02-27 Richard Biener <rguenther@suse.de>
6027 PR tree-optimization/84512
6028 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6029 Do not use the estimate returned from record_stmt_cost for
6030 the scalar iteration cost but sum properly using add_stmt_cost.
6032 2018-02-27 Richard Biener <rguenther@suse.de>
6034 PR tree-optimization/84466
6035 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6036 Adjust last change to less strictly validate use operands.
6038 2018-02-27 Martin Liska <mliska@suse.cz>
6040 PR gcov-profile/84548
6041 * gcov.c (process_file): Allow partial overlap and consider it
6042 also as group functions.
6043 (output_lines): Properly calculate range of lines for a group.
6045 2018-02-27 Martin Liska <mliska@suse.cz>
6047 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6048 'ggc' suffixes. Change first column width.
6049 (timer::print): Fix formatting of the column.
6051 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6053 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6054 preserve inline entry blocks for the sake of debug inline
6055 entry point markers alone.
6056 (remove_unused_locals): Suggest in comments a better place to
6057 force the preservation of inline entry blocks that are
6058 otherwise unused, but do not preserve them.
6060 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6062 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6064 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6067 * config/i386/constraints.md (Bs): Replace
6068 ix86_indirect_branch_register with
6069 TARGET_INDIRECT_BRANCH_REGISTER.
6071 * config/i386/i386.md (indirect_jump): Likewise.
6072 (tablejump): Likewise.
6073 (*sibcall_memory): Likewise.
6074 (*sibcall_value_memory): Likewise.
6075 Peepholes of indirect call and jump via memory: Likewise.
6076 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6077 (*sibcall_value_GOT_32): Likewise.
6078 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6079 (GOT_memory_operand): Likewise.
6080 (call_insn_operand): Likewise.
6081 (sibcall_insn_operand): Likewise.
6082 (GOT32_symbol_operand): Likewise.
6083 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6085 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6087 PR rtl-optimization/83496
6088 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6089 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6090 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6091 redundant insn, if any.
6092 (relax_delay_slots): Likewise.
6093 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6095 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6097 PR tree-optimization/83965
6098 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6099 that grouped statements are part of a reduction chain. Return
6100 true if the statement is not marked as a reduction itself but
6102 (vect_recog_dot_prod_pattern): Don't check whether the statement
6103 is part of a group here.
6104 (vect_recog_sad_pattern): Likewise.
6105 (vect_recog_widen_sum_pattern): Likewise.
6107 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6110 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6112 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6115 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6117 (ix86_output_indirect_function_return): New prototype.
6118 (ix86_split_simple_return_pop_internal): Likewise.
6119 * config/i386/i386.c (indirect_return_via_cx): New.
6120 (indirect_return_via_cx_bnd): Likewise.
6121 (indirect_thunk_name): Handle return va CX_REG.
6122 (output_indirect_thunk_function): Create alias for
6123 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6124 (ix86_output_indirect_jmp): Remove the bool argument.
6125 (ix86_output_indirect_function_return): New function.
6126 (ix86_split_simple_return_pop_internal): Likewise.
6127 * config/i386/i386.md (*indirect_jump): Don't pass false
6128 to ix86_output_indirect_jmp.
6129 (*tablejump_1): Likewise.
6130 (simple_return_pop_internal): Change it to define_insn_and_split.
6131 Call ix86_split_simple_return_pop_internal to split it for
6133 (simple_return_indirect_internal): Call
6134 ix86_output_indirect_function_return instead of
6135 ix86_output_indirect_jmp.
6137 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6140 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6141 memset and value initialization afterwards.
6143 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6145 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6147 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6150 * common/config/aarch64/aarch64-common.c
6151 (aarch_option_optimization_table[]): Switch
6152 off fomit-frame-pointer
6154 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6155 Chung-Ju Wu <jasonwucj@gmail.com>
6157 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6159 (store_multiple): Ditto.
6161 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6163 * config.gcc: Add --with-cpu support for nds32 target.
6164 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6165 * config/nds32/nds32.opt: Add -mcpu= option.
6167 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6169 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6170 isel=yes): Warn for these deprecated options.
6172 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6174 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6175 ISA_2_5_MASKS_EMBEDDED.
6177 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6179 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6180 p->max as pointers rather than using iterative_hash_expr.
6182 2018-02-23 Carl Love <cel@us.ibm.com>
6184 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6185 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6187 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6188 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6189 P8V_BUILTIN_VEC_VUNSIGNED2.
6191 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6194 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6195 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6197 * lra-constraints.c (curr_insn_transform): Set up
6198 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6200 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6201 * lra-eliminations.c (spill_pseudos): Ditto.
6202 (process_insn_for_elimination): Ditto.
6203 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6204 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6205 LRA_NON_CLOBBERED_ALT.
6207 2018-02-22 Martin Sebor <msebor@redhat.com>
6209 PR tree-optimization/84480
6210 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6211 to maybe_diag_stxncpy_trunc. Call it.
6212 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6213 from gimple_fold_builtin_strcpy. Print inlining stack.
6214 (handle_builtin_stxncpy): Print inlining stack.
6215 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6217 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6220 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6221 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6222 and -fcheck-pointer-bounds are used together.
6223 (indirect_thunk_prefix): New enum.
6224 (indirect_thunk_need_prefix): New function.
6225 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6226 "_nt" instead of "_bnd" for NOTRACK prefix.
6227 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6228 (output_indirect_thunk_function): Likewise.
6230 (ix86_code_end): Update output_indirect_thunk_function calls.
6231 (ix86_output_indirect_branch_via_reg): Replace
6232 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6233 (ix86_output_indirect_branch_via_push): Likewise.
6234 (ix86_output_function_return): Likewise.
6235 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6236 incompatible with -fcf-protection=branch and
6237 -fcheck-pointer-bounds.
6239 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6242 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6243 Change gcc_assert call to output_operand_lossage.
6245 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6247 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6249 2018-02-22 DJ Delorie <dj@redhat.com>
6250 Sebastian Perta <sebastian.perta@renesas.com>
6251 Oleg Endo <olegendo@gcc.gnu.org>
6253 * config/rx/rx.c (rx_rtx_costs): New function.
6254 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6256 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6258 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6260 2018-02-22 Martin Liska <mliska@suse.cz>
6263 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6264 Add "native" as a possible value.
6266 2018-02-22 Martin Liska <mliska@suse.cz>
6269 * config/i386/i386.c (ix86_option_override_internal):
6270 Add "native" as a possible value for -march and -mtune.
6272 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6275 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6276 to all type variants.
6278 PR tree-optimization/84503
6279 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6280 width as info->bitpos + info->bitsize - start.
6281 (merged_store_group::merge_overlapping): Simplify width computation.
6282 (check_no_overlap): New function.
6283 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6284 start + width and last_order of the group, fail if check_no_overlap
6286 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6287 to group if check_no_overlap fails.
6289 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6291 * config/rs6000/altivec.md: Delete contraint arguments to
6292 define_expand, define_split, and define_peephole2, and in
6293 define_insn_and_split if always unused.
6294 * config/rs6000/darwin.md: Ditto.
6295 * config/rs6000/dfp.md: Ditto.
6296 * config/rs6000/rs6000.md: Ditto.
6297 * config/rs6000/sync.md: Ditto.
6298 * config/rs6000/vector.md: Ditto.
6299 * config/rs6000/vsx.md: Ditto.
6301 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6303 * config/rs6000/altivec.md: Write output control strings as braced
6304 blocks instead of double-quoted strings.
6305 * config/rs6000/darwin.md: Ditto.
6306 * config/rs6000/rs6000.md: Ditto.
6307 * config/rs6000/vector.md: Ditto.
6308 * config/rs6000/vsx.md: Ditto.
6310 2018-02-21 Jason Merrill <jason@redhat.com>
6312 PR c++/84314 - ICE with templates and fastcall attribute.
6313 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6315 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6317 * ipa-cp.c (determine_versionability): Fix comment typos.
6319 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6322 * ipa-cp.c (determine_versionability): Do not version functions caling
6325 2018-02-21 Martin Liska <mliska@suse.cz>
6328 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6329 Add "native" as a possible value.
6330 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6331 the macro when native cpu detection is available.
6333 2018-02-21 Martin Liska <mliska@suse.cz>
6336 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6337 Add "native" as a possible value.
6338 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6339 when native cpu detection is available.
6341 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6342 Martin Sebor <msebor@redhat.com>
6344 PR tree-optimization/84478
6345 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6347 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6348 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6349 support which is conservatively correct, for 2 only stay conservative
6350 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6351 argument to the 2 argument get_range_strlen, adjust 6 arg
6352 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6354 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6355 (gimple_fold_builtin_strlen): Pass true as last argument to
6358 2018-02-20 Martin Sebor <msebor@redhat.com>
6361 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6362 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6363 (builtin_memref::builtin_memref): Factor out parts into
6364 set_base_and_offset and call it.
6366 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6369 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6370 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6371 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6372 search at the associated MODE_INT.
6374 2018-02-20 Jeff Law <law@redhat.com>
6377 PR tree-optimization/81592
6379 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6380 for range data rather than using global data.
6381 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6382 range data rather than using global data.
6383 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6384 pass it to children as needed.
6385 (struct directive::fmtresult): Similarly.
6386 (struct directive::set_width): Similarly.
6387 (struct directive::set_precision): Similarly.
6388 (format_integer, format_directive, parse_directive): Similarly.
6389 (format_none): Accept unnamed vr_values parameter.
6390 (format_percent, format_floating, format_character): Similarly.
6391 (format_string, format_plain): Similarly.
6392 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6393 the EVRP range analyzer for range data rather than using global data.
6394 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6395 gimple-ssa-evrp-analyze.h
6396 (class sprintf_dom_walker): Add after_dom_children member function.
6397 Add evrp_range_analyzer member.
6398 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6399 range analyzer as needed.
6400 (sprintf_dom_walker::after_dom_children): New member function.
6401 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6403 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6404 (evrp_range_analyzer::pop_to_marker): Likewise.
6406 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6408 PR tree-optimization/84419
6409 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6410 with the required type if its current type is compatible but
6413 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6416 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6417 after vectorization.
6419 2018-02-20 Martin Liska <mliska@suse.cz>
6422 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6423 possible values if we don't have a hint.
6425 2018-02-20 Martin Liska <mliska@suse.cz>
6429 * final.c (shorten_branches): Build align_tab array with one
6431 * opts.c (finish_options): Add alignment option limit check.
6432 (MAX_CODE_ALIGN): Likewise.
6433 (MAX_CODE_ALIGN_VALUE): Likewise.
6434 * doc/invoke.texi: Document maximum allowed option value for
6435 all -falign-* options.
6437 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6440 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6441 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6442 * var-tracking.c (emit_note_insn_var_location): Remove all references
6443 to NOTE_INSN_CALL_ARG_LOCATION.
6444 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6445 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6446 Use copy_rtx_if_shared.
6447 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6448 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6449 (dwarf2out_var_location): Remove handling of
6450 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6452 * final.c (final_scan_insn): Remove all references to
6453 NOTE_INSN_CALL_ARG_LOCATION.
6454 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6455 before dumping final insns.
6456 * except.c (emit_note_eh_region_end): Remove all references to
6457 NOTE_INSN_CALL_ARG_LOCATION.
6458 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6459 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6460 * config/arc/arc.c (hwloop_optimize): Likewise.
6461 * config/arm/arm.c (create_fix_barrier): Likewise.
6462 * config/s390/s390.c (s390_chunkify_start): Likewise.
6463 * config/sh/sh.c (find_barrier): Likewise.
6464 * config/i386/i386.c (rest_of_insert_endbranch,
6465 ix86_seh_fixup_eh_fallthru): Likewise.
6466 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6467 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6468 * config/frv/frv.c (frv_function_prologue): Likewise.
6469 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6471 (note_outside_basic_block_p): Remove all references to
6472 NOTE_INSN_CALL_ARG_LOCATION.
6473 * gengtype.c (adjust_field_rtx_def): Likewise.
6474 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6476 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6477 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6480 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6483 PR tree-optimization/84452
6484 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6485 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6488 2018-02-19 Martin Liska <mliska@suse.cz>
6491 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6493 2018-02-19 Martin Liska <mliska@suse.cz>
6494 Richard Sandiford <richard.sandiford@linaro.org>
6496 PR tree-optimization/82491
6497 * gimple-fold.c (get_base_constructor): Make earlier bail out
6500 2018-02-19 Carl Love <cel@us.ibm.com>
6502 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6503 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6505 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6506 P8V_BUILTIN_VEC_NEG.
6508 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6510 * config/rl78/rl78.md (movdf): New define expand.
6512 2018-02-19 Martin Liska <mliska@suse.cz>
6515 * doc/invoke.texi: Fix typo.
6516 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6518 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6520 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6521 handle rs6000_single_float and rs6000_double_float specially for
6524 2018-02-16 Jeff Law <law@redhat.com>
6526 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6529 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6532 * ipa-inline.c (inline_small_functions): Fix a typo.
6534 2018-02-16 Nathan Sidwell <nathan@acm.org>
6536 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6538 2018-02-16 Carl Love <cel@us.ibm.com>
6540 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6541 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6542 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6543 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6544 expansion to P8V_BUILTIN_VEC_FLOAT2.
6546 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6548 PR rtl-optimization/70023
6549 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6550 src_regno into account.
6552 2018-02-16 Carl Love <cel@us.ibm.com>
6554 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6555 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6556 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6557 * config/rs6000/rs6000.c: Remove case statements for
6558 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6559 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6560 and P9V_BUILTIN_VEC_VINSERT4B.
6561 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6562 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6563 * config/rs6000/vsx.md:
6564 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6567 2018-02-16 Carl Love <cel@us.ibm.com>
6569 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6571 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6573 * config/rs6000/rs6000-c.c: Add the definitions for
6574 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6575 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6576 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6577 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6578 definition for insert4b and define insn *insert3b_internal.
6579 * doc/extend.texi: Add documentation for vec_extract4b.
6581 2018-02-16 Nathan Sidwell <nathan@acm.org>
6583 * doc/extend.texi (Backwards Compatibility): Mention friend
6584 injection. Note for-scope is deprecated.
6585 * doc/invoke.texi (-ffriend-injection): Deprecate.
6587 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6589 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6590 that moved to I2, also allow destinations that are a paradoxical
6591 subreg (instead of a normal reg).
6593 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6596 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6599 2018-02-16 Richard Biener <rguenther@suse.de>
6601 PR tree-optimization/84037
6602 PR tree-optimization/84016
6604 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6605 Adjust vec_construct for the fact we need additional higher latency
6606 128bit inserts for AVX256 and AVX512 vector builds.
6607 (ix86_add_stmt_cost): Scale vector construction cost for
6610 2018-02-16 Richard Biener <rguenther@suse.de>
6612 PR tree-optimization/84417
6613 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6614 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6615 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6617 2018-02-16 Martin Liska <mliska@suse.cz>
6620 * internal-fn.def (ASAN_CHECK): Set proper flags.
6621 (ASAN_MARK): Likewise.
6623 2018-02-16 Julia Koval <julia.koval@intel.com>
6625 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6626 from PTA_CANNONLAKE.
6628 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6631 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6632 Use ++iter rather than iter++ for std::list iterators.
6633 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6634 defer deleting them until all nodes in the forest are processed. Do
6635 free even leaf nodes. Change to_process into auto_vec.
6638 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6639 * vec.h (vec_default_construct): Use memset instead of placement new
6640 if BROKEN_VALUE_INITIALIZATION is defined.
6641 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6642 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6645 PR rtl-optimization/83723
6646 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6647 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6648 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6650 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6652 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6654 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6656 PR rtl-optimization/81443
6657 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6658 from inner REGs to paradoxical SUBREGs.
6660 2018-02-16 Richard Biener <rguenther@suse.de>
6662 PR tree-optimization/84399
6663 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6664 For operands we can analyze at their definition make sure we can
6665 analyze them at each use as well.
6667 2018-02-16 Richard Biener <rguenther@suse.de>
6669 PR tree-optimization/84190
6670 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6671 volatile accesses if the decl isn't volatile.
6673 2018-02-15 Jason Merrill <jason@redhat.com>
6675 PR c++/84314 - ICE with templates and fastcall attribute.
6676 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6677 TYPE_CANONICAL on an existing type.
6679 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6681 PR tree-optimization/84383
6682 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6683 dstoff nor call operand_equal_p if dstbase is NULL.
6685 PR tree-optimization/84334
6686 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6687 also a CONSTANT_CLASS_P, punt.
6689 2018-02-14 Jim Wilson <jimw@sifive.com>
6691 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6692 first SMALL_OPERAND check. New local min_second_step. Move assert
6693 to where locals are set. Add TARGET_RVC support.
6694 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6696 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6698 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6700 2018-02-14 Martin Sebor <msebor@redhat.com>
6702 PR tree-optimization/83698
6703 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6704 arrays constrain the offset range to their bounds.
6705 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6706 (builtin_access::overlap): Avoid setting the size of overlap if it's
6708 (maybe_diag_overlap): Also consider arrays when deciding what values
6709 of offsets to include in diagnostics.
6711 2018-02-14 Martin Sebor <msebor@redhat.com>
6714 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6715 that correspond to the kind of a declaration.
6717 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6720 * config/pa/pa.md: Load address of PIC label using the linkage table
6721 if the label is nonlocal.
6723 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6725 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6726 warning message if user requests -maltivec=be.
6727 * doc/invoke.texi: Document deprecation of -maltivec=be.
6729 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6732 * config/rs6000/rs6000-c.c: Update definitions for
6733 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6734 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6736 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6739 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6740 add _get_ssp intrinsics. Remove argument from
6741 __builtin_ia32_rdssp[d|q].
6742 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6743 * config/i386/i386-builtin.def: Remove argument from
6744 __builtin_ia32_rdssp[d|q].
6745 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6746 ix86_expand_special_args_builtin for _rdssp[d|q].
6747 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6748 Clear register before usage.
6749 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6750 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6752 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6754 PR tree-optimization/84357
6755 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6756 operand 1 of an ARRAY_REF too.
6758 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6761 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6762 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6764 (set_of_reg): New struct.
6765 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6766 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6767 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6769 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6770 Split into bitclr, bitset, bitinvert patterns if appropriate.
6771 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6772 use rx_fuse_in_memory_bitop.
6773 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6774 to named insn, correct maximum insn length.
6776 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6779 * machmode.def: Define a complex mode for PARTIAL_INT.
6780 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6782 * doc/rtl.texi: Document CSPImode.
6783 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6785 (msp430_hard_regno_nregs_with_padding): Likewise.
6787 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6790 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6792 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6794 PR rtl-optimization/84169
6795 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6796 we generated a parallel as new i3 and we split that to new i2 and i3
6797 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6798 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6799 those to i2, not i1. Partially rewrite this scan code.
6801 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6804 * stor-layout.c (place_field): For variable length fields, adjust
6805 offset_align afterwards not just based on the field's alignment,
6806 but also on the size.
6809 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6810 of exps and logs in the use_exp2 case.
6812 2018-02-13 Jeff Law <law@redhat.com>
6814 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6817 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6820 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6824 * common.opt (gas-loc-support, gas-locview-support): New.
6825 (ginline-points, ginternal-reset-location-views): New.
6826 * doc/invoke.texi: Document them. Use @itemx where intended.
6827 (gvariable-location-views): Adjust.
6828 * target.def (reset_location_view): New.
6829 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6830 (TARGET_RESET_LOCATION_VIEW): New.
6831 * doc/tm.texi: Rebuilt.
6832 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6833 (dwarf2out_default_as_locview_support): New.
6834 (output_asm_line_debug_info): Use option variables.
6835 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6836 (output_loc_list): Likewise.
6837 (add_high_low_attributes): Check option variables.
6838 Don't output entry view attribute in strict mode.
6839 (gen_inlined_subroutine_die): Check option variables.
6840 (dwarf2out_inline_entry): Likewise.
6841 (init_sections_and_labels): Likewise.
6842 (dwarf2out_early_finish): Likewise.
6843 (maybe_reset_location_view): New, from...
6844 (dwarf2out_var_location): ... here. Call it.
6845 * debug.h (dwarf2out_default_as_loc_support): Declare.
6846 (dwarf2out_default_as_locview_support): Declare.
6847 * hooks.c (hook_int_rtx_insn_0): New.
6848 * hooks.h (hook_int_rtx_insn_0): Declare.
6849 * toplev.c (process_options): Take -gas-loc-support and
6850 -gas-locview-support from dwarf2out. Enable
6851 -gvariable-location-views by default only with locview
6852 assembler support. Enable -ginternal-reset-location-views by
6853 default only if the target defines the corresponding hook.
6854 Enable -ginline-points by default if location views are
6855 enabled; force it disabled if statement frontiers are
6857 * tree-inline.c (expand_call_inline): Check option variables.
6858 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6860 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6862 PR tree-optimization/84321
6863 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6864 handling. Also check whether the anti-range contains any values
6865 that satisfy the mask; switch to a VR_RANGE if not.
6867 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6870 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6872 2018-02-13 Martin Jambor <mjambor@suse.cz>
6875 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6876 of call statements, also set location of a load to a temporary.
6878 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6880 * config/rl78/rl78.c (add_vector_labels): New function.
6881 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6882 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6883 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6884 which checks that no arguments are passed.
6885 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6886 * doc/extend.texi: Documentation for the new attribute.
6888 2018-02-13 Andreas Schwab <schwab@suse.de>
6890 * config/riscv/linux.h (CPP_SPEC): Define.
6892 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6895 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6896 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6897 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6898 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6899 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6900 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6901 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6902 temporarily for AES and PCLMUL builtins.
6904 PR tree-optimization/84339
6905 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6906 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6910 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6911 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6912 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6914 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6916 * omp-simd-clone.h: New file.
6917 * omp-simd-clone.c: Include omp-simd-clone.h.
6918 (expand_simd_clones): No longer static.
6919 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6920 cgraph.h and omp-simd-clone.h.
6921 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6922 (vect_recog_widen_shift_pattern): Formatting fix.
6923 (vect_pattern_recog_1): Don't check optab for calls.
6926 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6927 operands[2] into a REG before using gen_lowpart on it.
6929 2018-02-12 Jeff Law <law@redhat.com>
6932 * config/sh/sh.c (find_barrier): Consider a sibling call
6935 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6936 successfully back substituting a reg.
6938 2018-02-12 Richard Biener <rguenther@suse.de>
6940 PR tree-optimization/84037
6941 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6942 parameter, move visited init to caller.
6943 (vect_slp_analyze_operations): Separate cost from validity
6944 check, initialize visited once for all instances.
6945 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6947 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6949 (vect_model_promotion_demotion_cost): Likewise.
6950 (vectorizable_bswap): Guard cost modeling with !slp_node
6951 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6953 (vectorizable_call): Likewise.
6954 (vectorizable_conversion): Likewise.
6955 (vectorizable_assignment): Likewise.
6956 (vectorizable_shift): Likewise.
6957 (vectorizable_operation): Likewise.
6958 (vectorizable_store): Likewise.
6959 (vectorizable_load): Likewise.
6960 (vectorizable_condition): Likewise.
6961 (vectorizable_comparison): Likewise.
6963 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6966 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6967 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6968 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6970 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6973 * params.def (inline-min-speedup): Increase from 8 to 15.
6974 (max-inline-insns-auto): Decrease from 40 to 30.
6975 * ipa-split.c (consider_split): Add some buffer for function to
6976 be considered inlining candidate.
6977 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6980 2018-02-12 Richard Biener <rguenther@suse.de>
6982 PR tree-optimization/84037
6983 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6984 matched stmts if we cannot swap the non-matched ones.
6986 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6988 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6989 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6990 _mm_maskz_scalef_round_ss): New intrinsics.
6991 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6992 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6993 __builtin_ia32_scalefss_round): Remove.
6994 (__builtin_ia32_scalefsd_mask_round,
6995 __builtin_ia32_scalefss_mask_round): New intrinsics.
6996 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6997 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6998 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6999 "<round_constraint>")): Changed to ...
7000 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7001 "<round_scalar_constraint>")): ... this.
7002 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7003 %0, %1, %2<round_op3>}"): Changed to ...
7004 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7005 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7006 %2<round_scalar_mask_op3>}"): ... this.
7007 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7009 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7011 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7012 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7013 (_mm_maskz_sqrt_round_ss): New intrinsics.
7014 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7015 (__builtin_ia32_sqrtsd_mask_round)
7016 (__builtin_ia32_sqrtss_mask_round): New builtins.
7017 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7018 (__builtin_ia32_sqrtss_round): Remove.
7019 (__builtin_ia32_sqrtsd_mask_round)
7020 (__builtin_ia32_sqrtss_mask_round): New builtins.
7021 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7022 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7023 ((match_operand:VF_128 1 "vector_operand"
7024 "xBm,<round_constraint>")): Changed to ...
7025 ((match_operand:VF_128 1 "vector_operand"
7026 "xBm,<round_scalar_constraint>")): ... this.
7027 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7028 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7029 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7030 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7031 %<iptr>1<round_scalar_mask_op3>}): ... this.
7032 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7033 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7035 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7038 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7039 Cast vec_cmpeq result to correct type.
7040 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7041 Cast vec_cmpgt result to correct type.
7043 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7045 * final.c (final_scan_insn_1): Renamed from...
7046 (final_scan_insn): ... this. New wrapper, to recover
7047 seen from the outermost call in recursive ones.
7048 * config/sparc/sparc.c (output_return): Drop seen from call.
7049 (output_sibcall): Likewise.
7050 * config/visium/visium.c (output_branch): Likewise.
7052 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7054 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7057 2018-02-10 Alan Modra <amodra@gmail.com>
7060 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7061 Specify LR as an input.
7063 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7066 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7067 remove_member_access_dummy_vars): New functions.
7068 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7069 lower_omp_1, execute_lower_omp): Use them.
7071 PR rtl-optimization/84308
7072 * shrink-wrap.c (spread_components): Release todo vector.
7074 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7076 PR rtl-optimization/57193
7077 * ira-color.c (struct allocno_color_data): Add member
7078 conflict_allocno_hard_prefs.
7079 (update_conflict_allocno_hard_prefs): New.
7080 (bucket_allocno_compare_func): Add a preference based on
7081 conflict_allocno_hard_prefs.
7082 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7083 (color_allocnos): Remove a dead code. Initiate
7084 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7086 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7089 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7090 constraint from =wa to wa. Avoid a subreg on the output operand,
7091 instead use a pseudo and subreg it in a move.
7092 (p9_xxbrd_<mode>): Changed to ...
7093 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7094 (p9_xxbrd_v2df): New expander.
7095 (p9_xxbrw_<mode>): Changed to ...
7096 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7097 (p9_xxbrw_v4sf): New expander.
7099 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7101 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7103 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7106 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7107 multiply in 32-bit mode.
7108 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7109 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7112 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7114 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7115 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7116 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7117 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7119 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7122 * dwarf2out.c (is_trivial_indirect_ref): New function.
7123 (dwarf2out_late_global_decl): Do not generate a location
7124 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7125 and that are not defined in the current unit.
7127 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7129 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7130 instead of a libcall for UNORDERED.
7132 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7135 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7136 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7138 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7141 * config/s390/s390.c (s390_set_current_function): Invoke
7142 s390_indirect_branch_settings also if fndecl didn't change.
7144 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7146 * config/rs6000/rs6000.md (blockage): Set length to zero.
7148 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7150 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7152 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7155 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7156 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7160 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7161 PARALLEL incoming that failed vt_get_decl_and_offset check.
7164 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7165 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7167 (get_variable_section): For decls in named .bss* sections pass true as
7168 second argument to bss_initializer_p.
7170 2018-02-09 Marek Polacek <polacek@redhat.com>
7171 Jakub Jelinek <jakub@redhat.com>
7174 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7175 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7176 Sync some changes from cxx_fold_indirect_ref.
7178 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7180 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7182 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7183 (BLOCK_INLINE_ENTRY_LABEL): New.
7184 (dwarf2out_var_location): Disregard inline entry markers.
7185 (inline_entry_data): New struct.
7186 (inline_entry_data_hasher): New hashtable type.
7187 (inline_entry_data_hasher::hash): New.
7188 (inline_entry_data_hasher::equal): New.
7189 (inline_entry_data_table): New variable.
7190 (add_high_low_attributes): Add DW_AT_entry_pc and
7191 DW_AT_GNU_entry_view attributes if a pending entry is found
7192 in inline_entry_data_table. Add old entry_pc attribute only
7193 if debug nonbinding markers are disabled.
7194 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7195 markers are enabled.
7196 (block_within_block_p, dwarf2out_inline_entry): New.
7197 (dwarf2out_finish): Check that no entries remained in
7198 inline_entry_data_table.
7199 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7200 (final_scan_insn, notice_source_line): Likewise.
7201 (rest_of_clean_state): Skip inline entry markers.
7202 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7204 * gimple.c (gimple_build_debug_inline_entry): New.
7205 * gimple.h (enum gimple_debug_subcode): Add
7206 GIMPLE_DEBUG_INLINE_ENTRY.
7207 (gimple_build_debug_inline_entry): Declare.
7208 (gimple_debug_inline_entry_p): New.
7209 (gimple_debug_nonbind_marker_p): Adjust.
7210 * insn-notes.def (INLINE_ENTRY): New.
7211 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7212 inline entry marker notes.
7213 (print_insn): Likewise.
7214 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7215 (INSN_DEBUG_MARKER_KIND): Likewise.
7216 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7217 * tree-inline.c (expand_call_inline): Build and insert
7218 debug_inline_entry stmt.
7219 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7220 inline entry blocks early, if nonbind markers are enabled.
7221 (dump_scope_block): Dump fragment info.
7222 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7223 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7224 (gimple_build_debug_inline_entry): New.
7225 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7226 Enable/disable inline entry points too.
7227 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7228 (DEBUG_INSN): Describe inline entry markers.
7230 * common.opt (gvariable-location-views): New.
7231 (gvariable-location-views=incompat5): New.
7232 * config.in: Rebuilt.
7233 * configure: Rebuilt.
7234 * configure.ac: Test assembler for view support.
7235 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7236 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7237 * dwarf2out.c (var_loc_view): New typedef.
7238 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7239 (dwarf2out_locviews_in_attribute): New.
7240 (dwarf2out_locviews_in_loclist): New.
7241 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7242 (enum dw_line_info_opcode): Add LI_adv_address.
7243 (struct dw_line_info_table): Add view.
7244 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7245 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7246 (zero_view_p): New variable.
7247 (ZERO_VIEW_P): New macro.
7248 (output_asm_line_debug_info): New.
7249 (struct var_loc_node): Add view.
7250 (add_AT_view_list, AT_loc_list): New.
7251 (add_var_loc_to_decl): Add view param. Test it against last.
7252 (new_loc_list): Add view params. Record them.
7253 (AT_loc_list_ptr): Handle loc and view lists.
7254 (view_list_to_loc_list_val_node): New.
7255 (print_dw_val): Handle dw_val_class_view_list.
7256 (size_of_die): Likewise.
7257 (value_format): Likewise.
7258 (loc_list_has_views): New.
7259 (gen_llsym): Set vl_symbol too.
7260 (maybe_gen_llsym, skip_loc_list_entry): New.
7261 (dwarf2out_maybe_output_loclist_view_pair): New.
7262 (output_loc_list): Output view list or entries too.
7263 (output_view_list_offset): New.
7264 (output_die): Handle dw_val_class_view_list.
7265 (output_dwarf_version): New.
7266 (output_compilation_unit_header): Use it.
7267 (output_skeleton_debug_sections): Likewise.
7268 (output_rnglists, output_line_info): Likewise.
7269 (output_pubnames, output_aranges): Update version comments.
7270 (output_one_line_info_table): Output view numbers in asm comments.
7271 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7272 Call maybe_gen_llsym.
7273 (loc_list_from_tree_1): Adjust.
7274 (add_AT_location_description): Create view list attribute if
7275 needed, check it's absent otherwise.
7276 (convert_cfa_to_fb_loc_list): Adjust.
7277 (maybe_emit_file): Call output_asm_line_debug_info for test.
7278 (dwarf2out_var_location): Reset views as needed. Precompute
7279 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7280 attribute. Set view.
7281 (new_line_info_table): Reset next view.
7282 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7283 (dwarf2out_source_line): Likewise. Output view resets and labels to
7284 the assembler, or select appropriate line info opcodes.
7285 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7286 (optimize_string_length): Catch it. Adjust.
7287 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7288 dw_val_class_view_list, and remove it if no longer needed.
7289 (hash_loc_list): Hash view numbers.
7290 (loc_list_hasher::equal): Compare them.
7291 (optimize_location_lists): Check whether a view list symbol is
7292 needed, and whether the locview attribute is present, and
7293 whether they match. Remove the locview attribute if no longer
7295 (index_location_lists): Call skip_loc_list_entry for test.
7296 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7297 Use output_dwarf_version.
7298 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7299 (struct dw_val_node): Add val_view_list.
7300 * final.c (SEEN_NEXT_VIEW): New.
7301 (set_next_view_needed): New.
7302 (clear_next_view_needed): New.
7303 (maybe_output_next_view): New.
7304 (final_start_function): Rename to...
7305 (final_start_function_1): ... this. Take pointer to FIRST,
7306 add SEEN parameter. Emit param bindings in the initial view.
7307 (final_start_function): Reintroduce SEEN-less interface.
7308 (final): Rename to...
7309 (final_1): ... this. Take SEEN parameter. Output final pending
7310 next view at the end.
7311 (final): Reintroduce seen-less interface.
7312 (final_scan_insn): Output pending next view before switching
7313 sections or ending a block. Mark the next view as needed when
7314 outputting variable locations. Notify debug backend of section
7315 changes, and of location view changes.
7316 (rest_of_handle_final): Adjust.
7317 * toplev.c (process_options): Autodetect value for debug variable
7318 location views option. Warn on incompat5 without -gdwarf-5.
7319 * doc/invoke.texi (gvariable-location-views): New.
7320 (gvariable-location-views=incompat5): New.
7321 (gno-variable-location-views): New.
7323 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7325 PR tree-optimization/84136
7326 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7327 that the result of find_edge is non-NULL.
7329 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7332 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7333 storing integer register in SImode. Fix cost of 256 and 512
7334 byte aligned SSE register store.
7336 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7338 * config/i386/i386.c (ix86_multiplication_cost): Fix
7339 multiplication cost for TARGET_AVX512DQ.
7341 2018-02-08 Marek Polacek <polacek@redhat.com>
7343 PR tree-optimization/84238
7344 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7347 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7349 PR tree-optimization/84265
7350 * tree-vect-stmts.c (vectorizable_store): Don't treat
7351 VMAT_CONTIGUOUS accesses as grouped.
7352 (vectorizable_load): Likewise.
7354 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7356 PR tree-optimization/81635
7357 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7358 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7359 (test_round_for_mask): New functions.
7360 (wide_int_cc_tests): Call test_round_for_mask.
7361 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7362 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7363 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7364 range returned by get_range_info.
7366 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7369 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7370 * symtab.c: Include builtins.h
7371 (symtab_node::output_to_lto_symbol_table_p): Move here
7372 from lto-streamer-out.c:output_symbol_p.
7373 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7374 (output_symbol_p): Move all logic to symtab.c
7375 (produce_symtab): Update.
7377 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7379 * config/s390/s390-opts.h (enum indirect_branch): Define.
7380 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7381 (s390_indirect_branch_via_thunk)
7382 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7383 (enum s390_indirect_branch_type): Define.
7384 * config/s390/s390.c (struct s390_frame_layout, struct
7385 machine_function): Remove.
7386 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7387 (indirect_branch_table_label_no, indirect_branch_table_name):
7389 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7390 (enum s390_indirect_branch_option): Define.
7391 (s390_return_addr_from_memory): New function.
7392 (s390_handle_string_attribute): New function.
7393 (s390_attribute_table): Add new attribute handler.
7394 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7395 (s390_indirect_branch_via_thunk): New function.
7396 (s390_indirect_branch_via_inline_thunk): New function.
7397 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7398 sibling call optimization for non z10 compiles.
7399 (s390_emit_call): Force indirect branch target to be a single
7400 register. Add r1 clobber for non-z10 compiles.
7401 (s390_emit_epilogue): Emit return jump via return_use expander.
7402 (s390_reorg): Handle JUMP_INSNs as execute targets.
7403 (s390_option_override_internal): Perform validity checks for the
7404 new command line options.
7405 (s390_indirect_branch_attrvalue): New function.
7406 (s390_indirect_branch_settings): New function.
7407 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7408 (s390_output_indirect_thunk_function): New function.
7409 (s390_code_end): Implement target hook.
7410 (s390_case_values_threshold): Implement target hook.
7411 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7413 * config/s390/s390.h (struct s390_frame_layout)
7414 (struct machine_function): Move here from s390.c.
7415 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7416 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7417 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7418 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7419 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7420 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7421 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7422 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7423 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7424 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7425 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7426 (mnemonic attribute): Add values which aren't recognized
7428 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7429 pattern for branch conversion. Fix mnemonic attribute.
7430 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7431 indirect branch via thunk if requested.
7432 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7433 ("*indirect_jump"): Disable for branch conversion using out of
7435 ("indirect_jump_via_thunk<mode>_z10")
7436 ("indirect_jump_via_thunk<mode>")
7437 ("indirect_jump_via_inlinethunk<mode>_z10")
7438 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7439 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7440 ("casesi_jump_via_inlinethunk<mode>_z10")
7441 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7442 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7443 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7444 ("*indirect2_jump"): Disable for branch conversion.
7445 ("casesi_jump"): Turn into expander and expand patterns for branch
7447 ("return_use"): New expander.
7448 ("*return"): Emit return via thunk and rename it to ...
7449 ("*return<mode>"): ... this one.
7450 * config/s390/s390.opt: Add new options and and enum for the
7453 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7455 * lra-constraints.c (match_reload): Unconditionally use
7456 gen_lowpart_SUBREG, rather than selecting between that
7457 and equivalent gen_rtx_SUBREG code.
7459 2018-02-08 Richard Biener <rguenther@suse.de>
7461 PR tree-optimization/84233
7462 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7463 changed flag instead of boguously re-using phi_inserted.
7465 2018-02-08 Martin Jambor <mjambor@suse.cz>
7467 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7468 static local variables.
7470 2018-02-08 Richard Biener <rguenther@suse.de>
7472 PR tree-optimization/84278
7473 * tree-vect-stmts.c (vectorizable_store): When looking for
7474 smaller vector types to perform grouped strided loads/stores
7475 make sure the mode is supported by the target.
7476 (vectorizable_load): Likewise.
7478 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7480 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7481 Increase LDP/STP opportunities by adding adjacent callee-saves.
7483 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7485 PR rtl-optimization/84068
7486 PR rtl-optimization/83459
7487 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7489 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7491 PR tree-optimization/84224
7492 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7493 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7496 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7499 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7500 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7501 count, remove one USE.
7503 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7505 * doc/install.texi (Configuration): Document the
7506 --with-long-double-format={ibm,ieee} PowerPC configuration
7510 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7511 Convert from define_expand to be define_insn_and_split. Rework
7512 float/double/_Float128 conversions to QI/HI/SImode to work with
7513 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7514 conversions to QI/HImode types did a store and then a load to
7515 truncate the value. For conversions to VSX registers, don't split
7516 the insn, instead emit the code directly. Use the code iterator
7517 any_fix to combine signed and unsigned conversions.
7518 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7519 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7520 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7521 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7522 (fix_<mode>di2_hw): Likewise.
7523 (fixuns_<mode>di2_hw): Likewise.
7524 (fix_<mode>si2_hw): Likewise.
7525 (fixuns_<mode>si2_hw): Likewise.
7526 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7527 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7528 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7529 fix<uns>_trunc<SFDF:mode>si2_p8.
7530 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7532 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7533 (fix<uns>_<mode>_mem): Likewise.
7534 (fctiw<u>z_<mode>_mem): Likewise.
7535 (fix<uns>_<mode>_mem): Likewise.
7536 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7537 the register allocator from doing a direct move to the GPRs to do
7538 a store, and instead use the ISA 3.0 store byte/half-word from
7539 vector register instruction. For IEEE 128-bit floating point,
7540 also optimize stores of 32-bit ints.
7541 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7543 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7545 * genextract.c (push_pathstr_operand): New function to support
7547 (walk_rtx): Call push_pathstr_operand.
7548 (print_path): Support [a-zA-Z].
7550 2018-02-07 Richard Biener <rguenther@suse.de>
7552 PR tree-optimization/84037
7553 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7554 (cse_and_gimplify_to_preheader): Declare.
7555 (vect_get_place_in_interleaving_chain): Likewise.
7556 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7558 (_loop_vec_info::~_loop_vec_info): Delete it.
7559 (cse_and_gimplify_to_preheader): New function.
7560 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7561 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7562 (vectorizable_load): Likewise. For grouped stores always base
7563 the IV on the first element.
7564 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7565 condition before gimplifying.
7567 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7569 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7570 *DIV_EXPR and *MOD_EXPR.
7572 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7575 * config/i386/i386.c (ix86_option_override_internal): Mask out
7576 the CF_SET bit when checking -fcf-protection.
7578 2018-02-07 Tom de Vries <tom@codesourcery.com>
7581 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7584 2018-02-07 Richard Biener <rguenther@suse.de>
7586 PR tree-optimization/84204
7587 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7590 PR tree-optimization/84205
7591 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7592 special-case isl_ast_op_zdiv_r.
7594 PR tree-optimization/84223
7595 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7596 Only add conditions from within the region.
7597 (gather_bbs::after_dom_children): Adjust.
7599 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7602 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7603 * config/avr/avr.md: Only post-reload split REG-REG moves if
7604 either register is GENERAL_REG_P.
7606 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7608 PR tree-optimization/84235
7609 * tree-ssa-scopedtables.c
7610 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7611 if the subtraction is performed in floating point type where NaNs are
7612 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7613 build 1. Formatting fix.
7615 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7618 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7619 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7620 and skip it regardless of bb boundaries. Use CALL_P macro,
7621 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7622 unnecessarily, formatting fix.
7624 2018-02-06 Michael Collison <michael.collison@arm.com>
7626 * config/arm/thumb2.md:
7627 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7628 (*thumb_mov_notscc): Ditto.
7630 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7633 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7634 unsigned_fix, not "s".
7636 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7638 * configure.ac (gcc_fn_eh_frame_ro): New function.
7639 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7640 correct .eh_frame permissions.
7641 * configure: Regenerate.
7643 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7645 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7648 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7650 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7651 Display warning message for -mno-speculate-indirect-jumps.
7653 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7655 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7657 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7659 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7661 PR tree-optimization/84225
7662 * tree-eh.c (find_trapping_overflow): Only call
7663 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7665 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7668 * config/i386/i386.c: Reimplement the check of possible options
7669 -mibt/-mshstk conbination. Change error messages.
7670 * doc/invoke.texi: Fix a typo: remove extra '='.
7672 2018-02-06 Marek Polacek <polacek@redhat.com>
7674 PR tree-optimization/84228
7675 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7677 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7680 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7681 emitted arch directives.
7682 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7683 __ARM_FEATURE_COPROC before changing architectures.
7685 2018-02-06 Richard Biener <rguenther@suse.de>
7687 * config/i386/i386.c (print_reg): Fix typo.
7688 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7690 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7692 * configure: Regenerate.
7694 2018-02-05 Martin Sebor <msebor@redhat.com>
7696 PR tree-optimization/83369
7697 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7700 2018-02-05 Martin Liska <mliska@suse.cz>
7702 * doc/invoke.texi: Cherry-pick upstream r323995.
7704 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7706 * ira.c (ira_init_register_move_cost): Adjust comment.
7708 2018-02-05 Martin Liska <mliska@suse.cz>
7710 PR gcov-profile/84137
7711 * doc/gcov.texi: Fix typo in documentation.
7713 2018-02-05 Martin Liska <mliska@suse.cz>
7715 PR gcov-profile/83879
7716 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7717 using dlopen functionality.
7719 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7721 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7722 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7723 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7724 _mm_maskz_range_round_ss): New intrinsics.
7725 (__builtin_ia32_rangesd128_round)
7726 (__builtin_ia32_rangess128_round): Remove.
7727 (__builtin_ia32_rangesd128_mask_round,
7728 __builtin_ia32_rangess128_mask_round): New builtins.
7729 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7730 __builtin_ia32_rangess128_round): Remove.
7731 (__builtin_ia32_rangesd128_mask_round,
7732 __builtin_ia32_rangess128_mask_round): New builtins.
7733 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7734 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7735 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7736 "<round_saeonly_constraint>")): Changed to ...
7737 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7738 "<round_saeonly_scalar_constraint>")): ... this.
7739 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7740 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7741 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7742 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7743 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7745 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7747 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7749 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7750 Remove all values except native, 8540 and 8548.
7752 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7754 * config/i386/i386.c (ix86_output_function_return): Pass
7755 INVALID_REGNUM, instead of -1, as invalid register number to
7756 indirect_thunk_name and output_indirect_thunk.
7758 2018-02-02 Julia Koval <julia.koval@intel.com>
7760 * config.gcc: Add -march=icelake.
7761 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7762 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7763 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7764 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7765 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7766 (processor_target_table): Add icelake.
7767 (ix86_option_override_internal): Handle new PTAs.
7768 (get_builtin_code_for_version): Handle icelake.
7769 (M_INTEL_COREI7_ICELAKE): New.
7770 (fold_builtin_cpu): Handle icelake.
7771 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7772 * doc/invoke.texi: Add -march=icelake.
7774 2018-02-02 Julia Koval <julia.koval@intel.com>
7776 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7777 to wide_int_bitmask.
7778 * wide-int-bitmask.h: New.
7780 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7783 * config/i386/i386.md: Replace Pmode with word_mode in
7784 builtin_setjmp_setup and builtin_longjmp to support x32.
7786 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7790 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7792 (rs6000_supported_cpu_names): New static variable.
7793 (linux_cpu_translation_table): Likewise.
7794 (elf_platform) <cpu>: Define new static variable and use it.
7795 Translate kernel AT_PLATFORM name to canonical name if needed.
7796 Error if platform name is unknown.
7798 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7801 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7803 2018-02-01 Jeff Law <law@redhat.com>
7806 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7807 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7808 the scratch if RELEASE_VIA_POP is false.
7809 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7810 If we have to save a temporary register, decrement SIZE appropriately.
7811 Pass new arguments to release_scratch_register_on_entry.
7812 (ix86_adjust_stack_and_probe): Likewise.
7813 (ix86_emit_probe_stack_range): Pass new arguments to
7814 release_scratch_register_on_entry.
7816 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7818 PR rtl-optimization/84157
7819 * combine.c (change_zero_ext): Use REG_P predicate in
7820 front of HARD_REGISTER_P predicate.
7822 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7824 * config/avr/avr.c (avr_option_override): Move disabling of
7825 -fdelete-null-pointer-checks to...
7826 * common/config/avr/avr-common.c (avr_option_optimization_table):
7829 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7831 PR tree-optimization/81635
7832 * tree-data-ref.c (split_constant_offset_1): For types that
7833 wrap on overflow, try to use range info to prove that wrapping
7836 2018-02-01 Renlin Li <renlin.li@arm.com>
7839 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7841 (aarch64_register_move_cost): Likewise.
7842 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7844 (REG_CLASS_NAMES): Likewise.
7845 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7846 TAILCALL_ADDR_REGS. Remove IP registers.
7847 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7849 2018-02-01 Richard Biener <rguenther@suse.de>
7851 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7852 for specifying RPO order and allow NULL for that.
7853 * domwalk.c (dom_walker::dom_walker): Likewise.
7854 (dom_walker::walk): Handle NULL RPO order.
7855 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7857 (rewrite_update_dom_walker): Likewise.
7858 (mark_def_dom_walker): Likewise.
7860 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7862 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7863 (aarch64_maybe_expand_sve_subreg_move): Declare.
7864 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7865 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7867 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7868 that are semantically a reverse operation.
7869 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7870 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7871 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7873 (aarch64_can_change_mode_class): For big-endian, forbid changes
7874 between two SVE modes if they have different element sizes.
7876 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7878 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7879 the TImode handling for big-endian targets.
7881 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7883 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7884 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7886 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7887 Remove BSWAP handing for big-endian targets and use the form of
7888 LD1RQ appropariate for the mode.
7890 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7892 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7893 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7896 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7899 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7900 check for operands that need to go through aarch64_sve_reload_be.
7902 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7904 PR tree-optimization/81661
7905 PR tree-optimization/84117
7906 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7907 * tree-eh.c: Include gimplify.h.
7908 (find_trapping_overflow, replace_trapping_overflow,
7909 rewrite_to_non_trapping_overflow): New functions.
7910 * tree-vect-loop.c: Include tree-eh.h.
7911 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7912 * tree-data-ref.c: Include tree-eh.h.
7913 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7915 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7917 PR rtl-optimization/84123
7918 * combine.c (change_zero_ext): Check if hard register satisfies
7919 can_change_dest_mode before calling gen_lowpart_SUBREG.
7921 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7924 * ira.c (ira_init_register_move_cost): Remove assert.
7926 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7928 PR rtl-optimization/84071
7929 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7930 * doc/tm.texi: Regenerate.
7932 2018-01-31 Richard Biener <rguenther@suse.de>
7934 PR tree-optimization/84132
7935 * tree-data-ref.c (analyze_miv_subscript): Properly
7936 check whether evolution_function_is_affine_multivariate_p
7937 before calling gcd_of_steps_may_divide_p.
7939 2018-01-31 Julia Koval <julia.koval@intel.com>
7942 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7943 * config/i386/i386.md (rdpid_rex64) New.
7944 (rdpid): Make 32bit only.
7946 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7949 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7950 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7952 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7955 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7957 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7959 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7961 PR rtl-optimization/84071
7962 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7963 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7965 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7967 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7968 (arc_attribute_table): Add 'aux' attribute.
7969 (arc_in_small_data_p): Consider aux like variables.
7970 (arc_is_aux_reg_p): New function.
7971 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7972 (arc_get_aux_arg): New function.
7973 (prepare_move_operands): Handle aux-register access.
7974 (arc_handle_aux_attribute): New function.
7975 * doc/extend.texi (ARC Variable attributes): Add subsection.
7977 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7979 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7980 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7981 (arc_attribute_table): Add 'uncached' attribute.
7982 (arc_print_operand): Print '.di' flag for uncached memory
7984 (arc_in_small_data_p): Do not consider for small data the uncached
7986 (arc_is_uncached_mem_p): New function.
7987 * config/arc/predicates.md (compact_store_memory_operand): Check
7988 for uncached memory accesses.
7989 (nonvol_nonimm_operand): Likewise.
7990 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7992 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7995 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7996 falign-loops=): Add Optimization flag.
7998 2018-01-30 Jeff Law <law@redhat.com>
8001 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8002 INT_REGISTERS_SAVED. Check it prior to calling
8003 get_scratch_register_on_entry.
8004 (ix86_adjust_stack_and_probe): Similarly.
8005 (ix86_emit_probe_stack_range): Similarly.
8006 (ix86_expand_prologue): Corresponding changes.
8008 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8011 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8012 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8014 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8017 * lra-constraints.c (curr_insn_transform): Process AND in the
8020 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8022 PR rtl-optimization/83986
8023 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8024 dependence against last_pending_memory_flush in addition to
8027 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8029 PR tree-optimization/81611
8030 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8033 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8036 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8039 2018-01-30 Richard Biener <rguenther@suse.de>
8040 Jakub Jelinek <jakub@redhat.com>
8042 PR tree-optimization/84111
8043 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8044 inner loops added during recursion, as they don't have up-to-date
8047 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8050 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8051 (can_inline_edge_by_limits_p): ... here.
8052 (can_early_inline_edge_p, check_callers,
8053 update_caller_keys, update_callee_keys, recursive_inlining,
8054 add_new_edges_to_heap, speculation_useful_p,
8055 inline_small_functions,
8056 inline_small_functions, flatten_function,
8057 inline_to_all_callers_1): Update.
8059 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8061 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8064 2018-01-30 Richard Biener <rguenther@suse.de>
8066 PR tree-optimization/83008
8067 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8068 invariant and constant vector uses in stmts when they need
8071 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8074 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8075 * configure: Regenerate.
8077 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8079 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8081 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8082 Use gen_rtx_REG rather than gen_lowpart.
8084 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8086 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8087 rather than 0 when creating partial subregs.
8089 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8091 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8094 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8097 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8098 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8099 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8100 flags. This restores the settings used before the 2017-07-24.
8101 Turning off pre increment/decrement/modify allows IVOPTS to
8102 optimize DF/SF loops where the index is an int.
8104 2018-01-29 Richard Biener <rguenther@suse.de>
8105 Kelvin Nilsen <kelvin@gcc.gnu.org>
8108 * tree-vect-stmts.c (vectorizable_call): Don't call
8109 targetm.vectorize_builtin_md_vectorized_function if callee is
8112 2018-01-22 Carl Love <cel@us.ibm.com>
8114 * doc/extend.tex: Fix typo in second arg in
8115 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8117 2018-01-29 Richard Biener <rguenther@suse.de>
8119 PR tree-optimization/84086
8120 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8121 (flush_ssaname_freelist): When SSA names were released reset
8122 the SCEV hash table.
8124 2018-01-29 Richard Biener <rguenther@suse.de>
8126 PR tree-optimization/84057
8127 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8128 removed paths when removing edges.
8130 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8132 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8133 -mfunction-return=@var{choice}.
8135 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8138 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8140 (layout::print_source_line): Likewise.
8141 (test_get_line_width_without_trailing_whitespace): Add test cases.
8143 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8146 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8149 2018-01-26 Jim Wilson <jimw@sifive.com>
8151 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8153 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8156 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8158 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8159 and CMP + SUB-immediate -> SUBS.
8161 2018-01-26 Martin Sebor <msebor@redhat.com>
8163 PR tree-optimization/83896
8164 * tree-ssa-strlen.c (get_string_len): Rename...
8165 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8166 Avoid assuming length is constant.
8167 (handle_char_store): Use HOST_WIDE_INT for string length.
8169 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8172 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8173 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8175 2018-01-26 Richard Biener <rguenther@suse.de>
8177 PR rtl-optimization/84003
8178 * dse.c (record_store): Only record redundant stores when
8179 the earlier store aliases at least all accesses the later one does.
8181 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8183 PR rtl-optimization/83985
8184 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8185 REG_CFA_RESTORE insns.
8186 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8187 REG_CFA_RESTORE insns here.
8190 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8191 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8193 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8195 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8196 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8197 (arc_init): Likewise.
8198 (arc_override_options): Likewise.
8199 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8201 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8203 * config/arc/arc.h (TARGET_DBNZ): Define.
8204 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8205 properly set the tune attribute.
8206 (dbnz): Use TARGET_DBNZ guard.
8207 * config/arc/arc.opt (mtune): Add core3 option.
8209 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8211 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8212 recognize new pic like addresses.
8213 (arc_delegitimize_address): Clean up.
8215 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8217 * config/arc/arc-arches.def: Option mrf16 valid for all
8219 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8220 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8221 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8222 * config/arc/arc-tables.opt: Regenerate.
8223 * config/arc/arc.c (arc_conditional_register_usage): Handle
8224 reduced register file case.
8225 (arc_file_start): Set must have build attributes.
8226 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8228 * config/arc/arc.opt (mrf16): Add new option.
8229 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8230 * config/arc/genmultilib.awk: Handle new mrf16 option.
8231 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8232 * config/arc/t-multilib: Regenerate.
8233 * doc/invoke.texi (ARC Options): Document mrf16 option.
8235 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8237 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8238 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8239 (arc_attribute_table): Add 'secure_call' attribute.
8240 (arc_print_operand): Print secure call operand.
8241 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8243 (arc_is_secure_call_p): New function. * config/arc/arc.md
8244 (call_i): Add support for sjli instruction.
8245 (call_value_i): Likewise.
8246 * config/arc/constraints.md (Csc): New constraint.
8248 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8249 John Eric Martin <John.Martin@emmicro-us.com>
8251 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8252 * config/arc/arc.c (_arc_jli_section): New struct.
8253 (arc_jli_section): New type.
8254 (rc_jli_sections): New static variable.
8255 (arc_handle_jli_attribute): New function.
8256 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8257 (arc_file_end): New function.
8258 (TARGET_ASM_FILE_END): Define.
8259 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8260 (arc_add_jli_section): New function.
8261 (jli_call_scan): Likewise.
8262 (arc_reorg): Call jli_call_scan.
8263 (arc_output_addsi): Remove 'S' from printing asm operand.
8264 (arc_is_jli_call_p): New function.
8265 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8267 (movhi_insn): Likewise.
8268 (movsi_insn): Likewise.
8269 (movsi_set_cc_insn): Likewise.
8270 (loadqi_update): Likewise.
8271 (load_zeroextendqisi_update): Likewise.
8272 (load_signextendqisi_update): Likewise.
8273 (loadhi_update): Likewise.
8274 (load_zeroextendhisi_update): Likewise.
8275 (load_signextendhisi_update): Likewise.
8276 (loadsi_update): Likewise.
8277 (loadsf_update): Likewise.
8278 (movsicc_insn): Likewise.
8279 (bset_insn): Likewise.
8280 (bxor_insn): Likewise.
8281 (bclr_insn): Likewise.
8282 (bmsk_insn): Likewise.
8283 (bicsi3_insn): Likewise.
8284 (cmpsi_cc_c_insn): Likewise.
8285 (movsi_ne): Likewise.
8286 (movsi_cond_exec): Likewise.
8287 (clrsbsi2): Likewise.
8298 (call_i): Remove 'S' asm letter, add jli instruction.
8299 (call_value_i): Likewise.
8300 * config/arc/arc.op (mjli-always): New option.
8301 * config/arc/constraints.md (Cji): New constraint.
8302 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8304 (subsf3_fpx): Likewise.
8305 (mulsf3_fpx): Likewise.
8306 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8308 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8310 * doc/invoke.texi (ARC): Document mjli-always option.
8312 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8314 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8315 avoid addition with 0 and use incw and decw where possible.
8317 2018-01-26 Richard Biener <rguenther@suse.de>
8319 PR tree-optimization/81082
8320 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8321 association if it requires casting to unsigned.
8322 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8323 from fold_plusminus_mult_expr to catch important cases late when
8324 range info is available.
8326 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8328 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8329 * configure.ac (hidden_linkonce): New test.
8330 * configure: Regenerate.
8331 * config.in: Regenerate.
8333 2018-01-26 Julia Koval <julia.koval@intel.com>
8335 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8336 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8337 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8338 _mm_mask_bitshuffle_epi64_mask): Fix type.
8339 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8340 USI_FTYPE_V4DI_V4DI_USI): Remove.
8341 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8342 __builtin_ia32_vpshufbitqmb256_mask,
8343 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8344 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8345 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8347 2018-01-26 Alan Modra <amodra@gmail.com>
8350 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8351 UNSPEC_VBPERMQ. Sort other unspecs.
8353 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8355 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8357 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8360 * predict.c (drop_profile): Do not push/pop cfun; update also
8362 (handle_missing_profiles): Fix logic looking for zero profiles.
8364 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8367 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8368 on functions with #pragma omp declare simd or functions with simd
8370 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8371 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8372 Remove trailing \n from warning_at calls.
8374 2018-01-25 Tom de Vries <tom@codesourcery.com>
8377 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8378 for neutered workers.
8380 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8383 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8384 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8386 2018-01-24 Jeff Law <law@redhat.com>
8389 * i386.c (get_probe_interval): Move to earlier point.
8390 (ix86_compute_frame_layout): If -fstack-clash-protection and
8391 the frame is larger than the probe interval, then use pushes
8392 to save registers rather than reg->mem moves.
8393 (ix86_expand_prologue): Remove conditional for int_registers_saved
8396 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8399 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8400 min/max for never referenced object.
8402 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8405 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8407 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8408 attributes from DECL_ATTRIBUTES (decl) without affecting
8409 DECL_ATTRIBUTES (current_function_decl).
8410 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8411 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8413 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8415 PR tree-optimization/83979
8416 * fold-const.c (fold_comparison): Use constant_boolean_node
8417 instead of boolean_{true,false}_node.
8419 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8421 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8424 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8426 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8427 Simplify the clause that sets the length attribute.
8428 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8429 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8430 clause that sets the length attribute.
8431 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8433 2018-01-24 Tom de Vries <tom@codesourcery.com>
8436 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8437 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8438 Add strict parameter.
8439 (prevent_branch_around_nothing): Insert dummy insn between branch to
8440 label and label with no ptx insn inbetween.
8441 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8443 2018-01-24 Tom de Vries <tom@codesourcery.com>
8446 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8447 for neutered threads in warp.
8448 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8450 2018-01-24 Richard Biener <rguenther@suse.de>
8452 PR tree-optimization/83176
8453 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8456 2018-01-24 Richard Biener <rguenther@suse.de>
8458 PR tree-optimization/82819
8459 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8460 code generating pluses that are no-ops in the target precision.
8462 2018-01-24 Richard Biener <rguenther@suse.de>
8465 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8467 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8469 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8470 to merge probabilities.
8471 * predict.c (probably_never_executed): Also mark as cold functions
8472 with global 0 profile and guessed local profile.
8473 * profile-count.c (profile_probability::combine_with_count): New
8475 * profile-count.h (profile_probability::operator*,
8476 profile_probability::operator*=, profile_probability::operator/,
8477 profile_probability::operator/=): Reduce precision to adjusted
8478 and set value to guessed on contradictory divisions.
8479 (profile_probability::combine_with_freq): Remove.
8480 (profile_probability::combine_wiht_count): Declare.
8481 (profile_count::force_nonzero):: Set to adjusted.
8482 (profile_count::probability_in):: Set quality to adjusted.
8483 * tree-ssa-tail-merge.c (replace_block_by): Use
8486 2018-01-23 Andrew Waterman <andrew@sifive.com>
8487 Jim Wilson <jimw@sifive.com>
8489 * config/riscv/riscv.c (riscv_stack_boundary): New.
8490 (riscv_option_override): Set riscv_stack_boundary. Handle
8491 riscv_preferred_stack_boundary_arg.
8492 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8493 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8494 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8495 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8496 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8497 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8499 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8502 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8503 of struct ix86_frame.
8504 (ix86_expand_epilogue): Likewise. Add a local variable for
8505 the reg_save_offset field in struct ix86_frame.
8507 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8509 PR tree-optimization/82604
8510 * tree-loop-distribution.c (enum partition_kind): New enum item
8511 PKIND_PARTIAL_MEMSET.
8512 (partition_builtin_p): Support above new enum item.
8513 (generate_code_for_partition): Ditto.
8514 (compute_access_range): Differentiate cases that equality can be
8515 proven at all loops, the innermost loops or no loops.
8516 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8517 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8518 (finalize_partitions, distribute_loop): Don't fuse partition of
8519 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8520 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8523 2018-01-23 Martin Liska <mliska@suse.cz>
8525 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8526 order to ignore the predictor.
8527 (PRED_POLYMORPHIC_CALL): Likewise.
8528 (PRED_RECURSIVE_CALL): Likewise.
8530 2018-01-23 Martin Liska <mliska@suse.cz>
8532 * tree-profile.c (tree_profiling): Print function header to
8533 aware reader which function we are working on.
8534 * value-prof.c (gimple_find_values_to_profile): Do not print
8535 not interesting value histograms.
8537 2018-01-23 Martin Liska <mliska@suse.cz>
8539 * profile-count.h (enum profile_quality): Add
8540 profile_uninitialized as the first value. Do not number values
8541 as they are zero based.
8542 (profile_count::verify): Update sanity check.
8543 (profile_probability::verify): Likewise.
8545 2018-01-23 Nathan Sidwell <nathan@acm.org>
8547 * doc/invoke.texi (ffor-scope): Deprecate.
8549 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8551 PR tree-optimization/83510
8552 * domwalk.c (set_all_edges_as_executable): New function.
8553 (dom_walker::dom_walker): Convert bool param
8554 "skip_unreachable_blocks" to enum reachability. Move setup of
8555 edge flags to set_all_edges_as_executable and only do it when
8556 reachability is REACHABLE_BLOCKS.
8557 * domwalk.h (enum dom_walker::reachability): New enum.
8558 (dom_walker::dom_walker): Convert bool param
8559 "skip_unreachable_blocks" to enum reachability.
8560 (set_all_edges_as_executable): New decl.
8561 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8562 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8564 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8565 but converting true to REACHABLE_BLOCKS.
8566 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8568 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8569 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8570 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8572 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8573 if check_all_array_refs will be called.
8575 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8577 * tree.c (selftest::test_location_wrappers): Add more test
8580 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8582 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8583 (selftest::test_bit_in_range): Likewise.
8585 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8588 * doc/sourcebuild.texi (vect_float): Say that the selector
8589 only describes the situation when -funsafe-math-optimizations is on.
8590 (vect_float_strict): Document.
8592 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8594 PR tree-optimization/83965
8595 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8596 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8597 instead of checking only for a reduction.
8598 (vect_recog_widen_sum_pattern): Likewise.
8600 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8602 * predict.c (probably_never_executed): Only use precise profile info.
8603 (compute_function_frequency): Skip after inlining hack since we now
8604 have quality checking.
8606 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8608 * profile-count.h (profile_probability::very_unlikely,
8609 profile_probability::unlikely, profile_probability::even): Set
8610 precision to guessed.
8612 2018-01-23 Richard Biener <rguenther@suse.de>
8614 PR tree-optimization/83963
8615 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8616 Properly terminate dominator walk when crossing the exit edge not
8617 when visiting its source block.
8619 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8622 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8623 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8625 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8627 PR tree-optimization/83957
8628 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8629 semicolon after for body surrounded by braces.
8631 PR tree-optimization/83081
8632 * profile-count.h (profile_probability::split): New method.
8633 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8634 Use profile_probability::split.
8635 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8636 when splitting a single conditional jump into 2.
8638 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8640 PR tree-optimization/69452
8641 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8644 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8646 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8647 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8648 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8650 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8652 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8653 * config/rl78/rl78.md (movdi): New define_expand.
8654 * config/rl78/rl78.c (rl78_split_movdi): New function.
8656 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8659 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8661 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8662 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8663 128-bit to produce an UNSPEC move to get the double word with the
8664 signbit and then a shift directly to do signbit.
8665 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8666 implementation with a new version that just does either a direct
8667 move or a regular move. Move memory interface to separate insns.
8668 Move insns so they are next to the expander.
8669 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8670 with signbit move. Split big and little endian case.
8671 (signbit<mode>2_dm_mem_le): Likewise.
8672 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8673 (signbit<mode>2_dm2): Likewise.
8675 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8677 * config/rl78/rl78.md (anddi3): New define_expand.
8679 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8681 * config/rl78/rl78.md (umindi3): New define_expand.
8683 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8685 * config/rl78/rl78.md (smindi3): New define_expand.
8687 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8689 * config/rl78/rl78.md (smaxdi3): New define_expand.
8691 2018-01-22 Carl Love <cel@us.ibm.com>
8693 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8694 LVX_V1TI): Add macro expansion.
8695 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8696 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8697 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8698 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8699 Change check to determine if the instruction is a byte reversing
8700 entry. Fix typo in comment.
8701 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8702 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8703 Add def_builtin calls for new builtins.
8704 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8705 Add define_insn expansion.
8707 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8709 * config/rl78/rl78.md (umaxdi3): New define_expand.
8711 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8713 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8714 for non-QImode registers.
8716 2018-01-22 Richard Biener <rguenther@suse.de>
8718 PR tree-optimization/83963
8719 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8720 including the loop exit block.
8721 (scop_detection::merge_sese): Likewise.
8722 (scop_detection::add_scop): Do it here instead.
8724 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8726 * doc/sourcebuild.texi (arm_softfloat): Document.
8728 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8731 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8732 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8733 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8735 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8736 David Edelsohn <dje.gcc@gmail.com>
8739 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8740 Change "crset eq" to "crset 2".
8741 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8742 (*call_indirect_aix<mode>_nospec): Likewise.
8743 (*call_value_indirect_aix<mode>_nospec): Likewise.
8744 (*call_indirect_elfv2<mode>_nospec): Likewise.
8745 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8746 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8747 change assembly output from . to $.
8748 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8749 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8750 (*tablejump<mode>_internal1_nospec): Likewise.
8752 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8755 * config/sh/sh_optimize_sett_clrt.cc:
8756 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8758 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8760 PR tree-optimization/83940
8761 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8762 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8763 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8764 use it to pass back the definition type.
8765 (vect_check_store_rhs): Likewise rhs_dt_out.
8766 (vect_build_gather_load_calls): Add a mask_dt argument and use
8767 it instead of a call to vect_is_simple_use.
8768 (vectorizable_store): Update calls to vect_check_load_store_mask
8769 and vect_check_store_rhs. Use the dt returned by the latter instead
8770 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8771 instead of calls to vect_is_simple_use. Pass the scalar rather
8772 than the vector operand to vect_is_simple_use when handling
8773 second and subsequent copies of an rhs value.
8774 (vectorizable_load): Update calls to vect_check_load_store_mask
8775 and vect_build_gather_load_calls. Use the cached mask_dt and
8776 gs_info.offset_dt instead of calls to vect_is_simple_use.
8778 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8781 * tree-emutls.c: Include gimplify.h.
8782 (lower_emutls_2): New function.
8783 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8784 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8785 it before further processing.
8788 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8789 UINTVAL (trueop1) instead of INTVAL (op1).
8791 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8795 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8796 INCOMING_FRAME_SP_OFFSET if not defined.
8797 (scan_trace): Add ENTRY argument. If true and
8798 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8799 emit a note to adjust the CFA offset.
8800 (create_cfi_notes): Adjust scan_trace callers.
8801 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8802 INCOMING_FRAME_SP_OFFSET in the CIE.
8803 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8804 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8806 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8807 * doc/tm.texi: Regenerated.
8809 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8811 PR rtl-optimization/83147
8812 * lra-constraints.c (remove_inheritance_pseudos): Use
8813 lra_substitute_pseudo_within_insn.
8815 2018-01-19 Tom de Vries <tom@codesourcery.com>
8816 Cesar Philippidis <cesar@codesourcery.com>
8819 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8821 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8824 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8825 spaces for function labels.
8827 2018-01-19 Martin Liska <mliska@suse.cz>
8829 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8830 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8831 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8832 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8833 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8834 (PRED_CONST_RETURN): Change from 69 to 65.
8835 (PRED_NULL_RETURN): Change from 91 to 71.
8836 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8837 (PRED_LOOP_GUARD): Change from 66 to 73.
8839 2018-01-19 Martin Liska <mliska@suse.cz>
8841 * predict.c (predict_insn_def): Add new assert.
8842 (struct branch_predictor): Change type to signed integer.
8843 (test_prediction_value_range): Amend test to cover
8845 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8846 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8847 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8848 (PRED_LOOP_IV_COMPARE): Likewise.
8849 * predict.h (PROB_UNINITIALIZED): Define new constant.
8851 2018-01-19 Martin Liska <mliska@suse.cz>
8853 * predict.c (dump_prediction): Add new format for
8854 analyze_brprob.py script which is enabled with -details
8856 * profile-count.h (precise_p): New function.
8858 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8860 PR tree-optimization/83922
8861 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8862 there are no statements that need masking.
8863 (vect_active_double_reduction_p): New function.
8864 (vect_analyze_loop_operations): Use it when handling phis that
8865 are not in the loop header.
8867 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8869 PR tree-optimization/83914
8870 * tree-vect-loop.c (vectorizable_induction): Don't convert
8871 init_expr or apply the peeling adjustment for inductions
8872 that are nested within the vectorized loop.
8874 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8876 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8879 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8883 * function.h (gimplify_parameters): Add gimple_seq * argument.
8884 * function.c: Include gimple.h and options.h.
8885 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8886 for the added local temporaries if needed.
8887 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8888 if there are any parameter cleanups, wrap whole body into a
8889 try/finally with the cleanups.
8891 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8894 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8895 Use GET_MODE_CLASS for scalar floating point.
8897 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8901 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8902 Fix call of call_cgraph_insertion_hooks.
8904 2018-01-18 Martin Sebor <msebor@redhat.com>
8906 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8908 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8911 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8914 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8917 * common.opt: (-ffile-prefix-map): New option.
8918 * opts.c (common_handle_option): Defer it.
8919 * opts-global.c (handle_common_deferred_options): Handle it.
8920 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8921 * file-prefix-map.h: New file.
8922 (remap_debug_filename, add_debug_prefix_map): ...here.
8923 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8924 * final.c (debug_prefix_map, add_debug_prefix_map
8925 remap_debug_filename): Move to...
8926 * file-prefix-map.c: New file.
8927 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8928 generalize, get rid of alloca(), use strrchr() instead of strchr().
8929 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8930 Implement in terms of add_prefix_map().
8931 (remap_macro_filename, remap_debug_filename): Implement in term of
8933 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8934 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8935 * dbxout.c: Include file-prefix-map.h.
8936 * varasm.c: Likewise.
8937 * vmsdbgout.c: Likewise.
8938 * xcoffout.c: Likewise.
8939 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8940 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8941 * doc/invoke.texi (-ffile-prefix-map): Document.
8942 (-fdebug-prefix-map): Update description.
8944 2018-01-18 Martin Liska <mliska@suse.cz>
8946 * config/i386/i386.c (indirect_thunk_name): Document that also
8948 (output_indirect_thunk): Document why both instructions
8949 (pause and lfence) are generated.
8951 2018-01-18 Richard Biener <rguenther@suse.de>
8953 PR tree-optimization/83887
8954 * graphite-scop-detection.c
8955 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8956 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8957 (scop_detection::merge_sese): Re-implement with a flood-fill
8958 algorithm that properly finds a SESE region if it exists.
8960 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8963 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8964 pointer_diff optimizations use view_convert instead of convert.
8966 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8968 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8969 Generate different code for -mno-speculate-indirect-jumps.
8970 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8971 (*call_indirect_aix<mode>): Disable for
8972 -mno-speculate-indirect-jumps.
8973 (*call_indirect_aix<mode>_nospec): New define_insn.
8974 (*call_value_indirect_aix<mode>): Disable for
8975 -mno-speculate-indirect-jumps.
8976 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8977 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8978 -mno-speculate-indirect-jumps.
8979 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8981 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8983 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8984 long double type, set the flags for noting the default long double
8985 type, even if we don't pass or return a long double type.
8987 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8990 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8993 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8995 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8996 support for merge[hl].
8997 (fold_mergehl_helper): New helper function.
8998 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8999 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9000 (altivec_vmrglw_direct): Add xxmrglw insn.
9002 2018-01-17 Andrew Waterman <andrew@sifive.com>
9004 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9005 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9007 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9010 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9011 call the lto_location_cache before reading the
9012 DECL_SOURCE_LOCATION of the types.
9014 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9015 Richard Sandiford <richard.sandiford@linaro.org>
9017 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9018 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9019 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9020 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9021 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9023 * config/aarch64/constraints.md (aarch64_movti_operand):
9025 * config/aarch64/predicates.md (Uti): Add new constraint.
9027 2018-01-17 Carl Love <cel@us.ibm.com>
9029 * config/rs6000/vsx.md (define_expand xl_len_r,
9030 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9031 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9033 (define_expand, define_insn): Move the shift left from the
9034 define_insn to the define_expand for lxvl and stxvl instructions.
9035 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9036 and XL_LEN_R definitions to PURE.
9038 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9040 * config/i386/i386.c (indirect_thunk_name): Declare regno
9041 as unsigned int. Compare regno with INVALID_REGNUM.
9042 (output_indirect_thunk): Ditto.
9043 (output_indirect_thunk_function): Ditto.
9044 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9045 in the call to output_indirect_thunk_function.
9047 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9050 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9051 rather than the size of inner_type to determine the stack slot size
9052 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9054 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9057 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9060 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9062 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9063 endian Linux systems to optionally enable multilibs for selecting
9064 the long double type if the user configured an explicit type.
9065 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9066 have no long double multilibs if not defined.
9067 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9068 warn if the user used -mabi={ieee,ibm}longdouble and we built
9069 multilibs for long double.
9070 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9071 appropriate multilib option.
9072 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9074 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9075 for building long double multilibs.
9076 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9078 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9080 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9083 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9085 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9088 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9091 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9094 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9096 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9097 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9099 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9101 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9102 different rtl trees depending on TARGET_64BIT.
9103 (rs6000_gen_lvx): Likewise.
9105 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9107 * config/visium/visium.md (nop): Tweak comment.
9108 (hazard_nop): Likewise.
9110 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9112 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9113 -mspeculate-indirect-jumps.
9114 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9115 for -mno-speculate-indirect-jumps.
9116 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9117 (*call_value_indirect_elfv2<mode>): Disable for
9118 -mno-speculate-indirect-jumps.
9119 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9120 (indirect_jump): Emit different RTL for
9121 -mno-speculate-indirect-jumps.
9122 (*indirect_jump<mode>): Disable for
9123 -mno-speculate-indirect-jumps.
9124 (*indirect_jump<mode>_nospec): New define_insn.
9125 (tablejump): Emit different RTL for
9126 -mno-speculate-indirect-jumps.
9127 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9128 (tablejumpsi_nospec): New define_expand.
9129 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9130 (tablejumpdi_nospec): New define_expand.
9131 (*tablejump<mode>_internal1): Disable for
9132 -mno-speculate-indirect-jumps.
9133 (*tablejump<mode>_internal1_nospec): New define_insn.
9134 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9137 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9139 * caller-save.c (insert_save): Drop unnecessary parameter. All
9142 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9143 Richard Biener <rguenth@suse.de>
9146 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9147 return early, inline manually is_gimple_sizepos. Make sure if we
9148 call gimplify_expr we don't end up with a gimple constant.
9149 * tree.c (variably_modified_type_p): Don't return true for
9150 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9151 * gimplify.h (is_gimple_sizepos): Remove.
9153 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9155 PR tree-optimization/83857
9156 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9157 vectorizable_live_operation for pure SLP statements.
9158 (vectorizable_live_operation): Handle PHIs.
9160 2018-01-16 Richard Biener <rguenther@suse.de>
9162 PR tree-optimization/83867
9163 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9164 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9166 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9169 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9170 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9171 If off is not INTEGER_CST, issue a may not be aligned warning
9172 rather than isn't aligned. Use isn%'t rather than isn't.
9173 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9175 <case MULT_EXPR>: Improve the case when bottom and one of the
9176 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9177 operand, in that case check if the other operand is multiple of
9178 bottom divided by the INTEGER_CST operand.
9180 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9183 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9184 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9185 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9186 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9187 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9188 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9189 (pa_function_arg_size): New function.
9191 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9193 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9194 in a separate statement.
9196 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9198 PR tree-optimization/83847
9199 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9200 group gathers and scatters.
9202 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9204 PR rtl-optimization/86620
9205 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9207 PR rtl-optimization/83213
9208 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9209 to last if both are JUMP_INSNs.
9211 PR tree-optimization/83843
9212 * gimple-ssa-store-merging.c
9213 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9214 store_immediate_info for bswap/nop orig_stores.
9216 2018-01-15 Andrew Waterman <andrew@sifive.com>
9218 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9220 <UDIV>: Increase cost if !TARGET_DIV.
9222 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9224 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9225 (define_attr "cr_logical_3op"): New.
9226 (cceq_ior_compare): Adjust.
9227 (cceq_ior_compare_complement): Adjust.
9228 (*cceq_rev_compare): Adjust.
9229 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9230 (is_cracked_insn): Adjust.
9231 (insn_must_be_first_in_group): Adjust.
9232 * config/rs6000/40x.md: Adjust.
9233 * config/rs6000/440.md: Adjust.
9234 * config/rs6000/476.md: Adjust.
9235 * config/rs6000/601.md: Adjust.
9236 * config/rs6000/603.md: Adjust.
9237 * config/rs6000/6xx.md: Adjust.
9238 * config/rs6000/7450.md: Adjust.
9239 * config/rs6000/7xx.md: Adjust.
9240 * config/rs6000/8540.md: Adjust.
9241 * config/rs6000/cell.md: Adjust.
9242 * config/rs6000/e300c2c3.md: Adjust.
9243 * config/rs6000/e500mc.md: Adjust.
9244 * config/rs6000/e500mc64.md: Adjust.
9245 * config/rs6000/e5500.md: Adjust.
9246 * config/rs6000/e6500.md: Adjust.
9247 * config/rs6000/mpc.md: Adjust.
9248 * config/rs6000/power4.md: Adjust.
9249 * config/rs6000/power5.md: Adjust.
9250 * config/rs6000/power6.md: Adjust.
9251 * config/rs6000/power7.md: Adjust.
9252 * config/rs6000/power8.md: Adjust.
9253 * config/rs6000/power9.md: Adjust.
9254 * config/rs6000/rs64.md: Adjust.
9255 * config/rs6000/titan.md: Adjust.
9257 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9259 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9260 ix86_indirect_branch_register logic.
9262 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9264 * config/i386/constraints.md (Bs): Update
9265 ix86_indirect_branch_register check. Don't check
9266 ix86_indirect_branch_register with GOT_memory_operand.
9268 * config/i386/predicates.md (GOT_memory_operand): Don't check
9269 ix86_indirect_branch_register here.
9270 (GOT32_symbol_operand): Likewise.
9272 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9274 * config/i386/predicates.md (constant_call_address_operand):
9275 Rewrite ix86_indirect_branch_register logic.
9276 (sibcall_insn_operand): Likewise.
9278 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9280 * config/i386/constraints.md (Bs): Replace
9281 ix86_indirect_branch_thunk_register with
9282 ix86_indirect_branch_register.
9284 * config/i386/i386.md (indirect_jump): Likewise.
9285 (tablejump): Likewise.
9286 (*sibcall_memory): Likewise.
9287 (*sibcall_value_memory): Likewise.
9288 Peepholes of indirect call and jump via memory: Likewise.
9289 * config/i386/i386.opt: Likewise.
9290 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9291 (GOT_memory_operand): Likewise.
9292 (call_insn_operand): Likewise.
9293 (sibcall_insn_operand): Likewise.
9294 (GOT32_symbol_operand): Likewise.
9296 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9299 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9300 type rather than type addr's type points to.
9301 (expand_omp_atomic_mutex): Likewise.
9302 (expand_omp_atomic): Likewise.
9304 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9307 * config/i386/i386.c (output_indirect_thunk_function): Use
9308 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9309 for __x86_return_thunk.
9311 2018-01-15 Richard Biener <rguenther@suse.de>
9314 * expmed.c (extract_bit_field_1): Fix typo.
9316 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9319 * config/arm/iterators.md (VF): New mode iterator.
9320 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9321 Remove integer-related logic from pattern.
9322 (neon_vabd<mode>_3): Likewise.
9324 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9327 * common.opt (fstrict-overflow): No longer an alias.
9328 (fwrapv-pointer): New option.
9329 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9330 also for pointer types based on flag_wrapv_pointer.
9331 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9332 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9333 opts->x_flag_wrapv got set.
9334 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9335 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9336 POINTER_TYPE_OVERFLOW_UNDEFINED.
9337 * match.pd: Likewise in address comparison pattern.
9338 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9340 2018-01-15 Richard Biener <rguenther@suse.de>
9343 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9344 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9345 Reset type names to their identifier if their TYPE_DECL doesn't
9346 have linkage (and thus is used for ODR and devirt).
9347 (save_debug_info_for_decl): Remove.
9348 (save_debug_info_for_type): Likewise.
9349 (add_tree_to_fld_list): Adjust.
9350 * tree-pretty-print.c (dump_generic_node): Make dumping of
9351 type names more robust.
9353 2018-01-15 Richard Biener <rguenther@suse.de>
9355 * BASE-VER: Bump to 8.0.1.
9357 2018-01-14 Martin Sebor <msebor@redhat.com>
9360 * builtins.c (check_access): Avoid warning when the no-warning bit
9363 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9365 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9366 * ira-color (allocno_hard_regs_compare): Likewise.
9368 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9371 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9372 Use .pushsection/.popsection.
9374 2018-01-14 Martin Sebor <msebor@redhat.com>
9377 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9379 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9381 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9382 entry from extra_headers.
9383 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9384 extra_headers, make the list bitwise identical to the i?86-*-* one.
9386 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9388 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9389 -mcmodel=large with -mindirect-branch=thunk,
9390 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9391 -mfunction-return=thunk-extern.
9392 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9393 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9394 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9396 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9398 * config/i386/i386.c (print_reg): Print the name of the full
9399 integer register without '%'.
9400 (ix86_print_operand): Handle 'V'.
9401 * doc/extend.texi: Document 'V' modifier.
9403 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9405 * config/i386/constraints.md (Bs): Disallow memory operand for
9406 -mindirect-branch-register.
9408 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9409 (GOT_memory_operand): Likewise.
9410 (call_insn_operand): Likewise.
9411 (sibcall_insn_operand): Likewise.
9412 (GOT32_symbol_operand): Likewise.
9413 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9414 for -mindirect-branch-register.
9415 (tablejump): Likewise.
9416 (*sibcall_memory): Likewise.
9417 (*sibcall_value_memory): Likewise.
9418 Disallow peepholes of indirect call and jump via memory for
9419 -mindirect-branch-register.
9420 (*call_pop): Replace m with Bw.
9421 (*call_value_pop): Likewise.
9422 (*sibcall_pop_memory): Replace m with Bs.
9423 * config/i386/i386.opt (mindirect-branch-register): New option.
9424 * doc/invoke.texi: Document -mindirect-branch-register option.
9426 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9428 * config/i386/i386-protos.h (ix86_output_function_return): New.
9429 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9430 set function_return_type.
9431 (indirect_thunk_name): Add ret_p to indicate thunk for function
9433 (output_indirect_thunk_function): Pass false to
9434 indirect_thunk_name.
9435 (ix86_output_indirect_branch_via_reg): Likewise.
9436 (ix86_output_indirect_branch_via_push): Likewise.
9437 (output_indirect_thunk_function): Create alias for function
9438 return thunk if regno < 0.
9439 (ix86_output_function_return): New function.
9440 (ix86_handle_fndecl_attribute): Handle function_return.
9441 (ix86_attribute_table): Add function_return.
9442 * config/i386/i386.h (machine_function): Add
9443 function_return_type.
9444 * config/i386/i386.md (simple_return_internal): Use
9445 ix86_output_function_return.
9446 (simple_return_internal_long): Likewise.
9447 * config/i386/i386.opt (mfunction-return=): New option.
9448 (indirect_branch): Mention -mfunction-return=.
9449 * doc/extend.texi: Document function_return function attribute.
9450 * doc/invoke.texi: Document -mfunction-return= option.
9452 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9454 * config/i386/i386-opts.h (indirect_branch): New.
9455 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9456 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9457 with local indirect jump when converting indirect call and jump.
9458 (ix86_set_indirect_branch_type): New.
9459 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9460 (indirectlabelno): New.
9461 (indirect_thunk_needed): Likewise.
9462 (indirect_thunk_bnd_needed): Likewise.
9463 (indirect_thunks_used): Likewise.
9464 (indirect_thunks_bnd_used): Likewise.
9465 (INDIRECT_LABEL): Likewise.
9466 (indirect_thunk_name): Likewise.
9467 (output_indirect_thunk): Likewise.
9468 (output_indirect_thunk_function): Likewise.
9469 (ix86_output_indirect_branch_via_reg): Likewise.
9470 (ix86_output_indirect_branch_via_push): Likewise.
9471 (ix86_output_indirect_branch): Likewise.
9472 (ix86_output_indirect_jmp): Likewise.
9473 (ix86_code_end): Call output_indirect_thunk_function if needed.
9474 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9476 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9477 (ix86_attribute_table): Add indirect_branch.
9478 * config/i386/i386.h (machine_function): Add indirect_branch_type
9479 and has_local_indirect_jump.
9480 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9482 (tablejump): Likewise.
9483 (*indirect_jump): Use ix86_output_indirect_jmp.
9484 (*tablejump_1): Likewise.
9485 (simple_return_indirect_internal): Likewise.
9486 * config/i386/i386.opt (mindirect-branch=): New option.
9487 (indirect_branch): New.
9490 (thunk-inline): Likewise.
9491 (thunk-extern): Likewise.
9492 * doc/extend.texi: Document indirect_branch function attribute.
9493 * doc/invoke.texi: Document -mindirect-branch= option.
9495 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9498 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9500 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9502 * ipa-inline.c (want_inline_small_function_p): Return false if
9503 inlining has already failed with CIF_FINAL_ERROR.
9504 (update_caller_keys): Call want_inline_small_function_p before
9506 (update_callee_keys): Likewise.
9508 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9510 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9512 (rs6000_quadword_masked_address_p): Likewise.
9513 (quad_aligned_load_p): Likewise.
9514 (quad_aligned_store_p): Likewise.
9515 (const_load_sequence_p): Add comment to describe the outer-most loop.
9516 (mimic_memory_attributes_and_flags): New function.
9517 (rs6000_gen_stvx): Likewise.
9518 (replace_swapped_aligned_store): Likewise.
9519 (rs6000_gen_lvx): Likewise.
9520 (replace_swapped_aligned_load): Likewise.
9521 (replace_swapped_load_constant): Capitalize argument name in
9522 comment describing this function.
9523 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9524 and stores that access quad-word aligned addresses and replace
9525 with stvx or lvx instructions when appropriate.
9526 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9527 New function prototype.
9528 (rs6000_quadword_masked_address_p): Likewise.
9529 (rs6000_gen_lvx): Likewise.
9530 (rs6000_gen_stvx): Likewise.
9531 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9532 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9533 when memory address is aligned.
9534 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9535 this split to select lvx instruction when memory address is aligned.
9536 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9537 instruction when memory address is aligned.
9538 (*vsx_le_perm_load_v16qi): Likewise.
9539 (four unnamed splitters): Modify to select the stvx instruction
9540 when memory is aligned.
9542 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9544 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9545 which appears in the queue multiple times.
9547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9548 Alan Hayward <alan.hayward@arm.com>
9549 David Sherwood <david.sherwood@arm.com>
9551 * tree-vectorizer.h (vec_lower_bound): New structure.
9552 (_loop_vec_info): Add check_nonzero and lower_bounds.
9553 (LOOP_VINFO_CHECK_NONZERO): New macro.
9554 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9555 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9556 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9557 fields. Make seg_len the distance travelled, not including the
9559 (dr_direction_indicator): Declare.
9560 (dr_zero_step_indicator): Likewise.
9561 (dr_known_forward_stride_p): Likewise.
9562 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9564 (runtime_alias_check_p): Allow runtime alias checks with
9566 (operator ==): Compare access_size and align.
9567 (prune_runtime_alias_test_list): Rework for new distinction between
9568 the access_size and seg_len.
9569 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9571 (get_segment_min_max): New function.
9572 (create_intersect_range_checks): Use it.
9573 (dr_step_indicator): New function.
9574 (dr_direction_indicator): Likewise.
9575 (dr_zero_step_indicator): Likewise.
9576 (dr_known_forward_stride_p): Likewise.
9577 * tree-loop-distribution.c (data_ref_segment_size): Return
9578 DR_STEP * (niters - 1).
9579 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9581 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9582 (vect_preserves_scalar_order_p): New function, split out from...
9583 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9584 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9585 (vect_vfa_access_size): New function.
9586 (vect_vfa_align): Likewise.
9587 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9588 (dump_lower_bound): New function.
9589 (vect_check_lower_bound): Likewise.
9590 (vect_small_gap_p): Likewise.
9591 (vectorizable_with_step_bound_p): Likewise.
9592 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9593 depencies if the vectorization factor is 1. Convert the checks
9594 for nonzero steps into checks on the bounds of DR_STEP. Try using
9595 a bunds check for variable steps if the minimum required step is
9596 relatively small. Update calls to the dr_with_seg_len
9597 constructor and to vect_compile_time_alias.
9598 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9600 (vect_loop_versioning): Call it.
9601 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9603 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9605 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9606 Alan Hayward <alan.hayward@arm.com>
9607 David Sherwood <david.sherwood@arm.com>
9609 * doc/sourcebuild.texi (vect_scatter_store): Document.
9610 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9612 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9614 * genopinit.c (main): Add supports_vec_scatter_store and
9615 supports_vec_scatter_store_cached to target_optabs.
9616 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9617 IFN_MASK_SCATTER_STORE.
9618 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9620 * internal-fn.h (internal_store_fn_p): Declare.
9621 (internal_fn_stored_value_index): Likewise.
9622 * internal-fn.c (scatter_store_direct): New macro.
9623 (expand_scatter_store_optab_fn): New function.
9624 (direct_scatter_store_optab_supported_p): New macro.
9625 (internal_store_fn_p): New function.
9626 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9627 IFN_MASK_SCATTER_STORE.
9628 (internal_fn_mask_index): Likewise.
9629 (internal_fn_stored_value_index): New function.
9630 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9632 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9633 * optabs-query.c (supports_vec_scatter_store_p): New function.
9634 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9635 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9636 true for scatter stores.
9637 (vect_gather_scatter_fn_p): Handle scatter stores too.
9638 (vect_check_gather_scatter): Consider using scatter stores if
9639 supports_vec_scatter_store_p.
9640 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9642 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9643 internal_fn_stored_value_index.
9644 (check_load_store_masking): Handle scatter stores too.
9645 (vect_get_store_rhs): Make public.
9646 (vectorizable_call): Use internal_store_fn_p.
9647 (vectorizable_store): Handle scatter store internal functions.
9648 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9649 when deciding whether the end of the group has been reached.
9650 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9651 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9652 (mask_scatter_store<mode>): New insns.
9654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9655 Alan Hayward <alan.hayward@arm.com>
9656 David Sherwood <david.sherwood@arm.com>
9658 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9659 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9660 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9662 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9663 Use vect_truncate_gather_scatter_offset if we can't treat the
9664 operation as a normal gather load or scatter store.
9665 (get_group_load_store_type): Take the gather_scatter_info
9666 as argument. Try using a gather load or scatter store for
9667 single-element groups.
9668 (get_load_store_type): Update calls to get_group_load_store_type
9669 and vect_use_strided_gather_scatters_p.
9671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9672 Alan Hayward <alan.hayward@arm.com>
9673 David Sherwood <david.sherwood@arm.com>
9675 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9676 optional tree argument.
9677 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9679 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9680 but continue to use the current value as a fallback.
9681 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9682 to compare the updates.
9683 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9684 (get_load_store_type): Use it when handling a strided access.
9685 (vect_get_strided_load_store_ops): New function.
9686 (vect_get_data_ptr_increment): Likewise.
9687 (vectorizable_load): Handle strided gather loads. Always pass
9688 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9690 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9691 Alan Hayward <alan.hayward@arm.com>
9692 David Sherwood <david.sherwood@arm.com>
9694 * doc/md.texi (gather_load@var{m}): Document.
9695 (mask_gather_load@var{m}): Likewise.
9696 * genopinit.c (main): Add supports_vec_gather_load and
9697 supports_vec_gather_load_cached to target_optabs.
9698 * optabs-tree.c (init_tree_optimization_optabs): Use
9699 ggc_cleared_alloc to allocate target_optabs.
9700 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9701 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9703 * internal-fn.h (internal_load_fn_p): Declare.
9704 (internal_gather_scatter_fn_p): Likewise.
9705 (internal_fn_mask_index): Likewise.
9706 (internal_gather_scatter_fn_supported_p): Likewise.
9707 * internal-fn.c (gather_load_direct): New macro.
9708 (expand_gather_load_optab_fn): New function.
9709 (direct_gather_load_optab_supported_p): New macro.
9710 (direct_internal_fn_optab): New function.
9711 (internal_load_fn_p): Likewise.
9712 (internal_gather_scatter_fn_p): Likewise.
9713 (internal_fn_mask_index): Likewise.
9714 (internal_gather_scatter_fn_supported_p): Likewise.
9715 * optabs-query.c (supports_at_least_one_mode_p): New function.
9716 (supports_vec_gather_load_p): Likewise.
9717 * optabs-query.h (supports_vec_gather_load_p): Declare.
9718 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9719 and memory_type field.
9720 (NUM_PATTERNS): Bump to 15.
9721 * tree-vect-data-refs.c: Include internal-fn.h.
9722 (vect_gather_scatter_fn_p): New function.
9723 (vect_describe_gather_scatter_call): Likewise.
9724 (vect_check_gather_scatter): Try using internal functions for
9725 gather loads. Recognize existing calls to a gather load function.
9726 (vect_analyze_data_refs): Consider using gather loads if
9727 supports_vec_gather_load_p.
9728 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9729 (vect_get_gather_scatter_offset_type): Likewise.
9730 (vect_convert_mask_for_vectype): Likewise.
9731 (vect_add_conversion_to_patterm): Likewise.
9732 (vect_try_gather_scatter_pattern): Likewise.
9733 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9734 (vect_vect_recog_func_ptrs): Add it.
9735 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9736 internal_fn_mask_index and internal_gather_scatter_fn_p.
9737 (check_load_store_masking): Take the gather_scatter_info as an
9738 argument and handle gather loads.
9739 (vect_get_gather_scatter_ops): New function.
9740 (vectorizable_call): Check internal_load_fn_p.
9741 (vectorizable_load): Likewise. Handle gather load internal
9743 (vectorizable_store): Update call to check_load_store_masking.
9744 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9745 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9746 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9747 (aarch64_gather_scale_operand_d): New predicates.
9748 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9749 (mask_gather_load<mode>): New insns.
9751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9752 Alan Hayward <alan.hayward@arm.com>
9753 David Sherwood <david.sherwood@arm.com>
9755 * optabs.def (fold_left_plus_optab): New optab.
9756 * doc/md.texi (fold_left_plus_@var{m}): Document.
9757 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9758 * internal-fn.c (fold_left_direct): Define.
9759 (expand_fold_left_optab_fn): Likewise.
9760 (direct_fold_left_optab_supported_p): Likewise.
9761 * fold-const-call.c (fold_const_fold_left): New function.
9762 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9763 * tree-parloops.c (valid_reduction_p): New function.
9764 (gather_scalar_reductions): Use it.
9765 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9766 (vect_finish_replace_stmt): Declare.
9767 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9768 (needs_fold_left_reduction_p): New function, split out from...
9769 (vect_is_simple_reduction): ...here. Accept reductions that
9770 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9771 (vect_force_simple_reduction): Also store the reduction type in
9772 the assignment's STMT_VINFO_REDUC_TYPE.
9773 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9774 (merge_with_identity): New function.
9775 (vect_expand_fold_left): Likewise.
9776 (vectorize_fold_left_reduction): Likewise.
9777 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9778 scalar phi in place for it. Check for target support and reject
9779 cases that would reassociate the operation. Defer the transform
9780 phase to vectorize_fold_left_reduction.
9781 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9782 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9783 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9785 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9787 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9788 call to ifc_temp_var.
9790 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9791 Alan Hayward <alan.hayward@arm.com>
9792 David Sherwood <david.sherwood@arm.com>
9794 * target.def (legitimize_address_displacement): Take the original
9795 offset as a poly_int.
9796 * targhooks.h (default_legitimize_address_displacement): Update
9798 * targhooks.c (default_legitimize_address_displacement): Likewise.
9799 * doc/tm.texi: Regenerate.
9800 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9801 as an argument, moving assert of ad->disp == ad->disp_term to...
9802 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9803 Try calling targetm.legitimize_address_displacement before expanding
9804 the address rather than afterwards, and adjust for the new interface.
9805 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9806 Match the new hook interface. Handle SVE addresses.
9807 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9812 * Makefile.in (OBJS): Add early-remat.o.
9813 * target.def (select_early_remat_modes): New hook.
9814 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9815 * doc/tm.texi: Regenerate.
9816 * targhooks.h (default_select_early_remat_modes): Declare.
9817 * targhooks.c (default_select_early_remat_modes): New function.
9818 * timevar.def (TV_EARLY_REMAT): New timevar.
9819 * passes.def (pass_early_remat): New pass.
9820 * tree-pass.h (make_pass_early_remat): Declare.
9821 * early-remat.c: New file.
9822 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9824 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9826 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9827 Alan Hayward <alan.hayward@arm.com>
9828 David Sherwood <david.sherwood@arm.com>
9830 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9831 vfm1 with a bound_epilog parameter.
9832 (vect_do_peeling): Update calls accordingly, and move the prologue
9833 call earlier in the function. Treat the base bound_epilog as 0 for
9834 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9835 this base when peeling for gaps.
9836 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9837 with fully-masked loops.
9838 (vect_estimate_min_profitable_iters): Handle the single peeled
9839 iteration in that case.
9841 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9842 Alan Hayward <alan.hayward@arm.com>
9843 David Sherwood <david.sherwood@arm.com>
9845 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9846 single-element interleaving even if the size is not a power of 2.
9847 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9848 accesses for single-element interleaving if the group size is
9851 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9852 Alan Hayward <alan.hayward@arm.com>
9853 David Sherwood <david.sherwood@arm.com>
9855 * doc/md.texi (fold_extract_last_@var{m}): Document.
9856 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9857 * optabs.def (fold_extract_last_optab): New optab.
9858 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9859 * internal-fn.c (fold_extract_direct): New macro.
9860 (expand_fold_extract_optab_fn): Likewise.
9861 (direct_fold_extract_optab_supported_p): Likewise.
9862 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9863 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9864 EXTRACT_LAST_REDUCTION.
9865 (get_initial_def_for_reduction): Do not create an initial vector
9866 for EXTRACT_LAST_REDUCTION reductions.
9867 (vectorizable_reduction): Leave the scalar phi in place for
9868 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9869 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9870 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9871 transform phase to vectorizable_condition.
9872 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9874 (vect_finish_stmt_generation): ...here.
9875 (vect_finish_replace_stmt): New function.
9876 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9877 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9879 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9881 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9882 Alan Hayward <alan.hayward@arm.com>
9883 David Sherwood <david.sherwood@arm.com>
9885 * doc/md.texi (extract_last_@var{m}): Document.
9886 * optabs.def (extract_last_optab): New optab.
9887 * internal-fn.def (EXTRACT_LAST): New internal function.
9888 * internal-fn.c (cond_unary_direct): New macro.
9889 (expand_cond_unary_optab_fn): Likewise.
9890 (direct_cond_unary_optab_supported_p): Likewise.
9891 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9892 loops using EXTRACT_LAST.
9893 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9894 (extract_last_<mode>): ...this optab.
9895 (vec_extract<mode><Vel>): Update accordingly.
9897 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9898 Alan Hayward <alan.hayward@arm.com>
9899 David Sherwood <david.sherwood@arm.com>
9901 * target.def (empty_mask_is_expensive): New hook.
9902 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9903 * doc/tm.texi: Regenerate.
9904 * targhooks.h (default_empty_mask_is_expensive): Declare.
9905 * targhooks.c (default_empty_mask_is_expensive): New function.
9906 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9907 if the target says that empty masks are expensive.
9908 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9910 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9913 Alan Hayward <alan.hayward@arm.com>
9914 David Sherwood <david.sherwood@arm.com>
9916 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9917 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9918 (vect_use_loop_mask_for_alignment_p): New function.
9919 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9920 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9921 niters_skip argument. Make sure that the first niters_skip elements
9922 of the first iteration are inactive.
9923 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9924 Update call to vect_set_loop_masks_directly.
9925 (get_misalign_in_elems): New function, split out from...
9926 (vect_gen_prolog_loop_niters): ...here.
9927 (vect_update_init_of_dr): Take a code argument that specifies whether
9928 the adjustment should be added or subtracted.
9929 (vect_update_init_of_drs): Likewise.
9930 (vect_prepare_for_masked_peels): New function.
9931 (vect_do_peeling): Skip prologue peeling if we're using a mask
9932 instead. Update call to vect_update_inits_of_drs.
9933 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9935 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9936 alignment. Do not include the number of peeled iterations in
9937 the minimum threshold in that case.
9938 (vectorizable_induction): Adjust the start value down by
9939 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9940 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9941 Take the number of skipped iterations into account when calculating
9943 * tree-vect-stmts.c (vect_gen_while_not): New function.
9945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9946 Alan Hayward <alan.hayward@arm.com>
9947 David Sherwood <david.sherwood@arm.com>
9949 * doc/sourcebuild.texi (vect_fully_masked): Document.
9950 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9952 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9954 (vect_analyze_loop_2): ...here. Don't check the vectorization
9955 factor against the number of loop iterations if the loop is
9958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9959 Alan Hayward <alan.hayward@arm.com>
9960 David Sherwood <david.sherwood@arm.com>
9962 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9963 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9964 (dump_groups): Update accordingly.
9965 (iv_use::mem_type): New member variable.
9966 (address_p): New function.
9967 (record_use): Add a mem_type argument and initialize the new
9969 (record_group_use): Add a mem_type argument. Use address_p.
9970 Remove obsolete null checks of base_object. Update call to record_use.
9971 (find_interesting_uses_op): Update call to record_group_use.
9972 (find_interesting_uses_cond): Likewise.
9973 (find_interesting_uses_address): Likewise.
9974 (get_mem_type_for_internal_fn): New function.
9975 (find_address_like_use): Likewise.
9976 (find_interesting_uses_stmt): Try find_address_like_use before
9977 calling find_interesting_uses_op.
9978 (addr_offset_valid_p): Use the iv mem_type field as the type
9979 of the addressed memory.
9980 (add_autoinc_candidates): Likewise.
9981 (get_address_cost): Likewise.
9982 (split_small_address_groups_p): Use address_p.
9983 (split_address_groups): Likewise.
9984 (add_iv_candidate_for_use): Likewise.
9985 (autoinc_possible_for_pair): Likewise.
9986 (rewrite_groups): Likewise.
9987 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9988 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9989 (get_alias_ptr_type_for_ptr_address): New function.
9990 (rewrite_use_address): Rewrite address uses in calls that were
9991 identified by find_address_like_use.
9993 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9994 Alan Hayward <alan.hayward@arm.com>
9995 David Sherwood <david.sherwood@arm.com>
9997 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9999 * gimple-expr.h (is_gimple_addressable: Likewise.
10000 * gimple-expr.c (is_gimple_address): Likewise.
10001 * internal-fn.c (expand_call_mem_ref): New function.
10002 (expand_mask_load_optab_fn): Use it.
10003 (expand_mask_store_optab_fn): Likewise.
10005 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10006 Alan Hayward <alan.hayward@arm.com>
10007 David Sherwood <david.sherwood@arm.com>
10009 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10010 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10011 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10012 (cond_umax@var{mode}): Document.
10013 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10014 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10015 (cond_umin_optab, cond_umax_optab): New optabs.
10016 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10017 (COND_IOR, COND_XOR): New internal functions.
10018 * internal-fn.h (get_conditional_internal_fn): Declare.
10019 * internal-fn.c (cond_binary_direct): New macro.
10020 (expand_cond_binary_optab_fn): Likewise.
10021 (direct_cond_binary_optab_supported_p): Likewise.
10022 (get_conditional_internal_fn): New function.
10023 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10024 Cope with reduction statements that are vectorized as calls rather
10026 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10027 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10028 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10029 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10030 (UNSPEC_COND_EOR): New unspecs.
10031 (optab): Add mappings for them.
10032 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10033 (sve_int_op, sve_fp_op): New int attributes.
10035 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10036 Alan Hayward <alan.hayward@arm.com>
10037 David Sherwood <david.sherwood@arm.com>
10039 * optabs.def (while_ult_optab): New optab.
10040 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10041 * internal-fn.def (WHILE_ULT): New internal function.
10042 * internal-fn.h (direct_internal_fn_supported_p): New override
10043 that takes two types as argument.
10044 * internal-fn.c (while_direct): New macro.
10045 (expand_while_optab_fn): New function.
10046 (convert_optab_supported_p): Likewise.
10047 (direct_while_optab_supported_p): New macro.
10048 * wide-int.h (wi::udiv_ceil): New function.
10049 * tree-vectorizer.h (rgroup_masks): New structure.
10050 (vec_loop_masks): New typedef.
10051 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10052 and fully_masked_p.
10053 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10054 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10055 (vect_max_vf): New function.
10056 (slpeel_make_loop_iterate_ntimes): Delete.
10057 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10058 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10059 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10060 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10061 internal-fn.h, stor-layout.h and optabs-query.h.
10062 (vect_set_loop_mask): New function.
10063 (add_preheader_seq): Likewise.
10064 (add_header_seq): Likewise.
10065 (interleave_supported_p): Likewise.
10066 (vect_maybe_permute_loop_masks): Likewise.
10067 (vect_set_loop_masks_directly): Likewise.
10068 (vect_set_loop_condition_masked): Likewise.
10069 (vect_set_loop_condition_unmasked): New function, split out from
10070 slpeel_make_loop_iterate_ntimes.
10071 (slpeel_make_loop_iterate_ntimes): Rename to..
10072 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10073 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10074 (vect_do_peeling): Update call accordingly.
10075 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10077 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10078 mask_compare_type, can_fully_mask_p and fully_masked_p.
10079 (release_vec_loop_masks): New function.
10080 (_loop_vec_info): Use it to free the loop masks.
10081 (can_produce_all_loop_masks_p): New function.
10082 (vect_get_max_nscalars_per_iter): Likewise.
10083 (vect_verify_full_masking): Likewise.
10084 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10085 retries, and free the mask rgroups before retrying. Check loop-wide
10086 reasons for disallowing fully-masked loops. Make the final decision
10087 about whether use a fully-masked loop or not.
10088 (vect_estimate_min_profitable_iters): Do not assume that peeling
10089 for the number of iterations will be needed for fully-masked loops.
10090 (vectorizable_reduction): Disable fully-masked loops.
10091 (vectorizable_live_operation): Likewise.
10092 (vect_halve_mask_nunits): New function.
10093 (vect_double_mask_nunits): Likewise.
10094 (vect_record_loop_mask): Likewise.
10095 (vect_get_loop_mask): Likewise.
10096 (vect_transform_loop): Handle the case in which the final loop
10097 iteration might handle a partial vector. Call vect_set_loop_condition
10098 instead of slpeel_make_loop_iterate_ntimes.
10099 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10100 (check_load_store_masking): New function.
10101 (prepare_load_store_mask): Likewise.
10102 (vectorizable_store): Handle fully-masked loops.
10103 (vectorizable_load): Likewise.
10104 (supportable_widening_operation): Use vect_halve_mask_nunits for
10106 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10107 (vect_gen_while): New function.
10108 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10109 (aarch64_uqdec<mode>): New insn.
10111 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10112 Alan Hayward <alan.hayward@arm.com>
10113 David Sherwood <david.sherwood@arm.com>
10115 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10116 (reduc_xor_scal_optab): New optabs.
10117 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10118 (reduc_xor_scal_@var{m}): Document.
10119 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10120 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10121 internal functions.
10122 * fold-const-call.c (fold_const_call): Handle them.
10123 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10124 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10125 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10126 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10127 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10128 (UNSPEC_XORV): New unspecs.
10129 (optab): Add entries for them.
10130 (BITWISEV): New int iterator.
10131 (bit_reduc_op): New int attributes.
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 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10138 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10139 * optabs.def (vec_shl_insert_optab): New optab.
10140 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10141 (duplicate_and_interleave): Likewise.
10142 * tree-vect-loop.c: Include internal-fn.h.
10143 (neutral_op_for_slp_reduction): New function, split out from
10144 get_initial_defs_for_reduction.
10145 (get_initial_def_for_reduction): Handle option 2 for variable-length
10146 vectors by loading the neutral value into a vector and then shifting
10147 the initial value into element 0.
10148 (get_initial_defs_for_reduction): Replace the code argument with
10149 the neutral value calculated by neutral_op_for_slp_reduction.
10150 Use gimple_build_vector for constant-length vectors.
10151 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10152 but the first group_size elements have a neutral value.
10153 Use duplicate_and_interleave otherwise.
10154 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10155 Update call to get_initial_defs_for_reduction. Handle SLP
10156 reductions for variable-length vectors by creating one vector
10157 result for each scalar result, with the elements associated
10158 with other scalar results stubbed out with the neutral value.
10159 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10160 Require IFN_VEC_SHL_INSERT for double reductions on
10161 variable-length vectors, or SLP reductions that have
10162 a neutral value. Require can_duplicate_and_interleave_p
10163 support for variable-length unchained SLP reductions if there
10164 is no neutral value, such as for MIN/MAX reductions. Also require
10165 the number of vector elements to be a multiple of the number of
10166 SLP statements when doing variable-length unchained SLP reductions.
10167 Update call to vect_create_epilog_for_reduction.
10168 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10169 and remove initial values.
10170 (duplicate_and_interleave): Make public.
10171 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10172 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10174 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10175 Alan Hayward <alan.hayward@arm.com>
10176 David Sherwood <david.sherwood@arm.com>
10178 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10179 (can_duplicate_and_interleave_p): New function.
10180 (vect_get_and_check_slp_defs): Take the vector of statements
10181 rather than just the current one. Remove excess parentheses.
10182 Restriction rejectinon of vect_constant_def and vect_external_def
10183 for variable-length vectors to boolean types, or types for which
10184 can_duplicate_and_interleave_p is false.
10185 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10186 (duplicate_and_interleave): New function.
10187 (vect_get_constant_vectors): Use gimple_build_vector for
10188 constant-length vectors and suitable variable-length constant
10189 vectors. Use duplicate_and_interleave for other variable-length
10190 vectors. Don't defer the update when inserting new statements.
10192 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10193 Alan Hayward <alan.hayward@arm.com>
10194 David Sherwood <david.sherwood@arm.com>
10196 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10197 min_profitable_iters doesn't go negative.
10199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10200 Alan Hayward <alan.hayward@arm.com>
10201 David Sherwood <david.sherwood@arm.com>
10203 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10204 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10205 * optabs.def (vec_mask_load_lanes_optab): New optab.
10206 (vec_mask_store_lanes_optab): Likewise.
10207 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10208 (MASK_STORE_LANES): Likewise.
10209 * internal-fn.c (mask_load_lanes_direct): New macro.
10210 (mask_store_lanes_direct): Likewise.
10211 (expand_mask_load_optab_fn): Handle masked operations.
10212 (expand_mask_load_lanes_optab_fn): New macro.
10213 (expand_mask_store_optab_fn): Handle masked operations.
10214 (expand_mask_store_lanes_optab_fn): New macro.
10215 (direct_mask_load_lanes_optab_supported_p): Likewise.
10216 (direct_mask_store_lanes_optab_supported_p): Likewise.
10217 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10219 (vect_load_lanes_supported): Likewise.
10220 * tree-vect-data-refs.c (strip_conversion): New function.
10221 (can_group_stmts_p): Likewise.
10222 (vect_analyze_data_ref_accesses): Use it instead of checking
10223 for a pair of assignments.
10224 (vect_store_lanes_supported): Take a masked_p parameter.
10225 (vect_load_lanes_supported): Likewise.
10226 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10227 vect_store_lanes_supported and vect_load_lanes_supported.
10228 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10229 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10230 parameter. Don't allow gaps for masked accesses.
10231 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10232 and vect_load_lanes_supported.
10233 (get_load_store_type): Take a masked_p parameter and update
10234 call to get_group_load_store_type.
10235 (vectorizable_store): Update call to get_load_store_type.
10236 Handle IFN_MASK_STORE_LANES.
10237 (vectorizable_load): Update call to get_load_store_type.
10238 Handle IFN_MASK_LOAD_LANES.
10240 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10241 Alan Hayward <alan.hayward@arm.com>
10242 David Sherwood <david.sherwood@arm.com>
10244 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10246 * config/aarch64/aarch64-protos.h
10247 (aarch64_sve_struct_memory_operand_p): Declare.
10248 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10249 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10250 (VPRED, vpred): Handle SVE structure modes.
10251 * config/aarch64/constraints.md (Utx): New constraint.
10252 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10253 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10254 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10255 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10256 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10257 structure modes. Split into pieces after RA.
10258 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10259 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10261 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10262 SVE structure modes.
10263 (aarch64_classify_address): Likewise.
10264 (sizetochar): Move earlier in file.
10265 (aarch64_print_operand): Handle SVE register lists.
10266 (aarch64_array_mode): New function.
10267 (aarch64_sve_struct_memory_operand_p): Likewise.
10268 (TARGET_ARRAY_MODE): Redefine.
10270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10271 Alan Hayward <alan.hayward@arm.com>
10272 David Sherwood <david.sherwood@arm.com>
10274 * target.def (array_mode): New target hook.
10275 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10276 * doc/tm.texi: Regenerate.
10277 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10278 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10279 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10280 targetm.array_mode.
10281 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10284 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10285 Alan Hayward <alan.hayward@arm.com>
10286 David Sherwood <david.sherwood@arm.com>
10288 * fold-const.c (fold_binary_loc): Check the argument types
10289 rather than the result type when testing for a vector operation.
10291 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10293 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10294 * doc/tm.texi: Regenerate.
10296 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10297 Alan Hayward <alan.hayward@arm.com>
10298 David Sherwood <david.sherwood@arm.com>
10300 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10301 (sve): Document new AArch64 extension.
10302 * doc/md.texi (w): Extend the description of the AArch64
10303 constraint to include SVE vectors.
10304 (Upl, Upa): Document new AArch64 predicate constraints.
10305 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10307 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10308 (msve-vector-bits=): New option.
10309 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10310 SVE when these are disabled.
10311 (sve): New extension.
10312 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10313 modes. Adjust their number of units based on aarch64_sve_vg.
10314 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10315 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10316 aarch64_addr_query_type.
10317 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10318 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10319 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10320 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10321 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10322 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10323 (aarch64_simd_imm_zero_p): Delete.
10324 (aarch64_check_zero_based_sve_index_immediate): Declare.
10325 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10326 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10327 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10328 (aarch64_sve_float_mul_immediate_p): Likewise.
10329 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10330 rather than an rtx.
10331 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10332 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10333 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10334 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10335 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10336 (aarch64_regmode_natural_size): Likewise.
10337 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10338 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10340 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10341 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10342 for VG and the SVE predicate registers.
10343 (V_ALIASES): Add a "z"-prefixed alias.
10344 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10345 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10346 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10347 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10348 (REG_CLASS_NAMES): Add entries for them.
10349 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10350 and the predicate registers.
10351 (aarch64_sve_vg): Declare.
10352 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10353 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10354 (REGMODE_NATURAL_SIZE): Define.
10355 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10357 * config/aarch64/aarch64.c: Include cfgrtl.h.
10358 (simd_immediate_info): Add a constructor for series vectors,
10359 and an associated step field.
10360 (aarch64_sve_vg): New variable.
10361 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10362 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10363 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10364 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10365 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10366 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10367 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10368 (aarch64_get_mask_mode): New functions.
10369 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10370 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10371 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10372 predicate modes and predicate registers. Explicitly restrict
10373 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10374 to store a vector mode if it is recognized by
10375 aarch64_classify_vector_mode.
10376 (aarch64_regmode_natural_size): New function.
10377 (aarch64_hard_regno_caller_save_mode): Return the original mode
10379 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10380 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10381 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10382 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10384 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10385 does not overlap dest if the function is frame-related. Handle
10387 (aarch64_split_add_offset): New function.
10388 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10389 them aarch64_add_offset.
10390 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10391 and update call to aarch64_sub_sp.
10392 (aarch64_add_cfa_expression): New function.
10393 (aarch64_expand_prologue): Pass extra temporary registers to the
10394 functions above. Handle the case in which we need to emit new
10395 DW_CFA_expressions for registers that were originally saved
10396 relative to the stack pointer, but now have to be expressed
10397 relative to the frame pointer.
10398 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10400 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10401 IP0 and IP1 values for SVE frames.
10402 (aarch64_expand_vec_series): New function.
10403 (aarch64_expand_sve_widened_duplicate): Likewise.
10404 (aarch64_expand_sve_const_vector): Likewise.
10405 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10406 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10407 into the register, rather than emitting a SET directly.
10408 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10409 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10410 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10411 (offset_9bit_signed_scaled_p): New functions.
10412 (aarch64_replicate_bitmask_imm): New function.
10413 (aarch64_bitmask_imm): Use it.
10414 (aarch64_cannot_force_const_mem): Reject expressions involving
10415 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10416 (aarch64_classify_index): Handle SVE indices, by requiring
10417 a plain register index with a scale that matches the element size.
10418 (aarch64_classify_address): Handle SVE addresses. Assert that
10419 the mode of the address is VOIDmode or an integer mode.
10420 Update call to aarch64_classify_symbol.
10421 (aarch64_classify_symbolic_expression): Update call to
10422 aarch64_classify_symbol.
10423 (aarch64_const_vec_all_in_range_p): New function.
10424 (aarch64_print_vector_float_operand): Likewise.
10425 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10426 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10427 and the FP immediates 1.0 and 0.5.
10428 (aarch64_print_address_internal): Handle SVE addresses.
10429 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10430 (aarch64_regno_regclass): Handle predicate registers.
10431 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10433 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10434 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10435 (aarch64_convert_sve_vector_bits): New function.
10436 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10437 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10438 rather than an rtx.
10439 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10440 Handle SVE vector and predicate modes. Accept VL-based constants
10441 that need only one temporary register, and VL offsets that require
10442 no temporary registers.
10443 (aarch64_conditional_register_usage): Mark the predicate registers
10444 as fixed if SVE isn't available.
10445 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10446 Return true for SVE vector and predicate modes.
10447 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10448 rather than an unsigned int. Handle SVE modes.
10449 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10451 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10453 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10454 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10455 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10456 (aarch64_sve_float_mul_immediate_p): New functions.
10457 (aarch64_sve_valid_immediate): New function.
10458 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10459 Explicitly reject structure modes. Check for INDEX constants.
10460 Handle PTRUE and PFALSE constants.
10461 (aarch64_check_zero_based_sve_index_immediate): New function.
10462 (aarch64_simd_imm_zero_p): Delete.
10463 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10464 vector modes. Accept constants in the range of CNT[BHWD].
10465 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10466 ask for an Advanced SIMD mode.
10467 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10468 (aarch64_simd_vector_alignment): Handle SVE predicates.
10469 (aarch64_vectorize_preferred_vector_alignment): New function.
10470 (aarch64_simd_vector_alignment_reachable): Use it instead of
10472 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10473 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10475 (MAX_VECT_LEN): Delete.
10476 (expand_vec_perm_d): Add a vec_flags field.
10477 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10478 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10479 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10481 (aarch64_evpc_rev): Rename to...
10482 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10483 (aarch64_evpc_rev_global): New function.
10484 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10485 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10487 (aarch64_evpc_sve_tbl): New function.
10488 (aarch64_expand_vec_perm_const_1): Update after rename of
10489 aarch64_evpc_rev. Handle SVE permutes too, trying
10490 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10491 than aarch64_evpc_tbl.
10492 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10493 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10494 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10495 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10496 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10497 (aarch64_expand_sve_vcond): New functions.
10498 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10499 of aarch64_vector_mode_p.
10500 (aarch64_dwarf_poly_indeterminate_value): New function.
10501 (aarch64_compute_pressure_classes): Likewise.
10502 (aarch64_can_change_mode_class): Likewise.
10503 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10504 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10505 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10506 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10507 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10508 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10509 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10510 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10512 (Dn, Dl, Dr): Accept const as well as const_vector.
10513 (Dz): Likewise. Compare against CONST0_RTX.
10514 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10515 of "vector" where appropriate.
10516 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10517 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10518 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10519 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10520 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10521 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10522 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10523 (v_int_equiv): Extend to SVE modes.
10524 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10526 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10527 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10528 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10529 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10530 (SVE_COND_FP_CMP): New int iterators.
10531 (perm_hilo): Handle the new unpack unspecs.
10532 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10534 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10535 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10536 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10537 (aarch64_equality_operator, aarch64_constant_vector_operand)
10538 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10539 (aarch64_sve_nonimmediate_operand): Likewise.
10540 (aarch64_sve_general_operand): Likewise.
10541 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10542 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10543 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10544 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10545 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10546 (aarch64_sve_float_arith_immediate): Likewise.
10547 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10548 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10549 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10550 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10551 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10552 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10553 (aarch64_sve_float_arith_operand): Likewise.
10554 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10555 (aarch64_sve_float_mul_operand): Likewise.
10556 (aarch64_sve_vec_perm_operand): Likewise.
10557 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10558 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10559 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10560 as well as const_vector.
10561 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10562 in file. Use CONST0_RTX and CONSTM1_RTX.
10563 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10564 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10565 Use aarch64_simd_imm_zero.
10566 * config/aarch64/aarch64-sve.md: New file.
10567 * config/aarch64/aarch64.md: Include it.
10568 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10569 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10570 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10571 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10572 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10573 (sve): New attribute.
10574 (enabled): Disable instructions with the sve attribute unless
10576 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10577 aarch64_expand_mov_immediate.
10578 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10579 CNT[BHSD] immediates.
10580 (movti): Split CONST_POLY_INT moves into two halves.
10581 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10582 Split additions that need a temporary here if the destination
10583 is the stack pointer.
10584 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10585 (*add<mode>3_poly_1): New instruction.
10586 (set_clobber_cc): New expander.
10588 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10590 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10591 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10592 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10593 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10594 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10595 Change innermode from fixed_mode_size to machine_mode.
10596 (simplify_subreg): Update call accordingly. Handle a constant-sized
10597 subreg of a variable-length CONST_VECTOR.
10599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10600 Alan Hayward <alan.hayward@arm.com>
10601 David Sherwood <david.sherwood@arm.com>
10603 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10604 (add_offset_to_base): New function, split out from...
10605 (create_mem_ref): ...here. When handling a scale other than 1,
10606 check first whether the address is valid without the offset.
10607 Add it into the base if so, leaving the index and scale as-is.
10609 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10612 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10613 fold_for_warn before checking if arg2 is INTEGER_CST.
10615 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10617 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10618 (store_multiple_operation): Delete.
10619 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10620 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10621 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10622 guarded by TARGET_STRING.
10623 (rs6000_output_load_multiple): Delete.
10624 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10625 OPTION_MASK_STRING / TARGET_STRING handling.
10626 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10627 (const rs6000_opt_masks) <"string">: Change mask to 0.
10628 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10629 (MASK_STRING): Delete.
10630 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10632 (load_multiple): Delete.
10639 (store_multiple): Delete.
10646 (movmemsi_8reg): Delete.
10647 (corresponding unnamed define_insn): Delete.
10648 (movmemsi_6reg): Delete.
10649 (corresponding unnamed define_insn): Delete.
10650 (movmemsi_4reg): Delete.
10651 (corresponding unnamed define_insn): Delete.
10652 (movmemsi_2reg): Delete.
10653 (corresponding unnamed define_insn): Delete.
10654 (movmemsi_1reg): Delete.
10655 (corresponding unnamed define_insn): Delete.
10656 * config/rs6000/rs6000.opt (mno-string): New.
10657 (mstring): Replace by deprecation warning stub.
10658 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10660 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10662 * regrename.c (regrename_do_replace): If replacing the same
10663 reg multiple times, try to reuse last created gen_raw_REG.
10666 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10667 main to workaround a bug in GDB.
10669 2018-01-12 Tom de Vries <tom@codesourcery.com>
10672 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10674 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10676 PR rtl-optimization/80481
10677 * ira-color.c (get_cap_member): New function.
10678 (allocnos_conflict_by_live_ranges_p): Use it.
10679 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10680 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10682 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10685 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10686 (*saddl_se_1): Ditto.
10687 (*ssubsi_1): Ditto.
10688 (*ssubl_se_1): Ditto.
10690 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10692 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10693 rather than wi::to_widest for DR_INITs.
10694 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10695 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10696 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10698 (vect_analyze_group_access_1): Note that here.
10700 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10702 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10703 polynomial type sizes.
10705 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10707 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10708 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10709 (gimple_add_tmp_var): Likewise.
10711 2018-01-12 Martin Liska <mliska@suse.cz>
10713 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10714 (gimple_alloc_sizes): Likewise.
10715 (dump_gimple_statistics): Use PRIu64 in printf format.
10716 * gimple.h: Change uint64_t to int.
10718 2018-01-12 Martin Liska <mliska@suse.cz>
10720 * tree-core.h: Use uint64_t instead of int.
10721 * tree.c (tree_node_counts): Likewise.
10722 (tree_node_sizes): Likewise.
10723 (dump_tree_statistics): Use PRIu64 in printf format.
10725 2018-01-12 Martin Liska <mliska@suse.cz>
10727 * Makefile.in: As qsort_chk is implemented in vec.c, add
10728 vec.o to linkage of gencfn-macros.
10729 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10730 passing the info to record_node_allocation_statistics.
10731 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10733 * ggc-common.c (struct ggc_usage): Add operator== and use
10734 it in operator< and compare function.
10735 * mem-stats.h (struct mem_usage): Likewise.
10736 * vec.c (struct vec_usage): Remove operator< and compare
10737 function. Can be simply inherited.
10739 2018-01-12 Martin Jambor <mjambor@suse.cz>
10742 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10743 * tree-ssa-math-opts.c: Include domwalk.h.
10744 (convert_mult_to_fma_1): New function.
10745 (fma_transformation_info): New type.
10746 (fma_deferring_state): Likewise.
10747 (cancel_fma_deferring): New function.
10748 (result_of_phi): Likewise.
10749 (last_fma_candidate_feeds_initial_phi): Likewise.
10750 (convert_mult_to_fma): Added deferring logic, split actual
10751 transformation to convert_mult_to_fma_1.
10752 (math_opts_dom_walker): New type.
10753 (math_opts_dom_walker::after_dom_children): New method, body moved
10754 here from pass_optimize_widening_mul::execute, added deferring logic
10756 (pass_optimize_widening_mul::execute): Moved most of code to
10757 math_opts_dom_walker::after_dom_children.
10758 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10759 * config/i386/i386.c (ix86_option_override_internal): Added
10760 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10762 2018-01-12 Richard Biener <rguenther@suse.de>
10765 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10766 inline instance vars.
10768 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10771 * config/rx/rx.c (rx_is_restricted_memory_address):
10772 Handle SUBREG case.
10774 2018-01-12 Richard Biener <rguenther@suse.de>
10776 PR tree-optimization/80846
10777 * target.def (split_reduction): New target hook.
10778 * targhooks.c (default_split_reduction): New function.
10779 * targhooks.h (default_split_reduction): Declare.
10780 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10781 target requests first reduce vectors by combining low and high
10783 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10784 (get_vectype_for_scalar_type_and_size): Export.
10785 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10786 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10787 * doc/tm.texi: Regenerate.
10788 * config/i386/i386.c (ix86_split_reduction): Implement
10789 TARGET_VECTORIZE_SPLIT_REDUCTION.
10791 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10794 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10795 in PIC mode except for TARGET_VXWORKS_RTP.
10796 * config/sparc/sparc.c: Include cfgrtl.h.
10797 (TARGET_INIT_PIC_REG): Define.
10798 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10799 (sparc_pic_register_p): New predicate.
10800 (sparc_legitimate_address_p): Use it.
10801 (sparc_legitimize_pic_address): Likewise.
10802 (sparc_delegitimize_address): Likewise.
10803 (sparc_mode_dependent_address_p): Likewise.
10804 (gen_load_pcrel_sym): Remove 4th parameter.
10805 (load_got_register): Adjust call to above. Remove obsolete stuff.
10806 (sparc_expand_prologue): Do not call load_got_register here.
10807 (sparc_flat_expand_prologue): Likewise.
10808 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10809 (sparc_use_pseudo_pic_reg): New function.
10810 (sparc_init_pic_reg): Likewise.
10811 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10812 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10814 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10816 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10817 Add item for branch_cost.
10819 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10821 PR rtl-optimization/83565
10822 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10823 not extend the result to a larger mode for rotate operations.
10824 (num_sign_bit_copies1): Likewise.
10826 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10829 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10831 Use values-Xc.o for -pedantic.
10832 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10834 2018-01-12 Martin Liska <mliska@suse.cz>
10837 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10839 (possible_polymorphic_call_targets): Use it.
10840 (ipa_devirt): Likewise.
10842 2018-01-12 Martin Liska <mliska@suse.cz>
10844 * profile-count.h (enum profile_quality): Use 0 as invalid
10845 enum value of profile_quality.
10847 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10849 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10850 -mext-string options.
10852 2018-01-12 Richard Biener <rguenther@suse.de>
10854 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10855 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10856 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10858 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10860 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10862 * configure.ac (--with-long-double-format): Add support for the
10863 configuration option to change the default long double format on
10865 * config.gcc (powerpc*-linux*-*): Likewise.
10866 * configure: Regenerate.
10867 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10868 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10869 used without modification.
10871 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10873 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10874 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10875 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10876 MISC_BUILTIN_SPEC_BARRIER.
10877 (rs6000_init_builtins): Likewise.
10878 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10880 (speculation_barrier): New define_insn.
10881 * doc/extend.texi: Document __builtin_speculation_barrier.
10883 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10886 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10887 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10888 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10890 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10891 integral modes instead of "ss" and "sd".
10892 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10893 vectors with 32-bit and 64-bit elements.
10894 (vecdupssescalarmodesuffix): New mode attribute.
10895 (vec_dup<mode>): Use it.
10897 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10900 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10901 frame if argument is passed on stack.
10903 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10906 * ree.c (combine_reaching_defs): Optimize also
10907 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10908 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10910 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10912 PR middle-end/83189
10913 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10915 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10917 PR middle-end/83718
10918 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10919 after they are computed.
10921 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10923 PR tree-optimization/83695
10924 * gimple-loop-linterchange.cc
10925 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10926 reset cached scev information after interchange.
10927 (pass_linterchange::execute): Remove call to scev_reset_htab.
10929 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10931 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10932 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10933 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10934 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10935 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10936 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10937 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10938 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10939 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10940 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10941 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10942 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10943 (V_lane_reg): Likewise.
10944 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10946 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10947 (vfmal_lane_low<mode>_intrinsic,
10948 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10949 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10950 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10951 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10952 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10953 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10955 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10957 * config/arm/arm-cpus.in (fp16fml): New feature.
10958 (ALL_SIMD): Add fp16fml.
10959 (armv8.2-a): Add fp16fml as an option.
10960 (armv8.3-a): Likewise.
10961 (armv8.4-a): Add fp16fml as part of fp16.
10962 * config/arm/arm.h (TARGET_FP16FML): Define.
10963 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10965 * config/arm/arm-modes.def (V2HF): Define.
10966 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10967 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10968 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10969 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10970 vfmsl_low, vfmsl_high): New set of builtins.
10971 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10972 (vfml_op): New code attribute.
10973 (VFMLHALVES): New int iterator.
10974 (VFML, VFMLSEL): New mode attributes.
10975 (V_reg): Define mapping for V2HF.
10976 (V_hi, V_lo): New mode attributes.
10977 (VF_constraint): Likewise.
10978 (vfml_half, vfml_half_selector): New int attributes.
10979 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10981 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10982 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10984 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10985 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10986 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10987 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10989 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10990 Document new effective target and option set.
10992 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10994 * config/arm/arm-cpus.in (armv8_4): New feature.
10995 (ARMv8_4a): New fgroup.
10996 (armv8.4-a): New arch.
10997 * config/arm/arm-tables.opt: Regenerate.
10998 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10999 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11000 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11001 Add matching rules for -march=armv8.4-a and extensions.
11002 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11004 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11007 * config/rx/rx.md (BW): New mode attribute.
11008 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11010 2018-01-11 Richard Biener <rguenther@suse.de>
11012 PR tree-optimization/83435
11013 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11014 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11015 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11017 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11018 Alan Hayward <alan.hayward@arm.com>
11019 David Sherwood <david.sherwood@arm.com>
11021 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11023 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11024 (aarch64_print_address_internal): Use it to check for a zero offset.
11026 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11027 Alan Hayward <alan.hayward@arm.com>
11028 David Sherwood <david.sherwood@arm.com>
11030 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11031 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11032 Return a poly_int64 rather than a HOST_WIDE_INT.
11033 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11034 rather than a HOST_WIDE_INT.
11035 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11036 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11037 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11038 final_offset from HOST_WIDE_INT to poly_int64.
11039 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11040 to_constant when getting the number of units in an Advanced SIMD
11042 (aarch64_builtin_vectorized_function): Check for a constant number
11044 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11046 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11047 attribute instead of GET_MODE_NUNITS.
11048 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11049 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11050 GET_MODE_SIZE for fixed-size registers.
11051 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11052 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11053 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11054 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11055 (aarch64_print_operand, aarch64_print_address_internal)
11056 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11057 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11058 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11059 Handle polynomial GET_MODE_SIZE.
11060 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11061 wider than SImode without modification.
11062 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11063 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11064 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11065 passing and returning SVE modes.
11066 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11067 rather than GEN_INT.
11068 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11069 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11070 (aarch64_allocate_and_probe_stack_space): Likewise.
11071 (aarch64_layout_frame): Cope with polynomial offsets.
11072 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11073 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11074 polynomial offsets.
11075 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11076 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11077 poly_int64 rather than a HOST_WIDE_INT.
11078 (aarch64_get_separate_components, aarch64_process_components)
11079 (aarch64_expand_prologue, aarch64_expand_epilogue)
11080 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11081 (aarch64_anchor_offset): New function, split out from...
11082 (aarch64_legitimize_address): ...here.
11083 (aarch64_builtin_vectorization_cost): Handle polynomial
11084 TYPE_VECTOR_SUBPARTS.
11085 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11087 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11088 number of elements from the PARALLEL rather than the mode.
11089 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11090 rather than GET_MODE_BITSIZE.
11091 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11092 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11093 (aarch64_expand_vec_perm_const_1): Handle polynomial
11094 d->perm.length () and d->perm elements.
11095 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11096 Apply to_constant to d->perm elements.
11097 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11098 polynomial CONST_VECTOR_NUNITS.
11099 (aarch64_move_pointer): Take amount as a poly_int64 rather
11101 (aarch64_progress_pointer): Avoid temporary variable.
11102 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11103 the mode attribute instead of GET_MODE.
11105 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11106 Alan Hayward <alan.hayward@arm.com>
11107 David Sherwood <david.sherwood@arm.com>
11109 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11110 x exists before using it.
11111 (aarch64_add_constant_internal): Rename to...
11112 (aarch64_add_offset_1): ...this. Replace regnum with separate
11113 src and dest rtxes. Handle the case in which they're different,
11114 including when the offset is zero. Replace scratchreg with an rtx.
11115 Use 2 additions if there is no spare register into which we can
11116 move a 16-bit constant.
11117 (aarch64_add_constant): Delete.
11118 (aarch64_add_offset): Replace reg with separate src and dest
11119 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11120 Use aarch64_add_offset_1.
11121 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11122 an rtx rather than an int. Take the delta as a poly_int64
11123 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11124 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11125 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11126 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11127 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11128 and aarch64_add_sp.
11129 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11130 aarch64_add_constant.
11132 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11134 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11135 Use scalar_float_mode.
11137 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11139 * config/aarch64/aarch64-simd.md
11140 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11141 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11142 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11143 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11144 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11145 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11146 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11147 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11148 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11149 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11151 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11154 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11155 targ_options->x_arm_arch_string is non NULL.
11157 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11159 * config/aarch64/aarch64.h
11160 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11162 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11165 * expmed.c (emit_store_flag_force): Swap if const op0
11166 and change VOIDmode to mode of op0.
11168 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11170 PR rtl-optimization/83761
11171 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11172 than bytes to mode_for_size.
11174 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11176 PR middle-end/83189
11177 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11178 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11181 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11183 PR middle-end/83575
11184 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11185 when in layout mode.
11186 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11187 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11190 2018-01-10 Michael Collison <michael.collison@arm.com>
11192 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11193 * config/aarch64/aarch64-option-extension.def: Add
11194 AARCH64_OPT_EXTENSION of 'fp16fml'.
11195 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11196 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11197 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11198 * config/aarch64/constraints.md (Ui7): New constraint.
11199 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11200 (VFMLA_SEL_W): Ditto.
11203 (VFMLA16_LOW): New int iterator.
11204 (VFMLA16_HIGH): Ditto.
11205 (UNSPEC_FMLAL): New unspec.
11206 (UNSPEC_FMLSL): Ditto.
11207 (UNSPEC_FMLAL2): Ditto.
11208 (UNSPEC_FMLSL2): Ditto.
11209 (f16mac): New code attribute.
11210 * config/aarch64/aarch64-simd-builtins.def
11211 (aarch64_fmlal_lowv2sf): Ditto.
11212 (aarch64_fmlsl_lowv2sf): Ditto.
11213 (aarch64_fmlalq_lowv4sf): Ditto.
11214 (aarch64_fmlslq_lowv4sf): Ditto.
11215 (aarch64_fmlal_highv2sf): Ditto.
11216 (aarch64_fmlsl_highv2sf): Ditto.
11217 (aarch64_fmlalq_highv4sf): Ditto.
11218 (aarch64_fmlslq_highv4sf): Ditto.
11219 (aarch64_fmlal_lane_lowv2sf): Ditto.
11220 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11221 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11222 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11223 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11224 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11225 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11226 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11227 (aarch64_fmlal_lane_highv2sf): Ditto.
11228 (aarch64_fmlsl_lane_highv2sf): Ditto.
11229 (aarch64_fmlal_laneq_highv2sf): Ditto.
11230 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11231 (aarch64_fmlalq_lane_highv4sf): Ditto.
11232 (aarch64_fmlsl_lane_highv4sf): Ditto.
11233 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11234 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11235 * config/aarch64/aarch64-simd.md:
11236 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11237 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11238 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11239 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11240 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11241 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11242 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11243 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11244 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11245 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11246 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11247 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11248 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11249 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11250 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11251 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11252 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11253 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11254 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11255 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11256 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11257 (vfmlsl_low_u32): Ditto.
11258 (vfmlalq_low_u32): Ditto.
11259 (vfmlslq_low_u32): Ditto.
11260 (vfmlal_high_u32): Ditto.
11261 (vfmlsl_high_u32): Ditto.
11262 (vfmlalq_high_u32): Ditto.
11263 (vfmlslq_high_u32): Ditto.
11264 (vfmlal_lane_low_u32): Ditto.
11265 (vfmlsl_lane_low_u32): Ditto.
11266 (vfmlal_laneq_low_u32): Ditto.
11267 (vfmlsl_laneq_low_u32): Ditto.
11268 (vfmlalq_lane_low_u32): Ditto.
11269 (vfmlslq_lane_low_u32): Ditto.
11270 (vfmlalq_laneq_low_u32): Ditto.
11271 (vfmlslq_laneq_low_u32): Ditto.
11272 (vfmlal_lane_high_u32): Ditto.
11273 (vfmlsl_lane_high_u32): Ditto.
11274 (vfmlal_laneq_high_u32): Ditto.
11275 (vfmlsl_laneq_high_u32): Ditto.
11276 (vfmlalq_lane_high_u32): Ditto.
11277 (vfmlslq_lane_high_u32): Ditto.
11278 (vfmlalq_laneq_high_u32): Ditto.
11279 (vfmlslq_laneq_high_u32): Ditto.
11280 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11281 (AARCH64_FL_FOR_ARCH8_4): New.
11282 (AARCH64_ISA_F16FML): New ISA flag.
11283 (TARGET_F16FML): New feature flag for fp16fml.
11284 (doc/invoke.texi): Document new fp16fml option.
11286 2018-01-10 Michael Collison <michael.collison@arm.com>
11288 * config/aarch64/aarch64-builtins.c:
11289 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11290 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11291 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11292 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11293 (AARCH64_ISA_SHA3): New ISA flag.
11294 (TARGET_SHA3): New feature flag for sha3.
11295 * config/aarch64/iterators.md (sha512_op): New int attribute.
11296 (CRYPTO_SHA512): New int iterator.
11297 (UNSPEC_SHA512H): New unspec.
11298 (UNSPEC_SHA512H2): Ditto.
11299 (UNSPEC_SHA512SU0): Ditto.
11300 (UNSPEC_SHA512SU1): Ditto.
11301 * config/aarch64/aarch64-simd-builtins.def
11302 (aarch64_crypto_sha512hqv2di): New builtin.
11303 (aarch64_crypto_sha512h2qv2di): Ditto.
11304 (aarch64_crypto_sha512su0qv2di): Ditto.
11305 (aarch64_crypto_sha512su1qv2di): Ditto.
11306 (aarch64_eor3qv8hi): Ditto.
11307 (aarch64_rax1qv2di): Ditto.
11308 (aarch64_xarqv2di): Ditto.
11309 (aarch64_bcaxqv8hi): Ditto.
11310 * config/aarch64/aarch64-simd.md:
11311 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11312 (aarch64_crypto_sha512su0qv2di): Ditto.
11313 (aarch64_crypto_sha512su1qv2di): Ditto.
11314 (aarch64_eor3qv8hi): Ditto.
11315 (aarch64_rax1qv2di): Ditto.
11316 (aarch64_xarqv2di): Ditto.
11317 (aarch64_bcaxqv8hi): Ditto.
11318 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11319 (vsha512h2q_u64): Ditto.
11320 (vsha512su0q_u64): Ditto.
11321 (vsha512su1q_u64): Ditto.
11322 (veor3q_u16): Ditto.
11323 (vrax1q_u64): Ditto.
11324 (vxarq_u64): Ditto.
11325 (vbcaxq_u16): Ditto.
11326 * config/arm/types.md (crypto_sha512): New type attribute.
11327 (crypto_sha3): Ditto.
11328 (doc/invoke.texi): Document new sha3 option.
11330 2018-01-10 Michael Collison <michael.collison@arm.com>
11332 * config/aarch64/aarch64-builtins.c:
11333 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11334 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11335 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11336 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11337 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11338 (AARCH64_ISA_SM4): New ISA flag.
11339 (TARGET_SM4): New feature flag for sm4.
11340 * config/aarch64/aarch64-simd-builtins.def
11341 (aarch64_sm3ss1qv4si): Ditto.
11342 (aarch64_sm3tt1aq4si): Ditto.
11343 (aarch64_sm3tt1bq4si): Ditto.
11344 (aarch64_sm3tt2aq4si): Ditto.
11345 (aarch64_sm3tt2bq4si): Ditto.
11346 (aarch64_sm3partw1qv4si): Ditto.
11347 (aarch64_sm3partw2qv4si): Ditto.
11348 (aarch64_sm4eqv4si): Ditto.
11349 (aarch64_sm4ekeyqv4si): Ditto.
11350 * config/aarch64/aarch64-simd.md:
11351 (aarch64_sm3ss1qv4si): Ditto.
11352 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11353 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11354 (aarch64_sm4eqv4si): Ditto.
11355 (aarch64_sm4ekeyqv4si): Ditto.
11356 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11357 (sm3part_op): Ditto.
11358 (CRYPTO_SM3TT): Ditto.
11359 (CRYPTO_SM3PART): Ditto.
11360 (UNSPEC_SM3SS1): New unspec.
11361 (UNSPEC_SM3TT1A): Ditto.
11362 (UNSPEC_SM3TT1B): Ditto.
11363 (UNSPEC_SM3TT2A): Ditto.
11364 (UNSPEC_SM3TT2B): Ditto.
11365 (UNSPEC_SM3PARTW1): Ditto.
11366 (UNSPEC_SM3PARTW2): Ditto.
11367 (UNSPEC_SM4E): Ditto.
11368 (UNSPEC_SM4EKEY): Ditto.
11369 * config/aarch64/constraints.md (Ui2): New constraint.
11370 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11371 * config/arm/types.md (crypto_sm3): New type attribute.
11372 (crypto_sm4): Ditto.
11373 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11374 (vsm3tt1aq_u32): Ditto.
11375 (vsm3tt1bq_u32): Ditto.
11376 (vsm3tt2aq_u32): Ditto.
11377 (vsm3tt2bq_u32): Ditto.
11378 (vsm3partw1q_u32): Ditto.
11379 (vsm3partw2q_u32): Ditto.
11380 (vsm4eq_u32): Ditto.
11381 (vsm4ekeyq_u32): Ditto.
11382 (doc/invoke.texi): Document new sm4 option.
11384 2018-01-10 Michael Collison <michael.collison@arm.com>
11386 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11387 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11388 (AARCH64_FL_FOR_ARCH8_4): New.
11389 (AARCH64_FL_V8_4): New flag.
11390 (doc/invoke.texi): Document new armv8.4-a option.
11392 2018-01-10 Michael Collison <michael.collison@arm.com>
11394 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11395 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11396 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11397 * config/aarch64/aarch64-option-extension.def: Add
11398 AARCH64_OPT_EXTENSION of 'sha2'.
11399 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11400 (crypto): Disable sha2 and aes if crypto disabled.
11401 (crypto): Enable aes and sha2 if enabled.
11402 (simd): Disable sha2 and aes if simd disabled.
11403 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11405 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11406 (TARGET_SHA2): New feature flag for sha2.
11407 (TARGET_AES): New feature flag for aes.
11408 * config/aarch64/aarch64-simd.md:
11409 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11410 conditional on TARGET_AES.
11411 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11412 (aarch64_crypto_sha1hsi): Make pattern conditional
11414 (aarch64_crypto_sha1hv4si): Ditto.
11415 (aarch64_be_crypto_sha1hv4si): Ditto.
11416 (aarch64_crypto_sha1su1v4si): Ditto.
11417 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11418 (aarch64_crypto_sha1su0v4si): Ditto.
11419 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11420 (aarch64_crypto_sha256su0v4si): Ditto.
11421 (aarch64_crypto_sha256su1v4si): Ditto.
11422 (doc/invoke.texi): Document new aes and sha2 options.
11424 2018-01-10 Martin Sebor <msebor@redhat.com>
11426 PR tree-optimization/83781
11427 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11430 2018-01-11 Martin Sebor <msebor@gmail.com>
11431 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11433 PR tree-optimization/83501
11434 PR tree-optimization/81703
11436 * tree-ssa-strlen.c (get_string_cst): Rename...
11437 (get_string_len): ...to this. Handle global constants.
11438 (handle_char_store): Adjust.
11440 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11441 Jim Wilson <jimw@sifive.com>
11443 * config/riscv/riscv-protos.h (riscv_output_return): New.
11444 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11445 (riscv_attribute_table, riscv_output_return),
11446 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11447 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11448 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11450 (riscv_expand_prologue): Add early return for naked function.
11451 (riscv_expand_epilogue): Likewise.
11452 (riscv_function_ok_for_sibcall): Return false for naked function.
11453 (riscv_set_current_function): New.
11454 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11455 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11456 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11457 * doc/extend.texi (RISC-V Function Attributes): New.
11459 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11461 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11462 check for 128-bit long double before checking TCmode.
11463 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11464 128-bit long doubles before checking TFmode or TCmode.
11465 (FLOAT128_IBM_P): Likewise.
11467 2018-01-10 Martin Sebor <msebor@redhat.com>
11469 PR tree-optimization/83671
11470 * builtins.c (c_strlen): Unconditionally return zero for the empty
11472 Use -Warray-bounds for warnings.
11473 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11474 for non-constant array indices with COMPONENT_REF, arrays of
11475 arrays, and pointers to arrays.
11476 (gimple_fold_builtin_strlen): Determine and set length range for
11477 non-constant character arrays.
11479 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11481 PR middle-end/81897
11482 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11485 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11487 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11489 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11492 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11493 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11494 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11495 indexed_or_indirect_operand predicate.
11496 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11497 (*vsx_le_perm_load_v8hi): Likewise.
11498 (*vsx_le_perm_load_v16qi): Likewise.
11499 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11500 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11501 (*vsx_le_perm_store_v8hi): Likewise.
11502 (*vsx_le_perm_store_v16qi): Likewise.
11503 (eight unnamed splitters): Likewise.
11505 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11507 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11508 * config/rs6000/emmintrin.h: Likewise.
11509 * config/rs6000/mmintrin.h: Likewise.
11510 * config/rs6000/xmmintrin.h: Likewise.
11512 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11515 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11517 * tree.c (tree_nop_conversion): Return true for location wrapper
11519 (maybe_wrap_with_location): New function.
11520 (selftest::check_strip_nops): New function.
11521 (selftest::test_location_wrappers): New function.
11522 (selftest::tree_c_tests): Call it.
11523 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11524 (maybe_wrap_with_location): New decl.
11525 (EXPR_LOCATION_WRAPPER_P): New macro.
11526 (location_wrapper_p): New inline function.
11527 (tree_strip_any_location_wrapper): New inline function.
11529 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11532 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11533 stack_realign_offset for the largest alignment of stack slot
11535 (ix86_find_max_used_stack_alignment): New function.
11536 (ix86_finalize_stack_frame_flags): Use it. Set
11537 max_used_stack_alignment if we don't realign stack.
11538 * config/i386/i386.h (machine_function): Add
11539 max_used_stack_alignment.
11541 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11543 * config/arm/arm.opt (-mbranch-cost): New option.
11544 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11547 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11550 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11551 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11553 2018-01-10 Richard Biener <rguenther@suse.de>
11556 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11557 early out so it also covers the case where we have a non-NULL
11560 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11562 PR tree-optimization/83753
11563 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11564 for non-strided grouped accesses if the number of elements is 1.
11566 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11569 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11570 * i386.h (TARGET_USE_GATHER): Define.
11571 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11573 2018-01-10 Martin Liska <mliska@suse.cz>
11576 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11577 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11579 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11580 CLEANUP_NO_PARTITIONING is not set.
11582 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11584 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11585 for vectors, as a partial revert of r254296.
11586 * rtl.h (const_vec_p): Delete.
11587 (const_vec_duplicate_p): Don't test for vector CONSTs.
11588 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11589 * expmed.c (make_tree): Likewise.
11592 * common.md (E, F): Use CONSTANT_P instead of checking for
11594 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11595 checking for CONST_VECTOR.
11597 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11599 PR middle-end/83575
11600 * predict.c (force_edge_cold): Handle in more sane way edges
11601 with no prediction.
11603 2018-01-09 Carl Love <cel@us.ibm.com>
11605 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11607 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11608 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11609 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11610 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11611 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11612 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11613 * config/rs6000/rs6000-protos.h: Add extern defition for
11614 rs6000_generate_float2_double_code.
11615 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11617 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11618 (float2_v2df): Add define_expand.
11620 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11623 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11624 op_mode in the force_to_mode call.
11626 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11628 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11629 instead of checking each element individually.
11630 (aarch64_evpc_uzp): Likewise.
11631 (aarch64_evpc_zip): Likewise.
11632 (aarch64_evpc_ext): Likewise.
11633 (aarch64_evpc_rev): Likewise.
11634 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11635 instead of checking each element individually. Return true without
11637 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11638 whether all selected elements come from the same input, instead of
11639 checking each element individually. Remove calls to gen_rtx_REG,
11640 start_sequence and end_sequence and instead assert that no rtl is
11643 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11645 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11646 order of HIGH and CONST checks.
11648 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11650 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11651 if the destination isn't an SSA_NAME.
11653 2018-01-09 Richard Biener <rguenther@suse.de>
11655 PR tree-optimization/83668
11656 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11658 (canonicalize_loop_form): ... here, renamed from ...
11659 (canonicalize_loop_closed_ssa_form): ... this and amended to
11660 swap successor edges for loop exit blocks to make us use
11661 the RPO order we need for initial schedule generation.
11663 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11665 PR tree-optimization/64811
11666 * match.pd: When optimizing comparisons with Inf, avoid
11667 introducing or losing exceptions from comparisons with NaN.
11669 2018-01-09 Martin Liska <mliska@suse.cz>
11672 * asan.c (shadow_mem_size): Add gcc_assert.
11674 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11676 Don't save registers in main().
11679 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11680 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11681 * config/avr/avr.c (avr_set_current_function): Don't error if
11682 naked, OS_task or OS_main are specified at the same time.
11683 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11685 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11687 * common/config/avr/avr-common.c (avr_option_optimization_table):
11688 Switch on -mmain-is-OS_task for optimizing compilations.
11690 2018-01-09 Richard Biener <rguenther@suse.de>
11692 PR tree-optimization/83572
11693 * graphite.c: Include cfganal.h.
11694 (graphite_transform_loops): Connect infinite loops to exit
11695 and remove fake edges at the end.
11697 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11699 * ipa-inline.c (edge_badness): Revert accidental checkin.
11701 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11704 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11705 symbols; not inline clones.
11707 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11710 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11711 hard registers. Formatting fixes.
11713 PR preprocessor/83722
11714 * gcc.c (try_generate_repro): Pass
11715 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11716 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11719 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11720 Kito Cheng <kito.cheng@gmail.com>
11722 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11723 (riscv_leaf_function_p): Delete.
11724 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11726 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11728 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11730 (do_ifelse): New function.
11731 (do_isel): New function.
11732 (do_sub3): New function.
11733 (do_add3): New function.
11734 (do_load_mask_compare): New function.
11735 (do_overlap_load_compare): New function.
11736 (expand_compare_loop): New function.
11737 (expand_block_compare): Call expand_compare_loop() when appropriate.
11738 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11739 option description.
11740 (-mblock-compare-inline-loop-limit): New option.
11742 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11745 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11746 Reverse order of second and third operands in first alternative.
11747 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11748 of first and second elements in UNSPEC_VPERMR vector.
11749 (altivec_expand_vec_perm_le): Likewise.
11751 2018-01-08 Jeff Law <law@redhat.com>
11753 PR rtl-optimizatin/81308
11754 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11755 (process_switch): If group_case_labels makes a change, then set
11757 (pass_convert_switch::execute): If a switch is converted, then
11758 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11760 PR rtl-optimization/81308
11761 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11764 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11766 PR target/83663 - Revert r255946
11767 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11768 generation for cases where splatting a value is not useful.
11769 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11770 across a vec_duplicate and a paradoxical subreg forming a vector
11771 mode to a vec_concat.
11773 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11775 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11776 -march=armv8.3-a variants.
11777 * config/arm/t-multilib: Likewise.
11778 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11780 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11782 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11784 (cceq_ior_compare_complement): Give it a name so I can use it, and
11785 change boolean_or_operator predicate to boolean_operator so it can
11786 be used to generate a crand.
11787 (eqne): New code iterator.
11788 (bd/bd_neg): New code_attrs.
11789 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11790 a single define_insn.
11791 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11792 decrement (bdnzt/bdnzf/bdzt/bdzf).
11793 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11794 with the new names of the branch decrement patterns, and added the
11795 names of the branch decrement conditional patterns.
11797 2018-01-08 Richard Biener <rguenther@suse.de>
11799 PR tree-optimization/83563
11800 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11803 2018-01-08 Richard Biener <rguenther@suse.de>
11805 PR middle-end/83713
11806 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11808 2018-01-08 Richard Biener <rguenther@suse.de>
11810 PR tree-optimization/83685
11811 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11812 references to abnormals.
11814 2018-01-08 Richard Biener <rguenther@suse.de>
11817 * dwarf2out.c (output_indirect_strings): Handle empty
11818 skeleton_debug_str_hash.
11819 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11821 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11823 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11824 (emit_store_direct): Likewise.
11825 (arc_trampoline_adjust_address): Likewise.
11826 (arc_asm_trampoline_template): New function.
11827 (arc_initialize_trampoline): Use asm_trampoline_template.
11828 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11829 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11830 * config/arc/arc.md (flush_icache): Delete pattern.
11832 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11834 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11835 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11838 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11841 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11842 by not USED_FOR_TARGET.
11843 (make_pass_resolve_sw_modes): Likewise.
11845 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11847 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11850 2018-01-08 Richard Biener <rguenther@suse.de>
11852 PR middle-end/83580
11853 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11855 2018-01-08 Richard Biener <rguenther@suse.de>
11857 PR middle-end/83517
11858 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11860 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11862 PR middle-end/81897
11863 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11864 basic blocks with a small number of successors.
11865 (convert_control_dep_chain_into_preds): Improve handling of
11867 (dump_predicates): Split apart into...
11868 (dump_pred_chain): ...here...
11869 (dump_pred_info): ...and here.
11870 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11871 (can_chain_union_be_invalidated_p): Improve check for invalidation
11873 (uninit_uses_cannot_happen): Avoid unnecessary if
11874 convert_control_dep_chain_into_preds yielded nothing.
11876 2018-01-06 Martin Sebor <msebor@redhat.com>
11878 PR tree-optimization/83640
11879 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11880 subtracting negative offset from size.
11881 (builtin_access::overlap): Adjust offset bounds of the access to fall
11882 within the size of the object if possible.
11884 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11886 PR rtl-optimization/83699
11887 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11888 extract_bit_field_as_subreg to cases in which the extracted
11889 value is also a vector.
11891 * lra-constraints.c (process_alt_operands): Test for the equivalence
11892 substitutions when detecting a possible reload cycle.
11894 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11897 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11898 by default if flag_selective_schedling{,2}. Formatting fixes.
11900 PR rtl-optimization/83682
11901 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11902 if it has non-VECTOR_MODE element mode.
11903 (vec_duplicate_p): Likewise.
11905 PR middle-end/83694
11906 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11907 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11909 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11912 * config/i386/i386-builtin.def
11913 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11914 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11915 Require also OPTION_MASK_ISA_AVX512F in addition to
11916 OPTION_MASK_ISA_GFNI.
11917 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11918 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11919 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11920 to OPTION_MASK_ISA_GFNI.
11921 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11922 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11923 OPTION_MASK_ISA_AVX512BW.
11924 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11925 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11926 addition to OPTION_MASK_ISA_GFNI.
11927 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11928 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11929 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11930 to OPTION_MASK_ISA_GFNI.
11931 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11932 a requirement for all ISAs rather than any of them with a few
11934 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11936 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11937 bitmasks to be enabled with 3 exceptions, instead of requiring any
11938 enabled ISA with lots of exceptions.
11939 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11940 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11941 Change avx512bw in isa attribute to avx512f.
11942 * config/i386/sgxintrin.h: Add license boilerplate.
11943 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11944 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11945 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11946 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11948 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11949 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11950 temporarily sse2 rather than sse if not enabled already.
11953 * config/i386/sse.md (VI248_VLBW): Rename to ...
11954 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11955 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11956 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11957 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11958 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11959 mode iterator instead of VI248_VLBW.
11961 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11963 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11964 (record_modified): Skip clobbers; add debug output.
11965 (param_change_prob): Use sreal frequencies.
11967 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11969 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11970 punt for user-aligned variables.
11972 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11974 * tree-chrec.c (chrec_contains_symbols): Return true for
11977 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11980 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11981 of (x|y) == x for BICS pattern.
11983 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11985 PR tree-optimization/83605
11986 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11987 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11990 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11992 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11993 * config/epiphany/rtems.h: New file.
11995 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11996 Uros Bizjak <ubizjak@gmail.com>
11999 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12000 QIreg_operand instead of register_operand predicate.
12001 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12002 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12003 comments instead of -fmitigate[-_]rop.
12005 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12008 * cgraphunit.c (symbol_table::compile): Switch to text_section
12009 before calling assembly_start debug hook.
12010 * run-rtl-passes.c (run_rtl_passes): Likewise.
12013 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12015 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12016 range_int_cst_p rather than !symbolic_range_p before calling
12017 extract_range_from_multiplicative_op_1.
12019 2018-01-04 Jeff Law <law@redhat.com>
12021 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12022 redundant test in assertion.
12024 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12026 * doc/rtl.texi: Document machine_mode wrapper classes.
12028 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12030 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12031 using tree_to_uhwi.
12033 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12035 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12036 the VEC_PERM_EXPR fold to fail.
12038 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12041 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12042 to switched_sections.
12044 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12047 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12048 test for d.testing.
12050 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12053 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12054 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12056 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12059 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12060 is BLKmode and bitpos not zero or mode change is needed.
12062 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12065 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12068 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12071 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12072 instead of MULT rtx. Update all corresponding splitters.
12073 (*saddl_se): Ditto.
12074 (*ssub<modesuffix>): Ditto.
12075 (*ssubl_se): Ditto.
12076 (*cmp_sadd_di): Update split patterns.
12077 (*cmp_sadd_si): Ditto.
12078 (*cmp_sadd_sidi): Ditto.
12079 (*cmp_ssub_di): Ditto.
12080 (*cmp_ssub_si): Ditto.
12081 (*cmp_ssub_sidi): Ditto.
12082 * config/alpha/predicates.md (const23_operand): New predicate.
12083 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12084 Look for ASHIFT, not MULT inner operand.
12085 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12087 2018-01-04 Martin Liska <mliska@suse.cz>
12089 PR gcov-profile/83669
12090 * gcov.c (output_intermediate_file): Add version to intermediate
12092 * doc/gcov.texi: Document new field 'version' in intermediate
12093 file format. Fix location of '-k' option of gcov command.
12095 2018-01-04 Martin Liska <mliska@suse.cz>
12098 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12100 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12102 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12104 2018-01-03 Martin Sebor <msebor@redhat.com>
12106 PR tree-optimization/83655
12107 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12108 checking calls with invalid arguments.
12110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12112 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12113 (vectorizable_mask_load_store): Delete.
12114 (vectorizable_call): Return false for masked loads and stores.
12115 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12116 instead of gimple_assign_rhs1.
12117 (vectorizable_load): Handle IFN_MASK_LOAD.
12118 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12122 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12124 (vectorizable_mask_load_store): ...here.
12125 (vectorizable_load): ...and here.
12127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12129 * tree-vect-stmts.c (vect_build_all_ones_mask)
12130 (vect_build_zero_merge_argument): New functions, split out from...
12131 (vectorizable_load): ...here.
12133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12135 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12137 (vectorizable_mask_load_store): ...here.
12138 (vectorizable_store): ...and here.
12140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12142 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12144 (vectorizable_mask_load_store): ...here.
12146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12148 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12149 (vect_model_store_cost): Take a vec_load_store_type instead of a
12151 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12152 (vect_model_store_cost): Take a vec_load_store_type instead of a
12154 (vectorizable_mask_load_store): Update accordingly.
12155 (vectorizable_store): Likewise.
12156 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12160 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12161 IFN_MASK_LOAD calls here rather than...
12162 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12165 Alan Hayward <alan.hayward@arm.com>
12166 David Sherwood <david.sherwood@arm.com>
12168 * expmed.c (extract_bit_field_1): For vector extracts,
12169 fall back to extract_bit_field_as_subreg if vec_extract
12172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12173 Alan Hayward <alan.hayward@arm.com>
12174 David Sherwood <david.sherwood@arm.com>
12176 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12177 they are variable or constant sized.
12178 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12179 slots for constant-sized data.
12181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12182 Alan Hayward <alan.hayward@arm.com>
12183 David Sherwood <david.sherwood@arm.com>
12185 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12186 handling COND_EXPRs with boolean comparisons, try to find a better
12187 basis for the mask type than the boolean itself.
12189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12191 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12192 is calculated and how it can be overridden.
12193 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12194 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12196 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12200 Alan Hayward <alan.hayward@arm.com>
12201 David Sherwood <david.sherwood@arm.com>
12203 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12204 Remove the mode argument.
12205 (aarch64_simd_valid_immediate): Remove the mode and inverse
12207 * config/aarch64/iterators.md (bitsize): New iterator.
12208 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12209 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12210 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12211 aarch64_simd_valid_immediate.
12212 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12213 (aarch64_reg_or_bic_imm): Likewise.
12214 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12215 with an insn_type enum and msl with a modifier_type enum.
12216 Replace element_width with a scalar_mode. Change the shift
12217 to unsigned int. Add constructors for scalar_float_mode and
12218 scalar_int_mode elements.
12219 (aarch64_vect_float_const_representable_p): Delete.
12220 (aarch64_can_const_movi_rtx_p)
12221 (aarch64_simd_scalar_immediate_valid_for_move)
12222 (aarch64_simd_make_constant): Update call to
12223 aarch64_simd_valid_immediate.
12224 (aarch64_advsimd_valid_immediate_hs): New function.
12225 (aarch64_advsimd_valid_immediate): Likewise.
12226 (aarch64_simd_valid_immediate): Remove mode and inverse
12227 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12228 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12229 and aarch64_float_const_representable_p on the result.
12230 (aarch64_output_simd_mov_immediate): Remove mode argument.
12231 Update call to aarch64_simd_valid_immediate and use of
12232 simd_immediate_info.
12233 (aarch64_output_scalar_simd_mov_immediate): Update call
12236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12237 Alan Hayward <alan.hayward@arm.com>
12238 David Sherwood <david.sherwood@arm.com>
12240 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12241 (mode_nunits): Likewise CONST_MODE_NUNITS.
12242 * machmode.def (ADJUST_NUNITS): Document.
12243 * genmodes.c (mode_data::need_nunits_adj): New field.
12244 (blank_mode): Update accordingly.
12245 (adj_nunits): New variable.
12246 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12248 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12249 listed in adj_nunits.
12250 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12251 listed in adj_nunits. Don't emit case statements for such modes.
12252 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12253 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12254 nothing if adj_nunits is nonnull.
12255 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12256 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12257 (emit_mode_fbit): Update use of print_maybe_const_decl.
12258 (emit_move_size): Likewise. Treat the array as non-const
12260 (emit_mode_adjustments): Handle adj_nunits.
12262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12264 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12265 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12266 (VECTOR_MODES): Use it.
12267 (make_vector_modes): Take the prefix as an argument.
12269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12270 Alan Hayward <alan.hayward@arm.com>
12271 David Sherwood <david.sherwood@arm.com>
12273 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12274 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12275 for MODE_VECTOR_BOOL.
12276 * machmode.def (VECTOR_BOOL_MODE): Document.
12277 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12278 (make_vector_bool_mode): New function.
12279 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12281 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12282 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12284 * stor-layout.c (int_mode_for_mode): Likewise.
12285 * tree.c (build_vector_type_for_mode): Likewise.
12286 * varasm.c (output_constant_pool_2): Likewise.
12287 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12288 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12289 for MODE_VECTOR_BOOL.
12290 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12291 of mode class checks.
12292 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12293 instead of a list of mode class checks.
12294 (expand_vector_scalar_condition): Likewise.
12295 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12298 Alan Hayward <alan.hayward@arm.com>
12299 David Sherwood <david.sherwood@arm.com>
12301 * machmode.h (mode_size): Change from unsigned short to
12303 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12304 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12305 or if measurement_type is not polynomial.
12306 (fixed_size_mode::includes_p): Check for constant-sized modes.
12307 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12308 return a poly_uint16 rather than an unsigned short.
12309 (emit_mode_size): Change the type of mode_size from unsigned short
12310 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12311 (emit_mode_adjustments): Cope with polynomial vector sizes.
12312 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12314 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12316 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12317 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12318 * caller-save.c (setup_save_areas): Likewise.
12319 (replace_reg_with_saved_mem): Likewise.
12320 * calls.c (emit_library_call_value_1): Likewise.
12321 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12322 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12323 (gen_lowpart_for_combine): Likewise.
12324 * convert.c (convert_to_integer_1): Likewise.
12325 * cse.c (equiv_constant, cse_insn): Likewise.
12326 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12327 (cselib_subst_to_values): Likewise.
12328 * dce.c (word_dce_process_block): Likewise.
12329 * df-problems.c (df_word_lr_mark_ref): Likewise.
12330 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12331 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12332 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12333 (rtl_for_decl_location): Likewise.
12334 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12335 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12336 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12337 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12338 (expand_expr_real_1): Likewise.
12339 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12340 (pad_below): Likewise.
12341 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12342 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12343 * ira.c (get_subreg_tracking_sizes): Likewise.
12344 * ira-build.c (ira_create_allocno_objects): Likewise.
12345 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12346 (ira_sort_regnos_for_alter_reg): Likewise.
12347 * ira-costs.c (record_operand_costs): Likewise.
12348 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12349 (resolve_simple_move): Likewise.
12350 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12351 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12352 (lra_constraints): Likewise.
12353 (CONST_POOL_OK_P): Reject variable-sized modes.
12354 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12355 (add_pseudo_to_slot, lra_spill): Likewise.
12356 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12357 * optabs-query.c (get_best_extraction_insn): Likewise.
12358 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12359 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12360 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12361 * recog.c (offsettable_address_addr_space_p): Likewise.
12362 * regcprop.c (maybe_mode_change): Likewise.
12363 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12364 * regrename.c (build_def_use): Likewise.
12365 * regstat.c (dump_reg_info): Likewise.
12366 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12367 (find_reloads, find_reloads_subreg_address): Likewise.
12368 * reload1.c (eliminate_regs_1): Likewise.
12369 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12370 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12371 (simplify_binary_operation_1, simplify_subreg): Likewise.
12372 * targhooks.c (default_function_arg_padding): Likewise.
12373 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12374 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12375 (verify_gimple_assign_ternary): Likewise.
12376 * tree-inline.c (estimate_move_cost): Likewise.
12377 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12378 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12379 (get_address_cost_ainc): Likewise.
12380 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12381 (vect_supportable_dr_alignment): Likewise.
12382 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12383 (vectorizable_reduction): Likewise.
12384 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12385 (vectorizable_operation, vectorizable_load): Likewise.
12386 * tree.c (build_same_sized_truth_vector_type): Likewise.
12387 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12388 * var-tracking.c (emit_note_insn_var_location): Likewise.
12389 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12390 (ADDR_VEC_ALIGN): Likewise.
12392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12393 Alan Hayward <alan.hayward@arm.com>
12394 David Sherwood <david.sherwood@arm.com>
12396 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12398 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12399 or if measurement_type is polynomial.
12400 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12401 * combine.c (make_extraction): Likewise.
12402 * dse.c (find_shift_sequence): Likewise.
12403 * dwarf2out.c (mem_loc_descriptor): Likewise.
12404 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12405 (extract_bit_field, extract_low_bits): Likewise.
12406 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12407 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12408 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12409 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12410 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12411 * reload.c (find_reloads): Likewise.
12412 * reload1.c (alter_reg): Likewise.
12413 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12414 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12415 * tree-if-conv.c (predicate_mem_writes): Likewise.
12416 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12417 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12418 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12419 * valtrack.c (dead_debug_insert_temp): Likewise.
12420 * varasm.c (mergeable_constant_section): Likewise.
12421 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12424 Alan Hayward <alan.hayward@arm.com>
12425 David Sherwood <david.sherwood@arm.com>
12427 * expr.c (expand_assignment): Cope with polynomial mode sizes
12428 when assigning to a CONCAT.
12430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12431 Alan Hayward <alan.hayward@arm.com>
12432 David Sherwood <david.sherwood@arm.com>
12434 * machmode.h (mode_precision): Change from unsigned short to
12436 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12438 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12439 or if measurement_type is not polynomial.
12440 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12441 in which the mode is already known to be a scalar_int_mode.
12442 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12443 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12445 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12446 for GET_MODE_PRECISION.
12447 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12448 for GET_MODE_PRECISION.
12449 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12451 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12452 (expand_field_assignment, make_extraction): Likewise.
12453 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12454 (get_last_value): Likewise.
12455 * convert.c (convert_to_integer_1): Likewise.
12456 * cse.c (cse_insn): Likewise.
12457 * expr.c (expand_expr_real_1): Likewise.
12458 * lra-constraints.c (simplify_operand_subreg): Likewise.
12459 * optabs-query.c (can_atomic_load_p): Likewise.
12460 * optabs.c (expand_atomic_load): Likewise.
12461 (expand_atomic_store): Likewise.
12462 * ree.c (combine_reaching_defs): Likewise.
12463 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12464 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12465 * tree.h (type_has_mode_precision_p): Likewise.
12466 * ubsan.c (instrument_si_overflow): Likewise.
12468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12469 Alan Hayward <alan.hayward@arm.com>
12470 David Sherwood <david.sherwood@arm.com>
12472 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12473 polynomial numbers of units.
12474 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12475 (valid_vector_subparts_p): New function.
12476 (build_vector_type): Remove temporary shim and take the number
12477 of units as a poly_uint64 rather than an int.
12478 (build_opaque_vector_type): Take the number of units as a
12479 poly_uint64 rather than an int.
12480 * tree.c (build_vector_from_ctor): Handle polynomial
12481 TYPE_VECTOR_SUBPARTS.
12482 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12483 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12484 (build_vector_from_val): If the number of units is variable,
12485 use build_vec_duplicate_cst for constant operands and
12486 VEC_DUPLICATE_EXPR otherwise.
12487 (make_vector_type): Remove temporary is_constant ().
12488 (build_vector_type, build_opaque_vector_type): Take the number of
12489 units as a poly_uint64 rather than an int.
12490 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12492 * cfgexpand.c (expand_debug_expr): Likewise.
12493 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12494 (store_constructor, expand_expr_real_1): Likewise.
12495 (const_scalar_mask_from_tree): Likewise.
12496 * fold-const-call.c (fold_const_reduction): Likewise.
12497 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12498 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12499 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12500 (fold_relational_const): Likewise.
12501 (native_interpret_vector): Likewise. Change the size from an
12502 int to an unsigned int.
12503 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12504 TYPE_VECTOR_SUBPARTS.
12505 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12506 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12507 duplicating a non-constant operand into a variable-length vector.
12508 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12509 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12510 * ipa-icf.c (sem_variable::equals): Likewise.
12511 * match.pd: Likewise.
12512 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12513 * print-tree.c (print_node): Likewise.
12514 * stor-layout.c (layout_type): Likewise.
12515 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12516 * tree-cfg.c (verify_gimple_comparison): Likewise.
12517 (verify_gimple_assign_binary): Likewise.
12518 (verify_gimple_assign_ternary): Likewise.
12519 (verify_gimple_assign_single): Likewise.
12520 * tree-pretty-print.c (dump_generic_node): Likewise.
12521 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12522 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12523 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12524 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12525 (vect_shift_permute_load_chain): Likewise.
12526 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12527 (expand_vector_condition, optimize_vector_constructor): Likewise.
12528 (lower_vec_perm, get_compute_type): Likewise.
12529 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12530 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12531 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12532 (vect_recog_mask_conversion_pattern): Likewise.
12533 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12534 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12535 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12536 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12537 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12538 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12539 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12540 (supportable_widening_operation): Likewise.
12541 (supportable_narrowing_operation): Likewise.
12542 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12544 * varasm.c (output_constant): Likewise.
12546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12547 Alan Hayward <alan.hayward@arm.com>
12548 David Sherwood <david.sherwood@arm.com>
12550 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12551 so that both the length == 3 and length != 3 cases set up their
12552 own permute vectors. Add comments explaining why we know the
12553 number of elements is constant.
12554 (vect_permute_load_chain): Likewise.
12556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12557 Alan Hayward <alan.hayward@arm.com>
12558 David Sherwood <david.sherwood@arm.com>
12560 * machmode.h (mode_nunits): Change from unsigned char to
12562 (ONLY_FIXED_SIZE_MODES): New macro.
12563 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12564 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12565 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12567 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12568 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12569 or if measurement_type is not polynomial.
12570 * genmodes.c (ZERO_COEFFS): New macro.
12571 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12573 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12574 Use ZERO_COEFFS when emitting initializers.
12575 * data-streamer.h (bp_pack_poly_value): New function.
12576 (bp_unpack_poly_value): Likewise.
12577 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12578 for GET_MODE_NUNITS.
12579 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12580 for GET_MODE_NUNITS.
12581 * tree.c (make_vector_type): Remove temporary shim and make
12582 the real function take the number of units as a poly_uint64
12583 rather than an int.
12584 (build_vector_type_for_mode): Handle polynomial nunits.
12585 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12586 * emit-rtl.c (const_vec_series_p_1): Likewise.
12587 (gen_rtx_CONST_VECTOR): Likewise.
12588 * fold-const.c (test_vec_duplicate_folding): Likewise.
12589 * genrecog.c (validate_pattern): Likewise.
12590 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12591 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12592 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12593 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12594 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12595 * rtlanal.c (subreg_get_info): Likewise.
12596 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12597 (vect_grouped_load_supported): Likewise.
12598 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12599 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12600 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12601 (simplify_const_unary_operation, simplify_binary_operation_1)
12602 (simplify_const_binary_operation, simplify_ternary_operation)
12603 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12604 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12605 instead of CONST_VECTOR_NUNITS.
12606 * varasm.c (output_constant_pool_2): Likewise.
12607 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12608 explicit-encoded elements in the XVEC for variable-length vectors.
12610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12612 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12615 Alan Hayward <alan.hayward@arm.com>
12616 David Sherwood <david.sherwood@arm.com>
12618 * coretypes.h (fixed_size_mode): Declare.
12619 (fixed_size_mode_pod): New typedef.
12620 * builtins.h (target_builtins::x_apply_args_mode)
12621 (target_builtins::x_apply_result_mode): Change type to
12622 fixed_size_mode_pod.
12623 * builtins.c (apply_args_size, apply_result_size, result_vector)
12624 (expand_builtin_apply_args_1, expand_builtin_apply)
12625 (expand_builtin_return): Update accordingly.
12627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12629 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12630 * cselib.c (cselib_hash_rtx): Likewise.
12631 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12632 CONST_VECTOR encoding.
12634 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12635 Jeff Law <law@redhat.com>
12638 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12639 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12640 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12641 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12644 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12645 explicitly probe *sp in a noreturn function if there were any callee
12646 register saves or frame pointer is needed.
12648 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12651 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12652 BLKmode for ternary, binary or unary expressions.
12655 * var-tracking.c (delete_vta_debug_insn): New inline function.
12656 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12657 insns from get_insns () to NULL instead of each bb separately.
12658 Use delete_vta_debug_insn. No longer static.
12659 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12660 delete_vta_debug_insns callers.
12661 * rtl.h (delete_vta_debug_insns): Declare.
12662 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12663 instead of variable_tracking_main.
12665 2018-01-03 Martin Sebor <msebor@redhat.com>
12667 PR tree-optimization/83603
12668 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12669 arguments past the endof the argument list in functions declared
12670 without a prototype.
12671 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12672 Avoid checking when arguments are null.
12674 2018-01-03 Martin Sebor <msebor@redhat.com>
12677 * doc/extend.texi (attribute const): Fix a typo.
12678 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12679 issuing -Wsuggest-attribute for void functions.
12681 2018-01-03 Martin Sebor <msebor@redhat.com>
12683 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12684 offset_int::from instead of wide_int::to_shwi.
12685 (maybe_diag_overlap): Remove assertion.
12686 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12687 * gimple-ssa-sprintf.c (format_directive): Same.
12688 (parse_directive): Same.
12689 (sprintf_dom_walker::compute_format_length): Same.
12690 (try_substitute_return_value): Same.
12692 2018-01-03 Jeff Law <law@redhat.com>
12694 PR middle-end/83654
12695 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12696 non-constant residual for zero at runtime and avoid probing in
12697 that case. Reorganize code for trailing problem to mirror handling
12700 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12702 PR tree-optimization/83501
12703 * tree-ssa-strlen.c (get_string_cst): New.
12704 (handle_char_store): Call get_string_cst.
12706 2018-01-03 Martin Liska <mliska@suse.cz>
12708 PR tree-optimization/83593
12709 * tree-ssa-strlen.c: Include tree-cfg.h.
12710 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12711 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12712 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12714 (strlen_dom_walker::before_dom_children): Call
12715 gimple_purge_dead_eh_edges. Dump tranformation with details
12717 (strlen_dom_walker::before_dom_children): Update call by adding
12718 new argument cleanup_eh.
12719 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12721 2018-01-03 Martin Liska <mliska@suse.cz>
12724 * cif-code.def (VARIADIC_THUNK): New enum value.
12725 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12728 2018-01-03 Jan Beulich <jbeulich@suse.com>
12730 * sse.md (mov<mode>_internal): Tighten condition for when to use
12731 vmovdqu<ssescalarsize> for TI and OI modes.
12733 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12735 Update copyright years.
12737 2018-01-03 Martin Liska <mliska@suse.cz>
12740 * ipa-visibility.c (function_and_variable_visibility): Skip
12741 functions with noipa attribure.
12743 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12745 * gcc.c (process_command): Update copyright notice dates.
12746 * gcov-dump.c (print_version): Ditto.
12747 * gcov.c (print_version): Ditto.
12748 * gcov-tool.c (print_version): Ditto.
12749 * gengtype.c (create_file): Ditto.
12750 * doc/cpp.texi: Bump @copying's copyright year.
12751 * doc/cppinternals.texi: Ditto.
12752 * doc/gcc.texi: Ditto.
12753 * doc/gccint.texi: Ditto.
12754 * doc/gcov.texi: Ditto.
12755 * doc/install.texi: Ditto.
12756 * doc/invoke.texi: Ditto.
12758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12760 * vector-builder.h (vector_builder::m_full_nelts): Change from
12761 unsigned int to poly_uint64.
12762 (vector_builder::full_nelts): Update prototype accordingly.
12763 (vector_builder::new_vector): Likewise.
12764 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12765 (vector_builder::operator ==): Likewise.
12766 (vector_builder::finalize): Likewise.
12767 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12768 Take the number of elements as a poly_uint64 rather than an
12770 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12771 from unsigned int to poly_uint64.
12772 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12773 (vec_perm_indices::new_vector): Likewise.
12774 (vec_perm_indices::length): Likewise.
12775 (vec_perm_indices::nelts_per_input): Likewise.
12776 (vec_perm_indices::input_nelts): Likewise.
12777 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12778 number of elements per input as a poly_uint64 rather than an
12779 unsigned int. Use the original encoding for variable-length
12780 vectors, rather than clamping each individual element.
12781 For the second and subsequent elements in each pattern,
12782 clamp the step and base before clamping their sum.
12783 (vec_perm_indices::series_p): Handle polynomial element counts.
12784 (vec_perm_indices::all_in_range_p): Likewise.
12785 (vec_perm_indices_to_tree): Likewise.
12786 (vec_perm_indices_to_rtx): Likewise.
12787 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12788 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12789 (tree_vector_builder::new_binary_operation): Handle polynomial
12790 element counts. Return false if we need to know the number
12791 of elements at compile time.
12792 * fold-const.c (fold_vec_perm): Punt if the number of elements
12793 isn't known at compile time.
12795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12797 * vec-perm-indices.h (vec_perm_builder): Change element type
12798 from HOST_WIDE_INT to poly_int64.
12799 (vec_perm_indices::element_type): Update accordingly.
12800 (vec_perm_indices::clamp): Handle polynomial element_types.
12801 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12802 (vec_perm_indices::all_in_range_p): Likewise.
12803 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12805 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12806 polynomial vec_perm_indices element types.
12807 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12808 * fold-const.c (fold_vec_perm): Likewise.
12809 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12810 * tree-vect-generic.c (lower_vec_perm): Likewise.
12811 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12812 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12813 element type to HOST_WIDE_INT.
12815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12816 Alan Hayward <alan.hayward@arm.com>
12817 David Sherwood <david.sherwood@arm.com>
12819 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12820 rather than an int. Use plus_constant.
12821 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12822 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12825 Alan Hayward <alan.hayward@arm.com>
12826 David Sherwood <david.sherwood@arm.com>
12828 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12829 a HOST_WIDE_INT to a poly_int64.
12831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12832 Alan Hayward <alan.hayward@arm.com>
12833 David Sherwood <david.sherwood@arm.com>
12835 * calls.c (load_register_parameters): Cope with polynomial
12836 mode sizes. Require a constant size for BLKmode parameters
12837 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12838 forces a parameter to be padded at the lsb end in order to
12839 fill a complete number of words, require the parameter size
12840 to be ordered wrt UNITS_PER_WORD.
12842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12843 Alan Hayward <alan.hayward@arm.com>
12844 David Sherwood <david.sherwood@arm.com>
12846 * reload1.c (spill_stack_slot_width): Change element type
12847 from unsigned int to poly_uint64_pod.
12848 (alter_reg): Treat mode sizes as polynomial.
12850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12851 Alan Hayward <alan.hayward@arm.com>
12852 David Sherwood <david.sherwood@arm.com>
12854 * reload.c (complex_word_subreg_p): New function.
12855 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
12862 operands whose sizes aren't ordered.
12863 (match_reload): Refer to this check here.
12865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12866 Alan Hayward <alan.hayward@arm.com>
12867 David Sherwood <david.sherwood@arm.com>
12869 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12870 that the mode size is in the set {1, 2, 4, 8, 16}.
12872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12873 Alan Hayward <alan.hayward@arm.com>
12874 David Sherwood <david.sherwood@arm.com>
12876 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12877 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12884 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12885 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12886 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12887 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12888 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12889 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12890 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12891 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12892 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12894 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12895 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12896 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12897 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12898 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12899 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12900 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12901 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12902 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12903 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12905 * expr.c (emit_move_resolve_push): Treat the input and result
12906 of PUSH_ROUNDING as a poly_int64.
12907 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12908 (emit_push_insn): Likewise.
12909 * lra-eliminations.c (mark_not_eliminable): Likewise.
12910 * recog.c (push_operand): Likewise.
12911 * reload1.c (elimination_effects): Likewise.
12912 * rtlanal.c (nonzero_bits1): Likewise.
12913 * calls.c (store_one_arg): Likewise. Require the padding to be
12914 known at compile time.
12916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12917 Alan Hayward <alan.hayward@arm.com>
12918 David Sherwood <david.sherwood@arm.com>
12920 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12921 Use plus_constant instead of gen_rtx_PLUS.
12923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12924 Alan Hayward <alan.hayward@arm.com>
12925 David Sherwood <david.sherwood@arm.com>
12927 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12928 rather than an int.
12930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12931 Alan Hayward <alan.hayward@arm.com>
12932 David Sherwood <david.sherwood@arm.com>
12934 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12935 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12936 via stack temporaries. Treat the mode size as polynomial too.
12938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12939 Alan Hayward <alan.hayward@arm.com>
12940 David Sherwood <david.sherwood@arm.com>
12942 * expr.c (expand_expr_real_2): When handling conversions involving
12943 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12944 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12945 as a poly_uint64 too.
12947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12948 Alan Hayward <alan.hayward@arm.com>
12949 David Sherwood <david.sherwood@arm.com>
12951 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12954 Alan Hayward <alan.hayward@arm.com>
12955 David Sherwood <david.sherwood@arm.com>
12957 * combine.c (can_change_dest_mode): Handle polynomial
12958 REGMODE_NATURAL_SIZE.
12959 * expmed.c (store_bit_field_1): Likewise.
12960 * expr.c (store_constructor): Likewise.
12961 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12962 and polynomial REGMODE_NATURAL_SIZE.
12963 (gen_lowpart_common): Likewise.
12964 * reginfo.c (record_subregs_of_mode): Likewise.
12965 * rtlanal.c (read_modify_subreg_p): 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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12972 numbers of elements.
12974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12975 Alan Hayward <alan.hayward@arm.com>
12976 David Sherwood <david.sherwood@arm.com>
12978 * match.pd: Cope with polynomial numbers of vector elements.
12980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12981 Alan Hayward <alan.hayward@arm.com>
12982 David Sherwood <david.sherwood@arm.com>
12984 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12985 in a POINTER_PLUS_EXPR.
12987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12988 Alan Hayward <alan.hayward@arm.com>
12989 David Sherwood <david.sherwood@arm.com>
12991 * omp-simd-clone.c (simd_clone_subparts): New function.
12992 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12993 (ipa_simd_modify_function_body): Likewise.
12995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12996 Alan Hayward <alan.hayward@arm.com>
12997 David Sherwood <david.sherwood@arm.com>
12999 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13000 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13001 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13002 (expand_vector_condition, vector_element): Likewise.
13003 (subparts_gt): New function.
13004 (get_compute_type): Use subparts_gt.
13005 (count_type_subparts): Delete.
13006 (expand_vector_operations_1): Use subparts_gt instead of
13007 count_type_subparts.
13009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13010 Alan Hayward <alan.hayward@arm.com>
13011 David Sherwood <david.sherwood@arm.com>
13013 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13014 (vect_compile_time_alias): ...this new function. Do the calculation
13015 on poly_ints rather than trees.
13016 (vect_prune_runtime_alias_test_list): Update call accordingly.
13018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13019 Alan Hayward <alan.hayward@arm.com>
13020 David Sherwood <david.sherwood@arm.com>
13022 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13024 (vect_schedule_slp_instance): Likewise.
13026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13027 Alan Hayward <alan.hayward@arm.com>
13028 David Sherwood <david.sherwood@arm.com>
13030 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13031 constant and extern definitions for variable-length vectors.
13032 (vect_get_constant_vectors): Note that the number of units
13033 is known to be constant.
13035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13036 Alan Hayward <alan.hayward@arm.com>
13037 David Sherwood <david.sherwood@arm.com>
13039 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13040 of units as polynomial. Choose between WIDE and NARROW based
13043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13044 Alan Hayward <alan.hayward@arm.com>
13045 David Sherwood <david.sherwood@arm.com>
13047 * tree-vect-stmts.c (simd_clone_subparts): New function.
13048 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13051 Alan Hayward <alan.hayward@arm.com>
13052 David Sherwood <david.sherwood@arm.com>
13054 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13055 vectors as polynomial. Use build_index_vector for
13056 IFN_GOMP_SIMD_LANE.
13058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13059 Alan Hayward <alan.hayward@arm.com>
13060 David Sherwood <david.sherwood@arm.com>
13062 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13063 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13064 for variable-length vectors.
13065 (vectorizable_mask_load_store): Treat the number of units as
13066 polynomial, asserting that it is constant if the condition has
13067 already been enforced.
13068 (vectorizable_store, vectorizable_load): Likewise.
13070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13071 Alan Hayward <alan.hayward@arm.com>
13072 David Sherwood <david.sherwood@arm.com>
13074 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13075 of units as polynomial. Punt if we can't tell at compile time
13076 which vector contains the final result.
13078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13079 Alan Hayward <alan.hayward@arm.com>
13080 David Sherwood <david.sherwood@arm.com>
13082 * tree-vect-loop.c (vectorizable_induction): Treat the number
13083 of units as polynomial. Punt on SLP inductions. Use an integer
13084 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13085 cast of such a series for variable-length floating-point
13088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13089 Alan Hayward <alan.hayward@arm.com>
13090 David Sherwood <david.sherwood@arm.com>
13092 * tree.h (build_index_vector): Declare.
13093 * tree.c (build_index_vector): New function.
13094 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13095 of units as polynomial, forcibly converting it to a constant if
13096 vectorizable_reduction has already enforced the condition.
13097 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13098 to create a {1,2,3,...} vector.
13099 (vectorizable_reduction): Treat the number of units as polynomial.
13100 Choose vectype_in based on the largest scalar element size rather
13101 than the smallest number of units. Enforce the restrictions
13104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13105 Alan Hayward <alan.hayward@arm.com>
13106 David Sherwood <david.sherwood@arm.com>
13108 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13109 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13116 * target.def (autovectorize_vector_sizes): Return the vector sizes
13117 by pointer, using vector_sizes rather than a bitmask.
13118 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13119 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13120 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13122 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13123 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13124 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13125 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13126 * omp-general.c (omp_max_vf): Likewise.
13127 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13128 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13129 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13130 * tree-vect-slp.c (vect_slp_bb): Likewise.
13131 * doc/tm.texi: Regenerate.
13132 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13134 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13135 the vector size as a poly_uint64 rather than an unsigned int.
13136 (current_vector_size): Change from an unsigned int to a poly_uint64.
13137 (get_vectype_for_scalar_type): Update accordingly.
13138 * tree.h (build_truth_vector_type): Take the size and number of
13139 units as a poly_uint64 rather than an unsigned int.
13140 (build_vector_type): Add a temporary overload that takes
13141 the number of units as a poly_uint64 rather than an unsigned int.
13142 * tree.c (make_vector_type): Likewise.
13143 (build_truth_vector_type): Take the number of units as a poly_uint64
13144 rather than an unsigned int.
13146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13147 Alan Hayward <alan.hayward@arm.com>
13148 David Sherwood <david.sherwood@arm.com>
13150 * target.def (get_mask_mode): Take the number of units and length
13151 as poly_uint64s rather than unsigned ints.
13152 * targhooks.h (default_get_mask_mode): Update accordingly.
13153 * targhooks.c (default_get_mask_mode): Likewise.
13154 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13155 * doc/tm.texi: Regenerate.
13157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13158 Alan Hayward <alan.hayward@arm.com>
13159 David Sherwood <david.sherwood@arm.com>
13161 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13162 * omp-general.c (omp_max_vf): Likewise.
13163 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13164 (expand_omp_simd): Handle polynomial safelen.
13165 * omp-low.c (omplow_simd_context): Add a default constructor.
13166 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13167 (lower_rec_simd_input_clauses): Update accordingly.
13168 (lower_rec_input_clauses): Likewise.
13170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13171 Alan Hayward <alan.hayward@arm.com>
13172 David Sherwood <david.sherwood@arm.com>
13174 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13175 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13176 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13177 (vect_analyze_slp_cost): Likewise.
13178 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13179 (vect_model_load_cost): Likewise.
13181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13182 Alan Hayward <alan.hayward@arm.com>
13183 David Sherwood <david.sherwood@arm.com>
13185 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13186 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13187 from an unsigned int * to a poly_uint64_pod *.
13188 (calculate_unrolling_factor): New function.
13189 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13192 Alan Hayward <alan.hayward@arm.com>
13193 David Sherwood <david.sherwood@arm.com>
13195 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13196 from an unsigned int to a poly_uint64.
13197 (_loop_vec_info::slp_unrolling_factor): Likewise.
13198 (_loop_vec_info::vectorization_factor): Change from an int
13200 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13201 (vect_get_num_vectors): New function.
13202 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13203 (vect_get_num_copies): Use vect_get_num_vectors.
13204 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13205 to an unsigned int *.
13206 (vect_analyze_data_refs): Change min_vf from an int * to a
13208 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13209 than an unsigned HOST_WIDE_INT.
13210 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13211 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13212 to an unsigned int *.
13213 (vect_analyze_data_ref_dependences): Likewise.
13214 (vect_compute_data_ref_alignment): Handle polynomial vf.
13215 (vect_enhance_data_refs_alignment): Likewise.
13216 (vect_prune_runtime_alias_test_list): Likewise.
13217 (vect_shift_permute_load_chain): Likewise.
13218 (vect_supportable_dr_alignment): Likewise.
13219 (dependence_distance_ge_vf): Take the vectorization factor as a
13220 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13221 (vect_analyze_data_refs): Change min_vf from an int * to a
13223 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13224 vfm1 as a poly_uint64 rather than an int. Make the same change
13225 for the returned bound_scalar.
13226 (vect_gen_vector_loop_niters): Handle polynomial vf.
13227 (vect_do_peeling): Likewise. Update call to
13228 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13229 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13231 * tree-vect-loop.c (vect_determine_vectorization_factor)
13232 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13233 (vect_get_known_peeling_cost): Likewise.
13234 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13235 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13236 (vect_transform_loop): Likewise. Use the lowest possible VF when
13237 updating the upper bounds of the loop.
13238 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13239 rather than an int.
13240 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13241 polynomial unroll factors.
13242 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13243 (vect_make_slp_decision): Likewise.
13244 (vect_supported_load_permutation_p): Likewise, and polynomial
13246 (vect_analyze_slp_cost): Handle polynomial vf.
13247 (vect_slp_analyze_node_operations): Likewise.
13248 (vect_slp_analyze_bb_1): Likewise.
13249 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13250 than an unsigned HOST_WIDE_INT.
13251 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13252 (vectorizable_load): Handle polynomial vf.
13253 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13255 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13258 Alan Hayward <alan.hayward@arm.com>
13259 David Sherwood <david.sherwood@arm.com>
13261 * match.pd: Handle bit operations involving three constants
13262 and try to fold one pair.
13264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13266 * tree-vect-loop-manip.c: Include gimple-fold.h.
13267 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13268 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13269 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13270 Add a path that uses a step of VF instead of 1, but disable it
13272 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13273 and niters_no_overflow parameters. Update calls to
13274 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13275 Create a new SSA name if the latter choses to use a ste other
13276 than zero, and return it via niters_vector_mult_vf_var.
13277 * tree-vect-loop.c (vect_transform_loop): Update calls to
13278 vect_do_peeling, vect_gen_vector_loop_niters and
13279 slpeel_make_loop_iterate_ntimes.
13280 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13281 (vect_gen_vector_loop_niters): Update declarations after above changes.
13283 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13285 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13286 128-bit round to integer instructions.
13287 (ceil<mode>2): Likewise.
13288 (btrunc<mode>2): Likewise.
13289 (round<mode>2): Likewise.
13291 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13293 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13294 unaligned VSX load/store on P8/P9.
13295 (expand_block_clear): Allow the use of unaligned VSX
13296 load/store on P8/P9.
13298 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13300 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13302 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13303 swap associated with both a load and a store.
13305 2018-01-02 Andrew Waterman <andrew@sifive.com>
13307 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13308 * config/riscv/riscv.md (clear_cache): Use it.
13310 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13312 * web.c: Remove out-of-date comment.
13314 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13316 * expr.c (fixup_args_size_notes): Check that any existing
13317 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13318 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13319 (emit_single_push_insn): ...here.
13321 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13323 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13324 (const_vector_encoded_nelts): New function.
13325 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13326 (const_vector_int_elt, const_vector_elt): Declare.
13327 * emit-rtl.c (const_vector_int_elt_1): New function.
13328 (const_vector_elt): Likewise.
13329 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13330 of CONST_VECTOR_ELT.
13332 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13334 * expr.c: Include rtx-vector-builder.h.
13335 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13336 directly on the tree encoding.
13337 (const_vector_from_tree): Likewise.
13338 * optabs.c: Include rtx-vector-builder.h.
13339 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13340 sequence of "u" values.
13341 * vec-perm-indices.c: Include rtx-vector-builder.h.
13342 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13343 directly on the vec_perm_indices encoding.
13345 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13347 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13348 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13349 * rtx-vector-builder.h: New file.
13350 * rtx-vector-builder.c: Likewise.
13351 * rtl.h (rtx_def::u2): Add a const_vector field.
13352 (CONST_VECTOR_NPATTERNS): New macro.
13353 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13354 (CONST_VECTOR_DUPLICATE_P): Likewise.
13355 (CONST_VECTOR_STEPPED_P): Likewise.
13356 (CONST_VECTOR_ENCODED_ELT): Likewise.
13357 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13358 (unwrap_const_vec_duplicate): Likewise.
13359 (const_vec_series_p): Check for a non-duplicated vector encoding.
13360 Say that the function only returns true for integer vectors.
13361 * emit-rtl.c: Include rtx-vector-builder.h.
13362 (gen_const_vec_duplicate_1): Delete.
13363 (gen_const_vector): Call gen_const_vec_duplicate instead of
13364 gen_const_vec_duplicate_1.
13365 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13366 (gen_const_vec_duplicate): Use rtx_vector_builder.
13367 (gen_const_vec_series): Likewise.
13368 (gen_rtx_CONST_VECTOR): Likewise.
13369 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13370 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13371 Build a new vector rather than modifying a CONST_VECTOR in-place.
13372 (handle_special_swappables): Update call accordingly.
13373 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13374 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13375 Build a new vector rather than modifying a CONST_VECTOR in-place.
13376 (handle_special_swappables): Update call accordingly.
13378 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13380 * simplify-rtx.c (simplify_const_binary_operation): Use
13381 CONST_VECTOR_ELT instead of XVECEXP.
13383 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13385 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13386 the selector elements to be different from the data elements
13387 if the selector is a VECTOR_CST.
13388 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13389 ssizetype for the selector.
13391 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13393 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13394 before testing each element individually.
13395 * tree-vect-generic.c (lower_vec_perm): Likewise.
13397 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13399 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13400 * selftest-run-tests.c (selftest::run_tests): Call it.
13401 * vector-builder.h (vector_builder::operator ==): New function.
13402 (vector_builder::operator !=): Likewise.
13403 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13404 (vec_perm_indices::all_from_input_p): New function.
13405 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13406 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13407 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13408 instead of reading the VECTOR_CST directly. Detect whether both
13409 vector inputs are the same before constructing the vec_perm_indices,
13410 and update the number of inputs argument accordingly. Use the
13411 utility functions added above. Only construct sel2 if we need to.
13413 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13415 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13416 the broadcast of the low byte.
13417 (expand_mult_highpart): Use an explicit encoding for the permutes.
13418 * optabs-query.c (can_mult_highpart_p): Likewise.
13419 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13420 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13421 (vectorizable_bswap): Likewise.
13422 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13423 explicit encoding for the power-of-2 permutes.
13424 (vect_permute_store_chain): Likewise.
13425 (vect_grouped_load_supported): Likewise.
13426 (vect_permute_load_chain): Likewise.
13428 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13430 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13431 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13432 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13433 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13434 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13435 (vect_gen_perm_mask_any): Likewise.
13437 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13439 * int-vector-builder.h: New file.
13440 * vec-perm-indices.h: Include int-vector-builder.h.
13441 (vec_perm_indices): Redefine as an int_vector_builder.
13442 (auto_vec_perm_indices): Delete.
13443 (vec_perm_builder): Redefine as a stand-alone class.
13444 (vec_perm_indices::vec_perm_indices): New function.
13445 (vec_perm_indices::clamp): Likewise.
13446 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13447 (vec_perm_indices::new_vector): New function.
13448 (vec_perm_indices::new_expanded_vector): Update for new
13449 vec_perm_indices class.
13450 (vec_perm_indices::rotate_inputs): New function.
13451 (vec_perm_indices::all_in_range_p): Operate directly on the
13452 encoded form, without computing elided elements.
13453 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13454 encoding. Update for new vec_perm_indices class.
13455 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13456 the given vec_perm_builder.
13457 (expand_vec_perm_var): Update vec_perm_builder constructor.
13458 (expand_mult_highpart): Use vec_perm_builder instead of
13459 auto_vec_perm_indices.
13460 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13461 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13462 or double series encoding as appropriate.
13463 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13464 vec_perm_indices instead of auto_vec_perm_indices.
13465 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13466 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13467 (vect_permute_store_chain): Likewise.
13468 (vect_grouped_load_supported): Likewise.
13469 (vect_permute_load_chain): Likewise.
13470 (vect_shift_permute_load_chain): Likewise.
13471 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13472 (vect_transform_slp_perm_load): Likewise.
13473 (vect_schedule_slp_instance): Likewise.
13474 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13475 (vectorizable_mask_load_store): Likewise.
13476 (vectorizable_bswap): Likewise.
13477 (vectorizable_store): Likewise.
13478 (vectorizable_load): Likewise.
13479 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13480 vec_perm_indices instead of auto_vec_perm_indices. Use
13481 tree_to_vec_perm_builder to read the vector from a tree.
13482 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13483 vec_perm_builder instead of a vec_perm_indices.
13484 (have_whole_vector_shift): Use vec_perm_builder and
13485 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13486 truncation to calc_vec_perm_mask_for_shift.
13487 (vect_create_epilog_for_reduction): Likewise.
13488 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13489 from auto_vec_perm_indices to vec_perm_indices.
13490 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13491 instead of changing individual elements.
13492 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13493 the vector in d.perm.
13494 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13495 from auto_vec_perm_indices to vec_perm_indices.
13496 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13497 instead of changing individual elements.
13498 (arm_vectorize_vec_perm_const): Use new_vector to install
13499 the vector in d.perm.
13500 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13501 Update vec_perm_builder constructor.
13502 (rs6000_expand_interleave): Likewise.
13503 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13504 (rs6000_expand_interleave): Likewise.
13506 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13508 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13509 to qimode could truncate the indices.
13510 * optabs.c (expand_vec_perm_var): Likewise.
13512 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13514 * Makefile.in (OBJS): Add vec-perm-indices.o.
13515 * vec-perm-indices.h: New file.
13516 * vec-perm-indices.c: Likewise.
13517 * target.h (vec_perm_indices): Replace with a forward class
13519 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13520 * optabs.h: Include vec-perm-indices.h.
13521 (expand_vec_perm): Delete.
13522 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13523 (expand_vec_perm_const): Declare.
13524 * target.def (vec_perm_const_ok): Replace with...
13525 (vec_perm_const): ...this new hook.
13526 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13527 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13528 * doc/tm.texi: Regenerate.
13529 * optabs.def (vec_perm_const): Delete.
13530 * doc/md.texi (vec_perm_const): Likewise.
13531 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13532 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13533 expand_vec_perm for constant permutation vectors. Assert that
13534 the mode of variable permutation vectors is the integer equivalent
13535 of the mode that is being permuted.
13536 * optabs-query.h (selector_fits_mode_p): Declare.
13537 * optabs-query.c: Include vec-perm-indices.h.
13538 (selector_fits_mode_p): New function.
13539 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13540 is defined, instead of checking whether the vec_perm_const_optab
13541 exists. Use targetm.vectorize.vec_perm_const instead of
13542 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13543 fit in the vector mode before using a variable permute.
13544 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13545 vec_perm_indices instead of an rtx.
13546 (expand_vec_perm): Replace with...
13547 (expand_vec_perm_const): ...this new function. Take the selector
13548 as a vec_perm_indices rather than an rtx. Also take the mode of
13549 the selector. Update call to shift_amt_for_vec_perm_mask.
13550 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13551 Use vec_perm_indices::new_expanded_vector to expand the original
13552 selector into bytes. Check whether the indices fit in the vector
13553 mode before using a variable permute.
13554 (expand_vec_perm_var): Make global.
13555 (expand_mult_highpart): Use expand_vec_perm_const.
13556 * fold-const.c: Includes vec-perm-indices.h.
13557 * tree-ssa-forwprop.c: Likewise.
13558 * tree-vect-data-refs.c: Likewise.
13559 * tree-vect-generic.c: Likewise.
13560 * tree-vect-loop.c: Likewise.
13561 * tree-vect-slp.c: Likewise.
13562 * tree-vect-stmts.c: Likewise.
13563 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13565 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13566 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13567 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13568 (aarch64_vectorize_vec_perm_const): ...this new function.
13569 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13570 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13571 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13572 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13573 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13574 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13575 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13577 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13578 check for NEON modes.
13579 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13580 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13581 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13582 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13584 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13585 the old VEC_PERM_CONST conditions.
13586 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13587 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13588 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13589 (ia64_vectorize_vec_perm_const_ok): Merge into...
13590 (ia64_vectorize_vec_perm_const): ...this new function.
13591 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13592 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13593 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13594 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13595 * config/mips/mips.c (mips_expand_vec_perm_const)
13596 (mips_vectorize_vec_perm_const_ok): Merge into...
13597 (mips_vectorize_vec_perm_const): ...this new function.
13598 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13599 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13600 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13601 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13602 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13603 (rs6000_expand_vec_perm_const): Delete.
13604 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13606 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13607 (altivec_expand_vec_perm_const_le): Take each operand individually.
13608 Operate on constant selectors rather than rtxes.
13609 (altivec_expand_vec_perm_const): Likewise. Update call to
13610 altivec_expand_vec_perm_const_le.
13611 (rs6000_expand_vec_perm_const): Delete.
13612 (rs6000_vectorize_vec_perm_const_ok): Delete.
13613 (rs6000_vectorize_vec_perm_const): New function.
13614 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13615 an element count and rtx array.
13616 (rs6000_expand_extract_even): Update call accordingly.
13617 (rs6000_expand_interleave): Likewise.
13618 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13619 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13620 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13621 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13622 (rs6000_expand_vec_perm_const): Delete.
13623 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13624 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13625 (altivec_expand_vec_perm_const_le): Take each operand individually.
13626 Operate on constant selectors rather than rtxes.
13627 (altivec_expand_vec_perm_const): Likewise. Update call to
13628 altivec_expand_vec_perm_const_le.
13629 (rs6000_expand_vec_perm_const): Delete.
13630 (rs6000_vectorize_vec_perm_const_ok): Delete.
13631 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13632 reference to the SPE evmerge intructions.
13633 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13634 an element count and rtx array.
13635 (rs6000_expand_extract_even): Update call accordingly.
13636 (rs6000_expand_interleave): Likewise.
13637 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13638 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13640 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13642 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13644 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13645 vector mode and that that mode matches the mode of the data
13647 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13648 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13649 directly using expand_vec_perm_1 when forcing selectors into
13651 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13653 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13655 * optabs-query.h (can_vec_perm_p): Delete.
13656 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13657 * optabs-query.c (can_vec_perm_p): Split into...
13658 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13659 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13660 particular selector is valid.
13661 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13662 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13663 (vect_grouped_load_supported): Likewise.
13664 (vect_shift_permute_load_chain): Likewise.
13665 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13666 (vect_transform_slp_perm_load): Likewise.
13667 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13668 (vectorizable_bswap): Likewise.
13669 (vect_gen_perm_mask_checked): Likewise.
13670 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13671 implementations of variable permutation vectors into account
13672 when deciding which selector to use.
13673 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13674 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13675 with a false third argument.
13676 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13677 to test whether the constant selector is valid and can_vec_perm_var_p
13678 to test whether a variable selector is valid.
13680 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13682 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13683 * optabs-query.c (can_vec_perm_p): Likewise.
13684 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13685 instead of vec_perm_indices.
13686 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13687 (vect_gen_perm_mask_checked): Likewise,
13688 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13689 (vect_gen_perm_mask_checked): Likewise,
13691 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13693 * optabs-query.h (qimode_for_vec_perm): Declare.
13694 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13695 (qimode_for_vec_perm): ...this new function.
13696 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13698 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13700 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13701 does not have a conditional at the top.
13703 2018-01-02 Richard Biener <rguenther@suse.de>
13705 * ipa-inline.c (big_speedup_p): Fix expression.
13707 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13710 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13713 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13717 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13718 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13719 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13720 cond_taken_branch_cost 3->4.
13722 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13724 PR tree-optimization/83581
13725 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13726 TODO_cleanup_cfg if any changes have been made.
13728 PR middle-end/83608
13729 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13730 convert_modes if target mode has the right side, but different mode
13733 PR middle-end/83609
13734 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13735 last argument when extracting from CONCAT. If either from_real or
13736 from_imag is NULL, use expansion through memory. If result is not
13737 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13738 the parts directly to inner mode, if even that fails, use expansion
13741 PR middle-end/83623
13742 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13743 check for bswap in mode rather than HImode and use that in expand_unop
13746 Copyright (C) 2018 Free Software Foundation, Inc.
13748 Copying and distribution of this file, with or without modification,
13749 are permitted in any medium without royalty provided the copyright
13750 notice and this notice are preserved.