1 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3 * gimple-match.h (gimple_match_op): New class.
4 (mprts_hook): Replace parameters with a gimple_match_op *.
5 (maybe_build_generic_op): Likewise.
6 (gimple_simplified_result_is_gimple_val): Replace parameters with
7 a const gimple_match_op *.
8 (gimple_simplify): Replace code_helper * and tree * parameters with
9 a gimple_match_op * parameter.
10 (gimple_resimplify1): Replace code_helper *, tree and tree *
11 parameters with a gimple_match_op * parameter.
12 (gimple_resimplify2): Likewise.
13 (gimple_resimplify3): Likewise.
14 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
15 parameters with a gimple_match_op * parameter.
16 * gimple-match-head.c (gimple_simplify): Change prototypes of
17 auto-generated functions to take a gimple_match_op * instead of
18 separate code_helper * and tree * parameters. Make the same
19 change in the top-level overload and update calls to the
20 gimple_resimplify routines. Update calls to the auto-generated
21 functions and to maybe_push_res_to_seq in the publicly-facing
22 operation-specific gimple_simplify overloads.
23 (gimple_match_op::MAX_NUM_OPS): Define.
24 (gimple_resimplify1): Replace rcode and ops with a single res_op
25 parameter. Update call to gimple_simplify.
26 (gimple_resimplify2): Likewise.
27 (gimple_resimplify3): Likewise.
28 (mprts_hook): Replace parameters with a gimple_match_op *.
29 (maybe_build_generic_op): Likewise.
30 (build_call_internal): Replace type, nargs and ops with
32 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
33 with a single gimple_match_op *. Update calls to mprts_hook,
34 build_call_internal and gimple_simplified_result_is_gimple_val.
35 Factor out code that is common to the tree_code and combined_fn cases.
36 * genmatch.c (expr::gen_transform): Replace tem_code and
37 tem_ops with a gimple_match_op called tem_op. Update calls
38 to the gimple_resimplify functions and maybe_push_res_to_seq.
39 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
40 res_ops. Update call to the gimple_resimplify functions.
41 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
42 (decision_tree::gen): Make the functions take a gimple_match_op *
43 called res_op instead of separate res_code and res_ops parameters.
44 Update call accordingly.
45 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
46 and ops with a single res_op parameter. Update calls to
47 maybe_build_generic_op and maybe_push_res_to_seq.
48 (fold_stmt_1): Update calls to gimple_simplify and
49 replace_stmt_with_simplification.
50 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
51 and gimple_simplified_result_is_gimple_val.
52 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
54 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
55 with a gimple_match_op *.
56 (vn_nary_build_or_lookup): Likewise. Update call to
57 vn_nary_build_or_lookup_1.
58 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
59 gimple_match_op *. Update calls to the gimple_resimplify routines
60 and to gimple_simplified_result_is_gimple_val.
61 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
62 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
63 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
64 (visit_nary_op): Likewise.
65 (visit_reference_op_load): Likewise.
67 2018-05-23 Luis Machado <luis.machado@linaro.org>
69 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
70 modifier for printing the step amount.
72 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
75 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
78 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
80 * doc/sourcebuild.texi (Endianness): New subsubsection.
82 2018-05-23 Luis Machado <luis.machado@linaro.org>
84 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
85 <prefetch_dynamic_strides>: New const bool field.
86 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
87 prefetch_dynamic_strides.
88 (exynosm1_prefetch_tune): Likewise.
89 (thunderxt88_prefetch_tune): Likewise.
90 (thunderx_prefetch_tune): Likewise.
91 (thunderx2t99_prefetch_tune): Likewise.
92 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
94 (aarch64_override_options_internal): Update to set
95 PARAM_PREFETCH_DYNAMIC_STRIDES.
96 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
97 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
98 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
99 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
100 prefetch-dynamic-strides setting.
102 2018-05-23 Luis Machado <luis.machado@linaro.org>
104 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
105 <minimum_stride>: New const int field.
106 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
107 minimum_stride field defaulting to -1.
108 (exynosm1_prefetch_tune): Likewise.
109 (thunderxt88_prefetch_tune): Likewise.
110 (thunderx_prefetch_tune): Likewise.
111 (thunderx2t99_prefetch_tune): Likewise.
112 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
113 <default_opt_level>: Set to 3.
114 (aarch64_override_options_internal): Update to set
115 PARAM_PREFETCH_MINIMUM_STRIDE.
116 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
117 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
118 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
119 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
120 stride is constant and is below the minimum stride threshold.
122 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
124 * config/arm/arm-cpus.in (mode26): Delete.
125 (armv4): Delete mode26 reference.
126 * config/arm/arm.c (arm_configure_build_target): Delete use of
129 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
131 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
133 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
134 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
136 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
137 (floatunsdidf2): Ditto.
139 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
141 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
142 (fixuns_trunc<mode>si2_avx512f): Ditto.
143 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
144 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
145 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
147 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
149 PR rtl-optimization/79985
150 * df-scan.c (df_insn_refs_collect): Remove special case for
151 global registers and asm statements.
153 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
155 * extend.texi (Global Register Variables): Rewrite the bullet list.
156 Note that the register is available for allocation. Note that access
157 via inline asm must use constraints. Add note about async-signal
158 handlers. Remove paragraph about automagic register selection.
160 2018-05-23 Richard Biener <rguenther@suse.de>
162 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
163 of fixed offset from memset VN.
165 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
167 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
169 (alloc_cand_and_find_basis): Initialize first_interp field.
170 (slsr_process_mul): Modify first_interp field.
171 (slsr_process_add): Likewise.
172 (slsr_process_cast): Modify first_interp field for each new
174 (slsr_process_copy): Likewise.
175 (dump_candidate): Dump first_interp field.
176 (replace_mult_candidate): Process all interpretations, not just
178 (replace_rhs_if_not_dup): Likewise.
179 (replace_one_candidate): Likewise.
181 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
183 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
185 (aarch64_needs_frame_chain): New function.
186 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
188 2018-05-23 Sudakshina Das <sudi.das@arm.com>
191 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
192 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
193 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
194 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
195 as true for strict-align.
196 (aarch64_can_inline_p): Perform checks even when callee has no
197 attributes to check for strict alignment.
198 * doc/extend.texi (AArch64 Function Attributes): Document
200 * doc/invoke.texi: (AArch64 Options): Likewise.
202 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
204 PR tree-optimization/85853
205 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
206 the handling of the root of the node to...
207 (vect_slp_analyze_node_operations_1): ...this new function,
208 and run the whole thing with the child nodes' def types
209 set according to their SLP node's def type.
211 2018-05-23 Richard Biener <rguenther@suse.de>
214 * tree-data-ref.c (create_runtime_alias_checks): Defer
215 and ignore overflow warnings.
217 2018-05-23 Yury Gribov <tetra2005@gmail.com>
219 PR tree-optimization/85822
220 * tree-vrp.c (is_masked_range_test): Fix handling of negative
223 2018-05-23 Richard Biener <rguenther@suse.de>
225 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
226 memset constants via native_interpret_expr.
228 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
231 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
233 (cgraph_node::create_alias): Likewise.
234 (cgraph_node::get_availability): Check ifunc_resolver instead
235 of looking up ifunc attribute.
236 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
237 * varasm.c (do_assemble_alias): Likewise.
238 (assemble_alias): Likewise.
239 (default_binds_local_p_3): Likewise.
240 * cgraph.h (cgraph_node): Add ifunc_resolver.
241 (cgraph_node::only_called_directly_or_aliased_p): Return false
243 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
245 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
246 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
247 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
248 instead of looking up ifunc attribute.
250 2018-05-22 Luis Machado <luis.machado@linaro.org>
252 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
254 2018-05-22 Martin Sebor <msebor@redhat.com>
257 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
258 only when expasion succeeds.
259 (expand_builtin_strcmp): Same.
260 (expand_builtin_strncmp): Same.
262 2018-05-22 Martin Sebor <msebor@redhat.com>
264 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
266 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
267 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
269 * config/aarch64/aarch64-ldpstp.md: Replace uses of
270 aarch64_mem_pair_operand with memory_operand and delete operand swapping
272 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
273 Add check for legitimate_address.
274 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
275 (aarch64_swap_ldrstr_operands): New.
276 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
279 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
280 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
282 * config/aarch64/aarch64.md: New patterns to generate stp
284 (store_pair_sw, store_pair_dw): New patterns to generate stp for
285 single words and double words.
286 (load_pair_sw, load_pair_dw): Likewise.
287 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
289 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
291 * config/aarch64/aarch64-ldpstp.md: Modify peephole
292 for different mode ldpstp and add peephole for merged zero stores.
294 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
296 (aarch64_gen_store_pair): Rename calls to match new patterns.
297 (aarch64_gen_load_pair): Rename calls to match new patterns.
298 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
299 (load_pair<DREG:mode><DREG2:mode>): ... This.
300 (store_pair<mode>): Rename to...
301 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
302 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
304 (V_INT_EQUIV): Handle SImode.
305 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
308 2018-05-22 Martin Sebor <msebor@redhat.com>
311 * calls.c (maybe_warn_nonstring_arg): Use string length to set
312 or ajust the presumed bound on an operation to avoid unnecessary
315 2018-05-22 Martin Sebor <msebor@redhat.com>
317 PR tree-optimization/85826
318 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
319 assuming that a DECL necesarily has a constant size.
321 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
324 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
326 2018-05-22 Richard Biener <rguenther@suse.de>
328 PR tree-optimization/85834
329 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
330 non-constant and non-zero memset arguments.
332 2018-05-22 Martin Liska <mliska@suse.cz>
335 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
337 2018-05-22 Richard Biener <rguenther@suse.de>
339 PR tree-optimization/85863
340 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
341 comparisons when vectype is specified.
342 (vectorizable_condition): Do not specify vectype for
343 vect_is_simple_cond when SLP vectorizing.
345 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
348 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
349 define __ibm128 as long double.
350 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
351 as a distinct type when IEEE 128-bit support is enabled.
352 (init_float128_ieee): Fix up conversions between IFmode and IEEE
353 128-bit types to use the correct functions.
354 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
355 convert between 128-bit floating point types that have different
356 modes but the same representation, instead of using gen_lowpart to
358 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
360 (IFKF_reg): New attributes to give the register constraints for
362 (extend<mode>tf2_internal): New insns to mark an explicit
363 conversion between 128-bit floating point types that have a
364 different mode but share the same representation.
366 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
368 PR tree-optimization/85814
369 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
370 a null return from get_strinfo when unsharing the next
371 strinfo in the chain.
373 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
376 * varasm.c (weak_finish): Clean up weak_decls.
378 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
380 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
381 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
382 UNSPEC_UADALP values.
383 * config/aarch64/iterators.md (ABAL): New int iterator.
386 (sur): Add mappings for the above.
387 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
389 (aarch64_<sur>abal<mode>_4): Likewise.
390 (aarch64_<sur>adalp<mode>_3): Likewise.
391 (<sur>sadv16qi): New define_expand.
393 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
395 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
396 (*movdf_internal): Ditto.
397 (*rcpsf2_sse): Ditto.
398 (*rsqrtsf2_sse): Ditto.
399 (*sqrt<mode>2_sse): Ditto.
401 2018-05-21 Tamar Christina <tamar.christina@arm.com>
403 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
405 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
406 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
407 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
408 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
410 * config/aarch64/arm_neon.h: Likewise.
411 * config/aarch64/iterators.md (VQ_I): New.
413 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
415 * config.gcc: Add arc/t-multilib-linux to tmake_file for
417 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
420 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
422 * config/nds32/constraints.md (S): New constraint.
423 * config/nds32/nds32.md (call_internal): Use constraint S.
424 (call_value_internal): Likewise.
425 (sibcall_internal): Likewise.
426 (sibcall_value_internal): Likewise.
428 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
429 Chung-Ju Wu <jasonwucj@gmail.com>
431 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
434 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
435 Chung-Ju Wu <jasonwucj@gmail.com>
437 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
438 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
439 (nds32_rtx_costs_impl): Simplify.
440 (nds32_address_cost_impl): Simplify.
441 (nds32_init_rtx_costs): New function.
442 (nds32_rtx_costs_speed_prefer): Likewise.
443 (nds32_rtx_costs_size_prefer): Likewise.
444 (nds32_address_cost_speed_prefer): Likewise.
445 (nds32_address_cost_speed_fwprop): Likewise.
446 (nds32_address_cost_size_prefer): Likewise.
447 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
448 * config/nds32/nds32.c (nds32_option_override): Use
449 nds32_init_rtx_costs function.
451 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
453 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
454 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
455 (TARGET_PIPELINE_N8): Likewise.
456 (TARGET_PIPELINE_N10): Likewise.
457 (TARGET_PIPELINE_N13): Likewise.
458 (TARGET_PIPELINE_GRAYWOLF): Likewise.
460 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
462 * config/nds32/nds32-fpu.md: Update copyright year.
464 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
466 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
468 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
470 * config/nds32/nds32.c
471 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
472 * config/nds32/nds32.opt (minline-asm-r15): New option.
474 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
476 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
478 * config/nds32/nds32.md (abssi2): New pattern.
480 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
482 * config/i386/i386.md (rex64namesuffix): New mode attribute.
483 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
484 Merge insn pattern from sse_cvtsi2ss<round_name> and
485 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
486 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
487 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
488 using SWI48 mode iterator.
489 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
490 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
491 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
492 pattern from sse_cvttss2si<round_saeonly_name>
493 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
494 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
495 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
496 using SWI48 mode iterator.
497 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
498 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
499 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
500 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
501 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
502 using SWI48 mode iterator.
503 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
504 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
505 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
506 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
507 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
509 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
510 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
511 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
512 pattern from sse_cvttsd2si<round_saeonly_name>
513 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
515 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
517 * config/nds32/nds32-md-auxiliary.c
518 (nds32_valid_smw_lwm_base_p): Refine.
519 (nds32_output_smw_single_word): Refine.
520 (nds32_output_smw_double_word): New.
521 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
523 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
525 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
526 (nds32_output_stack_pop): Refine.
527 (nds32_expand_unaligned_load): Refine.
528 (nds32_expand_unaligned_store): Refine.
530 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
531 Chung-Ju Wu <jasonwucj@gmail.com>
533 * config/nds32/constants.md: Add TP_REGNUM constant.
534 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
535 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
537 * config/nds32/nds32-doubleword.md: Consider flag_pic.
538 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
539 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
540 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
541 and PIC code generation.
542 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
544 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
546 * config/nds32/nds32.md: Support TLS and PIC.
547 * config/nds32/nds32.c: Support TLS and PIC.
548 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
549 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
552 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
554 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
557 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
558 Chung-Ju Wu <jasonwucj@gmail.com>
560 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
561 * config/nds32/nds32-md-auxiliary.c
562 (symbolic_reference_mentioned_p): New.
563 (nds32_legitimize_ict_address): New.
564 (nds32_expand_ict_move): New.
565 (nds32_indirect_call_referenced_p): New.
566 (nds32_symbol_binds_local_p): Delete.
567 (nds32_long_call_p): Modify.
568 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
569 * config/nds32/nds32-protos.h
570 (symbolic_reference_mentioned_p): Declare.
571 (nds32_legitimize_ict_address): Declare.
572 (nds32_expand_ict_move): Declare.
573 (nds32_indirect_call_referenced_p): Declare.
574 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
575 (nds32_relax_group): Use nds32_ict_const_p as condition.
576 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
577 (nds32_asm_file_start): Output ict_model directive in asm code.
578 (nds32_legitimate_address_p): Consider indirect call.
579 (nds32_print_operand): Consider indirect call.
580 (nds32_print_operand_address): Consider indirect call.
581 (nds32_insert_attributes): Handle "indirect_call" attribute.
582 (TARGET_LEGITIMATE_ADDRESS_P): Define.
583 (TARGET_LEGITIMATE_CONSTANT_P): Define.
584 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
585 (TARGET_DELEGITIMIZE_ADDRESS): Define.
586 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
587 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
588 (TARGET_ICT_MODEL_SMALL): Define.
589 (TARGET_ICT_MODEL_LARGE): Define.
590 * config/nds32/nds32.md (movsi): Consider ict model.
591 (call, call_value): Consider ict model.
592 (sibcall, sibcall_value): Consider ict model.
593 * config/nds32/nds32.opt (mict-model): New option.
594 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
597 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
598 Monk Chiang <sh.chiang04@gmail.com>
599 Jim Wilson <jimw@sifive.com>
601 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
602 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
603 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
604 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
605 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
606 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
607 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
608 compute save_libcall_adjustment properly.
609 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
610 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
611 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
612 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
613 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
614 (ABI_SPEC): Handle mabi=ilp32e.
615 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
617 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
618 <-march>: Add rv32e as an example.
620 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
623 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
624 (intra_create_variable_infos): Handle C++ constructors.
626 2018-05-18 Martin Liska <mliska@suse.cz>
628 * passes.def: Remove a redundant pass.
630 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
633 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
635 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
637 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
639 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
641 (armv2, armv2a, armv3, armv3m): Delete architectures.
642 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
643 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
644 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
646 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
647 (*mulsidi3adddi): Likewise.
648 (mulsidi3): Likewise.
649 (*mulsidi3_nov6): Likewise.
650 (umulsidi3): Likewise.
651 (umulsidi3_nov6): Likewise.
652 (umaddsidi4): Likewise.
653 (*umulsidi3adddi): Likewise.
654 (smulsi3_highpart): Likewise.
655 (*smulsi3_highpart_nov6): Likewise.
656 (umulsi3_highpart): Likewise.
657 (*umulsi3_highpart_nov6): Likewise.
658 * config/arm/arm.h (arm_arch3m): Delete.
659 * config/arm/arm.c (arm_arch3m): Delete.
660 (arm_option_override_internal): Update armv3-related comment.
661 (arm_configure_build_target): Delete use of isa_bit_mode32.
662 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
663 (arm_rtx_costs_internal): Delete check of arm_arch3m.
664 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
667 * config/arm/arm-protos.h (arm_arch3m): Delete.
668 * config/arm/arm-tables.opt: Regenerate.
669 * config/arm/arm-tune.md: Likewise.
670 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
671 deleted architectures.
673 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
675 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
676 (armv5t, armv5te): New features.
677 (ARMv5, ARMv5e): Delete fgroups.
678 (ARMv5t, ARMv5te): Adjust for above changes.
680 (armv5, armv5e): Delete arches.
681 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
683 (*call_reg_arm): Likewise.
684 (*call_value_reg_armv5): Likewise.
685 (*call_value_reg_arm): Likewise.
686 (*call_symbol): Likewise.
687 (*call_value_symbol): Likewise.
688 (*sibcall_insn): Likewise.
689 (*sibcall_value_insn): Likewise.
691 (prefetch): Likewise.
692 (define_split and define_peephole2 dependent on arm_arch5):
694 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
696 (TARGET_ARM_QBIT): Likewise.
697 (TARGET_DSP_MULTIPLY): Likewise.
698 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
699 (arm_arch5, arm_arch5e): Delete.
700 (arm_arch5t, arm_arch5te): Declare.
701 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
702 (arm_arch5t): Declare.
703 (arm_option_reconfigure_globals): Update for the above.
704 (arm_options_perform_arch_sanity_checks): Update comment, replace
705 use of arm_arch5 with arm_arch5t.
706 (use_return_insn): Likewise.
707 (arm_emit_call_insn): Likewise.
708 (output_return_instruction): Likewise.
709 (arm_final_prescan_insn): Likewise.
710 (arm_coproc_builtin_available): Likewise.
711 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
712 arm_arch5e with arm_arch5t and arm_arch5te.
713 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
714 (arm_arch5t, arm_arch5te): Declare.
715 * config/arm/arm-tables.opt: Regenerate.
716 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
717 * config/arm/t-multilib: Likewise.
718 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
719 instead of arm_arch5.
720 (*call_reg_thumb1): Likewise.
721 (*call_value_reg_thumb1_v5): Likewise.
722 (*call_value_reg_thumb1): Likewise.
723 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
725 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
727 2018-05-18 Martin Liska <mliska@suse.cz>
729 PR gcov-profile/84846
730 * doc/gcov.texi: Document -t option of gcov tool.
732 2018-05-18 Martin Liska <mliska@suse.cz>
734 PR gcov-profile/84846
735 * gcov.c (print_usage): Add new -t option.
736 (process_args): Handle the option.
737 (generate_results): Use stdout as output when requested by
740 2018-05-18 Martin Liska <mliska@suse.cz>
742 PR gcov-profile/84846
743 * coverage.c (coverage_init): Write PWD to .gcno file.
744 * doc/gcov.texi: Document how working directory is printed.
745 * gcov-dump.c (dump_gcov_file): Print PWD.
746 * gcov.c (output_intermediate_file): Likewise.
747 (read_graph_file): Read PWD string.
748 (output_lines): Print PWD.
750 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
753 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
754 for retval and return false if all args to phi are zero.
756 2018-05-18 Richard Biener <rguenther@suse.de>
758 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
760 (evrp_dom_walker::before_dom_children): Call it.
762 2018-05-18 Richard Biener <rguenther@suse.de>
764 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
765 results when processing array refs with variable index.
767 2018-05-18 Toon Moene <toon@moene.org>
769 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
770 directly after that of -floop-interchange. Indicate that both
771 options are enabled by default when specifying -O3.
773 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
775 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
776 iterator. Delete separate integer-mode vec_set<mode> expander.
777 (aarch64_simd_vec_setv2di): Delete.
778 (vec_setv2di): Delete.
779 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
780 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
781 the "w, r" alternative.
783 2018-05-18 Martin Liska <mliska@suse.cz>
785 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
786 * tree-pass.h (make_pass_lower_switch_O0): New function.
787 * tree-switch-conversion.c (node_has_low_bound): Remove.
788 (node_has_high_bound): Likewise.
789 (node_is_bounded): Likewise.
790 (class pass_lower_switch): Make it a template type and create
792 (pass_lower_switch::execute): Add template argument.
793 (make_pass_lower_switch): New function.
794 (make_pass_lower_switch_O0): New function.
795 (do_jump_if_equal): Remove.
796 (emit_case_nodes): Simplify to just handle all 3 cases and leave
797 all the hard work to tree optimization passes.
799 2018-05-18 Martin Liska <mliska@suse.cz>
801 * dbgcnt.c (limit_low): Renamed from limit.
802 (limit_high): New variable.
803 (dbg_cnt_is_enabled): Check for upper limit.
804 (dbg_cnt): Adjust dumping.
805 (dbg_cnt_set_limit_by_index): Add new argument for high
807 (dbg_cnt_set_limit_by_name): Likewise.
808 (dbg_cnt_process_single_pair): Parse new format.
809 (dbg_cnt_process_opt): Use strtok.
810 (dbg_cnt_list_all_counters): Remove 'value' and add
812 * doc/invoke.texi: Document changes.
814 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
816 * doc/sourcebuild.texi (scalar_all_fma): Document.
817 * tree.def (FMA_EXPR): Delete.
818 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
819 * internal-fn.c (ternary_direct): New macro.
820 (expand_ternary_optab_fn): Likewise.
821 (direct_ternary_optab_supported_p): Likewise.
822 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
823 * builtins.c (fold_builtin_fma): Delete.
824 (fold_builtin_3): Don't call it.
825 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
826 * expr.c (expand_expr_real_2): Likewise.
827 * fold-const.c (operand_equal_p): Likewise.
828 (fold_ternary_loc): Likewise.
829 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
830 * gimple.c (DEFTREECODE): Likewise.
831 * gimplify.c (gimplify_expr): Likewise.
832 * optabs-tree.c (optab_for_tree_code): Likewise.
833 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
834 * tree-eh.c (operation_could_trap_p): Likewise.
835 (stmt_could_throw_1_p): Likewise.
836 * tree-inline.c (estimate_operator_cost): Likewise.
837 * tree-pretty-print.c (dump_generic_node): Likewise.
838 (op_code_prio): Likewise.
839 * tree-ssa-loop-im.c (stmt_cost): Likewise.
840 * tree-ssa-operands.c (get_expr_operands): Likewise.
841 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
842 * fold-const-call.h (fold_fma): Delete.
843 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
844 CFN_FNMA and CFN_FNMS.
846 * genmatch.c (combined_fn): New enum.
847 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
848 (commutative_op): New function.
849 (commutate): Use it. Handle more than 2 operands.
850 (dt_operand::gen_gimple_expr): Use commutative_op.
851 (parser::parse_expr): Allow :c to be used with non-binary
852 operators if the commutative operand is known.
853 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
854 CFN_FMS, CFN_FNMA and CFN_FNMS.
855 (backprop::process_assign_use): Remove FMA_EXPR handling.
856 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
857 (gen_hsa_fma): New function.
858 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
859 IFN_FNMA and IFN_FNMS.
860 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
861 * gimple-fold.h (follow_all_ssa_edges): Declare.
862 * gimple-fold.c (follow_all_ssa_edges): New function.
863 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
864 gimple_build interface and use follow_all_ssa_edges to fold the result.
865 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
866 instead of checking for optabs directly.
867 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
868 rather than FMA_EXPRs.
869 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
870 call to IFN_FMA instead of an FMA_EXPR.
872 2018-05-17 Jim Wilson <jimw@sifive.com>
874 * expr.c (do_tablejump): When converting index to Pmode, if we have a
875 sign extended promoted subreg, and the range does not have the sign bit
876 set, then do a sign extend.
878 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
879 test, check for sign extended subreg and/or constant operands, and
880 do a sign extend in that case.
882 2018-05-17 Steve Ellcey <sellcey@cavium.com>
884 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
885 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
887 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
888 Change logics_shift_reg to logics_shift_imm.
889 (thunderx2t99_fp_loadpair_basic): Delete.
890 (thunderx2t99_fp_storepair_basic): Delete.
891 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
892 (thunderx2t99_asimd_polynomial): Delete.
893 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
894 and neon_fp_mul_d_scalar_q.
895 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
896 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
897 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
898 (thunderx2t99_asimd_lut): Add missing tbl types.
899 (thunderx2t99_asimd_ext): Delete.
900 (thunderx2t99_asimd_load1_1_mult): Delete.
901 (thunderx2t99_asimd_load1_2_mult): Delete.
902 (thunderx2t99_asimd_load1_ldp): New.
903 (thunderx2t99_asimd_load1): New.
904 (thunderx2t99_asimd_load2): Add missing *load2* types.
905 (thunderx2t99_asimd_load3): New.
906 (thunderx2t99_asimd_load4): New.
907 (thunderx2t99_asimd_store1_1_mult): Delete.
908 (thunderx2t99_asimd_store1_2_mult): Delete.
909 (thunderx2t99_asimd_store2_mult): Delete.
910 (thunderx2t99_asimd_store2_onelane): Delete.
911 (thunderx2t99_asimd_store_stp): New.
912 (thunderx2t99_asimd_store1): New.
913 (thunderx2t99_asimd_store2): New.
914 (thunderx2t99_asimd_store3): New.
915 (thunderx2t99_asimd_store4): New.
917 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
919 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
920 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
922 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
923 Segher Boessenkool <segher@kernel.crashing.org>
926 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
929 2018-05-17 Richard Biener <rguenther@suse.de>
931 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
932 for pruning loop and prune defs feeding only already visited PHIs.
934 2018-05-17 Richard Biener <rguenther@suse.de>
936 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
938 2018-05-17 Bin Cheng <bin.cheng@arm.com>
939 Richard Biener <rguenther@suse.de>
941 PR tree-optimization/85793
942 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
943 for VMAT_ELEMENTWISE.
945 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
947 * internal-fn.h (lookup_internal_fn): Declare
948 * internal-fn.c (lookup_internal_fn): New function.
949 * gimple.c (gimple_build_call_from_tree): Handle calls to
951 * gimple-pretty-print.c (dump_gimple_call): Print "." before
952 internal function names.
953 * tree-pretty-print.c (dump_generic_node): Likewise.
954 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
956 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
958 * gimple-fold.h (gimple_build): Make the function forms take
959 combined_fn rather than built_in_function.
960 (gimple_simplify): Likewise.
961 * gimple-match-head.c (gimple_simplify): Likewise.
962 * gimple-fold.c (gimple_build): Likewise.
963 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
964 rather than gimple_build_call_internal.
965 (get_initial_defs_for_reduction): Likewise.
966 (vect_create_epilog_for_reduction): Likewise.
967 (vectorizable_live_operation): Likewise.
969 2018-05-17 Martin Liska <mliska@suse.cz>
971 * gimple-ssa-sprintf.c (format_directive): Do not use
972 space in between 'G_' and '('.
974 2018-05-17 Jakub Jelinek <jakub@redhat.com>
977 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
978 even if the mask is not all ones.
981 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
983 (ix86_gimple_fold_builtin): Likewise.
986 * config/i386/i386.c: Include tree-vector-builder.h.
987 (ix86_vector_shift_count): New function.
988 (ix86_fold_builtin): Fold shift builtins by scalar count.
989 (ix86_gimple_fold_builtin): Likewise.
991 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
992 _mm512_setzero): New intrinsics.
994 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
995 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
997 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
998 code generation for cases where splatting a value is not useful.
999 * simplify-rtx.c (simplify_ternary_operation): Simplify
1000 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1001 mode to a vec_concat.
1003 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1005 * config.gcc: Support "goldmont-plus".
1006 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1008 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1009 PROCESSOR_GOLDMONT_PLUS.
1010 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1011 (processor_target_table): Add "goldmont-plus".
1012 (PTA_GOLDMONT_PLUS): Define.
1013 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1014 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1015 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1016 (fold_builtin_cpu): Add "goldmont-plus".
1017 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1018 (ix86_option_override_internal): Add "goldmont-plus".
1019 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1020 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1021 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1022 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1024 2018-05-17 Richard Biener <rguenther@suse.de>
1026 PR tree-optimization/85757
1027 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1028 remove defs that only feed that PHI from further processing.
1030 2018-05-16 Jim Wilson <jimw@sifive.com>
1032 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1034 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1036 2018-05-16 Mark Wielaard <mark@klomp.org>
1038 * dwarf2out.c (count_index_strings): New function.
1039 (output_indirect_strings): Call count_index_strings and generate
1040 header for dwarf_version >= 5.
1042 2018-05-16 Mark Wielaard <mark@klomp.org>
1044 * dwarf2out.c (dwarf_FORM): New function.
1045 (set_indirect_string): Use dwarf_FORM.
1046 (reset_indirect_string): Likewise.
1047 (size_of_die): Likewise.
1048 (value_format): Likewise.
1049 (output_die): Likewise.
1050 (add_skeleton_AT_string): Likewise.
1051 (output_macinfo_op): Likewise.
1052 (index_string): Likewise.
1053 (output_index_string_offset): Likewise.
1054 (output_index_string): Likewise.
1055 (count_index_strings): Likewise.
1057 2018-05-16 Carl Love <cel@us.ibm.com>
1059 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1060 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1062 2018-05-16 Martin Jambor <mjambor@suse.cz>
1064 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1065 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1067 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1069 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1070 (fnma<mode>4): Likewise.
1071 (fms<mode>4): Likewise.
1072 (fnms<mode>4): Likewise.
1073 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1074 (aarch64_fnma<mode>4): Likewise.
1075 (aarch64_fms<mode>4): Likewise.
1076 (aarch64_fnms<mode>4): Likewise.
1077 (aarch64_fnmadd<mode>4): Likewise.
1079 2018-05-16 Jason Merrill <jason@redhat.com>
1081 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1083 2018-05-16 Richard Biener <rguenther@suse.de>
1085 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1086 (dump_stmt_cost): Declare.
1087 (add_stmt_cost): Dump cost we add.
1088 (add_stmt_costs): New function.
1089 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1091 (vect_analyze_stmt): Adjust prototype.
1092 (vectorizable_condition): Likewise.
1093 (vectorizable_live_operation): Likewise.
1094 (vectorizable_reduction): Likewise.
1095 (vectorizable_induction): Likewise.
1096 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1097 cost vector to pass to vectorizable_ and record afterwards.
1098 (vect_model_reduction_cost): Take cost vector argument and adjust.
1099 (vect_model_induction_cost): Likewise.
1100 (vectorizable_reduction): Likewise.
1101 (vectorizable_induction): Likewise.
1102 (vectorizable_live_operation): Likewise.
1103 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1104 SLP_TREE_NUMBER_OF_VEC_STMTS.
1105 (vect_analyze_slp_cost_1): Remove.
1106 (vect_analyze_slp_cost): Likewise.
1107 (vect_slp_analyze_node_operations): Take visited args and
1108 a target cost vector. Avoid processing already visited stmt sets.
1109 (vect_slp_analyze_operations): Use a local cost vector to gather
1110 costs and register those of non-discarded instances.
1111 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1112 (vect_schedule_slp_instance): Remove copying of
1113 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1115 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1116 adding cost. Record cost entry location.
1117 (vect_prologue_cost_for_slp_op): Function to compute cost of
1118 a constant or invariant generated for SLP vect in the prologue,
1119 split out from vect_analyze_slp_cost_1.
1120 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1121 (vect_model_promotion_demotion_cost): Likewise.
1122 (vect_model_store_cost): Likewise, make static.
1123 (vect_model_load_cost): Likewise.
1124 (vectorizable_bswap): Add cost vector arg and adjust.
1125 (vectorizable_call): Likewise.
1126 (vectorizable_simd_clone_call): Likewise.
1127 (vectorizable_conversion): Likewise.
1128 (vectorizable_assignment): Likewise.
1129 (vectorizable_shift): Likewise.
1130 (vectorizable_operation): Likewise.
1131 (vectorizable_store): Likewise.
1132 (vectorizable_load): Likewise.
1133 (vectorizable_condition): Likewise.
1134 (vectorizable_comparison): Likewise.
1135 (can_vectorize_live_stmts): Likewise.
1136 (vect_analyze_stmt): Likewise.
1137 (vect_transform_stmt): Adjust calls to vectorizable_*.
1138 * tree-vectorizer.c: Include gimple-pretty-print.h.
1139 (dump_stmt_cost): New function.
1141 2018-05-16 Richard Biener <rguenther@suse.de>
1143 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1144 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1145 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1146 (check_name): New callback.
1147 (dse_classify_store): Track cycles via a visited bitmap of PHI
1148 defs and simplify handling of in-loop and across loop dead stores
1149 and properly fail for loop-variant refs. Handle byte-tracking with
1150 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1153 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1155 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1156 (vect_get_mask_type_for_stmt): Likewise.
1157 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1159 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1160 to determine the statement's vector type and the vector type that
1161 should be used for calculating nunits. Deal with cases in which
1162 the type has to be deferred.
1163 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1164 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1165 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1166 (vect_determine_vf_for_stmt): New functions, split out from...
1167 (vect_determine_vectorization_factor): ...here.
1168 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1169 (vect_get_mask_type_for_stmt): New functions, split out from
1170 vect_determine_vectorization_factor.
1172 2018-05-16 Richard Biener <rguenther@suse.de>
1174 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1175 verify the [VEC_]COND_EXPR embedded comparison.
1177 2018-05-15 Martin Sebor <msebor@redhat.com>
1179 PR tree-optimization/85753
1180 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1181 RECORD_TYPE in addition to ARRAY_TYPE.
1183 2018-05-15 Martin Sebor <msebor@redhat.com>
1186 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1188 2018-05-15 Richard Biener <rguenther@suse.de>
1190 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1191 add by_clobber_p one. Change algorithm to collect all defs
1192 representing uses we need to walk and try reducing them to
1193 a single one before failing.
1194 (dse_dom_walker::dse_optimize_stmt): Adjust.
1196 2018-05-13 Mark Wielaard <mark@klomp.org>
1198 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1199 (size_of_loc_descr): Likewise.
1200 (output_loc_operands): Likewise.
1201 (output_loc_operands_raw): Likewise.
1202 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1203 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1204 (hash_loc_operands): Likewise.
1205 (compare_loc_operands): Likewise.
1207 2018-05-14 Mark Wielaard <mark@klomp.org>
1209 * dwarf2out.c (count_index_addrs): New function.
1210 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1212 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1214 PR tree-optimization/83648
1215 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1216 return value as malloc candidate.
1218 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1221 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1222 as true in call to suggest_attribute.
1224 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1226 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1227 -mreadonly-in-sdata.
1229 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1231 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1233 (aarch64_crypto_aesd_fused): Likewise.
1235 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1237 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1238 (movsi_aarch64): Likewise.
1239 (load_pairsi): Likewise.
1240 (load_pairdi): Likewise.
1241 (store_pairsi): Likewise.
1242 (store_pairdi): Likewise.
1243 (load_pairsf): Likewise.
1244 (load_pairdf): Likewise.
1245 (store_pairsf): Likewise.
1246 (store_pairdf): Likewise.
1247 (zero_extend): Likewise.
1248 (trunc): Swap alternatives.
1249 (fcvt_target): Add '?' to prefer w over r.
1251 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1254 * config/i386/i386.md: Disallow non-commutative arithmetics in
1255 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1256 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1257 in the peephole2 before it.
1259 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1261 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1262 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1263 (ix86_handle_option): Handle -mcldemote.
1264 * config.gcc: New header.
1265 * config/i386/cldemoteintrin.h: New file.
1266 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1267 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1269 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1270 OPTION_MASK_ISA_CLDEMOTE.
1271 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1272 (ix86_valid_target_attribute_inner_p): Ditto.
1273 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1274 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1275 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1276 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1277 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1279 * config/i386/i386.opt: Add -mcldemote.
1280 * config/i386/x86intrin.h: New header.
1281 * doc/invoke.texi: Add -mcldemote.
1283 2018-05-14 Richard Biener <rguenther@suse.de>
1285 * doc/match-and-simplify.texi: Adjust :s documentation.
1287 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1289 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1290 intended memcpy size.
1291 (REORDER_45): Likewise.
1293 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1295 * sort.cc: New file.
1296 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1297 * vec.c (qsort_chk): Use gcc_qsort.
1298 * Makefile.in (OBJS-libcommon): Add sort.o.
1299 (build/sort.o): New target. Use it...
1300 (BUILD_RTL): ... here, and...
1301 (build/gencfn-macros): ... here, and...
1302 (build/genmatch): ... here.
1304 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1305 Chung-Ju Wu <jasonwucj@gmail.com>
1307 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1308 * config/nds32/nds32-graywolf.md: New file.
1309 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1310 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1312 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1313 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1314 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1315 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1316 * config/nds32/pipelines.md: Include n15 settings.
1318 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1319 Chung-Ju Wu <jasonwucj@gmail.com>
1321 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1322 * config/nds32/nds32-n13.md: New file.
1323 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1324 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1326 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1327 * config/nds32/nds32.md (pipeline_model): Add n13.
1328 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1329 * config/nds32/pipelines.md: Include n13 settings.
1331 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1332 Chung-Ju Wu <jasonwucj@gmail.com>
1334 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1335 * config/nds32/nds32-n10.md: New file.
1336 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1337 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1339 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1340 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1341 * config/nds32/nds32.md (pipeline_model): Add n10.
1342 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1343 * config/nds32/pipelines.md: Include n10 settings.
1345 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1346 Kito Cheng <kito.cheng@gmail.com>
1347 Chung-Ju Wu <jasonwucj@gmail.com>
1349 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1350 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1351 Add enum values for DSP extension instructions.
1352 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1354 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1355 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1357 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1358 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1359 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1360 * config/nds32/nds32-intrinsic.md: Likewise.
1361 * config/nds32/nds32_intrinsic.h: Likewise.
1362 * config/nds32/nds32-md-auxiliary.c: Likewise.
1363 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1364 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1365 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1366 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1367 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1368 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1369 TYPE_DMAC in switch statement.
1370 * config/nds32/nds32.c: New checking and implementation for DSP
1371 extension instructions.
1372 * config/nds32/nds32.h: Likewise.
1373 * config/nds32/nds32.md: Likewise.
1374 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1375 * config/nds32/predicates.md: Implement new predicates for DSP
1378 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1380 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1381 Reformat alternatives and attributes so it is easier to identify
1382 which constraints/attributes go with which instruction.
1383 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1384 (mov<mode>_softfloat32, FMOVE64): Likewise.
1385 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1386 (mov<mode>_softfloat64, FMOVE64): Likewise.
1388 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1390 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1392 (Basic PowerPC Built-in Functions): The new name of the
1393 subsection previously known as "PowerPC Built-in Functions".
1394 (Basic PowerPC Built-in Functions Available on all Configurations):
1396 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1397 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1398 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1399 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1401 2018-05-11 Martin Jambor <mjambor@suse.cz>
1404 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1407 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1410 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1412 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1414 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1415 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1416 (ix86_handle_option): Handle -mwaitpkg.
1417 * config.gcc: New header.
1418 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1419 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1420 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1422 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1423 OPTION_MASK_ISA_WAITPKG.
1424 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1425 (ix86_option_override_internal): Add PTA_WAITPKG.
1426 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1427 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1428 IX86_BUILTIN_TPAUSE.
1429 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1430 __builtin_ia32_umwait and __builtin_ia32_tpause.
1431 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1432 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1433 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1434 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1435 UNSPECV_TPAUSE): New.
1436 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1437 * config/i386/i386.opt: Add -mwaitpkg.
1438 * config/i386/waitpkgintrin.h: New file.
1439 * config/i386/x86intrin.h: New header.
1440 * doc/invoke.texi: Add -mwaitpkg.
1442 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1445 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
1447 (cortex-m0): Use armv6s-m isa.
1448 (cortex-m0plus): Likewise.
1449 (cortex-m1): Likewise.
1450 (cortex-m0.small-multiply): Likewise.
1451 (cortex-m0plus.small-multiply): Likewise.
1452 (cortex-m1.small-multiply): Likewise.
1454 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
1455 Jakub Jelinek <jakub@redhat.com>
1457 PR tree-optimization/85692
1458 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
1459 source permute as well.
1461 2018-05-11 Martin Liska <mliska@suse.cz>
1464 * doc/extend.texi: Document LLVM style format for no_sanitize
1467 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
1469 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
1470 mode_supports_vsx_dform_quad to mode_supports_dq_form.
1471 (mode_supports_vsx_dform_quad): Likewise.
1472 (mode_supports_vmx_dform): Move these functions to be next to the
1473 other mode_supports functions.
1474 (mode_supports_dq_form): Likewise.
1475 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
1476 mode_supports_dq_form.
1477 (reg_offset_addressing_ok_p): Likewise.
1478 (offsettable_ok_by_alignment): Likewise.
1479 (rs6000_legitimate_offset_address_p): Likewise.
1480 (legitimate_lo_sum_address_p): Likewise.
1481 (rs6000_legitimize_address): Likewise.
1482 (rs6000_legitimize_reload_address): Likewise.
1483 (rs6000_secondary_reload_inner): Likewise.
1484 (rs6000_preferred_reload_class): Likewise.
1485 (rs6000_output_move_128bit): Likewise.
1487 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1489 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
1490 Generate SImode target register for null target.
1491 <case IX86_BUILTIN_XGETBV>: Ditto.
1492 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
1493 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
1495 2018-05-10 Carl Love <cel@us.ibm.com>
1497 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1498 dcbtt and dcbtstt if operands[2] is 0.
1500 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
1503 * config/i386/sse.md (usadv64qi): New expander.
1505 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
1507 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
1508 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
1509 -maltivec=be support.
1510 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
1511 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
1512 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
1513 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
1514 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
1515 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
1516 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
1517 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
1518 altivec_vsumsws): Adjust.
1519 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
1520 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
1521 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
1523 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
1524 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
1525 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
1526 (altivec_lve<VI_char>x): Delete expand.
1527 (*altivec_lve<VI_char>x_internal): Rename to...
1528 (altivec_lve<VI_char>x): ... this.
1529 (altivec_lvxl_<mode>): Delete expand.
1530 (*altivec_lvxl_<mode>_internal): Rename to ...
1531 (altivec_lvxl_<mode>): ... this.
1532 (altivec_stvxl_<mode>): Delete expand.
1533 (*altivec_stvxl_<mode>_internal): Rename to ...
1534 (altivec_stvxl_<mode>): ... this.
1535 (altivec_stve<VI_char>x): Delete expand.
1536 (*altivec_stve<VI_char>x_internal): Rename to ...
1537 (altivec_stve<VI_char>x): ... this.
1538 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
1539 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
1540 reduc_plus_scal_<mode>): Adjust.
1541 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
1543 (rs6000_cpu_cpp_builtins): Adjust.
1544 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
1545 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
1546 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
1547 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
1548 -maltivec=be support.
1549 (rs6000_split_vec_extract_var): Adjust.
1550 (rs6000_split_v4si_init): Adjust.
1551 (swap_selector_for_mode): Delete.
1552 (altivec_expand_lvx_be, altivec_expand_stvx_be,
1553 altivec_expand_stvex_be): Delete.
1554 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
1555 -maltivec=be support.
1556 (rs6000_gimple_fold_builtin): Ditto.
1557 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
1559 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
1560 (TARGET_DIRECT_MOVE_64BIT): Adjust.
1561 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
1562 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
1563 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
1564 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
1565 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
1566 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
1567 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
1568 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
1569 anonymous split): Adjust.
1570 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
1571 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
1573 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
1575 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
1576 when --with-gxx-include-dir is also specified.
1577 * configure: Regenerate.
1579 2018-05-09 Jim Wilson <jimw@sifive.com>
1582 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
1583 * config/riscv/t-withmultilib: New.
1584 * config/riscv/withmultilib.h: New.
1585 * doc/install.texi: Document RISC-V --with-multilib-list support.
1587 2018-05-09 Richard Biener <rguenther@suse.de>
1589 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
1591 (vect_bb_vectorization_profitable_p): Adjust. Compute
1592 actual scalar cost using the cost vector and the add_stmt_cost
1595 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1597 PR rtl-optimization/85645
1598 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
1599 in the REG_CFA_REGISTER note for LR, don't leave it empty.
1601 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1603 PR rtl-optimization/85645
1604 * shrink-wrap.c (spread_components): Return a boolean saying if
1605 anything was changed.
1606 (try_shrink_wrapping_separate): Iterate spread_components until
1607 nothing changes anymore.
1609 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1611 PR rtl-optimization/85645
1612 * regrename.c (build_def_use): Also kill the chains that include the
1613 destination of a REG_CFA_REGISTER note.
1615 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
1617 PR rtl-optimization/85645
1618 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
1619 insn that has a REG_CFA_REGISTER note.
1621 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
1623 * cfgexpand.c (expand_clobber): New function.
1624 (expand_gimple_stmt_1): Use it.
1625 * tree-vect-stmts.c (vect_clobber_variable): New function,
1627 (vectorizable_simd_clone_call): ...here.
1628 (vectorizable_store): Emit a clobber either side of an
1629 IFN_STORE_LANES sequence.
1630 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1632 2018-05-09 Tom de Vries <tom@codesourcery.com>
1635 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1636 (define_insn "trap_if_false"): Add exit after trap.
1638 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1640 PR rtl-optimization/85638
1641 * bb-reorder.c: Include common/common-target.h.
1642 (create_forwarder_block): New function extracted from...
1643 (fix_up_crossing_landing_pad): ...here. Rename into...
1644 (dw2_fix_up_crossing_landing_pad): ...this.
1645 (sjlj_fix_up_crossing_landing_pad): New function.
1646 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1647 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1648 from both partitions and exit the loop after one iteration.
1650 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1653 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1655 (Basic PowerPC Built-in Functions): The new name of the
1656 subsection previously known as "PowerPC Built-in Functions".
1657 (Basic PowerPC Built-in Functions Available on all Configurations):
1659 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1661 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1663 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1665 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1668 2018-05-08 Jim Wilson <jimw@sifive.com>
1670 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1671 (LD_EMUL_SUFFIX): New.
1672 (LINK_SPEC): Use it.
1674 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1676 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1678 (Basic PowerPC Built-in Functions): The new name of the
1679 subsection previously known as "PowerPC Built-in Functions".
1680 (Basic PowerPC Built-in Functions Available on all Configurations):
1682 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1684 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1686 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1688 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1691 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1694 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1695 after cmpelim optimization.
1697 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1699 * config.gcc: Support "goldmont".
1700 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1701 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1703 * config/i386/i386.c (m_GOLDMONT): Define.
1704 (processor_target_table): Add "goldmont".
1705 (PTA_GOLDMONT): Define.
1706 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1707 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1708 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1709 (fold_builtin_cpu): Add "goldmont".
1710 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1711 (ix86_option_override_internal): Add "goldmont".
1712 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1713 (processor_type): Add PROCESSOR_GOLDMONT.
1714 * config/i386/i386.md: Add CPU "glm".
1715 * config/i386/glm.md: New file.
1716 * config/i386/x86-tune.def: Add m_GOLDMONT.
1717 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1719 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1722 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1724 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1725 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1726 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1729 * config/i386/i386.c (ix86_fold_builtin): Handle
1730 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1733 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1734 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1736 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1739 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1740 (check_arch): Likewise.
1741 (check_fpu): Return the result rather than printing it.
1742 (end arch): Fix operator precedence.
1743 (end cpu): Likewise.
1744 (END): Print the result from check_fpu.
1746 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1747 Alan Hayward <alan.hayward@arm.com>
1748 David Sherwood <david.sherwood@arm.com>
1750 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1751 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1752 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1753 (*fcmuo<mode>_and): New patterns.
1755 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1757 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1758 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1759 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1760 (cmp_op, sve_imm_con): New code attributes.
1761 (SVE_COND_INT_CMP, imm_con): Delete.
1762 (cmp_op): Remove above unspecs from int attribute.
1763 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1765 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1766 comparison-specific unspecs.
1767 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1768 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1769 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1770 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1771 (*vec_fcm<cmp_op><mode>): Rename to...
1772 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1773 (*vec_fcmuo<mode>): Rename to...
1774 (*fcmuo<mode>): ...this and adjust likewise.
1775 (*pred_fcm<cmp_op><mode>): New pattern.
1776 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1777 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1779 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1781 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1782 (aarch64_emit_sve_predicated_cond): New function.
1783 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1784 (aarch64_emit_unspec_cond_or): Replace with...
1785 (aarch64_emit_sve_or_conds): ...this new function. Use
1786 aarch64_emit_sve_ptrue_op for the individual comparisons and
1787 aarch64_emit_binop to OR them together.
1788 (aarch64_emit_inverted_unspec_cond): Replace with...
1789 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1790 aarch64_emit_sve_ptrue_op for the comparison and
1791 aarch64_emit_unop to invert the result.
1792 (aarch64_expand_sve_vec_cmp_float): Update after the above
1793 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1795 2018-05-07 Nathan Sidwell <nathan@acm.org>
1797 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1798 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1799 (Backwards Compatibility): Likewise.
1801 2018-05-07 Luis Machado <luis.machado@linaro.org>
1805 2018-05-07 Luis Machado <luis.machado@linaro.org>
1807 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1808 <prefetch_dynamic_strides>: New const bool field.
1809 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1810 prefetch_dynamic_strides.
1811 (exynosm1_prefetch_tune): Likewise.
1812 (thunderxt88_prefetch_tune): Likewise.
1813 (thunderx_prefetch_tune): Likewise.
1814 (thunderx2t99_prefetch_tune): Likewise.
1815 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1816 (aarch64_override_options_internal): Update to set
1817 PARAM_PREFETCH_DYNAMIC_STRIDES.
1818 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1819 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1820 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1821 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1822 prefetch-dynamic-strides setting.
1824 2018-05-07 Luis Machado <luis.machado@linaro.org>
1826 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1827 <minimum_stride>: New const int field.
1828 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1829 minimum_stride field.
1830 (exynosm1_prefetch_tune): Likewise.
1831 (thunderxt88_prefetch_tune): Likewise.
1832 (thunderx_prefetch_tune): Likewise.
1833 (thunderx2t99_prefetch_tune): Likewise.
1834 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1835 (aarch64_override_options_internal): Update to set
1836 PARAM_PREFETCH_MINIMUM_STRIDE.
1837 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1838 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1839 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1840 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1841 stride is constant and is below the minimum stride threshold.
1843 2018-05-07 Luis Machado <luis.machado@linaro.org>
1845 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1848 2018-05-07 Luis Machado <luis.machado@linaro.org>
1850 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1851 <prefetch_dynamic_strides>: New const bool field.
1852 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1853 prefetch_dynamic_strides.
1854 (exynosm1_prefetch_tune): Likewise.
1855 (thunderxt88_prefetch_tune): Likewise.
1856 (thunderx_prefetch_tune): Likewise.
1857 (thunderx2t99_prefetch_tune): Likewise.
1858 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1859 (aarch64_override_options_internal): Update to set
1860 PARAM_PREFETCH_DYNAMIC_STRIDES.
1861 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1862 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1863 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1864 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1865 prefetch-dynamic-strides setting.
1867 2018-05-07 Luis Machado <luis.machado@linaro.org>
1869 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1870 <minimum_stride>: New const int field.
1871 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1872 minimum_stride field.
1873 (exynosm1_prefetch_tune): Likewise.
1874 (thunderxt88_prefetch_tune): Likewise.
1875 (thunderx_prefetch_tune): Likewise.
1876 (thunderx2t99_prefetch_tune): Likewise.
1877 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1878 (aarch64_override_options_internal): Update to set
1879 PARAM_PREFETCH_MINIMUM_STRIDE.
1880 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1881 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1882 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1883 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1884 stride is constant and is below the minimum stride threshold.
1886 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1889 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1890 the type is addressable. Don't force op into register if it has
1893 2018-05-05 Roland McGrath <mcgrathr@google.com>
1896 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1897 any section for which we don't know a specific type it should have,
1898 regardless of name. Previously this was done only for the exact
1899 names ".init_array", ".fini_array", and ".preinit_array".
1900 (default_elf_asm_named_section): Add comment about
1901 relationship with default_section_type_flags and SECTION_NOTYPE.
1902 (get_section): Don't consider it a type conflict if one side has
1903 SECTION_NOTYPE and the other doesn't, as long as neither has the
1904 SECTION_BSS et al used in the default_section_type_flags logic.
1906 2018-05-05 Tom de Vries <tom@codesourcery.com>
1909 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1910 (workaround_barsyncs): New function.
1911 (nvptx_reorg): Use workaround_barsyncs.
1912 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1913 (define_expand "nvptx_membar_cta"): New define_expand.
1914 (define_insn "*nvptx_membar_cta"): New insn.
1916 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1918 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1919 To improve optimization opportunities.
1920 * builtin-types.def: The new needed builtin types for the above.
1922 2018-05-04 Richard Biener <rguenther@suse.de>
1924 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1925 * gimple-ssa-store-merging.c
1926 (imm_store_chain_info::output_merged_store): Remove redundant create,
1927 release split_store vector contents on failure.
1928 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1929 scalar stmt vector on cache hit.
1931 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1933 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1935 * config.gcc (powerpc-xilinx-eabi*): Remove.
1936 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1938 (fusion_addis_mem_combo_load): Ditto.
1939 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1941 (rs6000_cpu_cpp_builtins): Ditto.
1942 * config/rs6000/rs6000-linux.c
1943 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1944 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1945 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1947 (rs6000_setup_reg_addr_masks): Ditto.
1948 (rs6000_init_hard_regno_mode_ok): Ditto.
1949 (rs6000_option_override_internal): Ditto.
1950 (legitimate_lo_sum_address_p): Ditto.
1951 (rs6000_legitimize_address): Ditto.
1952 (rs6000_legitimize_reload_address): Ditto.
1953 (rs6000_legitimate_address_p): Ditto.
1954 (abi_v4_pass_in_fpr): Ditto.
1955 (setup_incoming_varargs): Ditto.
1956 (rs6000_gimplify_va_arg): Ditto.
1957 (rs6000_split_multireg_move): Ditto.
1958 (rs6000_savres_strategy): Ditto.
1959 (rs6000_emit_prologue_components): Ditto.
1960 (rs6000_emit_epilogue_components): Ditto.
1961 (rs6000_emit_prologue): Ditto.
1962 (rs6000_emit_epilogue): Ditto.
1963 (rs6000_elf_file_end): Ditto.
1964 (rs6000_function_value): Ditto.
1965 (rs6000_libcall_value): Ditto.
1966 * config/rs6000/rs6000.h: Ditto.
1967 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1968 (TARGET_MINMAX): ... this. New.
1969 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1970 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1971 (*movsi_internal1_single): Delete.
1972 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1973 mfpu=, mxilinx-fpu): Delete.
1974 * config/rs6000/singlefp.h: Delete.
1975 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1976 * config/rs6000/t-rs6000: Ditto.
1977 * config/rs6000/t-xilinx: Delete.
1978 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1979 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1980 (VStype_simple): Delete.
1981 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1982 * config/rs6000/xfpu.h: Delete.
1983 * config/rs6000/xfpu.md: Delete.
1984 * config/rs6000/xilinx.h: Delete.
1985 * config/rs6000/xilinx.opt: Delete.
1986 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1987 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1989 2018-05-04 Tom de Vries <tom@codesourcery.com>
1992 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1995 2018-05-04 Richard Biener <rguenther@suse.de>
1998 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1999 (expand_complex_div_wide): Likewise.
2000 (expand_complex_operations_1): Likewise.
2001 (expand_complex_libcall): Preserve EH info of the original stmt.
2002 (tree_lower_complex): Handle removed blocks.
2003 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2004 on complex multiplication and division libcall builtins.
2006 2018-05-04 Richard Biener <rguenther@suse.de>
2009 * fold-const.c (negate_expr_p): Restrict negation of operand
2010 zero of a division to when we know that can happen without
2012 (fold_negate_expr_1): Likewise.
2014 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2017 * real.h (real_nextafter): Declare.
2018 * real.c (real_nextafter): New function.
2019 * fold-const-call.c (fold_const_nextafter): New function.
2020 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2021 CASE_CFN_NEXTTOWARD.
2022 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2023 even when arg1_mode is different from arg0_mode.
2025 2018-05-03 Nathan Sidwell <nathan@acm.org>
2027 * doc/extend.texi (Deprecated Features): Remove
2029 (Backwards Compatibility): Likewise.
2030 * doc/invoke.texi (C++ Language Options): Likewise.
2031 (C++ Dialect Options): Likewise.
2033 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2036 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2037 _mm512_mask_mullox_epi64): New intrinsics.
2039 2018-05-03 Tom de Vries <tom@codesourcery.com>
2042 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2043 dump files): Add offload-tree.
2045 2018-05-03 Richard Biener <rguenther@suse.de>
2047 PR tree-optimization/85615
2048 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2049 to loops not nested in BBs loop father to avoid creating multi-entry
2052 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2054 PR tree-optimization/70291
2055 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2056 arguments. Change return type to tree. Emit libcall as a new
2057 statement rather than replacing existing one when inplace_p is true.
2058 (expand_complex_multiplication_components): New function.
2059 (expand_complex_multiplication): Expand floating-point complex
2060 multiplication using the above.
2061 (expand_complex_division): Rename inner_type parameter to type.
2062 Update expand_complex_libcall call-site.
2063 (expand_complex_operations_1): Update expand_complex_multiplication
2064 and expand_complex_division call-sites.
2066 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2069 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2070 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2071 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2072 the highest significant bit of the shift count mask is clear. In
2073 check whether and[sq]i3 is needed verify that all significant bits
2074 of the shift count other than the highest are set.
2076 2018-05-02 Tom de Vries <tom@codesourcery.com>
2079 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2080 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2081 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2082 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2083 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2084 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2085 __builtin_goacc_parlevel_size.
2087 2018-05-02 Richard Biener <rguenther@suse.de>
2089 PR tree-optimization/85597
2090 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2091 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2094 2018-05-02 Tom de Vries <tom@codesourcery.com>
2097 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2098 dump files): Add ltrans-tree.
2100 2018-05-02 Tom de Vries <tom@codesourcery.com>
2103 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2104 dump files): Add wpa-ipa.
2106 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2108 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2109 powerpc*-*-linux*paired* target.
2110 * config/rs6000/750cl.h: Delete.
2111 * config/rs6000/paired.h: Delete.
2112 * config/rs6000/paired.md: Delete.
2113 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2115 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2116 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2117 comment. Remove paired float support.
2118 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2119 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2121 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2122 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2124 * config/rs6000/rs6000.c: Remove paired float support.
2125 (paired_expand_vector_init, paired_expand_vector_move,
2126 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2127 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2128 paired_expand_builtin, paired_expand_predicate_builtin,
2129 paired_init_builtins): Delete.
2130 * config/rs6000/rs6000.h: Remove paired float support.
2131 * config/rs6000/rs6000.md: Remove paired float support.
2132 (move_from_CR_ov_bit): Delete.
2133 * config/rs6000/rs6000.opt (mpaired): Delete.
2134 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2135 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2137 2018-05-02 Richard Biener <rguenther@suse.de>
2140 * gimplify.c (gimplify_save_expr): When in SSA form allow
2141 SAVE_EXPRs to compute to SSA vars.
2143 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2146 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2147 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2148 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2149 clobber operands[2], instead use a new pseudo. Formatting fixes.
2151 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2153 PR tree-optimization/85586
2154 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2155 exit early for statements in the same group if the accesses are
2158 2018-05-02 Tom de Vries <tom@codesourcery.com>
2161 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2164 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2166 PR tree-optimization/85143
2167 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2169 2018-05-01 Tom de Vries <tom@codesourcery.com>
2172 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2173 not found" error message.
2175 2018-05-01 Tom de Vries <tom@codesourcery.com>
2178 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2179 * vec.c (test_ordered_remove_if): New function.
2180 (vec_c_tests): Call test_ordered_remove_if.
2181 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2182 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2183 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2184 VEC_ORDERED_REMOVE_IF.
2186 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2188 PR tree-optimization/82665
2189 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2190 pointer subtraction where arguments come from a memchr call.
2192 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2194 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2195 --push-state --as-needed and --pop-state instead of --as-needed and
2196 --no-as-needed if ld supports it.
2197 * configure: Regenerated.
2200 * doc/install.texi2html: Replace _002d with - and _002a with * in
2201 generated html files using sed.
2203 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2206 * gcc-rich-location.c (blank_line_before_p): New function.
2207 (use_new_line): New function.
2208 (gcc_rich_location::add_fixit_insert_formatted): New function.
2209 * gcc-rich-location.h
2210 (gcc_rich_location::add_fixit_insert_formatted): New function.
2212 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2214 * selftest.c (assert_streq): Rename "expected" and "actual" to
2215 "val1" and "val2". Extend NULL-handling to cover both inputs
2216 symmetrically, while still requiring both to be non-NULL for a pass.
2217 * selftest.h (assert_streq): Rename "expected" and "actual" to
2219 (ASSERT_EQ): Likewise.
2220 (ASSERT_EQ_AT): Likewise.
2221 (ASSERT_KNOWN_EQ): Likewise.
2222 (ASSERT_KNOWN_EQ_AT): Likewise.
2223 (ASSERT_NE): Likewise.
2224 (ASSERT_MAYBE_NE): Likewise.
2225 (ASSERT_MAYBE_NE_AT): Likewise.
2226 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2227 the assertion to pass.
2228 (ASSERT_STREQ_AT): Likewise.
2230 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2232 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2233 interaction with -pie.
2235 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2237 * selftest.h: Fix alphabetization of per-source-file selftest
2240 2018-04-30 Jason Merrill <jason@redhat.com>
2242 PR c++/61982 - dead stores to destroyed objects.
2243 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2246 2018-04-30 Jason Merrill <jason@redhat.com>
2248 * tree.c (build_clobber): New.
2249 * tree.h: Declare it.
2250 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2252 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2254 * diagnostic-show-locus.c (layout::layout): Update for
2255 location_get_source_line returning a char_span.
2256 (struct char_span): Move to input.h.
2257 (struct correction): Update for fields in char_span becoming
2259 (struct source_line): Update for location_get_source_line
2260 returning a char_span.
2261 (layout::print_line): Likewise.
2262 * edit-context.c (edited_file::print_content): Likewise.
2263 (edited_file::print_diff_hunk): Likewise.
2264 (edited_file::print_run_of_changed_lines): Likewise.
2265 (edited_file::get_num_lines): Likewise.
2266 (edited_line::edited_line): Likewise.
2267 * final.c (asm_show_source): Likewise.
2268 * input.c (location_get_source_line): Convert return type
2269 from const char * to char_span, losing the final "line_len"
2271 (dump_location_info): Update for the above.
2272 (get_substring_ranges_for_loc): Likewise. Use a char_span
2273 when handling the literal within the line.
2274 (test_reading_source_line): Update for location_get_source_line
2275 returning a char_span.
2276 * input.h (class char_span): Move here from
2277 diagnostic-show-locus.c, converting from a struct to a class.
2278 Make data members private.
2279 (char_span::operator bool): New.
2280 (char_span::length): New.
2281 (char_span::get_buffer): New.
2282 (char_span::operator[]): New.
2283 (char_span::subspan): Make const.
2284 (char_span::xstrdup): New.
2285 (location_get_source_line): Convert return type from const char *
2286 to char_span, losing the final "line_size" param.
2288 2018-04-30 Jan Hubicka <jh@suse.cz>
2290 * lto-wrapper.c (ltrans_priorities): New static var.
2291 (cmp_priority): New.
2292 (run_gcc): Read priorities and if doing parallel build order
2293 the Makefile by them.
2295 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2297 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2299 2018-04-30 Richard Biener <rguenther@suse.de>
2301 * tree-cfg.c (verify_address): Remove base argument, add
2302 flag whether to check TREE_ADDRESSABLE and do that.
2303 (verify_expr): Remove.
2304 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2305 (verify_gimple_assign_single): Likewise.
2306 (verify_gimple_switch): Likewise.
2307 (verify_expr_location_1): Dereference tp once. Add (disabled)
2308 piece from verify_expr.
2309 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2311 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2313 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2315 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2317 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2318 (small_data_pattern): Likewise.
2319 (arc_rewrite_small_data): Likewise.
2320 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2321 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2322 (get_symbol_alignment): New function.
2323 (legitimate_small_data_address_p): Likewise.
2324 (legitimate_scaled_address): Update, call
2325 legitimate_small_data_address_p.
2326 (output_sdata): New static variable.
2327 (arc_print_operand): Update how we handle small data operands.
2328 (arc_print_operand_address): Likewise.
2329 (arc_legitimate_address_p): Update, use
2330 legitimate_small_data_address_p.
2331 (arc_rewrite_small_data_p): Remove.
2332 (arc_rewrite_small_data_1): Likewise.
2333 (arc_rewrite_small_data): Likewise.
2334 (small_data_pattern): Likewise.
2335 (compact_sda_memory_operand): Update to use
2336 legitimate_small_data_address_p and get_symbol_alignment.
2337 (prepare_move_operands): Don't rewite sdata pattern.
2338 (prepare_extend_operands): Remove.
2339 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2341 (zero_extendqisi2): Likewise.
2342 (zero_extendhisi2): Likewise.
2343 (extendqihi2): Likewise.
2344 (extendqisi2): Likewise.
2345 (extendhisi2): Likewise.
2349 * config/arc/constraints.md (Usd): Change it to memory constraint.
2351 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2353 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2354 as source of std instructions.
2355 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2356 allow 6-bit constants as source for store instructions.
2357 (movdi_insn): Update instruction pattern to allow 6-bit constants
2358 as source for store instructions.
2360 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2362 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2364 2018-04-30 Nathan Sidwell <nathan@acm.org>
2365 Sandra Loosemore <sandra@codesourcery.com>
2367 * dumpfile.c (dump_open): Allow '-' for stdout.
2368 * doc/invoke.texi (Developer Options): Document dump filename
2369 determination early. Document stdin/stdout selection.
2371 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2373 Microblaze Target: PIC data text relative
2375 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2376 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2378 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2379 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2380 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2381 New addressing mode for data-text relative position indepenedent code.
2382 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2383 'ADDRESS_SYMBOLIC_TXT_REL'.
2384 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2385 (microblaze_legitimate_pic_operand): Exclude function calls from
2386 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2387 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2389 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2390 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2391 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2392 for 'address + offset'.
2393 (microblaze_expand_prologue): Add new function prologue call for
2395 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2396 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2397 table in case of TARGET_PIC_DATA_TEXT_REL.
2398 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2399 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2400 Add new macros 'UNSPEC_TEXT',
2401 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2402 + exclude function calls from 'UNSPEC_PLT' in case of data text
2404 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2405 new target hook for generating address diff vector tables in case of
2407 * doc/tm.texi : Regenerate.
2408 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2409 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2410 of addr diff vector generation.
2411 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2412 target hook definition.
2413 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2414 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2415 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2416 option for data text relative.
2418 2018-04-30 Richard Biener <rguenther@suse.de>
2420 * tree-chrec.h (evolution_function_is_constant_p): Remove
2422 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2424 2018-04-30 Richard Biener <rguenther@suse.de>
2427 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2429 2018-04-30 Richard Biener <rguenther@suse.de>
2431 PR tree-optimization/28364
2432 PR tree-optimization/85275
2433 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2434 copying first exit test.
2436 2018-04-28 Mark Wielaard <mark@klomp.org>
2438 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2440 (dwarf_AT): Handle DW_AT_addr_base.
2441 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
2443 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
2446 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
2447 (*ashl<dwi>3_doubleword_mask_1): Ditto.
2448 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
2449 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
2451 2018-04-28 Richard Biener <rguenther@suse.de>
2453 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
2454 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
2455 to reflect use. Only add interesting stmts.
2457 2018-04-27 Martin Jambor <mjambor@suse.cz>
2460 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
2461 the jump function allows for passing through aggregate values.
2463 2018-04-27 David Malcolm <dmalcolm@redhat.com>
2465 * input.h (in_system_header_at): Convert from macro to inline
2467 (from_macro_expansion_at): Likewise.
2468 (from_macro_definition_at): Likewise.
2470 2018-04-27 Jeff Law <law@redhat.com>
2472 * config.gcc: Mark tile* targets as deprecated/obsolete.
2474 2018-04-27 Richard Biener <rguenther@suse.de>
2476 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
2479 2018-04-27 Richard Biener <rguenther@suse.de>
2481 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
2483 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
2485 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
2486 with Yd constraint. Set "preferred_for_speed" attribute from
2487 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
2489 (*movdi_internal): Ditto.
2490 (movti_interunit splitters): Remove
2491 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
2492 (movdi_interunit splitters): Ditto.
2493 * config/i386/constraints.md (Ye): Remove.
2494 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
2496 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2499 * config/aarch64/constraints.md (Usg): Limit to 31.
2502 2018-04-27 Jakub Jelinek <jakub@redhat.com>
2504 PR tree-optimization/85529
2505 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
2506 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
2507 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
2508 zero extension or masking of the MSB bit.
2509 (optimize_range_tests): Add FIRST_BB argument, pass it through
2510 to optimize_range_tests_var_bound.
2511 (maybe_optimize_range_tests, reassociate_bb): Adjust
2512 optimize_range_tests callers.
2514 2018-04-26 Richard Biener <rguenther@suse.de>
2515 Jakub Jelinek <jakub@redhat.com>
2517 * cgraph.h (symbol_table): Just declare debug method here.
2518 * symtab.c (symbol_table::debug): Define.
2520 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
2522 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
2524 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
2526 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
2527 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
2528 (*movdi_internal): Substitute Yi and Yj constraint with x
2529 and Ym and Yn constraint with y constraint. Update "isa"
2530 attribute and set "preferred_for_speed" attribute from
2531 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
2532 (*movsi_internal): Ditto.
2533 (*movdf_internal): Ditto.
2534 (*movsf_internal): Ditto.
2535 (*zero_extendsidi2): Ditto.
2536 * config/i386/sse.md (vec_set<mode>_0): Ditto.
2537 (sse2_loadld): Ditto.
2538 (*vec_extract<ssevecmodelower>_0): Ditto.
2539 (*vec_extractv4si_0_zext_sse4): Ditto.
2540 (vec_concatv2di): Ditto.
2541 (*vec_dup<mode>): Ditto.
2542 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
2543 * config/i386/constraints.md (Yi): Remove.
2548 2018-04-26 Nathan Sidwell <nathan@acm.org>
2550 * dumpfile.c (dump_open): New.
2551 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
2552 (dump_finish): Detect stdio/stderr by value not name.
2554 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
2556 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
2558 2018-04-26 Tom de Vries <tom@codesourcery.com>
2561 * config/nvptx/nvptx.c (verify_neutering_jumps)
2562 (verify_neutering_labels): New function
2563 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
2565 2018-04-26 Tom de Vries <tom@codesourcery.com>
2568 * config/nvptx/nvptx.c (needs_neutering_p): New function.
2569 (nvptx_single): Use needs_neutering_p to skip over insns that do not
2572 2018-04-26 Richard Biener <rguenther@suse.de>
2573 Tom de Vries <tom@codesourcery.com>
2576 * lto-streamer-out.c (output_function): Fixup loops if required to match
2577 discovery done in the reader.
2579 2018-04-26 Richard Biener <rguenther@suse.de>
2581 PR tree-optimization/85116
2582 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
2583 have a loop exit from the single latch predecessor. Remove
2584 case of header with just condition.
2585 (ch_base::copy_headers): Exclude infinite loops from any
2587 (pass_ch::execute): Record exits.
2589 2018-04-26 Richard Biener <rguenther@suse.de>
2591 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
2592 prologue cost vector and pass it to vect_get_load_cost.
2593 (vect_get_peeling_costs_all_drs): Likewise.
2594 (vect_peeling_hash_get_lowest_cost): Likewise.
2595 (vect_enhance_data_refs_alignment): Likewise.
2597 2018-04-26 Richard Biener <rguenther@suse.de>
2600 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
2601 checking of integer<->pointer conversions.
2602 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
2603 sign-/zero-extending pointer types.
2604 (expand_omp_for_static_chunk): Likewise.
2606 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
2607 Jean Lee <xiaoyur347@gmail.com>
2609 * config/mips/mips.c (mips_asan_shadow_offset): New function.
2610 (TARGET_ASAN_SHADOW_OFFSET): Define.
2611 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
2612 true for -fsanitize=address.
2614 2018-04-25 Mark Wielaard <mark@klomp.org>
2616 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
2619 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2621 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
2622 than "alu", remove explicit "memory" and "imm_disp" attributes.
2623 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2626 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2627 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2628 gen_lowpart_no_emit.
2630 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2633 * config/i386/i386.c (ix86_expand_builtin): Change memory
2634 operand to XI, extend p0 to Pmode.
2635 * config/i386/i386.md: Change unspec volatile and operand
2636 1 mode to XI, change operand 0 mode to P.
2638 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2640 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2641 GET_MODE_MASK before any checking.
2642 (nds32_can_use_bset_p): Likewise.
2643 (nds32_can_use_btgl_p): Likewise.
2645 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2647 * config/nds32/nds32-doubleword.md: New define_split pattern for
2648 illegal register number.
2650 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2652 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2654 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2656 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2658 2018-04-25 Richard Biener <rguenther@suse.de>
2660 * lto-streamer.h (LTO_major_version): Bump to 8.
2662 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2664 * BASE-VER: Set to 9.0.0.
2666 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2668 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2669 in __abskf2 and __powikf2.
2671 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2674 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2675 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2676 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2677 Use the above on operand 2. Reindent.
2678 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2680 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2683 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2685 * config/i386/i386.opt (mcet): Removed.
2686 * doc/install.texi: Remove -mcet documentation.
2687 * doc/invoke.texi: Likewise.
2689 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2692 * doc/install.texi: Remove -mcet from bootstrap-cet.
2694 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2697 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2698 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2702 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2703 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2704 containing a CONST_VECTOR.
2706 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2708 * doc/install.texi: Update newlib dependency for nvptx.
2710 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2713 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2714 instead of INTVAL when shifting x left.
2716 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2718 PR tree-optimization/85478
2719 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2720 vect_grouped_store_supported for single element vectors.
2722 2018-04-24 Richard Biener <rguenther@suse.de>
2725 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2726 load cost increase to the case of non-constant step.
2728 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2731 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2732 destination if any_malformed_asm.
2734 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2737 * expr.c (store_field): In the bitfield case, if the value comes from
2738 a function call and is returned in registers by means of a PARALLEL,
2739 do not change the mode of the temporary unless BLKmode and VOIDmode.
2741 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2743 PR rtl-optimization/85423
2744 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2745 dependencies to debug insns when the previous insn is non-debug.
2747 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2749 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2750 enums into a single definition.
2751 (fls): Fix predicates and printing.
2754 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2756 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2757 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2758 and short u6 immediate.
2759 (check_if_valid_sleep_operand): Remove.
2760 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2762 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2764 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2765 flag_always_save_lp condition.
2766 * config/nds32/nds32.opt (malways-save-lp): New option.
2768 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2770 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2771 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2772 * config/nds32/nds32.h
2773 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2774 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2776 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2778 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2779 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2781 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2782 Chung-Ju Wu <jasonwucj@gmail.com>
2784 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2786 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2787 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2789 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2791 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2793 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2795 * config/nds32/nds32-protos.h (nds32_data_alignment,
2796 nds32_local_alignment): Declare.
2797 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2798 nds32_local_alignment): New functions.
2799 (TARGET_CONSTANT_ALIGNMENT): Define.
2800 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2802 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2804 * config/nds32/nds32.c
2805 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2806 (TARGET_MODES_TIEABLE_P): Likewise.
2808 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2810 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2813 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2814 Chung-Ju Wu <jasonwucj@gmail.com>
2816 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2817 for unaligned access.
2818 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2820 * config/nds32/nds32-intrinsic.md: Likewise.
2821 * config/nds32/nds32_intrinsic.h: Likewise.
2822 * config/nds32/nds32.h (nds32_builtins): Likewise.
2823 * config/nds32/nds32.opt (munaligned-access): New option.
2824 * config/nds32/nds32.c (nds32_asm_file_start): Display
2825 flag_unaligned_access status.
2827 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2829 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2830 -mno-relax is present.
2831 * config/riscv/linux.h (LINK_SPEC): Ditto.
2833 2018-04-20 Martin Sebor <msebor@redhat.com>
2836 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2838 (gimple_fold_builtin_stxcpy_chk): Same.
2839 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2841 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2844 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2845 __powikf2 when long double is IEEE 128-bit.
2847 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2849 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2850 step to make sure stack always aligned.
2852 2018-04-20 Carl Love <cel@us.ibm.com>
2855 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2856 size check for arg0.
2858 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2859 Tom de Vries <tom@codesourcery.com>
2862 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2863 Emit insns for calls too.
2864 (nvptx_find_par): Always look for worker-level predecessor insn.
2865 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2867 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2868 (nvptx_process_pars): Propagate frames for calls.
2870 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2873 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2875 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2876 (ix86_handle_option): Don't handle OPT_mibt.
2877 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2879 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2881 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2882 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2883 (ix86_target_macros): Define __CET__ with flag_cf_protection
2884 for -fcf-protection.
2885 * config/i386/i386.c (isa2_opts): Remove -mibt.
2886 * config/i386/i386.h (TARGET_IBT): Removed.
2887 (TARGET_IBT_P): Likewise.
2888 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2889 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2890 * config/i386/i386.opt (mcet): Update help message.
2893 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2894 -mcet as an alias for -mshstk.
2896 2018-04-20 Richard Biener <rguenther@suse.de>
2899 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2900 complexity by forcing a single use of the multiply operand.
2902 2018-04-20 Martin Jambor <mjambor@suse.cz>
2905 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2906 recursion dependency to only apply to non-clones.
2908 2018-04-20 Martin Jambor <mjambor@suse.cz>
2911 * ipa-cp.c (create_specialized_node): Check that clones of
2912 self-recursive edges exist during IPA-CP.
2914 2018-04-19 Toon Moene <toon@moene.org>
2916 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2919 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2921 PR tree-optimization/85467
2922 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2923 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2924 VECTOR_CST element to type.
2926 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2929 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2930 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2931 (builtin_longjmp): Likewise.
2932 (save_stack_nonlocal): New pattern.
2933 (restore_stack_nonlocal): Likewise.
2935 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2938 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2939 Replace ASM_OUTPUT_LABEL with fprintf.
2941 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2944 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2945 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2946 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2947 define __IBT__ and __SHSTK__ for -fcf-protection.
2948 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2950 (ix86_trampoline_init): Likewise.
2951 (x86_output_mi_thunk): Likewise.
2952 (ix86_notrack_prefixed_insn_p): Likewise.
2953 (ix86_option_override_internal): Don't disallow -fcf-protection.
2954 * config/i386/i386.md (rdssp<mode>): Also enable for
2956 (incssp<mode>): Likewise.
2957 (nop_endbr): Likewise.
2958 * config/i386/i386.opt (mcet): Change help message to built-in
2962 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2963 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2964 enable CET built-in functions.
2966 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2968 * common/config/i386/i386-common.c
2969 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2970 OPTION_MASK_ISA_MOVDIRI_UNSET,
2971 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2972 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2973 * config.gcc (movdirintrin.h): New header.
2974 * config/i386/cpuid.h (bit_MOVDIRI,
2975 bit_MOVDIR64B): New bits.
2976 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2978 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2979 (VOID, PVOID, PCVOID)): New function types.
2980 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2981 __builtin_ia32_directstoreu_u64,
2982 __builtin_ia32_movdir64b): New builtins.
2983 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2984 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2986 (ix86_valid_target_attribute_inner_p): Ditto.
2987 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2988 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2989 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2990 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2991 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2992 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2993 (movdiri<mode>, movdir64b_<mode>): New.
2994 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2995 * config/i386/immintrin.h: Include movdirintrin.h.
2996 * config/i386/movdirintrin.h: New file.
2997 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2999 2018-04-19 Richard Biener <rguenther@suse.de>
3002 * cfg.c (clear_bb_flags): When loop state says we have
3003 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3005 2018-04-19 Richard Biener <rguenther@suse.de>
3007 PR tree-optimization/84737
3008 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3009 copying restrict info.
3010 (vect_setup_realignment): Use it.
3011 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3012 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3013 the first DR to all generated stores.
3014 (vectorizable_load): Likewise for loads.
3016 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3018 PR tree-optimization/85446
3019 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3020 the integral and pointer types to have the same precision.
3022 * doc/install.texi: Document --disable-cet being the default and
3025 2018-04-18 Martin Liska <mliska@suse.cz>
3027 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3030 2018-04-18 Martin Liska <mliska@suse.cz>
3033 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3036 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3037 arguments if they are comparable.
3039 2018-04-18 Martin Liska <mliska@suse.cz>
3042 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3045 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3048 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3051 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3052 ENDBR after calling __morestack.
3054 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3057 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3058 by using gcc_base_ver to generate a gcc_driver_version, and use
3059 it when generating GCC_DRIVER_NAME.
3060 * configure: Regenerate.
3062 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3065 * config.gcc: Obsolete powerpc*-*-*spe*.
3067 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3070 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3071 (stabstr_D): Change type of unum from unsigned int to
3072 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3075 2018-04-17 Jim Wilson <jimw@sifive.com>
3078 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3079 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3080 Set arg_pointer_offset after using pretend_args_size.
3082 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3084 PR rtl-optimization/85431
3085 * dse.c (record_store): Ignore zero width stores.
3088 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3089 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3090 __builtin_stack_restore rather than after it.
3091 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3092 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3093 argument instead of virtual_dynamic_stack_rtx.
3095 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3097 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3099 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3100 Add note to error message to explain internal mapping of overloaded
3101 built-in function name to non-overloaded built-in function name.
3102 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3105 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3108 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3109 where the inputs overlap with the output.
3111 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3113 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3114 (=v, v) alternative and explicit "memory" attribute.
3115 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3116 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3118 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3119 "sselog1" type instead of "sselog".
3120 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3121 "sselog". Remove explicit "memory" attribute.
3122 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3123 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3125 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3126 "sselog1" type instead of "sselog". Remove explicit "memory"
3128 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3129 use "sselog1" type instead of "sselog". Remove explicit "memory"
3131 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3132 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3134 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3135 "sselog1" type instead of "sselog". Remove explicit "memory"
3137 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3138 use "sselog1" type instead of "sselog". Remove explicit "memory"
3142 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3145 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3148 2018-04-17 Martin Jambor <mjambor@suse.cz>
3151 * ipa-cp.c (create_specialized_node): Call
3152 expand_all_artificial_thunks if necessary.
3154 2018-04-17 Martin Liska <mliska@suse.cz>
3157 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3158 in message, remote space in between '_G' and '('.
3160 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3163 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3164 avx512f_vmcmp<mode>3<round_saeonly_name>,
3165 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3166 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3167 avx512f_rndscale<mode><round_saeonly_name>,
3168 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3169 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3170 Use %<iptr>2 instead of %2 for -masm=intel.
3171 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3172 avx512f_vcvttss2usi<round_saeonly_name>,
3173 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3175 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3176 avx512f_vcvttsd2usi<round_saeonly_name>,
3177 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3178 Use %q1 instead of %1 for -masm=intel.
3179 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3180 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3181 of %3 for -masm=intel.
3182 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3184 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3186 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3188 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3189 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3191 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3193 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3194 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3195 %g1 and one with %0 and %1.
3196 (avx512er_vmrcp28<mode><round_saeonly_name>,
3197 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3199 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3200 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3201 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3202 of %0 and %{%4%} for -masm=intel.
3203 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3204 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3205 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3206 order of %0 and %{%5%}%{z%} for -masm=intel.
3208 2018-04-17 Jan Hubicka <jh@suse.cz>
3211 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3213 2018-04-17 Martin Liska <mliska@suse.cz>
3216 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3217 for target_clone error message. Make default implementation
3218 clone to be a local declaration.
3219 (separate_attrs): Add new argument and check for an empty
3221 (expand_target_clones): Handle it.
3222 (ipa_target_clone): Make redirection just for target_clones
3225 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3226 Tom de Vries <tom@codesourcery.com>
3229 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3230 tiled basic blocks without omp continue statements.
3232 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3235 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3236 vec_extract expression as having side effects to make sure it gets
3239 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3242 * config/i386/i386.c (get_builtin_code_for_version): Check
3245 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3248 * gcc/config.gcc: Support "skylake".
3249 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3251 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3252 (processor_target_table): Add "skylake".
3253 (ix86_option_override_internal): Add "skylake".
3254 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3255 PROCESSOR_CANNONLAKE.
3256 (get_builtin_code_for_version): Fix priority for
3257 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3258 PROCESSOR_SKYLAKE-AVX512.
3259 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3260 (processor_type): Add PROCESSOR_SKYLAKE.
3262 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3263 Jason Merrill <jason@redhat.com>
3266 * convert.c (convert_to_integer_1): Use direct recursion for
3267 enumeral types and types with a precision less than the number
3268 of bits in their mode.
3270 2018-04-16 Julia Koval <julia.koval@intel.com>
3273 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3274 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3276 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3279 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3280 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3281 and -mno-direct-move.
3283 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3286 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3287 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3288 Ensure negative shifts result in {0}.
3290 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3292 PR rtl-optimization/79916
3293 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3294 regs (if any) to define how to gnerate SD moves when LRA is in
3297 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3299 PR rtl-optimization/85393
3300 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3301 * except.c (expand_dw2_landing_pad_for_region): Make static.
3302 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3303 a label and unconditional jump to old_bb, rather than
3304 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3307 PR rtl-optimization/85376
3308 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3309 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3310 instead of a specific value.
3312 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3313 Bin Cheng <bin.cheng@arm.com>
3315 PR tree-optimization/82965
3316 PR tree-optimization/83991
3317 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3318 by_profile_only parameter.
3319 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3320 information if the loop was predicted to iterate too many times.
3321 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3323 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3326 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3329 2018-04-13 Martin Liska <mliska@suse.cz>
3330 Jakub Jelinek <jakub@redhat.com>
3333 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3334 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3335 * builtins.c (expand_builtin_memory_copy_args): Use
3336 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3337 handle dest_addr == pc_rtx.
3339 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3342 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3343 asked to not generate direct moves.
3344 (fix_trunc<mode>si2_stfiwx): Similar.
3345 (fix_trunc<mode>si2_internal): Similar.
3347 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3350 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3351 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3352 lookup if dest in some wider mode is known to be const0_rtx and
3353 if so, record permanent equivalence for it to be ZERO_EXTEND of
3354 the narrower mode destination.
3356 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3358 * lto-streamer-out.c (output_function): Revert 259346.
3359 * omp-expand.c (expand_oacc_for): Likewise.
3361 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3363 PR rtl-optimization/85354
3364 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3365 * sel-sched.c (sel_global_init): ... here.
3367 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3370 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3371 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3372 mode for PE-COFF targets.
3373 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3374 (i386_pe_asm_lto_end): Likewise.
3375 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3376 (TARGET_ASM_LTO_END): Likewise.
3377 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3378 (i386_pe_asm_lto_start): New function.
3379 (i386_pe_asm_lto_end): Likewise.
3381 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3382 Richard Biener <rguenther@suse.de>
3385 * lto-streamer-out.c (output_function): Fix CFG loop state before
3387 * omp-expand.c (expand_oacc_for): Handle calls to internal
3388 functions like regular functions.
3390 2018-04-12 Richard Biener <rguenther@suse.de>
3393 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3394 for the early LTO debug to properly generate references to it
3395 during DIE emission. Do not re-use that for the skeleton for
3397 (dwarf2out_early_finish): Likewise.
3399 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3402 * config/i386/sse.md
3403 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3404 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3405 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3406 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3407 and output is a reg, avoid creating invalid lowpart subreg, but
3408 instead split into a 512-bit move. Don't split if not AVX512VL,
3409 input is xmm16+ reg and output is a mem.
3410 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3411 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3412 xmm16+ reg and output is a mem.
3414 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3416 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3417 also for flag_dwarf2_cfi_asm.
3419 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3421 PR rtl-optimization/85342
3422 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3423 a bool scalar var inside of the loop instead. Don't try to update
3424 recog_data.operand after failed apply_change_group.
3426 2018-04-12 Tom de Vries <tom@codesourcery.com>
3429 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3430 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3431 array with flexible array member as array without given dimension.
3432 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3433 argument for undefined param to true.
3435 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3438 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3439 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3440 from PowerPC section.
3441 * config/rs6000/sysv4.opt (mcall-): Improve help text.
3442 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
3443 help text that is too long.
3444 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
3445 help text that is too long.
3446 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
3447 help text that is too long.
3449 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
3451 * config/alpha/alpha.md (stack_probe_internal): Rename
3452 from "probe_stack". Update all callers.
3454 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3456 PR rtl-optimization/84566
3457 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
3458 sched_macro_fuse_insns.
3460 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3463 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
3464 (compute_block_dependences): ... from here.
3466 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3468 PR tree-optimization/85331
3469 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
3470 from int to HOST_WIDE_INT.
3472 2018-04-11 Martin Jambor <mjambor@suse.cz>
3475 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
3476 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
3477 not the same as the source val.
3478 (cgraph_edge_brings_value_p): New parameter.
3479 (gather_edges_for_value): Pass destination value to
3480 cgraph_edge_brings_value_p.
3481 (perhaps_add_new_callers): Likewise.
3482 (get_info_about_necessary_edges): Likewise and exclude values brought
3483 only by self-recursive edges.
3484 (create_specialized_node): Redirect only clones of self-calling edges.
3485 (+self_recursive_pass_through_p): New function.
3486 (find_more_scalar_values_for_callers_subset): Use it.
3487 (find_aggregate_values_for_callers_subset): Likewise.
3488 (known_aggs_to_agg_replacement_list): Removed.
3489 (decide_whether_version_node): Re-calculate known constants for all
3490 remaining context clones.
3492 2018-04-11 Richard Biener <rguenther@suse.de>
3495 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
3496 from early DWARF output.
3497 (dwarf2out_early_finish): Output line info unconditionally into
3498 early DWARF and add reference to it.
3500 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3503 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
3504 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
3505 other than V2DFmode using iptr mode attribute.
3506 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
3508 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
3510 PR rtl-optimization/84659
3511 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
3513 2018-04-11 Jakub Jelinek <jakub@redhat.com>
3516 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
3518 (output_loc_list): Pass address of a dummy size variable even in the
3519 locview handling loop.
3520 (index_location_lists): Add comment on why skip_loc_list_entry can't
3523 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
3526 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
3529 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
3532 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3533 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
3534 and -mstring-compare-inline-limit.
3536 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3539 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
3540 for stack clash protection in a register whenever we need it to be in
3543 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3545 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
3546 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
3548 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
3551 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
3553 (mlong-double-): Ditto.
3554 * config/rs6000/sysv4.opt (msdata=): Ditto.
3555 (mtls-size=): Ditto.
3557 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3559 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3560 erroneous entries for
3561 "vector int vec_ldl (int, long int *)", and
3562 "vector unsigned int vec_ldl (int, unsigned long int *)".
3563 Add comments and entries for
3564 "vector bool char vec_ldl (int, bool char *)",
3565 "vector bool short vec_ldl (int, bool short *)",
3566 "vector bool int vec_ldl (int, bool int *)",
3567 "vector bool long long vec_ldl (int, bool long long *)",
3568 "vector pixel vec_ldl (int, pixel *)",
3569 "vector long long vec_ldl (int, long long *)",
3570 "vector unsigned long long vec_ldl (int, unsigned long long *)".
3571 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
3572 type tree bool_long_long_type_node and correct definition of
3573 bool_V2DI_type_node to make reference to this new type tree.
3574 (rs6000_mangle_type): Replace erroneous reference to
3575 bool_long_type_node with bool_long_long_type_node.
3576 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
3577 comments to emphasize sign distinctions for char and int types and
3578 replace RS6000_BTI_bool_long constant with
3579 RS6000_BTI_bool_long_long constant. Also add comment to restrict
3580 use of RS6000_BTI_pixel.
3581 (bool_long_type_node): Remove this macro definition.
3582 (bool_long_long_type_node): New macro definition
3584 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3586 PR rtl-optimization/85300
3587 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
3588 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
3589 simplify_unary_operation fails.
3591 2018-04-10 Martin Liska <mliska@suse.cz>
3593 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
3594 cgraph_edge and ipa_ref.
3596 2018-04-10 Jakub Jelinek <jakub@redhat.com>
3600 * config/i386/sse.md
3601 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
3602 computation of the VEC_MERGE selector from mask.
3603 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
3604 Fix decoding of the VEC_MERGE selector into mask.
3606 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3608 PR tree-optimization/85286
3609 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
3611 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
3613 * final.c (final_1): Set insn_last_address as well as
3614 insn_current_address.
3616 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3619 * explow.c (emit_stack_probe): Call validize_mem on memory location
3620 before passing it to gen_probe_stack. Create address operand and
3621 legitimize it for the probe_stack_address case.
3623 2018-04-09 Jan Hubicka <jh@suse.cz>
3626 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3627 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3628 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3630 (free_lang_data): Rebuild type inheritance graph.
3632 2018-04-09 Martin Sebor <msebor@redhat.com>
3634 * invoke.texi (-finline-small-functions): Mention other optimization
3636 (-findirect-inlining, -fpartial-inlining): Same.
3637 (-finline-functions-called-once): Same.
3638 (-freorder-blocks-and-partition): Same.
3640 2018-04-09 Jan Hubicka <jh@suse.cz>
3643 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3644 jumps; choose last target that matches the criteria (i.e.
3645 no partition changes for non-crossing jumps).
3646 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3647 support for redirecting crossing jumps to non-crossing.
3649 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3651 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3652 also for naked functions.
3654 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3656 * config/arc/arc.md (add_shift): New pattern.
3657 (add_shift2): Likewise.
3658 (sub_shift): Likewise.
3659 (sub_shift_cmp0_noout): Likewise.
3660 (compare_si_ashiftsi): Likewise.
3661 (xbfu_cmp0_noout): New combine pattern.
3662 (xbfu_cmp0"): Likewise.
3663 (movsi_set_cc_insn): Place the predicable variant first.
3664 (commutative_binary_cmp0_noout): Remove clobber.
3665 (commutative_binary_cmp0): New pattern.
3666 (noncommutative_binary_cmp0): Likewise.
3667 (noncommutative_binary_cmp0_noout): Likewise.
3668 (noncommutative_binary_comparison_result_used): Removed.
3669 (rsub_cmp0): New pattern.
3670 (rsub_cmp0_noout): Likewise.
3671 (extzvsi): Changed, keep only meaningful variants.
3672 (SQH, SEZ): New iterators.
3673 (SQH_postfix): New mode attribute.
3674 (SEZ_prefix): New code attribute.
3675 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3676 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3677 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3679 (noncommutative_operator): Check the availability of barrel
3682 2018-04-09 Richard Biener <rguenther@suse.de>
3684 PR tree-optimization/85284
3685 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3686 Only use the niter constraining form of simple_iv when the exit
3689 2018-04-09 Tom de Vries <tom@codesourcery.com>
3692 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3693 (define_expand "*memory_barrier"): New define_expand.
3694 (define_insn "memory_barrier"): New insn.
3696 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3698 PR rtl-optimization/80463
3699 PR rtl-optimization/83972
3700 PR rtl-optimization/83480
3702 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3703 correct producer for the insn.
3704 (tidy_control_flow): Fixup seqnos in case of debug insns.
3706 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3708 PR rtl-optimization/83913
3710 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3711 different sched-times when merging exprs.
3713 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3715 PR rtl-optimization/83962
3717 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3718 tidy_fallthru_edge and tidy_control_flow.
3720 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3722 PR rtl-optimization/83530
3724 * sel-sched.c (force_next_insn): New global variable.
3725 (remove_insn_for_debug): When force_next_insn is true, also leave only
3726 next insn in the ready list.
3727 (sel_sched_region): When the region wasn't scheduled, make another pass
3728 over it with force_next_insn set to 1.
3730 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3732 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3734 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3735 for interrupt control.
3736 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3737 functions for interrupt control.
3738 * config/nds32/nds32-intrinsic.md: Likewise.
3739 * config/nds32/nds32_intrinsic.h: Likewise.
3740 * config/nds32/nds32.h (nds32_builtins): Likewise.
3742 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3744 * config/nds32/nds32.c (nds32_init_machine_status,
3745 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3746 strict_aligned_p field.
3747 (nds32_expand_to_rtl_hook): New function.
3748 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3749 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3751 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3752 Chung-Ju Wu <jasonwucj@gmail.com>
3754 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3755 * config/nds32/nds32-n7.md: New file.
3756 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3757 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3759 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3760 * config/nds32/nds32.md (pipeline_model): Add n7.
3761 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3762 * config/nds32/pipelines.md: Include n7 settings.
3764 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3765 Chung-Ju Wu <jasonwucj@gmail.com>
3767 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3768 * config/nds32/nds32-e8.md: New file.
3769 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3770 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3772 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3773 * config/nds32/nds32.md (pipeline_model): Add e8.
3774 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3775 * config/nds32/pipelines.md: Include e8 settings.
3777 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3778 Chung-Ju Wu <jasonwucj@gmail.com>
3780 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3781 * config/nds32/nds32-n8.md: New file.
3782 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3783 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3785 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3786 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3787 * config/nds32/nds32.md (pipeline_model): Add n8.
3788 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3789 * config/nds32/pipelines.md: Include n8 settings.
3791 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3792 Chung-Ju Wu <jasonwucj@gmail.com>
3794 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3795 * config/nds32/nds32-n9-2r1w.md: New file.
3796 * config/nds32/nds32-n9-3r2w.md: New file.
3797 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3798 nds32_register_ports): New or modify for cpu n9.
3799 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3801 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3802 * config/nds32/nds32-utils.c: New file.
3803 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3804 TARGET_MUL_SLOW): Define.
3805 * config/nds32/nds32.md (pipeline_model): New attribute.
3806 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3807 New options that support cpu n9.
3808 * config/nds32/pipelines.md: Include n9 settings.
3809 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3811 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3813 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3814 information if necessary.
3815 (output_cond_branch_compare_zero): Likewise.
3816 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3817 (nds32_target_alignment): Refine for alignment.
3818 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3819 (FUNCTION_BOUNDARY): Modify.
3820 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3822 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3824 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3826 * config/nds32/constants.md (unspec_volatile_element): Add values for
3827 TLB operation and data prefetch.
3828 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3829 functions for TLB operation and data prefetch.
3830 * config/nds32/nds32-intrinsic.md: Likewise.
3831 * config/nds32/nds32_intrinsic.h: Likewise.
3832 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3833 (nds32_print_operand): Likewise.
3834 * config/nds32/nds32.h (nds32_builtins): Likewise.
3836 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3837 Andrew Pinski <pinsika@gcc.gnu.org>
3840 * match.pd: Use constant_boolean_node of correct type instead of
3841 boolean_true_node or boolean_false_node for simplifying
3842 pointer comparisons to zero.
3844 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3846 PR tree-optimization/80021
3847 * tree.c (verify_type_variant): Make error call in verify_variant_match
3848 translatable and remove final full stop.
3850 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3852 * config/nds32/constants.md (unspec_volatile_element): Add
3853 UNSPEC_VOLATILE_EH_RETURN.
3854 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3855 nds32_output_stack_pop): Support dwarf exception handling process.
3856 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3857 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3858 exception handling process.
3859 (nds32_compute_stack_frame): Likewise.
3860 (nds32_return_addr_rtx): Likewise.
3861 (nds32_initial_elimination_offset): Likewise.
3862 (nds32_expand_prologue): Likewise.
3863 (nds32_expand_epilogue): Likewise.
3864 (nds32_dynamic_chain_address): New function.
3865 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3867 (DYNAMIC_CHAIN_ADDRESS): Define.
3868 (EH_RETURN_DATA_REGNO): Define.
3869 (EH_RETURN_STACKADJ_RTX): Define.
3870 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3871 patterns for dwarf exception handling.
3873 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3875 * config/nds32/nds32.h: Clean up obsolete macros.
3877 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3879 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3880 Add enum values for particular instructions.
3881 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3882 particular intrinsic functions.
3883 * config/nds32/nds32-intrinsic.md: Likewise.
3884 * config/nds32/nds32_intrinsic.h: Likewise.
3885 * config/nds32/nds32.h (nds32_builtins): Likewise.
3886 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3887 (btst, ave): New patterns for particular instructions.
3889 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3891 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3892 Add enum values for atomic load/store and memory sync.
3893 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3895 * config/nds32/nds32-intrinsic.md: Likewise.
3896 * config/nds32/nds32_intrinsic.h: Likewise.
3897 * config/nds32/nds32.h (nds32_builtins): Likewise.
3899 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3901 PR tree-optimization/85257
3902 * fold-const.c (native_encode_vector): If not all elts could fit
3903 and off is -1, return 0 rather than offset.
3904 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3905 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3906 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3907 adjust buffer in native_interpret_expr call.
3909 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3911 * config/nds32/constants.md (unspec_volatile_element): Add cache
3912 control enum values.
3913 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3914 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3915 * config/nds32/nds32.c (nds32_cctl_names): New.
3916 (nds32_print_operand): Handle cache control register names.
3917 * config/nds32/nds32.h (nds32_builtins): New enum values.
3918 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3920 * config/nds32/nds32.md (type): Add mmu.
3921 * config/nds32/pipelines.md (simple_insn): Add mmu.
3923 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3925 * config/nds32/nds32.md (type): Remove call.
3926 * config/nds32/pipelines.md (simple_insn): Likewise.
3928 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3930 * config/nds32/constants.md (unspec_volatile_element): Add
3931 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3932 UNSPEC_VOLATILE_FMFCFG.
3933 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3934 description for fmfcfg and fmfcsr.
3935 (bdesc_1arg): Add fmtcsr.
3936 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3937 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3938 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3939 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3940 unspec_fmfcfg): New patterns.
3941 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3942 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3943 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3944 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3945 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3946 __nds32__fmfcfg): Define.
3948 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3950 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3951 intrinsic register names.
3952 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3953 intrinsic register enum values and macros.
3955 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3957 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3958 for load/store addressing form.
3959 (nds32_print_operand_address): Likewise.
3961 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3964 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3965 based on LABEL_REF. Remove useless assertion.
3966 (pic_address_needs_scratch): Fix formatting.
3967 (sparc_legitimize_pic_address): Minor tweaks.
3968 (sparc_delegitimize_address): Adjust assertion accordingly.
3969 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3970 into symbolic_operand.
3971 (movsi_high_pic_label_ref): Likewise.
3972 (movsi_lo_sum_pic_label_ref): Likewise.
3973 (movdi_pic_label_ref): Likewise.
3974 (movdi_high_pic_label_ref): Likewise.
3975 (movdi_lo_sum_pic_label_ref): Likewise.
3977 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3979 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3980 custom LIB_SPEC setup.
3982 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3983 Kito Cheng <kito.cheng@gmail.com>
3985 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3986 * config/riscv/freebsd.h: New.
3988 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3990 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3991 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3994 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3995 Kito Cheng <kito.cheng@gmail.com>
3997 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3998 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3999 * config/nds32/nds32-protos.h (nds32_output_call,
4000 nds32_output_return): Declare.
4001 * config/nds32/nds32.md: Refine all the call and return patterns.
4003 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4006 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4007 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4009 PR rtl-optimization/84872
4010 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4011 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4014 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4016 * expr.c (copy_blkmode_to_reg): Revert 254862.
4017 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4019 2018-04-06 Richard Biener <rguenther@suse.de>
4022 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4023 after seeing a component reference with an adjacent field. Treat
4024 refs to arrays at struct end of external decls similar to
4025 refs to unconstrained commons.
4027 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4030 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4031 look through SAVE_EXPRs with non-side-effects argument. Adjust
4033 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4036 2018-04-06 Richard Biener <rguenther@suse.de>
4039 * alias.c (find_base_term): New wrapper around find_base_term
4040 unwinding CSELIB_VAL_PTR changes.
4041 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4044 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4046 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4048 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4049 constant definitions.
4050 ("nop"): lr 0,0 -> nopr r0
4051 ("nop_lr0", "nop_lr1"): New insn definitions.
4053 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4055 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4056 NDS32_V3PUSH_AVAILABLE_P macro.
4058 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4059 Chung-Ju Wu <jasonwucj@gmail.com>
4061 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4062 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4063 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4064 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4065 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4066 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4067 * config/nds32/constraints.md: New constraints and checking for hard
4068 float configuration.
4069 * config/nds32/iterators.md: New mode iterator and attribute for hard
4070 float configuration.
4071 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4073 * config/nds32/nds32-fpu.md: New file.
4074 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4075 deal with hard float code generation.
4076 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4078 (abi_type, float_reg_number): New enum type.
4079 * config/nds32/nds32-predicates.c: New predicates for hard float.
4080 * config/nds32/nds32-protos.h: Declare functions for hard float.
4081 * config/nds32/nds32.c: Implementation for hard float configuration.
4082 * config/nds32/nds32.h: Definitions for hard float configuration.
4083 * config/nds32/nds32.md: Include hard float machine description and
4084 modify patterns for hard float configuration.
4085 * config/nds32/nds32.opt: New options for hard float configuration.
4086 * config/nds32/predicates.md: New predicates for hard float
4089 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4091 * common/config/nds32/nds32-common.c
4092 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4094 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4097 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4098 CONSTRUCTOR_ELT (ctor, ...)->value.
4100 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4103 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4105 2018-04-05 Tom de Vries <tom@codesourcery.com>
4108 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4111 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4112 Kito Cheng <kito.cheng@gmail.com>
4114 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4115 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4116 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4117 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4119 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4120 Kito Cheng <kito.cheng@gmail.com>
4122 * config/nds32/constraints.md (Ufe): New memory constraint.
4123 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4124 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4125 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4127 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4128 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4130 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4132 * config/nds32/nds32.md: Use optimize_size in the condition for
4133 alu-shift instructions.
4135 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4137 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4139 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4141 * config/nds32/nds32.md (negsi2): Refine pattern.
4143 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4144 Chung-Ju Wu <jasonwucj@gmail.com>
4146 * config/nds32/iterators.md (shift_rotate): New code iterator.
4147 (shift): New code attribute.
4148 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4149 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4150 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4151 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4152 bit-wise operations.
4153 (andsi3, *andsi3): Ditto.
4154 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4155 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4156 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4157 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4158 nds32_ior_operand, nds32_xor_operand): New predicates.
4160 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4162 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4163 (addsi3, subsi3): ... this.
4165 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4167 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4169 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4171 * config/nds32/nds32.md: Adjust indention.
4173 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4175 * config/nds32/nds32.md (feature): New attribute.
4177 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4179 * config/nds32/nds32.md (subtype): New attribute.
4181 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4184 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4185 expansion to perform a bitwise AND of the argument followed by a
4186 boolean negation of the result.
4188 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4190 PR rtl-optimization/84878
4191 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4192 the basic block. Assert the use reference is not artificial and that
4193 it has an associated insn.
4195 2018-04-04 Michael Matz <matz@suse.de>
4197 * builtins.c (compute_objsize): Pass correct operand
4198 to array_at_struct_end_p.
4200 2018-04-04 Richard Biener <rguenther@suse.de>
4203 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4204 from contexts for DINFO_LEVEL_TERSE and below.
4206 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4208 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4209 resiter_operand condition.
4210 * config/nds32/nds32.md (*move<mode>): Ditto.
4212 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4213 Monk Chiang <sh.chiang04@gmail.com>
4215 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4217 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4219 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4221 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4222 Kito Cheng <kito.cheng@gmail.com>
4224 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4225 nds32_cond_code_str, output_cond_branch,
4226 output_cond_branch_compare_zero, nds32_expand_cbranch,
4227 nds32_expand_cstore, nds32_expand_movcc,
4228 nds32_output_cbranchsi4_equality_zero,
4229 nds32_output_cbranchsi4_equality_reg,
4230 nds32_output_cbranchsi4_equality_reg_or_const_int,
4231 nds32_output_cbranchsi4_greater_less_zero: New functions.
4232 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4233 nds32_expand_cstore, nds32_expand_movcc,
4234 nds32_output_cbranchsi4_equality_zero,
4235 nds32_output_cbranchsi4_equality_reg,
4236 nds32_output_cbranchsi4_equality_reg_or_const_int,
4237 nds32_output_cbranchsi4_greater_less_zero): Declare.
4238 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4239 nds32_rimm11s_operand): New predicates.
4240 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4241 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4244 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4246 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4247 * config/nds32/nds32.md: Ditto.
4248 * config/nds32/pipelines.md: Ditto.
4250 2018-04-04 Richard Biener <rguenther@suse.de>
4252 PR tree-optimization/85168
4253 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4254 propagating abnormals.
4256 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4258 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4260 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4261 Kito Cheng <kito.cheng@gmail.com>
4263 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4264 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4265 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4266 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4267 * config/nds32/nds32.md (sibcall_internal): New.
4268 (sibcall_register): Remove.
4269 (sibcall_immediate): Remove.
4270 (sibcall_value_internal): New.
4271 (sibcall_value_register): Remove.
4272 (sibcall_value_immediate): Remove.
4273 * config/nds32/predicates.md (nds32_general_register_operand): New.
4274 (nds32_call_address_operand): New.
4276 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4278 PR rtl-optimization/85167
4279 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4280 bb_defs if *split_p, instead preinitialize it to NULL.
4282 PR tree-optimization/85156
4283 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4284 evaluating the argument multiple times.
4286 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4288 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4290 (_mm_cvtpd_ps): Likewise.
4291 (_mm_cvttpd_epi32): Likewise.
4292 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4293 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4294 vector, pixel, and bool following altivec.h include.
4296 2018-04-03 Martin Sebor <msebor@redhat.com>
4298 * doc/extend.texi (Common Function Attributes): Clarify.
4299 (const attribute): Likewise.
4300 (pure attribute): Likewise.
4302 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4305 * config/i386/i386.c (ix86_expand_vector_set): Use
4306 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4308 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4310 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4311 instructions when changing rounding bits to preserve precision bits
4312 in the x87 control word.
4314 2018-04-03 Martin Liska <mliska@suse.cz>
4316 PR tree-optimization/82491
4317 * rtl.h (strip_offset_and_add): Replace += suboffset with
4318 poly_uint64 () + suboffset.
4320 2018-03-29 Martin Liska <mliska@suse.cz>
4321 Martin Jambor <mjambor@suse.cz>
4324 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4325 param_type is not an integral or pointer type.
4327 2018-04-03 Richard Biener <rguenther@suse.de>
4329 * sese.h (recompute_all_dominators): Remove.
4331 2018-04-02 Martin Sebor <msebor@redhat.com>
4333 * doc/invoke.texi (-Wrestrict): Fix typos.
4335 2018-04-02 Jim Wilson <jimw@sifive.com>
4337 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4338 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4339 (<optab>di3, <optab>si3_extend): Likewise.
4340 (<optab>si3_mask, <optab>si3_mask_1): New.
4341 (<optab>di3_mask, <optab>di3_mask_1): New.
4342 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4343 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4344 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4346 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4348 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4351 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4353 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4354 (nds32_canonicalize_comparison): New function.
4356 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4357 Kito Cheng <kito.cheng@gmail.com>
4358 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4360 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4361 * config/nds32/constants.md (unspec_volatile_element): Add
4362 UNSPEC_VOLATILE_RELAX_GROUP.
4363 * config/nds32/nds32-relax-opt.c: New file.
4364 * config/nds32/nds32-predicates.c
4365 (nds32_symbol_load_store_p): New function.
4366 * config/nds32/nds32-protos.h
4367 (nds32_symbol_load_store_p): Declare function.
4368 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4369 * config/nds32/nds32.c
4370 (nds32_register_pass): New function to register pass.
4371 (nds32_register_passes): New function to register passes.
4372 * config/nds32/nds32.md (relax_group): New pattern.
4373 * config/nds32/nds32.opt (mrelax-hint): New option.
4374 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4376 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4378 * config/nds32/t-nds32: Modify files dependency.
4380 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4382 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4383 (PROFILE_HOOK): Define its implementation.
4385 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4387 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4388 type and 32-bit size.
4390 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4393 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4394 (V_128_256): New mode iterator.
4395 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4396 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4397 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4399 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4400 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4402 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4405 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4406 NaN inputs correctly.
4408 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4411 * config/rs6000/vsx.md (??r): New mode attribute.
4412 (*vsx_mov<mode>_64bit): Use it.
4413 (*vsx_mov<mode>_32bit): Likewise.
4415 2018-03-30 Martin Sebor <msebor@redhat.com>
4417 PR tree-optimization/84818
4418 * builtins.c (check_access): Use warning_n.
4420 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4423 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4425 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4428 2018-03-30 Julia Koval <julia.koval@intel.com>
4431 * x86-tune.def (movx, partial_reg_dependency): Enable for
4434 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4437 * lra-constraints.c (process_alt_operands): Move setting
4438 this_alternative_matches below.
4440 2018-03-29 Martin Liska <mliska@suse.cz>
4443 * doc/invoke.texi: Document how LTO works with debug info.
4444 Describe auto-load support of binutils. Mention 'x86-64'
4445 as valid option value of -march option.
4447 2018-03-29 Jakub Jelinek <jakub@redhat.com>
4449 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
4452 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
4453 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
4454 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
4457 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4460 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
4461 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
4462 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
4463 for RS6000_BTM_POWERPC64.
4464 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
4465 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
4466 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
4471 2018-03-28 Carl Love <cel@us.ibm.com>
4474 2017-09-27 Carl Love <cel@us.ibm.com>
4476 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
4477 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
4478 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
4481 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4483 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
4484 instead of __vector bool.
4485 (_mm_max_pu8): Likewise.
4486 (_mm_min_pi16): Likewise.
4488 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
4491 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
4492 (DIVWEUO): Likewise.
4494 (DIVDEUO): Likewise.
4495 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
4496 DIVWEUO and DIVDEUO.
4497 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
4498 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
4499 (div_extend): Likewise.
4500 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
4502 (__builtin_divweuo): Likewise.
4503 (__builtin_divdeo): Likewise.
4504 (__builtin_divdeuo): Likewise.
4506 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4509 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
4510 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
4512 PR tree-optimization/82004
4513 * gimple-match-head.c (optimize_pow_to_exp): New function.
4514 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
4515 Don't fold to exp if optimize_pow_to_exp is false.
4517 2018-03-28 Martin Liska <mliska@suse.cz>
4520 * calls.c (initialize_argument_information): Fix trailing space.
4521 * common.opt: Fix typo and provide better explanation for
4522 -fsanitize-coverage option.
4523 * config/i386/i386.opt: Fix typo.
4525 2018-03-28 Jakub Jelinek <jakub@redhat.com>
4526 Martin Liska <mliska@suse.cz>
4529 * gimplify.c (asan_poison_variable): Don't do the check for
4530 gimplify_omp_ctxp here.
4531 (gimplify_decl_expr): Do it here.
4532 (gimplify_target_expr): Likewise.
4534 2018-03-28 Martin Liska <mliska@suse.cz>
4537 * config/i386/i386.c (ix86_function_arg_advance): Do not call
4538 chkp_type_bounds_count if MPX is not enabled.
4540 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
4542 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
4544 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
4547 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
4548 function to create the function decl for complex long double
4549 multiply and divide for -mabi=ieeelongdouble.
4550 (init_float128_ieee): Call it.
4552 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
4555 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
4556 -fcf-protection=branch -mibt.
4557 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
4559 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4562 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
4564 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
4567 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
4568 extern array declarations.
4570 2018-03-27 Richard Biener <rguenther@suse.de>
4573 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
4574 explicit single_use checks.
4576 2018-03-27 Richard Biener <rguenther@suse.de>
4578 PR tree-optimization/85082
4579 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
4582 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4584 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
4585 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
4586 Turn on fasynchronous-unwind-tables and funwind-tables.
4588 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
4591 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
4592 (*bmi_blsr_<mode>_ccz): Ditto.
4594 2018-03-26 Tom de Vries <tom@codesourcery.com>
4596 PR tree-optimization/85063
4597 * omp-general.c (offloading_function_p): New function. Factor out
4599 * omp-offload.c (pass_omp_target_link::gate): ... here.
4600 * omp-general.h (offloading_function_p): Declare.
4601 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
4602 with attribute omp declare target for offloading functions.
4604 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
4606 PR tree-optimization/84005
4607 * tree-data-ref.h (get_base_for_alignment): Declare.
4608 * tree-data-ref.c (get_base_for_alignment_1): New function.
4609 (get_base_for_alignment): Likewise.
4610 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
4611 get_base_for_alignment to find a suitable base object, instead
4612 of always using drb->base_address.
4614 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4617 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
4618 known size by default.
4620 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
4623 * lra-constraints.c (process_alt_operands): Don't match BLKmode
4624 and non BLKmode operands.
4626 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4629 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4630 Clean up attributes.
4632 2018-03-23 Richard Biener <rguenther@suse.de>
4635 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4636 we are going to emit early debug for LTO.
4638 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4641 * function.c (match_asm_constraints_1): Don't optimize if input
4642 doesn't satisfy general_operand predicate for output's mode.
4645 * alias.c (write_dependence_p): Don't require for x_canonicalized
4646 non-VOIDmode if x has VOIDmode.
4649 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4650 just don't try to optimize it rather than assert it never happens.
4652 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4654 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4655 macro expansions for definition of ST_INTERNAL_<mode> and
4656 LD_INTERNAL_<mode> builtins.
4657 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4659 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4661 (altivec_expand_st_builtin): Likewise.
4662 (altivec_expand_builtin): Remove calls to deleted functions.
4663 (rs6000_address_for_altivec): Delete this function.
4664 * config/rs6000/vector.md: Remove expands for
4665 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4667 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4670 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4671 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4672 re-computing once computed.
4673 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4674 (arm_init_machine_status): Initialize
4675 machine->static_chain_stack_bytes.
4677 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4680 * doc/extend.texi: Add four new prototypes for vec_ld.
4681 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4682 definitions for more logical presentation.
4683 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4684 entries for V1TI variants of __builtin_altivec_ld builtin.
4685 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4686 handling of V1TI variant of LVX icode pattern.
4687 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4688 (rs6000_gimple_fold_builtin): Likewise.
4689 (altivec_init_builtins): Add code to define
4690 __builtin_altivec_lvx_v1ti function.
4692 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4695 * function.c (match_asm_constraints_1): Don't do the optimization
4696 if input isn't a REG, SUBREG, MEM or constant.
4698 2018-03-22 Tom de Vries <tom@codesourcery.com>
4700 PR tree-optimization/84956
4701 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4702 bb_has_abnormal_pred.
4704 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4707 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4708 DECL_INITIAL (decl) to decl at the end.
4709 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4712 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4714 * doc/extend.texi (__builtin_tgmath): Document when complex
4715 integer types are treated as _Complex _Float64.
4717 2018-03-21 Tom de Vries <tom@codesourcery.com>
4719 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4721 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4723 PR tree-optimization/84960
4724 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4725 if it is ENTRY block, move them into single succ of ENTRY in that case.
4727 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4729 PR tree-optimization/84811
4730 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4731 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4732 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4733 (known_subrange_p): Update accordingly. Cast each value involved
4734 in the size comparison, rather than casting the result of the
4737 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4739 PR tree-optimization/84982
4740 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4741 by flipping the least significant bit rather than all bits from
4742 bitpos to bitpos + bitsize - 1.
4744 2018-03-21 Nathan Sidwell <nathan@acm.org>
4746 * doc/extend.texi (Deprecated Features): Remove mention of
4747 long-deleted deprecations.
4749 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4752 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4753 * configure: Regenerate.
4755 2018-03-21 Tom de Vries <tom@codesourcery.com>
4757 PR tree-optimization/83126
4758 * tree-parloops.c (num_phis): New function.
4759 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4761 2018-03-21 Nathan Sidwell <nathan@acm.org>
4763 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4764 mention anon-struct/union members and trailing attributes.
4766 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4768 PR tree-optimization/84969
4769 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4770 builtin memset partitions if they set different rhs values.
4772 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4774 PR rtl-optimization/84989
4775 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4776 VEC_DUPLICATE with scalar result mode.
4778 2018-03-21 Martin Liska <mliska@suse.cz>
4781 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4782 not intended return statement.
4784 2018-03-21 Martin Liska <mliska@suse.cz>
4787 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4788 (chkp_find_bound_slots_1): Limit number of iterations.
4790 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4793 * Minor grammar fixes for x86 options.
4795 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4798 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4799 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4801 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4804 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4805 (altivec_lvx_<mode>_1op): Likewise.
4806 (altivec_stvx_<mode>_2op): Likewise.
4807 (altivec_stvx_<mode>_1op): Likewise.
4808 (altivec_lvx_<VM2:mode>): New define_expand.
4809 (altivec_stvx_<VM2:mode>): Likewise.
4810 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4811 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4812 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4813 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4814 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4815 (rs6000_gen_lvx): Likewise.
4816 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4817 (altivec_expand_stv_builtin): Likewise.
4818 (altivec_expand_builtin): Likewise.
4819 * config/rs6000/vector.md: Likewise.
4821 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4824 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4827 2018-03-20 Richard Biener <rguenther@suse.de>
4830 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4831 sign-conversions as zero, fall back to standard scalar_stmt
4834 2018-03-20 Martin Liska <mliska@suse.cz>
4837 * predict.c (rebuild_frequencies): Handle case when we have
4838 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4840 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4843 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4844 flag_section_anchors.
4845 * varasm.c (use_blocks_for_decl_p): Remove hack for
4846 dw2_force_const_mem.
4849 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4851 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4852 be created, use lowpart_subreg of operands[0] rather than operands[0]
4854 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4855 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4856 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4857 and n constraint instead of aarch64_shift_imm_di and Usd.
4858 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4859 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4861 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4864 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4865 to favor GPR over NEON registers.
4866 (<shift>di3_neon): Likewise.
4868 2018-03-20 Tom de Vries <tom@codesourcery.com>
4871 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4872 (nvptx_process_pars): Emit bar.sync asap and alap.
4874 2018-03-20 Tom de Vries <tom@codesourcery.com>
4877 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4878 seen_label if seen_label is already set.
4880 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4883 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4884 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4885 Use 1U instead of 1. Formatting fixes.
4888 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4889 instead of TREE_TYPE (s1) for the return value.
4891 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4893 PR tree-optimization/84946
4894 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4895 bitsize + bitsize in poly_uint64 rather than poly_int64.
4898 * dwarf2asm.c: Include fold-const.c.
4899 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4900 of decl rather than decl itself.
4902 PR rtl-optimization/84643
4903 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4905 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4908 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4909 calling assemble_variable.
4911 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4914 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4915 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4917 2018-03-19 Jim Wilson <jimw@sifive.com>
4920 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4921 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4922 (riscv_first_stack_step): Likewise.
4923 (riscv_option_override): Use STACK_BOUNDARY instead of
4925 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4927 (BIGGEST_ALIGNMENT): Set to 128.
4928 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4929 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4932 2018-03-19 Richard Biener <rguenther@suse.de>
4934 PR tree-optimization/84933
4935 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4936 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4938 2018-03-19 Richard Biener <rguenther@suse.de>
4940 PR tree-optimization/84859
4941 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4942 (cond_if_else_store_replacement): Perform sinking operation on
4943 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4944 Generalize what a BB with a single eligible store is.
4946 2018-03-19 Richard Biener <rguenther@suse.de>
4948 PR tree-optimization/84929
4949 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4950 chrec_is_positive against non-chrec arg.
4952 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4955 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4957 2018-03-18 Martin Liska <mliska@suse.cz>
4959 PR rtl-optimization/84635
4960 * regrename.c (build_def_use): Use matches_mode only when
4963 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4965 PR tree-optimization/84913
4966 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4967 vectorize chains of COND_EXPRs.
4969 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4971 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4973 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4975 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4977 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4979 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4981 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4982 Kito Cheng <kito.cheng@gmail.com>
4984 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4985 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4986 (nds32_adjust_reg_alloc_order): New function.
4987 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4989 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4991 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4992 nds32_print_operand, nds32_print_operand_address): Use
4993 HOST_WIDE_INT_PRINT_DEC instead.
4995 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4997 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4999 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5002 * config/i386/i386.c (initial_ix86_tune_features,
5003 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5005 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5006 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5007 rather than 1u << ix86_tune. Formatting fix.
5008 (ix86_option_override_internal): Change ix86_arch_mask from
5009 unsigned int to unsigned HOST_WIDE_INT, initialize to
5010 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5011 (ix86_function_specific_restore): Likewise.
5013 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5016 * postreload.c (reload_combine_recognize_pattern): Perform
5017 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5018 truncate_int_for_mode the result for the destination's mode.
5021 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5022 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5024 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5027 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5030 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5031 regno_allocno_class_array and sorted_pseudos.
5032 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5033 insns where regno is used.
5035 2018-03-16 Martin Liska <mliska@suse.cz>
5038 * multiple_target.c (create_dispatcher_calls): Redirect
5039 reference in the symbol table.
5041 2018-03-16 Martin Liska <mliska@suse.cz>
5044 * multiple_target.c (create_dispatcher_calls): Redirect also
5047 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5051 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5052 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5053 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5055 2018-03-16 Julia Koval <julia.koval@intel.com>
5057 * doc/invoke.texi (Skylake Server): Add CLWB.
5058 Cannonlake): Remove CLWB.
5060 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5062 PR tree-optimization/84841
5063 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5065 (FLOAT_ONE_CONST_TYPE): Define.
5066 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5067 (sort_by_operand_rank): Put entries with higher constant_type last
5068 rather than first to match comments.
5070 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5072 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5075 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5078 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5079 check and use address of target always.
5081 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5084 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5085 (indirect_thunk_bnd_needed): Likewise.
5086 (indirect_thunks_used): Likewise.
5087 (indirect_thunks_bnd_used): Likewise.
5088 (indirect_return_needed): New.
5089 (indirect_return_bnd_needed): Likewise.
5090 (output_indirect_thunk_function): Add a bool argument for
5092 (output_indirect_thunk_function): Don't generate alias for
5093 function return thunk.
5094 (ix86_code_end): Call output_indirect_thunk_function to generate
5095 function return thunks.
5096 (ix86_output_function_return): Set indirect_return_bnd_needed
5097 and indirect_return_needed instead of indirect_thunk_bnd_needed
5098 and indirect_thunk_needed.
5100 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5102 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5103 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5104 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5106 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5107 Paul Hua <paul.hua.gm@gmail.com>
5110 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5112 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5114 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5115 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5118 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5121 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5122 instead of GET_MODE_SIZE when comparing Units.
5124 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5127 * varasm.c (hash_section): Return an unchangeble hash value
5128 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5129 Return !aarch64_can_use_per_function_literal_pools_p ().
5131 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5134 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5135 rather than address of cmode as last argument to prepare_cmp_insn.
5137 2018-03-15 Julia Koval <julia.koval@intel.com>
5139 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5140 F_AVX512VNNI, F_AVX512BITALG): New.
5142 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5145 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5146 insn for floating-point loads and stores.
5148 2018-03-14 Carl Love <cel@us.ibm.com>
5150 * config/rs6000/rs6000-c.c: Add macro definitions for
5151 ALTIVEC_BUILTIN_VEC_PERMXOR.
5152 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5153 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5154 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5155 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5157 * config/doc/extend.texi: Add prototypes for vec_permxor.
5159 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5162 * diagnostic-show-locus.c (class layout_point): Convert m_line
5163 from int to linenum_type.
5164 (line_span::comparator): Use linenum "compare" function when
5165 comparing line numbers.
5166 (test_line_span): New function.
5167 (layout_range::contains_point): Convert param "row" from int to
5169 (layout_range::intersects_line_p): Likewise.
5170 (layout::will_show_line_p): Likewise.
5171 (layout::print_source_line): Likewise.
5172 (layout::should_print_annotation_line_p): Likewise.
5173 (layout::print_annotation_line): Likewise.
5174 (layout::print_leading_fixits): Likewise.
5175 (layout::annotation_line_showed_range_p): Likewise.
5176 (struct line_corrections): Likewise for field m_row.
5177 (line_corrections::line_corrections): Likewise for param "row".
5178 (layout::print_trailing_fixits): Likewise.
5179 (layout::get_state_at_point): Likewise.
5180 (layout::get_x_bound_for_row): Likewise.
5181 (layout::print_line): Likewise.
5182 (diagnostic_show_locus): Likewise for locals "last_line" and
5184 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5185 * input.c (selftest::test_linenum_comparisons): New function.
5186 (selftest::input_c_tests): Call it.
5187 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5188 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5189 * selftest.h (ASSERT_GT): New macro.
5190 (ASSERT_GT_AT): New macro.
5191 (ASSERT_LT): New macro.
5192 (ASSERT_LT_AT): New macro.
5194 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5196 PR rtl-optimization/84780
5197 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5199 2018-03-14 Martin Liska <mliska@suse.cz>
5201 * tree.c (record_node_allocation_statistics): Use
5202 get_stats_node_kind.
5203 (get_stats_node_kind): New function extracted from
5204 record_node_allocation_statistics.
5205 (free_node): Use get_stats_node_kind.
5207 2018-03-14 Richard Biener <rguenther@suse.de>
5209 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5210 that the value-set of ANTIC_IN doesn't grow.
5213 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5215 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5216 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5218 2018-03-14 Julia Koval <julia.koval@intel.com>
5220 * config.gcc (icelake-client, icelake-server): New.
5222 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5223 (initial_ix86_arch_features): Ditto.
5224 (PTA_SKYLAKE): Add SGX.
5225 (PTA_ICELAKE): Remove.
5226 (PTA_ICELAKE_CLIENT): New.
5227 (PTA_ICELAKE_SERVER): New.
5228 (ix86_option_override_internal): Split up icelake on icelake client and
5230 (get_builtin_code_for_version): Ditto.
5231 (fold_builtin_cpu): Ditto.
5232 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5233 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5234 * config/i386/i386.h (processor_type): Ditto.
5235 * doc/invoke.texi: Ditto.
5237 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5240 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5241 INTEGER_CST offset, add it together with bitpos / 8 and
5242 sign extend based on POINTER_SIZE.
5246 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5249 * config/i386/constraints.md (Yc): New register constraint.
5250 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5251 Use Yc constraint for alternative 2 of operand 0. Remove
5252 preferred_for_speed attribute.
5254 2018-03-14 Richard Biener <rguenther@suse.de>
5256 PR tree-optimization/84830
5257 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5258 with the old one to avoid oscillations.
5260 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5263 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5265 (assign_by_spills): Return a flag of reload assignment failure.
5266 Do not process the reload assignment failures. Do not spill other
5267 reload pseudos if they has the same reg class. Update n if
5269 (lra_assign): Add a return arg. Set up from the result of
5270 assign_by_spills call.
5271 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5272 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5273 usage_insns if it is not NULL.
5274 (spill_hard_reg_in_range): New function.
5275 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5276 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5277 function prototypes.
5278 (lra_assign): Change prototype.
5279 * lra.c (lra): Add code to deal with fails by splitting hard reg
5282 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5284 * config/riscv/riscv.opt (mrelax): New option.
5285 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5286 "norelax" when riscv_mrelax is disabled.
5287 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5289 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5292 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5293 reassociation for int modes.
5295 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5297 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5298 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5300 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5301 * config/aarch64/aarch64-sve.md
5302 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5303 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5304 (*extend<mode><Vwide>2): Rename to...
5305 (aarch64_sve_extend<mode><Vwide>2): ...this.
5306 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5307 renaming the old pattern to...
5308 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5310 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5311 define_expand, renaming the old pattern to...
5312 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5313 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5314 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5315 account when deciding which SVE instruction the optab should use.
5316 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5318 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5320 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5321 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5322 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5323 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5324 (tlsdesc_small_<mode>): Turn a define_expand and use
5325 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5326 (tlsdesc_small_advsimd_<mode>): ...this.
5327 (tlsdesc_small_sve_<mode>): New pattern.
5329 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5331 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5332 (UNSPEC_UMUL_HIGHPART): New constants.
5333 (MUL_HIGHPART): New int iteraor.
5334 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5335 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5337 (*<su>mul<mode>3_highpart): New define_insn.
5339 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5342 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5345 2018-03-13 Martin Liska <mliska@suse.cz>
5348 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5350 (sem_item_optimizer::~sem_item_optimizer): Release it.
5351 (sem_item_optimizer::merge_classes): Register variable aliases.
5352 (sem_item_optimizer::fixup_pt_set): New function.
5353 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5354 * ipa-icf.h: Declare new variables and functions.
5356 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5359 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5360 integer_pow2p@2 and test integer_pow2p in condition.
5361 (A < 0 ? C : 0): Similarly for @1.
5364 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5365 characters starting at p contain '\0' character, don't look beyond
5369 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5370 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5373 * reg-stack.c (change_stack): Change update_end var from int to
5374 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5375 also call set_block_for_insn on the newly added insns and rescan.
5378 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5379 on the last operand.
5382 * tree.c (stabilize_reference_1): Return save_expr (e) for
5383 STATEMENT_LIST even if it doesn't have side-effects.
5385 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5387 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5389 2018-03-12 Renlin Li <renlin.li@arm.com>
5391 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5392 aarch64_output_scalar_simd_mov_immediate.
5394 2018-03-12 Martin Sebor <msebor@redhat.com>
5396 PR tree-optimization/83456
5397 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5398 for perfectly overlapping calls to memcpy.
5399 (gimple_fold_builtin_memory_chk): Same.
5400 (gimple_fold_builtin_strcpy): Handle no-warning.
5401 (gimple_fold_builtin_stxcpy_chk): Same.
5402 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5404 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5406 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5407 parameter. Use it for SFmode.
5408 (rs6000_function_arg_advance_1): Adjust.
5409 (rs6000_function_arg): Adjust.
5410 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5412 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5414 PR rtl-optimization/84169
5415 PR rtl-optimization/84780
5416 * combine.c (can_combine_p): Check for a 2-insn combination whether
5417 the destination register is used between the two insns, too.
5419 2018-03-12 Richard Biener <rguenther@suse.de>
5421 PR tree-optimization/84803
5422 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5423 for refs DR analysis didn't process.
5425 2018-03-12 Richard Biener <rguenther@suse.de>
5427 PR tree-optimization/84777
5428 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5429 force-vectorize loops ignore whether we are optimizing for size.
5431 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5433 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5434 (TARGET_MD_ASM_ADJUST): Define.
5436 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5437 Kito Cheng <kito.cheng@gmail.com>
5438 Chung-Ju Wu <jasonwucj@gmail.com>
5440 * config/nds32/nds32.c (nds32_compute_stack_frame,
5441 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
5442 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
5443 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
5444 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
5445 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
5446 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
5447 * config/nds32/nds32.md (prologue, epilogue): Use macro
5448 NDS32_V3PUSH_AVAILABLE_P to do checking.
5450 2018-03-11 Jakub Jelinek <jakub@redhat.com>
5453 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
5454 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
5455 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
5456 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
5457 addition of most attributes on !orig_type_die or the attribute not
5458 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
5460 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5461 Chung-Ju Wu <jasonwucj@gmail.com>
5463 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
5465 * config/nds32/nds32.opt (mvh): New option.
5467 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5468 Chung-Ju Wu <jasonwucj@gmail.com>
5470 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
5472 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
5473 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
5476 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
5477 Chung-Ju Wu <jasonwucj@gmail.com>
5479 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
5481 * config/nds32/nds32-multiple.md (strlensi): New pattern.
5482 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
5484 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
5485 Kito Cheng <kito.cheng@gmail.com>
5486 Chung-Ju Wu <jasonwucj@gmail.com>
5488 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
5489 UNSPEC_FFMISM and UNSPEC_FLMISM.
5490 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
5491 for ffb, ffmism and flmism.
5492 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
5493 (unspec_ffmism): Ditto.
5494 (unspec_flmism): Ditto.
5495 (nds32_expand_builtin_impl): Check if string extension is available.
5496 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
5497 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
5499 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
5502 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5505 * lra-assigns.c (assign_by_spills): Return a flag of reload
5506 assignment failure. Do not process the reload assignment
5507 failures. Do not spill other reload pseudos if they has the same
5509 (lra_assign): Add a return arg. Set up from the result of
5510 assign_by_spills call.
5511 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5512 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5513 usage_insns if it is not NULL.
5514 (spill_hard_reg_in_range): New function.
5515 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5516 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5517 function prototypes.
5518 (lra_assign): Change prototype.
5519 * lra.c (lra): Add code to deal with fails by splitting hard reg
5522 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
5525 * config/i386/i386.opt: Replace Enforcment with Enforcement.
5527 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
5530 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
5531 (dw_val_node): Add val_symbolic_view.
5532 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
5533 (symview_upper_bound): New.
5534 (new_line_info_table): Initialize symviews_since_reset.
5535 (dwarf2out_source_line): Count symviews_since_reset and set
5536 symview_upper_bound.
5537 (dw_val_equal_p): Handle symview.
5538 (add_AT_symview): New.
5539 (print_dw_val): Handle symview.
5540 (attr_checksum, attr_checksum_ordered): Likewise.
5541 (same_dw_val_p, size_of_die): Likewise.
5542 (value_format, output_die): Likewise.
5543 (add_high_low_attributes): Use add_AT_symview for entry_view.
5544 (dwarf2out_finish): Reset symview_upper_bound, clear
5547 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
5550 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
5551 Add strict argument and use it.
5552 (rs6000_split_multireg_move): Update for new strict argument.
5553 (mem_operand_gpr): Disallow all non-offsettable addresses.
5554 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
5556 2018-03-09 Jakub Jelinek <jakub@redhat.com>
5559 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
5560 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
5561 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
5564 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
5565 decl, use remap_type if we want to use the type.
5567 2018-03-09 Martin Sebor <msebor@redhat.com>
5569 PR tree-optimization/84526
5570 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5572 (builtin_access::generic_overlap): Be prepared to handle non-array
5575 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
5577 PR rtl-optimization/84682
5578 * lra-constraints.c (process_address_1): Check is_address flag
5579 for address constraints.
5580 (process_alt_operands): Likewise.
5581 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
5582 preprocess_constraints.
5583 * recog.h (preprocess_constraints): Add oploc parameter.
5585 * recog.c (preprocess_constraints): Test address_operand for
5586 CT_ADDRESS constraints.
5588 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
5591 * lra-assigns.c (assign_by_spills): Return a flag of reload
5592 assignment failure. Do not process the reload assignment
5593 failures. Do not spill other reload pseudos if they has the same
5595 (lra_assign): Add a return arg. Set up from the result of
5596 assign_by_spills call.
5597 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5598 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5599 usage_insns if it is not NULL.
5600 (spill_hard_reg_in_range): New function.
5601 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5602 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5603 function prototypes.
5604 (lra_assign): Change prototype.
5605 * lra.c (lra): Add code to deal with fails by splitting hard reg
5608 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5611 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
5612 Accept complain bool parameter. Only emit errors if it is true.
5613 (arm_parse_cpu_option_name): Likewise.
5614 (arm_target_thumb_only): Adjust callers of the above.
5615 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
5616 prototype to take a default true bool parameter.
5617 (arm_parse_arch_option_name): Likewise.
5619 2018-03-09 David Malcolm <dmalcolm@redhat.com>
5620 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
5624 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5625 * configure: Regenerate.
5626 * configure.ac ("linker --version-script option"): New.
5627 ("linker soname option"): New.
5629 2018-03-09 Richard Biener <rguenther@suse.de>
5631 PR tree-optimization/84775
5632 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5633 immediate uses of predicate stmts and mark them modified.
5636 PR tree-optimization/84178
5637 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5639 (version_loop_for_if_conversion): Delay update_ssa call.
5640 (tree_if_conversion): Delay update_ssa until after predicate
5643 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5646 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5647 when the function accesses prior frames.
5649 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5652 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5653 gen_llsym, otherwise call maybe_gen_llsym.
5656 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5657 has ',' character inside of it.
5659 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5662 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5663 as clobbering CC_REGNUM.
5665 2018-03-08 Richard Biener <rguenther@suse.de>
5668 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5669 (follow_copies_to_constant): Do not follow SSA names registered
5672 2018-03-08 Richard Biener <rguenther@suse.de>
5674 PR tree-optimization/84178
5675 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5677 (version_loop_for_if_conversion): Delay update_ssa call.
5678 (tree_if_conversion): Delay update_ssa until after predicate
5681 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5683 PR tree-optimization/84178
5684 * tree-if-conv.c (release_bb_predicate): Remove the
5685 the assertion that the stmts have NULL use_ops.
5686 Discard the statements, asserting that they haven't
5687 yet been added to a BB.
5689 2018-03-08 Richard Biener <rguenther@suse.de>
5691 PR tree-optimization/84746
5692 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5693 (phi_translate): Pass in destination ANTIC_OUT set.
5694 (phi_translate_1): Likewise. For a simplified result lookup
5695 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5696 (phi_translate_set): Adjust.
5697 (do_pre_regular_insertion): Likewise.
5698 (do_pre_partial_partial_insertion): Likewise.
5700 2018-03-08 Martin Liska <mliska@suse.cz>
5702 PR gcov-profile/84735
5703 * doc/gcov.texi: Document usage of profile files.
5704 * gcov-io.h: Document changes in the format.
5706 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5710 * dwarf2out.c (struct dw_line_info_table): Update comments for
5712 (FORCE_RESET_NEXT_VIEW): New.
5713 (FORCE_RESETTING_VIEW_P): New.
5714 (RESETTING_VIEW_P): Check for -1 too.
5715 (ZERO_VIEW_P): Likewise.
5716 (new_line_info_table): Force-reset next view.
5717 (dwarf2out_begin_function): Likewise.
5718 (dwarf2out_source_line): Simplify zero_view_p initialization.
5719 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5720 view directly. Omit view when omitting .loc at line 0.
5722 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5724 PR tree-optimization/84740
5725 * tree-switch-conversion.c (process_switch): Call build_constructors
5726 only if info.phi_count is non-zero.
5728 PR tree-optimization/84739
5729 * tree-tailcall.c (find_tail_calls): Check call arguments against
5730 DECL_ARGUMENTS (current_function_decl) rather than
5731 DECL_ARGUMENTS (func) when checking for tail recursion.
5733 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5735 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5736 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5737 Volker Reichelt's entry and add entries for people that perform
5738 GCC fuzzy testing and report numerous bugs.
5740 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5743 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5744 readonly data in sdata, if that is disabled.
5745 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5746 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5747 -mreadonly-in-sdata option.
5749 2018-03-07 Martin Sebor <msebor@redhat.com>
5751 PR tree-optimization/84468
5752 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5753 basic block when looking for nul assignment.
5755 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5758 * except.h (output_function_exception_table): Adjust prototype.
5759 * except.c (output_function_exception_table): Remove FNNAME parameter
5760 and add SECTION parameter. Ouput one part of the table at a time.
5761 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5762 the first part of the exception table and emit unwind directives.
5763 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5764 (i386_pe_seh_cold_init): Likewise.
5765 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5766 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5767 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5768 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5769 * config/i386/winnt.c: Include except.h.
5770 (struct seh_frame_state): Add reg_offset, after_prologue and
5771 in_cold_section fields.
5772 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5773 (i386_pe_seh_cold_init): New function.
5774 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5775 to seh->in_cold_section.
5776 (seh_emit_push): Record the offset of the push.
5777 (seh_emit_save): Record the offet of the save.
5778 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5779 Test seh->after_prologue to disregard the epilogue.
5780 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5781 (i386_pe_end_cold_function): New function.
5783 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5786 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5787 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5790 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5792 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5793 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5796 * multiple_target.c: Include tree-inline.h and intl.h.
5797 (expand_target_clones): Diagnose and fail if node->definition and
5798 !tree_versionable_function_p (node->decl).
5800 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5802 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5804 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5805 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5806 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5808 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5811 * combine.c (try_combine): Use reg_or_subregno instead of handling
5812 just paradoxical SUBREGs and REGs.
5814 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5816 * config/arc/arc.c (arc_finalize_pic): Remove function.
5817 (arc_must_save_register): We use single base PIC register, remove
5818 checks to save/restore the PIC register.
5819 (arc_expand_prologue): Likewise.
5820 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5822 (arc_verify_short): Likewise.
5823 (arc_attr_type): Likewise.
5824 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5825 (walk_stores): Likewise.
5826 (arc_address_cost): Make it static.
5827 (arc_verify_short): Likewise.
5828 (branch_dest): Likewise.
5829 (arc_attr_type): Likewise.
5830 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5831 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5832 (arc_final_prescan_insn): Remove inserting the nops due to
5833 hardware hazards. It is done in reorg step.
5834 (insn_length_variant_t): Remove.
5835 (insn_length_parameters_t): Likewise.
5836 (arc_insn_length_parameters): Likewise.
5837 (arc_get_insn_variants): Likewise.
5838 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5840 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5843 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5846 PR tree-optimization/84687
5847 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5849 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5851 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5853 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5854 Rename to ppc_speculation_barrier.
5855 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5856 __builtin_ppc_speculation_barrier.
5858 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5861 * combine.c (combine_simplify_rtx): Don't try to simplify if
5862 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5865 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5867 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5868 to 32 bytes when compiling for POWER9.
5870 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5873 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5874 regparm >= 3 with no arg reg available also for calls with
5875 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5878 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5880 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5882 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5885 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5887 2018-03-05 Richard Biener <rguenther@suse.de>
5889 PR tree-optimization/84486
5890 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5891 When inserting a __builtin_assume_aligned call set the LHS
5892 SSA name alignment info accordingly.
5894 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5896 PR tree-optimization/84114
5897 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5898 Avoid reassociation of FLOAT_MODE addition.
5900 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5902 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5903 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5904 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5905 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5906 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5907 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5908 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5910 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5911 __builtin_ia32_wbinvd): New builtins.
5912 (SPECIAL_ARGS2): New.
5913 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5914 (SPECIAL_ARGS2): New.
5915 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5916 (ix86_valid_target_attribute_inner_p): Ditto.
5917 (ix86_init_mmx_sse_builtins): Add special_args2.
5918 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5919 TARGET_WBNOINVD_P): New.
5920 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5921 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5922 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5923 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5924 * config/i386/pconfigintrin.h: New file.
5925 * config/i386/wbnoinvdintrin.h: Ditto.
5926 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5927 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5929 2018-03-05 Richard Biener <rguenther@suse.de>
5931 PR tree-optimization/84670
5932 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5934 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5935 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5936 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5937 doesn't grow if all successors (recursively) were visited at least
5940 2018-03-05 Richard Biener <rguenther@suse.de>
5942 PR tree-optimization/84650
5943 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5944 if executed in the loop pipeline.
5946 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5948 * doc/configfiles.texi (Configuration Files): Move info about
5949 conditionalizing $target-protos.h to...
5950 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5951 differs from $target-protos.h.
5953 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5954 Chung-Ju Wu <jasonwucj@gmail.com>
5956 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5957 * config/nds32/nds32-multiple.md (setmemsi): Define.
5958 * config/nds32/nds32-memory-manipulation.c
5959 (nds32_gen_dup_4_byte_to_word_value): New.
5960 (emit_setmem_word_loop): New.
5961 (emit_setmem_byte_loop): New.
5962 (nds32_expand_setmem_loop): New.
5963 (nds32_expand_setmem_loop_v3m): New.
5964 (nds32_expand_setmem_unroll): New.
5965 (nds32_expand_setmem): New.
5967 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5968 Chung-Ju Wu <jasonwucj@gmail.com>
5970 * config/nds32/nds32-memory-manipulation.c
5971 (nds32_emit_load_store): New.
5972 (nds32_emit_post_inc_load_store): New.
5973 (nds32_emit_mem_move): New.
5974 (nds32_emit_mem_move_block): New.
5975 (nds32_expand_movmemsi_loop_unknown_size): New.
5976 (nds32_expand_movmemsi_loop_known_size): New.
5977 (nds32_expand_movmemsi_loop): New.
5978 (nds32_expand_movmemsi_unroll): New.
5979 (nds32_expand_movmemqi): Rename ...
5980 (nds32_expand_movmemsi): ... to this.
5981 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5982 (movmemsi): ... to this.
5983 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5984 (nds32_expand_movmemsi): ... to this.
5986 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5987 Monk Chiang <sh.chiang04@gmail.com>
5988 Chung-Ju Wu <jasonwucj@gmail.com>
5990 * config/nds32/nds32-protos.h
5991 (nds32_expand_load_multiple): New arguments.
5992 (nds32_expand_store_multiple): Ditto.
5993 (nds32_valid_multiple_load_store): Rename ...
5994 (nds32_valid_multiple_load_store_p): ... to this.
5995 * config/nds32/nds32-memory-manipulation.c
5996 (nds32_expand_load_multiple): Refine implementation.
5997 (nds32_expand_store_multiple): Ditto.
5998 * config/nds32/nds32-multiple.md
5999 (load_multiple): Update nds32_expand_load_multiple interface.
6000 (store_multiple): Update nds32_expand_store_multiple interface.
6001 * config/nds32/nds32-predicates.c
6002 (nds32_valid_multiple_load_store): Rename ...
6003 (nds32_valid_multiple_load_store_p): ... to this and refine
6005 * config/nds32/predicates.md
6006 (nds32_load_multiple_and_update_address_operation): New predicate.
6007 (nds32_store_multiple_and_update_address_operation): New predicate.
6009 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6010 Chung-Ju Wu <jasonwucj@gmail.com>
6012 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6013 (combo): New attribute.
6014 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6016 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6018 * config/nds32/nds32.opt: Change -mcmodel= default value.
6020 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6021 Monk Chiang <sh.chiang04@gmail.com>
6022 Chung-Ju Wu <jasonwucj@gmail.com>
6024 * config/nds32/constants.md (unspec_element): New enum.
6025 * config/nds32/constraints.md (Umw): New constraint.
6026 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6027 * config/nds32/nds32-intrinsic.md: Likewise.
6028 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6029 (nds32_valid_smw_lwm_base_p): New.
6030 (nds32_output_smw_single_word): New.
6031 (nds32_output_lmw_single_word): New.
6032 (nds32_expand_unaligned_load): New.
6033 (nds32_expand_unaligned_store): New.
6034 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6035 (nds32_output_smw_single_word): Declare.
6036 (nds32_output_lmw_single_word): Declare.
6037 (nds32_expand_unaligned_load): Declare.
6038 (nds32_expand_unaligned_store): Declare.
6039 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6040 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6041 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6042 NDS32_BUILTIN_UASTORE_DW.
6043 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6046 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6047 Kito Cheng <kito.cheng@gmail.com>
6048 Chung-Ju Wu <jasonwucj@gmail.com>
6050 * config/nds32/nds32-intrinsic.c
6051 (nds32_expand_builtin_null_ftype_reg): Delete.
6052 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6053 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6054 (nds32_read_argument): New.
6055 (nds32_legitimize_target): Ditto.
6056 (nds32_legitimize_argument): Ditto.
6057 (nds32_check_constant_argument): Ditto.
6058 (nds32_expand_unop_builtin): Ditto.
6059 (nds32_expand_unopimm_builtin): Ditto.
6060 (nds32_expand_binop_builtin): Ditto.
6061 (nds32_builtin_decl_impl): Ditto.
6062 (builtin_description): Ditto.
6063 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6064 (nds32_init_builtins_impl): Ditto.
6065 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6066 (nds32_builtin_decl): New.
6067 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6068 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6070 2018-03-02 Jeff Law <law@redhat.com>
6072 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6073 (redundant_insn, fill_simple_delay_slots): Likewise.
6074 (fill_slots_from_thread): Likewise.
6075 * resource.c (mark_referenced_resources): Likewise.
6076 (mark_set_resources, find_dead_or_set_registers): Likewise.
6078 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6080 * substring-locations.h (format_warning_va): Formatting fix for
6082 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6084 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6085 * substring-locations.c: Include intl.h.
6086 (format_warning_va): Turned into small wrapper around
6087 format_warning_n_va, renamed to ...
6088 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6089 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6091 (format_warning_at_substring_n): New function.
6092 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6093 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6094 format_warning_at_substring with just a shorter name instead of
6095 const function pointer.
6096 (fmtwarn_n): New function.
6097 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6098 appropriate, get rid of all the fmtstr temporaries, move conditionals
6099 with G_() wrapped string literals directly into fmtwarn arguments,
6100 cast dir.len to (int), formatting fixes.
6102 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6104 * doc/invoke.texi: Remove "Cilk Plus" references.
6106 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6107 Richard Biener <rguenther@suse.de>
6110 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6111 for error or warning attributes if CALL_FROM_THUNK_P is set.
6114 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6117 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6118 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6121 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6122 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6124 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6125 instead of -1U in last predictors element's probability member.
6127 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6130 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6131 arguments if they are comparable.
6133 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6135 PR tree-optimization/84634
6136 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6137 masks and masked_loop_p with a single loop_masks, making sure it's
6138 null for bb vectorization.
6140 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6142 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6143 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6144 loop->force_vectorize to check whether there is no alias.
6146 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6149 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6151 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6153 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6154 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6155 instead of a loop around prev_real_insn.
6156 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6160 * config/i386/i386.c (ix86_print_operand): Use conditional
6161 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6164 2018-03-02 Richard Biener <rguenther@suse.de>
6166 PR tree-optimization/84427
6167 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6168 (bitmap_set_subtract_values): Rewrite to handle multiple
6171 (prune_clobbered_mems): Likewise.
6172 (phi_translate): Take edge instead of pred/phiblock.
6173 (phi_translate_1): Likewise.
6174 (phi_translate_set): Likewise. Insert all translated
6175 exprs for a value into the set, keeping possibly multiple
6176 expressions per value.
6177 (compute_antic_aux): Adjust for phi_translate changes.
6178 When intersecting union the expressions and prune those
6179 not in the final value set, keeping possibly multiple
6180 expressions per value. Do not use value-insertion
6181 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6182 all expressions. Add verification that the value-sets
6183 only shrink during iteration.
6184 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6185 (do_pre_regular_insertion): Likewise.
6186 (do_pre_partial_partial_insertion): Likewise.
6188 2018-03-02 Richard Biener <rguenther@suse.de>
6191 * config/darwin.c (saved_debug_info_level): New static global.
6192 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6193 (darwin_asm_lto_end): Restore debug info generation settings.
6195 2018-03-01 Martin Liska <mliska@suse.cz>
6198 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6201 2018-03-01 Richard Biener <rguenther@suse.de>
6204 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6205 type annotation with LTO when debug was disabled at compile-time.
6207 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6209 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6211 (mips_final_postscan_insn): Likewise.
6213 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6215 PR rtl-optimization/84528
6216 * alias.c (init_alias_target): Add commentary.
6217 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6218 a unique base value if the frame pointer is not eliminated
6219 to the stack pointer.
6221 2018-03-01 Tom de Vries <tom@codesourcery.com>
6223 PR rtl-optimization/83327
6224 * lra-int.h (hard_regs_spilled_into): Declare.
6225 * lra.c (hard_regs_spilled_into): Define.
6226 (init_reg_info): Init hard_regs_spilled_into.
6227 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6228 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6229 (process_bb_lives): Handle hard_regs_spilled_into.
6230 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6231 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6233 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6235 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6236 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6237 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6238 * config/rs6000/aix72.h: New file.
6240 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6242 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6243 instead of warning_at with conditional singular and plural messages
6247 * stor-layout.c (update_alignment_for_field): For
6248 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6249 && !DECL_PACKED (field), do the alignment update, just use
6250 only desired_align instead of MAX (type_align, desired_align)
6252 (place_field): Don't do known_align < desired_align handling
6253 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6254 is non-NULL, instead do it after rli->prev_field handling and
6255 only if not within a bitfield word. For DECL_PACKED (field)
6256 use type_align of BITS_PER_UNIT.
6258 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6260 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6261 superfluous parentheses and trailing spaces.
6263 2018-02-28 Richard Biener <rguenther@suse.de>
6265 PR tree-optimization/84584
6266 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6267 SCoPs with fake exit edge.
6269 2018-02-28 Martin Liska <mliska@suse.cz>
6272 * timevar.c (timer::print): Fix format to properly print 100%
6275 2018-02-28 Richard Biener <rguenther@suse.de>
6278 * genmatch.c (capture_info::walk_match): Do not mark
6279 captured expressions without operands as expr_p given
6280 they act more like predicates and should be subject to
6281 "lost tail" side-effect preserving.
6283 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6285 PR rtl-optimization/81611
6286 * auto-inc-dec.c (attempt_change): Move dead note from
6287 mem_insn if it's the next use of regno
6288 (find_address): Take address use of reg holding
6289 non-incremented value. Add parm to limit search to the named
6291 (merge_in_block): Attempt to use a mem insn that is the next
6292 use of the original regno.
6294 2018-02-27 Martin Sebor <msebor@redhat.com>
6297 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6298 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6300 2018-02-27 Martin Sebor <msebor@redhat.com>
6302 PR translation/84207
6303 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6304 n argument to unsigned HOST_WIDE_INT.
6305 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6306 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6307 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6308 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6310 2018-02-27 Richard Biener <rguenther@suse.de>
6312 PR tree-optimization/84512
6313 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6314 Do not use the estimate returned from record_stmt_cost for
6315 the scalar iteration cost but sum properly using add_stmt_cost.
6317 2018-02-27 Richard Biener <rguenther@suse.de>
6319 PR tree-optimization/84466
6320 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6321 Adjust last change to less strictly validate use operands.
6323 2018-02-27 Martin Liska <mliska@suse.cz>
6325 PR gcov-profile/84548
6326 * gcov.c (process_file): Allow partial overlap and consider it
6327 also as group functions.
6328 (output_lines): Properly calculate range of lines for a group.
6330 2018-02-27 Martin Liska <mliska@suse.cz>
6332 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6333 'ggc' suffixes. Change first column width.
6334 (timer::print): Fix formatting of the column.
6336 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6338 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6339 preserve inline entry blocks for the sake of debug inline
6340 entry point markers alone.
6341 (remove_unused_locals): Suggest in comments a better place to
6342 force the preservation of inline entry blocks that are
6343 otherwise unused, but do not preserve them.
6345 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6347 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6349 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6352 * config/i386/constraints.md (Bs): Replace
6353 ix86_indirect_branch_register with
6354 TARGET_INDIRECT_BRANCH_REGISTER.
6356 * config/i386/i386.md (indirect_jump): Likewise.
6357 (tablejump): Likewise.
6358 (*sibcall_memory): Likewise.
6359 (*sibcall_value_memory): Likewise.
6360 Peepholes of indirect call and jump via memory: Likewise.
6361 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6362 (*sibcall_value_GOT_32): Likewise.
6363 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6364 (GOT_memory_operand): Likewise.
6365 (call_insn_operand): Likewise.
6366 (sibcall_insn_operand): Likewise.
6367 (GOT32_symbol_operand): Likewise.
6368 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6370 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6372 PR rtl-optimization/83496
6373 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6374 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6375 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6376 redundant insn, if any.
6377 (relax_delay_slots): Likewise.
6378 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6380 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6382 PR tree-optimization/83965
6383 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6384 that grouped statements are part of a reduction chain. Return
6385 true if the statement is not marked as a reduction itself but
6387 (vect_recog_dot_prod_pattern): Don't check whether the statement
6388 is part of a group here.
6389 (vect_recog_sad_pattern): Likewise.
6390 (vect_recog_widen_sum_pattern): Likewise.
6392 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6395 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6397 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6400 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6402 (ix86_output_indirect_function_return): New prototype.
6403 (ix86_split_simple_return_pop_internal): Likewise.
6404 * config/i386/i386.c (indirect_return_via_cx): New.
6405 (indirect_return_via_cx_bnd): Likewise.
6406 (indirect_thunk_name): Handle return va CX_REG.
6407 (output_indirect_thunk_function): Create alias for
6408 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6409 (ix86_output_indirect_jmp): Remove the bool argument.
6410 (ix86_output_indirect_function_return): New function.
6411 (ix86_split_simple_return_pop_internal): Likewise.
6412 * config/i386/i386.md (*indirect_jump): Don't pass false
6413 to ix86_output_indirect_jmp.
6414 (*tablejump_1): Likewise.
6415 (simple_return_pop_internal): Change it to define_insn_and_split.
6416 Call ix86_split_simple_return_pop_internal to split it for
6418 (simple_return_indirect_internal): Call
6419 ix86_output_indirect_function_return instead of
6420 ix86_output_indirect_jmp.
6422 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6425 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6426 memset and value initialization afterwards.
6428 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6430 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6432 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6435 * common/config/aarch64/aarch64-common.c
6436 (aarch_option_optimization_table[]): Switch
6437 off fomit-frame-pointer
6439 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6440 Chung-Ju Wu <jasonwucj@gmail.com>
6442 * config/nds32/nds32-multiple.md (load_multiple): Disallow
6444 (store_multiple): Ditto.
6446 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6448 * config.gcc: Add --with-cpu support for nds32 target.
6449 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
6450 * config/nds32/nds32.opt: Add -mcpu= option.
6452 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
6454 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
6455 isel=yes): Warn for these deprecated options.
6457 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
6459 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
6460 ISA_2_5_MASKS_EMBEDDED.
6462 2018-02-23 Jakub Jelinek <jakub@redhat.com>
6464 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
6465 p->max as pointers rather than using iterative_hash_expr.
6467 2018-02-23 Carl Love <cel@us.ibm.com>
6469 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
6470 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
6472 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
6473 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
6474 P8V_BUILTIN_VEC_VUNSIGNED2.
6476 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
6479 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
6480 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
6482 * lra-constraints.c (curr_insn_transform): Set up
6483 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
6485 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
6486 * lra-eliminations.c (spill_pseudos): Ditto.
6487 (process_insn_for_elimination): Ditto.
6488 * lra-lives.c (reg_early_clobber_p): Use the new macros.
6489 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
6490 LRA_NON_CLOBBERED_ALT.
6492 2018-02-22 Martin Sebor <msebor@redhat.com>
6494 PR tree-optimization/84480
6495 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
6496 to maybe_diag_stxncpy_trunc. Call it.
6497 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
6498 from gimple_fold_builtin_strcpy. Print inlining stack.
6499 (handle_builtin_stxncpy): Print inlining stack.
6500 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
6502 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
6505 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
6506 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
6507 and -fcheck-pointer-bounds are used together.
6508 (indirect_thunk_prefix): New enum.
6509 (indirect_thunk_need_prefix): New function.
6510 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
6511 "_nt" instead of "_bnd" for NOTRACK prefix.
6512 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
6513 (output_indirect_thunk_function): Likewise.
6515 (ix86_code_end): Update output_indirect_thunk_function calls.
6516 (ix86_output_indirect_branch_via_reg): Replace
6517 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
6518 (ix86_output_indirect_branch_via_push): Likewise.
6519 (ix86_output_function_return): Likewise.
6520 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
6521 incompatible with -fcf-protection=branch and
6522 -fcheck-pointer-bounds.
6524 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6527 * config/aarch64/aarch64.c (aarch64_print_address_internal):
6528 Change gcc_assert call to output_operand_lossage.
6530 2018-02-22 Steve Ellcey <sellcey@cavium.com>
6532 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
6534 2018-02-22 DJ Delorie <dj@redhat.com>
6535 Sebastian Perta <sebastian.perta@renesas.com>
6536 Oleg Endo <olegendo@gcc.gnu.org>
6538 * config/rx/rx.c (rx_rtx_costs): New function.
6539 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
6541 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
6543 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
6545 2018-02-22 Martin Liska <mliska@suse.cz>
6548 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
6549 Add "native" as a possible value.
6551 2018-02-22 Martin Liska <mliska@suse.cz>
6554 * config/i386/i386.c (ix86_option_override_internal):
6555 Add "native" as a possible value for -march and -mtune.
6557 2018-02-22 Jakub Jelinek <jakub@redhat.com>
6560 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
6561 to all type variants.
6563 PR tree-optimization/84503
6564 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
6565 width as info->bitpos + info->bitsize - start.
6566 (merged_store_group::merge_overlapping): Simplify width computation.
6567 (check_no_overlap): New function.
6568 (imm_store_chain_info::try_coalesce_bswap): Compute expected
6569 start + width and last_order of the group, fail if check_no_overlap
6571 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
6572 to group if check_no_overlap fails.
6574 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6576 * config/rs6000/altivec.md: Delete contraint arguments to
6577 define_expand, define_split, and define_peephole2, and in
6578 define_insn_and_split if always unused.
6579 * config/rs6000/darwin.md: Ditto.
6580 * config/rs6000/dfp.md: Ditto.
6581 * config/rs6000/rs6000.md: Ditto.
6582 * config/rs6000/sync.md: Ditto.
6583 * config/rs6000/vector.md: Ditto.
6584 * config/rs6000/vsx.md: Ditto.
6586 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
6588 * config/rs6000/altivec.md: Write output control strings as braced
6589 blocks instead of double-quoted strings.
6590 * config/rs6000/darwin.md: Ditto.
6591 * config/rs6000/rs6000.md: Ditto.
6592 * config/rs6000/vector.md: Ditto.
6593 * config/rs6000/vsx.md: Ditto.
6595 2018-02-21 Jason Merrill <jason@redhat.com>
6597 PR c++/84314 - ICE with templates and fastcall attribute.
6598 * attribs.c (build_type_attribute_qual_variant): Remove assert.
6600 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6602 * ipa-cp.c (determine_versionability): Fix comment typos.
6604 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
6607 * ipa-cp.c (determine_versionability): Do not version functions caling
6610 2018-02-21 Martin Liska <mliska@suse.cz>
6613 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
6614 Add "native" as a possible value.
6615 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
6616 the macro when native cpu detection is available.
6618 2018-02-21 Martin Liska <mliska@suse.cz>
6621 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
6622 Add "native" as a possible value.
6623 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
6624 when native cpu detection is available.
6626 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6627 Martin Sebor <msebor@redhat.com>
6629 PR tree-optimization/84478
6630 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6632 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6633 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6634 support which is conservatively correct, for 2 only stay conservative
6635 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6636 argument to the 2 argument get_range_strlen, adjust 6 arg
6637 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6639 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6640 (gimple_fold_builtin_strlen): Pass true as last argument to
6643 2018-02-20 Martin Sebor <msebor@redhat.com>
6646 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6647 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6648 (builtin_memref::builtin_memref): Factor out parts into
6649 set_base_and_offset and call it.
6651 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6654 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6655 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6656 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6657 search at the associated MODE_INT.
6659 2018-02-20 Jeff Law <law@redhat.com>
6662 PR tree-optimization/81592
6664 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6665 for range data rather than using global data.
6666 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6667 range data rather than using global data.
6668 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6669 pass it to children as needed.
6670 (struct directive::fmtresult): Similarly.
6671 (struct directive::set_width): Similarly.
6672 (struct directive::set_precision): Similarly.
6673 (format_integer, format_directive, parse_directive): Similarly.
6674 (format_none): Accept unnamed vr_values parameter.
6675 (format_percent, format_floating, format_character): Similarly.
6676 (format_string, format_plain): Similarly.
6677 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6678 the EVRP range analyzer for range data rather than using global data.
6679 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6680 gimple-ssa-evrp-analyze.h
6681 (class sprintf_dom_walker): Add after_dom_children member function.
6682 Add evrp_range_analyzer member.
6683 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6684 range analyzer as needed.
6685 (sprintf_dom_walker::after_dom_children): New member function.
6686 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6688 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6689 (evrp_range_analyzer::pop_to_marker): Likewise.
6691 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6693 PR tree-optimization/84419
6694 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6695 with the required type if its current type is compatible but
6698 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6701 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6702 after vectorization.
6704 2018-02-20 Martin Liska <mliska@suse.cz>
6707 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6708 possible values if we don't have a hint.
6710 2018-02-20 Martin Liska <mliska@suse.cz>
6714 * final.c (shorten_branches): Build align_tab array with one
6716 * opts.c (finish_options): Add alignment option limit check.
6717 (MAX_CODE_ALIGN): Likewise.
6718 (MAX_CODE_ALIGN_VALUE): Likewise.
6719 * doc/invoke.texi: Document maximum allowed option value for
6720 all -falign-* options.
6722 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6725 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6726 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6727 * var-tracking.c (emit_note_insn_var_location): Remove all references
6728 to NOTE_INSN_CALL_ARG_LOCATION.
6729 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6730 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6731 Use copy_rtx_if_shared.
6732 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6733 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6734 (dwarf2out_var_location): Remove handling of
6735 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6737 * final.c (final_scan_insn): Remove all references to
6738 NOTE_INSN_CALL_ARG_LOCATION.
6739 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6740 before dumping final insns.
6741 * except.c (emit_note_eh_region_end): Remove all references to
6742 NOTE_INSN_CALL_ARG_LOCATION.
6743 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6744 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6745 * config/arc/arc.c (hwloop_optimize): Likewise.
6746 * config/arm/arm.c (create_fix_barrier): Likewise.
6747 * config/s390/s390.c (s390_chunkify_start): Likewise.
6748 * config/sh/sh.c (find_barrier): Likewise.
6749 * config/i386/i386.c (rest_of_insert_endbranch,
6750 ix86_seh_fixup_eh_fallthru): Likewise.
6751 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6752 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6753 * config/frv/frv.c (frv_function_prologue): Likewise.
6754 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6756 (note_outside_basic_block_p): Remove all references to
6757 NOTE_INSN_CALL_ARG_LOCATION.
6758 * gengtype.c (adjust_field_rtx_def): Likewise.
6759 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6761 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6762 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6765 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6768 PR tree-optimization/84452
6769 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6770 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6773 2018-02-19 Martin Liska <mliska@suse.cz>
6776 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6778 2018-02-19 Martin Liska <mliska@suse.cz>
6779 Richard Sandiford <richard.sandiford@linaro.org>
6781 PR tree-optimization/82491
6782 * gimple-fold.c (get_base_constructor): Make earlier bail out
6785 2018-02-19 Carl Love <cel@us.ibm.com>
6787 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6788 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6790 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6791 P8V_BUILTIN_VEC_NEG.
6793 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6795 * config/rl78/rl78.md (movdf): New define expand.
6797 2018-02-19 Martin Liska <mliska@suse.cz>
6800 * doc/invoke.texi: Fix typo.
6801 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6803 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6805 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6806 handle rs6000_single_float and rs6000_double_float specially for
6809 2018-02-16 Jeff Law <law@redhat.com>
6811 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6814 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6817 * ipa-inline.c (inline_small_functions): Fix a typo.
6819 2018-02-16 Nathan Sidwell <nathan@acm.org>
6821 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6823 2018-02-16 Carl Love <cel@us.ibm.com>
6825 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6826 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6827 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6828 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6829 expansion to P8V_BUILTIN_VEC_FLOAT2.
6831 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6833 PR rtl-optimization/70023
6834 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6835 src_regno into account.
6837 2018-02-16 Carl Love <cel@us.ibm.com>
6839 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6840 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6841 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6842 * config/rs6000/rs6000.c: Remove case statements for
6843 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6844 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6845 and P9V_BUILTIN_VEC_VINSERT4B.
6846 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6847 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6848 * config/rs6000/vsx.md:
6849 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6852 2018-02-16 Carl Love <cel@us.ibm.com>
6854 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6856 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6858 * config/rs6000/rs6000-c.c: Add the definitions for
6859 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6860 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6861 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6862 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6863 definition for insert4b and define insn *insert3b_internal.
6864 * doc/extend.texi: Add documentation for vec_extract4b.
6866 2018-02-16 Nathan Sidwell <nathan@acm.org>
6868 * doc/extend.texi (Backwards Compatibility): Mention friend
6869 injection. Note for-scope is deprecated.
6870 * doc/invoke.texi (-ffriend-injection): Deprecate.
6872 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6874 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6875 that moved to I2, also allow destinations that are a paradoxical
6876 subreg (instead of a normal reg).
6878 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6881 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6884 2018-02-16 Richard Biener <rguenther@suse.de>
6886 PR tree-optimization/84037
6887 PR tree-optimization/84016
6889 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6890 Adjust vec_construct for the fact we need additional higher latency
6891 128bit inserts for AVX256 and AVX512 vector builds.
6892 (ix86_add_stmt_cost): Scale vector construction cost for
6895 2018-02-16 Richard Biener <rguenther@suse.de>
6897 PR tree-optimization/84417
6898 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6899 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6900 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6902 2018-02-16 Martin Liska <mliska@suse.cz>
6905 * internal-fn.def (ASAN_CHECK): Set proper flags.
6906 (ASAN_MARK): Likewise.
6908 2018-02-16 Julia Koval <julia.koval@intel.com>
6910 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6911 from PTA_CANNONLAKE.
6913 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6916 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6917 Use ++iter rather than iter++ for std::list iterators.
6918 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6919 defer deleting them until all nodes in the forest are processed. Do
6920 free even leaf nodes. Change to_process into auto_vec.
6923 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6924 * vec.h (vec_default_construct): Use memset instead of placement new
6925 if BROKEN_VALUE_INITIALIZATION is defined.
6926 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6927 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6930 PR rtl-optimization/83723
6931 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6932 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6933 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6935 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6937 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6939 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6941 PR rtl-optimization/81443
6942 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6943 from inner REGs to paradoxical SUBREGs.
6945 2018-02-16 Richard Biener <rguenther@suse.de>
6947 PR tree-optimization/84399
6948 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6949 For operands we can analyze at their definition make sure we can
6950 analyze them at each use as well.
6952 2018-02-16 Richard Biener <rguenther@suse.de>
6954 PR tree-optimization/84190
6955 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6956 volatile accesses if the decl isn't volatile.
6958 2018-02-15 Jason Merrill <jason@redhat.com>
6960 PR c++/84314 - ICE with templates and fastcall attribute.
6961 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6962 TYPE_CANONICAL on an existing type.
6964 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6966 PR tree-optimization/84383
6967 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6968 dstoff nor call operand_equal_p if dstbase is NULL.
6970 PR tree-optimization/84334
6971 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6972 also a CONSTANT_CLASS_P, punt.
6974 2018-02-14 Jim Wilson <jimw@sifive.com>
6976 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6977 first SMALL_OPERAND check. New local min_second_step. Move assert
6978 to where locals are set. Add TARGET_RVC support.
6979 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6981 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6983 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6985 2018-02-14 Martin Sebor <msebor@redhat.com>
6987 PR tree-optimization/83698
6988 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6989 arrays constrain the offset range to their bounds.
6990 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6991 (builtin_access::overlap): Avoid setting the size of overlap if it's
6993 (maybe_diag_overlap): Also consider arrays when deciding what values
6994 of offsets to include in diagnostics.
6996 2018-02-14 Martin Sebor <msebor@redhat.com>
6999 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7000 that correspond to the kind of a declaration.
7002 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7005 * config/pa/pa.md: Load address of PIC label using the linkage table
7006 if the label is nonlocal.
7008 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7010 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7011 warning message if user requests -maltivec=be.
7012 * doc/invoke.texi: Document deprecation of -maltivec=be.
7014 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7017 * config/rs6000/rs6000-c.c: Update definitions for
7018 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7019 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7021 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7024 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7025 add _get_ssp intrinsics. Remove argument from
7026 __builtin_ia32_rdssp[d|q].
7027 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7028 * config/i386/i386-builtin.def: Remove argument from
7029 __builtin_ia32_rdssp[d|q].
7030 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7031 ix86_expand_special_args_builtin for _rdssp[d|q].
7032 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7033 Clear register before usage.
7034 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7035 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7037 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7039 PR tree-optimization/84357
7040 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7041 operand 1 of an ARRAY_REF too.
7043 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7046 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7047 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7049 (set_of_reg): New struct.
7050 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7051 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7052 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7054 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7055 Split into bitclr, bitset, bitinvert patterns if appropriate.
7056 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7057 use rx_fuse_in_memory_bitop.
7058 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7059 to named insn, correct maximum insn length.
7061 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7064 * machmode.def: Define a complex mode for PARTIAL_INT.
7065 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7067 * doc/rtl.texi: Document CSPImode.
7068 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7070 (msp430_hard_regno_nregs_with_padding): Likewise.
7072 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7075 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7077 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7079 PR rtl-optimization/84169
7080 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7081 we generated a parallel as new i3 and we split that to new i2 and i3
7082 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7083 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7084 those to i2, not i1. Partially rewrite this scan code.
7086 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7089 * stor-layout.c (place_field): For variable length fields, adjust
7090 offset_align afterwards not just based on the field's alignment,
7091 but also on the size.
7094 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7095 of exps and logs in the use_exp2 case.
7097 2018-02-13 Jeff Law <law@redhat.com>
7099 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7102 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7105 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7109 * common.opt (gas-loc-support, gas-locview-support): New.
7110 (ginline-points, ginternal-reset-location-views): New.
7111 * doc/invoke.texi: Document them. Use @itemx where intended.
7112 (gvariable-location-views): Adjust.
7113 * target.def (reset_location_view): New.
7114 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7115 (TARGET_RESET_LOCATION_VIEW): New.
7116 * doc/tm.texi: Rebuilt.
7117 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7118 (dwarf2out_default_as_locview_support): New.
7119 (output_asm_line_debug_info): Use option variables.
7120 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7121 (output_loc_list): Likewise.
7122 (add_high_low_attributes): Check option variables.
7123 Don't output entry view attribute in strict mode.
7124 (gen_inlined_subroutine_die): Check option variables.
7125 (dwarf2out_inline_entry): Likewise.
7126 (init_sections_and_labels): Likewise.
7127 (dwarf2out_early_finish): Likewise.
7128 (maybe_reset_location_view): New, from...
7129 (dwarf2out_var_location): ... here. Call it.
7130 * debug.h (dwarf2out_default_as_loc_support): Declare.
7131 (dwarf2out_default_as_locview_support): Declare.
7132 * hooks.c (hook_int_rtx_insn_0): New.
7133 * hooks.h (hook_int_rtx_insn_0): Declare.
7134 * toplev.c (process_options): Take -gas-loc-support and
7135 -gas-locview-support from dwarf2out. Enable
7136 -gvariable-location-views by default only with locview
7137 assembler support. Enable -ginternal-reset-location-views by
7138 default only if the target defines the corresponding hook.
7139 Enable -ginline-points by default if location views are
7140 enabled; force it disabled if statement frontiers are
7142 * tree-inline.c (expand_call_inline): Check option variables.
7143 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7145 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7147 PR tree-optimization/84321
7148 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7149 handling. Also check whether the anti-range contains any values
7150 that satisfy the mask; switch to a VR_RANGE if not.
7152 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7155 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7157 2018-02-13 Martin Jambor <mjambor@suse.cz>
7160 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7161 of call statements, also set location of a load to a temporary.
7163 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7165 * config/rl78/rl78.c (add_vector_labels): New function.
7166 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7167 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7168 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7169 which checks that no arguments are passed.
7170 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7171 * doc/extend.texi: Documentation for the new attribute.
7173 2018-02-13 Andreas Schwab <schwab@suse.de>
7175 * config/riscv/linux.h (CPP_SPEC): Define.
7177 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7180 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7181 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7182 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7183 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7184 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7185 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7186 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7187 temporarily for AES and PCLMUL builtins.
7189 PR tree-optimization/84339
7190 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7191 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7195 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7196 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7197 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7199 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7201 * omp-simd-clone.h: New file.
7202 * omp-simd-clone.c: Include omp-simd-clone.h.
7203 (expand_simd_clones): No longer static.
7204 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7205 cgraph.h and omp-simd-clone.h.
7206 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7207 (vect_recog_widen_shift_pattern): Formatting fix.
7208 (vect_pattern_recog_1): Don't check optab for calls.
7211 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7212 operands[2] into a REG before using gen_lowpart on it.
7214 2018-02-12 Jeff Law <law@redhat.com>
7217 * config/sh/sh.c (find_barrier): Consider a sibling call
7220 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7221 successfully back substituting a reg.
7223 2018-02-12 Richard Biener <rguenther@suse.de>
7225 PR tree-optimization/84037
7226 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7227 parameter, move visited init to caller.
7228 (vect_slp_analyze_operations): Separate cost from validity
7229 check, initialize visited once for all instances.
7230 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7232 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7234 (vect_model_promotion_demotion_cost): Likewise.
7235 (vectorizable_bswap): Guard cost modeling with !slp_node
7236 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7238 (vectorizable_call): Likewise.
7239 (vectorizable_conversion): Likewise.
7240 (vectorizable_assignment): Likewise.
7241 (vectorizable_shift): Likewise.
7242 (vectorizable_operation): Likewise.
7243 (vectorizable_store): Likewise.
7244 (vectorizable_load): Likewise.
7245 (vectorizable_condition): Likewise.
7246 (vectorizable_comparison): Likewise.
7248 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7251 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7252 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7253 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7255 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7258 * params.def (inline-min-speedup): Increase from 8 to 15.
7259 (max-inline-insns-auto): Decrease from 40 to 30.
7260 * ipa-split.c (consider_split): Add some buffer for function to
7261 be considered inlining candidate.
7262 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7265 2018-02-12 Richard Biener <rguenther@suse.de>
7267 PR tree-optimization/84037
7268 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7269 matched stmts if we cannot swap the non-matched ones.
7271 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7273 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7274 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7275 _mm_maskz_scalef_round_ss): New intrinsics.
7276 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7277 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7278 __builtin_ia32_scalefss_round): Remove.
7279 (__builtin_ia32_scalefsd_mask_round,
7280 __builtin_ia32_scalefss_mask_round): New intrinsics.
7281 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7282 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7283 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7284 "<round_constraint>")): Changed to ...
7285 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7286 "<round_scalar_constraint>")): ... this.
7287 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7288 %0, %1, %2<round_op3>}"): Changed to ...
7289 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7290 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7291 %2<round_scalar_mask_op3>}"): ... this.
7292 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7294 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7296 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7297 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7298 (_mm_maskz_sqrt_round_ss): New intrinsics.
7299 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7300 (__builtin_ia32_sqrtsd_mask_round)
7301 (__builtin_ia32_sqrtss_mask_round): New builtins.
7302 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7303 (__builtin_ia32_sqrtss_round): Remove.
7304 (__builtin_ia32_sqrtsd_mask_round)
7305 (__builtin_ia32_sqrtss_mask_round): New builtins.
7306 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7307 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7308 ((match_operand:VF_128 1 "vector_operand"
7309 "xBm,<round_constraint>")): Changed to ...
7310 ((match_operand:VF_128 1 "vector_operand"
7311 "xBm,<round_scalar_constraint>")): ... this.
7312 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7313 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7314 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7315 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7316 %<iptr>1<round_scalar_mask_op3>}): ... this.
7317 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7318 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7320 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7323 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7324 Cast vec_cmpeq result to correct type.
7325 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7326 Cast vec_cmpgt result to correct type.
7328 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7330 * final.c (final_scan_insn_1): Renamed from...
7331 (final_scan_insn): ... this. New wrapper, to recover
7332 seen from the outermost call in recursive ones.
7333 * config/sparc/sparc.c (output_return): Drop seen from call.
7334 (output_sibcall): Likewise.
7335 * config/visium/visium.c (output_branch): Likewise.
7337 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7339 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7342 2018-02-10 Alan Modra <amodra@gmail.com>
7345 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7346 Specify LR as an input.
7348 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7351 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7352 remove_member_access_dummy_vars): New functions.
7353 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7354 lower_omp_1, execute_lower_omp): Use them.
7356 PR rtl-optimization/84308
7357 * shrink-wrap.c (spread_components): Release todo vector.
7359 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7361 PR rtl-optimization/57193
7362 * ira-color.c (struct allocno_color_data): Add member
7363 conflict_allocno_hard_prefs.
7364 (update_conflict_allocno_hard_prefs): New.
7365 (bucket_allocno_compare_func): Add a preference based on
7366 conflict_allocno_hard_prefs.
7367 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7368 (color_allocnos): Remove a dead code. Initiate
7369 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7371 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7374 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7375 constraint from =wa to wa. Avoid a subreg on the output operand,
7376 instead use a pseudo and subreg it in a move.
7377 (p9_xxbrd_<mode>): Changed to ...
7378 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7379 (p9_xxbrd_v2df): New expander.
7380 (p9_xxbrw_<mode>): Changed to ...
7381 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7382 (p9_xxbrw_v4sf): New expander.
7384 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7386 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7388 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7391 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7392 multiply in 32-bit mode.
7393 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7394 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7397 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7399 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7400 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7401 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7402 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7404 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7407 * dwarf2out.c (is_trivial_indirect_ref): New function.
7408 (dwarf2out_late_global_decl): Do not generate a location
7409 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7410 and that are not defined in the current unit.
7412 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7414 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7415 instead of a libcall for UNORDERED.
7417 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7420 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7421 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7423 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7426 * config/s390/s390.c (s390_set_current_function): Invoke
7427 s390_indirect_branch_settings also if fndecl didn't change.
7429 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7431 * config/rs6000/rs6000.md (blockage): Set length to zero.
7433 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7435 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7437 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7440 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7441 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
7445 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
7446 PARALLEL incoming that failed vt_get_decl_and_offset check.
7449 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
7450 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
7452 (get_variable_section): For decls in named .bss* sections pass true as
7453 second argument to bss_initializer_p.
7455 2018-02-09 Marek Polacek <polacek@redhat.com>
7456 Jakub Jelinek <jakub@redhat.com>
7459 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
7460 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
7461 Sync some changes from cxx_fold_indirect_ref.
7463 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7465 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
7467 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
7468 (BLOCK_INLINE_ENTRY_LABEL): New.
7469 (dwarf2out_var_location): Disregard inline entry markers.
7470 (inline_entry_data): New struct.
7471 (inline_entry_data_hasher): New hashtable type.
7472 (inline_entry_data_hasher::hash): New.
7473 (inline_entry_data_hasher::equal): New.
7474 (inline_entry_data_table): New variable.
7475 (add_high_low_attributes): Add DW_AT_entry_pc and
7476 DW_AT_GNU_entry_view attributes if a pending entry is found
7477 in inline_entry_data_table. Add old entry_pc attribute only
7478 if debug nonbinding markers are disabled.
7479 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
7480 markers are enabled.
7481 (block_within_block_p, dwarf2out_inline_entry): New.
7482 (dwarf2out_finish): Check that no entries remained in
7483 inline_entry_data_table.
7484 * final.c (reemit_insn_block_notes): Handle inline entry notes.
7485 (final_scan_insn, notice_source_line): Likewise.
7486 (rest_of_clean_state): Skip inline entry markers.
7487 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
7489 * gimple.c (gimple_build_debug_inline_entry): New.
7490 * gimple.h (enum gimple_debug_subcode): Add
7491 GIMPLE_DEBUG_INLINE_ENTRY.
7492 (gimple_build_debug_inline_entry): Declare.
7493 (gimple_debug_inline_entry_p): New.
7494 (gimple_debug_nonbind_marker_p): Adjust.
7495 * insn-notes.def (INLINE_ENTRY): New.
7496 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
7497 inline entry marker notes.
7498 (print_insn): Likewise.
7499 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
7500 (INSN_DEBUG_MARKER_KIND): Likewise.
7501 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
7502 * tree-inline.c (expand_call_inline): Build and insert
7503 debug_inline_entry stmt.
7504 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
7505 inline entry blocks early, if nonbind markers are enabled.
7506 (dump_scope_block): Dump fragment info.
7507 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
7508 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
7509 (gimple_build_debug_inline_entry): New.
7510 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
7511 Enable/disable inline entry points too.
7512 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
7513 (DEBUG_INSN): Describe inline entry markers.
7515 * common.opt (gvariable-location-views): New.
7516 (gvariable-location-views=incompat5): New.
7517 * config.in: Rebuilt.
7518 * configure: Rebuilt.
7519 * configure.ac: Test assembler for view support.
7520 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
7521 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
7522 * dwarf2out.c (var_loc_view): New typedef.
7523 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
7524 (dwarf2out_locviews_in_attribute): New.
7525 (dwarf2out_locviews_in_loclist): New.
7526 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
7527 (enum dw_line_info_opcode): Add LI_adv_address.
7528 (struct dw_line_info_table): Add view.
7529 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
7530 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
7531 (zero_view_p): New variable.
7532 (ZERO_VIEW_P): New macro.
7533 (output_asm_line_debug_info): New.
7534 (struct var_loc_node): Add view.
7535 (add_AT_view_list, AT_loc_list): New.
7536 (add_var_loc_to_decl): Add view param. Test it against last.
7537 (new_loc_list): Add view params. Record them.
7538 (AT_loc_list_ptr): Handle loc and view lists.
7539 (view_list_to_loc_list_val_node): New.
7540 (print_dw_val): Handle dw_val_class_view_list.
7541 (size_of_die): Likewise.
7542 (value_format): Likewise.
7543 (loc_list_has_views): New.
7544 (gen_llsym): Set vl_symbol too.
7545 (maybe_gen_llsym, skip_loc_list_entry): New.
7546 (dwarf2out_maybe_output_loclist_view_pair): New.
7547 (output_loc_list): Output view list or entries too.
7548 (output_view_list_offset): New.
7549 (output_die): Handle dw_val_class_view_list.
7550 (output_dwarf_version): New.
7551 (output_compilation_unit_header): Use it.
7552 (output_skeleton_debug_sections): Likewise.
7553 (output_rnglists, output_line_info): Likewise.
7554 (output_pubnames, output_aranges): Update version comments.
7555 (output_one_line_info_table): Output view numbers in asm comments.
7556 (dw_loc_list): Determine current endview, pass it to new_loc_list.
7557 Call maybe_gen_llsym.
7558 (loc_list_from_tree_1): Adjust.
7559 (add_AT_location_description): Create view list attribute if
7560 needed, check it's absent otherwise.
7561 (convert_cfa_to_fb_loc_list): Adjust.
7562 (maybe_emit_file): Call output_asm_line_debug_info for test.
7563 (dwarf2out_var_location): Reset views as needed. Precompute
7564 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
7565 attribute. Set view.
7566 (new_line_info_table): Reset next view.
7567 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
7568 (dwarf2out_source_line): Likewise. Output view resets and labels to
7569 the assembler, or select appropriate line info opcodes.
7570 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
7571 (optimize_string_length): Catch it. Adjust.
7572 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
7573 dw_val_class_view_list, and remove it if no longer needed.
7574 (hash_loc_list): Hash view numbers.
7575 (loc_list_hasher::equal): Compare them.
7576 (optimize_location_lists): Check whether a view list symbol is
7577 needed, and whether the locview attribute is present, and
7578 whether they match. Remove the locview attribute if no longer
7580 (index_location_lists): Call skip_loc_list_entry for test.
7581 (dwarf2out_finish): Call output_asm_line_debug_info for test.
7582 Use output_dwarf_version.
7583 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
7584 (struct dw_val_node): Add val_view_list.
7585 * final.c (SEEN_NEXT_VIEW): New.
7586 (set_next_view_needed): New.
7587 (clear_next_view_needed): New.
7588 (maybe_output_next_view): New.
7589 (final_start_function): Rename to...
7590 (final_start_function_1): ... this. Take pointer to FIRST,
7591 add SEEN parameter. Emit param bindings in the initial view.
7592 (final_start_function): Reintroduce SEEN-less interface.
7593 (final): Rename to...
7594 (final_1): ... this. Take SEEN parameter. Output final pending
7595 next view at the end.
7596 (final): Reintroduce seen-less interface.
7597 (final_scan_insn): Output pending next view before switching
7598 sections or ending a block. Mark the next view as needed when
7599 outputting variable locations. Notify debug backend of section
7600 changes, and of location view changes.
7601 (rest_of_handle_final): Adjust.
7602 * toplev.c (process_options): Autodetect value for debug variable
7603 location views option. Warn on incompat5 without -gdwarf-5.
7604 * doc/invoke.texi (gvariable-location-views): New.
7605 (gvariable-location-views=incompat5): New.
7606 (gno-variable-location-views): New.
7608 2018-02-08 David Malcolm <dmalcolm@redhat.com>
7610 PR tree-optimization/84136
7611 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
7612 that the result of find_edge is non-NULL.
7614 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7617 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
7618 storing integer register in SImode. Fix cost of 256 and 512
7619 byte aligned SSE register store.
7621 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
7623 * config/i386/i386.c (ix86_multiplication_cost): Fix
7624 multiplication cost for TARGET_AVX512DQ.
7626 2018-02-08 Marek Polacek <polacek@redhat.com>
7628 PR tree-optimization/84238
7629 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7632 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7634 PR tree-optimization/84265
7635 * tree-vect-stmts.c (vectorizable_store): Don't treat
7636 VMAT_CONTIGUOUS accesses as grouped.
7637 (vectorizable_load): Likewise.
7639 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7641 PR tree-optimization/81635
7642 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7643 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7644 (test_round_for_mask): New functions.
7645 (wide_int_cc_tests): Call test_round_for_mask.
7646 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7647 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7648 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7649 range returned by get_range_info.
7651 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7654 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7655 * symtab.c: Include builtins.h
7656 (symtab_node::output_to_lto_symbol_table_p): Move here
7657 from lto-streamer-out.c:output_symbol_p.
7658 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7659 (output_symbol_p): Move all logic to symtab.c
7660 (produce_symtab): Update.
7662 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7664 * config/s390/s390-opts.h (enum indirect_branch): Define.
7665 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7666 (s390_indirect_branch_via_thunk)
7667 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7668 (enum s390_indirect_branch_type): Define.
7669 * config/s390/s390.c (struct s390_frame_layout, struct
7670 machine_function): Remove.
7671 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7672 (indirect_branch_table_label_no, indirect_branch_table_name):
7674 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7675 (enum s390_indirect_branch_option): Define.
7676 (s390_return_addr_from_memory): New function.
7677 (s390_handle_string_attribute): New function.
7678 (s390_attribute_table): Add new attribute handler.
7679 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7680 (s390_indirect_branch_via_thunk): New function.
7681 (s390_indirect_branch_via_inline_thunk): New function.
7682 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7683 sibling call optimization for non z10 compiles.
7684 (s390_emit_call): Force indirect branch target to be a single
7685 register. Add r1 clobber for non-z10 compiles.
7686 (s390_emit_epilogue): Emit return jump via return_use expander.
7687 (s390_reorg): Handle JUMP_INSNs as execute targets.
7688 (s390_option_override_internal): Perform validity checks for the
7689 new command line options.
7690 (s390_indirect_branch_attrvalue): New function.
7691 (s390_indirect_branch_settings): New function.
7692 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7693 (s390_output_indirect_thunk_function): New function.
7694 (s390_code_end): Implement target hook.
7695 (s390_case_values_threshold): Implement target hook.
7696 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7698 * config/s390/s390.h (struct s390_frame_layout)
7699 (struct machine_function): Move here from s390.c.
7700 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7701 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7702 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7703 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7704 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7705 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7706 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7707 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7708 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7709 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7710 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7711 (mnemonic attribute): Add values which aren't recognized
7713 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7714 pattern for branch conversion. Fix mnemonic attribute.
7715 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7716 indirect branch via thunk if requested.
7717 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7718 ("*indirect_jump"): Disable for branch conversion using out of
7720 ("indirect_jump_via_thunk<mode>_z10")
7721 ("indirect_jump_via_thunk<mode>")
7722 ("indirect_jump_via_inlinethunk<mode>_z10")
7723 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7724 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7725 ("casesi_jump_via_inlinethunk<mode>_z10")
7726 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7727 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7728 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7729 ("*indirect2_jump"): Disable for branch conversion.
7730 ("casesi_jump"): Turn into expander and expand patterns for branch
7732 ("return_use"): New expander.
7733 ("*return"): Emit return via thunk and rename it to ...
7734 ("*return<mode>"): ... this one.
7735 * config/s390/s390.opt: Add new options and and enum for the
7738 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7740 * lra-constraints.c (match_reload): Unconditionally use
7741 gen_lowpart_SUBREG, rather than selecting between that
7742 and equivalent gen_rtx_SUBREG code.
7744 2018-02-08 Richard Biener <rguenther@suse.de>
7746 PR tree-optimization/84233
7747 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7748 changed flag instead of boguously re-using phi_inserted.
7750 2018-02-08 Martin Jambor <mjambor@suse.cz>
7752 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7753 static local variables.
7755 2018-02-08 Richard Biener <rguenther@suse.de>
7757 PR tree-optimization/84278
7758 * tree-vect-stmts.c (vectorizable_store): When looking for
7759 smaller vector types to perform grouped strided loads/stores
7760 make sure the mode is supported by the target.
7761 (vectorizable_load): Likewise.
7763 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7765 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7766 Increase LDP/STP opportunities by adding adjacent callee-saves.
7768 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7770 PR rtl-optimization/84068
7771 PR rtl-optimization/83459
7772 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7774 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7776 PR tree-optimization/84224
7777 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7778 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7781 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7784 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7785 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7786 count, remove one USE.
7788 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7790 * doc/install.texi (Configuration): Document the
7791 --with-long-double-format={ibm,ieee} PowerPC configuration
7795 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7796 Convert from define_expand to be define_insn_and_split. Rework
7797 float/double/_Float128 conversions to QI/HI/SImode to work with
7798 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7799 conversions to QI/HImode types did a store and then a load to
7800 truncate the value. For conversions to VSX registers, don't split
7801 the insn, instead emit the code directly. Use the code iterator
7802 any_fix to combine signed and unsigned conversions.
7803 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7804 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7805 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7806 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7807 (fix_<mode>di2_hw): Likewise.
7808 (fixuns_<mode>di2_hw): Likewise.
7809 (fix_<mode>si2_hw): Likewise.
7810 (fixuns_<mode>si2_hw): Likewise.
7811 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7812 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7813 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7814 fix<uns>_trunc<SFDF:mode>si2_p8.
7815 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7817 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7818 (fix<uns>_<mode>_mem): Likewise.
7819 (fctiw<u>z_<mode>_mem): Likewise.
7820 (fix<uns>_<mode>_mem): Likewise.
7821 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7822 the register allocator from doing a direct move to the GPRs to do
7823 a store, and instead use the ISA 3.0 store byte/half-word from
7824 vector register instruction. For IEEE 128-bit floating point,
7825 also optimize stores of 32-bit ints.
7826 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7828 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7830 * genextract.c (push_pathstr_operand): New function to support
7832 (walk_rtx): Call push_pathstr_operand.
7833 (print_path): Support [a-zA-Z].
7835 2018-02-07 Richard Biener <rguenther@suse.de>
7837 PR tree-optimization/84037
7838 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7839 (cse_and_gimplify_to_preheader): Declare.
7840 (vect_get_place_in_interleaving_chain): Likewise.
7841 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7843 (_loop_vec_info::~_loop_vec_info): Delete it.
7844 (cse_and_gimplify_to_preheader): New function.
7845 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7846 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7847 (vectorizable_load): Likewise. For grouped stores always base
7848 the IV on the first element.
7849 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7850 condition before gimplifying.
7852 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7854 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7855 *DIV_EXPR and *MOD_EXPR.
7857 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7860 * config/i386/i386.c (ix86_option_override_internal): Mask out
7861 the CF_SET bit when checking -fcf-protection.
7863 2018-02-07 Tom de Vries <tom@codesourcery.com>
7866 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7869 2018-02-07 Richard Biener <rguenther@suse.de>
7871 PR tree-optimization/84204
7872 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7875 PR tree-optimization/84205
7876 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7877 special-case isl_ast_op_zdiv_r.
7879 PR tree-optimization/84223
7880 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7881 Only add conditions from within the region.
7882 (gather_bbs::after_dom_children): Adjust.
7884 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7887 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7888 * config/avr/avr.md: Only post-reload split REG-REG moves if
7889 either register is GENERAL_REG_P.
7891 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7893 PR tree-optimization/84235
7894 * tree-ssa-scopedtables.c
7895 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7896 if the subtraction is performed in floating point type where NaNs are
7897 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7898 build 1. Formatting fix.
7900 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7903 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7904 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7905 and skip it regardless of bb boundaries. Use CALL_P macro,
7906 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7907 unnecessarily, formatting fix.
7909 2018-02-06 Michael Collison <michael.collison@arm.com>
7911 * config/arm/thumb2.md:
7912 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7913 (*thumb_mov_notscc): Ditto.
7915 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7918 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7919 unsigned_fix, not "s".
7921 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7923 * configure.ac (gcc_fn_eh_frame_ro): New function.
7924 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7925 correct .eh_frame permissions.
7926 * configure: Regenerate.
7928 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7930 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7933 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7935 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7936 Display warning message for -mno-speculate-indirect-jumps.
7938 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7940 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7942 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7944 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7946 PR tree-optimization/84225
7947 * tree-eh.c (find_trapping_overflow): Only call
7948 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7950 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7953 * config/i386/i386.c: Reimplement the check of possible options
7954 -mibt/-mshstk conbination. Change error messages.
7955 * doc/invoke.texi: Fix a typo: remove extra '='.
7957 2018-02-06 Marek Polacek <polacek@redhat.com>
7959 PR tree-optimization/84228
7960 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7962 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7965 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7966 emitted arch directives.
7967 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7968 __ARM_FEATURE_COPROC before changing architectures.
7970 2018-02-06 Richard Biener <rguenther@suse.de>
7972 * config/i386/i386.c (print_reg): Fix typo.
7973 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7975 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7977 * configure: Regenerate.
7979 2018-02-05 Martin Sebor <msebor@redhat.com>
7981 PR tree-optimization/83369
7982 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7985 2018-02-05 Martin Liska <mliska@suse.cz>
7987 * doc/invoke.texi: Cherry-pick upstream r323995.
7989 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7991 * ira.c (ira_init_register_move_cost): Adjust comment.
7993 2018-02-05 Martin Liska <mliska@suse.cz>
7995 PR gcov-profile/84137
7996 * doc/gcov.texi: Fix typo in documentation.
7998 2018-02-05 Martin Liska <mliska@suse.cz>
8000 PR gcov-profile/83879
8001 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8002 using dlopen functionality.
8004 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8006 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8007 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8008 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8009 _mm_maskz_range_round_ss): New intrinsics.
8010 (__builtin_ia32_rangesd128_round)
8011 (__builtin_ia32_rangess128_round): Remove.
8012 (__builtin_ia32_rangesd128_mask_round,
8013 __builtin_ia32_rangess128_mask_round): New builtins.
8014 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8015 __builtin_ia32_rangess128_round): Remove.
8016 (__builtin_ia32_rangesd128_mask_round,
8017 __builtin_ia32_rangess128_mask_round): New builtins.
8018 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8019 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8020 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8021 "<round_saeonly_constraint>")): Changed to ...
8022 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8023 "<round_saeonly_scalar_constraint>")): ... this.
8024 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8025 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8026 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8027 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8028 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8030 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8032 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8034 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8035 Remove all values except native, 8540 and 8548.
8037 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8039 * config/i386/i386.c (ix86_output_function_return): Pass
8040 INVALID_REGNUM, instead of -1, as invalid register number to
8041 indirect_thunk_name and output_indirect_thunk.
8043 2018-02-02 Julia Koval <julia.koval@intel.com>
8045 * config.gcc: Add -march=icelake.
8046 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8047 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8048 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8049 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8050 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8051 (processor_target_table): Add icelake.
8052 (ix86_option_override_internal): Handle new PTAs.
8053 (get_builtin_code_for_version): Handle icelake.
8054 (M_INTEL_COREI7_ICELAKE): New.
8055 (fold_builtin_cpu): Handle icelake.
8056 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8057 * doc/invoke.texi: Add -march=icelake.
8059 2018-02-02 Julia Koval <julia.koval@intel.com>
8061 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8062 to wide_int_bitmask.
8063 * wide-int-bitmask.h: New.
8065 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8068 * config/i386/i386.md: Replace Pmode with word_mode in
8069 builtin_setjmp_setup and builtin_longjmp to support x32.
8071 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8075 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8077 (rs6000_supported_cpu_names): New static variable.
8078 (linux_cpu_translation_table): Likewise.
8079 (elf_platform) <cpu>: Define new static variable and use it.
8080 Translate kernel AT_PLATFORM name to canonical name if needed.
8081 Error if platform name is unknown.
8083 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8086 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8088 2018-02-01 Jeff Law <law@redhat.com>
8091 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8092 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8093 the scratch if RELEASE_VIA_POP is false.
8094 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8095 If we have to save a temporary register, decrement SIZE appropriately.
8096 Pass new arguments to release_scratch_register_on_entry.
8097 (ix86_adjust_stack_and_probe): Likewise.
8098 (ix86_emit_probe_stack_range): Pass new arguments to
8099 release_scratch_register_on_entry.
8101 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8103 PR rtl-optimization/84157
8104 * combine.c (change_zero_ext): Use REG_P predicate in
8105 front of HARD_REGISTER_P predicate.
8107 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8109 * config/avr/avr.c (avr_option_override): Move disabling of
8110 -fdelete-null-pointer-checks to...
8111 * common/config/avr/avr-common.c (avr_option_optimization_table):
8114 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8116 PR tree-optimization/81635
8117 * tree-data-ref.c (split_constant_offset_1): For types that
8118 wrap on overflow, try to use range info to prove that wrapping
8121 2018-02-01 Renlin Li <renlin.li@arm.com>
8124 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8126 (aarch64_register_move_cost): Likewise.
8127 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8129 (REG_CLASS_NAMES): Likewise.
8130 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8131 TAILCALL_ADDR_REGS. Remove IP registers.
8132 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8134 2018-02-01 Richard Biener <rguenther@suse.de>
8136 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8137 for specifying RPO order and allow NULL for that.
8138 * domwalk.c (dom_walker::dom_walker): Likewise.
8139 (dom_walker::walk): Handle NULL RPO order.
8140 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8142 (rewrite_update_dom_walker): Likewise.
8143 (mark_def_dom_walker): Likewise.
8145 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8147 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8148 (aarch64_maybe_expand_sve_subreg_move): Declare.
8149 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8150 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8152 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8153 that are semantically a reverse operation.
8154 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8155 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8156 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8158 (aarch64_can_change_mode_class): For big-endian, forbid changes
8159 between two SVE modes if they have different element sizes.
8161 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8163 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8164 the TImode handling for big-endian targets.
8166 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8168 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8169 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8171 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8172 Remove BSWAP handing for big-endian targets and use the form of
8173 LD1RQ appropariate for the mode.
8175 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8177 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8178 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8181 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8184 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8185 check for operands that need to go through aarch64_sve_reload_be.
8187 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8189 PR tree-optimization/81661
8190 PR tree-optimization/84117
8191 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8192 * tree-eh.c: Include gimplify.h.
8193 (find_trapping_overflow, replace_trapping_overflow,
8194 rewrite_to_non_trapping_overflow): New functions.
8195 * tree-vect-loop.c: Include tree-eh.h.
8196 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8197 * tree-data-ref.c: Include tree-eh.h.
8198 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8200 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8202 PR rtl-optimization/84123
8203 * combine.c (change_zero_ext): Check if hard register satisfies
8204 can_change_dest_mode before calling gen_lowpart_SUBREG.
8206 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8209 * ira.c (ira_init_register_move_cost): Remove assert.
8211 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8213 PR rtl-optimization/84071
8214 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8215 * doc/tm.texi: Regenerate.
8217 2018-01-31 Richard Biener <rguenther@suse.de>
8219 PR tree-optimization/84132
8220 * tree-data-ref.c (analyze_miv_subscript): Properly
8221 check whether evolution_function_is_affine_multivariate_p
8222 before calling gcd_of_steps_may_divide_p.
8224 2018-01-31 Julia Koval <julia.koval@intel.com>
8227 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8228 * config/i386/i386.md (rdpid_rex64) New.
8229 (rdpid): Make 32bit only.
8231 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8234 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8235 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8237 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8240 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8242 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8244 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8246 PR rtl-optimization/84071
8247 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8248 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8250 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8252 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8253 (arc_attribute_table): Add 'aux' attribute.
8254 (arc_in_small_data_p): Consider aux like variables.
8255 (arc_is_aux_reg_p): New function.
8256 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8257 (arc_get_aux_arg): New function.
8258 (prepare_move_operands): Handle aux-register access.
8259 (arc_handle_aux_attribute): New function.
8260 * doc/extend.texi (ARC Variable attributes): Add subsection.
8262 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8264 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8265 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8266 (arc_attribute_table): Add 'uncached' attribute.
8267 (arc_print_operand): Print '.di' flag for uncached memory
8269 (arc_in_small_data_p): Do not consider for small data the uncached
8271 (arc_is_uncached_mem_p): New function.
8272 * config/arc/predicates.md (compact_store_memory_operand): Check
8273 for uncached memory accesses.
8274 (nonvol_nonimm_operand): Likewise.
8275 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8277 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8280 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8281 falign-loops=): Add Optimization flag.
8283 2018-01-30 Jeff Law <law@redhat.com>
8286 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8287 INT_REGISTERS_SAVED. Check it prior to calling
8288 get_scratch_register_on_entry.
8289 (ix86_adjust_stack_and_probe): Similarly.
8290 (ix86_emit_probe_stack_range): Similarly.
8291 (ix86_expand_prologue): Corresponding changes.
8293 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8296 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8297 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8299 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8302 * lra-constraints.c (curr_insn_transform): Process AND in the
8305 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8307 PR rtl-optimization/83986
8308 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8309 dependence against last_pending_memory_flush in addition to
8312 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8314 PR tree-optimization/81611
8315 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8318 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8321 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8324 2018-01-30 Richard Biener <rguenther@suse.de>
8325 Jakub Jelinek <jakub@redhat.com>
8327 PR tree-optimization/84111
8328 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8329 inner loops added during recursion, as they don't have up-to-date
8332 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8335 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8336 (can_inline_edge_by_limits_p): ... here.
8337 (can_early_inline_edge_p, check_callers,
8338 update_caller_keys, update_callee_keys, recursive_inlining,
8339 add_new_edges_to_heap, speculation_useful_p,
8340 inline_small_functions,
8341 inline_small_functions, flatten_function,
8342 inline_to_all_callers_1): Update.
8344 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8346 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8349 2018-01-30 Richard Biener <rguenther@suse.de>
8351 PR tree-optimization/83008
8352 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8353 invariant and constant vector uses in stmts when they need
8356 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8359 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8360 * configure: Regenerate.
8362 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8364 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8366 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8367 Use gen_rtx_REG rather than gen_lowpart.
8369 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8371 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8372 rather than 0 when creating partial subregs.
8374 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8376 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8379 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8382 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8383 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8384 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8385 flags. This restores the settings used before the 2017-07-24.
8386 Turning off pre increment/decrement/modify allows IVOPTS to
8387 optimize DF/SF loops where the index is an int.
8389 2018-01-29 Richard Biener <rguenther@suse.de>
8390 Kelvin Nilsen <kelvin@gcc.gnu.org>
8393 * tree-vect-stmts.c (vectorizable_call): Don't call
8394 targetm.vectorize_builtin_md_vectorized_function if callee is
8397 2018-01-22 Carl Love <cel@us.ibm.com>
8399 * doc/extend.tex: Fix typo in second arg in
8400 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8402 2018-01-29 Richard Biener <rguenther@suse.de>
8404 PR tree-optimization/84086
8405 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8406 (flush_ssaname_freelist): When SSA names were released reset
8407 the SCEV hash table.
8409 2018-01-29 Richard Biener <rguenther@suse.de>
8411 PR tree-optimization/84057
8412 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8413 removed paths when removing edges.
8415 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8417 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8418 -mfunction-return=@var{choice}.
8420 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8423 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8425 (layout::print_source_line): Likewise.
8426 (test_get_line_width_without_trailing_whitespace): Add test cases.
8428 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8431 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8434 2018-01-26 Jim Wilson <jimw@sifive.com>
8436 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8438 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8441 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8443 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
8444 and CMP + SUB-immediate -> SUBS.
8446 2018-01-26 Martin Sebor <msebor@redhat.com>
8448 PR tree-optimization/83896
8449 * tree-ssa-strlen.c (get_string_len): Rename...
8450 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
8451 Avoid assuming length is constant.
8452 (handle_char_store): Use HOST_WIDE_INT for string length.
8454 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
8457 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
8458 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
8460 2018-01-26 Richard Biener <rguenther@suse.de>
8462 PR rtl-optimization/84003
8463 * dse.c (record_store): Only record redundant stores when
8464 the earlier store aliases at least all accesses the later one does.
8466 2018-01-26 Jakub Jelinek <jakub@redhat.com>
8468 PR rtl-optimization/83985
8469 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
8470 REG_CFA_RESTORE insns.
8471 (delete_unmarked_insns): Don't ignore separate shrink wrapping
8472 REG_CFA_RESTORE insns here.
8475 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
8476 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
8478 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8480 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
8481 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
8482 (arc_init): Likewise.
8483 (arc_override_options): Likewise.
8484 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
8486 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
8488 * config/arc/arc.h (TARGET_DBNZ): Define.
8489 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
8490 properly set the tune attribute.
8491 (dbnz): Use TARGET_DBNZ guard.
8492 * config/arc/arc.opt (mtune): Add core3 option.
8494 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8496 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
8497 recognize new pic like addresses.
8498 (arc_delegitimize_address): Clean up.
8500 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8502 * config/arc/arc-arches.def: Option mrf16 valid for all
8504 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
8505 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
8506 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
8507 * config/arc/arc-tables.opt: Regenerate.
8508 * config/arc/arc.c (arc_conditional_register_usage): Handle
8509 reduced register file case.
8510 (arc_file_start): Set must have build attributes.
8511 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
8513 * config/arc/arc.opt (mrf16): Add new option.
8514 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
8515 * config/arc/genmultilib.awk: Handle new mrf16 option.
8516 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
8517 * config/arc/t-multilib: Regenerate.
8518 * doc/invoke.texi (ARC Options): Document mrf16 option.
8520 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8522 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
8523 * config/arc/arc.c (arc_handle_secure_attribute): New function.
8524 (arc_attribute_table): Add 'secure_call' attribute.
8525 (arc_print_operand): Print secure call operand.
8526 (arc_function_ok_for_sibcall): Don't optimize tail calls when
8528 (arc_is_secure_call_p): New function. * config/arc/arc.md
8529 (call_i): Add support for sjli instruction.
8530 (call_value_i): Likewise.
8531 * config/arc/constraints.md (Csc): New constraint.
8533 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
8534 John Eric Martin <John.Martin@emmicro-us.com>
8536 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
8537 * config/arc/arc.c (_arc_jli_section): New struct.
8538 (arc_jli_section): New type.
8539 (rc_jli_sections): New static variable.
8540 (arc_handle_jli_attribute): New function.
8541 (arc_attribute_table): Add jli_always and jli_fixed attribute.
8542 (arc_file_end): New function.
8543 (TARGET_ASM_FILE_END): Define.
8544 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
8545 (arc_add_jli_section): New function.
8546 (jli_call_scan): Likewise.
8547 (arc_reorg): Call jli_call_scan.
8548 (arc_output_addsi): Remove 'S' from printing asm operand.
8549 (arc_is_jli_call_p): New function.
8550 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
8552 (movhi_insn): Likewise.
8553 (movsi_insn): Likewise.
8554 (movsi_set_cc_insn): Likewise.
8555 (loadqi_update): Likewise.
8556 (load_zeroextendqisi_update): Likewise.
8557 (load_signextendqisi_update): Likewise.
8558 (loadhi_update): Likewise.
8559 (load_zeroextendhisi_update): Likewise.
8560 (load_signextendhisi_update): Likewise.
8561 (loadsi_update): Likewise.
8562 (loadsf_update): Likewise.
8563 (movsicc_insn): Likewise.
8564 (bset_insn): Likewise.
8565 (bxor_insn): Likewise.
8566 (bclr_insn): Likewise.
8567 (bmsk_insn): Likewise.
8568 (bicsi3_insn): Likewise.
8569 (cmpsi_cc_c_insn): Likewise.
8570 (movsi_ne): Likewise.
8571 (movsi_cond_exec): Likewise.
8572 (clrsbsi2): Likewise.
8583 (call_i): Remove 'S' asm letter, add jli instruction.
8584 (call_value_i): Likewise.
8585 * config/arc/arc.op (mjli-always): New option.
8586 * config/arc/constraints.md (Cji): New constraint.
8587 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
8589 (subsf3_fpx): Likewise.
8590 (mulsf3_fpx): Likewise.
8591 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
8593 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
8595 * doc/invoke.texi (ARC): Document mjli-always option.
8597 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
8599 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
8600 avoid addition with 0 and use incw and decw where possible.
8602 2018-01-26 Richard Biener <rguenther@suse.de>
8604 PR tree-optimization/81082
8605 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
8606 association if it requires casting to unsigned.
8607 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
8608 from fold_plusminus_mult_expr to catch important cases late when
8609 range info is available.
8611 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8613 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
8614 * configure.ac (hidden_linkonce): New test.
8615 * configure: Regenerate.
8616 * config.in: Regenerate.
8618 2018-01-26 Julia Koval <julia.koval@intel.com>
8620 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
8621 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
8622 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
8623 _mm_mask_bitshuffle_epi64_mask): Fix type.
8624 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8625 USI_FTYPE_V4DI_V4DI_USI): Remove.
8626 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8627 __builtin_ia32_vpshufbitqmb256_mask,
8628 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8629 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8630 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8632 2018-01-26 Alan Modra <amodra@gmail.com>
8635 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8636 UNSPEC_VBPERMQ. Sort other unspecs.
8638 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8640 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8642 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8645 * predict.c (drop_profile): Do not push/pop cfun; update also
8647 (handle_missing_profiles): Fix logic looking for zero profiles.
8649 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8652 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8653 on functions with #pragma omp declare simd or functions with simd
8655 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8656 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8657 Remove trailing \n from warning_at calls.
8659 2018-01-25 Tom de Vries <tom@codesourcery.com>
8662 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8663 for neutered workers.
8665 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8668 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8669 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8671 2018-01-24 Jeff Law <law@redhat.com>
8674 * i386.c (get_probe_interval): Move to earlier point.
8675 (ix86_compute_frame_layout): If -fstack-clash-protection and
8676 the frame is larger than the probe interval, then use pushes
8677 to save registers rather than reg->mem moves.
8678 (ix86_expand_prologue): Remove conditional for int_registers_saved
8681 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8684 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8685 min/max for never referenced object.
8687 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8690 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8692 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8693 attributes from DECL_ATTRIBUTES (decl) without affecting
8694 DECL_ATTRIBUTES (current_function_decl).
8695 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8696 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8698 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8700 PR tree-optimization/83979
8701 * fold-const.c (fold_comparison): Use constant_boolean_node
8702 instead of boolean_{true,false}_node.
8704 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8706 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8709 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8711 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8712 Simplify the clause that sets the length attribute.
8713 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8714 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8715 clause that sets the length attribute.
8716 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8718 2018-01-24 Tom de Vries <tom@codesourcery.com>
8721 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8722 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8723 Add strict parameter.
8724 (prevent_branch_around_nothing): Insert dummy insn between branch to
8725 label and label with no ptx insn inbetween.
8726 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8728 2018-01-24 Tom de Vries <tom@codesourcery.com>
8731 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8732 for neutered threads in warp.
8733 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8735 2018-01-24 Richard Biener <rguenther@suse.de>
8737 PR tree-optimization/83176
8738 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8741 2018-01-24 Richard Biener <rguenther@suse.de>
8743 PR tree-optimization/82819
8744 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8745 code generating pluses that are no-ops in the target precision.
8747 2018-01-24 Richard Biener <rguenther@suse.de>
8750 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8752 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8754 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8755 to merge probabilities.
8756 * predict.c (probably_never_executed): Also mark as cold functions
8757 with global 0 profile and guessed local profile.
8758 * profile-count.c (profile_probability::combine_with_count): New
8760 * profile-count.h (profile_probability::operator*,
8761 profile_probability::operator*=, profile_probability::operator/,
8762 profile_probability::operator/=): Reduce precision to adjusted
8763 and set value to guessed on contradictory divisions.
8764 (profile_probability::combine_with_freq): Remove.
8765 (profile_probability::combine_wiht_count): Declare.
8766 (profile_count::force_nonzero):: Set to adjusted.
8767 (profile_count::probability_in):: Set quality to adjusted.
8768 * tree-ssa-tail-merge.c (replace_block_by): Use
8771 2018-01-23 Andrew Waterman <andrew@sifive.com>
8772 Jim Wilson <jimw@sifive.com>
8774 * config/riscv/riscv.c (riscv_stack_boundary): New.
8775 (riscv_option_override): Set riscv_stack_boundary. Handle
8776 riscv_preferred_stack_boundary_arg.
8777 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8778 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8779 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8780 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8781 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8782 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8784 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8787 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8788 of struct ix86_frame.
8789 (ix86_expand_epilogue): Likewise. Add a local variable for
8790 the reg_save_offset field in struct ix86_frame.
8792 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8794 PR tree-optimization/82604
8795 * tree-loop-distribution.c (enum partition_kind): New enum item
8796 PKIND_PARTIAL_MEMSET.
8797 (partition_builtin_p): Support above new enum item.
8798 (generate_code_for_partition): Ditto.
8799 (compute_access_range): Differentiate cases that equality can be
8800 proven at all loops, the innermost loops or no loops.
8801 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8802 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8803 (finalize_partitions, distribute_loop): Don't fuse partition of
8804 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8805 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8808 2018-01-23 Martin Liska <mliska@suse.cz>
8810 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8811 order to ignore the predictor.
8812 (PRED_POLYMORPHIC_CALL): Likewise.
8813 (PRED_RECURSIVE_CALL): Likewise.
8815 2018-01-23 Martin Liska <mliska@suse.cz>
8817 * tree-profile.c (tree_profiling): Print function header to
8818 aware reader which function we are working on.
8819 * value-prof.c (gimple_find_values_to_profile): Do not print
8820 not interesting value histograms.
8822 2018-01-23 Martin Liska <mliska@suse.cz>
8824 * profile-count.h (enum profile_quality): Add
8825 profile_uninitialized as the first value. Do not number values
8826 as they are zero based.
8827 (profile_count::verify): Update sanity check.
8828 (profile_probability::verify): Likewise.
8830 2018-01-23 Nathan Sidwell <nathan@acm.org>
8832 * doc/invoke.texi (ffor-scope): Deprecate.
8834 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8836 PR tree-optimization/83510
8837 * domwalk.c (set_all_edges_as_executable): New function.
8838 (dom_walker::dom_walker): Convert bool param
8839 "skip_unreachable_blocks" to enum reachability. Move setup of
8840 edge flags to set_all_edges_as_executable and only do it when
8841 reachability is REACHABLE_BLOCKS.
8842 * domwalk.h (enum dom_walker::reachability): New enum.
8843 (dom_walker::dom_walker): Convert bool param
8844 "skip_unreachable_blocks" to enum reachability.
8845 (set_all_edges_as_executable): New decl.
8846 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8847 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8849 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8850 but converting true to REACHABLE_BLOCKS.
8851 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8853 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8854 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8855 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8857 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8858 if check_all_array_refs will be called.
8860 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8862 * tree.c (selftest::test_location_wrappers): Add more test
8865 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8867 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8868 (selftest::test_bit_in_range): Likewise.
8870 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8873 * doc/sourcebuild.texi (vect_float): Say that the selector
8874 only describes the situation when -funsafe-math-optimizations is on.
8875 (vect_float_strict): Document.
8877 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8879 PR tree-optimization/83965
8880 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8881 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8882 instead of checking only for a reduction.
8883 (vect_recog_widen_sum_pattern): Likewise.
8885 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8887 * predict.c (probably_never_executed): Only use precise profile info.
8888 (compute_function_frequency): Skip after inlining hack since we now
8889 have quality checking.
8891 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8893 * profile-count.h (profile_probability::very_unlikely,
8894 profile_probability::unlikely, profile_probability::even): Set
8895 precision to guessed.
8897 2018-01-23 Richard Biener <rguenther@suse.de>
8899 PR tree-optimization/83963
8900 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8901 Properly terminate dominator walk when crossing the exit edge not
8902 when visiting its source block.
8904 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8907 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8908 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8910 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8912 PR tree-optimization/83957
8913 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8914 semicolon after for body surrounded by braces.
8916 PR tree-optimization/83081
8917 * profile-count.h (profile_probability::split): New method.
8918 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8919 Use profile_probability::split.
8920 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8921 when splitting a single conditional jump into 2.
8923 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8925 PR tree-optimization/69452
8926 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8929 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8931 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8932 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8933 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8935 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8937 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8938 * config/rl78/rl78.md (movdi): New define_expand.
8939 * config/rl78/rl78.c (rl78_split_movdi): New function.
8941 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8944 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8946 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8947 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8948 128-bit to produce an UNSPEC move to get the double word with the
8949 signbit and then a shift directly to do signbit.
8950 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8951 implementation with a new version that just does either a direct
8952 move or a regular move. Move memory interface to separate insns.
8953 Move insns so they are next to the expander.
8954 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8955 with signbit move. Split big and little endian case.
8956 (signbit<mode>2_dm_mem_le): Likewise.
8957 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8958 (signbit<mode>2_dm2): Likewise.
8960 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8962 * config/rl78/rl78.md (anddi3): New define_expand.
8964 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8966 * config/rl78/rl78.md (umindi3): New define_expand.
8968 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8970 * config/rl78/rl78.md (smindi3): New define_expand.
8972 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8974 * config/rl78/rl78.md (smaxdi3): New define_expand.
8976 2018-01-22 Carl Love <cel@us.ibm.com>
8978 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8979 LVX_V1TI): Add macro expansion.
8980 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8981 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8982 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8983 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8984 Change check to determine if the instruction is a byte reversing
8985 entry. Fix typo in comment.
8986 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8987 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8988 Add def_builtin calls for new builtins.
8989 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8990 Add define_insn expansion.
8992 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8994 * config/rl78/rl78.md (umaxdi3): New define_expand.
8996 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8998 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8999 for non-QImode registers.
9001 2018-01-22 Richard Biener <rguenther@suse.de>
9003 PR tree-optimization/83963
9004 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9005 including the loop exit block.
9006 (scop_detection::merge_sese): Likewise.
9007 (scop_detection::add_scop): Do it here instead.
9009 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9011 * doc/sourcebuild.texi (arm_softfloat): Document.
9013 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9016 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9017 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9018 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9020 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9021 David Edelsohn <dje.gcc@gmail.com>
9024 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9025 Change "crset eq" to "crset 2".
9026 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9027 (*call_indirect_aix<mode>_nospec): Likewise.
9028 (*call_value_indirect_aix<mode>_nospec): Likewise.
9029 (*call_indirect_elfv2<mode>_nospec): Likewise.
9030 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9031 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9032 change assembly output from . to $.
9033 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9034 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9035 (*tablejump<mode>_internal1_nospec): Likewise.
9037 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9040 * config/sh/sh_optimize_sett_clrt.cc:
9041 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9043 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9045 PR tree-optimization/83940
9046 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9047 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9048 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9049 use it to pass back the definition type.
9050 (vect_check_store_rhs): Likewise rhs_dt_out.
9051 (vect_build_gather_load_calls): Add a mask_dt argument and use
9052 it instead of a call to vect_is_simple_use.
9053 (vectorizable_store): Update calls to vect_check_load_store_mask
9054 and vect_check_store_rhs. Use the dt returned by the latter instead
9055 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9056 instead of calls to vect_is_simple_use. Pass the scalar rather
9057 than the vector operand to vect_is_simple_use when handling
9058 second and subsequent copies of an rhs value.
9059 (vectorizable_load): Update calls to vect_check_load_store_mask
9060 and vect_build_gather_load_calls. Use the cached mask_dt and
9061 gs_info.offset_dt instead of calls to vect_is_simple_use.
9063 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9066 * tree-emutls.c: Include gimplify.h.
9067 (lower_emutls_2): New function.
9068 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9069 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9070 it before further processing.
9073 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9074 UINTVAL (trueop1) instead of INTVAL (op1).
9076 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9080 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9081 INCOMING_FRAME_SP_OFFSET if not defined.
9082 (scan_trace): Add ENTRY argument. If true and
9083 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9084 emit a note to adjust the CFA offset.
9085 (create_cfi_notes): Adjust scan_trace callers.
9086 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9087 INCOMING_FRAME_SP_OFFSET in the CIE.
9088 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9089 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9091 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9092 * doc/tm.texi: Regenerated.
9094 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9096 PR rtl-optimization/83147
9097 * lra-constraints.c (remove_inheritance_pseudos): Use
9098 lra_substitute_pseudo_within_insn.
9100 2018-01-19 Tom de Vries <tom@codesourcery.com>
9101 Cesar Philippidis <cesar@codesourcery.com>
9104 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9106 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9109 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9110 spaces for function labels.
9112 2018-01-19 Martin Liska <mliska@suse.cz>
9114 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9115 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9116 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9117 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9118 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9119 (PRED_CONST_RETURN): Change from 69 to 65.
9120 (PRED_NULL_RETURN): Change from 91 to 71.
9121 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9122 (PRED_LOOP_GUARD): Change from 66 to 73.
9124 2018-01-19 Martin Liska <mliska@suse.cz>
9126 * predict.c (predict_insn_def): Add new assert.
9127 (struct branch_predictor): Change type to signed integer.
9128 (test_prediction_value_range): Amend test to cover
9130 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9131 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9132 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9133 (PRED_LOOP_IV_COMPARE): Likewise.
9134 * predict.h (PROB_UNINITIALIZED): Define new constant.
9136 2018-01-19 Martin Liska <mliska@suse.cz>
9138 * predict.c (dump_prediction): Add new format for
9139 analyze_brprob.py script which is enabled with -details
9141 * profile-count.h (precise_p): New function.
9143 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9145 PR tree-optimization/83922
9146 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9147 there are no statements that need masking.
9148 (vect_active_double_reduction_p): New function.
9149 (vect_analyze_loop_operations): Use it when handling phis that
9150 are not in the loop header.
9152 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9154 PR tree-optimization/83914
9155 * tree-vect-loop.c (vectorizable_induction): Don't convert
9156 init_expr or apply the peeling adjustment for inductions
9157 that are nested within the vectorized loop.
9159 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9161 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9164 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9168 * function.h (gimplify_parameters): Add gimple_seq * argument.
9169 * function.c: Include gimple.h and options.h.
9170 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9171 for the added local temporaries if needed.
9172 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9173 if there are any parameter cleanups, wrap whole body into a
9174 try/finally with the cleanups.
9176 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9179 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9180 Use GET_MODE_CLASS for scalar floating point.
9182 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9186 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9187 Fix call of call_cgraph_insertion_hooks.
9189 2018-01-18 Martin Sebor <msebor@redhat.com>
9191 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9193 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9196 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9199 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9202 * common.opt: (-ffile-prefix-map): New option.
9203 * opts.c (common_handle_option): Defer it.
9204 * opts-global.c (handle_common_deferred_options): Handle it.
9205 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9206 * file-prefix-map.h: New file.
9207 (remap_debug_filename, add_debug_prefix_map): ...here.
9208 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9209 * final.c (debug_prefix_map, add_debug_prefix_map
9210 remap_debug_filename): Move to...
9211 * file-prefix-map.c: New file.
9212 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9213 generalize, get rid of alloca(), use strrchr() instead of strchr().
9214 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9215 Implement in terms of add_prefix_map().
9216 (remap_macro_filename, remap_debug_filename): Implement in term of
9218 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9219 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9220 * dbxout.c: Include file-prefix-map.h.
9221 * varasm.c: Likewise.
9222 * vmsdbgout.c: Likewise.
9223 * xcoffout.c: Likewise.
9224 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9225 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9226 * doc/invoke.texi (-ffile-prefix-map): Document.
9227 (-fdebug-prefix-map): Update description.
9229 2018-01-18 Martin Liska <mliska@suse.cz>
9231 * config/i386/i386.c (indirect_thunk_name): Document that also
9233 (output_indirect_thunk): Document why both instructions
9234 (pause and lfence) are generated.
9236 2018-01-18 Richard Biener <rguenther@suse.de>
9238 PR tree-optimization/83887
9239 * graphite-scop-detection.c
9240 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9241 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9242 (scop_detection::merge_sese): Re-implement with a flood-fill
9243 algorithm that properly finds a SESE region if it exists.
9245 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9248 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9249 pointer_diff optimizations use view_convert instead of convert.
9251 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9253 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9254 Generate different code for -mno-speculate-indirect-jumps.
9255 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9256 (*call_indirect_aix<mode>): Disable for
9257 -mno-speculate-indirect-jumps.
9258 (*call_indirect_aix<mode>_nospec): New define_insn.
9259 (*call_value_indirect_aix<mode>): Disable for
9260 -mno-speculate-indirect-jumps.
9261 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9262 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9263 -mno-speculate-indirect-jumps.
9264 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9266 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9268 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9269 long double type, set the flags for noting the default long double
9270 type, even if we don't pass or return a long double type.
9272 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9275 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9278 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9280 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9281 support for merge[hl].
9282 (fold_mergehl_helper): New helper function.
9283 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9284 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9285 (altivec_vmrglw_direct): Add xxmrglw insn.
9287 2018-01-17 Andrew Waterman <andrew@sifive.com>
9289 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9290 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9292 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9295 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9296 call the lto_location_cache before reading the
9297 DECL_SOURCE_LOCATION of the types.
9299 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9300 Richard Sandiford <richard.sandiford@linaro.org>
9302 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9303 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9304 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9305 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9306 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9308 * config/aarch64/constraints.md (aarch64_movti_operand):
9310 * config/aarch64/predicates.md (Uti): Add new constraint.
9312 2018-01-17 Carl Love <cel@us.ibm.com>
9314 * config/rs6000/vsx.md (define_expand xl_len_r,
9315 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9316 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9318 (define_expand, define_insn): Move the shift left from the
9319 define_insn to the define_expand for lxvl and stxvl instructions.
9320 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9321 and XL_LEN_R definitions to PURE.
9323 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9325 * config/i386/i386.c (indirect_thunk_name): Declare regno
9326 as unsigned int. Compare regno with INVALID_REGNUM.
9327 (output_indirect_thunk): Ditto.
9328 (output_indirect_thunk_function): Ditto.
9329 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9330 in the call to output_indirect_thunk_function.
9332 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9335 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9336 rather than the size of inner_type to determine the stack slot size
9337 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9339 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9342 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9345 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9347 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9348 endian Linux systems to optionally enable multilibs for selecting
9349 the long double type if the user configured an explicit type.
9350 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9351 have no long double multilibs if not defined.
9352 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9353 warn if the user used -mabi={ieee,ibm}longdouble and we built
9354 multilibs for long double.
9355 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9356 appropriate multilib option.
9357 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9359 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9360 for building long double multilibs.
9361 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9363 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9365 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9368 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9370 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9373 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9376 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9379 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9381 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9382 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9384 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9386 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9387 different rtl trees depending on TARGET_64BIT.
9388 (rs6000_gen_lvx): Likewise.
9390 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9392 * config/visium/visium.md (nop): Tweak comment.
9393 (hazard_nop): Likewise.
9395 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9397 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9398 -mspeculate-indirect-jumps.
9399 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9400 for -mno-speculate-indirect-jumps.
9401 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9402 (*call_value_indirect_elfv2<mode>): Disable for
9403 -mno-speculate-indirect-jumps.
9404 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9405 (indirect_jump): Emit different RTL for
9406 -mno-speculate-indirect-jumps.
9407 (*indirect_jump<mode>): Disable for
9408 -mno-speculate-indirect-jumps.
9409 (*indirect_jump<mode>_nospec): New define_insn.
9410 (tablejump): Emit different RTL for
9411 -mno-speculate-indirect-jumps.
9412 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9413 (tablejumpsi_nospec): New define_expand.
9414 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9415 (tablejumpdi_nospec): New define_expand.
9416 (*tablejump<mode>_internal1): Disable for
9417 -mno-speculate-indirect-jumps.
9418 (*tablejump<mode>_internal1_nospec): New define_insn.
9419 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9422 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9424 * caller-save.c (insert_save): Drop unnecessary parameter. All
9427 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9428 Richard Biener <rguenth@suse.de>
9431 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9432 return early, inline manually is_gimple_sizepos. Make sure if we
9433 call gimplify_expr we don't end up with a gimple constant.
9434 * tree.c (variably_modified_type_p): Don't return true for
9435 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9436 * gimplify.h (is_gimple_sizepos): Remove.
9438 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9440 PR tree-optimization/83857
9441 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
9442 vectorizable_live_operation for pure SLP statements.
9443 (vectorizable_live_operation): Handle PHIs.
9445 2018-01-16 Richard Biener <rguenther@suse.de>
9447 PR tree-optimization/83867
9448 * tree-vect-stmts.c (vect_transform_stmt): Precompute
9449 nested_in_vect_loop_p since the scalar stmt may get invalidated.
9451 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9454 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
9455 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
9456 If off is not INTEGER_CST, issue a may not be aligned warning
9457 rather than isn't aligned. Use isn%'t rather than isn't.
9458 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
9460 <case MULT_EXPR>: Improve the case when bottom and one of the
9461 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
9462 operand, in that case check if the other operand is multiple of
9463 bottom divided by the INTEGER_CST operand.
9465 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9468 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
9469 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
9470 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
9471 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
9472 * config/pa/pa.c (pa_function_arg_advance): Likewise.
9473 (pa_function_arg, pa_arg_partial_bytes): Likewise.
9474 (pa_function_arg_size): New function.
9476 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9478 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
9479 in a separate statement.
9481 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9483 PR tree-optimization/83847
9484 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
9485 group gathers and scatters.
9487 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9489 PR rtl-optimization/86620
9490 * params.def (max-sched-ready-insns): Bump minimum value to 1.
9492 PR rtl-optimization/83213
9493 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
9494 to last if both are JUMP_INSNs.
9496 PR tree-optimization/83843
9497 * gimple-ssa-store-merging.c
9498 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
9499 store_immediate_info for bswap/nop orig_stores.
9501 2018-01-15 Andrew Waterman <andrew@sifive.com>
9503 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
9505 <UDIV>: Increase cost if !TARGET_DIV.
9507 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
9509 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
9510 (define_attr "cr_logical_3op"): New.
9511 (cceq_ior_compare): Adjust.
9512 (cceq_ior_compare_complement): Adjust.
9513 (*cceq_rev_compare): Adjust.
9514 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
9515 (is_cracked_insn): Adjust.
9516 (insn_must_be_first_in_group): Adjust.
9517 * config/rs6000/40x.md: Adjust.
9518 * config/rs6000/440.md: Adjust.
9519 * config/rs6000/476.md: Adjust.
9520 * config/rs6000/601.md: Adjust.
9521 * config/rs6000/603.md: Adjust.
9522 * config/rs6000/6xx.md: Adjust.
9523 * config/rs6000/7450.md: Adjust.
9524 * config/rs6000/7xx.md: Adjust.
9525 * config/rs6000/8540.md: Adjust.
9526 * config/rs6000/cell.md: Adjust.
9527 * config/rs6000/e300c2c3.md: Adjust.
9528 * config/rs6000/e500mc.md: Adjust.
9529 * config/rs6000/e500mc64.md: Adjust.
9530 * config/rs6000/e5500.md: Adjust.
9531 * config/rs6000/e6500.md: Adjust.
9532 * config/rs6000/mpc.md: Adjust.
9533 * config/rs6000/power4.md: Adjust.
9534 * config/rs6000/power5.md: Adjust.
9535 * config/rs6000/power6.md: Adjust.
9536 * config/rs6000/power7.md: Adjust.
9537 * config/rs6000/power8.md: Adjust.
9538 * config/rs6000/power9.md: Adjust.
9539 * config/rs6000/rs64.md: Adjust.
9540 * config/rs6000/titan.md: Adjust.
9542 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9544 * config/i386/predicates.md (indirect_branch_operand): Rewrite
9545 ix86_indirect_branch_register logic.
9547 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9549 * config/i386/constraints.md (Bs): Update
9550 ix86_indirect_branch_register check. Don't check
9551 ix86_indirect_branch_register with GOT_memory_operand.
9553 * config/i386/predicates.md (GOT_memory_operand): Don't check
9554 ix86_indirect_branch_register here.
9555 (GOT32_symbol_operand): Likewise.
9557 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9559 * config/i386/predicates.md (constant_call_address_operand):
9560 Rewrite ix86_indirect_branch_register logic.
9561 (sibcall_insn_operand): Likewise.
9563 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9565 * config/i386/constraints.md (Bs): Replace
9566 ix86_indirect_branch_thunk_register with
9567 ix86_indirect_branch_register.
9569 * config/i386/i386.md (indirect_jump): Likewise.
9570 (tablejump): Likewise.
9571 (*sibcall_memory): Likewise.
9572 (*sibcall_value_memory): Likewise.
9573 Peepholes of indirect call and jump via memory: Likewise.
9574 * config/i386/i386.opt: Likewise.
9575 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9576 (GOT_memory_operand): Likewise.
9577 (call_insn_operand): Likewise.
9578 (sibcall_insn_operand): Likewise.
9579 (GOT32_symbol_operand): Likewise.
9581 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9584 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
9585 type rather than type addr's type points to.
9586 (expand_omp_atomic_mutex): Likewise.
9587 (expand_omp_atomic): Likewise.
9589 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
9592 * config/i386/i386.c (output_indirect_thunk_function): Use
9593 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
9594 for __x86_return_thunk.
9596 2018-01-15 Richard Biener <rguenther@suse.de>
9599 * expmed.c (extract_bit_field_1): Fix typo.
9601 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9604 * config/arm/iterators.md (VF): New mode iterator.
9605 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
9606 Remove integer-related logic from pattern.
9607 (neon_vabd<mode>_3): Likewise.
9609 2018-01-15 Jakub Jelinek <jakub@redhat.com>
9612 * common.opt (fstrict-overflow): No longer an alias.
9613 (fwrapv-pointer): New option.
9614 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
9615 also for pointer types based on flag_wrapv_pointer.
9616 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
9617 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
9618 opts->x_flag_wrapv got set.
9619 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
9620 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
9621 POINTER_TYPE_OVERFLOW_UNDEFINED.
9622 * match.pd: Likewise in address comparison pattern.
9623 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9625 2018-01-15 Richard Biener <rguenther@suse.de>
9628 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9629 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9630 Reset type names to their identifier if their TYPE_DECL doesn't
9631 have linkage (and thus is used for ODR and devirt).
9632 (save_debug_info_for_decl): Remove.
9633 (save_debug_info_for_type): Likewise.
9634 (add_tree_to_fld_list): Adjust.
9635 * tree-pretty-print.c (dump_generic_node): Make dumping of
9636 type names more robust.
9638 2018-01-15 Richard Biener <rguenther@suse.de>
9640 * BASE-VER: Bump to 8.0.1.
9642 2018-01-14 Martin Sebor <msebor@redhat.com>
9645 * builtins.c (check_access): Avoid warning when the no-warning bit
9648 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9650 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9651 * ira-color (allocno_hard_regs_compare): Likewise.
9653 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9656 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9657 Use .pushsection/.popsection.
9659 2018-01-14 Martin Sebor <msebor@redhat.com>
9662 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9664 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9666 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9667 entry from extra_headers.
9668 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9669 extra_headers, make the list bitwise identical to the i?86-*-* one.
9671 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9673 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9674 -mcmodel=large with -mindirect-branch=thunk,
9675 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9676 -mfunction-return=thunk-extern.
9677 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9678 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9679 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9681 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9683 * config/i386/i386.c (print_reg): Print the name of the full
9684 integer register without '%'.
9685 (ix86_print_operand): Handle 'V'.
9686 * doc/extend.texi: Document 'V' modifier.
9688 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9690 * config/i386/constraints.md (Bs): Disallow memory operand for
9691 -mindirect-branch-register.
9693 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9694 (GOT_memory_operand): Likewise.
9695 (call_insn_operand): Likewise.
9696 (sibcall_insn_operand): Likewise.
9697 (GOT32_symbol_operand): Likewise.
9698 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9699 for -mindirect-branch-register.
9700 (tablejump): Likewise.
9701 (*sibcall_memory): Likewise.
9702 (*sibcall_value_memory): Likewise.
9703 Disallow peepholes of indirect call and jump via memory for
9704 -mindirect-branch-register.
9705 (*call_pop): Replace m with Bw.
9706 (*call_value_pop): Likewise.
9707 (*sibcall_pop_memory): Replace m with Bs.
9708 * config/i386/i386.opt (mindirect-branch-register): New option.
9709 * doc/invoke.texi: Document -mindirect-branch-register option.
9711 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9713 * config/i386/i386-protos.h (ix86_output_function_return): New.
9714 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9715 set function_return_type.
9716 (indirect_thunk_name): Add ret_p to indicate thunk for function
9718 (output_indirect_thunk_function): Pass false to
9719 indirect_thunk_name.
9720 (ix86_output_indirect_branch_via_reg): Likewise.
9721 (ix86_output_indirect_branch_via_push): Likewise.
9722 (output_indirect_thunk_function): Create alias for function
9723 return thunk if regno < 0.
9724 (ix86_output_function_return): New function.
9725 (ix86_handle_fndecl_attribute): Handle function_return.
9726 (ix86_attribute_table): Add function_return.
9727 * config/i386/i386.h (machine_function): Add
9728 function_return_type.
9729 * config/i386/i386.md (simple_return_internal): Use
9730 ix86_output_function_return.
9731 (simple_return_internal_long): Likewise.
9732 * config/i386/i386.opt (mfunction-return=): New option.
9733 (indirect_branch): Mention -mfunction-return=.
9734 * doc/extend.texi: Document function_return function attribute.
9735 * doc/invoke.texi: Document -mfunction-return= option.
9737 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9739 * config/i386/i386-opts.h (indirect_branch): New.
9740 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9741 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9742 with local indirect jump when converting indirect call and jump.
9743 (ix86_set_indirect_branch_type): New.
9744 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9745 (indirectlabelno): New.
9746 (indirect_thunk_needed): Likewise.
9747 (indirect_thunk_bnd_needed): Likewise.
9748 (indirect_thunks_used): Likewise.
9749 (indirect_thunks_bnd_used): Likewise.
9750 (INDIRECT_LABEL): Likewise.
9751 (indirect_thunk_name): Likewise.
9752 (output_indirect_thunk): Likewise.
9753 (output_indirect_thunk_function): Likewise.
9754 (ix86_output_indirect_branch_via_reg): Likewise.
9755 (ix86_output_indirect_branch_via_push): Likewise.
9756 (ix86_output_indirect_branch): Likewise.
9757 (ix86_output_indirect_jmp): Likewise.
9758 (ix86_code_end): Call output_indirect_thunk_function if needed.
9759 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9761 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9762 (ix86_attribute_table): Add indirect_branch.
9763 * config/i386/i386.h (machine_function): Add indirect_branch_type
9764 and has_local_indirect_jump.
9765 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9767 (tablejump): Likewise.
9768 (*indirect_jump): Use ix86_output_indirect_jmp.
9769 (*tablejump_1): Likewise.
9770 (simple_return_indirect_internal): Likewise.
9771 * config/i386/i386.opt (mindirect-branch=): New option.
9772 (indirect_branch): New.
9775 (thunk-inline): Likewise.
9776 (thunk-extern): Likewise.
9777 * doc/extend.texi: Document indirect_branch function attribute.
9778 * doc/invoke.texi: Document -mindirect-branch= option.
9780 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9783 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9785 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9787 * ipa-inline.c (want_inline_small_function_p): Return false if
9788 inlining has already failed with CIF_FINAL_ERROR.
9789 (update_caller_keys): Call want_inline_small_function_p before
9791 (update_callee_keys): Likewise.
9793 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9795 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9797 (rs6000_quadword_masked_address_p): Likewise.
9798 (quad_aligned_load_p): Likewise.
9799 (quad_aligned_store_p): Likewise.
9800 (const_load_sequence_p): Add comment to describe the outer-most loop.
9801 (mimic_memory_attributes_and_flags): New function.
9802 (rs6000_gen_stvx): Likewise.
9803 (replace_swapped_aligned_store): Likewise.
9804 (rs6000_gen_lvx): Likewise.
9805 (replace_swapped_aligned_load): Likewise.
9806 (replace_swapped_load_constant): Capitalize argument name in
9807 comment describing this function.
9808 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9809 and stores that access quad-word aligned addresses and replace
9810 with stvx or lvx instructions when appropriate.
9811 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9812 New function prototype.
9813 (rs6000_quadword_masked_address_p): Likewise.
9814 (rs6000_gen_lvx): Likewise.
9815 (rs6000_gen_stvx): Likewise.
9816 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9817 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9818 when memory address is aligned.
9819 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9820 this split to select lvx instruction when memory address is aligned.
9821 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9822 instruction when memory address is aligned.
9823 (*vsx_le_perm_load_v16qi): Likewise.
9824 (four unnamed splitters): Modify to select the stvx instruction
9825 when memory is aligned.
9827 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9829 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9830 which appears in the queue multiple times.
9832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9833 Alan Hayward <alan.hayward@arm.com>
9834 David Sherwood <david.sherwood@arm.com>
9836 * tree-vectorizer.h (vec_lower_bound): New structure.
9837 (_loop_vec_info): Add check_nonzero and lower_bounds.
9838 (LOOP_VINFO_CHECK_NONZERO): New macro.
9839 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9840 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9841 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9842 fields. Make seg_len the distance travelled, not including the
9844 (dr_direction_indicator): Declare.
9845 (dr_zero_step_indicator): Likewise.
9846 (dr_known_forward_stride_p): Likewise.
9847 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9849 (runtime_alias_check_p): Allow runtime alias checks with
9851 (operator ==): Compare access_size and align.
9852 (prune_runtime_alias_test_list): Rework for new distinction between
9853 the access_size and seg_len.
9854 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9856 (get_segment_min_max): New function.
9857 (create_intersect_range_checks): Use it.
9858 (dr_step_indicator): New function.
9859 (dr_direction_indicator): Likewise.
9860 (dr_zero_step_indicator): Likewise.
9861 (dr_known_forward_stride_p): Likewise.
9862 * tree-loop-distribution.c (data_ref_segment_size): Return
9863 DR_STEP * (niters - 1).
9864 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9866 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9867 (vect_preserves_scalar_order_p): New function, split out from...
9868 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9869 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9870 (vect_vfa_access_size): New function.
9871 (vect_vfa_align): Likewise.
9872 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9873 (dump_lower_bound): New function.
9874 (vect_check_lower_bound): Likewise.
9875 (vect_small_gap_p): Likewise.
9876 (vectorizable_with_step_bound_p): Likewise.
9877 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9878 depencies if the vectorization factor is 1. Convert the checks
9879 for nonzero steps into checks on the bounds of DR_STEP. Try using
9880 a bunds check for variable steps if the minimum required step is
9881 relatively small. Update calls to the dr_with_seg_len
9882 constructor and to vect_compile_time_alias.
9883 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9885 (vect_loop_versioning): Call it.
9886 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9888 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9891 Alan Hayward <alan.hayward@arm.com>
9892 David Sherwood <david.sherwood@arm.com>
9894 * doc/sourcebuild.texi (vect_scatter_store): Document.
9895 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9897 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9899 * genopinit.c (main): Add supports_vec_scatter_store and
9900 supports_vec_scatter_store_cached to target_optabs.
9901 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9902 IFN_MASK_SCATTER_STORE.
9903 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9905 * internal-fn.h (internal_store_fn_p): Declare.
9906 (internal_fn_stored_value_index): Likewise.
9907 * internal-fn.c (scatter_store_direct): New macro.
9908 (expand_scatter_store_optab_fn): New function.
9909 (direct_scatter_store_optab_supported_p): New macro.
9910 (internal_store_fn_p): New function.
9911 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9912 IFN_MASK_SCATTER_STORE.
9913 (internal_fn_mask_index): Likewise.
9914 (internal_fn_stored_value_index): New function.
9915 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9917 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9918 * optabs-query.c (supports_vec_scatter_store_p): New function.
9919 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9920 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9921 true for scatter stores.
9922 (vect_gather_scatter_fn_p): Handle scatter stores too.
9923 (vect_check_gather_scatter): Consider using scatter stores if
9924 supports_vec_scatter_store_p.
9925 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9927 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9928 internal_fn_stored_value_index.
9929 (check_load_store_masking): Handle scatter stores too.
9930 (vect_get_store_rhs): Make public.
9931 (vectorizable_call): Use internal_store_fn_p.
9932 (vectorizable_store): Handle scatter store internal functions.
9933 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9934 when deciding whether the end of the group has been reached.
9935 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9936 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9937 (mask_scatter_store<mode>): New insns.
9939 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9940 Alan Hayward <alan.hayward@arm.com>
9941 David Sherwood <david.sherwood@arm.com>
9943 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9944 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9945 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9947 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9948 Use vect_truncate_gather_scatter_offset if we can't treat the
9949 operation as a normal gather load or scatter store.
9950 (get_group_load_store_type): Take the gather_scatter_info
9951 as argument. Try using a gather load or scatter store for
9952 single-element groups.
9953 (get_load_store_type): Update calls to get_group_load_store_type
9954 and vect_use_strided_gather_scatters_p.
9956 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9957 Alan Hayward <alan.hayward@arm.com>
9958 David Sherwood <david.sherwood@arm.com>
9960 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9961 optional tree argument.
9962 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9964 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9965 but continue to use the current value as a fallback.
9966 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9967 to compare the updates.
9968 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9969 (get_load_store_type): Use it when handling a strided access.
9970 (vect_get_strided_load_store_ops): New function.
9971 (vect_get_data_ptr_increment): Likewise.
9972 (vectorizable_load): Handle strided gather loads. Always pass
9973 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9975 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9976 Alan Hayward <alan.hayward@arm.com>
9977 David Sherwood <david.sherwood@arm.com>
9979 * doc/md.texi (gather_load@var{m}): Document.
9980 (mask_gather_load@var{m}): Likewise.
9981 * genopinit.c (main): Add supports_vec_gather_load and
9982 supports_vec_gather_load_cached to target_optabs.
9983 * optabs-tree.c (init_tree_optimization_optabs): Use
9984 ggc_cleared_alloc to allocate target_optabs.
9985 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9986 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9988 * internal-fn.h (internal_load_fn_p): Declare.
9989 (internal_gather_scatter_fn_p): Likewise.
9990 (internal_fn_mask_index): Likewise.
9991 (internal_gather_scatter_fn_supported_p): Likewise.
9992 * internal-fn.c (gather_load_direct): New macro.
9993 (expand_gather_load_optab_fn): New function.
9994 (direct_gather_load_optab_supported_p): New macro.
9995 (direct_internal_fn_optab): New function.
9996 (internal_load_fn_p): Likewise.
9997 (internal_gather_scatter_fn_p): Likewise.
9998 (internal_fn_mask_index): Likewise.
9999 (internal_gather_scatter_fn_supported_p): Likewise.
10000 * optabs-query.c (supports_at_least_one_mode_p): New function.
10001 (supports_vec_gather_load_p): Likewise.
10002 * optabs-query.h (supports_vec_gather_load_p): Declare.
10003 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10004 and memory_type field.
10005 (NUM_PATTERNS): Bump to 15.
10006 * tree-vect-data-refs.c: Include internal-fn.h.
10007 (vect_gather_scatter_fn_p): New function.
10008 (vect_describe_gather_scatter_call): Likewise.
10009 (vect_check_gather_scatter): Try using internal functions for
10010 gather loads. Recognize existing calls to a gather load function.
10011 (vect_analyze_data_refs): Consider using gather loads if
10012 supports_vec_gather_load_p.
10013 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10014 (vect_get_gather_scatter_offset_type): Likewise.
10015 (vect_convert_mask_for_vectype): Likewise.
10016 (vect_add_conversion_to_patterm): Likewise.
10017 (vect_try_gather_scatter_pattern): Likewise.
10018 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10019 (vect_vect_recog_func_ptrs): Add it.
10020 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10021 internal_fn_mask_index and internal_gather_scatter_fn_p.
10022 (check_load_store_masking): Take the gather_scatter_info as an
10023 argument and handle gather loads.
10024 (vect_get_gather_scatter_ops): New function.
10025 (vectorizable_call): Check internal_load_fn_p.
10026 (vectorizable_load): Likewise. Handle gather load internal
10028 (vectorizable_store): Update call to check_load_store_masking.
10029 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10030 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10031 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10032 (aarch64_gather_scale_operand_d): New predicates.
10033 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10034 (mask_gather_load<mode>): New insns.
10036 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10037 Alan Hayward <alan.hayward@arm.com>
10038 David Sherwood <david.sherwood@arm.com>
10040 * optabs.def (fold_left_plus_optab): New optab.
10041 * doc/md.texi (fold_left_plus_@var{m}): Document.
10042 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10043 * internal-fn.c (fold_left_direct): Define.
10044 (expand_fold_left_optab_fn): Likewise.
10045 (direct_fold_left_optab_supported_p): Likewise.
10046 * fold-const-call.c (fold_const_fold_left): New function.
10047 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10048 * tree-parloops.c (valid_reduction_p): New function.
10049 (gather_scalar_reductions): Use it.
10050 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10051 (vect_finish_replace_stmt): Declare.
10052 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10053 (needs_fold_left_reduction_p): New function, split out from...
10054 (vect_is_simple_reduction): ...here. Accept reductions that
10055 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10056 (vect_force_simple_reduction): Also store the reduction type in
10057 the assignment's STMT_VINFO_REDUC_TYPE.
10058 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10059 (merge_with_identity): New function.
10060 (vect_expand_fold_left): Likewise.
10061 (vectorize_fold_left_reduction): Likewise.
10062 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10063 scalar phi in place for it. Check for target support and reject
10064 cases that would reassociate the operation. Defer the transform
10065 phase to vectorize_fold_left_reduction.
10066 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10067 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10068 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10070 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10072 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10073 call to ifc_temp_var.
10075 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10076 Alan Hayward <alan.hayward@arm.com>
10077 David Sherwood <david.sherwood@arm.com>
10079 * target.def (legitimize_address_displacement): Take the original
10080 offset as a poly_int.
10081 * targhooks.h (default_legitimize_address_displacement): Update
10083 * targhooks.c (default_legitimize_address_displacement): Likewise.
10084 * doc/tm.texi: Regenerate.
10085 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10086 as an argument, moving assert of ad->disp == ad->disp_term to...
10087 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10088 Try calling targetm.legitimize_address_displacement before expanding
10089 the address rather than afterwards, and adjust for the new interface.
10090 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10091 Match the new hook interface. Handle SVE addresses.
10092 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10093 new hook interface.
10095 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10097 * Makefile.in (OBJS): Add early-remat.o.
10098 * target.def (select_early_remat_modes): New hook.
10099 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10100 * doc/tm.texi: Regenerate.
10101 * targhooks.h (default_select_early_remat_modes): Declare.
10102 * targhooks.c (default_select_early_remat_modes): New function.
10103 * timevar.def (TV_EARLY_REMAT): New timevar.
10104 * passes.def (pass_early_remat): New pass.
10105 * tree-pass.h (make_pass_early_remat): Declare.
10106 * early-remat.c: New file.
10107 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10109 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
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 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10116 vfm1 with a bound_epilog parameter.
10117 (vect_do_peeling): Update calls accordingly, and move the prologue
10118 call earlier in the function. Treat the base bound_epilog as 0 for
10119 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10120 this base when peeling for gaps.
10121 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10122 with fully-masked loops.
10123 (vect_estimate_min_profitable_iters): Handle the single peeled
10124 iteration in that case.
10126 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10127 Alan Hayward <alan.hayward@arm.com>
10128 David Sherwood <david.sherwood@arm.com>
10130 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10131 single-element interleaving even if the size is not a power of 2.
10132 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10133 accesses for single-element interleaving if the group size is
10136 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10137 Alan Hayward <alan.hayward@arm.com>
10138 David Sherwood <david.sherwood@arm.com>
10140 * doc/md.texi (fold_extract_last_@var{m}): Document.
10141 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10142 * optabs.def (fold_extract_last_optab): New optab.
10143 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10144 * internal-fn.c (fold_extract_direct): New macro.
10145 (expand_fold_extract_optab_fn): Likewise.
10146 (direct_fold_extract_optab_supported_p): Likewise.
10147 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10148 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10149 EXTRACT_LAST_REDUCTION.
10150 (get_initial_def_for_reduction): Do not create an initial vector
10151 for EXTRACT_LAST_REDUCTION reductions.
10152 (vectorizable_reduction): Leave the scalar phi in place for
10153 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10154 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10155 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10156 transform phase to vectorizable_condition.
10157 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10159 (vect_finish_stmt_generation): ...here.
10160 (vect_finish_replace_stmt): New function.
10161 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10162 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10164 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10166 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10167 Alan Hayward <alan.hayward@arm.com>
10168 David Sherwood <david.sherwood@arm.com>
10170 * doc/md.texi (extract_last_@var{m}): Document.
10171 * optabs.def (extract_last_optab): New optab.
10172 * internal-fn.def (EXTRACT_LAST): New internal function.
10173 * internal-fn.c (cond_unary_direct): New macro.
10174 (expand_cond_unary_optab_fn): Likewise.
10175 (direct_cond_unary_optab_supported_p): Likewise.
10176 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10177 loops using EXTRACT_LAST.
10178 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10179 (extract_last_<mode>): ...this optab.
10180 (vec_extract<mode><Vel>): Update accordingly.
10182 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10183 Alan Hayward <alan.hayward@arm.com>
10184 David Sherwood <david.sherwood@arm.com>
10186 * target.def (empty_mask_is_expensive): New hook.
10187 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10188 * doc/tm.texi: Regenerate.
10189 * targhooks.h (default_empty_mask_is_expensive): Declare.
10190 * targhooks.c (default_empty_mask_is_expensive): New function.
10191 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10192 if the target says that empty masks are expensive.
10193 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10195 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10197 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10198 Alan Hayward <alan.hayward@arm.com>
10199 David Sherwood <david.sherwood@arm.com>
10201 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10202 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10203 (vect_use_loop_mask_for_alignment_p): New function.
10204 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10205 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10206 niters_skip argument. Make sure that the first niters_skip elements
10207 of the first iteration are inactive.
10208 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10209 Update call to vect_set_loop_masks_directly.
10210 (get_misalign_in_elems): New function, split out from...
10211 (vect_gen_prolog_loop_niters): ...here.
10212 (vect_update_init_of_dr): Take a code argument that specifies whether
10213 the adjustment should be added or subtracted.
10214 (vect_update_init_of_drs): Likewise.
10215 (vect_prepare_for_masked_peels): New function.
10216 (vect_do_peeling): Skip prologue peeling if we're using a mask
10217 instead. Update call to vect_update_inits_of_drs.
10218 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10220 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10221 alignment. Do not include the number of peeled iterations in
10222 the minimum threshold in that case.
10223 (vectorizable_induction): Adjust the start value down by
10224 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10225 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10226 Take the number of skipped iterations into account when calculating
10228 * tree-vect-stmts.c (vect_gen_while_not): New function.
10230 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10231 Alan Hayward <alan.hayward@arm.com>
10232 David Sherwood <david.sherwood@arm.com>
10234 * doc/sourcebuild.texi (vect_fully_masked): Document.
10235 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10236 default value to 0.
10237 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10239 (vect_analyze_loop_2): ...here. Don't check the vectorization
10240 factor against the number of loop iterations if the loop is
10243 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10244 Alan Hayward <alan.hayward@arm.com>
10245 David Sherwood <david.sherwood@arm.com>
10247 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10248 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10249 (dump_groups): Update accordingly.
10250 (iv_use::mem_type): New member variable.
10251 (address_p): New function.
10252 (record_use): Add a mem_type argument and initialize the new
10254 (record_group_use): Add a mem_type argument. Use address_p.
10255 Remove obsolete null checks of base_object. Update call to record_use.
10256 (find_interesting_uses_op): Update call to record_group_use.
10257 (find_interesting_uses_cond): Likewise.
10258 (find_interesting_uses_address): Likewise.
10259 (get_mem_type_for_internal_fn): New function.
10260 (find_address_like_use): Likewise.
10261 (find_interesting_uses_stmt): Try find_address_like_use before
10262 calling find_interesting_uses_op.
10263 (addr_offset_valid_p): Use the iv mem_type field as the type
10264 of the addressed memory.
10265 (add_autoinc_candidates): Likewise.
10266 (get_address_cost): Likewise.
10267 (split_small_address_groups_p): Use address_p.
10268 (split_address_groups): Likewise.
10269 (add_iv_candidate_for_use): Likewise.
10270 (autoinc_possible_for_pair): Likewise.
10271 (rewrite_groups): Likewise.
10272 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10273 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10274 (get_alias_ptr_type_for_ptr_address): New function.
10275 (rewrite_use_address): Rewrite address uses in calls that were
10276 identified by find_address_like_use.
10278 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10279 Alan Hayward <alan.hayward@arm.com>
10280 David Sherwood <david.sherwood@arm.com>
10282 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10284 * gimple-expr.h (is_gimple_addressable: Likewise.
10285 * gimple-expr.c (is_gimple_address): Likewise.
10286 * internal-fn.c (expand_call_mem_ref): New function.
10287 (expand_mask_load_optab_fn): Use it.
10288 (expand_mask_store_optab_fn): Likewise.
10290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10291 Alan Hayward <alan.hayward@arm.com>
10292 David Sherwood <david.sherwood@arm.com>
10294 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10295 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10296 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10297 (cond_umax@var{mode}): Document.
10298 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10299 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10300 (cond_umin_optab, cond_umax_optab): New optabs.
10301 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10302 (COND_IOR, COND_XOR): New internal functions.
10303 * internal-fn.h (get_conditional_internal_fn): Declare.
10304 * internal-fn.c (cond_binary_direct): New macro.
10305 (expand_cond_binary_optab_fn): Likewise.
10306 (direct_cond_binary_optab_supported_p): Likewise.
10307 (get_conditional_internal_fn): New function.
10308 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10309 Cope with reduction statements that are vectorized as calls rather
10311 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10312 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10313 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10314 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10315 (UNSPEC_COND_EOR): New unspecs.
10316 (optab): Add mappings for them.
10317 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10318 (sve_int_op, sve_fp_op): New int attributes.
10320 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10321 Alan Hayward <alan.hayward@arm.com>
10322 David Sherwood <david.sherwood@arm.com>
10324 * optabs.def (while_ult_optab): New optab.
10325 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10326 * internal-fn.def (WHILE_ULT): New internal function.
10327 * internal-fn.h (direct_internal_fn_supported_p): New override
10328 that takes two types as argument.
10329 * internal-fn.c (while_direct): New macro.
10330 (expand_while_optab_fn): New function.
10331 (convert_optab_supported_p): Likewise.
10332 (direct_while_optab_supported_p): New macro.
10333 * wide-int.h (wi::udiv_ceil): New function.
10334 * tree-vectorizer.h (rgroup_masks): New structure.
10335 (vec_loop_masks): New typedef.
10336 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10337 and fully_masked_p.
10338 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10339 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10340 (vect_max_vf): New function.
10341 (slpeel_make_loop_iterate_ntimes): Delete.
10342 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10343 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10344 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10345 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10346 internal-fn.h, stor-layout.h and optabs-query.h.
10347 (vect_set_loop_mask): New function.
10348 (add_preheader_seq): Likewise.
10349 (add_header_seq): Likewise.
10350 (interleave_supported_p): Likewise.
10351 (vect_maybe_permute_loop_masks): Likewise.
10352 (vect_set_loop_masks_directly): Likewise.
10353 (vect_set_loop_condition_masked): Likewise.
10354 (vect_set_loop_condition_unmasked): New function, split out from
10355 slpeel_make_loop_iterate_ntimes.
10356 (slpeel_make_loop_iterate_ntimes): Rename to..
10357 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10358 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10359 (vect_do_peeling): Update call accordingly.
10360 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10362 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10363 mask_compare_type, can_fully_mask_p and fully_masked_p.
10364 (release_vec_loop_masks): New function.
10365 (_loop_vec_info): Use it to free the loop masks.
10366 (can_produce_all_loop_masks_p): New function.
10367 (vect_get_max_nscalars_per_iter): Likewise.
10368 (vect_verify_full_masking): Likewise.
10369 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10370 retries, and free the mask rgroups before retrying. Check loop-wide
10371 reasons for disallowing fully-masked loops. Make the final decision
10372 about whether use a fully-masked loop or not.
10373 (vect_estimate_min_profitable_iters): Do not assume that peeling
10374 for the number of iterations will be needed for fully-masked loops.
10375 (vectorizable_reduction): Disable fully-masked loops.
10376 (vectorizable_live_operation): Likewise.
10377 (vect_halve_mask_nunits): New function.
10378 (vect_double_mask_nunits): Likewise.
10379 (vect_record_loop_mask): Likewise.
10380 (vect_get_loop_mask): Likewise.
10381 (vect_transform_loop): Handle the case in which the final loop
10382 iteration might handle a partial vector. Call vect_set_loop_condition
10383 instead of slpeel_make_loop_iterate_ntimes.
10384 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10385 (check_load_store_masking): New function.
10386 (prepare_load_store_mask): Likewise.
10387 (vectorizable_store): Handle fully-masked loops.
10388 (vectorizable_load): Likewise.
10389 (supportable_widening_operation): Use vect_halve_mask_nunits for
10391 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10392 (vect_gen_while): New function.
10393 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10394 (aarch64_uqdec<mode>): New insn.
10396 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10397 Alan Hayward <alan.hayward@arm.com>
10398 David Sherwood <david.sherwood@arm.com>
10400 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10401 (reduc_xor_scal_optab): New optabs.
10402 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10403 (reduc_xor_scal_@var{m}): Document.
10404 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10405 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10406 internal functions.
10407 * fold-const-call.c (fold_const_call): Handle them.
10408 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10409 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10410 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10411 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10412 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10413 (UNSPEC_XORV): New unspecs.
10414 (optab): Add entries for them.
10415 (BITWISEV): New int iterator.
10416 (bit_reduc_op): New int attributes.
10418 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10419 Alan Hayward <alan.hayward@arm.com>
10420 David Sherwood <david.sherwood@arm.com>
10422 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10423 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10424 * optabs.def (vec_shl_insert_optab): New optab.
10425 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10426 (duplicate_and_interleave): Likewise.
10427 * tree-vect-loop.c: Include internal-fn.h.
10428 (neutral_op_for_slp_reduction): New function, split out from
10429 get_initial_defs_for_reduction.
10430 (get_initial_def_for_reduction): Handle option 2 for variable-length
10431 vectors by loading the neutral value into a vector and then shifting
10432 the initial value into element 0.
10433 (get_initial_defs_for_reduction): Replace the code argument with
10434 the neutral value calculated by neutral_op_for_slp_reduction.
10435 Use gimple_build_vector for constant-length vectors.
10436 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10437 but the first group_size elements have a neutral value.
10438 Use duplicate_and_interleave otherwise.
10439 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10440 Update call to get_initial_defs_for_reduction. Handle SLP
10441 reductions for variable-length vectors by creating one vector
10442 result for each scalar result, with the elements associated
10443 with other scalar results stubbed out with the neutral value.
10444 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
10445 Require IFN_VEC_SHL_INSERT for double reductions on
10446 variable-length vectors, or SLP reductions that have
10447 a neutral value. Require can_duplicate_and_interleave_p
10448 support for variable-length unchained SLP reductions if there
10449 is no neutral value, such as for MIN/MAX reductions. Also require
10450 the number of vector elements to be a multiple of the number of
10451 SLP statements when doing variable-length unchained SLP reductions.
10452 Update call to vect_create_epilog_for_reduction.
10453 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
10454 and remove initial values.
10455 (duplicate_and_interleave): Make public.
10456 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
10457 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
10459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10460 Alan Hayward <alan.hayward@arm.com>
10461 David Sherwood <david.sherwood@arm.com>
10463 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
10464 (can_duplicate_and_interleave_p): New function.
10465 (vect_get_and_check_slp_defs): Take the vector of statements
10466 rather than just the current one. Remove excess parentheses.
10467 Restriction rejectinon of vect_constant_def and vect_external_def
10468 for variable-length vectors to boolean types, or types for which
10469 can_duplicate_and_interleave_p is false.
10470 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
10471 (duplicate_and_interleave): New function.
10472 (vect_get_constant_vectors): Use gimple_build_vector for
10473 constant-length vectors and suitable variable-length constant
10474 vectors. Use duplicate_and_interleave for other variable-length
10475 vectors. Don't defer the update when inserting new statements.
10477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10478 Alan Hayward <alan.hayward@arm.com>
10479 David Sherwood <david.sherwood@arm.com>
10481 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
10482 min_profitable_iters doesn't go negative.
10484 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10485 Alan Hayward <alan.hayward@arm.com>
10486 David Sherwood <david.sherwood@arm.com>
10488 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
10489 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
10490 * optabs.def (vec_mask_load_lanes_optab): New optab.
10491 (vec_mask_store_lanes_optab): Likewise.
10492 * internal-fn.def (MASK_LOAD_LANES): New internal function.
10493 (MASK_STORE_LANES): Likewise.
10494 * internal-fn.c (mask_load_lanes_direct): New macro.
10495 (mask_store_lanes_direct): Likewise.
10496 (expand_mask_load_optab_fn): Handle masked operations.
10497 (expand_mask_load_lanes_optab_fn): New macro.
10498 (expand_mask_store_optab_fn): Handle masked operations.
10499 (expand_mask_store_lanes_optab_fn): New macro.
10500 (direct_mask_load_lanes_optab_supported_p): Likewise.
10501 (direct_mask_store_lanes_optab_supported_p): Likewise.
10502 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
10504 (vect_load_lanes_supported): Likewise.
10505 * tree-vect-data-refs.c (strip_conversion): New function.
10506 (can_group_stmts_p): Likewise.
10507 (vect_analyze_data_ref_accesses): Use it instead of checking
10508 for a pair of assignments.
10509 (vect_store_lanes_supported): Take a masked_p parameter.
10510 (vect_load_lanes_supported): Likewise.
10511 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
10512 vect_store_lanes_supported and vect_load_lanes_supported.
10513 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10514 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
10515 parameter. Don't allow gaps for masked accesses.
10516 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
10517 and vect_load_lanes_supported.
10518 (get_load_store_type): Take a masked_p parameter and update
10519 call to get_group_load_store_type.
10520 (vectorizable_store): Update call to get_load_store_type.
10521 Handle IFN_MASK_STORE_LANES.
10522 (vectorizable_load): Update call to get_load_store_type.
10523 Handle IFN_MASK_LOAD_LANES.
10525 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10526 Alan Hayward <alan.hayward@arm.com>
10527 David Sherwood <david.sherwood@arm.com>
10529 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
10531 * config/aarch64/aarch64-protos.h
10532 (aarch64_sve_struct_memory_operand_p): Declare.
10533 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
10534 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
10535 (VPRED, vpred): Handle SVE structure modes.
10536 * config/aarch64/constraints.md (Utx): New constraint.
10537 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
10538 (aarch64_sve_struct_nonimmediate_operand): New predicates.
10539 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
10540 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
10541 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
10542 structure modes. Split into pieces after RA.
10543 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
10544 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
10546 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
10547 SVE structure modes.
10548 (aarch64_classify_address): Likewise.
10549 (sizetochar): Move earlier in file.
10550 (aarch64_print_operand): Handle SVE register lists.
10551 (aarch64_array_mode): New function.
10552 (aarch64_sve_struct_memory_operand_p): Likewise.
10553 (TARGET_ARRAY_MODE): Redefine.
10555 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10556 Alan Hayward <alan.hayward@arm.com>
10557 David Sherwood <david.sherwood@arm.com>
10559 * target.def (array_mode): New target hook.
10560 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
10561 * doc/tm.texi: Regenerate.
10562 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
10563 * hooks.c (hook_optmode_mode_uhwi_none): New function.
10564 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
10565 targetm.array_mode.
10566 * stor-layout.c (mode_for_array): Likewise. Support polynomial
10569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10570 Alan Hayward <alan.hayward@arm.com>
10571 David Sherwood <david.sherwood@arm.com>
10573 * fold-const.c (fold_binary_loc): Check the argument types
10574 rather than the result type when testing for a vector operation.
10576 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10578 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
10579 * doc/tm.texi: Regenerate.
10581 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10582 Alan Hayward <alan.hayward@arm.com>
10583 David Sherwood <david.sherwood@arm.com>
10585 * doc/invoke.texi (-msve-vector-bits=): Document new option.
10586 (sve): Document new AArch64 extension.
10587 * doc/md.texi (w): Extend the description of the AArch64
10588 constraint to include SVE vectors.
10589 (Upl, Upa): Document new AArch64 predicate constraints.
10590 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
10592 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
10593 (msve-vector-bits=): New option.
10594 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
10595 SVE when these are disabled.
10596 (sve): New extension.
10597 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
10598 modes. Adjust their number of units based on aarch64_sve_vg.
10599 (MAX_BITSIZE_MODE_ANY_MODE): Define.
10600 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
10601 aarch64_addr_query_type.
10602 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
10603 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
10604 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
10605 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
10606 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
10607 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
10608 (aarch64_simd_imm_zero_p): Delete.
10609 (aarch64_check_zero_based_sve_index_immediate): Declare.
10610 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10611 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10612 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10613 (aarch64_sve_float_mul_immediate_p): Likewise.
10614 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10615 rather than an rtx.
10616 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
10617 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
10618 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
10619 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
10620 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
10621 (aarch64_regmode_natural_size): Likewise.
10622 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
10623 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10625 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10626 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10627 for VG and the SVE predicate registers.
10628 (V_ALIASES): Add a "z"-prefixed alias.
10629 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10630 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10631 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10632 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10633 (REG_CLASS_NAMES): Add entries for them.
10634 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10635 and the predicate registers.
10636 (aarch64_sve_vg): Declare.
10637 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10638 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10639 (REGMODE_NATURAL_SIZE): Define.
10640 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10642 * config/aarch64/aarch64.c: Include cfgrtl.h.
10643 (simd_immediate_info): Add a constructor for series vectors,
10644 and an associated step field.
10645 (aarch64_sve_vg): New variable.
10646 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10647 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10648 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10649 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10650 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10651 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10652 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10653 (aarch64_get_mask_mode): New functions.
10654 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10655 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10656 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10657 predicate modes and predicate registers. Explicitly restrict
10658 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10659 to store a vector mode if it is recognized by
10660 aarch64_classify_vector_mode.
10661 (aarch64_regmode_natural_size): New function.
10662 (aarch64_hard_regno_caller_save_mode): Return the original mode
10664 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10665 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10666 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10667 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10669 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10670 does not overlap dest if the function is frame-related. Handle
10672 (aarch64_split_add_offset): New function.
10673 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10674 them aarch64_add_offset.
10675 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10676 and update call to aarch64_sub_sp.
10677 (aarch64_add_cfa_expression): New function.
10678 (aarch64_expand_prologue): Pass extra temporary registers to the
10679 functions above. Handle the case in which we need to emit new
10680 DW_CFA_expressions for registers that were originally saved
10681 relative to the stack pointer, but now have to be expressed
10682 relative to the frame pointer.
10683 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10685 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10686 IP0 and IP1 values for SVE frames.
10687 (aarch64_expand_vec_series): New function.
10688 (aarch64_expand_sve_widened_duplicate): Likewise.
10689 (aarch64_expand_sve_const_vector): Likewise.
10690 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10691 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10692 into the register, rather than emitting a SET directly.
10693 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10694 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10695 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10696 (offset_9bit_signed_scaled_p): New functions.
10697 (aarch64_replicate_bitmask_imm): New function.
10698 (aarch64_bitmask_imm): Use it.
10699 (aarch64_cannot_force_const_mem): Reject expressions involving
10700 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10701 (aarch64_classify_index): Handle SVE indices, by requiring
10702 a plain register index with a scale that matches the element size.
10703 (aarch64_classify_address): Handle SVE addresses. Assert that
10704 the mode of the address is VOIDmode or an integer mode.
10705 Update call to aarch64_classify_symbol.
10706 (aarch64_classify_symbolic_expression): Update call to
10707 aarch64_classify_symbol.
10708 (aarch64_const_vec_all_in_range_p): New function.
10709 (aarch64_print_vector_float_operand): Likewise.
10710 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10711 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10712 and the FP immediates 1.0 and 0.5.
10713 (aarch64_print_address_internal): Handle SVE addresses.
10714 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10715 (aarch64_regno_regclass): Handle predicate registers.
10716 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10718 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10719 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10720 (aarch64_convert_sve_vector_bits): New function.
10721 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10722 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10723 rather than an rtx.
10724 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10725 Handle SVE vector and predicate modes. Accept VL-based constants
10726 that need only one temporary register, and VL offsets that require
10727 no temporary registers.
10728 (aarch64_conditional_register_usage): Mark the predicate registers
10729 as fixed if SVE isn't available.
10730 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10731 Return true for SVE vector and predicate modes.
10732 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10733 rather than an unsigned int. Handle SVE modes.
10734 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10736 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10738 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10739 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10740 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10741 (aarch64_sve_float_mul_immediate_p): New functions.
10742 (aarch64_sve_valid_immediate): New function.
10743 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10744 Explicitly reject structure modes. Check for INDEX constants.
10745 Handle PTRUE and PFALSE constants.
10746 (aarch64_check_zero_based_sve_index_immediate): New function.
10747 (aarch64_simd_imm_zero_p): Delete.
10748 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10749 vector modes. Accept constants in the range of CNT[BHWD].
10750 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10751 ask for an Advanced SIMD mode.
10752 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10753 (aarch64_simd_vector_alignment): Handle SVE predicates.
10754 (aarch64_vectorize_preferred_vector_alignment): New function.
10755 (aarch64_simd_vector_alignment_reachable): Use it instead of
10757 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10758 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10760 (MAX_VECT_LEN): Delete.
10761 (expand_vec_perm_d): Add a vec_flags field.
10762 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10763 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10764 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10766 (aarch64_evpc_rev): Rename to...
10767 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10768 (aarch64_evpc_rev_global): New function.
10769 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10770 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10772 (aarch64_evpc_sve_tbl): New function.
10773 (aarch64_expand_vec_perm_const_1): Update after rename of
10774 aarch64_evpc_rev. Handle SVE permutes too, trying
10775 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10776 than aarch64_evpc_tbl.
10777 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10778 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10779 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10780 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10781 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10782 (aarch64_expand_sve_vcond): New functions.
10783 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10784 of aarch64_vector_mode_p.
10785 (aarch64_dwarf_poly_indeterminate_value): New function.
10786 (aarch64_compute_pressure_classes): Likewise.
10787 (aarch64_can_change_mode_class): Likewise.
10788 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10789 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10790 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10791 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10792 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10793 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10794 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10795 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10797 (Dn, Dl, Dr): Accept const as well as const_vector.
10798 (Dz): Likewise. Compare against CONST0_RTX.
10799 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10800 of "vector" where appropriate.
10801 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10802 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10803 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10804 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10805 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10806 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10807 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10808 (v_int_equiv): Extend to SVE modes.
10809 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10811 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10812 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10813 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10814 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10815 (SVE_COND_FP_CMP): New int iterators.
10816 (perm_hilo): Handle the new unpack unspecs.
10817 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10819 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10820 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10821 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10822 (aarch64_equality_operator, aarch64_constant_vector_operand)
10823 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10824 (aarch64_sve_nonimmediate_operand): Likewise.
10825 (aarch64_sve_general_operand): Likewise.
10826 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10827 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10828 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10829 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10830 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10831 (aarch64_sve_float_arith_immediate): Likewise.
10832 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10833 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10834 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10835 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10836 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10837 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10838 (aarch64_sve_float_arith_operand): Likewise.
10839 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10840 (aarch64_sve_float_mul_operand): Likewise.
10841 (aarch64_sve_vec_perm_operand): Likewise.
10842 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10843 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10844 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10845 as well as const_vector.
10846 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10847 in file. Use CONST0_RTX and CONSTM1_RTX.
10848 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10849 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10850 Use aarch64_simd_imm_zero.
10851 * config/aarch64/aarch64-sve.md: New file.
10852 * config/aarch64/aarch64.md: Include it.
10853 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10854 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10855 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10856 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10857 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10858 (sve): New attribute.
10859 (enabled): Disable instructions with the sve attribute unless
10861 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10862 aarch64_expand_mov_immediate.
10863 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10864 CNT[BHSD] immediates.
10865 (movti): Split CONST_POLY_INT moves into two halves.
10866 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10867 Split additions that need a temporary here if the destination
10868 is the stack pointer.
10869 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10870 (*add<mode>3_poly_1): New instruction.
10871 (set_clobber_cc): New expander.
10873 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10875 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10876 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10877 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10878 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10879 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10880 Change innermode from fixed_mode_size to machine_mode.
10881 (simplify_subreg): Update call accordingly. Handle a constant-sized
10882 subreg of a variable-length CONST_VECTOR.
10884 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10885 Alan Hayward <alan.hayward@arm.com>
10886 David Sherwood <david.sherwood@arm.com>
10888 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10889 (add_offset_to_base): New function, split out from...
10890 (create_mem_ref): ...here. When handling a scale other than 1,
10891 check first whether the address is valid without the offset.
10892 Add it into the base if so, leaving the index and scale as-is.
10894 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10897 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10898 fold_for_warn before checking if arg2 is INTEGER_CST.
10900 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10902 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10903 (store_multiple_operation): Delete.
10904 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10905 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10906 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10907 guarded by TARGET_STRING.
10908 (rs6000_output_load_multiple): Delete.
10909 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10910 OPTION_MASK_STRING / TARGET_STRING handling.
10911 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10912 (const rs6000_opt_masks) <"string">: Change mask to 0.
10913 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10914 (MASK_STRING): Delete.
10915 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10917 (load_multiple): Delete.
10924 (store_multiple): Delete.
10931 (movmemsi_8reg): Delete.
10932 (corresponding unnamed define_insn): Delete.
10933 (movmemsi_6reg): Delete.
10934 (corresponding unnamed define_insn): Delete.
10935 (movmemsi_4reg): Delete.
10936 (corresponding unnamed define_insn): Delete.
10937 (movmemsi_2reg): Delete.
10938 (corresponding unnamed define_insn): Delete.
10939 (movmemsi_1reg): Delete.
10940 (corresponding unnamed define_insn): Delete.
10941 * config/rs6000/rs6000.opt (mno-string): New.
10942 (mstring): Replace by deprecation warning stub.
10943 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10945 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10947 * regrename.c (regrename_do_replace): If replacing the same
10948 reg multiple times, try to reuse last created gen_raw_REG.
10951 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10952 main to workaround a bug in GDB.
10954 2018-01-12 Tom de Vries <tom@codesourcery.com>
10957 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10959 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10961 PR rtl-optimization/80481
10962 * ira-color.c (get_cap_member): New function.
10963 (allocnos_conflict_by_live_ranges_p): Use it.
10964 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10965 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10967 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10970 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10971 (*saddl_se_1): Ditto.
10972 (*ssubsi_1): Ditto.
10973 (*ssubl_se_1): Ditto.
10975 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10977 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10978 rather than wi::to_widest for DR_INITs.
10979 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10980 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10981 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10983 (vect_analyze_group_access_1): Note that here.
10985 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10987 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10988 polynomial type sizes.
10990 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10992 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10993 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10994 (gimple_add_tmp_var): Likewise.
10996 2018-01-12 Martin Liska <mliska@suse.cz>
10998 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10999 (gimple_alloc_sizes): Likewise.
11000 (dump_gimple_statistics): Use PRIu64 in printf format.
11001 * gimple.h: Change uint64_t to int.
11003 2018-01-12 Martin Liska <mliska@suse.cz>
11005 * tree-core.h: Use uint64_t instead of int.
11006 * tree.c (tree_node_counts): Likewise.
11007 (tree_node_sizes): Likewise.
11008 (dump_tree_statistics): Use PRIu64 in printf format.
11010 2018-01-12 Martin Liska <mliska@suse.cz>
11012 * Makefile.in: As qsort_chk is implemented in vec.c, add
11013 vec.o to linkage of gencfn-macros.
11014 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11015 passing the info to record_node_allocation_statistics.
11016 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11018 * ggc-common.c (struct ggc_usage): Add operator== and use
11019 it in operator< and compare function.
11020 * mem-stats.h (struct mem_usage): Likewise.
11021 * vec.c (struct vec_usage): Remove operator< and compare
11022 function. Can be simply inherited.
11024 2018-01-12 Martin Jambor <mjambor@suse.cz>
11027 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11028 * tree-ssa-math-opts.c: Include domwalk.h.
11029 (convert_mult_to_fma_1): New function.
11030 (fma_transformation_info): New type.
11031 (fma_deferring_state): Likewise.
11032 (cancel_fma_deferring): New function.
11033 (result_of_phi): Likewise.
11034 (last_fma_candidate_feeds_initial_phi): Likewise.
11035 (convert_mult_to_fma): Added deferring logic, split actual
11036 transformation to convert_mult_to_fma_1.
11037 (math_opts_dom_walker): New type.
11038 (math_opts_dom_walker::after_dom_children): New method, body moved
11039 here from pass_optimize_widening_mul::execute, added deferring logic
11041 (pass_optimize_widening_mul::execute): Moved most of code to
11042 math_opts_dom_walker::after_dom_children.
11043 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11044 * config/i386/i386.c (ix86_option_override_internal): Added
11045 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11047 2018-01-12 Richard Biener <rguenther@suse.de>
11050 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11051 inline instance vars.
11053 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11056 * config/rx/rx.c (rx_is_restricted_memory_address):
11057 Handle SUBREG case.
11059 2018-01-12 Richard Biener <rguenther@suse.de>
11061 PR tree-optimization/80846
11062 * target.def (split_reduction): New target hook.
11063 * targhooks.c (default_split_reduction): New function.
11064 * targhooks.h (default_split_reduction): Declare.
11065 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11066 target requests first reduce vectors by combining low and high
11068 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11069 (get_vectype_for_scalar_type_and_size): Export.
11070 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11071 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11072 * doc/tm.texi: Regenerate.
11073 * config/i386/i386.c (ix86_split_reduction): Implement
11074 TARGET_VECTORIZE_SPLIT_REDUCTION.
11076 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11079 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11080 in PIC mode except for TARGET_VXWORKS_RTP.
11081 * config/sparc/sparc.c: Include cfgrtl.h.
11082 (TARGET_INIT_PIC_REG): Define.
11083 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11084 (sparc_pic_register_p): New predicate.
11085 (sparc_legitimate_address_p): Use it.
11086 (sparc_legitimize_pic_address): Likewise.
11087 (sparc_delegitimize_address): Likewise.
11088 (sparc_mode_dependent_address_p): Likewise.
11089 (gen_load_pcrel_sym): Remove 4th parameter.
11090 (load_got_register): Adjust call to above. Remove obsolete stuff.
11091 (sparc_expand_prologue): Do not call load_got_register here.
11092 (sparc_flat_expand_prologue): Likewise.
11093 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11094 (sparc_use_pseudo_pic_reg): New function.
11095 (sparc_init_pic_reg): Likewise.
11096 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11097 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11099 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11101 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11102 Add item for branch_cost.
11104 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11106 PR rtl-optimization/83565
11107 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11108 not extend the result to a larger mode for rotate operations.
11109 (num_sign_bit_copies1): Likewise.
11111 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11114 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11116 Use values-Xc.o for -pedantic.
11117 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11119 2018-01-12 Martin Liska <mliska@suse.cz>
11122 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11124 (possible_polymorphic_call_targets): Use it.
11125 (ipa_devirt): Likewise.
11127 2018-01-12 Martin Liska <mliska@suse.cz>
11129 * profile-count.h (enum profile_quality): Use 0 as invalid
11130 enum value of profile_quality.
11132 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11134 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11135 -mext-string options.
11137 2018-01-12 Richard Biener <rguenther@suse.de>
11139 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11140 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11141 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11143 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11145 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11147 * configure.ac (--with-long-double-format): Add support for the
11148 configuration option to change the default long double format on
11150 * config.gcc (powerpc*-linux*-*): Likewise.
11151 * configure: Regenerate.
11152 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11153 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11154 used without modification.
11156 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11158 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11159 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11160 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11161 MISC_BUILTIN_SPEC_BARRIER.
11162 (rs6000_init_builtins): Likewise.
11163 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11165 (speculation_barrier): New define_insn.
11166 * doc/extend.texi: Document __builtin_speculation_barrier.
11168 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11171 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11172 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11173 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11175 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11176 integral modes instead of "ss" and "sd".
11177 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11178 vectors with 32-bit and 64-bit elements.
11179 (vecdupssescalarmodesuffix): New mode attribute.
11180 (vec_dup<mode>): Use it.
11182 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11185 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11186 frame if argument is passed on stack.
11188 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11191 * ree.c (combine_reaching_defs): Optimize also
11192 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11193 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11195 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11197 PR middle-end/83189
11198 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11200 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11202 PR middle-end/83718
11203 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11204 after they are computed.
11206 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11208 PR tree-optimization/83695
11209 * gimple-loop-linterchange.cc
11210 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11211 reset cached scev information after interchange.
11212 (pass_linterchange::execute): Remove call to scev_reset_htab.
11214 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11216 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11217 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11218 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11219 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11220 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11221 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11222 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11223 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11224 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11225 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11226 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11227 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11228 (V_lane_reg): Likewise.
11229 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11231 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11232 (vfmal_lane_low<mode>_intrinsic,
11233 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11234 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11235 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11236 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11237 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11238 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11240 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11242 * config/arm/arm-cpus.in (fp16fml): New feature.
11243 (ALL_SIMD): Add fp16fml.
11244 (armv8.2-a): Add fp16fml as an option.
11245 (armv8.3-a): Likewise.
11246 (armv8.4-a): Add fp16fml as part of fp16.
11247 * config/arm/arm.h (TARGET_FP16FML): Define.
11248 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11250 * config/arm/arm-modes.def (V2HF): Define.
11251 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11252 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11253 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11254 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11255 vfmsl_low, vfmsl_high): New set of builtins.
11256 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11257 (vfml_op): New code attribute.
11258 (VFMLHALVES): New int iterator.
11259 (VFML, VFMLSEL): New mode attributes.
11260 (V_reg): Define mapping for V2HF.
11261 (V_hi, V_lo): New mode attributes.
11262 (VF_constraint): Likewise.
11263 (vfml_half, vfml_half_selector): New int attributes.
11264 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11266 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11267 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11269 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11270 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11271 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11272 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11274 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11275 Document new effective target and option set.
11277 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11279 * config/arm/arm-cpus.in (armv8_4): New feature.
11280 (ARMv8_4a): New fgroup.
11281 (armv8.4-a): New arch.
11282 * config/arm/arm-tables.opt: Regenerate.
11283 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11284 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11285 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11286 Add matching rules for -march=armv8.4-a and extensions.
11287 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11289 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11292 * config/rx/rx.md (BW): New mode attribute.
11293 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11295 2018-01-11 Richard Biener <rguenther@suse.de>
11297 PR tree-optimization/83435
11298 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11299 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11300 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11302 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11303 Alan Hayward <alan.hayward@arm.com>
11304 David Sherwood <david.sherwood@arm.com>
11306 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11308 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11309 (aarch64_print_address_internal): Use it to check for a zero offset.
11311 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11312 Alan Hayward <alan.hayward@arm.com>
11313 David Sherwood <david.sherwood@arm.com>
11315 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11316 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11317 Return a poly_int64 rather than a HOST_WIDE_INT.
11318 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11319 rather than a HOST_WIDE_INT.
11320 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11321 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11322 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11323 final_offset from HOST_WIDE_INT to poly_int64.
11324 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11325 to_constant when getting the number of units in an Advanced SIMD
11327 (aarch64_builtin_vectorized_function): Check for a constant number
11329 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11331 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11332 attribute instead of GET_MODE_NUNITS.
11333 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11334 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11335 GET_MODE_SIZE for fixed-size registers.
11336 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11337 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11338 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11339 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11340 (aarch64_print_operand, aarch64_print_address_internal)
11341 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11342 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11343 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11344 Handle polynomial GET_MODE_SIZE.
11345 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11346 wider than SImode without modification.
11347 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11348 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11349 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11350 passing and returning SVE modes.
11351 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11352 rather than GEN_INT.
11353 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11354 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11355 (aarch64_allocate_and_probe_stack_space): Likewise.
11356 (aarch64_layout_frame): Cope with polynomial offsets.
11357 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11358 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11359 polynomial offsets.
11360 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11361 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11362 poly_int64 rather than a HOST_WIDE_INT.
11363 (aarch64_get_separate_components, aarch64_process_components)
11364 (aarch64_expand_prologue, aarch64_expand_epilogue)
11365 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11366 (aarch64_anchor_offset): New function, split out from...
11367 (aarch64_legitimize_address): ...here.
11368 (aarch64_builtin_vectorization_cost): Handle polynomial
11369 TYPE_VECTOR_SUBPARTS.
11370 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11372 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11373 number of elements from the PARALLEL rather than the mode.
11374 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11375 rather than GET_MODE_BITSIZE.
11376 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11377 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11378 (aarch64_expand_vec_perm_const_1): Handle polynomial
11379 d->perm.length () and d->perm elements.
11380 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11381 Apply to_constant to d->perm elements.
11382 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11383 polynomial CONST_VECTOR_NUNITS.
11384 (aarch64_move_pointer): Take amount as a poly_int64 rather
11386 (aarch64_progress_pointer): Avoid temporary variable.
11387 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11388 the mode attribute instead of GET_MODE.
11390 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11391 Alan Hayward <alan.hayward@arm.com>
11392 David Sherwood <david.sherwood@arm.com>
11394 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11395 x exists before using it.
11396 (aarch64_add_constant_internal): Rename to...
11397 (aarch64_add_offset_1): ...this. Replace regnum with separate
11398 src and dest rtxes. Handle the case in which they're different,
11399 including when the offset is zero. Replace scratchreg with an rtx.
11400 Use 2 additions if there is no spare register into which we can
11401 move a 16-bit constant.
11402 (aarch64_add_constant): Delete.
11403 (aarch64_add_offset): Replace reg with separate src and dest
11404 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11405 Use aarch64_add_offset_1.
11406 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11407 an rtx rather than an int. Take the delta as a poly_int64
11408 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11409 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11410 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11411 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11412 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11413 and aarch64_add_sp.
11414 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11415 aarch64_add_constant.
11417 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11419 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11420 Use scalar_float_mode.
11422 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11424 * config/aarch64/aarch64-simd.md
11425 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11426 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11427 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11428 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11429 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11430 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11431 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11432 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11433 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11434 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11436 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11439 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11440 targ_options->x_arm_arch_string is non NULL.
11442 2018-01-11 Tamar Christina <tamar.christina@arm.com>
11444 * config/aarch64/aarch64.h
11445 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
11447 2018-01-11 Sudakshina Das <sudi.das@arm.com>
11450 * expmed.c (emit_store_flag_force): Swap if const op0
11451 and change VOIDmode to mode of op0.
11453 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11455 PR rtl-optimization/83761
11456 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
11457 than bytes to mode_for_size.
11459 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11461 PR middle-end/83189
11462 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
11463 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
11466 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11468 PR middle-end/83575
11469 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
11470 when in layout mode.
11471 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
11472 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
11475 2018-01-10 Michael Collison <michael.collison@arm.com>
11477 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
11478 * config/aarch64/aarch64-option-extension.def: Add
11479 AARCH64_OPT_EXTENSION of 'fp16fml'.
11480 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11481 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
11482 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
11483 * config/aarch64/constraints.md (Ui7): New constraint.
11484 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
11485 (VFMLA_SEL_W): Ditto.
11488 (VFMLA16_LOW): New int iterator.
11489 (VFMLA16_HIGH): Ditto.
11490 (UNSPEC_FMLAL): New unspec.
11491 (UNSPEC_FMLSL): Ditto.
11492 (UNSPEC_FMLAL2): Ditto.
11493 (UNSPEC_FMLSL2): Ditto.
11494 (f16mac): New code attribute.
11495 * config/aarch64/aarch64-simd-builtins.def
11496 (aarch64_fmlal_lowv2sf): Ditto.
11497 (aarch64_fmlsl_lowv2sf): Ditto.
11498 (aarch64_fmlalq_lowv4sf): Ditto.
11499 (aarch64_fmlslq_lowv4sf): Ditto.
11500 (aarch64_fmlal_highv2sf): Ditto.
11501 (aarch64_fmlsl_highv2sf): Ditto.
11502 (aarch64_fmlalq_highv4sf): Ditto.
11503 (aarch64_fmlslq_highv4sf): Ditto.
11504 (aarch64_fmlal_lane_lowv2sf): Ditto.
11505 (aarch64_fmlsl_lane_lowv2sf): Ditto.
11506 (aarch64_fmlal_laneq_lowv2sf): Ditto.
11507 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
11508 (aarch64_fmlalq_lane_lowv4sf): Ditto.
11509 (aarch64_fmlsl_lane_lowv4sf): Ditto.
11510 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
11511 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
11512 (aarch64_fmlal_lane_highv2sf): Ditto.
11513 (aarch64_fmlsl_lane_highv2sf): Ditto.
11514 (aarch64_fmlal_laneq_highv2sf): Ditto.
11515 (aarch64_fmlsl_laneq_highv2sf): Ditto.
11516 (aarch64_fmlalq_lane_highv4sf): Ditto.
11517 (aarch64_fmlsl_lane_highv4sf): Ditto.
11518 (aarch64_fmlalq_laneq_highv4sf): Ditto.
11519 (aarch64_fmlsl_laneq_highv4sf): Ditto.
11520 * config/aarch64/aarch64-simd.md:
11521 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
11522 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11523 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
11524 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
11525 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
11526 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
11527 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
11528 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
11529 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
11530 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
11531 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
11532 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
11533 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
11534 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
11535 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
11536 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
11537 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
11538 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
11539 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
11540 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
11541 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
11542 (vfmlsl_low_u32): Ditto.
11543 (vfmlalq_low_u32): Ditto.
11544 (vfmlslq_low_u32): Ditto.
11545 (vfmlal_high_u32): Ditto.
11546 (vfmlsl_high_u32): Ditto.
11547 (vfmlalq_high_u32): Ditto.
11548 (vfmlslq_high_u32): Ditto.
11549 (vfmlal_lane_low_u32): Ditto.
11550 (vfmlsl_lane_low_u32): Ditto.
11551 (vfmlal_laneq_low_u32): Ditto.
11552 (vfmlsl_laneq_low_u32): Ditto.
11553 (vfmlalq_lane_low_u32): Ditto.
11554 (vfmlslq_lane_low_u32): Ditto.
11555 (vfmlalq_laneq_low_u32): Ditto.
11556 (vfmlslq_laneq_low_u32): Ditto.
11557 (vfmlal_lane_high_u32): Ditto.
11558 (vfmlsl_lane_high_u32): Ditto.
11559 (vfmlal_laneq_high_u32): Ditto.
11560 (vfmlsl_laneq_high_u32): Ditto.
11561 (vfmlalq_lane_high_u32): Ditto.
11562 (vfmlslq_lane_high_u32): Ditto.
11563 (vfmlalq_laneq_high_u32): Ditto.
11564 (vfmlslq_laneq_high_u32): Ditto.
11565 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
11566 (AARCH64_FL_FOR_ARCH8_4): New.
11567 (AARCH64_ISA_F16FML): New ISA flag.
11568 (TARGET_F16FML): New feature flag for fp16fml.
11569 (doc/invoke.texi): Document new fp16fml option.
11571 2018-01-10 Michael Collison <michael.collison@arm.com>
11573 * config/aarch64/aarch64-builtins.c:
11574 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
11575 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11576 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
11577 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
11578 (AARCH64_ISA_SHA3): New ISA flag.
11579 (TARGET_SHA3): New feature flag for sha3.
11580 * config/aarch64/iterators.md (sha512_op): New int attribute.
11581 (CRYPTO_SHA512): New int iterator.
11582 (UNSPEC_SHA512H): New unspec.
11583 (UNSPEC_SHA512H2): Ditto.
11584 (UNSPEC_SHA512SU0): Ditto.
11585 (UNSPEC_SHA512SU1): Ditto.
11586 * config/aarch64/aarch64-simd-builtins.def
11587 (aarch64_crypto_sha512hqv2di): New builtin.
11588 (aarch64_crypto_sha512h2qv2di): Ditto.
11589 (aarch64_crypto_sha512su0qv2di): Ditto.
11590 (aarch64_crypto_sha512su1qv2di): Ditto.
11591 (aarch64_eor3qv8hi): Ditto.
11592 (aarch64_rax1qv2di): Ditto.
11593 (aarch64_xarqv2di): Ditto.
11594 (aarch64_bcaxqv8hi): Ditto.
11595 * config/aarch64/aarch64-simd.md:
11596 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
11597 (aarch64_crypto_sha512su0qv2di): Ditto.
11598 (aarch64_crypto_sha512su1qv2di): Ditto.
11599 (aarch64_eor3qv8hi): Ditto.
11600 (aarch64_rax1qv2di): Ditto.
11601 (aarch64_xarqv2di): Ditto.
11602 (aarch64_bcaxqv8hi): Ditto.
11603 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
11604 (vsha512h2q_u64): Ditto.
11605 (vsha512su0q_u64): Ditto.
11606 (vsha512su1q_u64): Ditto.
11607 (veor3q_u16): Ditto.
11608 (vrax1q_u64): Ditto.
11609 (vxarq_u64): Ditto.
11610 (vbcaxq_u16): Ditto.
11611 * config/arm/types.md (crypto_sha512): New type attribute.
11612 (crypto_sha3): Ditto.
11613 (doc/invoke.texi): Document new sha3 option.
11615 2018-01-10 Michael Collison <michael.collison@arm.com>
11617 * config/aarch64/aarch64-builtins.c:
11618 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
11619 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11620 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
11621 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
11622 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
11623 (AARCH64_ISA_SM4): New ISA flag.
11624 (TARGET_SM4): New feature flag for sm4.
11625 * config/aarch64/aarch64-simd-builtins.def
11626 (aarch64_sm3ss1qv4si): Ditto.
11627 (aarch64_sm3tt1aq4si): Ditto.
11628 (aarch64_sm3tt1bq4si): Ditto.
11629 (aarch64_sm3tt2aq4si): Ditto.
11630 (aarch64_sm3tt2bq4si): Ditto.
11631 (aarch64_sm3partw1qv4si): Ditto.
11632 (aarch64_sm3partw2qv4si): Ditto.
11633 (aarch64_sm4eqv4si): Ditto.
11634 (aarch64_sm4ekeyqv4si): Ditto.
11635 * config/aarch64/aarch64-simd.md:
11636 (aarch64_sm3ss1qv4si): Ditto.
11637 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11638 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11639 (aarch64_sm4eqv4si): Ditto.
11640 (aarch64_sm4ekeyqv4si): Ditto.
11641 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11642 (sm3part_op): Ditto.
11643 (CRYPTO_SM3TT): Ditto.
11644 (CRYPTO_SM3PART): Ditto.
11645 (UNSPEC_SM3SS1): New unspec.
11646 (UNSPEC_SM3TT1A): Ditto.
11647 (UNSPEC_SM3TT1B): Ditto.
11648 (UNSPEC_SM3TT2A): Ditto.
11649 (UNSPEC_SM3TT2B): Ditto.
11650 (UNSPEC_SM3PARTW1): Ditto.
11651 (UNSPEC_SM3PARTW2): Ditto.
11652 (UNSPEC_SM4E): Ditto.
11653 (UNSPEC_SM4EKEY): Ditto.
11654 * config/aarch64/constraints.md (Ui2): New constraint.
11655 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11656 * config/arm/types.md (crypto_sm3): New type attribute.
11657 (crypto_sm4): Ditto.
11658 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11659 (vsm3tt1aq_u32): Ditto.
11660 (vsm3tt1bq_u32): Ditto.
11661 (vsm3tt2aq_u32): Ditto.
11662 (vsm3tt2bq_u32): Ditto.
11663 (vsm3partw1q_u32): Ditto.
11664 (vsm3partw2q_u32): Ditto.
11665 (vsm4eq_u32): Ditto.
11666 (vsm4ekeyq_u32): Ditto.
11667 (doc/invoke.texi): Document new sm4 option.
11669 2018-01-10 Michael Collison <michael.collison@arm.com>
11671 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11672 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11673 (AARCH64_FL_FOR_ARCH8_4): New.
11674 (AARCH64_FL_V8_4): New flag.
11675 (doc/invoke.texi): Document new armv8.4-a option.
11677 2018-01-10 Michael Collison <michael.collison@arm.com>
11679 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11680 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11681 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11682 * config/aarch64/aarch64-option-extension.def: Add
11683 AARCH64_OPT_EXTENSION of 'sha2'.
11684 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11685 (crypto): Disable sha2 and aes if crypto disabled.
11686 (crypto): Enable aes and sha2 if enabled.
11687 (simd): Disable sha2 and aes if simd disabled.
11688 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11690 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11691 (TARGET_SHA2): New feature flag for sha2.
11692 (TARGET_AES): New feature flag for aes.
11693 * config/aarch64/aarch64-simd.md:
11694 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11695 conditional on TARGET_AES.
11696 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11697 (aarch64_crypto_sha1hsi): Make pattern conditional
11699 (aarch64_crypto_sha1hv4si): Ditto.
11700 (aarch64_be_crypto_sha1hv4si): Ditto.
11701 (aarch64_crypto_sha1su1v4si): Ditto.
11702 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11703 (aarch64_crypto_sha1su0v4si): Ditto.
11704 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11705 (aarch64_crypto_sha256su0v4si): Ditto.
11706 (aarch64_crypto_sha256su1v4si): Ditto.
11707 (doc/invoke.texi): Document new aes and sha2 options.
11709 2018-01-10 Martin Sebor <msebor@redhat.com>
11711 PR tree-optimization/83781
11712 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11715 2018-01-11 Martin Sebor <msebor@gmail.com>
11716 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11718 PR tree-optimization/83501
11719 PR tree-optimization/81703
11721 * tree-ssa-strlen.c (get_string_cst): Rename...
11722 (get_string_len): ...to this. Handle global constants.
11723 (handle_char_store): Adjust.
11725 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11726 Jim Wilson <jimw@sifive.com>
11728 * config/riscv/riscv-protos.h (riscv_output_return): New.
11729 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11730 (riscv_attribute_table, riscv_output_return),
11731 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11732 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11733 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11735 (riscv_expand_prologue): Add early return for naked function.
11736 (riscv_expand_epilogue): Likewise.
11737 (riscv_function_ok_for_sibcall): Return false for naked function.
11738 (riscv_set_current_function): New.
11739 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11740 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11741 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11742 * doc/extend.texi (RISC-V Function Attributes): New.
11744 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11746 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11747 check for 128-bit long double before checking TCmode.
11748 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11749 128-bit long doubles before checking TFmode or TCmode.
11750 (FLOAT128_IBM_P): Likewise.
11752 2018-01-10 Martin Sebor <msebor@redhat.com>
11754 PR tree-optimization/83671
11755 * builtins.c (c_strlen): Unconditionally return zero for the empty
11757 Use -Warray-bounds for warnings.
11758 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11759 for non-constant array indices with COMPONENT_REF, arrays of
11760 arrays, and pointers to arrays.
11761 (gimple_fold_builtin_strlen): Determine and set length range for
11762 non-constant character arrays.
11764 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11766 PR middle-end/81897
11767 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11770 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11772 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11774 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11777 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11778 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11779 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11780 indexed_or_indirect_operand predicate.
11781 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11782 (*vsx_le_perm_load_v8hi): Likewise.
11783 (*vsx_le_perm_load_v16qi): Likewise.
11784 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11785 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11786 (*vsx_le_perm_store_v8hi): Likewise.
11787 (*vsx_le_perm_store_v16qi): Likewise.
11788 (eight unnamed splitters): Likewise.
11790 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11792 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11793 * config/rs6000/emmintrin.h: Likewise.
11794 * config/rs6000/mmintrin.h: Likewise.
11795 * config/rs6000/xmmintrin.h: Likewise.
11797 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11800 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11802 * tree.c (tree_nop_conversion): Return true for location wrapper
11804 (maybe_wrap_with_location): New function.
11805 (selftest::check_strip_nops): New function.
11806 (selftest::test_location_wrappers): New function.
11807 (selftest::tree_c_tests): Call it.
11808 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11809 (maybe_wrap_with_location): New decl.
11810 (EXPR_LOCATION_WRAPPER_P): New macro.
11811 (location_wrapper_p): New inline function.
11812 (tree_strip_any_location_wrapper): New inline function.
11814 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11817 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11818 stack_realign_offset for the largest alignment of stack slot
11820 (ix86_find_max_used_stack_alignment): New function.
11821 (ix86_finalize_stack_frame_flags): Use it. Set
11822 max_used_stack_alignment if we don't realign stack.
11823 * config/i386/i386.h (machine_function): Add
11824 max_used_stack_alignment.
11826 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11828 * config/arm/arm.opt (-mbranch-cost): New option.
11829 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11832 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11835 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11836 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11838 2018-01-10 Richard Biener <rguenther@suse.de>
11841 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11842 early out so it also covers the case where we have a non-NULL
11845 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11847 PR tree-optimization/83753
11848 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11849 for non-strided grouped accesses if the number of elements is 1.
11851 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11854 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11855 * i386.h (TARGET_USE_GATHER): Define.
11856 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11858 2018-01-10 Martin Liska <mliska@suse.cz>
11861 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11862 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11864 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11865 CLEANUP_NO_PARTITIONING is not set.
11867 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11869 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11870 for vectors, as a partial revert of r254296.
11871 * rtl.h (const_vec_p): Delete.
11872 (const_vec_duplicate_p): Don't test for vector CONSTs.
11873 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11874 * expmed.c (make_tree): Likewise.
11877 * common.md (E, F): Use CONSTANT_P instead of checking for
11879 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11880 checking for CONST_VECTOR.
11882 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11884 PR middle-end/83575
11885 * predict.c (force_edge_cold): Handle in more sane way edges
11886 with no prediction.
11888 2018-01-09 Carl Love <cel@us.ibm.com>
11890 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11892 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11893 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11894 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11895 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11896 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11897 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11898 * config/rs6000/rs6000-protos.h: Add extern defition for
11899 rs6000_generate_float2_double_code.
11900 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11902 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11903 (float2_v2df): Add define_expand.
11905 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11908 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11909 op_mode in the force_to_mode call.
11911 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11913 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11914 instead of checking each element individually.
11915 (aarch64_evpc_uzp): Likewise.
11916 (aarch64_evpc_zip): Likewise.
11917 (aarch64_evpc_ext): Likewise.
11918 (aarch64_evpc_rev): Likewise.
11919 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11920 instead of checking each element individually. Return true without
11922 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11923 whether all selected elements come from the same input, instead of
11924 checking each element individually. Remove calls to gen_rtx_REG,
11925 start_sequence and end_sequence and instead assert that no rtl is
11928 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11930 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11931 order of HIGH and CONST checks.
11933 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11935 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11936 if the destination isn't an SSA_NAME.
11938 2018-01-09 Richard Biener <rguenther@suse.de>
11940 PR tree-optimization/83668
11941 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11943 (canonicalize_loop_form): ... here, renamed from ...
11944 (canonicalize_loop_closed_ssa_form): ... this and amended to
11945 swap successor edges for loop exit blocks to make us use
11946 the RPO order we need for initial schedule generation.
11948 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11950 PR tree-optimization/64811
11951 * match.pd: When optimizing comparisons with Inf, avoid
11952 introducing or losing exceptions from comparisons with NaN.
11954 2018-01-09 Martin Liska <mliska@suse.cz>
11957 * asan.c (shadow_mem_size): Add gcc_assert.
11959 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11961 Don't save registers in main().
11964 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11965 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11966 * config/avr/avr.c (avr_set_current_function): Don't error if
11967 naked, OS_task or OS_main are specified at the same time.
11968 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11970 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11972 * common/config/avr/avr-common.c (avr_option_optimization_table):
11973 Switch on -mmain-is-OS_task for optimizing compilations.
11975 2018-01-09 Richard Biener <rguenther@suse.de>
11977 PR tree-optimization/83572
11978 * graphite.c: Include cfganal.h.
11979 (graphite_transform_loops): Connect infinite loops to exit
11980 and remove fake edges at the end.
11982 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11984 * ipa-inline.c (edge_badness): Revert accidental checkin.
11986 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11989 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11990 symbols; not inline clones.
11992 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11995 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11996 hard registers. Formatting fixes.
11998 PR preprocessor/83722
11999 * gcc.c (try_generate_repro): Pass
12000 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12001 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12004 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12005 Kito Cheng <kito.cheng@gmail.com>
12007 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12008 (riscv_leaf_function_p): Delete.
12009 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12011 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12013 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12015 (do_ifelse): New function.
12016 (do_isel): New function.
12017 (do_sub3): New function.
12018 (do_add3): New function.
12019 (do_load_mask_compare): New function.
12020 (do_overlap_load_compare): New function.
12021 (expand_compare_loop): New function.
12022 (expand_block_compare): Call expand_compare_loop() when appropriate.
12023 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12024 option description.
12025 (-mblock-compare-inline-loop-limit): New option.
12027 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12030 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12031 Reverse order of second and third operands in first alternative.
12032 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12033 of first and second elements in UNSPEC_VPERMR vector.
12034 (altivec_expand_vec_perm_le): Likewise.
12036 2018-01-08 Jeff Law <law@redhat.com>
12038 PR rtl-optimizatin/81308
12039 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12040 (process_switch): If group_case_labels makes a change, then set
12042 (pass_convert_switch::execute): If a switch is converted, then
12043 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12045 PR rtl-optimization/81308
12046 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12049 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12051 PR target/83663 - Revert r255946
12052 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12053 generation for cases where splatting a value is not useful.
12054 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12055 across a vec_duplicate and a paradoxical subreg forming a vector
12056 mode to a vec_concat.
12058 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12060 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12061 -march=armv8.3-a variants.
12062 * config/arm/t-multilib: Likewise.
12063 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12065 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12067 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12069 (cceq_ior_compare_complement): Give it a name so I can use it, and
12070 change boolean_or_operator predicate to boolean_operator so it can
12071 be used to generate a crand.
12072 (eqne): New code iterator.
12073 (bd/bd_neg): New code_attrs.
12074 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12075 a single define_insn.
12076 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12077 decrement (bdnzt/bdnzf/bdzt/bdzf).
12078 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12079 with the new names of the branch decrement patterns, and added the
12080 names of the branch decrement conditional patterns.
12082 2018-01-08 Richard Biener <rguenther@suse.de>
12084 PR tree-optimization/83563
12085 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12088 2018-01-08 Richard Biener <rguenther@suse.de>
12090 PR middle-end/83713
12091 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12093 2018-01-08 Richard Biener <rguenther@suse.de>
12095 PR tree-optimization/83685
12096 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12097 references to abnormals.
12099 2018-01-08 Richard Biener <rguenther@suse.de>
12102 * dwarf2out.c (output_indirect_strings): Handle empty
12103 skeleton_debug_str_hash.
12104 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12106 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12108 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12109 (emit_store_direct): Likewise.
12110 (arc_trampoline_adjust_address): Likewise.
12111 (arc_asm_trampoline_template): New function.
12112 (arc_initialize_trampoline): Use asm_trampoline_template.
12113 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12114 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12115 * config/arc/arc.md (flush_icache): Delete pattern.
12117 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12119 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12120 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12123 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12126 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12127 by not USED_FOR_TARGET.
12128 (make_pass_resolve_sw_modes): Likewise.
12130 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12132 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12135 2018-01-08 Richard Biener <rguenther@suse.de>
12137 PR middle-end/83580
12138 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12140 2018-01-08 Richard Biener <rguenther@suse.de>
12142 PR middle-end/83517
12143 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12145 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12147 PR middle-end/81897
12148 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12149 basic blocks with a small number of successors.
12150 (convert_control_dep_chain_into_preds): Improve handling of
12152 (dump_predicates): Split apart into...
12153 (dump_pred_chain): ...here...
12154 (dump_pred_info): ...and here.
12155 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12156 (can_chain_union_be_invalidated_p): Improve check for invalidation
12158 (uninit_uses_cannot_happen): Avoid unnecessary if
12159 convert_control_dep_chain_into_preds yielded nothing.
12161 2018-01-06 Martin Sebor <msebor@redhat.com>
12163 PR tree-optimization/83640
12164 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12165 subtracting negative offset from size.
12166 (builtin_access::overlap): Adjust offset bounds of the access to fall
12167 within the size of the object if possible.
12169 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12171 PR rtl-optimization/83699
12172 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12173 extract_bit_field_as_subreg to cases in which the extracted
12174 value is also a vector.
12176 * lra-constraints.c (process_alt_operands): Test for the equivalence
12177 substitutions when detecting a possible reload cycle.
12179 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12182 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12183 by default if flag_selective_schedling{,2}. Formatting fixes.
12185 PR rtl-optimization/83682
12186 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12187 if it has non-VECTOR_MODE element mode.
12188 (vec_duplicate_p): Likewise.
12190 PR middle-end/83694
12191 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12192 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12194 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12197 * config/i386/i386-builtin.def
12198 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12199 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12200 Require also OPTION_MASK_ISA_AVX512F in addition to
12201 OPTION_MASK_ISA_GFNI.
12202 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12203 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12204 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12205 to OPTION_MASK_ISA_GFNI.
12206 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12207 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12208 OPTION_MASK_ISA_AVX512BW.
12209 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12210 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12211 addition to OPTION_MASK_ISA_GFNI.
12212 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12213 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12214 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12215 to OPTION_MASK_ISA_GFNI.
12216 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12217 a requirement for all ISAs rather than any of them with a few
12219 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12221 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12222 bitmasks to be enabled with 3 exceptions, instead of requiring any
12223 enabled ISA with lots of exceptions.
12224 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12225 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12226 Change avx512bw in isa attribute to avx512f.
12227 * config/i386/sgxintrin.h: Add license boilerplate.
12228 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12229 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12230 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12231 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12233 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12234 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12235 temporarily sse2 rather than sse if not enabled already.
12238 * config/i386/sse.md (VI248_VLBW): Rename to ...
12239 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12240 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12241 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12242 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12243 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12244 mode iterator instead of VI248_VLBW.
12246 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12248 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12249 (record_modified): Skip clobbers; add debug output.
12250 (param_change_prob): Use sreal frequencies.
12252 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12254 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12255 punt for user-aligned variables.
12257 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12259 * tree-chrec.c (chrec_contains_symbols): Return true for
12262 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12265 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12266 of (x|y) == x for BICS pattern.
12268 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12270 PR tree-optimization/83605
12271 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12272 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12275 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12277 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12278 * config/epiphany/rtems.h: New file.
12280 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12281 Uros Bizjak <ubizjak@gmail.com>
12284 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12285 QIreg_operand instead of register_operand predicate.
12286 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12287 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12288 comments instead of -fmitigate[-_]rop.
12290 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12293 * cgraphunit.c (symbol_table::compile): Switch to text_section
12294 before calling assembly_start debug hook.
12295 * run-rtl-passes.c (run_rtl_passes): Likewise.
12298 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12300 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12301 range_int_cst_p rather than !symbolic_range_p before calling
12302 extract_range_from_multiplicative_op_1.
12304 2018-01-04 Jeff Law <law@redhat.com>
12306 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12307 redundant test in assertion.
12309 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12311 * doc/rtl.texi: Document machine_mode wrapper classes.
12313 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12315 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12316 using tree_to_uhwi.
12318 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12320 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12321 the VEC_PERM_EXPR fold to fail.
12323 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12326 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12327 to switched_sections.
12329 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12332 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12333 test for d.testing.
12335 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12338 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12339 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12341 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12344 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12345 is BLKmode and bitpos not zero or mode change is needed.
12347 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12350 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12353 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12356 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12357 instead of MULT rtx. Update all corresponding splitters.
12358 (*saddl_se): Ditto.
12359 (*ssub<modesuffix>): Ditto.
12360 (*ssubl_se): Ditto.
12361 (*cmp_sadd_di): Update split patterns.
12362 (*cmp_sadd_si): Ditto.
12363 (*cmp_sadd_sidi): Ditto.
12364 (*cmp_ssub_di): Ditto.
12365 (*cmp_ssub_si): Ditto.
12366 (*cmp_ssub_sidi): Ditto.
12367 * config/alpha/predicates.md (const23_operand): New predicate.
12368 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12369 Look for ASHIFT, not MULT inner operand.
12370 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12372 2018-01-04 Martin Liska <mliska@suse.cz>
12374 PR gcov-profile/83669
12375 * gcov.c (output_intermediate_file): Add version to intermediate
12377 * doc/gcov.texi: Document new field 'version' in intermediate
12378 file format. Fix location of '-k' option of gcov command.
12380 2018-01-04 Martin Liska <mliska@suse.cz>
12383 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12385 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12387 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12389 2018-01-03 Martin Sebor <msebor@redhat.com>
12391 PR tree-optimization/83655
12392 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12393 checking calls with invalid arguments.
12395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12397 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12398 (vectorizable_mask_load_store): Delete.
12399 (vectorizable_call): Return false for masked loads and stores.
12400 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12401 instead of gimple_assign_rhs1.
12402 (vectorizable_load): Handle IFN_MASK_LOAD.
12403 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12407 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12409 (vectorizable_mask_load_store): ...here.
12410 (vectorizable_load): ...and here.
12412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12414 * tree-vect-stmts.c (vect_build_all_ones_mask)
12415 (vect_build_zero_merge_argument): New functions, split out from...
12416 (vectorizable_load): ...here.
12418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12420 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12422 (vectorizable_mask_load_store): ...here.
12423 (vectorizable_store): ...and here.
12425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12427 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12429 (vectorizable_mask_load_store): ...here.
12431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12433 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12434 (vect_model_store_cost): Take a vec_load_store_type instead of a
12436 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12437 (vect_model_store_cost): Take a vec_load_store_type instead of a
12439 (vectorizable_mask_load_store): Update accordingly.
12440 (vectorizable_store): Likewise.
12441 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
12443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12445 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
12446 IFN_MASK_LOAD calls here rather than...
12447 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
12449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12450 Alan Hayward <alan.hayward@arm.com>
12451 David Sherwood <david.sherwood@arm.com>
12453 * expmed.c (extract_bit_field_1): For vector extracts,
12454 fall back to extract_bit_field_as_subreg if vec_extract
12457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12458 Alan Hayward <alan.hayward@arm.com>
12459 David Sherwood <david.sherwood@arm.com>
12461 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
12462 they are variable or constant sized.
12463 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
12464 slots for constant-sized data.
12466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12467 Alan Hayward <alan.hayward@arm.com>
12468 David Sherwood <david.sherwood@arm.com>
12470 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
12471 handling COND_EXPRs with boolean comparisons, try to find a better
12472 basis for the mask type than the boolean itself.
12474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12476 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
12477 is calculated and how it can be overridden.
12478 * genmodes.c (max_bitsize_mode_any_mode): New variable.
12479 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
12481 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
12484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12485 Alan Hayward <alan.hayward@arm.com>
12486 David Sherwood <david.sherwood@arm.com>
12488 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
12489 Remove the mode argument.
12490 (aarch64_simd_valid_immediate): Remove the mode and inverse
12492 * config/aarch64/iterators.md (bitsize): New iterator.
12493 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
12494 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
12495 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
12496 aarch64_simd_valid_immediate.
12497 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
12498 (aarch64_reg_or_bic_imm): Likewise.
12499 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
12500 with an insn_type enum and msl with a modifier_type enum.
12501 Replace element_width with a scalar_mode. Change the shift
12502 to unsigned int. Add constructors for scalar_float_mode and
12503 scalar_int_mode elements.
12504 (aarch64_vect_float_const_representable_p): Delete.
12505 (aarch64_can_const_movi_rtx_p)
12506 (aarch64_simd_scalar_immediate_valid_for_move)
12507 (aarch64_simd_make_constant): Update call to
12508 aarch64_simd_valid_immediate.
12509 (aarch64_advsimd_valid_immediate_hs): New function.
12510 (aarch64_advsimd_valid_immediate): Likewise.
12511 (aarch64_simd_valid_immediate): Remove mode and inverse
12512 arguments. Rewrite to use the above. Use const_vec_duplicate_p
12513 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
12514 and aarch64_float_const_representable_p on the result.
12515 (aarch64_output_simd_mov_immediate): Remove mode argument.
12516 Update call to aarch64_simd_valid_immediate and use of
12517 simd_immediate_info.
12518 (aarch64_output_scalar_simd_mov_immediate): Update call
12521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12522 Alan Hayward <alan.hayward@arm.com>
12523 David Sherwood <david.sherwood@arm.com>
12525 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
12526 (mode_nunits): Likewise CONST_MODE_NUNITS.
12527 * machmode.def (ADJUST_NUNITS): Document.
12528 * genmodes.c (mode_data::need_nunits_adj): New field.
12529 (blank_mode): Update accordingly.
12530 (adj_nunits): New variable.
12531 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
12533 (emit_mode_size_inline): Set need_bytesize_adj for all modes
12534 listed in adj_nunits.
12535 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
12536 listed in adj_nunits. Don't emit case statements for such modes.
12537 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
12538 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
12539 nothing if adj_nunits is nonnull.
12540 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
12541 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
12542 (emit_mode_fbit): Update use of print_maybe_const_decl.
12543 (emit_move_size): Likewise. Treat the array as non-const
12545 (emit_mode_adjustments): Handle adj_nunits.
12547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12549 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
12550 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
12551 (VECTOR_MODES): Use it.
12552 (make_vector_modes): Take the prefix as an argument.
12554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12555 Alan Hayward <alan.hayward@arm.com>
12556 David Sherwood <david.sherwood@arm.com>
12558 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
12559 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
12560 for MODE_VECTOR_BOOL.
12561 * machmode.def (VECTOR_BOOL_MODE): Document.
12562 * genmodes.c (VECTOR_BOOL_MODE): New macro.
12563 (make_vector_bool_mode): New function.
12564 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
12566 * lto-streamer-in.c (lto_input_mode_table): Likewise.
12567 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
12569 * stor-layout.c (int_mode_for_mode): Likewise.
12570 * tree.c (build_vector_type_for_mode): Likewise.
12571 * varasm.c (output_constant_pool_2): Likewise.
12572 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
12573 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
12574 for MODE_VECTOR_BOOL.
12575 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
12576 of mode class checks.
12577 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
12578 instead of a list of mode class checks.
12579 (expand_vector_scalar_condition): Likewise.
12580 (type_for_widest_vector_mode): Handle BImode as an inner mode.
12582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12583 Alan Hayward <alan.hayward@arm.com>
12584 David Sherwood <david.sherwood@arm.com>
12586 * machmode.h (mode_size): Change from unsigned short to
12588 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
12589 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12590 or if measurement_type is not polynomial.
12591 (fixed_size_mode::includes_p): Check for constant-sized modes.
12592 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
12593 return a poly_uint16 rather than an unsigned short.
12594 (emit_mode_size): Change the type of mode_size from unsigned short
12595 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
12596 (emit_mode_adjustments): Cope with polynomial vector sizes.
12597 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12599 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12601 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
12602 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
12603 * caller-save.c (setup_save_areas): Likewise.
12604 (replace_reg_with_saved_mem): Likewise.
12605 * calls.c (emit_library_call_value_1): Likewise.
12606 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
12607 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
12608 (gen_lowpart_for_combine): Likewise.
12609 * convert.c (convert_to_integer_1): Likewise.
12610 * cse.c (equiv_constant, cse_insn): Likewise.
12611 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
12612 (cselib_subst_to_values): Likewise.
12613 * dce.c (word_dce_process_block): Likewise.
12614 * df-problems.c (df_word_lr_mark_ref): Likewise.
12615 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
12616 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
12617 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
12618 (rtl_for_decl_location): Likewise.
12619 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
12620 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
12621 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
12622 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
12623 (expand_expr_real_1): Likewise.
12624 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12625 (pad_below): Likewise.
12626 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12627 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12628 * ira.c (get_subreg_tracking_sizes): Likewise.
12629 * ira-build.c (ira_create_allocno_objects): Likewise.
12630 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12631 (ira_sort_regnos_for_alter_reg): Likewise.
12632 * ira-costs.c (record_operand_costs): Likewise.
12633 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12634 (resolve_simple_move): Likewise.
12635 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12636 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12637 (lra_constraints): Likewise.
12638 (CONST_POOL_OK_P): Reject variable-sized modes.
12639 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12640 (add_pseudo_to_slot, lra_spill): Likewise.
12641 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12642 * optabs-query.c (get_best_extraction_insn): Likewise.
12643 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12644 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12645 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12646 * recog.c (offsettable_address_addr_space_p): Likewise.
12647 * regcprop.c (maybe_mode_change): Likewise.
12648 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12649 * regrename.c (build_def_use): Likewise.
12650 * regstat.c (dump_reg_info): Likewise.
12651 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12652 (find_reloads, find_reloads_subreg_address): Likewise.
12653 * reload1.c (eliminate_regs_1): Likewise.
12654 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12655 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12656 (simplify_binary_operation_1, simplify_subreg): Likewise.
12657 * targhooks.c (default_function_arg_padding): Likewise.
12658 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12659 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12660 (verify_gimple_assign_ternary): Likewise.
12661 * tree-inline.c (estimate_move_cost): Likewise.
12662 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12663 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12664 (get_address_cost_ainc): Likewise.
12665 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12666 (vect_supportable_dr_alignment): Likewise.
12667 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12668 (vectorizable_reduction): Likewise.
12669 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12670 (vectorizable_operation, vectorizable_load): Likewise.
12671 * tree.c (build_same_sized_truth_vector_type): Likewise.
12672 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12673 * var-tracking.c (emit_note_insn_var_location): Likewise.
12674 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12675 (ADDR_VEC_ALIGN): Likewise.
12677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12678 Alan Hayward <alan.hayward@arm.com>
12679 David Sherwood <david.sherwood@arm.com>
12681 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12683 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12684 or if measurement_type is polynomial.
12685 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12686 * combine.c (make_extraction): Likewise.
12687 * dse.c (find_shift_sequence): Likewise.
12688 * dwarf2out.c (mem_loc_descriptor): Likewise.
12689 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12690 (extract_bit_field, extract_low_bits): Likewise.
12691 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12692 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12693 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12694 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12695 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12696 * reload.c (find_reloads): Likewise.
12697 * reload1.c (alter_reg): Likewise.
12698 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12699 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12700 * tree-if-conv.c (predicate_mem_writes): Likewise.
12701 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12702 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12703 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12704 * valtrack.c (dead_debug_insert_temp): Likewise.
12705 * varasm.c (mergeable_constant_section): Likewise.
12706 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12709 Alan Hayward <alan.hayward@arm.com>
12710 David Sherwood <david.sherwood@arm.com>
12712 * expr.c (expand_assignment): Cope with polynomial mode sizes
12713 when assigning to a CONCAT.
12715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12716 Alan Hayward <alan.hayward@arm.com>
12717 David Sherwood <david.sherwood@arm.com>
12719 * machmode.h (mode_precision): Change from unsigned short to
12721 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12723 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12724 or if measurement_type is not polynomial.
12725 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12726 in which the mode is already known to be a scalar_int_mode.
12727 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12728 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12730 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12731 for GET_MODE_PRECISION.
12732 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12733 for GET_MODE_PRECISION.
12734 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12736 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12737 (expand_field_assignment, make_extraction): Likewise.
12738 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12739 (get_last_value): Likewise.
12740 * convert.c (convert_to_integer_1): Likewise.
12741 * cse.c (cse_insn): Likewise.
12742 * expr.c (expand_expr_real_1): Likewise.
12743 * lra-constraints.c (simplify_operand_subreg): Likewise.
12744 * optabs-query.c (can_atomic_load_p): Likewise.
12745 * optabs.c (expand_atomic_load): Likewise.
12746 (expand_atomic_store): Likewise.
12747 * ree.c (combine_reaching_defs): Likewise.
12748 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12749 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12750 * tree.h (type_has_mode_precision_p): Likewise.
12751 * ubsan.c (instrument_si_overflow): Likewise.
12753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12754 Alan Hayward <alan.hayward@arm.com>
12755 David Sherwood <david.sherwood@arm.com>
12757 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12758 polynomial numbers of units.
12759 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12760 (valid_vector_subparts_p): New function.
12761 (build_vector_type): Remove temporary shim and take the number
12762 of units as a poly_uint64 rather than an int.
12763 (build_opaque_vector_type): Take the number of units as a
12764 poly_uint64 rather than an int.
12765 * tree.c (build_vector_from_ctor): Handle polynomial
12766 TYPE_VECTOR_SUBPARTS.
12767 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12768 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12769 (build_vector_from_val): If the number of units is variable,
12770 use build_vec_duplicate_cst for constant operands and
12771 VEC_DUPLICATE_EXPR otherwise.
12772 (make_vector_type): Remove temporary is_constant ().
12773 (build_vector_type, build_opaque_vector_type): Take the number of
12774 units as a poly_uint64 rather than an int.
12775 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12777 * cfgexpand.c (expand_debug_expr): Likewise.
12778 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12779 (store_constructor, expand_expr_real_1): Likewise.
12780 (const_scalar_mask_from_tree): Likewise.
12781 * fold-const-call.c (fold_const_reduction): Likewise.
12782 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12783 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12784 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12785 (fold_relational_const): Likewise.
12786 (native_interpret_vector): Likewise. Change the size from an
12787 int to an unsigned int.
12788 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12789 TYPE_VECTOR_SUBPARTS.
12790 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12791 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12792 duplicating a non-constant operand into a variable-length vector.
12793 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12794 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12795 * ipa-icf.c (sem_variable::equals): Likewise.
12796 * match.pd: Likewise.
12797 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12798 * print-tree.c (print_node): Likewise.
12799 * stor-layout.c (layout_type): Likewise.
12800 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12801 * tree-cfg.c (verify_gimple_comparison): Likewise.
12802 (verify_gimple_assign_binary): Likewise.
12803 (verify_gimple_assign_ternary): Likewise.
12804 (verify_gimple_assign_single): Likewise.
12805 * tree-pretty-print.c (dump_generic_node): Likewise.
12806 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12807 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12808 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12809 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12810 (vect_shift_permute_load_chain): Likewise.
12811 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12812 (expand_vector_condition, optimize_vector_constructor): Likewise.
12813 (lower_vec_perm, get_compute_type): Likewise.
12814 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12815 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12816 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12817 (vect_recog_mask_conversion_pattern): Likewise.
12818 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12819 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12820 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12821 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12822 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12823 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12824 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12825 (supportable_widening_operation): Likewise.
12826 (supportable_narrowing_operation): Likewise.
12827 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12829 * varasm.c (output_constant): Likewise.
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 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12836 so that both the length == 3 and length != 3 cases set up their
12837 own permute vectors. Add comments explaining why we know the
12838 number of elements is constant.
12839 (vect_permute_load_chain): Likewise.
12841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12842 Alan Hayward <alan.hayward@arm.com>
12843 David Sherwood <david.sherwood@arm.com>
12845 * machmode.h (mode_nunits): Change from unsigned char to
12847 (ONLY_FIXED_SIZE_MODES): New macro.
12848 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12849 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12850 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12852 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12853 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12854 or if measurement_type is not polynomial.
12855 * genmodes.c (ZERO_COEFFS): New macro.
12856 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12858 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12859 Use ZERO_COEFFS when emitting initializers.
12860 * data-streamer.h (bp_pack_poly_value): New function.
12861 (bp_unpack_poly_value): Likewise.
12862 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12863 for GET_MODE_NUNITS.
12864 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12865 for GET_MODE_NUNITS.
12866 * tree.c (make_vector_type): Remove temporary shim and make
12867 the real function take the number of units as a poly_uint64
12868 rather than an int.
12869 (build_vector_type_for_mode): Handle polynomial nunits.
12870 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12871 * emit-rtl.c (const_vec_series_p_1): Likewise.
12872 (gen_rtx_CONST_VECTOR): Likewise.
12873 * fold-const.c (test_vec_duplicate_folding): Likewise.
12874 * genrecog.c (validate_pattern): Likewise.
12875 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12876 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12877 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12878 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12879 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12880 * rtlanal.c (subreg_get_info): Likewise.
12881 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12882 (vect_grouped_load_supported): Likewise.
12883 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12884 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12885 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12886 (simplify_const_unary_operation, simplify_binary_operation_1)
12887 (simplify_const_binary_operation, simplify_ternary_operation)
12888 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12889 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12890 instead of CONST_VECTOR_NUNITS.
12891 * varasm.c (output_constant_pool_2): Likewise.
12892 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12893 explicit-encoded elements in the XVEC for variable-length vectors.
12895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12897 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12900 Alan Hayward <alan.hayward@arm.com>
12901 David Sherwood <david.sherwood@arm.com>
12903 * coretypes.h (fixed_size_mode): Declare.
12904 (fixed_size_mode_pod): New typedef.
12905 * builtins.h (target_builtins::x_apply_args_mode)
12906 (target_builtins::x_apply_result_mode): Change type to
12907 fixed_size_mode_pod.
12908 * builtins.c (apply_args_size, apply_result_size, result_vector)
12909 (expand_builtin_apply_args_1, expand_builtin_apply)
12910 (expand_builtin_return): Update accordingly.
12912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12914 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12915 * cselib.c (cselib_hash_rtx): Likewise.
12916 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12917 CONST_VECTOR encoding.
12919 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12920 Jeff Law <law@redhat.com>
12923 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12924 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12925 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12926 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12929 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12930 explicitly probe *sp in a noreturn function if there were any callee
12931 register saves or frame pointer is needed.
12933 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12936 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12937 BLKmode for ternary, binary or unary expressions.
12940 * var-tracking.c (delete_vta_debug_insn): New inline function.
12941 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12942 insns from get_insns () to NULL instead of each bb separately.
12943 Use delete_vta_debug_insn. No longer static.
12944 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12945 delete_vta_debug_insns callers.
12946 * rtl.h (delete_vta_debug_insns): Declare.
12947 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12948 instead of variable_tracking_main.
12950 2018-01-03 Martin Sebor <msebor@redhat.com>
12952 PR tree-optimization/83603
12953 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12954 arguments past the endof the argument list in functions declared
12955 without a prototype.
12956 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12957 Avoid checking when arguments are null.
12959 2018-01-03 Martin Sebor <msebor@redhat.com>
12962 * doc/extend.texi (attribute const): Fix a typo.
12963 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12964 issuing -Wsuggest-attribute for void functions.
12966 2018-01-03 Martin Sebor <msebor@redhat.com>
12968 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12969 offset_int::from instead of wide_int::to_shwi.
12970 (maybe_diag_overlap): Remove assertion.
12971 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12972 * gimple-ssa-sprintf.c (format_directive): Same.
12973 (parse_directive): Same.
12974 (sprintf_dom_walker::compute_format_length): Same.
12975 (try_substitute_return_value): Same.
12977 2018-01-03 Jeff Law <law@redhat.com>
12979 PR middle-end/83654
12980 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12981 non-constant residual for zero at runtime and avoid probing in
12982 that case. Reorganize code for trailing problem to mirror handling
12985 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12987 PR tree-optimization/83501
12988 * tree-ssa-strlen.c (get_string_cst): New.
12989 (handle_char_store): Call get_string_cst.
12991 2018-01-03 Martin Liska <mliska@suse.cz>
12993 PR tree-optimization/83593
12994 * tree-ssa-strlen.c: Include tree-cfg.h.
12995 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12996 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12997 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12999 (strlen_dom_walker::before_dom_children): Call
13000 gimple_purge_dead_eh_edges. Dump tranformation with details
13002 (strlen_dom_walker::before_dom_children): Update call by adding
13003 new argument cleanup_eh.
13004 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13006 2018-01-03 Martin Liska <mliska@suse.cz>
13009 * cif-code.def (VARIADIC_THUNK): New enum value.
13010 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13013 2018-01-03 Jan Beulich <jbeulich@suse.com>
13015 * sse.md (mov<mode>_internal): Tighten condition for when to use
13016 vmovdqu<ssescalarsize> for TI and OI modes.
13018 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13020 Update copyright years.
13022 2018-01-03 Martin Liska <mliska@suse.cz>
13025 * ipa-visibility.c (function_and_variable_visibility): Skip
13026 functions with noipa attribure.
13028 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13030 * gcc.c (process_command): Update copyright notice dates.
13031 * gcov-dump.c (print_version): Ditto.
13032 * gcov.c (print_version): Ditto.
13033 * gcov-tool.c (print_version): Ditto.
13034 * gengtype.c (create_file): Ditto.
13035 * doc/cpp.texi: Bump @copying's copyright year.
13036 * doc/cppinternals.texi: Ditto.
13037 * doc/gcc.texi: Ditto.
13038 * doc/gccint.texi: Ditto.
13039 * doc/gcov.texi: Ditto.
13040 * doc/install.texi: Ditto.
13041 * doc/invoke.texi: Ditto.
13043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13045 * vector-builder.h (vector_builder::m_full_nelts): Change from
13046 unsigned int to poly_uint64.
13047 (vector_builder::full_nelts): Update prototype accordingly.
13048 (vector_builder::new_vector): Likewise.
13049 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13050 (vector_builder::operator ==): Likewise.
13051 (vector_builder::finalize): Likewise.
13052 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13053 Take the number of elements as a poly_uint64 rather than an
13055 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13056 from unsigned int to poly_uint64.
13057 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13058 (vec_perm_indices::new_vector): Likewise.
13059 (vec_perm_indices::length): Likewise.
13060 (vec_perm_indices::nelts_per_input): Likewise.
13061 (vec_perm_indices::input_nelts): Likewise.
13062 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13063 number of elements per input as a poly_uint64 rather than an
13064 unsigned int. Use the original encoding for variable-length
13065 vectors, rather than clamping each individual element.
13066 For the second and subsequent elements in each pattern,
13067 clamp the step and base before clamping their sum.
13068 (vec_perm_indices::series_p): Handle polynomial element counts.
13069 (vec_perm_indices::all_in_range_p): Likewise.
13070 (vec_perm_indices_to_tree): Likewise.
13071 (vec_perm_indices_to_rtx): Likewise.
13072 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13073 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13074 (tree_vector_builder::new_binary_operation): Handle polynomial
13075 element counts. Return false if we need to know the number
13076 of elements at compile time.
13077 * fold-const.c (fold_vec_perm): Punt if the number of elements
13078 isn't known at compile time.
13080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13082 * vec-perm-indices.h (vec_perm_builder): Change element type
13083 from HOST_WIDE_INT to poly_int64.
13084 (vec_perm_indices::element_type): Update accordingly.
13085 (vec_perm_indices::clamp): Handle polynomial element_types.
13086 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13087 (vec_perm_indices::all_in_range_p): Likewise.
13088 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13090 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13091 polynomial vec_perm_indices element types.
13092 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13093 * fold-const.c (fold_vec_perm): Likewise.
13094 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13095 * tree-vect-generic.c (lower_vec_perm): Likewise.
13096 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13097 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13098 element type to HOST_WIDE_INT.
13100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13101 Alan Hayward <alan.hayward@arm.com>
13102 David Sherwood <david.sherwood@arm.com>
13104 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13105 rather than an int. Use plus_constant.
13106 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13107 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13110 Alan Hayward <alan.hayward@arm.com>
13111 David Sherwood <david.sherwood@arm.com>
13113 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13114 a HOST_WIDE_INT to a poly_int64.
13116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13117 Alan Hayward <alan.hayward@arm.com>
13118 David Sherwood <david.sherwood@arm.com>
13120 * calls.c (load_register_parameters): Cope with polynomial
13121 mode sizes. Require a constant size for BLKmode parameters
13122 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13123 forces a parameter to be padded at the lsb end in order to
13124 fill a complete number of words, require the parameter size
13125 to be ordered wrt UNITS_PER_WORD.
13127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13128 Alan Hayward <alan.hayward@arm.com>
13129 David Sherwood <david.sherwood@arm.com>
13131 * reload1.c (spill_stack_slot_width): Change element type
13132 from unsigned int to poly_uint64_pod.
13133 (alter_reg): Treat mode sizes as polynomial.
13135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13136 Alan Hayward <alan.hayward@arm.com>
13137 David Sherwood <david.sherwood@arm.com>
13139 * reload.c (complex_word_subreg_p): New function.
13140 (reload_inner_reg_of_subreg, push_reload): Use it.
13142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13143 Alan Hayward <alan.hayward@arm.com>
13144 David Sherwood <david.sherwood@arm.com>
13146 * lra-constraints.c (process_alt_operands): Reject matched
13147 operands whose sizes aren't ordered.
13148 (match_reload): Refer to this check here.
13150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13151 Alan Hayward <alan.hayward@arm.com>
13152 David Sherwood <david.sherwood@arm.com>
13154 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13155 that the mode size is in the set {1, 2, 4, 8, 16}.
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 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13162 Use plus_constant instead of gen_rtx_PLUS.
13164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13165 Alan Hayward <alan.hayward@arm.com>
13166 David Sherwood <david.sherwood@arm.com>
13168 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13169 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13170 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13171 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13172 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13173 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13174 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13175 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13176 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13177 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13179 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13180 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13181 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13182 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13183 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13184 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13185 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13186 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13187 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13188 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13190 * expr.c (emit_move_resolve_push): Treat the input and result
13191 of PUSH_ROUNDING as a poly_int64.
13192 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13193 (emit_push_insn): Likewise.
13194 * lra-eliminations.c (mark_not_eliminable): Likewise.
13195 * recog.c (push_operand): Likewise.
13196 * reload1.c (elimination_effects): Likewise.
13197 * rtlanal.c (nonzero_bits1): Likewise.
13198 * calls.c (store_one_arg): Likewise. Require the padding to be
13199 known at compile time.
13201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13202 Alan Hayward <alan.hayward@arm.com>
13203 David Sherwood <david.sherwood@arm.com>
13205 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13206 Use plus_constant instead of gen_rtx_PLUS.
13208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13209 Alan Hayward <alan.hayward@arm.com>
13210 David Sherwood <david.sherwood@arm.com>
13212 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13213 rather than an int.
13215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13216 Alan Hayward <alan.hayward@arm.com>
13217 David Sherwood <david.sherwood@arm.com>
13219 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13220 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13221 via stack temporaries. Treat the mode size as polynomial too.
13223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13224 Alan Hayward <alan.hayward@arm.com>
13225 David Sherwood <david.sherwood@arm.com>
13227 * expr.c (expand_expr_real_2): When handling conversions involving
13228 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13229 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13230 as a poly_uint64 too.
13232 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13233 Alan Hayward <alan.hayward@arm.com>
13234 David Sherwood <david.sherwood@arm.com>
13236 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13239 Alan Hayward <alan.hayward@arm.com>
13240 David Sherwood <david.sherwood@arm.com>
13242 * combine.c (can_change_dest_mode): Handle polynomial
13243 REGMODE_NATURAL_SIZE.
13244 * expmed.c (store_bit_field_1): Likewise.
13245 * expr.c (store_constructor): Likewise.
13246 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13247 and polynomial REGMODE_NATURAL_SIZE.
13248 (gen_lowpart_common): Likewise.
13249 * reginfo.c (record_subregs_of_mode): Likewise.
13250 * rtlanal.c (read_modify_subreg_p): Likewise.
13252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13253 Alan Hayward <alan.hayward@arm.com>
13254 David Sherwood <david.sherwood@arm.com>
13256 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13257 numbers of elements.
13259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13260 Alan Hayward <alan.hayward@arm.com>
13261 David Sherwood <david.sherwood@arm.com>
13263 * match.pd: Cope with polynomial numbers of vector elements.
13265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13266 Alan Hayward <alan.hayward@arm.com>
13267 David Sherwood <david.sherwood@arm.com>
13269 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13270 in a POINTER_PLUS_EXPR.
13272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13273 Alan Hayward <alan.hayward@arm.com>
13274 David Sherwood <david.sherwood@arm.com>
13276 * omp-simd-clone.c (simd_clone_subparts): New function.
13277 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13278 (ipa_simd_modify_function_body): Likewise.
13280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13281 Alan Hayward <alan.hayward@arm.com>
13282 David Sherwood <david.sherwood@arm.com>
13284 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13285 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13286 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13287 (expand_vector_condition, vector_element): Likewise.
13288 (subparts_gt): New function.
13289 (get_compute_type): Use subparts_gt.
13290 (count_type_subparts): Delete.
13291 (expand_vector_operations_1): Use subparts_gt instead of
13292 count_type_subparts.
13294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13295 Alan Hayward <alan.hayward@arm.com>
13296 David Sherwood <david.sherwood@arm.com>
13298 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13299 (vect_compile_time_alias): ...this new function. Do the calculation
13300 on poly_ints rather than trees.
13301 (vect_prune_runtime_alias_test_list): Update call accordingly.
13303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13304 Alan Hayward <alan.hayward@arm.com>
13305 David Sherwood <david.sherwood@arm.com>
13307 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13309 (vect_schedule_slp_instance): Likewise.
13311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13312 Alan Hayward <alan.hayward@arm.com>
13313 David Sherwood <david.sherwood@arm.com>
13315 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13316 constant and extern definitions for variable-length vectors.
13317 (vect_get_constant_vectors): Note that the number of units
13318 is known to be constant.
13320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13321 Alan Hayward <alan.hayward@arm.com>
13322 David Sherwood <david.sherwood@arm.com>
13324 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13325 of units as polynomial. Choose between WIDE and NARROW based
13328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13329 Alan Hayward <alan.hayward@arm.com>
13330 David Sherwood <david.sherwood@arm.com>
13332 * tree-vect-stmts.c (simd_clone_subparts): New function.
13333 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13336 Alan Hayward <alan.hayward@arm.com>
13337 David Sherwood <david.sherwood@arm.com>
13339 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13340 vectors as polynomial. Use build_index_vector for
13341 IFN_GOMP_SIMD_LANE.
13343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13344 Alan Hayward <alan.hayward@arm.com>
13345 David Sherwood <david.sherwood@arm.com>
13347 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13348 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13349 for variable-length vectors.
13350 (vectorizable_mask_load_store): Treat the number of units as
13351 polynomial, asserting that it is constant if the condition has
13352 already been enforced.
13353 (vectorizable_store, vectorizable_load): Likewise.
13355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13356 Alan Hayward <alan.hayward@arm.com>
13357 David Sherwood <david.sherwood@arm.com>
13359 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13360 of units as polynomial. Punt if we can't tell at compile time
13361 which vector contains the final result.
13363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13364 Alan Hayward <alan.hayward@arm.com>
13365 David Sherwood <david.sherwood@arm.com>
13367 * tree-vect-loop.c (vectorizable_induction): Treat the number
13368 of units as polynomial. Punt on SLP inductions. Use an integer
13369 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13370 cast of such a series for variable-length floating-point
13373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13374 Alan Hayward <alan.hayward@arm.com>
13375 David Sherwood <david.sherwood@arm.com>
13377 * tree.h (build_index_vector): Declare.
13378 * tree.c (build_index_vector): New function.
13379 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13380 of units as polynomial, forcibly converting it to a constant if
13381 vectorizable_reduction has already enforced the condition.
13382 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13383 to create a {1,2,3,...} vector.
13384 (vectorizable_reduction): Treat the number of units as polynomial.
13385 Choose vectype_in based on the largest scalar element size rather
13386 than the smallest number of units. Enforce the restrictions
13389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13390 Alan Hayward <alan.hayward@arm.com>
13391 David Sherwood <david.sherwood@arm.com>
13393 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13394 number of units as polynomial.
13396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13397 Alan Hayward <alan.hayward@arm.com>
13398 David Sherwood <david.sherwood@arm.com>
13400 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13401 * target.def (autovectorize_vector_sizes): Return the vector sizes
13402 by pointer, using vector_sizes rather than a bitmask.
13403 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13404 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13405 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13407 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13408 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13409 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13410 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13411 * omp-general.c (omp_max_vf): Likewise.
13412 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13413 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13414 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13415 * tree-vect-slp.c (vect_slp_bb): Likewise.
13416 * doc/tm.texi: Regenerate.
13417 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13419 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13420 the vector size as a poly_uint64 rather than an unsigned int.
13421 (current_vector_size): Change from an unsigned int to a poly_uint64.
13422 (get_vectype_for_scalar_type): Update accordingly.
13423 * tree.h (build_truth_vector_type): Take the size and number of
13424 units as a poly_uint64 rather than an unsigned int.
13425 (build_vector_type): Add a temporary overload that takes
13426 the number of units as a poly_uint64 rather than an unsigned int.
13427 * tree.c (make_vector_type): Likewise.
13428 (build_truth_vector_type): Take the number of units as a poly_uint64
13429 rather than an unsigned int.
13431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13432 Alan Hayward <alan.hayward@arm.com>
13433 David Sherwood <david.sherwood@arm.com>
13435 * target.def (get_mask_mode): Take the number of units and length
13436 as poly_uint64s rather than unsigned ints.
13437 * targhooks.h (default_get_mask_mode): Update accordingly.
13438 * targhooks.c (default_get_mask_mode): Likewise.
13439 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13440 * doc/tm.texi: Regenerate.
13442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13443 Alan Hayward <alan.hayward@arm.com>
13444 David Sherwood <david.sherwood@arm.com>
13446 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
13447 * omp-general.c (omp_max_vf): Likewise.
13448 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
13449 (expand_omp_simd): Handle polynomial safelen.
13450 * omp-low.c (omplow_simd_context): Add a default constructor.
13451 (omplow_simd_context::max_vf): Change from int to poly_uint64.
13452 (lower_rec_simd_input_clauses): Update accordingly.
13453 (lower_rec_input_clauses): Likewise.
13455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13456 Alan Hayward <alan.hayward@arm.com>
13457 David Sherwood <david.sherwood@arm.com>
13459 * tree-vectorizer.h (vect_nunits_for_cost): New function.
13460 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
13461 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
13462 (vect_analyze_slp_cost): Likewise.
13463 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
13464 (vect_model_load_cost): Likewise.
13466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13467 Alan Hayward <alan.hayward@arm.com>
13468 David Sherwood <david.sherwood@arm.com>
13470 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
13471 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
13472 from an unsigned int * to a poly_uint64_pod *.
13473 (calculate_unrolling_factor): New function.
13474 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
13476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13477 Alan Hayward <alan.hayward@arm.com>
13478 David Sherwood <david.sherwood@arm.com>
13480 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
13481 from an unsigned int to a poly_uint64.
13482 (_loop_vec_info::slp_unrolling_factor): Likewise.
13483 (_loop_vec_info::vectorization_factor): Change from an int
13485 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
13486 (vect_get_num_vectors): New function.
13487 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
13488 (vect_get_num_copies): Use vect_get_num_vectors.
13489 (vect_analyze_data_ref_dependences): Change max_vf from an int *
13490 to an unsigned int *.
13491 (vect_analyze_data_refs): Change min_vf from an int * to a
13493 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13494 than an unsigned HOST_WIDE_INT.
13495 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
13496 (vect_analyze_data_ref_dependence): Change max_vf from an int *
13497 to an unsigned int *.
13498 (vect_analyze_data_ref_dependences): Likewise.
13499 (vect_compute_data_ref_alignment): Handle polynomial vf.
13500 (vect_enhance_data_refs_alignment): Likewise.
13501 (vect_prune_runtime_alias_test_list): Likewise.
13502 (vect_shift_permute_load_chain): Likewise.
13503 (vect_supportable_dr_alignment): Likewise.
13504 (dependence_distance_ge_vf): Take the vectorization factor as a
13505 poly_uint64 rather than an unsigned HOST_WIDE_INT.
13506 (vect_analyze_data_refs): Change min_vf from an int * to a
13508 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
13509 vfm1 as a poly_uint64 rather than an int. Make the same change
13510 for the returned bound_scalar.
13511 (vect_gen_vector_loop_niters): Handle polynomial vf.
13512 (vect_do_peeling): Likewise. Update call to
13513 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
13514 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
13516 * tree-vect-loop.c (vect_determine_vectorization_factor)
13517 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
13518 (vect_get_known_peeling_cost): Likewise.
13519 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
13520 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
13521 (vect_transform_loop): Likewise. Use the lowest possible VF when
13522 updating the upper bounds of the loop.
13523 (vect_min_worthwhile_factor): Make static. Return an unsigned int
13524 rather than an int.
13525 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
13526 polynomial unroll factors.
13527 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
13528 (vect_make_slp_decision): Likewise.
13529 (vect_supported_load_permutation_p): Likewise, and polynomial
13531 (vect_analyze_slp_cost): Handle polynomial vf.
13532 (vect_slp_analyze_node_operations): Likewise.
13533 (vect_slp_analyze_bb_1): Likewise.
13534 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
13535 than an unsigned HOST_WIDE_INT.
13536 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
13537 (vectorizable_load): Handle polynomial vf.
13538 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
13540 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
13542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13543 Alan Hayward <alan.hayward@arm.com>
13544 David Sherwood <david.sherwood@arm.com>
13546 * match.pd: Handle bit operations involving three constants
13547 and try to fold one pair.
13549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13551 * tree-vect-loop-manip.c: Include gimple-fold.h.
13552 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
13553 niters_maybe_zero parameters. Handle other cases besides a step of 1.
13554 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
13555 Add a path that uses a step of VF instead of 1, but disable it
13557 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
13558 and niters_no_overflow parameters. Update calls to
13559 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
13560 Create a new SSA name if the latter choses to use a ste other
13561 than zero, and return it via niters_vector_mult_vf_var.
13562 * tree-vect-loop.c (vect_transform_loop): Update calls to
13563 vect_do_peeling, vect_gen_vector_loop_niters and
13564 slpeel_make_loop_iterate_ntimes.
13565 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
13566 (vect_gen_vector_loop_niters): Update declarations after above changes.
13568 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
13570 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
13571 128-bit round to integer instructions.
13572 (ceil<mode>2): Likewise.
13573 (btrunc<mode>2): Likewise.
13574 (round<mode>2): Likewise.
13576 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13578 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
13579 unaligned VSX load/store on P8/P9.
13580 (expand_block_clear): Allow the use of unaligned VSX
13581 load/store on P8/P9.
13583 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13585 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
13587 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
13588 swap associated with both a load and a store.
13590 2018-01-02 Andrew Waterman <andrew@sifive.com>
13592 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
13593 * config/riscv/riscv.md (clear_cache): Use it.
13595 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
13597 * web.c: Remove out-of-date comment.
13599 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13601 * expr.c (fixup_args_size_notes): Check that any existing
13602 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
13603 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
13604 (emit_single_push_insn): ...here.
13606 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13608 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
13609 (const_vector_encoded_nelts): New function.
13610 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
13611 (const_vector_int_elt, const_vector_elt): Declare.
13612 * emit-rtl.c (const_vector_int_elt_1): New function.
13613 (const_vector_elt): Likewise.
13614 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
13615 of CONST_VECTOR_ELT.
13617 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13619 * expr.c: Include rtx-vector-builder.h.
13620 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
13621 directly on the tree encoding.
13622 (const_vector_from_tree): Likewise.
13623 * optabs.c: Include rtx-vector-builder.h.
13624 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13625 sequence of "u" values.
13626 * vec-perm-indices.c: Include rtx-vector-builder.h.
13627 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13628 directly on the vec_perm_indices encoding.
13630 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13632 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13633 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13634 * rtx-vector-builder.h: New file.
13635 * rtx-vector-builder.c: Likewise.
13636 * rtl.h (rtx_def::u2): Add a const_vector field.
13637 (CONST_VECTOR_NPATTERNS): New macro.
13638 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13639 (CONST_VECTOR_DUPLICATE_P): Likewise.
13640 (CONST_VECTOR_STEPPED_P): Likewise.
13641 (CONST_VECTOR_ENCODED_ELT): Likewise.
13642 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13643 (unwrap_const_vec_duplicate): Likewise.
13644 (const_vec_series_p): Check for a non-duplicated vector encoding.
13645 Say that the function only returns true for integer vectors.
13646 * emit-rtl.c: Include rtx-vector-builder.h.
13647 (gen_const_vec_duplicate_1): Delete.
13648 (gen_const_vector): Call gen_const_vec_duplicate instead of
13649 gen_const_vec_duplicate_1.
13650 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13651 (gen_const_vec_duplicate): Use rtx_vector_builder.
13652 (gen_const_vec_series): Likewise.
13653 (gen_rtx_CONST_VECTOR): Likewise.
13654 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13655 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13656 Build a new vector rather than modifying a CONST_VECTOR in-place.
13657 (handle_special_swappables): Update call accordingly.
13658 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13659 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13660 Build a new vector rather than modifying a CONST_VECTOR in-place.
13661 (handle_special_swappables): Update call accordingly.
13663 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13665 * simplify-rtx.c (simplify_const_binary_operation): Use
13666 CONST_VECTOR_ELT instead of XVECEXP.
13668 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13670 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13671 the selector elements to be different from the data elements
13672 if the selector is a VECTOR_CST.
13673 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13674 ssizetype for the selector.
13676 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13678 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13679 before testing each element individually.
13680 * tree-vect-generic.c (lower_vec_perm): Likewise.
13682 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13684 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13685 * selftest-run-tests.c (selftest::run_tests): Call it.
13686 * vector-builder.h (vector_builder::operator ==): New function.
13687 (vector_builder::operator !=): Likewise.
13688 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13689 (vec_perm_indices::all_from_input_p): New function.
13690 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13691 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13692 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13693 instead of reading the VECTOR_CST directly. Detect whether both
13694 vector inputs are the same before constructing the vec_perm_indices,
13695 and update the number of inputs argument accordingly. Use the
13696 utility functions added above. Only construct sel2 if we need to.
13698 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13700 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13701 the broadcast of the low byte.
13702 (expand_mult_highpart): Use an explicit encoding for the permutes.
13703 * optabs-query.c (can_mult_highpart_p): Likewise.
13704 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13705 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13706 (vectorizable_bswap): Likewise.
13707 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13708 explicit encoding for the power-of-2 permutes.
13709 (vect_permute_store_chain): Likewise.
13710 (vect_grouped_load_supported): Likewise.
13711 (vect_permute_load_chain): Likewise.
13713 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13715 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13716 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13717 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13718 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13719 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13720 (vect_gen_perm_mask_any): Likewise.
13722 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13724 * int-vector-builder.h: New file.
13725 * vec-perm-indices.h: Include int-vector-builder.h.
13726 (vec_perm_indices): Redefine as an int_vector_builder.
13727 (auto_vec_perm_indices): Delete.
13728 (vec_perm_builder): Redefine as a stand-alone class.
13729 (vec_perm_indices::vec_perm_indices): New function.
13730 (vec_perm_indices::clamp): Likewise.
13731 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13732 (vec_perm_indices::new_vector): New function.
13733 (vec_perm_indices::new_expanded_vector): Update for new
13734 vec_perm_indices class.
13735 (vec_perm_indices::rotate_inputs): New function.
13736 (vec_perm_indices::all_in_range_p): Operate directly on the
13737 encoded form, without computing elided elements.
13738 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13739 encoding. Update for new vec_perm_indices class.
13740 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13741 the given vec_perm_builder.
13742 (expand_vec_perm_var): Update vec_perm_builder constructor.
13743 (expand_mult_highpart): Use vec_perm_builder instead of
13744 auto_vec_perm_indices.
13745 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13746 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13747 or double series encoding as appropriate.
13748 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13749 vec_perm_indices instead of auto_vec_perm_indices.
13750 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13751 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13752 (vect_permute_store_chain): Likewise.
13753 (vect_grouped_load_supported): Likewise.
13754 (vect_permute_load_chain): Likewise.
13755 (vect_shift_permute_load_chain): Likewise.
13756 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13757 (vect_transform_slp_perm_load): Likewise.
13758 (vect_schedule_slp_instance): Likewise.
13759 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13760 (vectorizable_mask_load_store): Likewise.
13761 (vectorizable_bswap): Likewise.
13762 (vectorizable_store): Likewise.
13763 (vectorizable_load): Likewise.
13764 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13765 vec_perm_indices instead of auto_vec_perm_indices. Use
13766 tree_to_vec_perm_builder to read the vector from a tree.
13767 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13768 vec_perm_builder instead of a vec_perm_indices.
13769 (have_whole_vector_shift): Use vec_perm_builder and
13770 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13771 truncation to calc_vec_perm_mask_for_shift.
13772 (vect_create_epilog_for_reduction): Likewise.
13773 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13774 from auto_vec_perm_indices to vec_perm_indices.
13775 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13776 instead of changing individual elements.
13777 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13778 the vector in d.perm.
13779 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13780 from auto_vec_perm_indices to vec_perm_indices.
13781 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13782 instead of changing individual elements.
13783 (arm_vectorize_vec_perm_const): Use new_vector to install
13784 the vector in d.perm.
13785 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13786 Update vec_perm_builder constructor.
13787 (rs6000_expand_interleave): Likewise.
13788 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13789 (rs6000_expand_interleave): Likewise.
13791 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13793 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13794 to qimode could truncate the indices.
13795 * optabs.c (expand_vec_perm_var): Likewise.
13797 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13799 * Makefile.in (OBJS): Add vec-perm-indices.o.
13800 * vec-perm-indices.h: New file.
13801 * vec-perm-indices.c: Likewise.
13802 * target.h (vec_perm_indices): Replace with a forward class
13804 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13805 * optabs.h: Include vec-perm-indices.h.
13806 (expand_vec_perm): Delete.
13807 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13808 (expand_vec_perm_const): Declare.
13809 * target.def (vec_perm_const_ok): Replace with...
13810 (vec_perm_const): ...this new hook.
13811 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13812 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13813 * doc/tm.texi: Regenerate.
13814 * optabs.def (vec_perm_const): Delete.
13815 * doc/md.texi (vec_perm_const): Likewise.
13816 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13817 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13818 expand_vec_perm for constant permutation vectors. Assert that
13819 the mode of variable permutation vectors is the integer equivalent
13820 of the mode that is being permuted.
13821 * optabs-query.h (selector_fits_mode_p): Declare.
13822 * optabs-query.c: Include vec-perm-indices.h.
13823 (selector_fits_mode_p): New function.
13824 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13825 is defined, instead of checking whether the vec_perm_const_optab
13826 exists. Use targetm.vectorize.vec_perm_const instead of
13827 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13828 fit in the vector mode before using a variable permute.
13829 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13830 vec_perm_indices instead of an rtx.
13831 (expand_vec_perm): Replace with...
13832 (expand_vec_perm_const): ...this new function. Take the selector
13833 as a vec_perm_indices rather than an rtx. Also take the mode of
13834 the selector. Update call to shift_amt_for_vec_perm_mask.
13835 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13836 Use vec_perm_indices::new_expanded_vector to expand the original
13837 selector into bytes. Check whether the indices fit in the vector
13838 mode before using a variable permute.
13839 (expand_vec_perm_var): Make global.
13840 (expand_mult_highpart): Use expand_vec_perm_const.
13841 * fold-const.c: Includes vec-perm-indices.h.
13842 * tree-ssa-forwprop.c: Likewise.
13843 * tree-vect-data-refs.c: Likewise.
13844 * tree-vect-generic.c: Likewise.
13845 * tree-vect-loop.c: Likewise.
13846 * tree-vect-slp.c: Likewise.
13847 * tree-vect-stmts.c: Likewise.
13848 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13850 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13851 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13852 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13853 (aarch64_vectorize_vec_perm_const): ...this new function.
13854 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13855 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13856 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13857 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13858 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13859 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13860 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13862 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13863 check for NEON modes.
13864 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13865 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13866 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13867 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13869 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13870 the old VEC_PERM_CONST conditions.
13871 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13872 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13873 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13874 (ia64_vectorize_vec_perm_const_ok): Merge into...
13875 (ia64_vectorize_vec_perm_const): ...this new function.
13876 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13877 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13878 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13879 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13880 * config/mips/mips.c (mips_expand_vec_perm_const)
13881 (mips_vectorize_vec_perm_const_ok): Merge into...
13882 (mips_vectorize_vec_perm_const): ...this new function.
13883 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13884 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13885 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13886 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13887 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13888 (rs6000_expand_vec_perm_const): Delete.
13889 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13891 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13892 (altivec_expand_vec_perm_const_le): Take each operand individually.
13893 Operate on constant selectors rather than rtxes.
13894 (altivec_expand_vec_perm_const): Likewise. Update call to
13895 altivec_expand_vec_perm_const_le.
13896 (rs6000_expand_vec_perm_const): Delete.
13897 (rs6000_vectorize_vec_perm_const_ok): Delete.
13898 (rs6000_vectorize_vec_perm_const): New function.
13899 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13900 an element count and rtx array.
13901 (rs6000_expand_extract_even): Update call accordingly.
13902 (rs6000_expand_interleave): Likewise.
13903 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13904 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13905 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13906 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13907 (rs6000_expand_vec_perm_const): Delete.
13908 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13909 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13910 (altivec_expand_vec_perm_const_le): Take each operand individually.
13911 Operate on constant selectors rather than rtxes.
13912 (altivec_expand_vec_perm_const): Likewise. Update call to
13913 altivec_expand_vec_perm_const_le.
13914 (rs6000_expand_vec_perm_const): Delete.
13915 (rs6000_vectorize_vec_perm_const_ok): Delete.
13916 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13917 reference to the SPE evmerge intructions.
13918 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13919 an element count and rtx array.
13920 (rs6000_expand_extract_even): Update call accordingly.
13921 (rs6000_expand_interleave): Likewise.
13922 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13923 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13925 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13927 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13929 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13930 vector mode and that that mode matches the mode of the data
13932 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13933 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13934 directly using expand_vec_perm_1 when forcing selectors into
13936 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13938 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13940 * optabs-query.h (can_vec_perm_p): Delete.
13941 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13942 * optabs-query.c (can_vec_perm_p): Split into...
13943 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13944 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13945 particular selector is valid.
13946 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13947 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13948 (vect_grouped_load_supported): Likewise.
13949 (vect_shift_permute_load_chain): Likewise.
13950 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13951 (vect_transform_slp_perm_load): Likewise.
13952 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13953 (vectorizable_bswap): Likewise.
13954 (vect_gen_perm_mask_checked): Likewise.
13955 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13956 implementations of variable permutation vectors into account
13957 when deciding which selector to use.
13958 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13959 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13960 with a false third argument.
13961 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13962 to test whether the constant selector is valid and can_vec_perm_var_p
13963 to test whether a variable selector is valid.
13965 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13967 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13968 * optabs-query.c (can_vec_perm_p): Likewise.
13969 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13970 instead of vec_perm_indices.
13971 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13972 (vect_gen_perm_mask_checked): Likewise,
13973 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13974 (vect_gen_perm_mask_checked): Likewise,
13976 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13978 * optabs-query.h (qimode_for_vec_perm): Declare.
13979 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13980 (qimode_for_vec_perm): ...this new function.
13981 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13983 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13985 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13986 does not have a conditional at the top.
13988 2018-01-02 Richard Biener <rguenther@suse.de>
13990 * ipa-inline.c (big_speedup_p): Fix expression.
13992 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13995 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13998 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14002 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14003 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14004 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14005 cond_taken_branch_cost 3->4.
14007 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14009 PR tree-optimization/83581
14010 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14011 TODO_cleanup_cfg if any changes have been made.
14013 PR middle-end/83608
14014 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14015 convert_modes if target mode has the right side, but different mode
14018 PR middle-end/83609
14019 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14020 last argument when extracting from CONCAT. If either from_real or
14021 from_imag is NULL, use expansion through memory. If result is not
14022 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14023 the parts directly to inner mode, if even that fails, use expansion
14026 PR middle-end/83623
14027 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14028 check for bswap in mode rather than HImode and use that in expand_unop
14031 Copyright (C) 2018 Free Software Foundation, Inc.
14033 Copying and distribution of this file, with or without modification,
14034 are permitted in any medium without royalty provided the copyright
14035 notice and this notice are preserved.