1 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2 Segher Boessenkool <segher@kernel.crashing.org>
5 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
8 2018-05-17 Richard Biener <rguenther@suse.de>
10 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
11 for pruning loop and prune defs feeding only already visited PHIs.
13 2018-05-17 Richard Biener <rguenther@suse.de>
15 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
17 2018-05-17 Bin Cheng <bin.cheng@arm.com>
18 Richard Biener <rguenther@suse.de>
20 PR tree-optimization/85793
21 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
24 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
26 * internal-fn.h (lookup_internal_fn): Declare
27 * internal-fn.c (lookup_internal_fn): New function.
28 * gimple.c (gimple_build_call_from_tree): Handle calls to
30 * gimple-pretty-print.c (dump_gimple_call): Print "." before
31 internal function names.
32 * tree-pretty-print.c (dump_generic_node): Likewise.
33 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
35 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
37 * gimple-fold.h (gimple_build): Make the function forms take
38 combined_fn rather than built_in_function.
39 (gimple_simplify): Likewise.
40 * gimple-match-head.c (gimple_simplify): Likewise.
41 * gimple-fold.c (gimple_build): Likewise.
42 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
43 rather than gimple_build_call_internal.
44 (get_initial_defs_for_reduction): Likewise.
45 (vect_create_epilog_for_reduction): Likewise.
46 (vectorizable_live_operation): Likewise.
48 2018-05-17 Martin Liska <mliska@suse.cz>
50 * gimple-ssa-sprintf.c (format_directive): Do not use
51 space in between 'G_' and '('.
53 2018-05-17 Jakub Jelinek <jakub@redhat.com>
56 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
57 even if the mask is not all ones.
60 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
62 (ix86_gimple_fold_builtin): Likewise.
65 * config/i386/i386.c: Include tree-vector-builder.h.
66 (ix86_vector_shift_count): New function.
67 (ix86_fold_builtin): Fold shift builtins by scalar count.
68 (ix86_gimple_fold_builtin): Likewise.
70 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
71 _mm512_setzero): New intrinsics.
73 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
74 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
76 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
77 code generation for cases where splatting a value is not useful.
78 * simplify-rtx.c (simplify_ternary_operation): Simplify
79 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
82 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
84 * config.gcc: Support "goldmont-plus".
85 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
87 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
88 PROCESSOR_GOLDMONT_PLUS.
89 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
90 (processor_target_table): Add "goldmont-plus".
91 (PTA_GOLDMONT_PLUS): Define.
92 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
93 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
94 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
95 (fold_builtin_cpu): Add "goldmont-plus".
96 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
97 (ix86_option_override_internal): Add "goldmont-plus".
98 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
99 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
100 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
101 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
103 2018-05-17 Richard Biener <rguenther@suse.de>
105 PR tree-optimization/85757
106 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
107 remove defs that only feed that PHI from further processing.
109 2018-05-16 Jim Wilson <jimw@sifive.com>
111 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
113 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
115 2018-05-16 Mark Wielaard <mark@klomp.org>
117 * dwarf2out.c (count_index_strings): New function.
118 (output_indirect_strings): Call count_index_strings and generate
119 header for dwarf_version >= 5.
121 2018-05-16 Mark Wielaard <mark@klomp.org>
123 * dwarf2out.c (dwarf_FORM): New function.
124 (set_indirect_string): Use dwarf_FORM.
125 (reset_indirect_string): Likewise.
126 (size_of_die): Likewise.
127 (value_format): Likewise.
128 (output_die): Likewise.
129 (add_skeleton_AT_string): Likewise.
130 (output_macinfo_op): Likewise.
131 (index_string): Likewise.
132 (output_index_string_offset): Likewise.
133 (output_index_string): Likewise.
134 (count_index_strings): Likewise.
136 2018-05-16 Carl Love <cel@us.ibm.com>
138 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
139 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
141 2018-05-16 Martin Jambor <mjambor@suse.cz>
143 * ipa-prop.c (ipa_free_all_edge_args): Remove.
144 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
146 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
148 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
149 (fnma<mode>4): Likewise.
150 (fms<mode>4): Likewise.
151 (fnms<mode>4): Likewise.
152 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
153 (aarch64_fnma<mode>4): Likewise.
154 (aarch64_fms<mode>4): Likewise.
155 (aarch64_fnms<mode>4): Likewise.
156 (aarch64_fnmadd<mode>4): Likewise.
158 2018-05-16 Jason Merrill <jason@redhat.com>
160 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
162 2018-05-16 Richard Biener <rguenther@suse.de>
164 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
165 (dump_stmt_cost): Declare.
166 (add_stmt_cost): Dump cost we add.
167 (add_stmt_costs): New function.
168 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
170 (vect_analyze_stmt): Adjust prototype.
171 (vectorizable_condition): Likewise.
172 (vectorizable_live_operation): Likewise.
173 (vectorizable_reduction): Likewise.
174 (vectorizable_induction): Likewise.
175 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
176 cost vector to pass to vectorizable_ and record afterwards.
177 (vect_model_reduction_cost): Take cost vector argument and adjust.
178 (vect_model_induction_cost): Likewise.
179 (vectorizable_reduction): Likewise.
180 (vectorizable_induction): Likewise.
181 (vectorizable_live_operation): Likewise.
182 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
183 SLP_TREE_NUMBER_OF_VEC_STMTS.
184 (vect_analyze_slp_cost_1): Remove.
185 (vect_analyze_slp_cost): Likewise.
186 (vect_slp_analyze_node_operations): Take visited args and
187 a target cost vector. Avoid processing already visited stmt sets.
188 (vect_slp_analyze_operations): Use a local cost vector to gather
189 costs and register those of non-discarded instances.
190 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
191 (vect_schedule_slp_instance): Remove copying of
192 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
194 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
195 adding cost. Record cost entry location.
196 (vect_prologue_cost_for_slp_op): Function to compute cost of
197 a constant or invariant generated for SLP vect in the prologue,
198 split out from vect_analyze_slp_cost_1.
199 (vect_model_simple_cost): Make static. Adjust for SLP costing.
200 (vect_model_promotion_demotion_cost): Likewise.
201 (vect_model_store_cost): Likewise, make static.
202 (vect_model_load_cost): Likewise.
203 (vectorizable_bswap): Add cost vector arg and adjust.
204 (vectorizable_call): Likewise.
205 (vectorizable_simd_clone_call): Likewise.
206 (vectorizable_conversion): Likewise.
207 (vectorizable_assignment): Likewise.
208 (vectorizable_shift): Likewise.
209 (vectorizable_operation): Likewise.
210 (vectorizable_store): Likewise.
211 (vectorizable_load): Likewise.
212 (vectorizable_condition): Likewise.
213 (vectorizable_comparison): Likewise.
214 (can_vectorize_live_stmts): Likewise.
215 (vect_analyze_stmt): Likewise.
216 (vect_transform_stmt): Adjust calls to vectorizable_*.
217 * tree-vectorizer.c: Include gimple-pretty-print.h.
218 (dump_stmt_cost): New function.
220 2018-05-16 Richard Biener <rguenther@suse.de>
222 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
223 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
224 * tree-ssa-dse.c: Include tree-ssa-loop.h.
225 (check_name): New callback.
226 (dse_classify_store): Track cycles via a visited bitmap of PHI
227 defs and simplify handling of in-loop and across loop dead stores
228 and properly fail for loop-variant refs. Handle byte-tracking with
229 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
232 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
234 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
235 (vect_get_mask_type_for_stmt): Likewise.
236 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
238 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
239 to determine the statement's vector type and the vector type that
240 should be used for calculating nunits. Deal with cases in which
241 the type has to be deferred.
242 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
243 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
244 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
245 (vect_determine_vf_for_stmt): New functions, split out from...
246 (vect_determine_vectorization_factor): ...here.
247 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
248 (vect_get_mask_type_for_stmt): New functions, split out from
249 vect_determine_vectorization_factor.
251 2018-05-16 Richard Biener <rguenther@suse.de>
253 * tree-cfg.c (verify_gimple_assign_ternary): Properly
254 verify the [VEC_]COND_EXPR embedded comparison.
256 2018-05-15 Martin Sebor <msebor@redhat.com>
258 PR tree-optimization/85753
259 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
260 RECORD_TYPE in addition to ARRAY_TYPE.
262 2018-05-15 Martin Sebor <msebor@redhat.com>
265 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
267 2018-05-15 Richard Biener <rguenther@suse.de>
269 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
270 add by_clobber_p one. Change algorithm to collect all defs
271 representing uses we need to walk and try reducing them to
272 a single one before failing.
273 (dse_dom_walker::dse_optimize_stmt): Adjust.
275 2018-05-13 Mark Wielaard <mark@klomp.org>
277 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
278 (size_of_loc_descr): Likewise.
279 (output_loc_operands): Likewise.
280 (output_loc_operands_raw): Likewise.
281 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
282 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
283 (hash_loc_operands): Likewise.
284 (compare_loc_operands): Likewise.
286 2018-05-14 Mark Wielaard <mark@klomp.org>
288 * dwarf2out.c (count_index_addrs): New function.
289 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
291 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
293 PR tree-optimization/83648
294 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
295 return value as malloc candidate.
297 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
300 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
301 as true in call to suggest_attribute.
303 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
305 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
308 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
310 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
312 (aarch64_crypto_aesd_fused): Likewise.
314 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
316 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
317 (movsi_aarch64): Likewise.
318 (load_pairsi): Likewise.
319 (load_pairdi): Likewise.
320 (store_pairsi): Likewise.
321 (store_pairdi): Likewise.
322 (load_pairsf): Likewise.
323 (load_pairdf): Likewise.
324 (store_pairsf): Likewise.
325 (store_pairdf): Likewise.
326 (zero_extend): Likewise.
327 (trunc): Swap alternatives.
328 (fcvt_target): Add '?' to prefer w over r.
330 2018-05-14 Jakub Jelinek <jakub@redhat.com>
333 * config/i386/i386.md: Disallow non-commutative arithmetics in
334 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
335 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
336 in the peephole2 before it.
338 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
340 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
341 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
342 (ix86_handle_option): Handle -mcldemote.
343 * config.gcc: New header.
344 * config/i386/cldemoteintrin.h: New file.
345 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
346 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
348 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
349 OPTION_MASK_ISA_CLDEMOTE.
350 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
351 (ix86_valid_target_attribute_inner_p): Ditto.
352 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
353 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
354 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
355 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
356 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
358 * config/i386/i386.opt: Add -mcldemote.
359 * config/i386/x86intrin.h: New header.
360 * doc/invoke.texi: Add -mcldemote.
362 2018-05-14 Richard Biener <rguenther@suse.de>
364 * doc/match-and-simplify.texi: Adjust :s documentation.
366 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
368 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
369 intended memcpy size.
370 (REORDER_45): Likewise.
372 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
375 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
376 * vec.c (qsort_chk): Use gcc_qsort.
377 * Makefile.in (OBJS-libcommon): Add sort.o.
378 (build/sort.o): New target. Use it...
379 (BUILD_RTL): ... here, and...
380 (build/gencfn-macros): ... here, and...
381 (build/genmatch): ... here.
383 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
384 Chung-Ju Wu <jasonwucj@gmail.com>
386 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
387 * config/nds32/nds32-graywolf.md: New file.
388 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
389 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
391 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
392 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
393 * config/nds32/nds32.md (pipeline_model): Add graywolf.
394 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
395 * config/nds32/pipelines.md: Include n15 settings.
397 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
398 Chung-Ju Wu <jasonwucj@gmail.com>
400 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
401 * config/nds32/nds32-n13.md: New file.
402 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
403 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
405 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
406 * config/nds32/nds32.md (pipeline_model): Add n13.
407 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
408 * config/nds32/pipelines.md: Include n13 settings.
410 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
411 Chung-Ju Wu <jasonwucj@gmail.com>
413 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
414 * config/nds32/nds32-n10.md: New file.
415 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
416 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
418 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
419 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
420 * config/nds32/nds32.md (pipeline_model): Add n10.
421 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
422 * config/nds32/pipelines.md: Include n10 settings.
424 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
425 Kito Cheng <kito.cheng@gmail.com>
426 Chung-Ju Wu <jasonwucj@gmail.com>
428 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
429 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
430 Add enum values for DSP extension instructions.
431 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
433 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
434 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
436 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
437 * config/nds32/nds32-dspext.md: New file for DSP implementation.
438 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
439 * config/nds32/nds32-intrinsic.md: Likewise.
440 * config/nds32/nds32_intrinsic.h: Likewise.
441 * config/nds32/nds32-md-auxiliary.c: Likewise.
442 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
443 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
444 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
445 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
446 * config/nds32/nds32-protos.h: New declarations for DSP extension.
447 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
448 TYPE_DMAC in switch statement.
449 * config/nds32/nds32.c: New checking and implementation for DSP
450 extension instructions.
451 * config/nds32/nds32.h: Likewise.
452 * config/nds32/nds32.md: Likewise.
453 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
454 * config/nds32/predicates.md: Implement new predicates for DSP
457 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
459 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
460 Reformat alternatives and attributes so it is easier to identify
461 which constraints/attributes go with which instruction.
462 (mov<mode>_hardfloat32, FMOVE64): Likewise.
463 (mov<mode>_softfloat32, FMOVE64): Likewise.
464 (mov<mode>_hardfloat64, FMOVE64): Likewise.
465 (mov<mode>_softfloat64, FMOVE64): Likewise.
467 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
469 * doc/extend.texi (PowerPC Built-in Functions): Rename this
471 (Basic PowerPC Built-in Functions): The new name of the
472 subsection previously known as "PowerPC Built-in Functions".
473 (Basic PowerPC Built-in Functions Available on all Configurations):
475 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
476 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
477 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
478 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
480 2018-05-11 Martin Jambor <mjambor@suse.cz>
483 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
486 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
489 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
491 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
493 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
494 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
495 (ix86_handle_option): Handle -mwaitpkg.
496 * config.gcc: New header.
497 * config/i386/cpuid.h (bit_WAITPKG): New bit.
498 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
499 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
501 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
502 OPTION_MASK_ISA_WAITPKG.
503 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
504 (ix86_option_override_internal): Add PTA_WAITPKG.
505 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
506 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
508 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
509 __builtin_ia32_umwait and __builtin_ia32_tpause.
510 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
511 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
512 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
513 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
514 UNSPECV_TPAUSE): New.
515 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
516 * config/i386/i386.opt: Add -mwaitpkg.
517 * config/i386/waitpkgintrin.h: New file.
518 * config/i386/x86intrin.h: New header.
519 * doc/invoke.texi: Add -mwaitpkg.
521 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
524 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
526 (cortex-m0): Use armv6s-m isa.
527 (cortex-m0plus): Likewise.
528 (cortex-m1): Likewise.
529 (cortex-m0.small-multiply): Likewise.
530 (cortex-m0plus.small-multiply): Likewise.
531 (cortex-m1.small-multiply): Likewise.
533 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
534 Jakub Jelinek <jakub@redhat.com>
536 PR tree-optimization/85692
537 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
538 source permute as well.
540 2018-05-11 Martin Liska <mliska@suse.cz>
543 * doc/extend.texi: Document LLVM style format for no_sanitize
546 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
548 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
549 mode_supports_vsx_dform_quad to mode_supports_dq_form.
550 (mode_supports_vsx_dform_quad): Likewise.
551 (mode_supports_vmx_dform): Move these functions to be next to the
552 other mode_supports functions.
553 (mode_supports_dq_form): Likewise.
554 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
555 mode_supports_dq_form.
556 (reg_offset_addressing_ok_p): Likewise.
557 (offsettable_ok_by_alignment): Likewise.
558 (rs6000_legitimate_offset_address_p): Likewise.
559 (legitimate_lo_sum_address_p): Likewise.
560 (rs6000_legitimize_address): Likewise.
561 (rs6000_legitimize_reload_address): Likewise.
562 (rs6000_secondary_reload_inner): Likewise.
563 (rs6000_preferred_reload_class): Likewise.
564 (rs6000_output_move_128bit): Likewise.
566 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
568 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
569 Generate SImode target register for null target.
570 <case IX86_BUILTIN_XGETBV>: Ditto.
571 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
572 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
574 2018-05-10 Carl Love <cel@us.ibm.com>
576 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
577 dcbtt and dcbtstt if operands[2] is 0.
579 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
582 * config/i386/sse.md (usadv64qi): New expander.
584 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
586 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
587 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
588 -maltivec=be support.
589 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
590 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
591 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
592 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
593 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
594 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
595 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
596 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
597 altivec_vsumsws): Adjust.
598 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
599 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
600 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
602 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
603 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
604 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
605 (altivec_lve<VI_char>x): Delete expand.
606 (*altivec_lve<VI_char>x_internal): Rename to...
607 (altivec_lve<VI_char>x): ... this.
608 (altivec_lvxl_<mode>): Delete expand.
609 (*altivec_lvxl_<mode>_internal): Rename to ...
610 (altivec_lvxl_<mode>): ... this.
611 (altivec_stvxl_<mode>): Delete expand.
612 (*altivec_stvxl_<mode>_internal): Rename to ...
613 (altivec_stvxl_<mode>): ... this.
614 (altivec_stve<VI_char>x): Delete expand.
615 (*altivec_stve<VI_char>x_internal): Rename to ...
616 (altivec_stve<VI_char>x): ... this.
617 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
618 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
619 reduc_plus_scal_<mode>): Adjust.
620 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
622 (rs6000_cpu_cpp_builtins): Adjust.
623 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
624 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
625 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
626 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
627 -maltivec=be support.
628 (rs6000_split_vec_extract_var): Adjust.
629 (rs6000_split_v4si_init): Adjust.
630 (swap_selector_for_mode): Delete.
631 (altivec_expand_lvx_be, altivec_expand_stvx_be,
632 altivec_expand_stvex_be): Delete.
633 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
634 -maltivec=be support.
635 (rs6000_gimple_fold_builtin): Ditto.
636 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
638 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
639 (TARGET_DIRECT_MOVE_64BIT): Adjust.
640 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
641 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
642 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
643 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
644 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
645 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
646 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
647 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
648 anonymous split): Adjust.
649 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
650 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
652 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
654 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
655 when --with-gxx-include-dir is also specified.
656 * configure: Regenerate.
658 2018-05-09 Jim Wilson <jimw@sifive.com>
661 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
662 * config/riscv/t-withmultilib: New.
663 * config/riscv/withmultilib.h: New.
664 * doc/install.texi: Document RISC-V --with-multilib-list support.
666 2018-05-09 Richard Biener <rguenther@suse.de>
668 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
670 (vect_bb_vectorization_profitable_p): Adjust. Compute
671 actual scalar cost using the cost vector and the add_stmt_cost
674 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
676 PR rtl-optimization/85645
677 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
678 in the REG_CFA_REGISTER note for LR, don't leave it empty.
680 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
682 PR rtl-optimization/85645
683 * shrink-wrap.c (spread_components): Return a boolean saying if
684 anything was changed.
685 (try_shrink_wrapping_separate): Iterate spread_components until
686 nothing changes anymore.
688 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
690 PR rtl-optimization/85645
691 * regrename.c (build_def_use): Also kill the chains that include the
692 destination of a REG_CFA_REGISTER note.
694 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
696 PR rtl-optimization/85645
697 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
698 insn that has a REG_CFA_REGISTER note.
700 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
702 * cfgexpand.c (expand_clobber): New function.
703 (expand_gimple_stmt_1): Use it.
704 * tree-vect-stmts.c (vect_clobber_variable): New function,
706 (vectorizable_simd_clone_call): ...here.
707 (vectorizable_store): Emit a clobber either side of an
708 IFN_STORE_LANES sequence.
709 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
711 2018-05-09 Tom de Vries <tom@codesourcery.com>
714 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
715 (define_insn "trap_if_false"): Add exit after trap.
717 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
719 PR rtl-optimization/85638
720 * bb-reorder.c: Include common/common-target.h.
721 (create_forwarder_block): New function extracted from...
722 (fix_up_crossing_landing_pad): ...here. Rename into...
723 (dw2_fix_up_crossing_landing_pad): ...this.
724 (sjlj_fix_up_crossing_landing_pad): New function.
725 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
726 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
727 from both partitions and exit the loop after one iteration.
729 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
732 * doc/extend.texi (PowerPC Built-in Functions): Rename this
734 (Basic PowerPC Built-in Functions): The new name of the
735 subsection previously known as "PowerPC Built-in Functions".
736 (Basic PowerPC Built-in Functions Available on all Configurations):
738 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
740 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
742 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
744 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
747 2018-05-08 Jim Wilson <jimw@sifive.com>
749 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
750 (LD_EMUL_SUFFIX): New.
753 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
755 * doc/extend.texi (PowerPC Built-in Functions): Rename this
757 (Basic PowerPC Built-in Functions): The new name of the
758 subsection previously known as "PowerPC Built-in Functions".
759 (Basic PowerPC Built-in Functions Available on all Configurations):
761 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
763 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
765 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
767 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
770 2018-05-08 Jakub Jelinek <jakub@redhat.com>
773 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
774 after cmpelim optimization.
776 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
778 * config.gcc: Support "goldmont".
779 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
780 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
782 * config/i386/i386.c (m_GOLDMONT): Define.
783 (processor_target_table): Add "goldmont".
784 (PTA_GOLDMONT): Define.
785 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
786 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
787 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
788 (fold_builtin_cpu): Add "goldmont".
789 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
790 (ix86_option_override_internal): Add "goldmont".
791 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
792 (processor_type): Add PROCESSOR_GOLDMONT.
793 * config/i386/i386.md: Add CPU "glm".
794 * config/i386/glm.md: New file.
795 * config/i386/x86-tune.def: Add m_GOLDMONT.
796 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
798 2018-05-08 Jakub Jelinek <jakub@redhat.com>
801 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
803 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
804 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
805 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
808 * config/i386/i386.c (ix86_fold_builtin): Handle
809 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
812 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
813 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
815 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
818 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
819 (check_arch): Likewise.
820 (check_fpu): Return the result rather than printing it.
821 (end arch): Fix operator precedence.
823 (END): Print the result from check_fpu.
825 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
826 Alan Hayward <alan.hayward@arm.com>
827 David Sherwood <david.sherwood@arm.com>
829 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
830 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
831 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
832 (*fcmuo<mode>_and): New patterns.
834 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
836 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
837 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
838 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
839 (cmp_op, sve_imm_con): New code attributes.
840 (SVE_COND_INT_CMP, imm_con): Delete.
841 (cmp_op): Remove above unspecs from int attribute.
842 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
844 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
845 comparison-specific unspecs.
846 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
847 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
848 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
849 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
850 (*vec_fcm<cmp_op><mode>): Rename to...
851 (*fcm<cmp_op><mode>): ...this and adjust likewise.
852 (*vec_fcmuo<mode>): Rename to...
853 (*fcmuo<mode>): ...this and adjust likewise.
854 (*pred_fcm<cmp_op><mode>): New pattern.
855 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
856 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
858 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
860 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
861 (aarch64_emit_sve_predicated_cond): New function.
862 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
863 (aarch64_emit_unspec_cond_or): Replace with...
864 (aarch64_emit_sve_or_conds): ...this new function. Use
865 aarch64_emit_sve_ptrue_op for the individual comparisons and
866 aarch64_emit_binop to OR them together.
867 (aarch64_emit_inverted_unspec_cond): Replace with...
868 (aarch64_emit_sve_inverted_cond): ...this new function. Use
869 aarch64_emit_sve_ptrue_op for the comparison and
870 aarch64_emit_unop to invert the result.
871 (aarch64_expand_sve_vec_cmp_float): Update after the above
872 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
874 2018-05-07 Nathan Sidwell <nathan@acm.org>
876 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
877 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
878 (Backwards Compatibility): Likewise.
880 2018-05-07 Luis Machado <luis.machado@linaro.org>
884 2018-05-07 Luis Machado <luis.machado@linaro.org>
886 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
887 <prefetch_dynamic_strides>: New const bool field.
888 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
889 prefetch_dynamic_strides.
890 (exynosm1_prefetch_tune): Likewise.
891 (thunderxt88_prefetch_tune): Likewise.
892 (thunderx_prefetch_tune): Likewise.
893 (thunderx2t99_prefetch_tune): Likewise.
894 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
895 (aarch64_override_options_internal): Update to set
896 PARAM_PREFETCH_DYNAMIC_STRIDES.
897 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
898 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
899 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
900 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
901 prefetch-dynamic-strides setting.
903 2018-05-07 Luis Machado <luis.machado@linaro.org>
905 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
906 <minimum_stride>: New const int field.
907 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
908 minimum_stride field.
909 (exynosm1_prefetch_tune): Likewise.
910 (thunderxt88_prefetch_tune): Likewise.
911 (thunderx_prefetch_tune): Likewise.
912 (thunderx2t99_prefetch_tune): Likewise.
913 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
914 (aarch64_override_options_internal): Update to set
915 PARAM_PREFETCH_MINIMUM_STRIDE.
916 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
917 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
918 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
919 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
920 stride is constant and is below the minimum stride threshold.
922 2018-05-07 Luis Machado <luis.machado@linaro.org>
924 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
927 2018-05-07 Luis Machado <luis.machado@linaro.org>
929 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
930 <prefetch_dynamic_strides>: New const bool field.
931 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
932 prefetch_dynamic_strides.
933 (exynosm1_prefetch_tune): Likewise.
934 (thunderxt88_prefetch_tune): Likewise.
935 (thunderx_prefetch_tune): Likewise.
936 (thunderx2t99_prefetch_tune): Likewise.
937 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
938 (aarch64_override_options_internal): Update to set
939 PARAM_PREFETCH_DYNAMIC_STRIDES.
940 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
941 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
942 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
943 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
944 prefetch-dynamic-strides setting.
946 2018-05-07 Luis Machado <luis.machado@linaro.org>
948 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
949 <minimum_stride>: New const int field.
950 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
951 minimum_stride field.
952 (exynosm1_prefetch_tune): Likewise.
953 (thunderxt88_prefetch_tune): Likewise.
954 (thunderx_prefetch_tune): Likewise.
955 (thunderx2t99_prefetch_tune): Likewise.
956 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
957 (aarch64_override_options_internal): Update to set
958 PARAM_PREFETCH_MINIMUM_STRIDE.
959 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
960 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
961 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
962 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
963 stride is constant and is below the minimum stride threshold.
965 2018-05-06 Jakub Jelinek <jakub@redhat.com>
968 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
969 the type is addressable. Don't force op into register if it has
972 2018-05-05 Roland McGrath <mcgrathr@google.com>
975 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
976 any section for which we don't know a specific type it should have,
977 regardless of name. Previously this was done only for the exact
978 names ".init_array", ".fini_array", and ".preinit_array".
979 (default_elf_asm_named_section): Add comment about
980 relationship with default_section_type_flags and SECTION_NOTYPE.
981 (get_section): Don't consider it a type conflict if one side has
982 SECTION_NOTYPE and the other doesn't, as long as neither has the
983 SECTION_BSS et al used in the default_section_type_flags logic.
985 2018-05-05 Tom de Vries <tom@codesourcery.com>
988 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
989 (workaround_barsyncs): New function.
990 (nvptx_reorg): Use workaround_barsyncs.
991 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
992 (define_expand "nvptx_membar_cta"): New define_expand.
993 (define_insn "*nvptx_membar_cta"): New insn.
995 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
997 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
998 To improve optimization opportunities.
999 * builtin-types.def: The new needed builtin types for the above.
1001 2018-05-04 Richard Biener <rguenther@suse.de>
1003 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1004 * gimple-ssa-store-merging.c
1005 (imm_store_chain_info::output_merged_store): Remove redundant create,
1006 release split_store vector contents on failure.
1007 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1008 scalar stmt vector on cache hit.
1010 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1012 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1014 * config.gcc (powerpc-xilinx-eabi*): Remove.
1015 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1017 (fusion_addis_mem_combo_load): Ditto.
1018 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1020 (rs6000_cpu_cpp_builtins): Ditto.
1021 * config/rs6000/rs6000-linux.c
1022 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1023 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1024 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1026 (rs6000_setup_reg_addr_masks): Ditto.
1027 (rs6000_init_hard_regno_mode_ok): Ditto.
1028 (rs6000_option_override_internal): Ditto.
1029 (legitimate_lo_sum_address_p): Ditto.
1030 (rs6000_legitimize_address): Ditto.
1031 (rs6000_legitimize_reload_address): Ditto.
1032 (rs6000_legitimate_address_p): Ditto.
1033 (abi_v4_pass_in_fpr): Ditto.
1034 (setup_incoming_varargs): Ditto.
1035 (rs6000_gimplify_va_arg): Ditto.
1036 (rs6000_split_multireg_move): Ditto.
1037 (rs6000_savres_strategy): Ditto.
1038 (rs6000_emit_prologue_components): Ditto.
1039 (rs6000_emit_epilogue_components): Ditto.
1040 (rs6000_emit_prologue): Ditto.
1041 (rs6000_emit_epilogue): Ditto.
1042 (rs6000_elf_file_end): Ditto.
1043 (rs6000_function_value): Ditto.
1044 (rs6000_libcall_value): Ditto.
1045 * config/rs6000/rs6000.h: Ditto.
1046 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1047 (TARGET_MINMAX): ... this. New.
1048 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1049 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1050 (*movsi_internal1_single): Delete.
1051 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1052 mfpu=, mxilinx-fpu): Delete.
1053 * config/rs6000/singlefp.h: Delete.
1054 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1055 * config/rs6000/t-rs6000: Ditto.
1056 * config/rs6000/t-xilinx: Delete.
1057 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1058 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1059 (VStype_simple): Delete.
1060 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1061 * config/rs6000/xfpu.h: Delete.
1062 * config/rs6000/xfpu.md: Delete.
1063 * config/rs6000/xilinx.h: Delete.
1064 * config/rs6000/xilinx.opt: Delete.
1065 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1066 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1068 2018-05-04 Tom de Vries <tom@codesourcery.com>
1071 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1074 2018-05-04 Richard Biener <rguenther@suse.de>
1077 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1078 (expand_complex_div_wide): Likewise.
1079 (expand_complex_operations_1): Likewise.
1080 (expand_complex_libcall): Preserve EH info of the original stmt.
1081 (tree_lower_complex): Handle removed blocks.
1082 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1083 on complex multiplication and division libcall builtins.
1085 2018-05-04 Richard Biener <rguenther@suse.de>
1088 * fold-const.c (negate_expr_p): Restrict negation of operand
1089 zero of a division to when we know that can happen without
1091 (fold_negate_expr_1): Likewise.
1093 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1096 * real.h (real_nextafter): Declare.
1097 * real.c (real_nextafter): New function.
1098 * fold-const-call.c (fold_const_nextafter): New function.
1099 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1100 CASE_CFN_NEXTTOWARD.
1101 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1102 even when arg1_mode is different from arg0_mode.
1104 2018-05-03 Nathan Sidwell <nathan@acm.org>
1106 * doc/extend.texi (Deprecated Features): Remove
1108 (Backwards Compatibility): Likewise.
1109 * doc/invoke.texi (C++ Language Options): Likewise.
1110 (C++ Dialect Options): Likewise.
1112 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1115 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1116 _mm512_mask_mullox_epi64): New intrinsics.
1118 2018-05-03 Tom de Vries <tom@codesourcery.com>
1121 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1122 dump files): Add offload-tree.
1124 2018-05-03 Richard Biener <rguenther@suse.de>
1126 PR tree-optimization/85615
1127 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1128 to loops not nested in BBs loop father to avoid creating multi-entry
1131 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1133 PR tree-optimization/70291
1134 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1135 arguments. Change return type to tree. Emit libcall as a new
1136 statement rather than replacing existing one when inplace_p is true.
1137 (expand_complex_multiplication_components): New function.
1138 (expand_complex_multiplication): Expand floating-point complex
1139 multiplication using the above.
1140 (expand_complex_division): Rename inner_type parameter to type.
1141 Update expand_complex_libcall call-site.
1142 (expand_complex_operations_1): Update expand_complex_multiplication
1143 and expand_complex_division call-sites.
1145 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1148 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1149 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1150 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1151 the highest significant bit of the shift count mask is clear. In
1152 check whether and[sq]i3 is needed verify that all significant bits
1153 of the shift count other than the highest are set.
1155 2018-05-02 Tom de Vries <tom@codesourcery.com>
1158 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1159 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1160 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1161 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1162 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1163 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1164 __builtin_goacc_parlevel_size.
1166 2018-05-02 Richard Biener <rguenther@suse.de>
1168 PR tree-optimization/85597
1169 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1170 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1173 2018-05-02 Tom de Vries <tom@codesourcery.com>
1176 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1177 dump files): Add ltrans-tree.
1179 2018-05-02 Tom de Vries <tom@codesourcery.com>
1182 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1183 dump files): Add wpa-ipa.
1185 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1187 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1188 powerpc*-*-linux*paired* target.
1189 * config/rs6000/750cl.h: Delete.
1190 * config/rs6000/paired.h: Delete.
1191 * config/rs6000/paired.md: Delete.
1192 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1194 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1195 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1196 comment. Remove paired float support.
1197 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1198 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1200 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1201 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1203 * config/rs6000/rs6000.c: Remove paired float support.
1204 (paired_expand_vector_init, paired_expand_vector_move,
1205 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1206 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1207 paired_expand_builtin, paired_expand_predicate_builtin,
1208 paired_init_builtins): Delete.
1209 * config/rs6000/rs6000.h: Remove paired float support.
1210 * config/rs6000/rs6000.md: Remove paired float support.
1211 (move_from_CR_ov_bit): Delete.
1212 * config/rs6000/rs6000.opt (mpaired): Delete.
1213 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1214 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1216 2018-05-02 Richard Biener <rguenther@suse.de>
1219 * gimplify.c (gimplify_save_expr): When in SSA form allow
1220 SAVE_EXPRs to compute to SSA vars.
1222 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1225 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1226 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1227 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1228 clobber operands[2], instead use a new pseudo. Formatting fixes.
1230 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1232 PR tree-optimization/85586
1233 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1234 exit early for statements in the same group if the accesses are
1237 2018-05-02 Tom de Vries <tom@codesourcery.com>
1240 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1243 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1245 PR tree-optimization/85143
1246 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1248 2018-05-01 Tom de Vries <tom@codesourcery.com>
1251 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1252 not found" error message.
1254 2018-05-01 Tom de Vries <tom@codesourcery.com>
1257 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1258 * vec.c (test_ordered_remove_if): New function.
1259 (vec_c_tests): Call test_ordered_remove_if.
1260 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1261 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1262 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1263 VEC_ORDERED_REMOVE_IF.
1265 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1267 PR tree-optimization/82665
1268 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1269 pointer subtraction where arguments come from a memchr call.
1271 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1273 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1274 --push-state --as-needed and --pop-state instead of --as-needed and
1275 --no-as-needed if ld supports it.
1276 * configure: Regenerated.
1279 * doc/install.texi2html: Replace _002d with - and _002a with * in
1280 generated html files using sed.
1282 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1285 * gcc-rich-location.c (blank_line_before_p): New function.
1286 (use_new_line): New function.
1287 (gcc_rich_location::add_fixit_insert_formatted): New function.
1288 * gcc-rich-location.h
1289 (gcc_rich_location::add_fixit_insert_formatted): New function.
1291 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1293 * selftest.c (assert_streq): Rename "expected" and "actual" to
1294 "val1" and "val2". Extend NULL-handling to cover both inputs
1295 symmetrically, while still requiring both to be non-NULL for a pass.
1296 * selftest.h (assert_streq): Rename "expected" and "actual" to
1298 (ASSERT_EQ): Likewise.
1299 (ASSERT_EQ_AT): Likewise.
1300 (ASSERT_KNOWN_EQ): Likewise.
1301 (ASSERT_KNOWN_EQ_AT): Likewise.
1302 (ASSERT_NE): Likewise.
1303 (ASSERT_MAYBE_NE): Likewise.
1304 (ASSERT_MAYBE_NE_AT): Likewise.
1305 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1306 the assertion to pass.
1307 (ASSERT_STREQ_AT): Likewise.
1309 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1311 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1312 interaction with -pie.
1314 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1316 * selftest.h: Fix alphabetization of per-source-file selftest
1319 2018-04-30 Jason Merrill <jason@redhat.com>
1321 PR c++/61982 - dead stores to destroyed objects.
1322 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1325 2018-04-30 Jason Merrill <jason@redhat.com>
1327 * tree.c (build_clobber): New.
1328 * tree.h: Declare it.
1329 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1331 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1333 * diagnostic-show-locus.c (layout::layout): Update for
1334 location_get_source_line returning a char_span.
1335 (struct char_span): Move to input.h.
1336 (struct correction): Update for fields in char_span becoming
1338 (struct source_line): Update for location_get_source_line
1339 returning a char_span.
1340 (layout::print_line): Likewise.
1341 * edit-context.c (edited_file::print_content): Likewise.
1342 (edited_file::print_diff_hunk): Likewise.
1343 (edited_file::print_run_of_changed_lines): Likewise.
1344 (edited_file::get_num_lines): Likewise.
1345 (edited_line::edited_line): Likewise.
1346 * final.c (asm_show_source): Likewise.
1347 * input.c (location_get_source_line): Convert return type
1348 from const char * to char_span, losing the final "line_len"
1350 (dump_location_info): Update for the above.
1351 (get_substring_ranges_for_loc): Likewise. Use a char_span
1352 when handling the literal within the line.
1353 (test_reading_source_line): Update for location_get_source_line
1354 returning a char_span.
1355 * input.h (class char_span): Move here from
1356 diagnostic-show-locus.c, converting from a struct to a class.
1357 Make data members private.
1358 (char_span::operator bool): New.
1359 (char_span::length): New.
1360 (char_span::get_buffer): New.
1361 (char_span::operator[]): New.
1362 (char_span::subspan): Make const.
1363 (char_span::xstrdup): New.
1364 (location_get_source_line): Convert return type from const char *
1365 to char_span, losing the final "line_size" param.
1367 2018-04-30 Jan Hubicka <jh@suse.cz>
1369 * lto-wrapper.c (ltrans_priorities): New static var.
1370 (cmp_priority): New.
1371 (run_gcc): Read priorities and if doing parallel build order
1372 the Makefile by them.
1374 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1376 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1378 2018-04-30 Richard Biener <rguenther@suse.de>
1380 * tree-cfg.c (verify_address): Remove base argument, add
1381 flag whether to check TREE_ADDRESSABLE and do that.
1382 (verify_expr): Remove.
1383 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1384 (verify_gimple_assign_single): Likewise.
1385 (verify_gimple_switch): Likewise.
1386 (verify_expr_location_1): Dereference tp once. Add (disabled)
1387 piece from verify_expr.
1388 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1390 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1392 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1394 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1396 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1397 (small_data_pattern): Likewise.
1398 (arc_rewrite_small_data): Likewise.
1399 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1400 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1401 (get_symbol_alignment): New function.
1402 (legitimate_small_data_address_p): Likewise.
1403 (legitimate_scaled_address): Update, call
1404 legitimate_small_data_address_p.
1405 (output_sdata): New static variable.
1406 (arc_print_operand): Update how we handle small data operands.
1407 (arc_print_operand_address): Likewise.
1408 (arc_legitimate_address_p): Update, use
1409 legitimate_small_data_address_p.
1410 (arc_rewrite_small_data_p): Remove.
1411 (arc_rewrite_small_data_1): Likewise.
1412 (arc_rewrite_small_data): Likewise.
1413 (small_data_pattern): Likewise.
1414 (compact_sda_memory_operand): Update to use
1415 legitimate_small_data_address_p and get_symbol_alignment.
1416 (prepare_move_operands): Don't rewite sdata pattern.
1417 (prepare_extend_operands): Remove.
1418 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1420 (zero_extendqisi2): Likewise.
1421 (zero_extendhisi2): Likewise.
1422 (extendqihi2): Likewise.
1423 (extendqisi2): Likewise.
1424 (extendhisi2): Likewise.
1428 * config/arc/constraints.md (Usd): Change it to memory constraint.
1430 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1432 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1433 as source of std instructions.
1434 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1435 allow 6-bit constants as source for store instructions.
1436 (movdi_insn): Update instruction pattern to allow 6-bit constants
1437 as source for store instructions.
1439 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1441 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1443 2018-04-30 Nathan Sidwell <nathan@acm.org>
1444 Sandra Loosemore <sandra@codesourcery.com>
1446 * dumpfile.c (dump_open): Allow '-' for stdout.
1447 * doc/invoke.texi (Developer Options): Document dump filename
1448 determination early. Document stdin/stdout selection.
1450 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1452 Microblaze Target: PIC data text relative
1454 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1455 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1457 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1458 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1459 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1460 New addressing mode for data-text relative position indepenedent code.
1461 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1462 'ADDRESS_SYMBOLIC_TXT_REL'.
1463 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1464 (microblaze_legitimate_pic_operand): Exclude function calls from
1465 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1466 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1468 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1469 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1470 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1471 for 'address + offset'.
1472 (microblaze_expand_prologue): Add new function prologue call for
1474 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1475 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1476 table in case of TARGET_PIC_DATA_TEXT_REL.
1477 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1478 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1479 Add new macros 'UNSPEC_TEXT',
1480 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1481 + exclude function calls from 'UNSPEC_PLT' in case of data text
1483 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1484 new target hook for generating address diff vector tables in case of
1486 * doc/tm.texi : Regenerate.
1487 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1488 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1489 of addr diff vector generation.
1490 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1491 target hook definition.
1492 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1493 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1494 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1495 option for data text relative.
1497 2018-04-30 Richard Biener <rguenther@suse.de>
1499 * tree-chrec.h (evolution_function_is_constant_p): Remove
1501 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1503 2018-04-30 Richard Biener <rguenther@suse.de>
1506 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1508 2018-04-30 Richard Biener <rguenther@suse.de>
1510 PR tree-optimization/28364
1511 PR tree-optimization/85275
1512 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1513 copying first exit test.
1515 2018-04-28 Mark Wielaard <mark@klomp.org>
1517 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1519 (dwarf_AT): Handle DW_AT_addr_base.
1520 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1522 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1525 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1526 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1527 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1528 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1530 2018-04-28 Richard Biener <rguenther@suse.de>
1532 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1533 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1534 to reflect use. Only add interesting stmts.
1536 2018-04-27 Martin Jambor <mjambor@suse.cz>
1539 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1540 the jump function allows for passing through aggregate values.
1542 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1544 * input.h (in_system_header_at): Convert from macro to inline
1546 (from_macro_expansion_at): Likewise.
1547 (from_macro_definition_at): Likewise.
1549 2018-04-27 Jeff Law <law@redhat.com>
1551 * config.gcc: Mark tile* targets as deprecated/obsolete.
1553 2018-04-27 Richard Biener <rguenther@suse.de>
1555 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1558 2018-04-27 Richard Biener <rguenther@suse.de>
1560 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1562 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1564 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1565 with Yd constraint. Set "preferred_for_speed" attribute from
1566 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1568 (*movdi_internal): Ditto.
1569 (movti_interunit splitters): Remove
1570 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1571 (movdi_interunit splitters): Ditto.
1572 * config/i386/constraints.md (Ye): Remove.
1573 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1575 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1578 * config/aarch64/constraints.md (Usg): Limit to 31.
1581 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1583 PR tree-optimization/85529
1584 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1585 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1586 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1587 zero extension or masking of the MSB bit.
1588 (optimize_range_tests): Add FIRST_BB argument, pass it through
1589 to optimize_range_tests_var_bound.
1590 (maybe_optimize_range_tests, reassociate_bb): Adjust
1591 optimize_range_tests callers.
1593 2018-04-26 Richard Biener <rguenther@suse.de>
1594 Jakub Jelinek <jakub@redhat.com>
1596 * cgraph.h (symbol_table): Just declare debug method here.
1597 * symtab.c (symbol_table::debug): Define.
1599 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1601 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1603 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1605 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1606 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1607 (*movdi_internal): Substitute Yi and Yj constraint with x
1608 and Ym and Yn constraint with y constraint. Update "isa"
1609 attribute and set "preferred_for_speed" attribute from
1610 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1611 (*movsi_internal): Ditto.
1612 (*movdf_internal): Ditto.
1613 (*movsf_internal): Ditto.
1614 (*zero_extendsidi2): Ditto.
1615 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1616 (sse2_loadld): Ditto.
1617 (*vec_extract<ssevecmodelower>_0): Ditto.
1618 (*vec_extractv4si_0_zext_sse4): Ditto.
1619 (vec_concatv2di): Ditto.
1620 (*vec_dup<mode>): Ditto.
1621 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1622 * config/i386/constraints.md (Yi): Remove.
1627 2018-04-26 Nathan Sidwell <nathan@acm.org>
1629 * dumpfile.c (dump_open): New.
1630 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1631 (dump_finish): Detect stdio/stderr by value not name.
1633 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1635 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1637 2018-04-26 Tom de Vries <tom@codesourcery.com>
1640 * config/nvptx/nvptx.c (verify_neutering_jumps)
1641 (verify_neutering_labels): New function
1642 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1644 2018-04-26 Tom de Vries <tom@codesourcery.com>
1647 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1648 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1651 2018-04-26 Richard Biener <rguenther@suse.de>
1652 Tom de Vries <tom@codesourcery.com>
1655 * lto-streamer-out.c (output_function): Fixup loops if required to match
1656 discovery done in the reader.
1658 2018-04-26 Richard Biener <rguenther@suse.de>
1660 PR tree-optimization/85116
1661 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1662 have a loop exit from the single latch predecessor. Remove
1663 case of header with just condition.
1664 (ch_base::copy_headers): Exclude infinite loops from any
1666 (pass_ch::execute): Record exits.
1668 2018-04-26 Richard Biener <rguenther@suse.de>
1670 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1671 prologue cost vector and pass it to vect_get_load_cost.
1672 (vect_get_peeling_costs_all_drs): Likewise.
1673 (vect_peeling_hash_get_lowest_cost): Likewise.
1674 (vect_enhance_data_refs_alignment): Likewise.
1676 2018-04-26 Richard Biener <rguenther@suse.de>
1679 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1680 checking of integer<->pointer conversions.
1681 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1682 sign-/zero-extending pointer types.
1683 (expand_omp_for_static_chunk): Likewise.
1685 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1686 Jean Lee <xiaoyur347@gmail.com>
1688 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1689 (TARGET_ASAN_SHADOW_OFFSET): Define.
1690 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1691 true for -fsanitize=address.
1693 2018-04-25 Mark Wielaard <mark@klomp.org>
1695 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1698 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1700 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1701 than "alu", remove explicit "memory" and "imm_disp" attributes.
1702 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1705 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1706 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1707 gen_lowpart_no_emit.
1709 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1712 * config/i386/i386.c (ix86_expand_builtin): Change memory
1713 operand to XI, extend p0 to Pmode.
1714 * config/i386/i386.md: Change unspec volatile and operand
1715 1 mode to XI, change operand 0 mode to P.
1717 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1719 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1720 GET_MODE_MASK before any checking.
1721 (nds32_can_use_bset_p): Likewise.
1722 (nds32_can_use_btgl_p): Likewise.
1724 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1726 * config/nds32/nds32-doubleword.md: New define_split pattern for
1727 illegal register number.
1729 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1731 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1733 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1735 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1737 2018-04-25 Richard Biener <rguenther@suse.de>
1739 * lto-streamer.h (LTO_major_version): Bump to 8.
1741 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1743 * BASE-VER: Set to 9.0.0.
1745 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1747 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1748 in __abskf2 and __powikf2.
1750 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1753 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1754 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1755 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1756 Use the above on operand 2. Reindent.
1757 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1759 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1762 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1764 * config/i386/i386.opt (mcet): Removed.
1765 * doc/install.texi: Remove -mcet documentation.
1766 * doc/invoke.texi: Likewise.
1768 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1771 * doc/install.texi: Remove -mcet from bootstrap-cet.
1773 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1776 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1777 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1781 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1782 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1783 containing a CONST_VECTOR.
1785 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1787 * doc/install.texi: Update newlib dependency for nvptx.
1789 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1792 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1793 instead of INTVAL when shifting x left.
1795 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1797 PR tree-optimization/85478
1798 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1799 vect_grouped_store_supported for single element vectors.
1801 2018-04-24 Richard Biener <rguenther@suse.de>
1804 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1805 load cost increase to the case of non-constant step.
1807 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1810 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1811 destination if any_malformed_asm.
1813 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1816 * expr.c (store_field): In the bitfield case, if the value comes from
1817 a function call and is returned in registers by means of a PARALLEL,
1818 do not change the mode of the temporary unless BLKmode and VOIDmode.
1820 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1822 PR rtl-optimization/85423
1823 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1824 dependencies to debug insns when the previous insn is non-debug.
1826 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1828 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1829 enums into a single definition.
1830 (fls): Fix predicates and printing.
1833 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1835 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1836 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1837 and short u6 immediate.
1838 (check_if_valid_sleep_operand): Remove.
1839 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1841 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1843 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1844 flag_always_save_lp condition.
1845 * config/nds32/nds32.opt (malways-save-lp): New option.
1847 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1849 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1850 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1851 * config/nds32/nds32.h
1852 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1853 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1855 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1857 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1858 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1860 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1861 Chung-Ju Wu <jasonwucj@gmail.com>
1863 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1865 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1866 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1868 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1870 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1872 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1874 * config/nds32/nds32-protos.h (nds32_data_alignment,
1875 nds32_local_alignment): Declare.
1876 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1877 nds32_local_alignment): New functions.
1878 (TARGET_CONSTANT_ALIGNMENT): Define.
1879 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1881 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1883 * config/nds32/nds32.c
1884 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1885 (TARGET_MODES_TIEABLE_P): Likewise.
1887 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1889 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1892 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1893 Chung-Ju Wu <jasonwucj@gmail.com>
1895 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1896 for unaligned access.
1897 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1899 * config/nds32/nds32-intrinsic.md: Likewise.
1900 * config/nds32/nds32_intrinsic.h: Likewise.
1901 * config/nds32/nds32.h (nds32_builtins): Likewise.
1902 * config/nds32/nds32.opt (munaligned-access): New option.
1903 * config/nds32/nds32.c (nds32_asm_file_start): Display
1904 flag_unaligned_access status.
1906 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1908 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1909 -mno-relax is present.
1910 * config/riscv/linux.h (LINK_SPEC): Ditto.
1912 2018-04-20 Martin Sebor <msebor@redhat.com>
1915 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1917 (gimple_fold_builtin_stxcpy_chk): Same.
1918 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1920 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1923 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1924 __powikf2 when long double is IEEE 128-bit.
1926 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1928 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1929 step to make sure stack always aligned.
1931 2018-04-20 Carl Love <cel@us.ibm.com>
1934 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1935 size check for arg0.
1937 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1938 Tom de Vries <tom@codesourcery.com>
1941 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1942 Emit insns for calls too.
1943 (nvptx_find_par): Always look for worker-level predecessor insn.
1944 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1946 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1947 (nvptx_process_pars): Propagate frames for calls.
1949 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1952 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1954 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1955 (ix86_handle_option): Don't handle OPT_mibt.
1956 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1958 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1960 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1961 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1962 (ix86_target_macros): Define __CET__ with flag_cf_protection
1963 for -fcf-protection.
1964 * config/i386/i386.c (isa2_opts): Remove -mibt.
1965 * config/i386/i386.h (TARGET_IBT): Removed.
1966 (TARGET_IBT_P): Likewise.
1967 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1968 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1969 * config/i386/i386.opt (mcet): Update help message.
1972 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1973 -mcet as an alias for -mshstk.
1975 2018-04-20 Richard Biener <rguenther@suse.de>
1978 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1979 complexity by forcing a single use of the multiply operand.
1981 2018-04-20 Martin Jambor <mjambor@suse.cz>
1984 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1985 recursion dependency to only apply to non-clones.
1987 2018-04-20 Martin Jambor <mjambor@suse.cz>
1990 * ipa-cp.c (create_specialized_node): Check that clones of
1991 self-recursive edges exist during IPA-CP.
1993 2018-04-19 Toon Moene <toon@moene.org>
1995 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1998 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2000 PR tree-optimization/85467
2001 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2002 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2003 VECTOR_CST element to type.
2005 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2008 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2009 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2010 (builtin_longjmp): Likewise.
2011 (save_stack_nonlocal): New pattern.
2012 (restore_stack_nonlocal): Likewise.
2014 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2017 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2018 Replace ASM_OUTPUT_LABEL with fprintf.
2020 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2023 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2024 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2025 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2026 define __IBT__ and __SHSTK__ for -fcf-protection.
2027 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2029 (ix86_trampoline_init): Likewise.
2030 (x86_output_mi_thunk): Likewise.
2031 (ix86_notrack_prefixed_insn_p): Likewise.
2032 (ix86_option_override_internal): Don't disallow -fcf-protection.
2033 * config/i386/i386.md (rdssp<mode>): Also enable for
2035 (incssp<mode>): Likewise.
2036 (nop_endbr): Likewise.
2037 * config/i386/i386.opt (mcet): Change help message to built-in
2041 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2042 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2043 enable CET built-in functions.
2045 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2047 * common/config/i386/i386-common.c
2048 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2049 OPTION_MASK_ISA_MOVDIRI_UNSET,
2050 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2051 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2052 * config.gcc (movdirintrin.h): New header.
2053 * config/i386/cpuid.h (bit_MOVDIRI,
2054 bit_MOVDIR64B): New bits.
2055 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2057 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2058 (VOID, PVOID, PCVOID)): New function types.
2059 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2060 __builtin_ia32_directstoreu_u64,
2061 __builtin_ia32_movdir64b): New builtins.
2062 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2063 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2065 (ix86_valid_target_attribute_inner_p): Ditto.
2066 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2067 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2068 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2069 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2070 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2071 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2072 (movdiri<mode>, movdir64b_<mode>): New.
2073 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2074 * config/i386/immintrin.h: Include movdirintrin.h.
2075 * config/i386/movdirintrin.h: New file.
2076 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2078 2018-04-19 Richard Biener <rguenther@suse.de>
2081 * cfg.c (clear_bb_flags): When loop state says we have
2082 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2084 2018-04-19 Richard Biener <rguenther@suse.de>
2086 PR tree-optimization/84737
2087 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2088 copying restrict info.
2089 (vect_setup_realignment): Use it.
2090 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2091 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2092 the first DR to all generated stores.
2093 (vectorizable_load): Likewise for loads.
2095 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2097 PR tree-optimization/85446
2098 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2099 the integral and pointer types to have the same precision.
2101 * doc/install.texi: Document --disable-cet being the default and
2104 2018-04-18 Martin Liska <mliska@suse.cz>
2106 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2109 2018-04-18 Martin Liska <mliska@suse.cz>
2112 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2115 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2116 arguments if they are comparable.
2118 2018-04-18 Martin Liska <mliska@suse.cz>
2121 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2124 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2127 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2130 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2131 ENDBR after calling __morestack.
2133 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2136 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2137 by using gcc_base_ver to generate a gcc_driver_version, and use
2138 it when generating GCC_DRIVER_NAME.
2139 * configure: Regenerate.
2141 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2144 * config.gcc: Obsolete powerpc*-*-*spe*.
2146 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2149 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2150 (stabstr_D): Change type of unum from unsigned int to
2151 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2154 2018-04-17 Jim Wilson <jimw@sifive.com>
2157 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2158 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2159 Set arg_pointer_offset after using pretend_args_size.
2161 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2163 PR rtl-optimization/85431
2164 * dse.c (record_store): Ignore zero width stores.
2167 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2168 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2169 __builtin_stack_restore rather than after it.
2170 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2171 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2172 argument instead of virtual_dynamic_stack_rtx.
2174 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2176 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2178 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2179 Add note to error message to explain internal mapping of overloaded
2180 built-in function name to non-overloaded built-in function name.
2181 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2184 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2187 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2188 where the inputs overlap with the output.
2190 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2192 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2193 (=v, v) alternative and explicit "memory" attribute.
2194 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2195 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2197 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2198 "sselog1" type instead of "sselog".
2199 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2200 "sselog". Remove explicit "memory" attribute.
2201 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2202 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2204 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2205 "sselog1" type instead of "sselog". Remove explicit "memory"
2207 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2208 use "sselog1" type instead of "sselog". Remove explicit "memory"
2210 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2211 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2213 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2214 "sselog1" type instead of "sselog". Remove explicit "memory"
2216 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2217 use "sselog1" type instead of "sselog". Remove explicit "memory"
2221 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2224 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2227 2018-04-17 Martin Jambor <mjambor@suse.cz>
2230 * ipa-cp.c (create_specialized_node): Call
2231 expand_all_artificial_thunks if necessary.
2233 2018-04-17 Martin Liska <mliska@suse.cz>
2236 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2237 in message, remote space in between '_G' and '('.
2239 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2242 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2243 avx512f_vmcmp<mode>3<round_saeonly_name>,
2244 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2245 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2246 avx512f_rndscale<mode><round_saeonly_name>,
2247 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2248 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2249 Use %<iptr>2 instead of %2 for -masm=intel.
2250 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2251 avx512f_vcvttss2usi<round_saeonly_name>,
2252 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2254 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2255 avx512f_vcvttsd2usi<round_saeonly_name>,
2256 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2257 Use %q1 instead of %1 for -masm=intel.
2258 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2259 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2260 of %3 for -masm=intel.
2261 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2263 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2265 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2267 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2268 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2270 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2272 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2273 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2274 %g1 and one with %0 and %1.
2275 (avx512er_vmrcp28<mode><round_saeonly_name>,
2276 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2278 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2279 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2280 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2281 of %0 and %{%4%} for -masm=intel.
2282 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2283 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2284 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2285 order of %0 and %{%5%}%{z%} for -masm=intel.
2287 2018-04-17 Jan Hubicka <jh@suse.cz>
2290 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2292 2018-04-17 Martin Liska <mliska@suse.cz>
2295 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2296 for target_clone error message. Make default implementation
2297 clone to be a local declaration.
2298 (separate_attrs): Add new argument and check for an empty
2300 (expand_target_clones): Handle it.
2301 (ipa_target_clone): Make redirection just for target_clones
2304 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2305 Tom de Vries <tom@codesourcery.com>
2308 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2309 tiled basic blocks without omp continue statements.
2311 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2314 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2315 vec_extract expression as having side effects to make sure it gets
2318 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2321 * config/i386/i386.c (get_builtin_code_for_version): Check
2324 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2327 * gcc/config.gcc: Support "skylake".
2328 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2330 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2331 (processor_target_table): Add "skylake".
2332 (ix86_option_override_internal): Add "skylake".
2333 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2334 PROCESSOR_CANNONLAKE.
2335 (get_builtin_code_for_version): Fix priority for
2336 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2337 PROCESSOR_SKYLAKE-AVX512.
2338 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2339 (processor_type): Add PROCESSOR_SKYLAKE.
2341 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2342 Jason Merrill <jason@redhat.com>
2345 * convert.c (convert_to_integer_1): Use direct recursion for
2346 enumeral types and types with a precision less than the number
2347 of bits in their mode.
2349 2018-04-16 Julia Koval <julia.koval@intel.com>
2352 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2353 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2355 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2358 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2359 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2360 and -mno-direct-move.
2362 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2365 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2366 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2367 Ensure negative shifts result in {0}.
2369 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2371 PR rtl-optimization/79916
2372 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2373 regs (if any) to define how to gnerate SD moves when LRA is in
2376 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2378 PR rtl-optimization/85393
2379 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2380 * except.c (expand_dw2_landing_pad_for_region): Make static.
2381 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2382 a label and unconditional jump to old_bb, rather than
2383 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2386 PR rtl-optimization/85376
2387 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2388 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2389 instead of a specific value.
2391 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2392 Bin Cheng <bin.cheng@arm.com>
2394 PR tree-optimization/82965
2395 PR tree-optimization/83991
2396 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2397 by_profile_only parameter.
2398 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2399 information if the loop was predicted to iterate too many times.
2400 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2402 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2405 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2408 2018-04-13 Martin Liska <mliska@suse.cz>
2409 Jakub Jelinek <jakub@redhat.com>
2412 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2413 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2414 * builtins.c (expand_builtin_memory_copy_args): Use
2415 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2416 handle dest_addr == pc_rtx.
2418 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2421 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2422 asked to not generate direct moves.
2423 (fix_trunc<mode>si2_stfiwx): Similar.
2424 (fix_trunc<mode>si2_internal): Similar.
2426 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2429 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2430 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2431 lookup if dest in some wider mode is known to be const0_rtx and
2432 if so, record permanent equivalence for it to be ZERO_EXTEND of
2433 the narrower mode destination.
2435 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2437 * lto-streamer-out.c (output_function): Revert 259346.
2438 * omp-expand.c (expand_oacc_for): Likewise.
2440 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2442 PR rtl-optimization/85354
2443 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2444 * sel-sched.c (sel_global_init): ... here.
2446 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2449 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2450 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2451 mode for PE-COFF targets.
2452 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2453 (i386_pe_asm_lto_end): Likewise.
2454 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2455 (TARGET_ASM_LTO_END): Likewise.
2456 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2457 (i386_pe_asm_lto_start): New function.
2458 (i386_pe_asm_lto_end): Likewise.
2460 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2461 Richard Biener <rguenther@suse.de>
2464 * lto-streamer-out.c (output_function): Fix CFG loop state before
2466 * omp-expand.c (expand_oacc_for): Handle calls to internal
2467 functions like regular functions.
2469 2018-04-12 Richard Biener <rguenther@suse.de>
2472 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2473 for the early LTO debug to properly generate references to it
2474 during DIE emission. Do not re-use that for the skeleton for
2476 (dwarf2out_early_finish): Likewise.
2478 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2481 * config/i386/sse.md
2482 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2483 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2484 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2485 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2486 and output is a reg, avoid creating invalid lowpart subreg, but
2487 instead split into a 512-bit move. Don't split if not AVX512VL,
2488 input is xmm16+ reg and output is a mem.
2489 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2490 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2491 xmm16+ reg and output is a mem.
2493 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2495 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2496 also for flag_dwarf2_cfi_asm.
2498 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2500 PR rtl-optimization/85342
2501 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2502 a bool scalar var inside of the loop instead. Don't try to update
2503 recog_data.operand after failed apply_change_group.
2505 2018-04-12 Tom de Vries <tom@codesourcery.com>
2508 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2509 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2510 array with flexible array member as array without given dimension.
2511 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2512 argument for undefined param to true.
2514 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2517 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2518 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2519 from PowerPC section.
2520 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2521 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2522 help text that is too long.
2523 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2524 help text that is too long.
2525 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2526 help text that is too long.
2528 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2530 * config/alpha/alpha.md (stack_probe_internal): Rename
2531 from "probe_stack". Update all callers.
2533 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2535 PR rtl-optimization/84566
2536 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2537 sched_macro_fuse_insns.
2539 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2542 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2543 (compute_block_dependences): ... from here.
2545 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2547 PR tree-optimization/85331
2548 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2549 from int to HOST_WIDE_INT.
2551 2018-04-11 Martin Jambor <mjambor@suse.cz>
2554 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2555 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2556 not the same as the source val.
2557 (cgraph_edge_brings_value_p): New parameter.
2558 (gather_edges_for_value): Pass destination value to
2559 cgraph_edge_brings_value_p.
2560 (perhaps_add_new_callers): Likewise.
2561 (get_info_about_necessary_edges): Likewise and exclude values brought
2562 only by self-recursive edges.
2563 (create_specialized_node): Redirect only clones of self-calling edges.
2564 (+self_recursive_pass_through_p): New function.
2565 (find_more_scalar_values_for_callers_subset): Use it.
2566 (find_aggregate_values_for_callers_subset): Likewise.
2567 (known_aggs_to_agg_replacement_list): Removed.
2568 (decide_whether_version_node): Re-calculate known constants for all
2569 remaining context clones.
2571 2018-04-11 Richard Biener <rguenther@suse.de>
2574 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2575 from early DWARF output.
2576 (dwarf2out_early_finish): Output line info unconditionally into
2577 early DWARF and add reference to it.
2579 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2582 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2583 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2584 other than V2DFmode using iptr mode attribute.
2585 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2587 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2589 PR rtl-optimization/84659
2590 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2592 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2595 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2597 (output_loc_list): Pass address of a dummy size variable even in the
2598 locview handling loop.
2599 (index_location_lists): Add comment on why skip_loc_list_entry can't
2602 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2605 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2608 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2611 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2612 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2613 and -mstring-compare-inline-limit.
2615 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2618 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2619 for stack clash protection in a register whenever we need it to be in
2622 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2624 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2625 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2627 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2630 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2632 (mlong-double-): Ditto.
2633 * config/rs6000/sysv4.opt (msdata=): Ditto.
2634 (mtls-size=): Ditto.
2636 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2638 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2639 erroneous entries for
2640 "vector int vec_ldl (int, long int *)", and
2641 "vector unsigned int vec_ldl (int, unsigned long int *)".
2642 Add comments and entries for
2643 "vector bool char vec_ldl (int, bool char *)",
2644 "vector bool short vec_ldl (int, bool short *)",
2645 "vector bool int vec_ldl (int, bool int *)",
2646 "vector bool long long vec_ldl (int, bool long long *)",
2647 "vector pixel vec_ldl (int, pixel *)",
2648 "vector long long vec_ldl (int, long long *)",
2649 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2650 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2651 type tree bool_long_long_type_node and correct definition of
2652 bool_V2DI_type_node to make reference to this new type tree.
2653 (rs6000_mangle_type): Replace erroneous reference to
2654 bool_long_type_node with bool_long_long_type_node.
2655 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2656 comments to emphasize sign distinctions for char and int types and
2657 replace RS6000_BTI_bool_long constant with
2658 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2659 use of RS6000_BTI_pixel.
2660 (bool_long_type_node): Remove this macro definition.
2661 (bool_long_long_type_node): New macro definition
2663 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2665 PR rtl-optimization/85300
2666 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2667 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2668 simplify_unary_operation fails.
2670 2018-04-10 Martin Liska <mliska@suse.cz>
2672 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2673 cgraph_edge and ipa_ref.
2675 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2679 * config/i386/sse.md
2680 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2681 computation of the VEC_MERGE selector from mask.
2682 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2683 Fix decoding of the VEC_MERGE selector into mask.
2685 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2687 PR tree-optimization/85286
2688 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2690 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2692 * final.c (final_1): Set insn_last_address as well as
2693 insn_current_address.
2695 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2698 * explow.c (emit_stack_probe): Call validize_mem on memory location
2699 before passing it to gen_probe_stack. Create address operand and
2700 legitimize it for the probe_stack_address case.
2702 2018-04-09 Jan Hubicka <jh@suse.cz>
2705 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2706 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2707 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2709 (free_lang_data): Rebuild type inheritance graph.
2711 2018-04-09 Martin Sebor <msebor@redhat.com>
2713 * invoke.texi (-finline-small-functions): Mention other optimization
2715 (-findirect-inlining, -fpartial-inlining): Same.
2716 (-finline-functions-called-once): Same.
2717 (-freorder-blocks-and-partition): Same.
2719 2018-04-09 Jan Hubicka <jh@suse.cz>
2722 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2723 jumps; choose last target that matches the criteria (i.e.
2724 no partition changes for non-crossing jumps).
2725 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2726 support for redirecting crossing jumps to non-crossing.
2728 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2730 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2731 also for naked functions.
2733 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2735 * config/arc/arc.md (add_shift): New pattern.
2736 (add_shift2): Likewise.
2737 (sub_shift): Likewise.
2738 (sub_shift_cmp0_noout): Likewise.
2739 (compare_si_ashiftsi): Likewise.
2740 (xbfu_cmp0_noout): New combine pattern.
2741 (xbfu_cmp0"): Likewise.
2742 (movsi_set_cc_insn): Place the predicable variant first.
2743 (commutative_binary_cmp0_noout): Remove clobber.
2744 (commutative_binary_cmp0): New pattern.
2745 (noncommutative_binary_cmp0): Likewise.
2746 (noncommutative_binary_cmp0_noout): Likewise.
2747 (noncommutative_binary_comparison_result_used): Removed.
2748 (rsub_cmp0): New pattern.
2749 (rsub_cmp0_noout): Likewise.
2750 (extzvsi): Changed, keep only meaningful variants.
2751 (SQH, SEZ): New iterators.
2752 (SQH_postfix): New mode attribute.
2753 (SEZ_prefix): New code attribute.
2754 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2755 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2756 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2758 (noncommutative_operator): Check the availability of barrel
2761 2018-04-09 Richard Biener <rguenther@suse.de>
2763 PR tree-optimization/85284
2764 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2765 Only use the niter constraining form of simple_iv when the exit
2768 2018-04-09 Tom de Vries <tom@codesourcery.com>
2771 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2772 (define_expand "*memory_barrier"): New define_expand.
2773 (define_insn "memory_barrier"): New insn.
2775 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2777 PR rtl-optimization/80463
2778 PR rtl-optimization/83972
2779 PR rtl-optimization/83480
2781 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2782 correct producer for the insn.
2783 (tidy_control_flow): Fixup seqnos in case of debug insns.
2785 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2787 PR rtl-optimization/83913
2789 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2790 different sched-times when merging exprs.
2792 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2794 PR rtl-optimization/83962
2796 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2797 tidy_fallthru_edge and tidy_control_flow.
2799 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2801 PR rtl-optimization/83530
2803 * sel-sched.c (force_next_insn): New global variable.
2804 (remove_insn_for_debug): When force_next_insn is true, also leave only
2805 next insn in the ready list.
2806 (sel_sched_region): When the region wasn't scheduled, make another pass
2807 over it with force_next_insn set to 1.
2809 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2811 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2813 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2814 for interrupt control.
2815 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2816 functions for interrupt control.
2817 * config/nds32/nds32-intrinsic.md: Likewise.
2818 * config/nds32/nds32_intrinsic.h: Likewise.
2819 * config/nds32/nds32.h (nds32_builtins): Likewise.
2821 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2823 * config/nds32/nds32.c (nds32_init_machine_status,
2824 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2825 strict_aligned_p field.
2826 (nds32_expand_to_rtl_hook): New function.
2827 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2828 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2830 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2831 Chung-Ju Wu <jasonwucj@gmail.com>
2833 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2834 * config/nds32/nds32-n7.md: New file.
2835 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2836 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2838 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2839 * config/nds32/nds32.md (pipeline_model): Add n7.
2840 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2841 * config/nds32/pipelines.md: Include n7 settings.
2843 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2844 Chung-Ju Wu <jasonwucj@gmail.com>
2846 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2847 * config/nds32/nds32-e8.md: New file.
2848 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2849 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2851 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2852 * config/nds32/nds32.md (pipeline_model): Add e8.
2853 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2854 * config/nds32/pipelines.md: Include e8 settings.
2856 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2857 Chung-Ju Wu <jasonwucj@gmail.com>
2859 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2860 * config/nds32/nds32-n8.md: New file.
2861 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2862 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2864 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2865 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2866 * config/nds32/nds32.md (pipeline_model): Add n8.
2867 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2868 * config/nds32/pipelines.md: Include n8 settings.
2870 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2871 Chung-Ju Wu <jasonwucj@gmail.com>
2873 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2874 * config/nds32/nds32-n9-2r1w.md: New file.
2875 * config/nds32/nds32-n9-3r2w.md: New file.
2876 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2877 nds32_register_ports): New or modify for cpu n9.
2878 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2880 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2881 * config/nds32/nds32-utils.c: New file.
2882 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2883 TARGET_MUL_SLOW): Define.
2884 * config/nds32/nds32.md (pipeline_model): New attribute.
2885 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2886 New options that support cpu n9.
2887 * config/nds32/pipelines.md: Include n9 settings.
2888 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2890 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2892 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2893 information if necessary.
2894 (output_cond_branch_compare_zero): Likewise.
2895 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2896 (nds32_target_alignment): Refine for alignment.
2897 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2898 (FUNCTION_BOUNDARY): Modify.
2899 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2901 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2903 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2905 * config/nds32/constants.md (unspec_volatile_element): Add values for
2906 TLB operation and data prefetch.
2907 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2908 functions for TLB operation and data prefetch.
2909 * config/nds32/nds32-intrinsic.md: Likewise.
2910 * config/nds32/nds32_intrinsic.h: Likewise.
2911 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2912 (nds32_print_operand): Likewise.
2913 * config/nds32/nds32.h (nds32_builtins): Likewise.
2915 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2916 Andrew Pinski <pinsika@gcc.gnu.org>
2919 * match.pd: Use constant_boolean_node of correct type instead of
2920 boolean_true_node or boolean_false_node for simplifying
2921 pointer comparisons to zero.
2923 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2925 PR tree-optimization/80021
2926 * tree.c (verify_type_variant): Make error call in verify_variant_match
2927 translatable and remove final full stop.
2929 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2931 * config/nds32/constants.md (unspec_volatile_element): Add
2932 UNSPEC_VOLATILE_EH_RETURN.
2933 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2934 nds32_output_stack_pop): Support dwarf exception handling process.
2935 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2936 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2937 exception handling process.
2938 (nds32_compute_stack_frame): Likewise.
2939 (nds32_return_addr_rtx): Likewise.
2940 (nds32_initial_elimination_offset): Likewise.
2941 (nds32_expand_prologue): Likewise.
2942 (nds32_expand_epilogue): Likewise.
2943 (nds32_dynamic_chain_address): New function.
2944 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2946 (DYNAMIC_CHAIN_ADDRESS): Define.
2947 (EH_RETURN_DATA_REGNO): Define.
2948 (EH_RETURN_STACKADJ_RTX): Define.
2949 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2950 patterns for dwarf exception handling.
2952 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2954 * config/nds32/nds32.h: Clean up obsolete macros.
2956 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2958 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2959 Add enum values for particular instructions.
2960 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2961 particular intrinsic functions.
2962 * config/nds32/nds32-intrinsic.md: Likewise.
2963 * config/nds32/nds32_intrinsic.h: Likewise.
2964 * config/nds32/nds32.h (nds32_builtins): Likewise.
2965 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2966 (btst, ave): New patterns for particular instructions.
2968 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2970 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2971 Add enum values for atomic load/store and memory sync.
2972 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2974 * config/nds32/nds32-intrinsic.md: Likewise.
2975 * config/nds32/nds32_intrinsic.h: Likewise.
2976 * config/nds32/nds32.h (nds32_builtins): Likewise.
2978 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2980 PR tree-optimization/85257
2981 * fold-const.c (native_encode_vector): If not all elts could fit
2982 and off is -1, return 0 rather than offset.
2983 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2984 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2985 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2986 adjust buffer in native_interpret_expr call.
2988 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2990 * config/nds32/constants.md (unspec_volatile_element): Add cache
2991 control enum values.
2992 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2993 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2994 * config/nds32/nds32.c (nds32_cctl_names): New.
2995 (nds32_print_operand): Handle cache control register names.
2996 * config/nds32/nds32.h (nds32_builtins): New enum values.
2997 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2999 * config/nds32/nds32.md (type): Add mmu.
3000 * config/nds32/pipelines.md (simple_insn): Add mmu.
3002 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3004 * config/nds32/nds32.md (type): Remove call.
3005 * config/nds32/pipelines.md (simple_insn): Likewise.
3007 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3009 * config/nds32/constants.md (unspec_volatile_element): Add
3010 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3011 UNSPEC_VOLATILE_FMFCFG.
3012 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3013 description for fmfcfg and fmfcsr.
3014 (bdesc_1arg): Add fmtcsr.
3015 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3016 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3017 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3018 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3019 unspec_fmfcfg): New patterns.
3020 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3021 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3022 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3023 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3024 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3025 __nds32__fmfcfg): Define.
3027 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3029 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3030 intrinsic register names.
3031 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3032 intrinsic register enum values and macros.
3034 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3036 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3037 for load/store addressing form.
3038 (nds32_print_operand_address): Likewise.
3040 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3043 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3044 based on LABEL_REF. Remove useless assertion.
3045 (pic_address_needs_scratch): Fix formatting.
3046 (sparc_legitimize_pic_address): Minor tweaks.
3047 (sparc_delegitimize_address): Adjust assertion accordingly.
3048 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3049 into symbolic_operand.
3050 (movsi_high_pic_label_ref): Likewise.
3051 (movsi_lo_sum_pic_label_ref): Likewise.
3052 (movdi_pic_label_ref): Likewise.
3053 (movdi_high_pic_label_ref): Likewise.
3054 (movdi_lo_sum_pic_label_ref): Likewise.
3056 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3058 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3059 custom LIB_SPEC setup.
3061 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3062 Kito Cheng <kito.cheng@gmail.com>
3064 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3065 * config/riscv/freebsd.h: New.
3067 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3069 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3070 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3073 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3074 Kito Cheng <kito.cheng@gmail.com>
3076 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3077 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3078 * config/nds32/nds32-protos.h (nds32_output_call,
3079 nds32_output_return): Declare.
3080 * config/nds32/nds32.md: Refine all the call and return patterns.
3082 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3085 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3086 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3088 PR rtl-optimization/84872
3089 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3090 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3093 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3095 * expr.c (copy_blkmode_to_reg): Revert 254862.
3096 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3098 2018-04-06 Richard Biener <rguenther@suse.de>
3101 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3102 after seeing a component reference with an adjacent field. Treat
3103 refs to arrays at struct end of external decls similar to
3104 refs to unconstrained commons.
3106 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3109 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3110 look through SAVE_EXPRs with non-side-effects argument. Adjust
3112 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3115 2018-04-06 Richard Biener <rguenther@suse.de>
3118 * alias.c (find_base_term): New wrapper around find_base_term
3119 unwinding CSELIB_VAL_PTR changes.
3120 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3123 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3125 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3127 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3128 constant definitions.
3129 ("nop"): lr 0,0 -> nopr r0
3130 ("nop_lr0", "nop_lr1"): New insn definitions.
3132 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3134 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3135 NDS32_V3PUSH_AVAILABLE_P macro.
3137 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3138 Chung-Ju Wu <jasonwucj@gmail.com>
3140 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3141 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3142 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3143 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3144 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3145 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3146 * config/nds32/constraints.md: New constraints and checking for hard
3147 float configuration.
3148 * config/nds32/iterators.md: New mode iterator and attribute for hard
3149 float configuration.
3150 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3152 * config/nds32/nds32-fpu.md: New file.
3153 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3154 deal with hard float code generation.
3155 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3157 (abi_type, float_reg_number): New enum type.
3158 * config/nds32/nds32-predicates.c: New predicates for hard float.
3159 * config/nds32/nds32-protos.h: Declare functions for hard float.
3160 * config/nds32/nds32.c: Implementation for hard float configuration.
3161 * config/nds32/nds32.h: Definitions for hard float configuration.
3162 * config/nds32/nds32.md: Include hard float machine description and
3163 modify patterns for hard float configuration.
3164 * config/nds32/nds32.opt: New options for hard float configuration.
3165 * config/nds32/predicates.md: New predicates for hard float
3168 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3170 * common/config/nds32/nds32-common.c
3171 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3173 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3176 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3177 CONSTRUCTOR_ELT (ctor, ...)->value.
3179 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3182 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3184 2018-04-05 Tom de Vries <tom@codesourcery.com>
3187 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3190 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3191 Kito Cheng <kito.cheng@gmail.com>
3193 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3194 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3195 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3196 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3198 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3199 Kito Cheng <kito.cheng@gmail.com>
3201 * config/nds32/constraints.md (Ufe): New memory constraint.
3202 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3203 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3204 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3206 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3207 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3209 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3211 * config/nds32/nds32.md: Use optimize_size in the condition for
3212 alu-shift instructions.
3214 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3216 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3218 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3220 * config/nds32/nds32.md (negsi2): Refine pattern.
3222 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3223 Chung-Ju Wu <jasonwucj@gmail.com>
3225 * config/nds32/iterators.md (shift_rotate): New code iterator.
3226 (shift): New code attribute.
3227 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3228 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3229 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3230 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3231 bit-wise operations.
3232 (andsi3, *andsi3): Ditto.
3233 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3234 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3235 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3236 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3237 nds32_ior_operand, nds32_xor_operand): New predicates.
3239 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3241 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3242 (addsi3, subsi3): ... this.
3244 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3246 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3248 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3250 * config/nds32/nds32.md: Adjust indention.
3252 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3254 * config/nds32/nds32.md (feature): New attribute.
3256 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3258 * config/nds32/nds32.md (subtype): New attribute.
3260 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3263 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3264 expansion to perform a bitwise AND of the argument followed by a
3265 boolean negation of the result.
3267 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3269 PR rtl-optimization/84878
3270 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3271 the basic block. Assert the use reference is not artificial and that
3272 it has an associated insn.
3274 2018-04-04 Michael Matz <matz@suse.de>
3276 * builtins.c (compute_objsize): Pass correct operand
3277 to array_at_struct_end_p.
3279 2018-04-04 Richard Biener <rguenther@suse.de>
3282 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3283 from contexts for DINFO_LEVEL_TERSE and below.
3285 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3287 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3288 resiter_operand condition.
3289 * config/nds32/nds32.md (*move<mode>): Ditto.
3291 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3292 Monk Chiang <sh.chiang04@gmail.com>
3294 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3296 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3298 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3300 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3301 Kito Cheng <kito.cheng@gmail.com>
3303 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3304 nds32_cond_code_str, output_cond_branch,
3305 output_cond_branch_compare_zero, nds32_expand_cbranch,
3306 nds32_expand_cstore, nds32_expand_movcc,
3307 nds32_output_cbranchsi4_equality_zero,
3308 nds32_output_cbranchsi4_equality_reg,
3309 nds32_output_cbranchsi4_equality_reg_or_const_int,
3310 nds32_output_cbranchsi4_greater_less_zero: New functions.
3311 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3312 nds32_expand_cstore, nds32_expand_movcc,
3313 nds32_output_cbranchsi4_equality_zero,
3314 nds32_output_cbranchsi4_equality_reg,
3315 nds32_output_cbranchsi4_equality_reg_or_const_int,
3316 nds32_output_cbranchsi4_greater_less_zero): Declare.
3317 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3318 nds32_rimm11s_operand): New predicates.
3319 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3320 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3323 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3325 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3326 * config/nds32/nds32.md: Ditto.
3327 * config/nds32/pipelines.md: Ditto.
3329 2018-04-04 Richard Biener <rguenther@suse.de>
3331 PR tree-optimization/85168
3332 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3333 propagating abnormals.
3335 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3337 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3339 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3340 Kito Cheng <kito.cheng@gmail.com>
3342 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3343 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3344 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3345 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3346 * config/nds32/nds32.md (sibcall_internal): New.
3347 (sibcall_register): Remove.
3348 (sibcall_immediate): Remove.
3349 (sibcall_value_internal): New.
3350 (sibcall_value_register): Remove.
3351 (sibcall_value_immediate): Remove.
3352 * config/nds32/predicates.md (nds32_general_register_operand): New.
3353 (nds32_call_address_operand): New.
3355 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3357 PR rtl-optimization/85167
3358 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3359 bb_defs if *split_p, instead preinitialize it to NULL.
3361 PR tree-optimization/85156
3362 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3363 evaluating the argument multiple times.
3365 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3367 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3369 (_mm_cvtpd_ps): Likewise.
3370 (_mm_cvttpd_epi32): Likewise.
3371 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3372 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3373 vector, pixel, and bool following altivec.h include.
3375 2018-04-03 Martin Sebor <msebor@redhat.com>
3377 * doc/extend.texi (Common Function Attributes): Clarify.
3378 (const attribute): Likewise.
3379 (pure attribute): Likewise.
3381 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3384 * config/i386/i386.c (ix86_expand_vector_set): Use
3385 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3387 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3389 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3390 instructions when changing rounding bits to preserve precision bits
3391 in the x87 control word.
3393 2018-04-03 Martin Liska <mliska@suse.cz>
3395 PR tree-optimization/82491
3396 * rtl.h (strip_offset_and_add): Replace += suboffset with
3397 poly_uint64 () + suboffset.
3399 2018-03-29 Martin Liska <mliska@suse.cz>
3400 Martin Jambor <mjambor@suse.cz>
3403 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3404 param_type is not an integral or pointer type.
3406 2018-04-03 Richard Biener <rguenther@suse.de>
3408 * sese.h (recompute_all_dominators): Remove.
3410 2018-04-02 Martin Sebor <msebor@redhat.com>
3412 * doc/invoke.texi (-Wrestrict): Fix typos.
3414 2018-04-02 Jim Wilson <jimw@sifive.com>
3416 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3417 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3418 (<optab>di3, <optab>si3_extend): Likewise.
3419 (<optab>si3_mask, <optab>si3_mask_1): New.
3420 (<optab>di3_mask, <optab>di3_mask_1): New.
3421 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3422 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3423 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3425 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3427 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3430 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3432 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3433 (nds32_canonicalize_comparison): New function.
3435 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3436 Kito Cheng <kito.cheng@gmail.com>
3437 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3439 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3440 * config/nds32/constants.md (unspec_volatile_element): Add
3441 UNSPEC_VOLATILE_RELAX_GROUP.
3442 * config/nds32/nds32-relax-opt.c: New file.
3443 * config/nds32/nds32-predicates.c
3444 (nds32_symbol_load_store_p): New function.
3445 * config/nds32/nds32-protos.h
3446 (nds32_symbol_load_store_p): Declare function.
3447 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3448 * config/nds32/nds32.c
3449 (nds32_register_pass): New function to register pass.
3450 (nds32_register_passes): New function to register passes.
3451 * config/nds32/nds32.md (relax_group): New pattern.
3452 * config/nds32/nds32.opt (mrelax-hint): New option.
3453 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3455 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3457 * config/nds32/t-nds32: Modify files dependency.
3459 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3461 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3462 (PROFILE_HOOK): Define its implementation.
3464 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3466 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3467 type and 32-bit size.
3469 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3472 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3473 (V_128_256): New mode iterator.
3474 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3475 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3476 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3478 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3479 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3481 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3484 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3485 NaN inputs correctly.
3487 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3490 * config/rs6000/vsx.md (??r): New mode attribute.
3491 (*vsx_mov<mode>_64bit): Use it.
3492 (*vsx_mov<mode>_32bit): Likewise.
3494 2018-03-30 Martin Sebor <msebor@redhat.com>
3496 PR tree-optimization/84818
3497 * builtins.c (check_access): Use warning_n.
3499 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3502 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3504 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3507 2018-03-30 Julia Koval <julia.koval@intel.com>
3510 * x86-tune.def (movx, partial_reg_dependency): Enable for
3513 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3516 * lra-constraints.c (process_alt_operands): Move setting
3517 this_alternative_matches below.
3519 2018-03-29 Martin Liska <mliska@suse.cz>
3522 * doc/invoke.texi: Document how LTO works with debug info.
3523 Describe auto-load support of binutils. Mention 'x86-64'
3524 as valid option value of -march option.
3526 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3528 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3531 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3532 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3533 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3536 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3539 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3540 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3541 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3542 for RS6000_BTM_POWERPC64.
3543 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3544 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3545 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3550 2018-03-28 Carl Love <cel@us.ibm.com>
3553 2017-09-27 Carl Love <cel@us.ibm.com>
3555 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3556 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3557 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3560 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3562 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3563 instead of __vector bool.
3564 (_mm_max_pu8): Likewise.
3565 (_mm_min_pi16): Likewise.
3567 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3570 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3571 (DIVWEUO): Likewise.
3573 (DIVDEUO): Likewise.
3574 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3575 DIVWEUO and DIVDEUO.
3576 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3577 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3578 (div_extend): Likewise.
3579 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3581 (__builtin_divweuo): Likewise.
3582 (__builtin_divdeo): Likewise.
3583 (__builtin_divdeuo): Likewise.
3585 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3588 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3589 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3591 PR tree-optimization/82004
3592 * gimple-match-head.c (optimize_pow_to_exp): New function.
3593 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3594 Don't fold to exp if optimize_pow_to_exp is false.
3596 2018-03-28 Martin Liska <mliska@suse.cz>
3599 * calls.c (initialize_argument_information): Fix trailing space.
3600 * common.opt: Fix typo and provide better explanation for
3601 -fsanitize-coverage option.
3602 * config/i386/i386.opt: Fix typo.
3604 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3605 Martin Liska <mliska@suse.cz>
3608 * gimplify.c (asan_poison_variable): Don't do the check for
3609 gimplify_omp_ctxp here.
3610 (gimplify_decl_expr): Do it here.
3611 (gimplify_target_expr): Likewise.
3613 2018-03-28 Martin Liska <mliska@suse.cz>
3616 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3617 chkp_type_bounds_count if MPX is not enabled.
3619 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3621 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3623 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3626 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3627 function to create the function decl for complex long double
3628 multiply and divide for -mabi=ieeelongdouble.
3629 (init_float128_ieee): Call it.
3631 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3634 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3635 -fcf-protection=branch -mibt.
3636 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3638 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3641 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3643 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3646 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3647 extern array declarations.
3649 2018-03-27 Richard Biener <rguenther@suse.de>
3652 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3653 explicit single_use checks.
3655 2018-03-27 Richard Biener <rguenther@suse.de>
3657 PR tree-optimization/85082
3658 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3661 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3663 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3664 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3665 Turn on fasynchronous-unwind-tables and funwind-tables.
3667 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3670 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3671 (*bmi_blsr_<mode>_ccz): Ditto.
3673 2018-03-26 Tom de Vries <tom@codesourcery.com>
3675 PR tree-optimization/85063
3676 * omp-general.c (offloading_function_p): New function. Factor out
3678 * omp-offload.c (pass_omp_target_link::gate): ... here.
3679 * omp-general.h (offloading_function_p): Declare.
3680 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3681 with attribute omp declare target for offloading functions.
3683 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3685 PR tree-optimization/84005
3686 * tree-data-ref.h (get_base_for_alignment): Declare.
3687 * tree-data-ref.c (get_base_for_alignment_1): New function.
3688 (get_base_for_alignment): Likewise.
3689 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3690 get_base_for_alignment to find a suitable base object, instead
3691 of always using drb->base_address.
3693 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3696 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3697 known size by default.
3699 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3702 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3703 and non BLKmode operands.
3705 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3708 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3709 Clean up attributes.
3711 2018-03-23 Richard Biener <rguenther@suse.de>
3714 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3715 we are going to emit early debug for LTO.
3717 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3720 * function.c (match_asm_constraints_1): Don't optimize if input
3721 doesn't satisfy general_operand predicate for output's mode.
3724 * alias.c (write_dependence_p): Don't require for x_canonicalized
3725 non-VOIDmode if x has VOIDmode.
3728 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3729 just don't try to optimize it rather than assert it never happens.
3731 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3733 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3734 macro expansions for definition of ST_INTERNAL_<mode> and
3735 LD_INTERNAL_<mode> builtins.
3736 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3738 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3740 (altivec_expand_st_builtin): Likewise.
3741 (altivec_expand_builtin): Remove calls to deleted functions.
3742 (rs6000_address_for_altivec): Delete this function.
3743 * config/rs6000/vector.md: Remove expands for
3744 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3746 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3749 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3750 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3751 re-computing once computed.
3752 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3753 (arm_init_machine_status): Initialize
3754 machine->static_chain_stack_bytes.
3756 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3759 * doc/extend.texi: Add four new prototypes for vec_ld.
3760 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3761 definitions for more logical presentation.
3762 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3763 entries for V1TI variants of __builtin_altivec_ld builtin.
3764 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3765 handling of V1TI variant of LVX icode pattern.
3766 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3767 (rs6000_gimple_fold_builtin): Likewise.
3768 (altivec_init_builtins): Add code to define
3769 __builtin_altivec_lvx_v1ti function.
3771 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3774 * function.c (match_asm_constraints_1): Don't do the optimization
3775 if input isn't a REG, SUBREG, MEM or constant.
3777 2018-03-22 Tom de Vries <tom@codesourcery.com>
3779 PR tree-optimization/84956
3780 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3781 bb_has_abnormal_pred.
3783 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3786 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3787 DECL_INITIAL (decl) to decl at the end.
3788 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3791 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3793 * doc/extend.texi (__builtin_tgmath): Document when complex
3794 integer types are treated as _Complex _Float64.
3796 2018-03-21 Tom de Vries <tom@codesourcery.com>
3798 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3800 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3802 PR tree-optimization/84960
3803 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3804 if it is ENTRY block, move them into single succ of ENTRY in that case.
3806 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3808 PR tree-optimization/84811
3809 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3810 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3811 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3812 (known_subrange_p): Update accordingly. Cast each value involved
3813 in the size comparison, rather than casting the result of the
3816 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3818 PR tree-optimization/84982
3819 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3820 by flipping the least significant bit rather than all bits from
3821 bitpos to bitpos + bitsize - 1.
3823 2018-03-21 Nathan Sidwell <nathan@acm.org>
3825 * doc/extend.texi (Deprecated Features): Remove mention of
3826 long-deleted deprecations.
3828 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3831 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3832 * configure: Regenerate.
3834 2018-03-21 Tom de Vries <tom@codesourcery.com>
3836 PR tree-optimization/83126
3837 * tree-parloops.c (num_phis): New function.
3838 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3840 2018-03-21 Nathan Sidwell <nathan@acm.org>
3842 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3843 mention anon-struct/union members and trailing attributes.
3845 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3847 PR tree-optimization/84969
3848 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3849 builtin memset partitions if they set different rhs values.
3851 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3853 PR rtl-optimization/84989
3854 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3855 VEC_DUPLICATE with scalar result mode.
3857 2018-03-21 Martin Liska <mliska@suse.cz>
3860 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3861 not intended return statement.
3863 2018-03-21 Martin Liska <mliska@suse.cz>
3866 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3867 (chkp_find_bound_slots_1): Limit number of iterations.
3869 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3872 * Minor grammar fixes for x86 options.
3874 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3877 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3878 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3880 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3883 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3884 (altivec_lvx_<mode>_1op): Likewise.
3885 (altivec_stvx_<mode>_2op): Likewise.
3886 (altivec_stvx_<mode>_1op): Likewise.
3887 (altivec_lvx_<VM2:mode>): New define_expand.
3888 (altivec_stvx_<VM2:mode>): Likewise.
3889 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3890 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3891 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3892 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3893 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3894 (rs6000_gen_lvx): Likewise.
3895 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3896 (altivec_expand_stv_builtin): Likewise.
3897 (altivec_expand_builtin): Likewise.
3898 * config/rs6000/vector.md: Likewise.
3900 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3903 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3906 2018-03-20 Richard Biener <rguenther@suse.de>
3909 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3910 sign-conversions as zero, fall back to standard scalar_stmt
3913 2018-03-20 Martin Liska <mliska@suse.cz>
3916 * predict.c (rebuild_frequencies): Handle case when we have
3917 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3919 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3922 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3923 flag_section_anchors.
3924 * varasm.c (use_blocks_for_decl_p): Remove hack for
3925 dw2_force_const_mem.
3928 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3930 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3931 be created, use lowpart_subreg of operands[0] rather than operands[0]
3933 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3934 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3935 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3936 and n constraint instead of aarch64_shift_imm_di and Usd.
3937 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3938 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3940 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3943 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3944 to favor GPR over NEON registers.
3945 (<shift>di3_neon): Likewise.
3947 2018-03-20 Tom de Vries <tom@codesourcery.com>
3950 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3951 (nvptx_process_pars): Emit bar.sync asap and alap.
3953 2018-03-20 Tom de Vries <tom@codesourcery.com>
3956 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3957 seen_label if seen_label is already set.
3959 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3962 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3963 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3964 Use 1U instead of 1. Formatting fixes.
3967 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3968 instead of TREE_TYPE (s1) for the return value.
3970 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3972 PR tree-optimization/84946
3973 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3974 bitsize + bitsize in poly_uint64 rather than poly_int64.
3977 * dwarf2asm.c: Include fold-const.c.
3978 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3979 of decl rather than decl itself.
3981 PR rtl-optimization/84643
3982 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3984 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3987 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3988 calling assemble_variable.
3990 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3993 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3994 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3996 2018-03-19 Jim Wilson <jimw@sifive.com>
3999 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4000 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4001 (riscv_first_stack_step): Likewise.
4002 (riscv_option_override): Use STACK_BOUNDARY instead of
4004 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4006 (BIGGEST_ALIGNMENT): Set to 128.
4007 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4008 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4011 2018-03-19 Richard Biener <rguenther@suse.de>
4013 PR tree-optimization/84933
4014 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4015 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4017 2018-03-19 Richard Biener <rguenther@suse.de>
4019 PR tree-optimization/84859
4020 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4021 (cond_if_else_store_replacement): Perform sinking operation on
4022 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4023 Generalize what a BB with a single eligible store is.
4025 2018-03-19 Richard Biener <rguenther@suse.de>
4027 PR tree-optimization/84929
4028 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4029 chrec_is_positive against non-chrec arg.
4031 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4034 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4036 2018-03-18 Martin Liska <mliska@suse.cz>
4038 PR rtl-optimization/84635
4039 * regrename.c (build_def_use): Use matches_mode only when
4042 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4044 PR tree-optimization/84913
4045 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4046 vectorize chains of COND_EXPRs.
4048 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4050 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4052 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4054 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4056 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4058 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4060 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4061 Kito Cheng <kito.cheng@gmail.com>
4063 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4064 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4065 (nds32_adjust_reg_alloc_order): New function.
4066 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4068 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4070 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4071 nds32_print_operand, nds32_print_operand_address): Use
4072 HOST_WIDE_INT_PRINT_DEC instead.
4074 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4076 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4078 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4081 * config/i386/i386.c (initial_ix86_tune_features,
4082 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4084 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4085 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4086 rather than 1u << ix86_tune. Formatting fix.
4087 (ix86_option_override_internal): Change ix86_arch_mask from
4088 unsigned int to unsigned HOST_WIDE_INT, initialize to
4089 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4090 (ix86_function_specific_restore): Likewise.
4092 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4095 * postreload.c (reload_combine_recognize_pattern): Perform
4096 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4097 truncate_int_for_mode the result for the destination's mode.
4100 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4101 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4103 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4106 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4109 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4110 regno_allocno_class_array and sorted_pseudos.
4111 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4112 insns where regno is used.
4114 2018-03-16 Martin Liska <mliska@suse.cz>
4117 * multiple_target.c (create_dispatcher_calls): Redirect
4118 reference in the symbol table.
4120 2018-03-16 Martin Liska <mliska@suse.cz>
4123 * multiple_target.c (create_dispatcher_calls): Redirect also
4126 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4130 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4131 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4132 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4134 2018-03-16 Julia Koval <julia.koval@intel.com>
4136 * doc/invoke.texi (Skylake Server): Add CLWB.
4137 Cannonlake): Remove CLWB.
4139 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4141 PR tree-optimization/84841
4142 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4144 (FLOAT_ONE_CONST_TYPE): Define.
4145 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4146 (sort_by_operand_rank): Put entries with higher constant_type last
4147 rather than first to match comments.
4149 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4151 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4154 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4157 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4158 check and use address of target always.
4160 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4163 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4164 (indirect_thunk_bnd_needed): Likewise.
4165 (indirect_thunks_used): Likewise.
4166 (indirect_thunks_bnd_used): Likewise.
4167 (indirect_return_needed): New.
4168 (indirect_return_bnd_needed): Likewise.
4169 (output_indirect_thunk_function): Add a bool argument for
4171 (output_indirect_thunk_function): Don't generate alias for
4172 function return thunk.
4173 (ix86_code_end): Call output_indirect_thunk_function to generate
4174 function return thunks.
4175 (ix86_output_function_return): Set indirect_return_bnd_needed
4176 and indirect_return_needed instead of indirect_thunk_bnd_needed
4177 and indirect_thunk_needed.
4179 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4181 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4182 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4183 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4185 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4186 Paul Hua <paul.hua.gm@gmail.com>
4189 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4191 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4193 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4194 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4197 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4200 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4201 instead of GET_MODE_SIZE when comparing Units.
4203 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4206 * varasm.c (hash_section): Return an unchangeble hash value
4207 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4208 Return !aarch64_can_use_per_function_literal_pools_p ().
4210 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4213 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4214 rather than address of cmode as last argument to prepare_cmp_insn.
4216 2018-03-15 Julia Koval <julia.koval@intel.com>
4218 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4219 F_AVX512VNNI, F_AVX512BITALG): New.
4221 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4224 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4225 insn for floating-point loads and stores.
4227 2018-03-14 Carl Love <cel@us.ibm.com>
4229 * config/rs6000/rs6000-c.c: Add macro definitions for
4230 ALTIVEC_BUILTIN_VEC_PERMXOR.
4231 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4232 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4233 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4234 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4236 * config/doc/extend.texi: Add prototypes for vec_permxor.
4238 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4241 * diagnostic-show-locus.c (class layout_point): Convert m_line
4242 from int to linenum_type.
4243 (line_span::comparator): Use linenum "compare" function when
4244 comparing line numbers.
4245 (test_line_span): New function.
4246 (layout_range::contains_point): Convert param "row" from int to
4248 (layout_range::intersects_line_p): Likewise.
4249 (layout::will_show_line_p): Likewise.
4250 (layout::print_source_line): Likewise.
4251 (layout::should_print_annotation_line_p): Likewise.
4252 (layout::print_annotation_line): Likewise.
4253 (layout::print_leading_fixits): Likewise.
4254 (layout::annotation_line_showed_range_p): Likewise.
4255 (struct line_corrections): Likewise for field m_row.
4256 (line_corrections::line_corrections): Likewise for param "row".
4257 (layout::print_trailing_fixits): Likewise.
4258 (layout::get_state_at_point): Likewise.
4259 (layout::get_x_bound_for_row): Likewise.
4260 (layout::print_line): Likewise.
4261 (diagnostic_show_locus): Likewise for locals "last_line" and
4263 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4264 * input.c (selftest::test_linenum_comparisons): New function.
4265 (selftest::input_c_tests): Call it.
4266 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4267 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4268 * selftest.h (ASSERT_GT): New macro.
4269 (ASSERT_GT_AT): New macro.
4270 (ASSERT_LT): New macro.
4271 (ASSERT_LT_AT): New macro.
4273 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4275 PR rtl-optimization/84780
4276 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4278 2018-03-14 Martin Liska <mliska@suse.cz>
4280 * tree.c (record_node_allocation_statistics): Use
4281 get_stats_node_kind.
4282 (get_stats_node_kind): New function extracted from
4283 record_node_allocation_statistics.
4284 (free_node): Use get_stats_node_kind.
4286 2018-03-14 Richard Biener <rguenther@suse.de>
4288 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4289 that the value-set of ANTIC_IN doesn't grow.
4292 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4294 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4295 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4297 2018-03-14 Julia Koval <julia.koval@intel.com>
4299 * config.gcc (icelake-client, icelake-server): New.
4301 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4302 (initial_ix86_arch_features): Ditto.
4303 (PTA_SKYLAKE): Add SGX.
4304 (PTA_ICELAKE): Remove.
4305 (PTA_ICELAKE_CLIENT): New.
4306 (PTA_ICELAKE_SERVER): New.
4307 (ix86_option_override_internal): Split up icelake on icelake client and
4309 (get_builtin_code_for_version): Ditto.
4310 (fold_builtin_cpu): Ditto.
4311 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4312 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4313 * config/i386/i386.h (processor_type): Ditto.
4314 * doc/invoke.texi: Ditto.
4316 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4319 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4320 INTEGER_CST offset, add it together with bitpos / 8 and
4321 sign extend based on POINTER_SIZE.
4325 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4328 * config/i386/constraints.md (Yc): New register constraint.
4329 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4330 Use Yc constraint for alternative 2 of operand 0. Remove
4331 preferred_for_speed attribute.
4333 2018-03-14 Richard Biener <rguenther@suse.de>
4335 PR tree-optimization/84830
4336 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4337 with the old one to avoid oscillations.
4339 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4342 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4344 (assign_by_spills): Return a flag of reload assignment failure.
4345 Do not process the reload assignment failures. Do not spill other
4346 reload pseudos if they has the same reg class. Update n if
4348 (lra_assign): Add a return arg. Set up from the result of
4349 assign_by_spills call.
4350 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4351 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4352 usage_insns if it is not NULL.
4353 (spill_hard_reg_in_range): New function.
4354 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4355 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4356 function prototypes.
4357 (lra_assign): Change prototype.
4358 * lra.c (lra): Add code to deal with fails by splitting hard reg
4361 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4363 * config/riscv/riscv.opt (mrelax): New option.
4364 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4365 "norelax" when riscv_mrelax is disabled.
4366 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4368 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4371 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4372 reassociation for int modes.
4374 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4376 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4377 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4379 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4380 * config/aarch64/aarch64-sve.md
4381 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4382 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4383 (*extend<mode><Vwide>2): Rename to...
4384 (aarch64_sve_extend<mode><Vwide>2): ...this.
4385 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4386 renaming the old pattern to...
4387 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4389 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4390 define_expand, renaming the old pattern to...
4391 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4392 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4393 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4394 account when deciding which SVE instruction the optab should use.
4395 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4397 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4399 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4400 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4401 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4402 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4403 (tlsdesc_small_<mode>): Turn a define_expand and use
4404 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4405 (tlsdesc_small_advsimd_<mode>): ...this.
4406 (tlsdesc_small_sve_<mode>): New pattern.
4408 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4410 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4411 (UNSPEC_UMUL_HIGHPART): New constants.
4412 (MUL_HIGHPART): New int iteraor.
4413 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4414 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4416 (*<su>mul<mode>3_highpart): New define_insn.
4418 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4421 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4424 2018-03-13 Martin Liska <mliska@suse.cz>
4427 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4429 (sem_item_optimizer::~sem_item_optimizer): Release it.
4430 (sem_item_optimizer::merge_classes): Register variable aliases.
4431 (sem_item_optimizer::fixup_pt_set): New function.
4432 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4433 * ipa-icf.h: Declare new variables and functions.
4435 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4438 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4439 integer_pow2p@2 and test integer_pow2p in condition.
4440 (A < 0 ? C : 0): Similarly for @1.
4443 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4444 characters starting at p contain '\0' character, don't look beyond
4448 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4449 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4452 * reg-stack.c (change_stack): Change update_end var from int to
4453 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4454 also call set_block_for_insn on the newly added insns and rescan.
4457 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4458 on the last operand.
4461 * tree.c (stabilize_reference_1): Return save_expr (e) for
4462 STATEMENT_LIST even if it doesn't have side-effects.
4464 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4466 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4468 2018-03-12 Renlin Li <renlin.li@arm.com>
4470 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4471 aarch64_output_scalar_simd_mov_immediate.
4473 2018-03-12 Martin Sebor <msebor@redhat.com>
4475 PR tree-optimization/83456
4476 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4477 for perfectly overlapping calls to memcpy.
4478 (gimple_fold_builtin_memory_chk): Same.
4479 (gimple_fold_builtin_strcpy): Handle no-warning.
4480 (gimple_fold_builtin_stxcpy_chk): Same.
4481 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4483 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4485 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4486 parameter. Use it for SFmode.
4487 (rs6000_function_arg_advance_1): Adjust.
4488 (rs6000_function_arg): Adjust.
4489 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4491 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4493 PR rtl-optimization/84169
4494 PR rtl-optimization/84780
4495 * combine.c (can_combine_p): Check for a 2-insn combination whether
4496 the destination register is used between the two insns, too.
4498 2018-03-12 Richard Biener <rguenther@suse.de>
4500 PR tree-optimization/84803
4501 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4502 for refs DR analysis didn't process.
4504 2018-03-12 Richard Biener <rguenther@suse.de>
4506 PR tree-optimization/84777
4507 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4508 force-vectorize loops ignore whether we are optimizing for size.
4510 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4512 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4513 (TARGET_MD_ASM_ADJUST): Define.
4515 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4516 Kito Cheng <kito.cheng@gmail.com>
4517 Chung-Ju Wu <jasonwucj@gmail.com>
4519 * config/nds32/nds32.c (nds32_compute_stack_frame,
4520 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4521 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4522 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4523 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4524 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4525 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4526 * config/nds32/nds32.md (prologue, epilogue): Use macro
4527 NDS32_V3PUSH_AVAILABLE_P to do checking.
4529 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4532 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4533 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4534 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4535 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4536 addition of most attributes on !orig_type_die or the attribute not
4537 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4539 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4540 Chung-Ju Wu <jasonwucj@gmail.com>
4542 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4544 * config/nds32/nds32.opt (mvh): New option.
4546 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4547 Chung-Ju Wu <jasonwucj@gmail.com>
4549 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4551 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4552 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4555 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4556 Chung-Ju Wu <jasonwucj@gmail.com>
4558 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4560 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4561 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4563 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4564 Kito Cheng <kito.cheng@gmail.com>
4565 Chung-Ju Wu <jasonwucj@gmail.com>
4567 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4568 UNSPEC_FFMISM and UNSPEC_FLMISM.
4569 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4570 for ffb, ffmism and flmism.
4571 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4572 (unspec_ffmism): Ditto.
4573 (unspec_flmism): Ditto.
4574 (nds32_expand_builtin_impl): Check if string extension is available.
4575 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4576 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4578 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4581 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4584 * lra-assigns.c (assign_by_spills): Return a flag of reload
4585 assignment failure. Do not process the reload assignment
4586 failures. Do not spill other reload pseudos if they has the same
4588 (lra_assign): Add a return arg. Set up from the result of
4589 assign_by_spills call.
4590 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4591 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4592 usage_insns if it is not NULL.
4593 (spill_hard_reg_in_range): New function.
4594 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4595 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4596 function prototypes.
4597 (lra_assign): Change prototype.
4598 * lra.c (lra): Add code to deal with fails by splitting hard reg
4601 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4604 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4606 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4609 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4610 (dw_val_node): Add val_symbolic_view.
4611 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4612 (symview_upper_bound): New.
4613 (new_line_info_table): Initialize symviews_since_reset.
4614 (dwarf2out_source_line): Count symviews_since_reset and set
4615 symview_upper_bound.
4616 (dw_val_equal_p): Handle symview.
4617 (add_AT_symview): New.
4618 (print_dw_val): Handle symview.
4619 (attr_checksum, attr_checksum_ordered): Likewise.
4620 (same_dw_val_p, size_of_die): Likewise.
4621 (value_format, output_die): Likewise.
4622 (add_high_low_attributes): Use add_AT_symview for entry_view.
4623 (dwarf2out_finish): Reset symview_upper_bound, clear
4626 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4629 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4630 Add strict argument and use it.
4631 (rs6000_split_multireg_move): Update for new strict argument.
4632 (mem_operand_gpr): Disallow all non-offsettable addresses.
4633 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4635 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4638 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4639 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4640 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4643 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4644 decl, use remap_type if we want to use the type.
4646 2018-03-09 Martin Sebor <msebor@redhat.com>
4648 PR tree-optimization/84526
4649 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4651 (builtin_access::generic_overlap): Be prepared to handle non-array
4654 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4656 PR rtl-optimization/84682
4657 * lra-constraints.c (process_address_1): Check is_address flag
4658 for address constraints.
4659 (process_alt_operands): Likewise.
4660 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4661 preprocess_constraints.
4662 * recog.h (preprocess_constraints): Add oploc parameter.
4664 * recog.c (preprocess_constraints): Test address_operand for
4665 CT_ADDRESS constraints.
4667 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4670 * lra-assigns.c (assign_by_spills): Return a flag of reload
4671 assignment failure. Do not process the reload assignment
4672 failures. Do not spill other reload pseudos if they has the same
4674 (lra_assign): Add a return arg. Set up from the result of
4675 assign_by_spills call.
4676 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4677 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4678 usage_insns if it is not NULL.
4679 (spill_hard_reg_in_range): New function.
4680 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4681 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4682 function prototypes.
4683 (lra_assign): Change prototype.
4684 * lra.c (lra): Add code to deal with fails by splitting hard reg
4687 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4690 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4691 Accept complain bool parameter. Only emit errors if it is true.
4692 (arm_parse_cpu_option_name): Likewise.
4693 (arm_target_thumb_only): Adjust callers of the above.
4694 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4695 prototype to take a default true bool parameter.
4696 (arm_parse_arch_option_name): Likewise.
4698 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4699 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4703 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4704 * configure: Regenerate.
4705 * configure.ac ("linker --version-script option"): New.
4706 ("linker soname option"): New.
4708 2018-03-09 Richard Biener <rguenther@suse.de>
4710 PR tree-optimization/84775
4711 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4712 immediate uses of predicate stmts and mark them modified.
4715 PR tree-optimization/84178
4716 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4718 (version_loop_for_if_conversion): Delay update_ssa call.
4719 (tree_if_conversion): Delay update_ssa until after predicate
4722 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4725 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4726 when the function accesses prior frames.
4728 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4731 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4732 gen_llsym, otherwise call maybe_gen_llsym.
4735 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4736 has ',' character inside of it.
4738 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4741 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4742 as clobbering CC_REGNUM.
4744 2018-03-08 Richard Biener <rguenther@suse.de>
4747 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4748 (follow_copies_to_constant): Do not follow SSA names registered
4751 2018-03-08 Richard Biener <rguenther@suse.de>
4753 PR tree-optimization/84178
4754 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4756 (version_loop_for_if_conversion): Delay update_ssa call.
4757 (tree_if_conversion): Delay update_ssa until after predicate
4760 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4762 PR tree-optimization/84178
4763 * tree-if-conv.c (release_bb_predicate): Remove the
4764 the assertion that the stmts have NULL use_ops.
4765 Discard the statements, asserting that they haven't
4766 yet been added to a BB.
4768 2018-03-08 Richard Biener <rguenther@suse.de>
4770 PR tree-optimization/84746
4771 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4772 (phi_translate): Pass in destination ANTIC_OUT set.
4773 (phi_translate_1): Likewise. For a simplified result lookup
4774 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4775 (phi_translate_set): Adjust.
4776 (do_pre_regular_insertion): Likewise.
4777 (do_pre_partial_partial_insertion): Likewise.
4779 2018-03-08 Martin Liska <mliska@suse.cz>
4781 PR gcov-profile/84735
4782 * doc/gcov.texi: Document usage of profile files.
4783 * gcov-io.h: Document changes in the format.
4785 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4789 * dwarf2out.c (struct dw_line_info_table): Update comments for
4791 (FORCE_RESET_NEXT_VIEW): New.
4792 (FORCE_RESETTING_VIEW_P): New.
4793 (RESETTING_VIEW_P): Check for -1 too.
4794 (ZERO_VIEW_P): Likewise.
4795 (new_line_info_table): Force-reset next view.
4796 (dwarf2out_begin_function): Likewise.
4797 (dwarf2out_source_line): Simplify zero_view_p initialization.
4798 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4799 view directly. Omit view when omitting .loc at line 0.
4801 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4803 PR tree-optimization/84740
4804 * tree-switch-conversion.c (process_switch): Call build_constructors
4805 only if info.phi_count is non-zero.
4807 PR tree-optimization/84739
4808 * tree-tailcall.c (find_tail_calls): Check call arguments against
4809 DECL_ARGUMENTS (current_function_decl) rather than
4810 DECL_ARGUMENTS (func) when checking for tail recursion.
4812 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4814 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4815 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4816 Volker Reichelt's entry and add entries for people that perform
4817 GCC fuzzy testing and report numerous bugs.
4819 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4822 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4823 readonly data in sdata, if that is disabled.
4824 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4825 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4826 -mreadonly-in-sdata option.
4828 2018-03-07 Martin Sebor <msebor@redhat.com>
4830 PR tree-optimization/84468
4831 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4832 basic block when looking for nul assignment.
4834 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4837 * except.h (output_function_exception_table): Adjust prototype.
4838 * except.c (output_function_exception_table): Remove FNNAME parameter
4839 and add SECTION parameter. Ouput one part of the table at a time.
4840 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4841 the first part of the exception table and emit unwind directives.
4842 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4843 (i386_pe_seh_cold_init): Likewise.
4844 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4845 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4846 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4847 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4848 * config/i386/winnt.c: Include except.h.
4849 (struct seh_frame_state): Add reg_offset, after_prologue and
4850 in_cold_section fields.
4851 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4852 (i386_pe_seh_cold_init): New function.
4853 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4854 to seh->in_cold_section.
4855 (seh_emit_push): Record the offset of the push.
4856 (seh_emit_save): Record the offet of the save.
4857 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4858 Test seh->after_prologue to disregard the epilogue.
4859 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4860 (i386_pe_end_cold_function): New function.
4862 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4865 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4866 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4869 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4871 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4872 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4875 * multiple_target.c: Include tree-inline.h and intl.h.
4876 (expand_target_clones): Diagnose and fail if node->definition and
4877 !tree_versionable_function_p (node->decl).
4879 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4881 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4883 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4884 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4885 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4887 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4890 * combine.c (try_combine): Use reg_or_subregno instead of handling
4891 just paradoxical SUBREGs and REGs.
4893 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4895 * config/arc/arc.c (arc_finalize_pic): Remove function.
4896 (arc_must_save_register): We use single base PIC register, remove
4897 checks to save/restore the PIC register.
4898 (arc_expand_prologue): Likewise.
4899 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4901 (arc_verify_short): Likewise.
4902 (arc_attr_type): Likewise.
4903 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4904 (walk_stores): Likewise.
4905 (arc_address_cost): Make it static.
4906 (arc_verify_short): Likewise.
4907 (branch_dest): Likewise.
4908 (arc_attr_type): Likewise.
4909 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4910 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4911 (arc_final_prescan_insn): Remove inserting the nops due to
4912 hardware hazards. It is done in reorg step.
4913 (insn_length_variant_t): Remove.
4914 (insn_length_parameters_t): Likewise.
4915 (arc_insn_length_parameters): Likewise.
4916 (arc_get_insn_variants): Likewise.
4917 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4919 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4922 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4925 PR tree-optimization/84687
4926 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4928 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4930 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4932 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4933 Rename to ppc_speculation_barrier.
4934 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4935 __builtin_ppc_speculation_barrier.
4937 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4940 * combine.c (combine_simplify_rtx): Don't try to simplify if
4941 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4944 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4946 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4947 to 32 bytes when compiling for POWER9.
4949 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4952 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4953 regparm >= 3 with no arg reg available also for calls with
4954 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4957 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4959 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4961 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4964 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4966 2018-03-05 Richard Biener <rguenther@suse.de>
4968 PR tree-optimization/84486
4969 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4970 When inserting a __builtin_assume_aligned call set the LHS
4971 SSA name alignment info accordingly.
4973 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4975 PR tree-optimization/84114
4976 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4977 Avoid reassociation of FLOAT_MODE addition.
4979 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4981 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4982 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4983 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4984 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4985 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4986 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4987 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4989 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4990 __builtin_ia32_wbinvd): New builtins.
4991 (SPECIAL_ARGS2): New.
4992 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4993 (SPECIAL_ARGS2): New.
4994 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4995 (ix86_valid_target_attribute_inner_p): Ditto.
4996 (ix86_init_mmx_sse_builtins): Add special_args2.
4997 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4998 TARGET_WBNOINVD_P): New.
4999 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5000 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5001 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5002 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5003 * config/i386/pconfigintrin.h: New file.
5004 * config/i386/wbnoinvdintrin.h: Ditto.
5005 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5006 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5008 2018-03-05 Richard Biener <rguenther@suse.de>
5010 PR tree-optimization/84670
5011 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5013 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5014 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5015 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5016 doesn't grow if all successors (recursively) were visited at least
5019 2018-03-05 Richard Biener <rguenther@suse.de>
5021 PR tree-optimization/84650
5022 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5023 if executed in the loop pipeline.
5025 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5027 * doc/configfiles.texi (Configuration Files): Move info about
5028 conditionalizing $target-protos.h to...
5029 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5030 differs from $target-protos.h.
5032 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5033 Chung-Ju Wu <jasonwucj@gmail.com>
5035 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5036 * config/nds32/nds32-multiple.md (setmemsi): Define.
5037 * config/nds32/nds32-memory-manipulation.c
5038 (nds32_gen_dup_4_byte_to_word_value): New.
5039 (emit_setmem_word_loop): New.
5040 (emit_setmem_byte_loop): New.
5041 (nds32_expand_setmem_loop): New.
5042 (nds32_expand_setmem_loop_v3m): New.
5043 (nds32_expand_setmem_unroll): New.
5044 (nds32_expand_setmem): New.
5046 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5047 Chung-Ju Wu <jasonwucj@gmail.com>
5049 * config/nds32/nds32-memory-manipulation.c
5050 (nds32_emit_load_store): New.
5051 (nds32_emit_post_inc_load_store): New.
5052 (nds32_emit_mem_move): New.
5053 (nds32_emit_mem_move_block): New.
5054 (nds32_expand_movmemsi_loop_unknown_size): New.
5055 (nds32_expand_movmemsi_loop_known_size): New.
5056 (nds32_expand_movmemsi_loop): New.
5057 (nds32_expand_movmemsi_unroll): New.
5058 (nds32_expand_movmemqi): Rename ...
5059 (nds32_expand_movmemsi): ... to this.
5060 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5061 (movmemsi): ... to this.
5062 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5063 (nds32_expand_movmemsi): ... to this.
5065 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5066 Monk Chiang <sh.chiang04@gmail.com>
5067 Chung-Ju Wu <jasonwucj@gmail.com>
5069 * config/nds32/nds32-protos.h
5070 (nds32_expand_load_multiple): New arguments.
5071 (nds32_expand_store_multiple): Ditto.
5072 (nds32_valid_multiple_load_store): Rename ...
5073 (nds32_valid_multiple_load_store_p): ... to this.
5074 * config/nds32/nds32-memory-manipulation.c
5075 (nds32_expand_load_multiple): Refine implementation.
5076 (nds32_expand_store_multiple): Ditto.
5077 * config/nds32/nds32-multiple.md
5078 (load_multiple): Update nds32_expand_load_multiple interface.
5079 (store_multiple): Update nds32_expand_store_multiple interface.
5080 * config/nds32/nds32-predicates.c
5081 (nds32_valid_multiple_load_store): Rename ...
5082 (nds32_valid_multiple_load_store_p): ... to this and refine
5084 * config/nds32/predicates.md
5085 (nds32_load_multiple_and_update_address_operation): New predicate.
5086 (nds32_store_multiple_and_update_address_operation): New predicate.
5088 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5089 Chung-Ju Wu <jasonwucj@gmail.com>
5091 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5092 (combo): New attribute.
5093 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5095 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5097 * config/nds32/nds32.opt: Change -mcmodel= default value.
5099 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5100 Monk Chiang <sh.chiang04@gmail.com>
5101 Chung-Ju Wu <jasonwucj@gmail.com>
5103 * config/nds32/constants.md (unspec_element): New enum.
5104 * config/nds32/constraints.md (Umw): New constraint.
5105 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5106 * config/nds32/nds32-intrinsic.md: Likewise.
5107 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5108 (nds32_valid_smw_lwm_base_p): New.
5109 (nds32_output_smw_single_word): New.
5110 (nds32_output_lmw_single_word): New.
5111 (nds32_expand_unaligned_load): New.
5112 (nds32_expand_unaligned_store): New.
5113 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5114 (nds32_output_smw_single_word): Declare.
5115 (nds32_output_lmw_single_word): Declare.
5116 (nds32_expand_unaligned_load): Declare.
5117 (nds32_expand_unaligned_store): Declare.
5118 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5119 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5120 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5121 NDS32_BUILTIN_UASTORE_DW.
5122 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5125 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5126 Kito Cheng <kito.cheng@gmail.com>
5127 Chung-Ju Wu <jasonwucj@gmail.com>
5129 * config/nds32/nds32-intrinsic.c
5130 (nds32_expand_builtin_null_ftype_reg): Delete.
5131 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5132 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5133 (nds32_read_argument): New.
5134 (nds32_legitimize_target): Ditto.
5135 (nds32_legitimize_argument): Ditto.
5136 (nds32_check_constant_argument): Ditto.
5137 (nds32_expand_unop_builtin): Ditto.
5138 (nds32_expand_unopimm_builtin): Ditto.
5139 (nds32_expand_binop_builtin): Ditto.
5140 (nds32_builtin_decl_impl): Ditto.
5141 (builtin_description): Ditto.
5142 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5143 (nds32_init_builtins_impl): Ditto.
5144 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5145 (nds32_builtin_decl): New.
5146 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5147 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5149 2018-03-02 Jeff Law <law@redhat.com>
5151 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5152 (redundant_insn, fill_simple_delay_slots): Likewise.
5153 (fill_slots_from_thread): Likewise.
5154 * resource.c (mark_referenced_resources): Likewise.
5155 (mark_set_resources, find_dead_or_set_registers): Likewise.
5157 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5159 * substring-locations.h (format_warning_va): Formatting fix for
5161 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5163 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5164 * substring-locations.c: Include intl.h.
5165 (format_warning_va): Turned into small wrapper around
5166 format_warning_n_va, renamed to ...
5167 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5168 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5170 (format_warning_at_substring_n): New function.
5171 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5172 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5173 format_warning_at_substring with just a shorter name instead of
5174 const function pointer.
5175 (fmtwarn_n): New function.
5176 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5177 appropriate, get rid of all the fmtstr temporaries, move conditionals
5178 with G_() wrapped string literals directly into fmtwarn arguments,
5179 cast dir.len to (int), formatting fixes.
5181 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5183 * doc/invoke.texi: Remove "Cilk Plus" references.
5185 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5186 Richard Biener <rguenther@suse.de>
5189 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5190 for error or warning attributes if CALL_FROM_THUNK_P is set.
5193 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5196 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5197 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5200 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5201 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5203 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5204 instead of -1U in last predictors element's probability member.
5206 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5209 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5210 arguments if they are comparable.
5212 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5214 PR tree-optimization/84634
5215 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5216 masks and masked_loop_p with a single loop_masks, making sure it's
5217 null for bb vectorization.
5219 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5221 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5222 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5223 loop->force_vectorize to check whether there is no alias.
5225 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5228 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5230 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5232 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5233 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5234 instead of a loop around prev_real_insn.
5235 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5239 * config/i386/i386.c (ix86_print_operand): Use conditional
5240 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5243 2018-03-02 Richard Biener <rguenther@suse.de>
5245 PR tree-optimization/84427
5246 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5247 (bitmap_set_subtract_values): Rewrite to handle multiple
5250 (prune_clobbered_mems): Likewise.
5251 (phi_translate): Take edge instead of pred/phiblock.
5252 (phi_translate_1): Likewise.
5253 (phi_translate_set): Likewise. Insert all translated
5254 exprs for a value into the set, keeping possibly multiple
5255 expressions per value.
5256 (compute_antic_aux): Adjust for phi_translate changes.
5257 When intersecting union the expressions and prune those
5258 not in the final value set, keeping possibly multiple
5259 expressions per value. Do not use value-insertion
5260 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5261 all expressions. Add verification that the value-sets
5262 only shrink during iteration.
5263 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5264 (do_pre_regular_insertion): Likewise.
5265 (do_pre_partial_partial_insertion): Likewise.
5267 2018-03-02 Richard Biener <rguenther@suse.de>
5270 * config/darwin.c (saved_debug_info_level): New static global.
5271 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5272 (darwin_asm_lto_end): Restore debug info generation settings.
5274 2018-03-01 Martin Liska <mliska@suse.cz>
5277 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5280 2018-03-01 Richard Biener <rguenther@suse.de>
5283 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5284 type annotation with LTO when debug was disabled at compile-time.
5286 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5288 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5290 (mips_final_postscan_insn): Likewise.
5292 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5294 PR rtl-optimization/84528
5295 * alias.c (init_alias_target): Add commentary.
5296 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5297 a unique base value if the frame pointer is not eliminated
5298 to the stack pointer.
5300 2018-03-01 Tom de Vries <tom@codesourcery.com>
5302 PR rtl-optimization/83327
5303 * lra-int.h (hard_regs_spilled_into): Declare.
5304 * lra.c (hard_regs_spilled_into): Define.
5305 (init_reg_info): Init hard_regs_spilled_into.
5306 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5307 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5308 (process_bb_lives): Handle hard_regs_spilled_into.
5309 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5310 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5312 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5314 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5315 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5316 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5317 * config/rs6000/aix72.h: New file.
5319 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5321 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5322 instead of warning_at with conditional singular and plural messages
5326 * stor-layout.c (update_alignment_for_field): For
5327 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5328 && !DECL_PACKED (field), do the alignment update, just use
5329 only desired_align instead of MAX (type_align, desired_align)
5331 (place_field): Don't do known_align < desired_align handling
5332 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5333 is non-NULL, instead do it after rli->prev_field handling and
5334 only if not within a bitfield word. For DECL_PACKED (field)
5335 use type_align of BITS_PER_UNIT.
5337 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5339 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5340 superfluous parentheses and trailing spaces.
5342 2018-02-28 Richard Biener <rguenther@suse.de>
5344 PR tree-optimization/84584
5345 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5346 SCoPs with fake exit edge.
5348 2018-02-28 Martin Liska <mliska@suse.cz>
5351 * timevar.c (timer::print): Fix format to properly print 100%
5354 2018-02-28 Richard Biener <rguenther@suse.de>
5357 * genmatch.c (capture_info::walk_match): Do not mark
5358 captured expressions without operands as expr_p given
5359 they act more like predicates and should be subject to
5360 "lost tail" side-effect preserving.
5362 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5364 PR rtl-optimization/81611
5365 * auto-inc-dec.c (attempt_change): Move dead note from
5366 mem_insn if it's the next use of regno
5367 (find_address): Take address use of reg holding
5368 non-incremented value. Add parm to limit search to the named
5370 (merge_in_block): Attempt to use a mem insn that is the next
5371 use of the original regno.
5373 2018-02-27 Martin Sebor <msebor@redhat.com>
5376 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5377 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5379 2018-02-27 Martin Sebor <msebor@redhat.com>
5381 PR translation/84207
5382 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5383 n argument to unsigned HOST_WIDE_INT.
5384 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5385 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5386 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5387 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5389 2018-02-27 Richard Biener <rguenther@suse.de>
5391 PR tree-optimization/84512
5392 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5393 Do not use the estimate returned from record_stmt_cost for
5394 the scalar iteration cost but sum properly using add_stmt_cost.
5396 2018-02-27 Richard Biener <rguenther@suse.de>
5398 PR tree-optimization/84466
5399 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5400 Adjust last change to less strictly validate use operands.
5402 2018-02-27 Martin Liska <mliska@suse.cz>
5404 PR gcov-profile/84548
5405 * gcov.c (process_file): Allow partial overlap and consider it
5406 also as group functions.
5407 (output_lines): Properly calculate range of lines for a group.
5409 2018-02-27 Martin Liska <mliska@suse.cz>
5411 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5412 'ggc' suffixes. Change first column width.
5413 (timer::print): Fix formatting of the column.
5415 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5417 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5418 preserve inline entry blocks for the sake of debug inline
5419 entry point markers alone.
5420 (remove_unused_locals): Suggest in comments a better place to
5421 force the preservation of inline entry blocks that are
5422 otherwise unused, but do not preserve them.
5424 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5426 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5428 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5431 * config/i386/constraints.md (Bs): Replace
5432 ix86_indirect_branch_register with
5433 TARGET_INDIRECT_BRANCH_REGISTER.
5435 * config/i386/i386.md (indirect_jump): Likewise.
5436 (tablejump): Likewise.
5437 (*sibcall_memory): Likewise.
5438 (*sibcall_value_memory): Likewise.
5439 Peepholes of indirect call and jump via memory: Likewise.
5440 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5441 (*sibcall_value_GOT_32): Likewise.
5442 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5443 (GOT_memory_operand): Likewise.
5444 (call_insn_operand): Likewise.
5445 (sibcall_insn_operand): Likewise.
5446 (GOT32_symbol_operand): Likewise.
5447 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5449 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5451 PR rtl-optimization/83496
5452 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5453 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5454 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5455 redundant insn, if any.
5456 (relax_delay_slots): Likewise.
5457 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5459 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5461 PR tree-optimization/83965
5462 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5463 that grouped statements are part of a reduction chain. Return
5464 true if the statement is not marked as a reduction itself but
5466 (vect_recog_dot_prod_pattern): Don't check whether the statement
5467 is part of a group here.
5468 (vect_recog_sad_pattern): Likewise.
5469 (vect_recog_widen_sum_pattern): Likewise.
5471 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5474 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5476 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5479 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5481 (ix86_output_indirect_function_return): New prototype.
5482 (ix86_split_simple_return_pop_internal): Likewise.
5483 * config/i386/i386.c (indirect_return_via_cx): New.
5484 (indirect_return_via_cx_bnd): Likewise.
5485 (indirect_thunk_name): Handle return va CX_REG.
5486 (output_indirect_thunk_function): Create alias for
5487 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5488 (ix86_output_indirect_jmp): Remove the bool argument.
5489 (ix86_output_indirect_function_return): New function.
5490 (ix86_split_simple_return_pop_internal): Likewise.
5491 * config/i386/i386.md (*indirect_jump): Don't pass false
5492 to ix86_output_indirect_jmp.
5493 (*tablejump_1): Likewise.
5494 (simple_return_pop_internal): Change it to define_insn_and_split.
5495 Call ix86_split_simple_return_pop_internal to split it for
5497 (simple_return_indirect_internal): Call
5498 ix86_output_indirect_function_return instead of
5499 ix86_output_indirect_jmp.
5501 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5504 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5505 memset and value initialization afterwards.
5507 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5509 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5511 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5514 * common/config/aarch64/aarch64-common.c
5515 (aarch_option_optimization_table[]): Switch
5516 off fomit-frame-pointer
5518 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5519 Chung-Ju Wu <jasonwucj@gmail.com>
5521 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5523 (store_multiple): Ditto.
5525 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5527 * config.gcc: Add --with-cpu support for nds32 target.
5528 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5529 * config/nds32/nds32.opt: Add -mcpu= option.
5531 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5533 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5534 isel=yes): Warn for these deprecated options.
5536 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5538 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5539 ISA_2_5_MASKS_EMBEDDED.
5541 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5543 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5544 p->max as pointers rather than using iterative_hash_expr.
5546 2018-02-23 Carl Love <cel@us.ibm.com>
5548 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5549 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5551 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5552 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5553 P8V_BUILTIN_VEC_VUNSIGNED2.
5555 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5558 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5559 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5561 * lra-constraints.c (curr_insn_transform): Set up
5562 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5564 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5565 * lra-eliminations.c (spill_pseudos): Ditto.
5566 (process_insn_for_elimination): Ditto.
5567 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5568 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5569 LRA_NON_CLOBBERED_ALT.
5571 2018-02-22 Martin Sebor <msebor@redhat.com>
5573 PR tree-optimization/84480
5574 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5575 to maybe_diag_stxncpy_trunc. Call it.
5576 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5577 from gimple_fold_builtin_strcpy. Print inlining stack.
5578 (handle_builtin_stxncpy): Print inlining stack.
5579 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5581 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5584 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5585 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5586 and -fcheck-pointer-bounds are used together.
5587 (indirect_thunk_prefix): New enum.
5588 (indirect_thunk_need_prefix): New function.
5589 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5590 "_nt" instead of "_bnd" for NOTRACK prefix.
5591 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5592 (output_indirect_thunk_function): Likewise.
5594 (ix86_code_end): Update output_indirect_thunk_function calls.
5595 (ix86_output_indirect_branch_via_reg): Replace
5596 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5597 (ix86_output_indirect_branch_via_push): Likewise.
5598 (ix86_output_function_return): Likewise.
5599 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5600 incompatible with -fcf-protection=branch and
5601 -fcheck-pointer-bounds.
5603 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5606 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5607 Change gcc_assert call to output_operand_lossage.
5609 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5611 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5613 2018-02-22 DJ Delorie <dj@redhat.com>
5614 Sebastian Perta <sebastian.perta@renesas.com>
5615 Oleg Endo <olegendo@gcc.gnu.org>
5617 * config/rx/rx.c (rx_rtx_costs): New function.
5618 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5620 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5622 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5624 2018-02-22 Martin Liska <mliska@suse.cz>
5627 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5628 Add "native" as a possible value.
5630 2018-02-22 Martin Liska <mliska@suse.cz>
5633 * config/i386/i386.c (ix86_option_override_internal):
5634 Add "native" as a possible value for -march and -mtune.
5636 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5639 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5640 to all type variants.
5642 PR tree-optimization/84503
5643 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5644 width as info->bitpos + info->bitsize - start.
5645 (merged_store_group::merge_overlapping): Simplify width computation.
5646 (check_no_overlap): New function.
5647 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5648 start + width and last_order of the group, fail if check_no_overlap
5650 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5651 to group if check_no_overlap fails.
5653 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5655 * config/rs6000/altivec.md: Delete contraint arguments to
5656 define_expand, define_split, and define_peephole2, and in
5657 define_insn_and_split if always unused.
5658 * config/rs6000/darwin.md: Ditto.
5659 * config/rs6000/dfp.md: Ditto.
5660 * config/rs6000/rs6000.md: Ditto.
5661 * config/rs6000/sync.md: Ditto.
5662 * config/rs6000/vector.md: Ditto.
5663 * config/rs6000/vsx.md: Ditto.
5665 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5667 * config/rs6000/altivec.md: Write output control strings as braced
5668 blocks instead of double-quoted strings.
5669 * config/rs6000/darwin.md: Ditto.
5670 * config/rs6000/rs6000.md: Ditto.
5671 * config/rs6000/vector.md: Ditto.
5672 * config/rs6000/vsx.md: Ditto.
5674 2018-02-21 Jason Merrill <jason@redhat.com>
5676 PR c++/84314 - ICE with templates and fastcall attribute.
5677 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5679 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5681 * ipa-cp.c (determine_versionability): Fix comment typos.
5683 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5686 * ipa-cp.c (determine_versionability): Do not version functions caling
5689 2018-02-21 Martin Liska <mliska@suse.cz>
5692 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5693 Add "native" as a possible value.
5694 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5695 the macro when native cpu detection is available.
5697 2018-02-21 Martin Liska <mliska@suse.cz>
5700 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5701 Add "native" as a possible value.
5702 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5703 when native cpu detection is available.
5705 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5706 Martin Sebor <msebor@redhat.com>
5708 PR tree-optimization/84478
5709 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5711 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5712 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5713 support which is conservatively correct, for 2 only stay conservative
5714 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5715 argument to the 2 argument get_range_strlen, adjust 6 arg
5716 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5718 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5719 (gimple_fold_builtin_strlen): Pass true as last argument to
5722 2018-02-20 Martin Sebor <msebor@redhat.com>
5725 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5726 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5727 (builtin_memref::builtin_memref): Factor out parts into
5728 set_base_and_offset and call it.
5730 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5733 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5734 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5735 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5736 search at the associated MODE_INT.
5738 2018-02-20 Jeff Law <law@redhat.com>
5741 PR tree-optimization/81592
5743 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5744 for range data rather than using global data.
5745 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5746 range data rather than using global data.
5747 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5748 pass it to children as needed.
5749 (struct directive::fmtresult): Similarly.
5750 (struct directive::set_width): Similarly.
5751 (struct directive::set_precision): Similarly.
5752 (format_integer, format_directive, parse_directive): Similarly.
5753 (format_none): Accept unnamed vr_values parameter.
5754 (format_percent, format_floating, format_character): Similarly.
5755 (format_string, format_plain): Similarly.
5756 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5757 the EVRP range analyzer for range data rather than using global data.
5758 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5759 gimple-ssa-evrp-analyze.h
5760 (class sprintf_dom_walker): Add after_dom_children member function.
5761 Add evrp_range_analyzer member.
5762 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5763 range analyzer as needed.
5764 (sprintf_dom_walker::after_dom_children): New member function.
5765 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5767 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5768 (evrp_range_analyzer::pop_to_marker): Likewise.
5770 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5772 PR tree-optimization/84419
5773 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5774 with the required type if its current type is compatible but
5777 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5780 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5781 after vectorization.
5783 2018-02-20 Martin Liska <mliska@suse.cz>
5786 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5787 possible values if we don't have a hint.
5789 2018-02-20 Martin Liska <mliska@suse.cz>
5793 * final.c (shorten_branches): Build align_tab array with one
5795 * opts.c (finish_options): Add alignment option limit check.
5796 (MAX_CODE_ALIGN): Likewise.
5797 (MAX_CODE_ALIGN_VALUE): Likewise.
5798 * doc/invoke.texi: Document maximum allowed option value for
5799 all -falign-* options.
5801 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5804 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5805 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5806 * var-tracking.c (emit_note_insn_var_location): Remove all references
5807 to NOTE_INSN_CALL_ARG_LOCATION.
5808 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5809 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5810 Use copy_rtx_if_shared.
5811 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5812 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5813 (dwarf2out_var_location): Remove handling of
5814 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5816 * final.c (final_scan_insn): Remove all references to
5817 NOTE_INSN_CALL_ARG_LOCATION.
5818 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5819 before dumping final insns.
5820 * except.c (emit_note_eh_region_end): Remove all references to
5821 NOTE_INSN_CALL_ARG_LOCATION.
5822 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5823 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5824 * config/arc/arc.c (hwloop_optimize): Likewise.
5825 * config/arm/arm.c (create_fix_barrier): Likewise.
5826 * config/s390/s390.c (s390_chunkify_start): Likewise.
5827 * config/sh/sh.c (find_barrier): Likewise.
5828 * config/i386/i386.c (rest_of_insert_endbranch,
5829 ix86_seh_fixup_eh_fallthru): Likewise.
5830 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5831 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5832 * config/frv/frv.c (frv_function_prologue): Likewise.
5833 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5835 (note_outside_basic_block_p): Remove all references to
5836 NOTE_INSN_CALL_ARG_LOCATION.
5837 * gengtype.c (adjust_field_rtx_def): Likewise.
5838 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5840 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5841 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5844 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5847 PR tree-optimization/84452
5848 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5849 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5852 2018-02-19 Martin Liska <mliska@suse.cz>
5855 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5857 2018-02-19 Martin Liska <mliska@suse.cz>
5858 Richard Sandiford <richard.sandiford@linaro.org>
5860 PR tree-optimization/82491
5861 * gimple-fold.c (get_base_constructor): Make earlier bail out
5864 2018-02-19 Carl Love <cel@us.ibm.com>
5866 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5867 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5869 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5870 P8V_BUILTIN_VEC_NEG.
5872 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5874 * config/rl78/rl78.md (movdf): New define expand.
5876 2018-02-19 Martin Liska <mliska@suse.cz>
5879 * doc/invoke.texi: Fix typo.
5880 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5882 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5884 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5885 handle rs6000_single_float and rs6000_double_float specially for
5888 2018-02-16 Jeff Law <law@redhat.com>
5890 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5893 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5896 * ipa-inline.c (inline_small_functions): Fix a typo.
5898 2018-02-16 Nathan Sidwell <nathan@acm.org>
5900 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5902 2018-02-16 Carl Love <cel@us.ibm.com>
5904 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5905 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5906 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5907 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5908 expansion to P8V_BUILTIN_VEC_FLOAT2.
5910 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5912 PR rtl-optimization/70023
5913 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5914 src_regno into account.
5916 2018-02-16 Carl Love <cel@us.ibm.com>
5918 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5919 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5920 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5921 * config/rs6000/rs6000.c: Remove case statements for
5922 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5923 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5924 and P9V_BUILTIN_VEC_VINSERT4B.
5925 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5926 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5927 * config/rs6000/vsx.md:
5928 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5931 2018-02-16 Carl Love <cel@us.ibm.com>
5933 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5935 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5937 * config/rs6000/rs6000-c.c: Add the definitions for
5938 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5939 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5940 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5941 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5942 definition for insert4b and define insn *insert3b_internal.
5943 * doc/extend.texi: Add documentation for vec_extract4b.
5945 2018-02-16 Nathan Sidwell <nathan@acm.org>
5947 * doc/extend.texi (Backwards Compatibility): Mention friend
5948 injection. Note for-scope is deprecated.
5949 * doc/invoke.texi (-ffriend-injection): Deprecate.
5951 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5953 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5954 that moved to I2, also allow destinations that are a paradoxical
5955 subreg (instead of a normal reg).
5957 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5960 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5963 2018-02-16 Richard Biener <rguenther@suse.de>
5965 PR tree-optimization/84037
5966 PR tree-optimization/84016
5968 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5969 Adjust vec_construct for the fact we need additional higher latency
5970 128bit inserts for AVX256 and AVX512 vector builds.
5971 (ix86_add_stmt_cost): Scale vector construction cost for
5974 2018-02-16 Richard Biener <rguenther@suse.de>
5976 PR tree-optimization/84417
5977 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5978 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5979 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5981 2018-02-16 Martin Liska <mliska@suse.cz>
5984 * internal-fn.def (ASAN_CHECK): Set proper flags.
5985 (ASAN_MARK): Likewise.
5987 2018-02-16 Julia Koval <julia.koval@intel.com>
5989 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5990 from PTA_CANNONLAKE.
5992 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5995 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5996 Use ++iter rather than iter++ for std::list iterators.
5997 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5998 defer deleting them until all nodes in the forest are processed. Do
5999 free even leaf nodes. Change to_process into auto_vec.
6002 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6003 * vec.h (vec_default_construct): Use memset instead of placement new
6004 if BROKEN_VALUE_INITIALIZATION is defined.
6005 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6006 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6009 PR rtl-optimization/83723
6010 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6011 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6012 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6014 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6016 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6018 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6020 PR rtl-optimization/81443
6021 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6022 from inner REGs to paradoxical SUBREGs.
6024 2018-02-16 Richard Biener <rguenther@suse.de>
6026 PR tree-optimization/84399
6027 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6028 For operands we can analyze at their definition make sure we can
6029 analyze them at each use as well.
6031 2018-02-16 Richard Biener <rguenther@suse.de>
6033 PR tree-optimization/84190
6034 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6035 volatile accesses if the decl isn't volatile.
6037 2018-02-15 Jason Merrill <jason@redhat.com>
6039 PR c++/84314 - ICE with templates and fastcall attribute.
6040 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6041 TYPE_CANONICAL on an existing type.
6043 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6045 PR tree-optimization/84383
6046 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6047 dstoff nor call operand_equal_p if dstbase is NULL.
6049 PR tree-optimization/84334
6050 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6051 also a CONSTANT_CLASS_P, punt.
6053 2018-02-14 Jim Wilson <jimw@sifive.com>
6055 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6056 first SMALL_OPERAND check. New local min_second_step. Move assert
6057 to where locals are set. Add TARGET_RVC support.
6058 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6060 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6062 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6064 2018-02-14 Martin Sebor <msebor@redhat.com>
6066 PR tree-optimization/83698
6067 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6068 arrays constrain the offset range to their bounds.
6069 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6070 (builtin_access::overlap): Avoid setting the size of overlap if it's
6072 (maybe_diag_overlap): Also consider arrays when deciding what values
6073 of offsets to include in diagnostics.
6075 2018-02-14 Martin Sebor <msebor@redhat.com>
6078 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6079 that correspond to the kind of a declaration.
6081 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6084 * config/pa/pa.md: Load address of PIC label using the linkage table
6085 if the label is nonlocal.
6087 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6089 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6090 warning message if user requests -maltivec=be.
6091 * doc/invoke.texi: Document deprecation of -maltivec=be.
6093 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6096 * config/rs6000/rs6000-c.c: Update definitions for
6097 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6098 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6100 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6103 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6104 add _get_ssp intrinsics. Remove argument from
6105 __builtin_ia32_rdssp[d|q].
6106 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6107 * config/i386/i386-builtin.def: Remove argument from
6108 __builtin_ia32_rdssp[d|q].
6109 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6110 ix86_expand_special_args_builtin for _rdssp[d|q].
6111 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6112 Clear register before usage.
6113 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6114 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6116 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6118 PR tree-optimization/84357
6119 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6120 operand 1 of an ARRAY_REF too.
6122 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6125 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6126 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6128 (set_of_reg): New struct.
6129 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6130 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6131 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6133 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6134 Split into bitclr, bitset, bitinvert patterns if appropriate.
6135 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6136 use rx_fuse_in_memory_bitop.
6137 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6138 to named insn, correct maximum insn length.
6140 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6143 * machmode.def: Define a complex mode for PARTIAL_INT.
6144 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6146 * doc/rtl.texi: Document CSPImode.
6147 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6149 (msp430_hard_regno_nregs_with_padding): Likewise.
6151 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6154 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6156 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6158 PR rtl-optimization/84169
6159 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6160 we generated a parallel as new i3 and we split that to new i2 and i3
6161 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6162 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6163 those to i2, not i1. Partially rewrite this scan code.
6165 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6168 * stor-layout.c (place_field): For variable length fields, adjust
6169 offset_align afterwards not just based on the field's alignment,
6170 but also on the size.
6173 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6174 of exps and logs in the use_exp2 case.
6176 2018-02-13 Jeff Law <law@redhat.com>
6178 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6181 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6184 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6188 * common.opt (gas-loc-support, gas-locview-support): New.
6189 (ginline-points, ginternal-reset-location-views): New.
6190 * doc/invoke.texi: Document them. Use @itemx where intended.
6191 (gvariable-location-views): Adjust.
6192 * target.def (reset_location_view): New.
6193 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6194 (TARGET_RESET_LOCATION_VIEW): New.
6195 * doc/tm.texi: Rebuilt.
6196 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6197 (dwarf2out_default_as_locview_support): New.
6198 (output_asm_line_debug_info): Use option variables.
6199 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6200 (output_loc_list): Likewise.
6201 (add_high_low_attributes): Check option variables.
6202 Don't output entry view attribute in strict mode.
6203 (gen_inlined_subroutine_die): Check option variables.
6204 (dwarf2out_inline_entry): Likewise.
6205 (init_sections_and_labels): Likewise.
6206 (dwarf2out_early_finish): Likewise.
6207 (maybe_reset_location_view): New, from...
6208 (dwarf2out_var_location): ... here. Call it.
6209 * debug.h (dwarf2out_default_as_loc_support): Declare.
6210 (dwarf2out_default_as_locview_support): Declare.
6211 * hooks.c (hook_int_rtx_insn_0): New.
6212 * hooks.h (hook_int_rtx_insn_0): Declare.
6213 * toplev.c (process_options): Take -gas-loc-support and
6214 -gas-locview-support from dwarf2out. Enable
6215 -gvariable-location-views by default only with locview
6216 assembler support. Enable -ginternal-reset-location-views by
6217 default only if the target defines the corresponding hook.
6218 Enable -ginline-points by default if location views are
6219 enabled; force it disabled if statement frontiers are
6221 * tree-inline.c (expand_call_inline): Check option variables.
6222 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6224 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6226 PR tree-optimization/84321
6227 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6228 handling. Also check whether the anti-range contains any values
6229 that satisfy the mask; switch to a VR_RANGE if not.
6231 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6234 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6236 2018-02-13 Martin Jambor <mjambor@suse.cz>
6239 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6240 of call statements, also set location of a load to a temporary.
6242 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6244 * config/rl78/rl78.c (add_vector_labels): New function.
6245 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6246 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6247 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6248 which checks that no arguments are passed.
6249 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6250 * doc/extend.texi: Documentation for the new attribute.
6252 2018-02-13 Andreas Schwab <schwab@suse.de>
6254 * config/riscv/linux.h (CPP_SPEC): Define.
6256 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6259 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6260 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6261 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6262 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6263 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6264 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6265 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6266 temporarily for AES and PCLMUL builtins.
6268 PR tree-optimization/84339
6269 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6270 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6274 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6275 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6276 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6278 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6280 * omp-simd-clone.h: New file.
6281 * omp-simd-clone.c: Include omp-simd-clone.h.
6282 (expand_simd_clones): No longer static.
6283 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6284 cgraph.h and omp-simd-clone.h.
6285 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6286 (vect_recog_widen_shift_pattern): Formatting fix.
6287 (vect_pattern_recog_1): Don't check optab for calls.
6290 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6291 operands[2] into a REG before using gen_lowpart on it.
6293 2018-02-12 Jeff Law <law@redhat.com>
6296 * config/sh/sh.c (find_barrier): Consider a sibling call
6299 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6300 successfully back substituting a reg.
6302 2018-02-12 Richard Biener <rguenther@suse.de>
6304 PR tree-optimization/84037
6305 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6306 parameter, move visited init to caller.
6307 (vect_slp_analyze_operations): Separate cost from validity
6308 check, initialize visited once for all instances.
6309 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6311 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6313 (vect_model_promotion_demotion_cost): Likewise.
6314 (vectorizable_bswap): Guard cost modeling with !slp_node
6315 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6317 (vectorizable_call): Likewise.
6318 (vectorizable_conversion): Likewise.
6319 (vectorizable_assignment): Likewise.
6320 (vectorizable_shift): Likewise.
6321 (vectorizable_operation): Likewise.
6322 (vectorizable_store): Likewise.
6323 (vectorizable_load): Likewise.
6324 (vectorizable_condition): Likewise.
6325 (vectorizable_comparison): Likewise.
6327 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6330 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6331 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6332 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6334 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6337 * params.def (inline-min-speedup): Increase from 8 to 15.
6338 (max-inline-insns-auto): Decrease from 40 to 30.
6339 * ipa-split.c (consider_split): Add some buffer for function to
6340 be considered inlining candidate.
6341 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6344 2018-02-12 Richard Biener <rguenther@suse.de>
6346 PR tree-optimization/84037
6347 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6348 matched stmts if we cannot swap the non-matched ones.
6350 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6352 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6353 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6354 _mm_maskz_scalef_round_ss): New intrinsics.
6355 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6356 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6357 __builtin_ia32_scalefss_round): Remove.
6358 (__builtin_ia32_scalefsd_mask_round,
6359 __builtin_ia32_scalefss_mask_round): New intrinsics.
6360 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6361 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6362 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6363 "<round_constraint>")): Changed to ...
6364 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6365 "<round_scalar_constraint>")): ... this.
6366 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6367 %0, %1, %2<round_op3>}"): Changed to ...
6368 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6369 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6370 %2<round_scalar_mask_op3>}"): ... this.
6371 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6373 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6375 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6376 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6377 (_mm_maskz_sqrt_round_ss): New intrinsics.
6378 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6379 (__builtin_ia32_sqrtsd_mask_round)
6380 (__builtin_ia32_sqrtss_mask_round): New builtins.
6381 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6382 (__builtin_ia32_sqrtss_round): Remove.
6383 (__builtin_ia32_sqrtsd_mask_round)
6384 (__builtin_ia32_sqrtss_mask_round): New builtins.
6385 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6386 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6387 ((match_operand:VF_128 1 "vector_operand"
6388 "xBm,<round_constraint>")): Changed to ...
6389 ((match_operand:VF_128 1 "vector_operand"
6390 "xBm,<round_scalar_constraint>")): ... this.
6391 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6392 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6393 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6394 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6395 %<iptr>1<round_scalar_mask_op3>}): ... this.
6396 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6397 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6399 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6402 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6403 Cast vec_cmpeq result to correct type.
6404 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6405 Cast vec_cmpgt result to correct type.
6407 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6409 * final.c (final_scan_insn_1): Renamed from...
6410 (final_scan_insn): ... this. New wrapper, to recover
6411 seen from the outermost call in recursive ones.
6412 * config/sparc/sparc.c (output_return): Drop seen from call.
6413 (output_sibcall): Likewise.
6414 * config/visium/visium.c (output_branch): Likewise.
6416 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6418 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6421 2018-02-10 Alan Modra <amodra@gmail.com>
6424 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6425 Specify LR as an input.
6427 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6430 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6431 remove_member_access_dummy_vars): New functions.
6432 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6433 lower_omp_1, execute_lower_omp): Use them.
6435 PR rtl-optimization/84308
6436 * shrink-wrap.c (spread_components): Release todo vector.
6438 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6440 PR rtl-optimization/57193
6441 * ira-color.c (struct allocno_color_data): Add member
6442 conflict_allocno_hard_prefs.
6443 (update_conflict_allocno_hard_prefs): New.
6444 (bucket_allocno_compare_func): Add a preference based on
6445 conflict_allocno_hard_prefs.
6446 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6447 (color_allocnos): Remove a dead code. Initiate
6448 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6450 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6453 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6454 constraint from =wa to wa. Avoid a subreg on the output operand,
6455 instead use a pseudo and subreg it in a move.
6456 (p9_xxbrd_<mode>): Changed to ...
6457 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6458 (p9_xxbrd_v2df): New expander.
6459 (p9_xxbrw_<mode>): Changed to ...
6460 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6461 (p9_xxbrw_v4sf): New expander.
6463 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6465 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6467 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6470 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6471 multiply in 32-bit mode.
6472 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6473 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6476 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6478 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6479 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6480 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6481 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6483 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6486 * dwarf2out.c (is_trivial_indirect_ref): New function.
6487 (dwarf2out_late_global_decl): Do not generate a location
6488 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6489 and that are not defined in the current unit.
6491 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6493 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6494 instead of a libcall for UNORDERED.
6496 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6499 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6500 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6502 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6505 * config/s390/s390.c (s390_set_current_function): Invoke
6506 s390_indirect_branch_settings also if fndecl didn't change.
6508 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6510 * config/rs6000/rs6000.md (blockage): Set length to zero.
6512 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6514 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6516 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6519 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6520 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6524 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6525 PARALLEL incoming that failed vt_get_decl_and_offset check.
6528 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6529 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6531 (get_variable_section): For decls in named .bss* sections pass true as
6532 second argument to bss_initializer_p.
6534 2018-02-09 Marek Polacek <polacek@redhat.com>
6535 Jakub Jelinek <jakub@redhat.com>
6538 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6539 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6540 Sync some changes from cxx_fold_indirect_ref.
6542 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6544 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6546 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6547 (BLOCK_INLINE_ENTRY_LABEL): New.
6548 (dwarf2out_var_location): Disregard inline entry markers.
6549 (inline_entry_data): New struct.
6550 (inline_entry_data_hasher): New hashtable type.
6551 (inline_entry_data_hasher::hash): New.
6552 (inline_entry_data_hasher::equal): New.
6553 (inline_entry_data_table): New variable.
6554 (add_high_low_attributes): Add DW_AT_entry_pc and
6555 DW_AT_GNU_entry_view attributes if a pending entry is found
6556 in inline_entry_data_table. Add old entry_pc attribute only
6557 if debug nonbinding markers are disabled.
6558 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6559 markers are enabled.
6560 (block_within_block_p, dwarf2out_inline_entry): New.
6561 (dwarf2out_finish): Check that no entries remained in
6562 inline_entry_data_table.
6563 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6564 (final_scan_insn, notice_source_line): Likewise.
6565 (rest_of_clean_state): Skip inline entry markers.
6566 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6568 * gimple.c (gimple_build_debug_inline_entry): New.
6569 * gimple.h (enum gimple_debug_subcode): Add
6570 GIMPLE_DEBUG_INLINE_ENTRY.
6571 (gimple_build_debug_inline_entry): Declare.
6572 (gimple_debug_inline_entry_p): New.
6573 (gimple_debug_nonbind_marker_p): Adjust.
6574 * insn-notes.def (INLINE_ENTRY): New.
6575 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6576 inline entry marker notes.
6577 (print_insn): Likewise.
6578 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6579 (INSN_DEBUG_MARKER_KIND): Likewise.
6580 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6581 * tree-inline.c (expand_call_inline): Build and insert
6582 debug_inline_entry stmt.
6583 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6584 inline entry blocks early, if nonbind markers are enabled.
6585 (dump_scope_block): Dump fragment info.
6586 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6587 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6588 (gimple_build_debug_inline_entry): New.
6589 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6590 Enable/disable inline entry points too.
6591 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6592 (DEBUG_INSN): Describe inline entry markers.
6594 * common.opt (gvariable-location-views): New.
6595 (gvariable-location-views=incompat5): New.
6596 * config.in: Rebuilt.
6597 * configure: Rebuilt.
6598 * configure.ac: Test assembler for view support.
6599 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6600 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6601 * dwarf2out.c (var_loc_view): New typedef.
6602 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6603 (dwarf2out_locviews_in_attribute): New.
6604 (dwarf2out_locviews_in_loclist): New.
6605 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6606 (enum dw_line_info_opcode): Add LI_adv_address.
6607 (struct dw_line_info_table): Add view.
6608 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6609 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6610 (zero_view_p): New variable.
6611 (ZERO_VIEW_P): New macro.
6612 (output_asm_line_debug_info): New.
6613 (struct var_loc_node): Add view.
6614 (add_AT_view_list, AT_loc_list): New.
6615 (add_var_loc_to_decl): Add view param. Test it against last.
6616 (new_loc_list): Add view params. Record them.
6617 (AT_loc_list_ptr): Handle loc and view lists.
6618 (view_list_to_loc_list_val_node): New.
6619 (print_dw_val): Handle dw_val_class_view_list.
6620 (size_of_die): Likewise.
6621 (value_format): Likewise.
6622 (loc_list_has_views): New.
6623 (gen_llsym): Set vl_symbol too.
6624 (maybe_gen_llsym, skip_loc_list_entry): New.
6625 (dwarf2out_maybe_output_loclist_view_pair): New.
6626 (output_loc_list): Output view list or entries too.
6627 (output_view_list_offset): New.
6628 (output_die): Handle dw_val_class_view_list.
6629 (output_dwarf_version): New.
6630 (output_compilation_unit_header): Use it.
6631 (output_skeleton_debug_sections): Likewise.
6632 (output_rnglists, output_line_info): Likewise.
6633 (output_pubnames, output_aranges): Update version comments.
6634 (output_one_line_info_table): Output view numbers in asm comments.
6635 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6636 Call maybe_gen_llsym.
6637 (loc_list_from_tree_1): Adjust.
6638 (add_AT_location_description): Create view list attribute if
6639 needed, check it's absent otherwise.
6640 (convert_cfa_to_fb_loc_list): Adjust.
6641 (maybe_emit_file): Call output_asm_line_debug_info for test.
6642 (dwarf2out_var_location): Reset views as needed. Precompute
6643 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6644 attribute. Set view.
6645 (new_line_info_table): Reset next view.
6646 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6647 (dwarf2out_source_line): Likewise. Output view resets and labels to
6648 the assembler, or select appropriate line info opcodes.
6649 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6650 (optimize_string_length): Catch it. Adjust.
6651 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6652 dw_val_class_view_list, and remove it if no longer needed.
6653 (hash_loc_list): Hash view numbers.
6654 (loc_list_hasher::equal): Compare them.
6655 (optimize_location_lists): Check whether a view list symbol is
6656 needed, and whether the locview attribute is present, and
6657 whether they match. Remove the locview attribute if no longer
6659 (index_location_lists): Call skip_loc_list_entry for test.
6660 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6661 Use output_dwarf_version.
6662 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6663 (struct dw_val_node): Add val_view_list.
6664 * final.c (SEEN_NEXT_VIEW): New.
6665 (set_next_view_needed): New.
6666 (clear_next_view_needed): New.
6667 (maybe_output_next_view): New.
6668 (final_start_function): Rename to...
6669 (final_start_function_1): ... this. Take pointer to FIRST,
6670 add SEEN parameter. Emit param bindings in the initial view.
6671 (final_start_function): Reintroduce SEEN-less interface.
6672 (final): Rename to...
6673 (final_1): ... this. Take SEEN parameter. Output final pending
6674 next view at the end.
6675 (final): Reintroduce seen-less interface.
6676 (final_scan_insn): Output pending next view before switching
6677 sections or ending a block. Mark the next view as needed when
6678 outputting variable locations. Notify debug backend of section
6679 changes, and of location view changes.
6680 (rest_of_handle_final): Adjust.
6681 * toplev.c (process_options): Autodetect value for debug variable
6682 location views option. Warn on incompat5 without -gdwarf-5.
6683 * doc/invoke.texi (gvariable-location-views): New.
6684 (gvariable-location-views=incompat5): New.
6685 (gno-variable-location-views): New.
6687 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6689 PR tree-optimization/84136
6690 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6691 that the result of find_edge is non-NULL.
6693 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6696 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6697 storing integer register in SImode. Fix cost of 256 and 512
6698 byte aligned SSE register store.
6700 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6702 * config/i386/i386.c (ix86_multiplication_cost): Fix
6703 multiplication cost for TARGET_AVX512DQ.
6705 2018-02-08 Marek Polacek <polacek@redhat.com>
6707 PR tree-optimization/84238
6708 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6711 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6713 PR tree-optimization/84265
6714 * tree-vect-stmts.c (vectorizable_store): Don't treat
6715 VMAT_CONTIGUOUS accesses as grouped.
6716 (vectorizable_load): Likewise.
6718 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6720 PR tree-optimization/81635
6721 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6722 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6723 (test_round_for_mask): New functions.
6724 (wide_int_cc_tests): Call test_round_for_mask.
6725 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6726 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6727 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6728 range returned by get_range_info.
6730 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6733 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6734 * symtab.c: Include builtins.h
6735 (symtab_node::output_to_lto_symbol_table_p): Move here
6736 from lto-streamer-out.c:output_symbol_p.
6737 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6738 (output_symbol_p): Move all logic to symtab.c
6739 (produce_symtab): Update.
6741 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6743 * config/s390/s390-opts.h (enum indirect_branch): Define.
6744 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6745 (s390_indirect_branch_via_thunk)
6746 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6747 (enum s390_indirect_branch_type): Define.
6748 * config/s390/s390.c (struct s390_frame_layout, struct
6749 machine_function): Remove.
6750 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6751 (indirect_branch_table_label_no, indirect_branch_table_name):
6753 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6754 (enum s390_indirect_branch_option): Define.
6755 (s390_return_addr_from_memory): New function.
6756 (s390_handle_string_attribute): New function.
6757 (s390_attribute_table): Add new attribute handler.
6758 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6759 (s390_indirect_branch_via_thunk): New function.
6760 (s390_indirect_branch_via_inline_thunk): New function.
6761 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6762 sibling call optimization for non z10 compiles.
6763 (s390_emit_call): Force indirect branch target to be a single
6764 register. Add r1 clobber for non-z10 compiles.
6765 (s390_emit_epilogue): Emit return jump via return_use expander.
6766 (s390_reorg): Handle JUMP_INSNs as execute targets.
6767 (s390_option_override_internal): Perform validity checks for the
6768 new command line options.
6769 (s390_indirect_branch_attrvalue): New function.
6770 (s390_indirect_branch_settings): New function.
6771 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6772 (s390_output_indirect_thunk_function): New function.
6773 (s390_code_end): Implement target hook.
6774 (s390_case_values_threshold): Implement target hook.
6775 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6777 * config/s390/s390.h (struct s390_frame_layout)
6778 (struct machine_function): Move here from s390.c.
6779 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6780 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6781 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6782 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6783 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6784 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6785 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6786 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6787 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6788 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6789 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6790 (mnemonic attribute): Add values which aren't recognized
6792 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6793 pattern for branch conversion. Fix mnemonic attribute.
6794 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6795 indirect branch via thunk if requested.
6796 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6797 ("*indirect_jump"): Disable for branch conversion using out of
6799 ("indirect_jump_via_thunk<mode>_z10")
6800 ("indirect_jump_via_thunk<mode>")
6801 ("indirect_jump_via_inlinethunk<mode>_z10")
6802 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6803 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6804 ("casesi_jump_via_inlinethunk<mode>_z10")
6805 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6806 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6807 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6808 ("*indirect2_jump"): Disable for branch conversion.
6809 ("casesi_jump"): Turn into expander and expand patterns for branch
6811 ("return_use"): New expander.
6812 ("*return"): Emit return via thunk and rename it to ...
6813 ("*return<mode>"): ... this one.
6814 * config/s390/s390.opt: Add new options and and enum for the
6817 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6819 * lra-constraints.c (match_reload): Unconditionally use
6820 gen_lowpart_SUBREG, rather than selecting between that
6821 and equivalent gen_rtx_SUBREG code.
6823 2018-02-08 Richard Biener <rguenther@suse.de>
6825 PR tree-optimization/84233
6826 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6827 changed flag instead of boguously re-using phi_inserted.
6829 2018-02-08 Martin Jambor <mjambor@suse.cz>
6831 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6832 static local variables.
6834 2018-02-08 Richard Biener <rguenther@suse.de>
6836 PR tree-optimization/84278
6837 * tree-vect-stmts.c (vectorizable_store): When looking for
6838 smaller vector types to perform grouped strided loads/stores
6839 make sure the mode is supported by the target.
6840 (vectorizable_load): Likewise.
6842 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6844 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6845 Increase LDP/STP opportunities by adding adjacent callee-saves.
6847 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6849 PR rtl-optimization/84068
6850 PR rtl-optimization/83459
6851 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6853 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6855 PR tree-optimization/84224
6856 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6857 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6860 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6863 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6864 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6865 count, remove one USE.
6867 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6869 * doc/install.texi (Configuration): Document the
6870 --with-long-double-format={ibm,ieee} PowerPC configuration
6874 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6875 Convert from define_expand to be define_insn_and_split. Rework
6876 float/double/_Float128 conversions to QI/HI/SImode to work with
6877 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6878 conversions to QI/HImode types did a store and then a load to
6879 truncate the value. For conversions to VSX registers, don't split
6880 the insn, instead emit the code directly. Use the code iterator
6881 any_fix to combine signed and unsigned conversions.
6882 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6883 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6884 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6885 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6886 (fix_<mode>di2_hw): Likewise.
6887 (fixuns_<mode>di2_hw): Likewise.
6888 (fix_<mode>si2_hw): Likewise.
6889 (fixuns_<mode>si2_hw): Likewise.
6890 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6891 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6892 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6893 fix<uns>_trunc<SFDF:mode>si2_p8.
6894 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6896 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6897 (fix<uns>_<mode>_mem): Likewise.
6898 (fctiw<u>z_<mode>_mem): Likewise.
6899 (fix<uns>_<mode>_mem): Likewise.
6900 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6901 the register allocator from doing a direct move to the GPRs to do
6902 a store, and instead use the ISA 3.0 store byte/half-word from
6903 vector register instruction. For IEEE 128-bit floating point,
6904 also optimize stores of 32-bit ints.
6905 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6907 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6909 * genextract.c (push_pathstr_operand): New function to support
6911 (walk_rtx): Call push_pathstr_operand.
6912 (print_path): Support [a-zA-Z].
6914 2018-02-07 Richard Biener <rguenther@suse.de>
6916 PR tree-optimization/84037
6917 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6918 (cse_and_gimplify_to_preheader): Declare.
6919 (vect_get_place_in_interleaving_chain): Likewise.
6920 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6922 (_loop_vec_info::~_loop_vec_info): Delete it.
6923 (cse_and_gimplify_to_preheader): New function.
6924 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6925 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6926 (vectorizable_load): Likewise. For grouped stores always base
6927 the IV on the first element.
6928 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6929 condition before gimplifying.
6931 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6933 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6934 *DIV_EXPR and *MOD_EXPR.
6936 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6939 * config/i386/i386.c (ix86_option_override_internal): Mask out
6940 the CF_SET bit when checking -fcf-protection.
6942 2018-02-07 Tom de Vries <tom@codesourcery.com>
6945 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6948 2018-02-07 Richard Biener <rguenther@suse.de>
6950 PR tree-optimization/84204
6951 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6954 PR tree-optimization/84205
6955 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6956 special-case isl_ast_op_zdiv_r.
6958 PR tree-optimization/84223
6959 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6960 Only add conditions from within the region.
6961 (gather_bbs::after_dom_children): Adjust.
6963 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6966 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6967 * config/avr/avr.md: Only post-reload split REG-REG moves if
6968 either register is GENERAL_REG_P.
6970 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6972 PR tree-optimization/84235
6973 * tree-ssa-scopedtables.c
6974 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6975 if the subtraction is performed in floating point type where NaNs are
6976 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6977 build 1. Formatting fix.
6979 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6982 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6983 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6984 and skip it regardless of bb boundaries. Use CALL_P macro,
6985 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6986 unnecessarily, formatting fix.
6988 2018-02-06 Michael Collison <michael.collison@arm.com>
6990 * config/arm/thumb2.md:
6991 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6992 (*thumb_mov_notscc): Ditto.
6994 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6997 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6998 unsigned_fix, not "s".
7000 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7002 * configure.ac (gcc_fn_eh_frame_ro): New function.
7003 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7004 correct .eh_frame permissions.
7005 * configure: Regenerate.
7007 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7009 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7012 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7014 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7015 Display warning message for -mno-speculate-indirect-jumps.
7017 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7019 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7021 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7023 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7025 PR tree-optimization/84225
7026 * tree-eh.c (find_trapping_overflow): Only call
7027 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7029 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7032 * config/i386/i386.c: Reimplement the check of possible options
7033 -mibt/-mshstk conbination. Change error messages.
7034 * doc/invoke.texi: Fix a typo: remove extra '='.
7036 2018-02-06 Marek Polacek <polacek@redhat.com>
7038 PR tree-optimization/84228
7039 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7041 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7044 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7045 emitted arch directives.
7046 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7047 __ARM_FEATURE_COPROC before changing architectures.
7049 2018-02-06 Richard Biener <rguenther@suse.de>
7051 * config/i386/i386.c (print_reg): Fix typo.
7052 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7054 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7056 * configure: Regenerate.
7058 2018-02-05 Martin Sebor <msebor@redhat.com>
7060 PR tree-optimization/83369
7061 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7064 2018-02-05 Martin Liska <mliska@suse.cz>
7066 * doc/invoke.texi: Cherry-pick upstream r323995.
7068 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7070 * ira.c (ira_init_register_move_cost): Adjust comment.
7072 2018-02-05 Martin Liska <mliska@suse.cz>
7074 PR gcov-profile/84137
7075 * doc/gcov.texi: Fix typo in documentation.
7077 2018-02-05 Martin Liska <mliska@suse.cz>
7079 PR gcov-profile/83879
7080 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7081 using dlopen functionality.
7083 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7085 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7086 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7087 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7088 _mm_maskz_range_round_ss): New intrinsics.
7089 (__builtin_ia32_rangesd128_round)
7090 (__builtin_ia32_rangess128_round): Remove.
7091 (__builtin_ia32_rangesd128_mask_round,
7092 __builtin_ia32_rangess128_mask_round): New builtins.
7093 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7094 __builtin_ia32_rangess128_round): Remove.
7095 (__builtin_ia32_rangesd128_mask_round,
7096 __builtin_ia32_rangess128_mask_round): New builtins.
7097 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7098 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7099 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7100 "<round_saeonly_constraint>")): Changed to ...
7101 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7102 "<round_saeonly_scalar_constraint>")): ... this.
7103 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7104 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7105 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7106 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7107 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7109 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7111 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7113 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7114 Remove all values except native, 8540 and 8548.
7116 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7118 * config/i386/i386.c (ix86_output_function_return): Pass
7119 INVALID_REGNUM, instead of -1, as invalid register number to
7120 indirect_thunk_name and output_indirect_thunk.
7122 2018-02-02 Julia Koval <julia.koval@intel.com>
7124 * config.gcc: Add -march=icelake.
7125 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7126 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7127 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7128 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7129 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7130 (processor_target_table): Add icelake.
7131 (ix86_option_override_internal): Handle new PTAs.
7132 (get_builtin_code_for_version): Handle icelake.
7133 (M_INTEL_COREI7_ICELAKE): New.
7134 (fold_builtin_cpu): Handle icelake.
7135 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7136 * doc/invoke.texi: Add -march=icelake.
7138 2018-02-02 Julia Koval <julia.koval@intel.com>
7140 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7141 to wide_int_bitmask.
7142 * wide-int-bitmask.h: New.
7144 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7147 * config/i386/i386.md: Replace Pmode with word_mode in
7148 builtin_setjmp_setup and builtin_longjmp to support x32.
7150 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7154 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7156 (rs6000_supported_cpu_names): New static variable.
7157 (linux_cpu_translation_table): Likewise.
7158 (elf_platform) <cpu>: Define new static variable and use it.
7159 Translate kernel AT_PLATFORM name to canonical name if needed.
7160 Error if platform name is unknown.
7162 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7165 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7167 2018-02-01 Jeff Law <law@redhat.com>
7170 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7171 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7172 the scratch if RELEASE_VIA_POP is false.
7173 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7174 If we have to save a temporary register, decrement SIZE appropriately.
7175 Pass new arguments to release_scratch_register_on_entry.
7176 (ix86_adjust_stack_and_probe): Likewise.
7177 (ix86_emit_probe_stack_range): Pass new arguments to
7178 release_scratch_register_on_entry.
7180 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7182 PR rtl-optimization/84157
7183 * combine.c (change_zero_ext): Use REG_P predicate in
7184 front of HARD_REGISTER_P predicate.
7186 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7188 * config/avr/avr.c (avr_option_override): Move disabling of
7189 -fdelete-null-pointer-checks to...
7190 * common/config/avr/avr-common.c (avr_option_optimization_table):
7193 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7195 PR tree-optimization/81635
7196 * tree-data-ref.c (split_constant_offset_1): For types that
7197 wrap on overflow, try to use range info to prove that wrapping
7200 2018-02-01 Renlin Li <renlin.li@arm.com>
7203 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7205 (aarch64_register_move_cost): Likewise.
7206 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7208 (REG_CLASS_NAMES): Likewise.
7209 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7210 TAILCALL_ADDR_REGS. Remove IP registers.
7211 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7213 2018-02-01 Richard Biener <rguenther@suse.de>
7215 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7216 for specifying RPO order and allow NULL for that.
7217 * domwalk.c (dom_walker::dom_walker): Likewise.
7218 (dom_walker::walk): Handle NULL RPO order.
7219 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7221 (rewrite_update_dom_walker): Likewise.
7222 (mark_def_dom_walker): Likewise.
7224 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7226 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7227 (aarch64_maybe_expand_sve_subreg_move): Declare.
7228 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7229 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7231 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7232 that are semantically a reverse operation.
7233 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7234 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7235 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7237 (aarch64_can_change_mode_class): For big-endian, forbid changes
7238 between two SVE modes if they have different element sizes.
7240 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7242 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7243 the TImode handling for big-endian targets.
7245 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7247 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7248 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7250 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7251 Remove BSWAP handing for big-endian targets and use the form of
7252 LD1RQ appropariate for the mode.
7254 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7256 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7257 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7260 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7263 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7264 check for operands that need to go through aarch64_sve_reload_be.
7266 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7268 PR tree-optimization/81661
7269 PR tree-optimization/84117
7270 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7271 * tree-eh.c: Include gimplify.h.
7272 (find_trapping_overflow, replace_trapping_overflow,
7273 rewrite_to_non_trapping_overflow): New functions.
7274 * tree-vect-loop.c: Include tree-eh.h.
7275 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7276 * tree-data-ref.c: Include tree-eh.h.
7277 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7279 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7281 PR rtl-optimization/84123
7282 * combine.c (change_zero_ext): Check if hard register satisfies
7283 can_change_dest_mode before calling gen_lowpart_SUBREG.
7285 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7288 * ira.c (ira_init_register_move_cost): Remove assert.
7290 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7292 PR rtl-optimization/84071
7293 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7294 * doc/tm.texi: Regenerate.
7296 2018-01-31 Richard Biener <rguenther@suse.de>
7298 PR tree-optimization/84132
7299 * tree-data-ref.c (analyze_miv_subscript): Properly
7300 check whether evolution_function_is_affine_multivariate_p
7301 before calling gcd_of_steps_may_divide_p.
7303 2018-01-31 Julia Koval <julia.koval@intel.com>
7306 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7307 * config/i386/i386.md (rdpid_rex64) New.
7308 (rdpid): Make 32bit only.
7310 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7313 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7314 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7316 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7319 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7321 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7323 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7325 PR rtl-optimization/84071
7326 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7327 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7329 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7331 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7332 (arc_attribute_table): Add 'aux' attribute.
7333 (arc_in_small_data_p): Consider aux like variables.
7334 (arc_is_aux_reg_p): New function.
7335 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7336 (arc_get_aux_arg): New function.
7337 (prepare_move_operands): Handle aux-register access.
7338 (arc_handle_aux_attribute): New function.
7339 * doc/extend.texi (ARC Variable attributes): Add subsection.
7341 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7343 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7344 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7345 (arc_attribute_table): Add 'uncached' attribute.
7346 (arc_print_operand): Print '.di' flag for uncached memory
7348 (arc_in_small_data_p): Do not consider for small data the uncached
7350 (arc_is_uncached_mem_p): New function.
7351 * config/arc/predicates.md (compact_store_memory_operand): Check
7352 for uncached memory accesses.
7353 (nonvol_nonimm_operand): Likewise.
7354 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7356 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7359 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7360 falign-loops=): Add Optimization flag.
7362 2018-01-30 Jeff Law <law@redhat.com>
7365 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7366 INT_REGISTERS_SAVED. Check it prior to calling
7367 get_scratch_register_on_entry.
7368 (ix86_adjust_stack_and_probe): Similarly.
7369 (ix86_emit_probe_stack_range): Similarly.
7370 (ix86_expand_prologue): Corresponding changes.
7372 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7375 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7376 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7378 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7381 * lra-constraints.c (curr_insn_transform): Process AND in the
7384 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7386 PR rtl-optimization/83986
7387 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7388 dependence against last_pending_memory_flush in addition to
7391 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7393 PR tree-optimization/81611
7394 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7397 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7400 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7403 2018-01-30 Richard Biener <rguenther@suse.de>
7404 Jakub Jelinek <jakub@redhat.com>
7406 PR tree-optimization/84111
7407 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7408 inner loops added during recursion, as they don't have up-to-date
7411 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7414 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7415 (can_inline_edge_by_limits_p): ... here.
7416 (can_early_inline_edge_p, check_callers,
7417 update_caller_keys, update_callee_keys, recursive_inlining,
7418 add_new_edges_to_heap, speculation_useful_p,
7419 inline_small_functions,
7420 inline_small_functions, flatten_function,
7421 inline_to_all_callers_1): Update.
7423 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7425 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7428 2018-01-30 Richard Biener <rguenther@suse.de>
7430 PR tree-optimization/83008
7431 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7432 invariant and constant vector uses in stmts when they need
7435 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7438 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7439 * configure: Regenerate.
7441 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7443 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7445 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7446 Use gen_rtx_REG rather than gen_lowpart.
7448 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7450 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7451 rather than 0 when creating partial subregs.
7453 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7455 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7458 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7461 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7462 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7463 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7464 flags. This restores the settings used before the 2017-07-24.
7465 Turning off pre increment/decrement/modify allows IVOPTS to
7466 optimize DF/SF loops where the index is an int.
7468 2018-01-29 Richard Biener <rguenther@suse.de>
7469 Kelvin Nilsen <kelvin@gcc.gnu.org>
7472 * tree-vect-stmts.c (vectorizable_call): Don't call
7473 targetm.vectorize_builtin_md_vectorized_function if callee is
7476 2018-01-22 Carl Love <cel@us.ibm.com>
7478 * doc/extend.tex: Fix typo in second arg in
7479 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7481 2018-01-29 Richard Biener <rguenther@suse.de>
7483 PR tree-optimization/84086
7484 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7485 (flush_ssaname_freelist): When SSA names were released reset
7486 the SCEV hash table.
7488 2018-01-29 Richard Biener <rguenther@suse.de>
7490 PR tree-optimization/84057
7491 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7492 removed paths when removing edges.
7494 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7496 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7497 -mfunction-return=@var{choice}.
7499 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7502 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7504 (layout::print_source_line): Likewise.
7505 (test_get_line_width_without_trailing_whitespace): Add test cases.
7507 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7510 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7513 2018-01-26 Jim Wilson <jimw@sifive.com>
7515 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7517 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7520 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7522 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7523 and CMP + SUB-immediate -> SUBS.
7525 2018-01-26 Martin Sebor <msebor@redhat.com>
7527 PR tree-optimization/83896
7528 * tree-ssa-strlen.c (get_string_len): Rename...
7529 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7530 Avoid assuming length is constant.
7531 (handle_char_store): Use HOST_WIDE_INT for string length.
7533 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7536 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7537 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7539 2018-01-26 Richard Biener <rguenther@suse.de>
7541 PR rtl-optimization/84003
7542 * dse.c (record_store): Only record redundant stores when
7543 the earlier store aliases at least all accesses the later one does.
7545 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7547 PR rtl-optimization/83985
7548 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7549 REG_CFA_RESTORE insns.
7550 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7551 REG_CFA_RESTORE insns here.
7554 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7555 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7557 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7559 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7560 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7561 (arc_init): Likewise.
7562 (arc_override_options): Likewise.
7563 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7565 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7567 * config/arc/arc.h (TARGET_DBNZ): Define.
7568 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7569 properly set the tune attribute.
7570 (dbnz): Use TARGET_DBNZ guard.
7571 * config/arc/arc.opt (mtune): Add core3 option.
7573 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7575 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7576 recognize new pic like addresses.
7577 (arc_delegitimize_address): Clean up.
7579 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7581 * config/arc/arc-arches.def: Option mrf16 valid for all
7583 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7584 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7585 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7586 * config/arc/arc-tables.opt: Regenerate.
7587 * config/arc/arc.c (arc_conditional_register_usage): Handle
7588 reduced register file case.
7589 (arc_file_start): Set must have build attributes.
7590 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7592 * config/arc/arc.opt (mrf16): Add new option.
7593 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7594 * config/arc/genmultilib.awk: Handle new mrf16 option.
7595 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7596 * config/arc/t-multilib: Regenerate.
7597 * doc/invoke.texi (ARC Options): Document mrf16 option.
7599 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7601 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7602 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7603 (arc_attribute_table): Add 'secure_call' attribute.
7604 (arc_print_operand): Print secure call operand.
7605 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7607 (arc_is_secure_call_p): New function. * config/arc/arc.md
7608 (call_i): Add support for sjli instruction.
7609 (call_value_i): Likewise.
7610 * config/arc/constraints.md (Csc): New constraint.
7612 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7613 John Eric Martin <John.Martin@emmicro-us.com>
7615 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7616 * config/arc/arc.c (_arc_jli_section): New struct.
7617 (arc_jli_section): New type.
7618 (rc_jli_sections): New static variable.
7619 (arc_handle_jli_attribute): New function.
7620 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7621 (arc_file_end): New function.
7622 (TARGET_ASM_FILE_END): Define.
7623 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7624 (arc_add_jli_section): New function.
7625 (jli_call_scan): Likewise.
7626 (arc_reorg): Call jli_call_scan.
7627 (arc_output_addsi): Remove 'S' from printing asm operand.
7628 (arc_is_jli_call_p): New function.
7629 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7631 (movhi_insn): Likewise.
7632 (movsi_insn): Likewise.
7633 (movsi_set_cc_insn): Likewise.
7634 (loadqi_update): Likewise.
7635 (load_zeroextendqisi_update): Likewise.
7636 (load_signextendqisi_update): Likewise.
7637 (loadhi_update): Likewise.
7638 (load_zeroextendhisi_update): Likewise.
7639 (load_signextendhisi_update): Likewise.
7640 (loadsi_update): Likewise.
7641 (loadsf_update): Likewise.
7642 (movsicc_insn): Likewise.
7643 (bset_insn): Likewise.
7644 (bxor_insn): Likewise.
7645 (bclr_insn): Likewise.
7646 (bmsk_insn): Likewise.
7647 (bicsi3_insn): Likewise.
7648 (cmpsi_cc_c_insn): Likewise.
7649 (movsi_ne): Likewise.
7650 (movsi_cond_exec): Likewise.
7651 (clrsbsi2): Likewise.
7662 (call_i): Remove 'S' asm letter, add jli instruction.
7663 (call_value_i): Likewise.
7664 * config/arc/arc.op (mjli-always): New option.
7665 * config/arc/constraints.md (Cji): New constraint.
7666 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7668 (subsf3_fpx): Likewise.
7669 (mulsf3_fpx): Likewise.
7670 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7672 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7674 * doc/invoke.texi (ARC): Document mjli-always option.
7676 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7678 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7679 avoid addition with 0 and use incw and decw where possible.
7681 2018-01-26 Richard Biener <rguenther@suse.de>
7683 PR tree-optimization/81082
7684 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7685 association if it requires casting to unsigned.
7686 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7687 from fold_plusminus_mult_expr to catch important cases late when
7688 range info is available.
7690 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7692 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7693 * configure.ac (hidden_linkonce): New test.
7694 * configure: Regenerate.
7695 * config.in: Regenerate.
7697 2018-01-26 Julia Koval <julia.koval@intel.com>
7699 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7700 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7701 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7702 _mm_mask_bitshuffle_epi64_mask): Fix type.
7703 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7704 USI_FTYPE_V4DI_V4DI_USI): Remove.
7705 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7706 __builtin_ia32_vpshufbitqmb256_mask,
7707 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7708 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7709 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7711 2018-01-26 Alan Modra <amodra@gmail.com>
7714 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7715 UNSPEC_VBPERMQ. Sort other unspecs.
7717 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7719 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7721 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7724 * predict.c (drop_profile): Do not push/pop cfun; update also
7726 (handle_missing_profiles): Fix logic looking for zero profiles.
7728 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7731 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7732 on functions with #pragma omp declare simd or functions with simd
7734 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7735 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7736 Remove trailing \n from warning_at calls.
7738 2018-01-25 Tom de Vries <tom@codesourcery.com>
7741 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7742 for neutered workers.
7744 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7747 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7748 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7750 2018-01-24 Jeff Law <law@redhat.com>
7753 * i386.c (get_probe_interval): Move to earlier point.
7754 (ix86_compute_frame_layout): If -fstack-clash-protection and
7755 the frame is larger than the probe interval, then use pushes
7756 to save registers rather than reg->mem moves.
7757 (ix86_expand_prologue): Remove conditional for int_registers_saved
7760 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7763 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7764 min/max for never referenced object.
7766 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7769 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7771 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7772 attributes from DECL_ATTRIBUTES (decl) without affecting
7773 DECL_ATTRIBUTES (current_function_decl).
7774 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7775 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7777 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7779 PR tree-optimization/83979
7780 * fold-const.c (fold_comparison): Use constant_boolean_node
7781 instead of boolean_{true,false}_node.
7783 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7785 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7788 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7790 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7791 Simplify the clause that sets the length attribute.
7792 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7793 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7794 clause that sets the length attribute.
7795 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7797 2018-01-24 Tom de Vries <tom@codesourcery.com>
7800 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7801 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7802 Add strict parameter.
7803 (prevent_branch_around_nothing): Insert dummy insn between branch to
7804 label and label with no ptx insn inbetween.
7805 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7807 2018-01-24 Tom de Vries <tom@codesourcery.com>
7810 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7811 for neutered threads in warp.
7812 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7814 2018-01-24 Richard Biener <rguenther@suse.de>
7816 PR tree-optimization/83176
7817 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7820 2018-01-24 Richard Biener <rguenther@suse.de>
7822 PR tree-optimization/82819
7823 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7824 code generating pluses that are no-ops in the target precision.
7826 2018-01-24 Richard Biener <rguenther@suse.de>
7829 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7831 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7833 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7834 to merge probabilities.
7835 * predict.c (probably_never_executed): Also mark as cold functions
7836 with global 0 profile and guessed local profile.
7837 * profile-count.c (profile_probability::combine_with_count): New
7839 * profile-count.h (profile_probability::operator*,
7840 profile_probability::operator*=, profile_probability::operator/,
7841 profile_probability::operator/=): Reduce precision to adjusted
7842 and set value to guessed on contradictory divisions.
7843 (profile_probability::combine_with_freq): Remove.
7844 (profile_probability::combine_wiht_count): Declare.
7845 (profile_count::force_nonzero):: Set to adjusted.
7846 (profile_count::probability_in):: Set quality to adjusted.
7847 * tree-ssa-tail-merge.c (replace_block_by): Use
7850 2018-01-23 Andrew Waterman <andrew@sifive.com>
7851 Jim Wilson <jimw@sifive.com>
7853 * config/riscv/riscv.c (riscv_stack_boundary): New.
7854 (riscv_option_override): Set riscv_stack_boundary. Handle
7855 riscv_preferred_stack_boundary_arg.
7856 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7857 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7858 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7859 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7860 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7861 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7863 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7866 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7867 of struct ix86_frame.
7868 (ix86_expand_epilogue): Likewise. Add a local variable for
7869 the reg_save_offset field in struct ix86_frame.
7871 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7873 PR tree-optimization/82604
7874 * tree-loop-distribution.c (enum partition_kind): New enum item
7875 PKIND_PARTIAL_MEMSET.
7876 (partition_builtin_p): Support above new enum item.
7877 (generate_code_for_partition): Ditto.
7878 (compute_access_range): Differentiate cases that equality can be
7879 proven at all loops, the innermost loops or no loops.
7880 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7881 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7882 (finalize_partitions, distribute_loop): Don't fuse partition of
7883 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7884 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7887 2018-01-23 Martin Liska <mliska@suse.cz>
7889 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7890 order to ignore the predictor.
7891 (PRED_POLYMORPHIC_CALL): Likewise.
7892 (PRED_RECURSIVE_CALL): Likewise.
7894 2018-01-23 Martin Liska <mliska@suse.cz>
7896 * tree-profile.c (tree_profiling): Print function header to
7897 aware reader which function we are working on.
7898 * value-prof.c (gimple_find_values_to_profile): Do not print
7899 not interesting value histograms.
7901 2018-01-23 Martin Liska <mliska@suse.cz>
7903 * profile-count.h (enum profile_quality): Add
7904 profile_uninitialized as the first value. Do not number values
7905 as they are zero based.
7906 (profile_count::verify): Update sanity check.
7907 (profile_probability::verify): Likewise.
7909 2018-01-23 Nathan Sidwell <nathan@acm.org>
7911 * doc/invoke.texi (ffor-scope): Deprecate.
7913 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7915 PR tree-optimization/83510
7916 * domwalk.c (set_all_edges_as_executable): New function.
7917 (dom_walker::dom_walker): Convert bool param
7918 "skip_unreachable_blocks" to enum reachability. Move setup of
7919 edge flags to set_all_edges_as_executable and only do it when
7920 reachability is REACHABLE_BLOCKS.
7921 * domwalk.h (enum dom_walker::reachability): New enum.
7922 (dom_walker::dom_walker): Convert bool param
7923 "skip_unreachable_blocks" to enum reachability.
7924 (set_all_edges_as_executable): New decl.
7925 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7926 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7928 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7929 but converting true to REACHABLE_BLOCKS.
7930 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7932 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7933 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7934 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7936 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7937 if check_all_array_refs will be called.
7939 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7941 * tree.c (selftest::test_location_wrappers): Add more test
7944 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7946 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7947 (selftest::test_bit_in_range): Likewise.
7949 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7952 * doc/sourcebuild.texi (vect_float): Say that the selector
7953 only describes the situation when -funsafe-math-optimizations is on.
7954 (vect_float_strict): Document.
7956 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7958 PR tree-optimization/83965
7959 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7960 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7961 instead of checking only for a reduction.
7962 (vect_recog_widen_sum_pattern): Likewise.
7964 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7966 * predict.c (probably_never_executed): Only use precise profile info.
7967 (compute_function_frequency): Skip after inlining hack since we now
7968 have quality checking.
7970 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7972 * profile-count.h (profile_probability::very_unlikely,
7973 profile_probability::unlikely, profile_probability::even): Set
7974 precision to guessed.
7976 2018-01-23 Richard Biener <rguenther@suse.de>
7978 PR tree-optimization/83963
7979 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7980 Properly terminate dominator walk when crossing the exit edge not
7981 when visiting its source block.
7983 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7986 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7987 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7989 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7991 PR tree-optimization/83957
7992 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7993 semicolon after for body surrounded by braces.
7995 PR tree-optimization/83081
7996 * profile-count.h (profile_probability::split): New method.
7997 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7998 Use profile_probability::split.
7999 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8000 when splitting a single conditional jump into 2.
8002 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8004 PR tree-optimization/69452
8005 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8008 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8010 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8011 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8012 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8014 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8016 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8017 * config/rl78/rl78.md (movdi): New define_expand.
8018 * config/rl78/rl78.c (rl78_split_movdi): New function.
8020 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8023 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8025 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8026 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8027 128-bit to produce an UNSPEC move to get the double word with the
8028 signbit and then a shift directly to do signbit.
8029 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8030 implementation with a new version that just does either a direct
8031 move or a regular move. Move memory interface to separate insns.
8032 Move insns so they are next to the expander.
8033 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8034 with signbit move. Split big and little endian case.
8035 (signbit<mode>2_dm_mem_le): Likewise.
8036 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8037 (signbit<mode>2_dm2): Likewise.
8039 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8041 * config/rl78/rl78.md (anddi3): New define_expand.
8043 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8045 * config/rl78/rl78.md (umindi3): New define_expand.
8047 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8049 * config/rl78/rl78.md (smindi3): New define_expand.
8051 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8053 * config/rl78/rl78.md (smaxdi3): New define_expand.
8055 2018-01-22 Carl Love <cel@us.ibm.com>
8057 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8058 LVX_V1TI): Add macro expansion.
8059 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8060 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8061 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8062 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8063 Change check to determine if the instruction is a byte reversing
8064 entry. Fix typo in comment.
8065 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8066 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8067 Add def_builtin calls for new builtins.
8068 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8069 Add define_insn expansion.
8071 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8073 * config/rl78/rl78.md (umaxdi3): New define_expand.
8075 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8077 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8078 for non-QImode registers.
8080 2018-01-22 Richard Biener <rguenther@suse.de>
8082 PR tree-optimization/83963
8083 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8084 including the loop exit block.
8085 (scop_detection::merge_sese): Likewise.
8086 (scop_detection::add_scop): Do it here instead.
8088 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8090 * doc/sourcebuild.texi (arm_softfloat): Document.
8092 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8095 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8096 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8097 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8099 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8100 David Edelsohn <dje.gcc@gmail.com>
8103 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8104 Change "crset eq" to "crset 2".
8105 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8106 (*call_indirect_aix<mode>_nospec): Likewise.
8107 (*call_value_indirect_aix<mode>_nospec): Likewise.
8108 (*call_indirect_elfv2<mode>_nospec): Likewise.
8109 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8110 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8111 change assembly output from . to $.
8112 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8113 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8114 (*tablejump<mode>_internal1_nospec): Likewise.
8116 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8119 * config/sh/sh_optimize_sett_clrt.cc:
8120 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8122 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8124 PR tree-optimization/83940
8125 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8126 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8127 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8128 use it to pass back the definition type.
8129 (vect_check_store_rhs): Likewise rhs_dt_out.
8130 (vect_build_gather_load_calls): Add a mask_dt argument and use
8131 it instead of a call to vect_is_simple_use.
8132 (vectorizable_store): Update calls to vect_check_load_store_mask
8133 and vect_check_store_rhs. Use the dt returned by the latter instead
8134 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8135 instead of calls to vect_is_simple_use. Pass the scalar rather
8136 than the vector operand to vect_is_simple_use when handling
8137 second and subsequent copies of an rhs value.
8138 (vectorizable_load): Update calls to vect_check_load_store_mask
8139 and vect_build_gather_load_calls. Use the cached mask_dt and
8140 gs_info.offset_dt instead of calls to vect_is_simple_use.
8142 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8145 * tree-emutls.c: Include gimplify.h.
8146 (lower_emutls_2): New function.
8147 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8148 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8149 it before further processing.
8152 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8153 UINTVAL (trueop1) instead of INTVAL (op1).
8155 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8159 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8160 INCOMING_FRAME_SP_OFFSET if not defined.
8161 (scan_trace): Add ENTRY argument. If true and
8162 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8163 emit a note to adjust the CFA offset.
8164 (create_cfi_notes): Adjust scan_trace callers.
8165 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8166 INCOMING_FRAME_SP_OFFSET in the CIE.
8167 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8168 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8170 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8171 * doc/tm.texi: Regenerated.
8173 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8175 PR rtl-optimization/83147
8176 * lra-constraints.c (remove_inheritance_pseudos): Use
8177 lra_substitute_pseudo_within_insn.
8179 2018-01-19 Tom de Vries <tom@codesourcery.com>
8180 Cesar Philippidis <cesar@codesourcery.com>
8183 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8185 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8188 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8189 spaces for function labels.
8191 2018-01-19 Martin Liska <mliska@suse.cz>
8193 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8194 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8195 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8196 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8197 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8198 (PRED_CONST_RETURN): Change from 69 to 65.
8199 (PRED_NULL_RETURN): Change from 91 to 71.
8200 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8201 (PRED_LOOP_GUARD): Change from 66 to 73.
8203 2018-01-19 Martin Liska <mliska@suse.cz>
8205 * predict.c (predict_insn_def): Add new assert.
8206 (struct branch_predictor): Change type to signed integer.
8207 (test_prediction_value_range): Amend test to cover
8209 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8210 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8211 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8212 (PRED_LOOP_IV_COMPARE): Likewise.
8213 * predict.h (PROB_UNINITIALIZED): Define new constant.
8215 2018-01-19 Martin Liska <mliska@suse.cz>
8217 * predict.c (dump_prediction): Add new format for
8218 analyze_brprob.py script which is enabled with -details
8220 * profile-count.h (precise_p): New function.
8222 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8224 PR tree-optimization/83922
8225 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8226 there are no statements that need masking.
8227 (vect_active_double_reduction_p): New function.
8228 (vect_analyze_loop_operations): Use it when handling phis that
8229 are not in the loop header.
8231 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8233 PR tree-optimization/83914
8234 * tree-vect-loop.c (vectorizable_induction): Don't convert
8235 init_expr or apply the peeling adjustment for inductions
8236 that are nested within the vectorized loop.
8238 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8240 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8243 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8247 * function.h (gimplify_parameters): Add gimple_seq * argument.
8248 * function.c: Include gimple.h and options.h.
8249 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8250 for the added local temporaries if needed.
8251 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8252 if there are any parameter cleanups, wrap whole body into a
8253 try/finally with the cleanups.
8255 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8258 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8259 Use GET_MODE_CLASS for scalar floating point.
8261 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8265 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8266 Fix call of call_cgraph_insertion_hooks.
8268 2018-01-18 Martin Sebor <msebor@redhat.com>
8270 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8272 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8275 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8278 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8281 * common.opt: (-ffile-prefix-map): New option.
8282 * opts.c (common_handle_option): Defer it.
8283 * opts-global.c (handle_common_deferred_options): Handle it.
8284 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8285 * file-prefix-map.h: New file.
8286 (remap_debug_filename, add_debug_prefix_map): ...here.
8287 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8288 * final.c (debug_prefix_map, add_debug_prefix_map
8289 remap_debug_filename): Move to...
8290 * file-prefix-map.c: New file.
8291 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8292 generalize, get rid of alloca(), use strrchr() instead of strchr().
8293 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8294 Implement in terms of add_prefix_map().
8295 (remap_macro_filename, remap_debug_filename): Implement in term of
8297 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8298 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8299 * dbxout.c: Include file-prefix-map.h.
8300 * varasm.c: Likewise.
8301 * vmsdbgout.c: Likewise.
8302 * xcoffout.c: Likewise.
8303 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8304 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8305 * doc/invoke.texi (-ffile-prefix-map): Document.
8306 (-fdebug-prefix-map): Update description.
8308 2018-01-18 Martin Liska <mliska@suse.cz>
8310 * config/i386/i386.c (indirect_thunk_name): Document that also
8312 (output_indirect_thunk): Document why both instructions
8313 (pause and lfence) are generated.
8315 2018-01-18 Richard Biener <rguenther@suse.de>
8317 PR tree-optimization/83887
8318 * graphite-scop-detection.c
8319 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8320 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8321 (scop_detection::merge_sese): Re-implement with a flood-fill
8322 algorithm that properly finds a SESE region if it exists.
8324 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8327 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8328 pointer_diff optimizations use view_convert instead of convert.
8330 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8332 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8333 Generate different code for -mno-speculate-indirect-jumps.
8334 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8335 (*call_indirect_aix<mode>): Disable for
8336 -mno-speculate-indirect-jumps.
8337 (*call_indirect_aix<mode>_nospec): New define_insn.
8338 (*call_value_indirect_aix<mode>): Disable for
8339 -mno-speculate-indirect-jumps.
8340 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8341 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8342 -mno-speculate-indirect-jumps.
8343 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8345 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8347 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8348 long double type, set the flags for noting the default long double
8349 type, even if we don't pass or return a long double type.
8351 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8354 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8357 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8359 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8360 support for merge[hl].
8361 (fold_mergehl_helper): New helper function.
8362 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8363 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8364 (altivec_vmrglw_direct): Add xxmrglw insn.
8366 2018-01-17 Andrew Waterman <andrew@sifive.com>
8368 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8369 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8371 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8374 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8375 call the lto_location_cache before reading the
8376 DECL_SOURCE_LOCATION of the types.
8378 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8379 Richard Sandiford <richard.sandiford@linaro.org>
8381 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8382 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8383 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8384 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8385 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8387 * config/aarch64/constraints.md (aarch64_movti_operand):
8389 * config/aarch64/predicates.md (Uti): Add new constraint.
8391 2018-01-17 Carl Love <cel@us.ibm.com>
8393 * config/rs6000/vsx.md (define_expand xl_len_r,
8394 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8395 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8397 (define_expand, define_insn): Move the shift left from the
8398 define_insn to the define_expand for lxvl and stxvl instructions.
8399 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8400 and XL_LEN_R definitions to PURE.
8402 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8404 * config/i386/i386.c (indirect_thunk_name): Declare regno
8405 as unsigned int. Compare regno with INVALID_REGNUM.
8406 (output_indirect_thunk): Ditto.
8407 (output_indirect_thunk_function): Ditto.
8408 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8409 in the call to output_indirect_thunk_function.
8411 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8414 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8415 rather than the size of inner_type to determine the stack slot size
8416 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8418 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8421 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8424 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8426 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8427 endian Linux systems to optionally enable multilibs for selecting
8428 the long double type if the user configured an explicit type.
8429 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8430 have no long double multilibs if not defined.
8431 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8432 warn if the user used -mabi={ieee,ibm}longdouble and we built
8433 multilibs for long double.
8434 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8435 appropriate multilib option.
8436 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8438 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8439 for building long double multilibs.
8440 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8442 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8444 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8447 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8449 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8452 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8455 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8458 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8460 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8461 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8463 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8465 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8466 different rtl trees depending on TARGET_64BIT.
8467 (rs6000_gen_lvx): Likewise.
8469 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8471 * config/visium/visium.md (nop): Tweak comment.
8472 (hazard_nop): Likewise.
8474 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8476 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8477 -mspeculate-indirect-jumps.
8478 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8479 for -mno-speculate-indirect-jumps.
8480 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8481 (*call_value_indirect_elfv2<mode>): Disable for
8482 -mno-speculate-indirect-jumps.
8483 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8484 (indirect_jump): Emit different RTL for
8485 -mno-speculate-indirect-jumps.
8486 (*indirect_jump<mode>): Disable for
8487 -mno-speculate-indirect-jumps.
8488 (*indirect_jump<mode>_nospec): New define_insn.
8489 (tablejump): Emit different RTL for
8490 -mno-speculate-indirect-jumps.
8491 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8492 (tablejumpsi_nospec): New define_expand.
8493 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8494 (tablejumpdi_nospec): New define_expand.
8495 (*tablejump<mode>_internal1): Disable for
8496 -mno-speculate-indirect-jumps.
8497 (*tablejump<mode>_internal1_nospec): New define_insn.
8498 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8501 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8503 * caller-save.c (insert_save): Drop unnecessary parameter. All
8506 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8507 Richard Biener <rguenth@suse.de>
8510 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8511 return early, inline manually is_gimple_sizepos. Make sure if we
8512 call gimplify_expr we don't end up with a gimple constant.
8513 * tree.c (variably_modified_type_p): Don't return true for
8514 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8515 * gimplify.h (is_gimple_sizepos): Remove.
8517 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8519 PR tree-optimization/83857
8520 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8521 vectorizable_live_operation for pure SLP statements.
8522 (vectorizable_live_operation): Handle PHIs.
8524 2018-01-16 Richard Biener <rguenther@suse.de>
8526 PR tree-optimization/83867
8527 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8528 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8530 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8533 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8534 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8535 If off is not INTEGER_CST, issue a may not be aligned warning
8536 rather than isn't aligned. Use isn%'t rather than isn't.
8537 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8539 <case MULT_EXPR>: Improve the case when bottom and one of the
8540 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8541 operand, in that case check if the other operand is multiple of
8542 bottom divided by the INTEGER_CST operand.
8544 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8547 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8548 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8549 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8550 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8551 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8552 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8553 (pa_function_arg_size): New function.
8555 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8557 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8558 in a separate statement.
8560 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8562 PR tree-optimization/83847
8563 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8564 group gathers and scatters.
8566 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8568 PR rtl-optimization/86620
8569 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8571 PR rtl-optimization/83213
8572 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8573 to last if both are JUMP_INSNs.
8575 PR tree-optimization/83843
8576 * gimple-ssa-store-merging.c
8577 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8578 store_immediate_info for bswap/nop orig_stores.
8580 2018-01-15 Andrew Waterman <andrew@sifive.com>
8582 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8584 <UDIV>: Increase cost if !TARGET_DIV.
8586 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8588 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8589 (define_attr "cr_logical_3op"): New.
8590 (cceq_ior_compare): Adjust.
8591 (cceq_ior_compare_complement): Adjust.
8592 (*cceq_rev_compare): Adjust.
8593 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8594 (is_cracked_insn): Adjust.
8595 (insn_must_be_first_in_group): Adjust.
8596 * config/rs6000/40x.md: Adjust.
8597 * config/rs6000/440.md: Adjust.
8598 * config/rs6000/476.md: Adjust.
8599 * config/rs6000/601.md: Adjust.
8600 * config/rs6000/603.md: Adjust.
8601 * config/rs6000/6xx.md: Adjust.
8602 * config/rs6000/7450.md: Adjust.
8603 * config/rs6000/7xx.md: Adjust.
8604 * config/rs6000/8540.md: Adjust.
8605 * config/rs6000/cell.md: Adjust.
8606 * config/rs6000/e300c2c3.md: Adjust.
8607 * config/rs6000/e500mc.md: Adjust.
8608 * config/rs6000/e500mc64.md: Adjust.
8609 * config/rs6000/e5500.md: Adjust.
8610 * config/rs6000/e6500.md: Adjust.
8611 * config/rs6000/mpc.md: Adjust.
8612 * config/rs6000/power4.md: Adjust.
8613 * config/rs6000/power5.md: Adjust.
8614 * config/rs6000/power6.md: Adjust.
8615 * config/rs6000/power7.md: Adjust.
8616 * config/rs6000/power8.md: Adjust.
8617 * config/rs6000/power9.md: Adjust.
8618 * config/rs6000/rs64.md: Adjust.
8619 * config/rs6000/titan.md: Adjust.
8621 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8623 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8624 ix86_indirect_branch_register logic.
8626 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8628 * config/i386/constraints.md (Bs): Update
8629 ix86_indirect_branch_register check. Don't check
8630 ix86_indirect_branch_register with GOT_memory_operand.
8632 * config/i386/predicates.md (GOT_memory_operand): Don't check
8633 ix86_indirect_branch_register here.
8634 (GOT32_symbol_operand): Likewise.
8636 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8638 * config/i386/predicates.md (constant_call_address_operand):
8639 Rewrite ix86_indirect_branch_register logic.
8640 (sibcall_insn_operand): Likewise.
8642 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8644 * config/i386/constraints.md (Bs): Replace
8645 ix86_indirect_branch_thunk_register with
8646 ix86_indirect_branch_register.
8648 * config/i386/i386.md (indirect_jump): Likewise.
8649 (tablejump): Likewise.
8650 (*sibcall_memory): Likewise.
8651 (*sibcall_value_memory): Likewise.
8652 Peepholes of indirect call and jump via memory: Likewise.
8653 * config/i386/i386.opt: Likewise.
8654 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8655 (GOT_memory_operand): Likewise.
8656 (call_insn_operand): Likewise.
8657 (sibcall_insn_operand): Likewise.
8658 (GOT32_symbol_operand): Likewise.
8660 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8663 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8664 type rather than type addr's type points to.
8665 (expand_omp_atomic_mutex): Likewise.
8666 (expand_omp_atomic): Likewise.
8668 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8671 * config/i386/i386.c (output_indirect_thunk_function): Use
8672 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8673 for __x86_return_thunk.
8675 2018-01-15 Richard Biener <rguenther@suse.de>
8678 * expmed.c (extract_bit_field_1): Fix typo.
8680 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8683 * config/arm/iterators.md (VF): New mode iterator.
8684 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8685 Remove integer-related logic from pattern.
8686 (neon_vabd<mode>_3): Likewise.
8688 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8691 * common.opt (fstrict-overflow): No longer an alias.
8692 (fwrapv-pointer): New option.
8693 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8694 also for pointer types based on flag_wrapv_pointer.
8695 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8696 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8697 opts->x_flag_wrapv got set.
8698 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8699 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8700 POINTER_TYPE_OVERFLOW_UNDEFINED.
8701 * match.pd: Likewise in address comparison pattern.
8702 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8704 2018-01-15 Richard Biener <rguenther@suse.de>
8707 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8708 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8709 Reset type names to their identifier if their TYPE_DECL doesn't
8710 have linkage (and thus is used for ODR and devirt).
8711 (save_debug_info_for_decl): Remove.
8712 (save_debug_info_for_type): Likewise.
8713 (add_tree_to_fld_list): Adjust.
8714 * tree-pretty-print.c (dump_generic_node): Make dumping of
8715 type names more robust.
8717 2018-01-15 Richard Biener <rguenther@suse.de>
8719 * BASE-VER: Bump to 8.0.1.
8721 2018-01-14 Martin Sebor <msebor@redhat.com>
8724 * builtins.c (check_access): Avoid warning when the no-warning bit
8727 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8729 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8730 * ira-color (allocno_hard_regs_compare): Likewise.
8732 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8735 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8736 Use .pushsection/.popsection.
8738 2018-01-14 Martin Sebor <msebor@redhat.com>
8741 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8743 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8745 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8746 entry from extra_headers.
8747 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8748 extra_headers, make the list bitwise identical to the i?86-*-* one.
8750 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8752 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8753 -mcmodel=large with -mindirect-branch=thunk,
8754 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8755 -mfunction-return=thunk-extern.
8756 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8757 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8758 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8760 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8762 * config/i386/i386.c (print_reg): Print the name of the full
8763 integer register without '%'.
8764 (ix86_print_operand): Handle 'V'.
8765 * doc/extend.texi: Document 'V' modifier.
8767 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8769 * config/i386/constraints.md (Bs): Disallow memory operand for
8770 -mindirect-branch-register.
8772 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8773 (GOT_memory_operand): Likewise.
8774 (call_insn_operand): Likewise.
8775 (sibcall_insn_operand): Likewise.
8776 (GOT32_symbol_operand): Likewise.
8777 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8778 for -mindirect-branch-register.
8779 (tablejump): Likewise.
8780 (*sibcall_memory): Likewise.
8781 (*sibcall_value_memory): Likewise.
8782 Disallow peepholes of indirect call and jump via memory for
8783 -mindirect-branch-register.
8784 (*call_pop): Replace m with Bw.
8785 (*call_value_pop): Likewise.
8786 (*sibcall_pop_memory): Replace m with Bs.
8787 * config/i386/i386.opt (mindirect-branch-register): New option.
8788 * doc/invoke.texi: Document -mindirect-branch-register option.
8790 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8792 * config/i386/i386-protos.h (ix86_output_function_return): New.
8793 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8794 set function_return_type.
8795 (indirect_thunk_name): Add ret_p to indicate thunk for function
8797 (output_indirect_thunk_function): Pass false to
8798 indirect_thunk_name.
8799 (ix86_output_indirect_branch_via_reg): Likewise.
8800 (ix86_output_indirect_branch_via_push): Likewise.
8801 (output_indirect_thunk_function): Create alias for function
8802 return thunk if regno < 0.
8803 (ix86_output_function_return): New function.
8804 (ix86_handle_fndecl_attribute): Handle function_return.
8805 (ix86_attribute_table): Add function_return.
8806 * config/i386/i386.h (machine_function): Add
8807 function_return_type.
8808 * config/i386/i386.md (simple_return_internal): Use
8809 ix86_output_function_return.
8810 (simple_return_internal_long): Likewise.
8811 * config/i386/i386.opt (mfunction-return=): New option.
8812 (indirect_branch): Mention -mfunction-return=.
8813 * doc/extend.texi: Document function_return function attribute.
8814 * doc/invoke.texi: Document -mfunction-return= option.
8816 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8818 * config/i386/i386-opts.h (indirect_branch): New.
8819 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8820 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8821 with local indirect jump when converting indirect call and jump.
8822 (ix86_set_indirect_branch_type): New.
8823 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8824 (indirectlabelno): New.
8825 (indirect_thunk_needed): Likewise.
8826 (indirect_thunk_bnd_needed): Likewise.
8827 (indirect_thunks_used): Likewise.
8828 (indirect_thunks_bnd_used): Likewise.
8829 (INDIRECT_LABEL): Likewise.
8830 (indirect_thunk_name): Likewise.
8831 (output_indirect_thunk): Likewise.
8832 (output_indirect_thunk_function): Likewise.
8833 (ix86_output_indirect_branch_via_reg): Likewise.
8834 (ix86_output_indirect_branch_via_push): Likewise.
8835 (ix86_output_indirect_branch): Likewise.
8836 (ix86_output_indirect_jmp): Likewise.
8837 (ix86_code_end): Call output_indirect_thunk_function if needed.
8838 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8840 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8841 (ix86_attribute_table): Add indirect_branch.
8842 * config/i386/i386.h (machine_function): Add indirect_branch_type
8843 and has_local_indirect_jump.
8844 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8846 (tablejump): Likewise.
8847 (*indirect_jump): Use ix86_output_indirect_jmp.
8848 (*tablejump_1): Likewise.
8849 (simple_return_indirect_internal): Likewise.
8850 * config/i386/i386.opt (mindirect-branch=): New option.
8851 (indirect_branch): New.
8854 (thunk-inline): Likewise.
8855 (thunk-extern): Likewise.
8856 * doc/extend.texi: Document indirect_branch function attribute.
8857 * doc/invoke.texi: Document -mindirect-branch= option.
8859 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8862 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8864 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8866 * ipa-inline.c (want_inline_small_function_p): Return false if
8867 inlining has already failed with CIF_FINAL_ERROR.
8868 (update_caller_keys): Call want_inline_small_function_p before
8870 (update_callee_keys): Likewise.
8872 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8874 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8876 (rs6000_quadword_masked_address_p): Likewise.
8877 (quad_aligned_load_p): Likewise.
8878 (quad_aligned_store_p): Likewise.
8879 (const_load_sequence_p): Add comment to describe the outer-most loop.
8880 (mimic_memory_attributes_and_flags): New function.
8881 (rs6000_gen_stvx): Likewise.
8882 (replace_swapped_aligned_store): Likewise.
8883 (rs6000_gen_lvx): Likewise.
8884 (replace_swapped_aligned_load): Likewise.
8885 (replace_swapped_load_constant): Capitalize argument name in
8886 comment describing this function.
8887 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8888 and stores that access quad-word aligned addresses and replace
8889 with stvx or lvx instructions when appropriate.
8890 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8891 New function prototype.
8892 (rs6000_quadword_masked_address_p): Likewise.
8893 (rs6000_gen_lvx): Likewise.
8894 (rs6000_gen_stvx): Likewise.
8895 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8896 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8897 when memory address is aligned.
8898 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8899 this split to select lvx instruction when memory address is aligned.
8900 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8901 instruction when memory address is aligned.
8902 (*vsx_le_perm_load_v16qi): Likewise.
8903 (four unnamed splitters): Modify to select the stvx instruction
8904 when memory is aligned.
8906 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8908 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8909 which appears in the queue multiple times.
8911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8912 Alan Hayward <alan.hayward@arm.com>
8913 David Sherwood <david.sherwood@arm.com>
8915 * tree-vectorizer.h (vec_lower_bound): New structure.
8916 (_loop_vec_info): Add check_nonzero and lower_bounds.
8917 (LOOP_VINFO_CHECK_NONZERO): New macro.
8918 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8919 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8920 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8921 fields. Make seg_len the distance travelled, not including the
8923 (dr_direction_indicator): Declare.
8924 (dr_zero_step_indicator): Likewise.
8925 (dr_known_forward_stride_p): Likewise.
8926 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8928 (runtime_alias_check_p): Allow runtime alias checks with
8930 (operator ==): Compare access_size and align.
8931 (prune_runtime_alias_test_list): Rework for new distinction between
8932 the access_size and seg_len.
8933 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8935 (get_segment_min_max): New function.
8936 (create_intersect_range_checks): Use it.
8937 (dr_step_indicator): New function.
8938 (dr_direction_indicator): Likewise.
8939 (dr_zero_step_indicator): Likewise.
8940 (dr_known_forward_stride_p): Likewise.
8941 * tree-loop-distribution.c (data_ref_segment_size): Return
8942 DR_STEP * (niters - 1).
8943 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8945 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8946 (vect_preserves_scalar_order_p): New function, split out from...
8947 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8948 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8949 (vect_vfa_access_size): New function.
8950 (vect_vfa_align): Likewise.
8951 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8952 (dump_lower_bound): New function.
8953 (vect_check_lower_bound): Likewise.
8954 (vect_small_gap_p): Likewise.
8955 (vectorizable_with_step_bound_p): Likewise.
8956 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8957 depencies if the vectorization factor is 1. Convert the checks
8958 for nonzero steps into checks on the bounds of DR_STEP. Try using
8959 a bunds check for variable steps if the minimum required step is
8960 relatively small. Update calls to the dr_with_seg_len
8961 constructor and to vect_compile_time_alias.
8962 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8964 (vect_loop_versioning): Call it.
8965 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8967 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8969 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8970 Alan Hayward <alan.hayward@arm.com>
8971 David Sherwood <david.sherwood@arm.com>
8973 * doc/sourcebuild.texi (vect_scatter_store): Document.
8974 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8976 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8978 * genopinit.c (main): Add supports_vec_scatter_store and
8979 supports_vec_scatter_store_cached to target_optabs.
8980 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8981 IFN_MASK_SCATTER_STORE.
8982 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8984 * internal-fn.h (internal_store_fn_p): Declare.
8985 (internal_fn_stored_value_index): Likewise.
8986 * internal-fn.c (scatter_store_direct): New macro.
8987 (expand_scatter_store_optab_fn): New function.
8988 (direct_scatter_store_optab_supported_p): New macro.
8989 (internal_store_fn_p): New function.
8990 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8991 IFN_MASK_SCATTER_STORE.
8992 (internal_fn_mask_index): Likewise.
8993 (internal_fn_stored_value_index): New function.
8994 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8996 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8997 * optabs-query.c (supports_vec_scatter_store_p): New function.
8998 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8999 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9000 true for scatter stores.
9001 (vect_gather_scatter_fn_p): Handle scatter stores too.
9002 (vect_check_gather_scatter): Consider using scatter stores if
9003 supports_vec_scatter_store_p.
9004 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9006 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9007 internal_fn_stored_value_index.
9008 (check_load_store_masking): Handle scatter stores too.
9009 (vect_get_store_rhs): Make public.
9010 (vectorizable_call): Use internal_store_fn_p.
9011 (vectorizable_store): Handle scatter store internal functions.
9012 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9013 when deciding whether the end of the group has been reached.
9014 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9015 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9016 (mask_scatter_store<mode>): New insns.
9018 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9019 Alan Hayward <alan.hayward@arm.com>
9020 David Sherwood <david.sherwood@arm.com>
9022 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9023 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9024 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9026 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9027 Use vect_truncate_gather_scatter_offset if we can't treat the
9028 operation as a normal gather load or scatter store.
9029 (get_group_load_store_type): Take the gather_scatter_info
9030 as argument. Try using a gather load or scatter store for
9031 single-element groups.
9032 (get_load_store_type): Update calls to get_group_load_store_type
9033 and vect_use_strided_gather_scatters_p.
9035 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9036 Alan Hayward <alan.hayward@arm.com>
9037 David Sherwood <david.sherwood@arm.com>
9039 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9040 optional tree argument.
9041 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9043 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9044 but continue to use the current value as a fallback.
9045 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9046 to compare the updates.
9047 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9048 (get_load_store_type): Use it when handling a strided access.
9049 (vect_get_strided_load_store_ops): New function.
9050 (vect_get_data_ptr_increment): Likewise.
9051 (vectorizable_load): Handle strided gather loads. Always pass
9052 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9055 Alan Hayward <alan.hayward@arm.com>
9056 David Sherwood <david.sherwood@arm.com>
9058 * doc/md.texi (gather_load@var{m}): Document.
9059 (mask_gather_load@var{m}): Likewise.
9060 * genopinit.c (main): Add supports_vec_gather_load and
9061 supports_vec_gather_load_cached to target_optabs.
9062 * optabs-tree.c (init_tree_optimization_optabs): Use
9063 ggc_cleared_alloc to allocate target_optabs.
9064 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9065 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9067 * internal-fn.h (internal_load_fn_p): Declare.
9068 (internal_gather_scatter_fn_p): Likewise.
9069 (internal_fn_mask_index): Likewise.
9070 (internal_gather_scatter_fn_supported_p): Likewise.
9071 * internal-fn.c (gather_load_direct): New macro.
9072 (expand_gather_load_optab_fn): New function.
9073 (direct_gather_load_optab_supported_p): New macro.
9074 (direct_internal_fn_optab): New function.
9075 (internal_load_fn_p): Likewise.
9076 (internal_gather_scatter_fn_p): Likewise.
9077 (internal_fn_mask_index): Likewise.
9078 (internal_gather_scatter_fn_supported_p): Likewise.
9079 * optabs-query.c (supports_at_least_one_mode_p): New function.
9080 (supports_vec_gather_load_p): Likewise.
9081 * optabs-query.h (supports_vec_gather_load_p): Declare.
9082 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9083 and memory_type field.
9084 (NUM_PATTERNS): Bump to 15.
9085 * tree-vect-data-refs.c: Include internal-fn.h.
9086 (vect_gather_scatter_fn_p): New function.
9087 (vect_describe_gather_scatter_call): Likewise.
9088 (vect_check_gather_scatter): Try using internal functions for
9089 gather loads. Recognize existing calls to a gather load function.
9090 (vect_analyze_data_refs): Consider using gather loads if
9091 supports_vec_gather_load_p.
9092 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9093 (vect_get_gather_scatter_offset_type): Likewise.
9094 (vect_convert_mask_for_vectype): Likewise.
9095 (vect_add_conversion_to_patterm): Likewise.
9096 (vect_try_gather_scatter_pattern): Likewise.
9097 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9098 (vect_vect_recog_func_ptrs): Add it.
9099 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9100 internal_fn_mask_index and internal_gather_scatter_fn_p.
9101 (check_load_store_masking): Take the gather_scatter_info as an
9102 argument and handle gather loads.
9103 (vect_get_gather_scatter_ops): New function.
9104 (vectorizable_call): Check internal_load_fn_p.
9105 (vectorizable_load): Likewise. Handle gather load internal
9107 (vectorizable_store): Update call to check_load_store_masking.
9108 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9109 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9110 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9111 (aarch64_gather_scale_operand_d): New predicates.
9112 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9113 (mask_gather_load<mode>): New insns.
9115 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9116 Alan Hayward <alan.hayward@arm.com>
9117 David Sherwood <david.sherwood@arm.com>
9119 * optabs.def (fold_left_plus_optab): New optab.
9120 * doc/md.texi (fold_left_plus_@var{m}): Document.
9121 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9122 * internal-fn.c (fold_left_direct): Define.
9123 (expand_fold_left_optab_fn): Likewise.
9124 (direct_fold_left_optab_supported_p): Likewise.
9125 * fold-const-call.c (fold_const_fold_left): New function.
9126 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9127 * tree-parloops.c (valid_reduction_p): New function.
9128 (gather_scalar_reductions): Use it.
9129 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9130 (vect_finish_replace_stmt): Declare.
9131 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9132 (needs_fold_left_reduction_p): New function, split out from...
9133 (vect_is_simple_reduction): ...here. Accept reductions that
9134 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9135 (vect_force_simple_reduction): Also store the reduction type in
9136 the assignment's STMT_VINFO_REDUC_TYPE.
9137 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9138 (merge_with_identity): New function.
9139 (vect_expand_fold_left): Likewise.
9140 (vectorize_fold_left_reduction): Likewise.
9141 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9142 scalar phi in place for it. Check for target support and reject
9143 cases that would reassociate the operation. Defer the transform
9144 phase to vectorize_fold_left_reduction.
9145 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9146 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9147 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9151 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9152 call to ifc_temp_var.
9154 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9155 Alan Hayward <alan.hayward@arm.com>
9156 David Sherwood <david.sherwood@arm.com>
9158 * target.def (legitimize_address_displacement): Take the original
9159 offset as a poly_int.
9160 * targhooks.h (default_legitimize_address_displacement): Update
9162 * targhooks.c (default_legitimize_address_displacement): Likewise.
9163 * doc/tm.texi: Regenerate.
9164 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9165 as an argument, moving assert of ad->disp == ad->disp_term to...
9166 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9167 Try calling targetm.legitimize_address_displacement before expanding
9168 the address rather than afterwards, and adjust for the new interface.
9169 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9170 Match the new hook interface. Handle SVE addresses.
9171 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9174 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9176 * Makefile.in (OBJS): Add early-remat.o.
9177 * target.def (select_early_remat_modes): New hook.
9178 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9179 * doc/tm.texi: Regenerate.
9180 * targhooks.h (default_select_early_remat_modes): Declare.
9181 * targhooks.c (default_select_early_remat_modes): New function.
9182 * timevar.def (TV_EARLY_REMAT): New timevar.
9183 * passes.def (pass_early_remat): New pass.
9184 * tree-pass.h (make_pass_early_remat): Declare.
9185 * early-remat.c: New file.
9186 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9188 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9191 Alan Hayward <alan.hayward@arm.com>
9192 David Sherwood <david.sherwood@arm.com>
9194 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9195 vfm1 with a bound_epilog parameter.
9196 (vect_do_peeling): Update calls accordingly, and move the prologue
9197 call earlier in the function. Treat the base bound_epilog as 0 for
9198 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9199 this base when peeling for gaps.
9200 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9201 with fully-masked loops.
9202 (vect_estimate_min_profitable_iters): Handle the single peeled
9203 iteration in that case.
9205 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9206 Alan Hayward <alan.hayward@arm.com>
9207 David Sherwood <david.sherwood@arm.com>
9209 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9210 single-element interleaving even if the size is not a power of 2.
9211 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9212 accesses for single-element interleaving if the group size is
9215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9216 Alan Hayward <alan.hayward@arm.com>
9217 David Sherwood <david.sherwood@arm.com>
9219 * doc/md.texi (fold_extract_last_@var{m}): Document.
9220 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9221 * optabs.def (fold_extract_last_optab): New optab.
9222 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9223 * internal-fn.c (fold_extract_direct): New macro.
9224 (expand_fold_extract_optab_fn): Likewise.
9225 (direct_fold_extract_optab_supported_p): Likewise.
9226 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9227 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9228 EXTRACT_LAST_REDUCTION.
9229 (get_initial_def_for_reduction): Do not create an initial vector
9230 for EXTRACT_LAST_REDUCTION reductions.
9231 (vectorizable_reduction): Leave the scalar phi in place for
9232 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9233 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9234 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9235 transform phase to vectorizable_condition.
9236 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9238 (vect_finish_stmt_generation): ...here.
9239 (vect_finish_replace_stmt): New function.
9240 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9241 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9243 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9246 Alan Hayward <alan.hayward@arm.com>
9247 David Sherwood <david.sherwood@arm.com>
9249 * doc/md.texi (extract_last_@var{m}): Document.
9250 * optabs.def (extract_last_optab): New optab.
9251 * internal-fn.def (EXTRACT_LAST): New internal function.
9252 * internal-fn.c (cond_unary_direct): New macro.
9253 (expand_cond_unary_optab_fn): Likewise.
9254 (direct_cond_unary_optab_supported_p): Likewise.
9255 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9256 loops using EXTRACT_LAST.
9257 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9258 (extract_last_<mode>): ...this optab.
9259 (vec_extract<mode><Vel>): Update accordingly.
9261 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9262 Alan Hayward <alan.hayward@arm.com>
9263 David Sherwood <david.sherwood@arm.com>
9265 * target.def (empty_mask_is_expensive): New hook.
9266 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9267 * doc/tm.texi: Regenerate.
9268 * targhooks.h (default_empty_mask_is_expensive): Declare.
9269 * targhooks.c (default_empty_mask_is_expensive): New function.
9270 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9271 if the target says that empty masks are expensive.
9272 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9274 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9276 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9277 Alan Hayward <alan.hayward@arm.com>
9278 David Sherwood <david.sherwood@arm.com>
9280 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9281 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9282 (vect_use_loop_mask_for_alignment_p): New function.
9283 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9284 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9285 niters_skip argument. Make sure that the first niters_skip elements
9286 of the first iteration are inactive.
9287 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9288 Update call to vect_set_loop_masks_directly.
9289 (get_misalign_in_elems): New function, split out from...
9290 (vect_gen_prolog_loop_niters): ...here.
9291 (vect_update_init_of_dr): Take a code argument that specifies whether
9292 the adjustment should be added or subtracted.
9293 (vect_update_init_of_drs): Likewise.
9294 (vect_prepare_for_masked_peels): New function.
9295 (vect_do_peeling): Skip prologue peeling if we're using a mask
9296 instead. Update call to vect_update_inits_of_drs.
9297 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9299 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9300 alignment. Do not include the number of peeled iterations in
9301 the minimum threshold in that case.
9302 (vectorizable_induction): Adjust the start value down by
9303 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9304 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9305 Take the number of skipped iterations into account when calculating
9307 * tree-vect-stmts.c (vect_gen_while_not): New function.
9309 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9310 Alan Hayward <alan.hayward@arm.com>
9311 David Sherwood <david.sherwood@arm.com>
9313 * doc/sourcebuild.texi (vect_fully_masked): Document.
9314 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9316 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9318 (vect_analyze_loop_2): ...here. Don't check the vectorization
9319 factor against the number of loop iterations if the loop is
9322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9323 Alan Hayward <alan.hayward@arm.com>
9324 David Sherwood <david.sherwood@arm.com>
9326 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9327 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9328 (dump_groups): Update accordingly.
9329 (iv_use::mem_type): New member variable.
9330 (address_p): New function.
9331 (record_use): Add a mem_type argument and initialize the new
9333 (record_group_use): Add a mem_type argument. Use address_p.
9334 Remove obsolete null checks of base_object. Update call to record_use.
9335 (find_interesting_uses_op): Update call to record_group_use.
9336 (find_interesting_uses_cond): Likewise.
9337 (find_interesting_uses_address): Likewise.
9338 (get_mem_type_for_internal_fn): New function.
9339 (find_address_like_use): Likewise.
9340 (find_interesting_uses_stmt): Try find_address_like_use before
9341 calling find_interesting_uses_op.
9342 (addr_offset_valid_p): Use the iv mem_type field as the type
9343 of the addressed memory.
9344 (add_autoinc_candidates): Likewise.
9345 (get_address_cost): Likewise.
9346 (split_small_address_groups_p): Use address_p.
9347 (split_address_groups): Likewise.
9348 (add_iv_candidate_for_use): Likewise.
9349 (autoinc_possible_for_pair): Likewise.
9350 (rewrite_groups): Likewise.
9351 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9352 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9353 (get_alias_ptr_type_for_ptr_address): New function.
9354 (rewrite_use_address): Rewrite address uses in calls that were
9355 identified by find_address_like_use.
9357 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9358 Alan Hayward <alan.hayward@arm.com>
9359 David Sherwood <david.sherwood@arm.com>
9361 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9363 * gimple-expr.h (is_gimple_addressable: Likewise.
9364 * gimple-expr.c (is_gimple_address): Likewise.
9365 * internal-fn.c (expand_call_mem_ref): New function.
9366 (expand_mask_load_optab_fn): Use it.
9367 (expand_mask_store_optab_fn): Likewise.
9369 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9370 Alan Hayward <alan.hayward@arm.com>
9371 David Sherwood <david.sherwood@arm.com>
9373 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9374 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9375 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9376 (cond_umax@var{mode}): Document.
9377 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9378 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9379 (cond_umin_optab, cond_umax_optab): New optabs.
9380 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9381 (COND_IOR, COND_XOR): New internal functions.
9382 * internal-fn.h (get_conditional_internal_fn): Declare.
9383 * internal-fn.c (cond_binary_direct): New macro.
9384 (expand_cond_binary_optab_fn): Likewise.
9385 (direct_cond_binary_optab_supported_p): Likewise.
9386 (get_conditional_internal_fn): New function.
9387 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9388 Cope with reduction statements that are vectorized as calls rather
9390 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9391 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9392 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9393 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9394 (UNSPEC_COND_EOR): New unspecs.
9395 (optab): Add mappings for them.
9396 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9397 (sve_int_op, sve_fp_op): New int attributes.
9399 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9400 Alan Hayward <alan.hayward@arm.com>
9401 David Sherwood <david.sherwood@arm.com>
9403 * optabs.def (while_ult_optab): New optab.
9404 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9405 * internal-fn.def (WHILE_ULT): New internal function.
9406 * internal-fn.h (direct_internal_fn_supported_p): New override
9407 that takes two types as argument.
9408 * internal-fn.c (while_direct): New macro.
9409 (expand_while_optab_fn): New function.
9410 (convert_optab_supported_p): Likewise.
9411 (direct_while_optab_supported_p): New macro.
9412 * wide-int.h (wi::udiv_ceil): New function.
9413 * tree-vectorizer.h (rgroup_masks): New structure.
9414 (vec_loop_masks): New typedef.
9415 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9417 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9418 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9419 (vect_max_vf): New function.
9420 (slpeel_make_loop_iterate_ntimes): Delete.
9421 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9422 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9423 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9424 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9425 internal-fn.h, stor-layout.h and optabs-query.h.
9426 (vect_set_loop_mask): New function.
9427 (add_preheader_seq): Likewise.
9428 (add_header_seq): Likewise.
9429 (interleave_supported_p): Likewise.
9430 (vect_maybe_permute_loop_masks): Likewise.
9431 (vect_set_loop_masks_directly): Likewise.
9432 (vect_set_loop_condition_masked): Likewise.
9433 (vect_set_loop_condition_unmasked): New function, split out from
9434 slpeel_make_loop_iterate_ntimes.
9435 (slpeel_make_loop_iterate_ntimes): Rename to..
9436 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9437 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9438 (vect_do_peeling): Update call accordingly.
9439 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9441 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9442 mask_compare_type, can_fully_mask_p and fully_masked_p.
9443 (release_vec_loop_masks): New function.
9444 (_loop_vec_info): Use it to free the loop masks.
9445 (can_produce_all_loop_masks_p): New function.
9446 (vect_get_max_nscalars_per_iter): Likewise.
9447 (vect_verify_full_masking): Likewise.
9448 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9449 retries, and free the mask rgroups before retrying. Check loop-wide
9450 reasons for disallowing fully-masked loops. Make the final decision
9451 about whether use a fully-masked loop or not.
9452 (vect_estimate_min_profitable_iters): Do not assume that peeling
9453 for the number of iterations will be needed for fully-masked loops.
9454 (vectorizable_reduction): Disable fully-masked loops.
9455 (vectorizable_live_operation): Likewise.
9456 (vect_halve_mask_nunits): New function.
9457 (vect_double_mask_nunits): Likewise.
9458 (vect_record_loop_mask): Likewise.
9459 (vect_get_loop_mask): Likewise.
9460 (vect_transform_loop): Handle the case in which the final loop
9461 iteration might handle a partial vector. Call vect_set_loop_condition
9462 instead of slpeel_make_loop_iterate_ntimes.
9463 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9464 (check_load_store_masking): New function.
9465 (prepare_load_store_mask): Likewise.
9466 (vectorizable_store): Handle fully-masked loops.
9467 (vectorizable_load): Likewise.
9468 (supportable_widening_operation): Use vect_halve_mask_nunits for
9470 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9471 (vect_gen_while): New function.
9472 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9473 (aarch64_uqdec<mode>): New insn.
9475 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9476 Alan Hayward <alan.hayward@arm.com>
9477 David Sherwood <david.sherwood@arm.com>
9479 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9480 (reduc_xor_scal_optab): New optabs.
9481 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9482 (reduc_xor_scal_@var{m}): Document.
9483 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9484 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9486 * fold-const-call.c (fold_const_call): Handle them.
9487 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9488 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9489 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9490 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9491 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9492 (UNSPEC_XORV): New unspecs.
9493 (optab): Add entries for them.
9494 (BITWISEV): New int iterator.
9495 (bit_reduc_op): New int attributes.
9497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9498 Alan Hayward <alan.hayward@arm.com>
9499 David Sherwood <david.sherwood@arm.com>
9501 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9502 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9503 * optabs.def (vec_shl_insert_optab): New optab.
9504 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9505 (duplicate_and_interleave): Likewise.
9506 * tree-vect-loop.c: Include internal-fn.h.
9507 (neutral_op_for_slp_reduction): New function, split out from
9508 get_initial_defs_for_reduction.
9509 (get_initial_def_for_reduction): Handle option 2 for variable-length
9510 vectors by loading the neutral value into a vector and then shifting
9511 the initial value into element 0.
9512 (get_initial_defs_for_reduction): Replace the code argument with
9513 the neutral value calculated by neutral_op_for_slp_reduction.
9514 Use gimple_build_vector for constant-length vectors.
9515 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9516 but the first group_size elements have a neutral value.
9517 Use duplicate_and_interleave otherwise.
9518 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9519 Update call to get_initial_defs_for_reduction. Handle SLP
9520 reductions for variable-length vectors by creating one vector
9521 result for each scalar result, with the elements associated
9522 with other scalar results stubbed out with the neutral value.
9523 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9524 Require IFN_VEC_SHL_INSERT for double reductions on
9525 variable-length vectors, or SLP reductions that have
9526 a neutral value. Require can_duplicate_and_interleave_p
9527 support for variable-length unchained SLP reductions if there
9528 is no neutral value, such as for MIN/MAX reductions. Also require
9529 the number of vector elements to be a multiple of the number of
9530 SLP statements when doing variable-length unchained SLP reductions.
9531 Update call to vect_create_epilog_for_reduction.
9532 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9533 and remove initial values.
9534 (duplicate_and_interleave): Make public.
9535 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9536 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9538 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9539 Alan Hayward <alan.hayward@arm.com>
9540 David Sherwood <david.sherwood@arm.com>
9542 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9543 (can_duplicate_and_interleave_p): New function.
9544 (vect_get_and_check_slp_defs): Take the vector of statements
9545 rather than just the current one. Remove excess parentheses.
9546 Restriction rejectinon of vect_constant_def and vect_external_def
9547 for variable-length vectors to boolean types, or types for which
9548 can_duplicate_and_interleave_p is false.
9549 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9550 (duplicate_and_interleave): New function.
9551 (vect_get_constant_vectors): Use gimple_build_vector for
9552 constant-length vectors and suitable variable-length constant
9553 vectors. Use duplicate_and_interleave for other variable-length
9554 vectors. Don't defer the update when inserting new statements.
9556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9557 Alan Hayward <alan.hayward@arm.com>
9558 David Sherwood <david.sherwood@arm.com>
9560 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9561 min_profitable_iters doesn't go negative.
9563 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9564 Alan Hayward <alan.hayward@arm.com>
9565 David Sherwood <david.sherwood@arm.com>
9567 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9568 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9569 * optabs.def (vec_mask_load_lanes_optab): New optab.
9570 (vec_mask_store_lanes_optab): Likewise.
9571 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9572 (MASK_STORE_LANES): Likewise.
9573 * internal-fn.c (mask_load_lanes_direct): New macro.
9574 (mask_store_lanes_direct): Likewise.
9575 (expand_mask_load_optab_fn): Handle masked operations.
9576 (expand_mask_load_lanes_optab_fn): New macro.
9577 (expand_mask_store_optab_fn): Handle masked operations.
9578 (expand_mask_store_lanes_optab_fn): New macro.
9579 (direct_mask_load_lanes_optab_supported_p): Likewise.
9580 (direct_mask_store_lanes_optab_supported_p): Likewise.
9581 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9583 (vect_load_lanes_supported): Likewise.
9584 * tree-vect-data-refs.c (strip_conversion): New function.
9585 (can_group_stmts_p): Likewise.
9586 (vect_analyze_data_ref_accesses): Use it instead of checking
9587 for a pair of assignments.
9588 (vect_store_lanes_supported): Take a masked_p parameter.
9589 (vect_load_lanes_supported): Likewise.
9590 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9591 vect_store_lanes_supported and vect_load_lanes_supported.
9592 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9593 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9594 parameter. Don't allow gaps for masked accesses.
9595 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9596 and vect_load_lanes_supported.
9597 (get_load_store_type): Take a masked_p parameter and update
9598 call to get_group_load_store_type.
9599 (vectorizable_store): Update call to get_load_store_type.
9600 Handle IFN_MASK_STORE_LANES.
9601 (vectorizable_load): Update call to get_load_store_type.
9602 Handle IFN_MASK_LOAD_LANES.
9604 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9605 Alan Hayward <alan.hayward@arm.com>
9606 David Sherwood <david.sherwood@arm.com>
9608 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9610 * config/aarch64/aarch64-protos.h
9611 (aarch64_sve_struct_memory_operand_p): Declare.
9612 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9613 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9614 (VPRED, vpred): Handle SVE structure modes.
9615 * config/aarch64/constraints.md (Utx): New constraint.
9616 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9617 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9618 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9619 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9620 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9621 structure modes. Split into pieces after RA.
9622 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9623 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9625 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9626 SVE structure modes.
9627 (aarch64_classify_address): Likewise.
9628 (sizetochar): Move earlier in file.
9629 (aarch64_print_operand): Handle SVE register lists.
9630 (aarch64_array_mode): New function.
9631 (aarch64_sve_struct_memory_operand_p): Likewise.
9632 (TARGET_ARRAY_MODE): Redefine.
9634 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9635 Alan Hayward <alan.hayward@arm.com>
9636 David Sherwood <david.sherwood@arm.com>
9638 * target.def (array_mode): New target hook.
9639 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9640 * doc/tm.texi: Regenerate.
9641 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9642 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9643 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9645 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9648 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9649 Alan Hayward <alan.hayward@arm.com>
9650 David Sherwood <david.sherwood@arm.com>
9652 * fold-const.c (fold_binary_loc): Check the argument types
9653 rather than the result type when testing for a vector operation.
9655 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9657 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9658 * doc/tm.texi: Regenerate.
9660 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9661 Alan Hayward <alan.hayward@arm.com>
9662 David Sherwood <david.sherwood@arm.com>
9664 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9665 (sve): Document new AArch64 extension.
9666 * doc/md.texi (w): Extend the description of the AArch64
9667 constraint to include SVE vectors.
9668 (Upl, Upa): Document new AArch64 predicate constraints.
9669 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9671 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9672 (msve-vector-bits=): New option.
9673 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9674 SVE when these are disabled.
9675 (sve): New extension.
9676 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9677 modes. Adjust their number of units based on aarch64_sve_vg.
9678 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9679 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9680 aarch64_addr_query_type.
9681 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9682 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9683 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9684 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9685 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9686 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9687 (aarch64_simd_imm_zero_p): Delete.
9688 (aarch64_check_zero_based_sve_index_immediate): Declare.
9689 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9690 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9691 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9692 (aarch64_sve_float_mul_immediate_p): Likewise.
9693 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9695 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9696 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9697 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9698 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9699 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9700 (aarch64_regmode_natural_size): Likewise.
9701 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9702 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9704 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9705 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9706 for VG and the SVE predicate registers.
9707 (V_ALIASES): Add a "z"-prefixed alias.
9708 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9709 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9710 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9711 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9712 (REG_CLASS_NAMES): Add entries for them.
9713 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9714 and the predicate registers.
9715 (aarch64_sve_vg): Declare.
9716 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9717 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9718 (REGMODE_NATURAL_SIZE): Define.
9719 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9721 * config/aarch64/aarch64.c: Include cfgrtl.h.
9722 (simd_immediate_info): Add a constructor for series vectors,
9723 and an associated step field.
9724 (aarch64_sve_vg): New variable.
9725 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9726 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9727 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9728 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9729 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9730 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9731 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9732 (aarch64_get_mask_mode): New functions.
9733 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9734 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9735 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9736 predicate modes and predicate registers. Explicitly restrict
9737 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9738 to store a vector mode if it is recognized by
9739 aarch64_classify_vector_mode.
9740 (aarch64_regmode_natural_size): New function.
9741 (aarch64_hard_regno_caller_save_mode): Return the original mode
9743 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9744 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9745 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9746 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9748 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9749 does not overlap dest if the function is frame-related. Handle
9751 (aarch64_split_add_offset): New function.
9752 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9753 them aarch64_add_offset.
9754 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9755 and update call to aarch64_sub_sp.
9756 (aarch64_add_cfa_expression): New function.
9757 (aarch64_expand_prologue): Pass extra temporary registers to the
9758 functions above. Handle the case in which we need to emit new
9759 DW_CFA_expressions for registers that were originally saved
9760 relative to the stack pointer, but now have to be expressed
9761 relative to the frame pointer.
9762 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9764 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9765 IP0 and IP1 values for SVE frames.
9766 (aarch64_expand_vec_series): New function.
9767 (aarch64_expand_sve_widened_duplicate): Likewise.
9768 (aarch64_expand_sve_const_vector): Likewise.
9769 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9770 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9771 into the register, rather than emitting a SET directly.
9772 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9773 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9774 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9775 (offset_9bit_signed_scaled_p): New functions.
9776 (aarch64_replicate_bitmask_imm): New function.
9777 (aarch64_bitmask_imm): Use it.
9778 (aarch64_cannot_force_const_mem): Reject expressions involving
9779 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9780 (aarch64_classify_index): Handle SVE indices, by requiring
9781 a plain register index with a scale that matches the element size.
9782 (aarch64_classify_address): Handle SVE addresses. Assert that
9783 the mode of the address is VOIDmode or an integer mode.
9784 Update call to aarch64_classify_symbol.
9785 (aarch64_classify_symbolic_expression): Update call to
9786 aarch64_classify_symbol.
9787 (aarch64_const_vec_all_in_range_p): New function.
9788 (aarch64_print_vector_float_operand): Likewise.
9789 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9790 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9791 and the FP immediates 1.0 and 0.5.
9792 (aarch64_print_address_internal): Handle SVE addresses.
9793 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9794 (aarch64_regno_regclass): Handle predicate registers.
9795 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9797 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9798 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9799 (aarch64_convert_sve_vector_bits): New function.
9800 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9801 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9803 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9804 Handle SVE vector and predicate modes. Accept VL-based constants
9805 that need only one temporary register, and VL offsets that require
9806 no temporary registers.
9807 (aarch64_conditional_register_usage): Mark the predicate registers
9808 as fixed if SVE isn't available.
9809 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9810 Return true for SVE vector and predicate modes.
9811 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9812 rather than an unsigned int. Handle SVE modes.
9813 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9815 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9817 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9818 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9819 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9820 (aarch64_sve_float_mul_immediate_p): New functions.
9821 (aarch64_sve_valid_immediate): New function.
9822 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9823 Explicitly reject structure modes. Check for INDEX constants.
9824 Handle PTRUE and PFALSE constants.
9825 (aarch64_check_zero_based_sve_index_immediate): New function.
9826 (aarch64_simd_imm_zero_p): Delete.
9827 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9828 vector modes. Accept constants in the range of CNT[BHWD].
9829 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9830 ask for an Advanced SIMD mode.
9831 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9832 (aarch64_simd_vector_alignment): Handle SVE predicates.
9833 (aarch64_vectorize_preferred_vector_alignment): New function.
9834 (aarch64_simd_vector_alignment_reachable): Use it instead of
9836 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9837 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9839 (MAX_VECT_LEN): Delete.
9840 (expand_vec_perm_d): Add a vec_flags field.
9841 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9842 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9843 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9845 (aarch64_evpc_rev): Rename to...
9846 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9847 (aarch64_evpc_rev_global): New function.
9848 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9849 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9851 (aarch64_evpc_sve_tbl): New function.
9852 (aarch64_expand_vec_perm_const_1): Update after rename of
9853 aarch64_evpc_rev. Handle SVE permutes too, trying
9854 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9855 than aarch64_evpc_tbl.
9856 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9857 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9858 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9859 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9860 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9861 (aarch64_expand_sve_vcond): New functions.
9862 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9863 of aarch64_vector_mode_p.
9864 (aarch64_dwarf_poly_indeterminate_value): New function.
9865 (aarch64_compute_pressure_classes): Likewise.
9866 (aarch64_can_change_mode_class): Likewise.
9867 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9868 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9869 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9870 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9871 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9872 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9873 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9874 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9876 (Dn, Dl, Dr): Accept const as well as const_vector.
9877 (Dz): Likewise. Compare against CONST0_RTX.
9878 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9879 of "vector" where appropriate.
9880 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9881 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9882 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9883 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9884 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9885 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9886 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9887 (v_int_equiv): Extend to SVE modes.
9888 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9890 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9891 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9892 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9893 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9894 (SVE_COND_FP_CMP): New int iterators.
9895 (perm_hilo): Handle the new unpack unspecs.
9896 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9898 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9899 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9900 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9901 (aarch64_equality_operator, aarch64_constant_vector_operand)
9902 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9903 (aarch64_sve_nonimmediate_operand): Likewise.
9904 (aarch64_sve_general_operand): Likewise.
9905 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9906 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9907 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9908 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9909 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9910 (aarch64_sve_float_arith_immediate): Likewise.
9911 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9912 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9913 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9914 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9915 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9916 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9917 (aarch64_sve_float_arith_operand): Likewise.
9918 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9919 (aarch64_sve_float_mul_operand): Likewise.
9920 (aarch64_sve_vec_perm_operand): Likewise.
9921 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9922 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9923 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9924 as well as const_vector.
9925 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9926 in file. Use CONST0_RTX and CONSTM1_RTX.
9927 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9928 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9929 Use aarch64_simd_imm_zero.
9930 * config/aarch64/aarch64-sve.md: New file.
9931 * config/aarch64/aarch64.md: Include it.
9932 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9933 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9934 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9935 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9936 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9937 (sve): New attribute.
9938 (enabled): Disable instructions with the sve attribute unless
9940 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9941 aarch64_expand_mov_immediate.
9942 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9943 CNT[BHSD] immediates.
9944 (movti): Split CONST_POLY_INT moves into two halves.
9945 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9946 Split additions that need a temporary here if the destination
9947 is the stack pointer.
9948 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9949 (*add<mode>3_poly_1): New instruction.
9950 (set_clobber_cc): New expander.
9952 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9954 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9955 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9956 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9957 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9958 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9959 Change innermode from fixed_mode_size to machine_mode.
9960 (simplify_subreg): Update call accordingly. Handle a constant-sized
9961 subreg of a variable-length CONST_VECTOR.
9963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9964 Alan Hayward <alan.hayward@arm.com>
9965 David Sherwood <david.sherwood@arm.com>
9967 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9968 (add_offset_to_base): New function, split out from...
9969 (create_mem_ref): ...here. When handling a scale other than 1,
9970 check first whether the address is valid without the offset.
9971 Add it into the base if so, leaving the index and scale as-is.
9973 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9976 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9977 fold_for_warn before checking if arg2 is INTEGER_CST.
9979 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9981 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9982 (store_multiple_operation): Delete.
9983 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9984 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9985 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9986 guarded by TARGET_STRING.
9987 (rs6000_output_load_multiple): Delete.
9988 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9989 OPTION_MASK_STRING / TARGET_STRING handling.
9990 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9991 (const rs6000_opt_masks) <"string">: Change mask to 0.
9992 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9993 (MASK_STRING): Delete.
9994 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9996 (load_multiple): Delete.
10003 (store_multiple): Delete.
10010 (movmemsi_8reg): Delete.
10011 (corresponding unnamed define_insn): Delete.
10012 (movmemsi_6reg): Delete.
10013 (corresponding unnamed define_insn): Delete.
10014 (movmemsi_4reg): Delete.
10015 (corresponding unnamed define_insn): Delete.
10016 (movmemsi_2reg): Delete.
10017 (corresponding unnamed define_insn): Delete.
10018 (movmemsi_1reg): Delete.
10019 (corresponding unnamed define_insn): Delete.
10020 * config/rs6000/rs6000.opt (mno-string): New.
10021 (mstring): Replace by deprecation warning stub.
10022 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10024 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10026 * regrename.c (regrename_do_replace): If replacing the same
10027 reg multiple times, try to reuse last created gen_raw_REG.
10030 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10031 main to workaround a bug in GDB.
10033 2018-01-12 Tom de Vries <tom@codesourcery.com>
10036 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10038 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10040 PR rtl-optimization/80481
10041 * ira-color.c (get_cap_member): New function.
10042 (allocnos_conflict_by_live_ranges_p): Use it.
10043 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10044 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10046 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10049 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10050 (*saddl_se_1): Ditto.
10051 (*ssubsi_1): Ditto.
10052 (*ssubl_se_1): Ditto.
10054 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10056 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10057 rather than wi::to_widest for DR_INITs.
10058 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10059 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10060 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10062 (vect_analyze_group_access_1): Note that here.
10064 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10066 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10067 polynomial type sizes.
10069 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10071 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10072 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10073 (gimple_add_tmp_var): Likewise.
10075 2018-01-12 Martin Liska <mliska@suse.cz>
10077 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10078 (gimple_alloc_sizes): Likewise.
10079 (dump_gimple_statistics): Use PRIu64 in printf format.
10080 * gimple.h: Change uint64_t to int.
10082 2018-01-12 Martin Liska <mliska@suse.cz>
10084 * tree-core.h: Use uint64_t instead of int.
10085 * tree.c (tree_node_counts): Likewise.
10086 (tree_node_sizes): Likewise.
10087 (dump_tree_statistics): Use PRIu64 in printf format.
10089 2018-01-12 Martin Liska <mliska@suse.cz>
10091 * Makefile.in: As qsort_chk is implemented in vec.c, add
10092 vec.o to linkage of gencfn-macros.
10093 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10094 passing the info to record_node_allocation_statistics.
10095 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10097 * ggc-common.c (struct ggc_usage): Add operator== and use
10098 it in operator< and compare function.
10099 * mem-stats.h (struct mem_usage): Likewise.
10100 * vec.c (struct vec_usage): Remove operator< and compare
10101 function. Can be simply inherited.
10103 2018-01-12 Martin Jambor <mjambor@suse.cz>
10106 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10107 * tree-ssa-math-opts.c: Include domwalk.h.
10108 (convert_mult_to_fma_1): New function.
10109 (fma_transformation_info): New type.
10110 (fma_deferring_state): Likewise.
10111 (cancel_fma_deferring): New function.
10112 (result_of_phi): Likewise.
10113 (last_fma_candidate_feeds_initial_phi): Likewise.
10114 (convert_mult_to_fma): Added deferring logic, split actual
10115 transformation to convert_mult_to_fma_1.
10116 (math_opts_dom_walker): New type.
10117 (math_opts_dom_walker::after_dom_children): New method, body moved
10118 here from pass_optimize_widening_mul::execute, added deferring logic
10120 (pass_optimize_widening_mul::execute): Moved most of code to
10121 math_opts_dom_walker::after_dom_children.
10122 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10123 * config/i386/i386.c (ix86_option_override_internal): Added
10124 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10126 2018-01-12 Richard Biener <rguenther@suse.de>
10129 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10130 inline instance vars.
10132 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10135 * config/rx/rx.c (rx_is_restricted_memory_address):
10136 Handle SUBREG case.
10138 2018-01-12 Richard Biener <rguenther@suse.de>
10140 PR tree-optimization/80846
10141 * target.def (split_reduction): New target hook.
10142 * targhooks.c (default_split_reduction): New function.
10143 * targhooks.h (default_split_reduction): Declare.
10144 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10145 target requests first reduce vectors by combining low and high
10147 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10148 (get_vectype_for_scalar_type_and_size): Export.
10149 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10150 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10151 * doc/tm.texi: Regenerate.
10152 * config/i386/i386.c (ix86_split_reduction): Implement
10153 TARGET_VECTORIZE_SPLIT_REDUCTION.
10155 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10158 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10159 in PIC mode except for TARGET_VXWORKS_RTP.
10160 * config/sparc/sparc.c: Include cfgrtl.h.
10161 (TARGET_INIT_PIC_REG): Define.
10162 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10163 (sparc_pic_register_p): New predicate.
10164 (sparc_legitimate_address_p): Use it.
10165 (sparc_legitimize_pic_address): Likewise.
10166 (sparc_delegitimize_address): Likewise.
10167 (sparc_mode_dependent_address_p): Likewise.
10168 (gen_load_pcrel_sym): Remove 4th parameter.
10169 (load_got_register): Adjust call to above. Remove obsolete stuff.
10170 (sparc_expand_prologue): Do not call load_got_register here.
10171 (sparc_flat_expand_prologue): Likewise.
10172 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10173 (sparc_use_pseudo_pic_reg): New function.
10174 (sparc_init_pic_reg): Likewise.
10175 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10176 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10178 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10180 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10181 Add item for branch_cost.
10183 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10185 PR rtl-optimization/83565
10186 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10187 not extend the result to a larger mode for rotate operations.
10188 (num_sign_bit_copies1): Likewise.
10190 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10193 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10195 Use values-Xc.o for -pedantic.
10196 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10198 2018-01-12 Martin Liska <mliska@suse.cz>
10201 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10203 (possible_polymorphic_call_targets): Use it.
10204 (ipa_devirt): Likewise.
10206 2018-01-12 Martin Liska <mliska@suse.cz>
10208 * profile-count.h (enum profile_quality): Use 0 as invalid
10209 enum value of profile_quality.
10211 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10213 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10214 -mext-string options.
10216 2018-01-12 Richard Biener <rguenther@suse.de>
10218 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10219 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10220 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10222 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10224 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10226 * configure.ac (--with-long-double-format): Add support for the
10227 configuration option to change the default long double format on
10229 * config.gcc (powerpc*-linux*-*): Likewise.
10230 * configure: Regenerate.
10231 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10232 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10233 used without modification.
10235 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10237 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10238 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10239 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10240 MISC_BUILTIN_SPEC_BARRIER.
10241 (rs6000_init_builtins): Likewise.
10242 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10244 (speculation_barrier): New define_insn.
10245 * doc/extend.texi: Document __builtin_speculation_barrier.
10247 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10250 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10251 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10252 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10254 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10255 integral modes instead of "ss" and "sd".
10256 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10257 vectors with 32-bit and 64-bit elements.
10258 (vecdupssescalarmodesuffix): New mode attribute.
10259 (vec_dup<mode>): Use it.
10261 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10264 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10265 frame if argument is passed on stack.
10267 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10270 * ree.c (combine_reaching_defs): Optimize also
10271 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10272 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10274 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10276 PR middle-end/83189
10277 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10279 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10281 PR middle-end/83718
10282 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10283 after they are computed.
10285 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10287 PR tree-optimization/83695
10288 * gimple-loop-linterchange.cc
10289 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10290 reset cached scev information after interchange.
10291 (pass_linterchange::execute): Remove call to scev_reset_htab.
10293 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10295 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10296 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10297 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10298 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10299 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10300 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10301 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10302 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10303 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10304 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10305 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10306 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10307 (V_lane_reg): Likewise.
10308 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10310 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10311 (vfmal_lane_low<mode>_intrinsic,
10312 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10313 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10314 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10315 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10316 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10317 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10319 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10321 * config/arm/arm-cpus.in (fp16fml): New feature.
10322 (ALL_SIMD): Add fp16fml.
10323 (armv8.2-a): Add fp16fml as an option.
10324 (armv8.3-a): Likewise.
10325 (armv8.4-a): Add fp16fml as part of fp16.
10326 * config/arm/arm.h (TARGET_FP16FML): Define.
10327 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10329 * config/arm/arm-modes.def (V2HF): Define.
10330 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10331 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10332 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10333 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10334 vfmsl_low, vfmsl_high): New set of builtins.
10335 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10336 (vfml_op): New code attribute.
10337 (VFMLHALVES): New int iterator.
10338 (VFML, VFMLSEL): New mode attributes.
10339 (V_reg): Define mapping for V2HF.
10340 (V_hi, V_lo): New mode attributes.
10341 (VF_constraint): Likewise.
10342 (vfml_half, vfml_half_selector): New int attributes.
10343 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10345 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10346 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10348 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10349 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10350 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10351 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10353 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10354 Document new effective target and option set.
10356 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10358 * config/arm/arm-cpus.in (armv8_4): New feature.
10359 (ARMv8_4a): New fgroup.
10360 (armv8.4-a): New arch.
10361 * config/arm/arm-tables.opt: Regenerate.
10362 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10363 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10364 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10365 Add matching rules for -march=armv8.4-a and extensions.
10366 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10368 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10371 * config/rx/rx.md (BW): New mode attribute.
10372 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10374 2018-01-11 Richard Biener <rguenther@suse.de>
10376 PR tree-optimization/83435
10377 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10378 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10379 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10381 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10382 Alan Hayward <alan.hayward@arm.com>
10383 David Sherwood <david.sherwood@arm.com>
10385 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10387 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10388 (aarch64_print_address_internal): Use it to check for a zero offset.
10390 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10391 Alan Hayward <alan.hayward@arm.com>
10392 David Sherwood <david.sherwood@arm.com>
10394 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10395 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10396 Return a poly_int64 rather than a HOST_WIDE_INT.
10397 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10398 rather than a HOST_WIDE_INT.
10399 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10400 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10401 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10402 final_offset from HOST_WIDE_INT to poly_int64.
10403 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10404 to_constant when getting the number of units in an Advanced SIMD
10406 (aarch64_builtin_vectorized_function): Check for a constant number
10408 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10410 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10411 attribute instead of GET_MODE_NUNITS.
10412 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10413 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10414 GET_MODE_SIZE for fixed-size registers.
10415 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10416 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10417 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10418 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10419 (aarch64_print_operand, aarch64_print_address_internal)
10420 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10421 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10422 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10423 Handle polynomial GET_MODE_SIZE.
10424 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10425 wider than SImode without modification.
10426 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10427 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10428 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10429 passing and returning SVE modes.
10430 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10431 rather than GEN_INT.
10432 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10433 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10434 (aarch64_allocate_and_probe_stack_space): Likewise.
10435 (aarch64_layout_frame): Cope with polynomial offsets.
10436 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10437 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10438 polynomial offsets.
10439 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10440 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10441 poly_int64 rather than a HOST_WIDE_INT.
10442 (aarch64_get_separate_components, aarch64_process_components)
10443 (aarch64_expand_prologue, aarch64_expand_epilogue)
10444 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10445 (aarch64_anchor_offset): New function, split out from...
10446 (aarch64_legitimize_address): ...here.
10447 (aarch64_builtin_vectorization_cost): Handle polynomial
10448 TYPE_VECTOR_SUBPARTS.
10449 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10451 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10452 number of elements from the PARALLEL rather than the mode.
10453 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10454 rather than GET_MODE_BITSIZE.
10455 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10456 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10457 (aarch64_expand_vec_perm_const_1): Handle polynomial
10458 d->perm.length () and d->perm elements.
10459 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10460 Apply to_constant to d->perm elements.
10461 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10462 polynomial CONST_VECTOR_NUNITS.
10463 (aarch64_move_pointer): Take amount as a poly_int64 rather
10465 (aarch64_progress_pointer): Avoid temporary variable.
10466 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10467 the mode attribute instead of GET_MODE.
10469 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10470 Alan Hayward <alan.hayward@arm.com>
10471 David Sherwood <david.sherwood@arm.com>
10473 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10474 x exists before using it.
10475 (aarch64_add_constant_internal): Rename to...
10476 (aarch64_add_offset_1): ...this. Replace regnum with separate
10477 src and dest rtxes. Handle the case in which they're different,
10478 including when the offset is zero. Replace scratchreg with an rtx.
10479 Use 2 additions if there is no spare register into which we can
10480 move a 16-bit constant.
10481 (aarch64_add_constant): Delete.
10482 (aarch64_add_offset): Replace reg with separate src and dest
10483 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10484 Use aarch64_add_offset_1.
10485 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10486 an rtx rather than an int. Take the delta as a poly_int64
10487 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10488 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10489 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10490 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10491 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10492 and aarch64_add_sp.
10493 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10494 aarch64_add_constant.
10496 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10498 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10499 Use scalar_float_mode.
10501 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10503 * config/aarch64/aarch64-simd.md
10504 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10505 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10506 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10507 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10508 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10509 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10510 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10511 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10512 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10513 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10515 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10518 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10519 targ_options->x_arm_arch_string is non NULL.
10521 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10523 * config/aarch64/aarch64.h
10524 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10526 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10529 * expmed.c (emit_store_flag_force): Swap if const op0
10530 and change VOIDmode to mode of op0.
10532 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10534 PR rtl-optimization/83761
10535 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10536 than bytes to mode_for_size.
10538 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10540 PR middle-end/83189
10541 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10542 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10545 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10547 PR middle-end/83575
10548 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10549 when in layout mode.
10550 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10551 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10554 2018-01-10 Michael Collison <michael.collison@arm.com>
10556 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10557 * config/aarch64/aarch64-option-extension.def: Add
10558 AARCH64_OPT_EXTENSION of 'fp16fml'.
10559 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10560 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10561 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10562 * config/aarch64/constraints.md (Ui7): New constraint.
10563 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10564 (VFMLA_SEL_W): Ditto.
10567 (VFMLA16_LOW): New int iterator.
10568 (VFMLA16_HIGH): Ditto.
10569 (UNSPEC_FMLAL): New unspec.
10570 (UNSPEC_FMLSL): Ditto.
10571 (UNSPEC_FMLAL2): Ditto.
10572 (UNSPEC_FMLSL2): Ditto.
10573 (f16mac): New code attribute.
10574 * config/aarch64/aarch64-simd-builtins.def
10575 (aarch64_fmlal_lowv2sf): Ditto.
10576 (aarch64_fmlsl_lowv2sf): Ditto.
10577 (aarch64_fmlalq_lowv4sf): Ditto.
10578 (aarch64_fmlslq_lowv4sf): Ditto.
10579 (aarch64_fmlal_highv2sf): Ditto.
10580 (aarch64_fmlsl_highv2sf): Ditto.
10581 (aarch64_fmlalq_highv4sf): Ditto.
10582 (aarch64_fmlslq_highv4sf): Ditto.
10583 (aarch64_fmlal_lane_lowv2sf): Ditto.
10584 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10585 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10586 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10587 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10588 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10589 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10590 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10591 (aarch64_fmlal_lane_highv2sf): Ditto.
10592 (aarch64_fmlsl_lane_highv2sf): Ditto.
10593 (aarch64_fmlal_laneq_highv2sf): Ditto.
10594 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10595 (aarch64_fmlalq_lane_highv4sf): Ditto.
10596 (aarch64_fmlsl_lane_highv4sf): Ditto.
10597 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10598 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10599 * config/aarch64/aarch64-simd.md:
10600 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10601 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10602 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10603 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10604 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10605 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10606 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10607 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10608 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10609 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10610 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10611 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10612 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10613 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10614 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10615 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10616 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10617 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10618 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10619 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10620 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10621 (vfmlsl_low_u32): Ditto.
10622 (vfmlalq_low_u32): Ditto.
10623 (vfmlslq_low_u32): Ditto.
10624 (vfmlal_high_u32): Ditto.
10625 (vfmlsl_high_u32): Ditto.
10626 (vfmlalq_high_u32): Ditto.
10627 (vfmlslq_high_u32): Ditto.
10628 (vfmlal_lane_low_u32): Ditto.
10629 (vfmlsl_lane_low_u32): Ditto.
10630 (vfmlal_laneq_low_u32): Ditto.
10631 (vfmlsl_laneq_low_u32): Ditto.
10632 (vfmlalq_lane_low_u32): Ditto.
10633 (vfmlslq_lane_low_u32): Ditto.
10634 (vfmlalq_laneq_low_u32): Ditto.
10635 (vfmlslq_laneq_low_u32): Ditto.
10636 (vfmlal_lane_high_u32): Ditto.
10637 (vfmlsl_lane_high_u32): Ditto.
10638 (vfmlal_laneq_high_u32): Ditto.
10639 (vfmlsl_laneq_high_u32): Ditto.
10640 (vfmlalq_lane_high_u32): Ditto.
10641 (vfmlslq_lane_high_u32): Ditto.
10642 (vfmlalq_laneq_high_u32): Ditto.
10643 (vfmlslq_laneq_high_u32): Ditto.
10644 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10645 (AARCH64_FL_FOR_ARCH8_4): New.
10646 (AARCH64_ISA_F16FML): New ISA flag.
10647 (TARGET_F16FML): New feature flag for fp16fml.
10648 (doc/invoke.texi): Document new fp16fml option.
10650 2018-01-10 Michael Collison <michael.collison@arm.com>
10652 * config/aarch64/aarch64-builtins.c:
10653 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10654 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10655 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10656 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10657 (AARCH64_ISA_SHA3): New ISA flag.
10658 (TARGET_SHA3): New feature flag for sha3.
10659 * config/aarch64/iterators.md (sha512_op): New int attribute.
10660 (CRYPTO_SHA512): New int iterator.
10661 (UNSPEC_SHA512H): New unspec.
10662 (UNSPEC_SHA512H2): Ditto.
10663 (UNSPEC_SHA512SU0): Ditto.
10664 (UNSPEC_SHA512SU1): Ditto.
10665 * config/aarch64/aarch64-simd-builtins.def
10666 (aarch64_crypto_sha512hqv2di): New builtin.
10667 (aarch64_crypto_sha512h2qv2di): Ditto.
10668 (aarch64_crypto_sha512su0qv2di): Ditto.
10669 (aarch64_crypto_sha512su1qv2di): Ditto.
10670 (aarch64_eor3qv8hi): Ditto.
10671 (aarch64_rax1qv2di): Ditto.
10672 (aarch64_xarqv2di): Ditto.
10673 (aarch64_bcaxqv8hi): Ditto.
10674 * config/aarch64/aarch64-simd.md:
10675 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10676 (aarch64_crypto_sha512su0qv2di): Ditto.
10677 (aarch64_crypto_sha512su1qv2di): Ditto.
10678 (aarch64_eor3qv8hi): Ditto.
10679 (aarch64_rax1qv2di): Ditto.
10680 (aarch64_xarqv2di): Ditto.
10681 (aarch64_bcaxqv8hi): Ditto.
10682 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10683 (vsha512h2q_u64): Ditto.
10684 (vsha512su0q_u64): Ditto.
10685 (vsha512su1q_u64): Ditto.
10686 (veor3q_u16): Ditto.
10687 (vrax1q_u64): Ditto.
10688 (vxarq_u64): Ditto.
10689 (vbcaxq_u16): Ditto.
10690 * config/arm/types.md (crypto_sha512): New type attribute.
10691 (crypto_sha3): Ditto.
10692 (doc/invoke.texi): Document new sha3 option.
10694 2018-01-10 Michael Collison <michael.collison@arm.com>
10696 * config/aarch64/aarch64-builtins.c:
10697 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10698 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10699 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10700 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10701 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10702 (AARCH64_ISA_SM4): New ISA flag.
10703 (TARGET_SM4): New feature flag for sm4.
10704 * config/aarch64/aarch64-simd-builtins.def
10705 (aarch64_sm3ss1qv4si): Ditto.
10706 (aarch64_sm3tt1aq4si): Ditto.
10707 (aarch64_sm3tt1bq4si): Ditto.
10708 (aarch64_sm3tt2aq4si): Ditto.
10709 (aarch64_sm3tt2bq4si): Ditto.
10710 (aarch64_sm3partw1qv4si): Ditto.
10711 (aarch64_sm3partw2qv4si): Ditto.
10712 (aarch64_sm4eqv4si): Ditto.
10713 (aarch64_sm4ekeyqv4si): Ditto.
10714 * config/aarch64/aarch64-simd.md:
10715 (aarch64_sm3ss1qv4si): Ditto.
10716 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10717 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10718 (aarch64_sm4eqv4si): Ditto.
10719 (aarch64_sm4ekeyqv4si): Ditto.
10720 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10721 (sm3part_op): Ditto.
10722 (CRYPTO_SM3TT): Ditto.
10723 (CRYPTO_SM3PART): Ditto.
10724 (UNSPEC_SM3SS1): New unspec.
10725 (UNSPEC_SM3TT1A): Ditto.
10726 (UNSPEC_SM3TT1B): Ditto.
10727 (UNSPEC_SM3TT2A): Ditto.
10728 (UNSPEC_SM3TT2B): Ditto.
10729 (UNSPEC_SM3PARTW1): Ditto.
10730 (UNSPEC_SM3PARTW2): Ditto.
10731 (UNSPEC_SM4E): Ditto.
10732 (UNSPEC_SM4EKEY): Ditto.
10733 * config/aarch64/constraints.md (Ui2): New constraint.
10734 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10735 * config/arm/types.md (crypto_sm3): New type attribute.
10736 (crypto_sm4): Ditto.
10737 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10738 (vsm3tt1aq_u32): Ditto.
10739 (vsm3tt1bq_u32): Ditto.
10740 (vsm3tt2aq_u32): Ditto.
10741 (vsm3tt2bq_u32): Ditto.
10742 (vsm3partw1q_u32): Ditto.
10743 (vsm3partw2q_u32): Ditto.
10744 (vsm4eq_u32): Ditto.
10745 (vsm4ekeyq_u32): Ditto.
10746 (doc/invoke.texi): Document new sm4 option.
10748 2018-01-10 Michael Collison <michael.collison@arm.com>
10750 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10751 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10752 (AARCH64_FL_FOR_ARCH8_4): New.
10753 (AARCH64_FL_V8_4): New flag.
10754 (doc/invoke.texi): Document new armv8.4-a option.
10756 2018-01-10 Michael Collison <michael.collison@arm.com>
10758 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10759 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10760 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10761 * config/aarch64/aarch64-option-extension.def: Add
10762 AARCH64_OPT_EXTENSION of 'sha2'.
10763 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10764 (crypto): Disable sha2 and aes if crypto disabled.
10765 (crypto): Enable aes and sha2 if enabled.
10766 (simd): Disable sha2 and aes if simd disabled.
10767 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10769 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10770 (TARGET_SHA2): New feature flag for sha2.
10771 (TARGET_AES): New feature flag for aes.
10772 * config/aarch64/aarch64-simd.md:
10773 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10774 conditional on TARGET_AES.
10775 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10776 (aarch64_crypto_sha1hsi): Make pattern conditional
10778 (aarch64_crypto_sha1hv4si): Ditto.
10779 (aarch64_be_crypto_sha1hv4si): Ditto.
10780 (aarch64_crypto_sha1su1v4si): Ditto.
10781 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10782 (aarch64_crypto_sha1su0v4si): Ditto.
10783 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10784 (aarch64_crypto_sha256su0v4si): Ditto.
10785 (aarch64_crypto_sha256su1v4si): Ditto.
10786 (doc/invoke.texi): Document new aes and sha2 options.
10788 2018-01-10 Martin Sebor <msebor@redhat.com>
10790 PR tree-optimization/83781
10791 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10794 2018-01-11 Martin Sebor <msebor@gmail.com>
10795 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10797 PR tree-optimization/83501
10798 PR tree-optimization/81703
10800 * tree-ssa-strlen.c (get_string_cst): Rename...
10801 (get_string_len): ...to this. Handle global constants.
10802 (handle_char_store): Adjust.
10804 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10805 Jim Wilson <jimw@sifive.com>
10807 * config/riscv/riscv-protos.h (riscv_output_return): New.
10808 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10809 (riscv_attribute_table, riscv_output_return),
10810 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10811 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10812 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10814 (riscv_expand_prologue): Add early return for naked function.
10815 (riscv_expand_epilogue): Likewise.
10816 (riscv_function_ok_for_sibcall): Return false for naked function.
10817 (riscv_set_current_function): New.
10818 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10819 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10820 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10821 * doc/extend.texi (RISC-V Function Attributes): New.
10823 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10825 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10826 check for 128-bit long double before checking TCmode.
10827 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10828 128-bit long doubles before checking TFmode or TCmode.
10829 (FLOAT128_IBM_P): Likewise.
10831 2018-01-10 Martin Sebor <msebor@redhat.com>
10833 PR tree-optimization/83671
10834 * builtins.c (c_strlen): Unconditionally return zero for the empty
10836 Use -Warray-bounds for warnings.
10837 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10838 for non-constant array indices with COMPONENT_REF, arrays of
10839 arrays, and pointers to arrays.
10840 (gimple_fold_builtin_strlen): Determine and set length range for
10841 non-constant character arrays.
10843 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10845 PR middle-end/81897
10846 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10849 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10851 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10853 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10856 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10857 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10858 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10859 indexed_or_indirect_operand predicate.
10860 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10861 (*vsx_le_perm_load_v8hi): Likewise.
10862 (*vsx_le_perm_load_v16qi): Likewise.
10863 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10864 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10865 (*vsx_le_perm_store_v8hi): Likewise.
10866 (*vsx_le_perm_store_v16qi): Likewise.
10867 (eight unnamed splitters): Likewise.
10869 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10871 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10872 * config/rs6000/emmintrin.h: Likewise.
10873 * config/rs6000/mmintrin.h: Likewise.
10874 * config/rs6000/xmmintrin.h: Likewise.
10876 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10879 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10881 * tree.c (tree_nop_conversion): Return true for location wrapper
10883 (maybe_wrap_with_location): New function.
10884 (selftest::check_strip_nops): New function.
10885 (selftest::test_location_wrappers): New function.
10886 (selftest::tree_c_tests): Call it.
10887 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10888 (maybe_wrap_with_location): New decl.
10889 (EXPR_LOCATION_WRAPPER_P): New macro.
10890 (location_wrapper_p): New inline function.
10891 (tree_strip_any_location_wrapper): New inline function.
10893 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10896 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10897 stack_realign_offset for the largest alignment of stack slot
10899 (ix86_find_max_used_stack_alignment): New function.
10900 (ix86_finalize_stack_frame_flags): Use it. Set
10901 max_used_stack_alignment if we don't realign stack.
10902 * config/i386/i386.h (machine_function): Add
10903 max_used_stack_alignment.
10905 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10907 * config/arm/arm.opt (-mbranch-cost): New option.
10908 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10911 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10914 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10915 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10917 2018-01-10 Richard Biener <rguenther@suse.de>
10920 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10921 early out so it also covers the case where we have a non-NULL
10924 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10926 PR tree-optimization/83753
10927 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10928 for non-strided grouped accesses if the number of elements is 1.
10930 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10933 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10934 * i386.h (TARGET_USE_GATHER): Define.
10935 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10937 2018-01-10 Martin Liska <mliska@suse.cz>
10940 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10941 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10943 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10944 CLEANUP_NO_PARTITIONING is not set.
10946 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10948 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10949 for vectors, as a partial revert of r254296.
10950 * rtl.h (const_vec_p): Delete.
10951 (const_vec_duplicate_p): Don't test for vector CONSTs.
10952 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10953 * expmed.c (make_tree): Likewise.
10956 * common.md (E, F): Use CONSTANT_P instead of checking for
10958 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10959 checking for CONST_VECTOR.
10961 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10963 PR middle-end/83575
10964 * predict.c (force_edge_cold): Handle in more sane way edges
10965 with no prediction.
10967 2018-01-09 Carl Love <cel@us.ibm.com>
10969 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10971 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10972 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10973 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10974 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10975 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10976 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10977 * config/rs6000/rs6000-protos.h: Add extern defition for
10978 rs6000_generate_float2_double_code.
10979 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10981 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10982 (float2_v2df): Add define_expand.
10984 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10987 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10988 op_mode in the force_to_mode call.
10990 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10992 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10993 instead of checking each element individually.
10994 (aarch64_evpc_uzp): Likewise.
10995 (aarch64_evpc_zip): Likewise.
10996 (aarch64_evpc_ext): Likewise.
10997 (aarch64_evpc_rev): Likewise.
10998 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10999 instead of checking each element individually. Return true without
11001 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11002 whether all selected elements come from the same input, instead of
11003 checking each element individually. Remove calls to gen_rtx_REG,
11004 start_sequence and end_sequence and instead assert that no rtl is
11007 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11009 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11010 order of HIGH and CONST checks.
11012 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11014 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11015 if the destination isn't an SSA_NAME.
11017 2018-01-09 Richard Biener <rguenther@suse.de>
11019 PR tree-optimization/83668
11020 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11022 (canonicalize_loop_form): ... here, renamed from ...
11023 (canonicalize_loop_closed_ssa_form): ... this and amended to
11024 swap successor edges for loop exit blocks to make us use
11025 the RPO order we need for initial schedule generation.
11027 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11029 PR tree-optimization/64811
11030 * match.pd: When optimizing comparisons with Inf, avoid
11031 introducing or losing exceptions from comparisons with NaN.
11033 2018-01-09 Martin Liska <mliska@suse.cz>
11036 * asan.c (shadow_mem_size): Add gcc_assert.
11038 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11040 Don't save registers in main().
11043 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11044 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11045 * config/avr/avr.c (avr_set_current_function): Don't error if
11046 naked, OS_task or OS_main are specified at the same time.
11047 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11049 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11051 * common/config/avr/avr-common.c (avr_option_optimization_table):
11052 Switch on -mmain-is-OS_task for optimizing compilations.
11054 2018-01-09 Richard Biener <rguenther@suse.de>
11056 PR tree-optimization/83572
11057 * graphite.c: Include cfganal.h.
11058 (graphite_transform_loops): Connect infinite loops to exit
11059 and remove fake edges at the end.
11061 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11063 * ipa-inline.c (edge_badness): Revert accidental checkin.
11065 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11068 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11069 symbols; not inline clones.
11071 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11074 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11075 hard registers. Formatting fixes.
11077 PR preprocessor/83722
11078 * gcc.c (try_generate_repro): Pass
11079 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11080 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11083 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11084 Kito Cheng <kito.cheng@gmail.com>
11086 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11087 (riscv_leaf_function_p): Delete.
11088 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11090 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11092 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11094 (do_ifelse): New function.
11095 (do_isel): New function.
11096 (do_sub3): New function.
11097 (do_add3): New function.
11098 (do_load_mask_compare): New function.
11099 (do_overlap_load_compare): New function.
11100 (expand_compare_loop): New function.
11101 (expand_block_compare): Call expand_compare_loop() when appropriate.
11102 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11103 option description.
11104 (-mblock-compare-inline-loop-limit): New option.
11106 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11109 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11110 Reverse order of second and third operands in first alternative.
11111 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11112 of first and second elements in UNSPEC_VPERMR vector.
11113 (altivec_expand_vec_perm_le): Likewise.
11115 2018-01-08 Jeff Law <law@redhat.com>
11117 PR rtl-optimizatin/81308
11118 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11119 (process_switch): If group_case_labels makes a change, then set
11121 (pass_convert_switch::execute): If a switch is converted, then
11122 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11124 PR rtl-optimization/81308
11125 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11128 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11130 PR target/83663 - Revert r255946
11131 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11132 generation for cases where splatting a value is not useful.
11133 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11134 across a vec_duplicate and a paradoxical subreg forming a vector
11135 mode to a vec_concat.
11137 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11139 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11140 -march=armv8.3-a variants.
11141 * config/arm/t-multilib: Likewise.
11142 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11144 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11146 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11148 (cceq_ior_compare_complement): Give it a name so I can use it, and
11149 change boolean_or_operator predicate to boolean_operator so it can
11150 be used to generate a crand.
11151 (eqne): New code iterator.
11152 (bd/bd_neg): New code_attrs.
11153 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11154 a single define_insn.
11155 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11156 decrement (bdnzt/bdnzf/bdzt/bdzf).
11157 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11158 with the new names of the branch decrement patterns, and added the
11159 names of the branch decrement conditional patterns.
11161 2018-01-08 Richard Biener <rguenther@suse.de>
11163 PR tree-optimization/83563
11164 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11167 2018-01-08 Richard Biener <rguenther@suse.de>
11169 PR middle-end/83713
11170 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11172 2018-01-08 Richard Biener <rguenther@suse.de>
11174 PR tree-optimization/83685
11175 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11176 references to abnormals.
11178 2018-01-08 Richard Biener <rguenther@suse.de>
11181 * dwarf2out.c (output_indirect_strings): Handle empty
11182 skeleton_debug_str_hash.
11183 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11185 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11187 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11188 (emit_store_direct): Likewise.
11189 (arc_trampoline_adjust_address): Likewise.
11190 (arc_asm_trampoline_template): New function.
11191 (arc_initialize_trampoline): Use asm_trampoline_template.
11192 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11193 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11194 * config/arc/arc.md (flush_icache): Delete pattern.
11196 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11198 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11199 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11202 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11205 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11206 by not USED_FOR_TARGET.
11207 (make_pass_resolve_sw_modes): Likewise.
11209 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11211 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11214 2018-01-08 Richard Biener <rguenther@suse.de>
11216 PR middle-end/83580
11217 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11219 2018-01-08 Richard Biener <rguenther@suse.de>
11221 PR middle-end/83517
11222 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11224 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11226 PR middle-end/81897
11227 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11228 basic blocks with a small number of successors.
11229 (convert_control_dep_chain_into_preds): Improve handling of
11231 (dump_predicates): Split apart into...
11232 (dump_pred_chain): ...here...
11233 (dump_pred_info): ...and here.
11234 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11235 (can_chain_union_be_invalidated_p): Improve check for invalidation
11237 (uninit_uses_cannot_happen): Avoid unnecessary if
11238 convert_control_dep_chain_into_preds yielded nothing.
11240 2018-01-06 Martin Sebor <msebor@redhat.com>
11242 PR tree-optimization/83640
11243 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11244 subtracting negative offset from size.
11245 (builtin_access::overlap): Adjust offset bounds of the access to fall
11246 within the size of the object if possible.
11248 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11250 PR rtl-optimization/83699
11251 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11252 extract_bit_field_as_subreg to cases in which the extracted
11253 value is also a vector.
11255 * lra-constraints.c (process_alt_operands): Test for the equivalence
11256 substitutions when detecting a possible reload cycle.
11258 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11261 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11262 by default if flag_selective_schedling{,2}. Formatting fixes.
11264 PR rtl-optimization/83682
11265 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11266 if it has non-VECTOR_MODE element mode.
11267 (vec_duplicate_p): Likewise.
11269 PR middle-end/83694
11270 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11271 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11273 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11276 * config/i386/i386-builtin.def
11277 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11278 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11279 Require also OPTION_MASK_ISA_AVX512F in addition to
11280 OPTION_MASK_ISA_GFNI.
11281 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11282 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11283 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11284 to OPTION_MASK_ISA_GFNI.
11285 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11286 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11287 OPTION_MASK_ISA_AVX512BW.
11288 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11289 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11290 addition to OPTION_MASK_ISA_GFNI.
11291 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11292 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11293 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11294 to OPTION_MASK_ISA_GFNI.
11295 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11296 a requirement for all ISAs rather than any of them with a few
11298 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11300 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11301 bitmasks to be enabled with 3 exceptions, instead of requiring any
11302 enabled ISA with lots of exceptions.
11303 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11304 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11305 Change avx512bw in isa attribute to avx512f.
11306 * config/i386/sgxintrin.h: Add license boilerplate.
11307 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11308 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11309 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11310 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11312 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11313 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11314 temporarily sse2 rather than sse if not enabled already.
11317 * config/i386/sse.md (VI248_VLBW): Rename to ...
11318 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11319 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11320 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11321 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11322 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11323 mode iterator instead of VI248_VLBW.
11325 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11327 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11328 (record_modified): Skip clobbers; add debug output.
11329 (param_change_prob): Use sreal frequencies.
11331 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11333 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11334 punt for user-aligned variables.
11336 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11338 * tree-chrec.c (chrec_contains_symbols): Return true for
11341 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11344 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11345 of (x|y) == x for BICS pattern.
11347 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11349 PR tree-optimization/83605
11350 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11351 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11354 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11356 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11357 * config/epiphany/rtems.h: New file.
11359 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11360 Uros Bizjak <ubizjak@gmail.com>
11363 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11364 QIreg_operand instead of register_operand predicate.
11365 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11366 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11367 comments instead of -fmitigate[-_]rop.
11369 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11372 * cgraphunit.c (symbol_table::compile): Switch to text_section
11373 before calling assembly_start debug hook.
11374 * run-rtl-passes.c (run_rtl_passes): Likewise.
11377 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11379 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11380 range_int_cst_p rather than !symbolic_range_p before calling
11381 extract_range_from_multiplicative_op_1.
11383 2018-01-04 Jeff Law <law@redhat.com>
11385 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11386 redundant test in assertion.
11388 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11390 * doc/rtl.texi: Document machine_mode wrapper classes.
11392 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11394 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11395 using tree_to_uhwi.
11397 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11399 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11400 the VEC_PERM_EXPR fold to fail.
11402 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11405 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11406 to switched_sections.
11408 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11411 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11412 test for d.testing.
11414 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11417 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11418 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11420 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11423 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11424 is BLKmode and bitpos not zero or mode change is needed.
11426 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11429 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11432 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11435 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11436 instead of MULT rtx. Update all corresponding splitters.
11437 (*saddl_se): Ditto.
11438 (*ssub<modesuffix>): Ditto.
11439 (*ssubl_se): Ditto.
11440 (*cmp_sadd_di): Update split patterns.
11441 (*cmp_sadd_si): Ditto.
11442 (*cmp_sadd_sidi): Ditto.
11443 (*cmp_ssub_di): Ditto.
11444 (*cmp_ssub_si): Ditto.
11445 (*cmp_ssub_sidi): Ditto.
11446 * config/alpha/predicates.md (const23_operand): New predicate.
11447 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11448 Look for ASHIFT, not MULT inner operand.
11449 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11451 2018-01-04 Martin Liska <mliska@suse.cz>
11453 PR gcov-profile/83669
11454 * gcov.c (output_intermediate_file): Add version to intermediate
11456 * doc/gcov.texi: Document new field 'version' in intermediate
11457 file format. Fix location of '-k' option of gcov command.
11459 2018-01-04 Martin Liska <mliska@suse.cz>
11462 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11464 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11466 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11468 2018-01-03 Martin Sebor <msebor@redhat.com>
11470 PR tree-optimization/83655
11471 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11472 checking calls with invalid arguments.
11474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11476 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11477 (vectorizable_mask_load_store): Delete.
11478 (vectorizable_call): Return false for masked loads and stores.
11479 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11480 instead of gimple_assign_rhs1.
11481 (vectorizable_load): Handle IFN_MASK_LOAD.
11482 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11486 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11488 (vectorizable_mask_load_store): ...here.
11489 (vectorizable_load): ...and here.
11491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11493 * tree-vect-stmts.c (vect_build_all_ones_mask)
11494 (vect_build_zero_merge_argument): New functions, split out from...
11495 (vectorizable_load): ...here.
11497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11499 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11501 (vectorizable_mask_load_store): ...here.
11502 (vectorizable_store): ...and here.
11504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11506 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11508 (vectorizable_mask_load_store): ...here.
11510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11512 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11513 (vect_model_store_cost): Take a vec_load_store_type instead of a
11515 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11516 (vect_model_store_cost): Take a vec_load_store_type instead of a
11518 (vectorizable_mask_load_store): Update accordingly.
11519 (vectorizable_store): Likewise.
11520 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11524 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11525 IFN_MASK_LOAD calls here rather than...
11526 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11529 Alan Hayward <alan.hayward@arm.com>
11530 David Sherwood <david.sherwood@arm.com>
11532 * expmed.c (extract_bit_field_1): For vector extracts,
11533 fall back to extract_bit_field_as_subreg if vec_extract
11536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11537 Alan Hayward <alan.hayward@arm.com>
11538 David Sherwood <david.sherwood@arm.com>
11540 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11541 they are variable or constant sized.
11542 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11543 slots for constant-sized data.
11545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11546 Alan Hayward <alan.hayward@arm.com>
11547 David Sherwood <david.sherwood@arm.com>
11549 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11550 handling COND_EXPRs with boolean comparisons, try to find a better
11551 basis for the mask type than the boolean itself.
11553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11555 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11556 is calculated and how it can be overridden.
11557 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11558 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11560 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11564 Alan Hayward <alan.hayward@arm.com>
11565 David Sherwood <david.sherwood@arm.com>
11567 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11568 Remove the mode argument.
11569 (aarch64_simd_valid_immediate): Remove the mode and inverse
11571 * config/aarch64/iterators.md (bitsize): New iterator.
11572 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11573 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11574 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11575 aarch64_simd_valid_immediate.
11576 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11577 (aarch64_reg_or_bic_imm): Likewise.
11578 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11579 with an insn_type enum and msl with a modifier_type enum.
11580 Replace element_width with a scalar_mode. Change the shift
11581 to unsigned int. Add constructors for scalar_float_mode and
11582 scalar_int_mode elements.
11583 (aarch64_vect_float_const_representable_p): Delete.
11584 (aarch64_can_const_movi_rtx_p)
11585 (aarch64_simd_scalar_immediate_valid_for_move)
11586 (aarch64_simd_make_constant): Update call to
11587 aarch64_simd_valid_immediate.
11588 (aarch64_advsimd_valid_immediate_hs): New function.
11589 (aarch64_advsimd_valid_immediate): Likewise.
11590 (aarch64_simd_valid_immediate): Remove mode and inverse
11591 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11592 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11593 and aarch64_float_const_representable_p on the result.
11594 (aarch64_output_simd_mov_immediate): Remove mode argument.
11595 Update call to aarch64_simd_valid_immediate and use of
11596 simd_immediate_info.
11597 (aarch64_output_scalar_simd_mov_immediate): Update call
11600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11601 Alan Hayward <alan.hayward@arm.com>
11602 David Sherwood <david.sherwood@arm.com>
11604 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11605 (mode_nunits): Likewise CONST_MODE_NUNITS.
11606 * machmode.def (ADJUST_NUNITS): Document.
11607 * genmodes.c (mode_data::need_nunits_adj): New field.
11608 (blank_mode): Update accordingly.
11609 (adj_nunits): New variable.
11610 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11612 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11613 listed in adj_nunits.
11614 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11615 listed in adj_nunits. Don't emit case statements for such modes.
11616 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11617 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11618 nothing if adj_nunits is nonnull.
11619 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11620 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11621 (emit_mode_fbit): Update use of print_maybe_const_decl.
11622 (emit_move_size): Likewise. Treat the array as non-const
11624 (emit_mode_adjustments): Handle adj_nunits.
11626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11628 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11629 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11630 (VECTOR_MODES): Use it.
11631 (make_vector_modes): Take the prefix as an argument.
11633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11634 Alan Hayward <alan.hayward@arm.com>
11635 David Sherwood <david.sherwood@arm.com>
11637 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11638 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11639 for MODE_VECTOR_BOOL.
11640 * machmode.def (VECTOR_BOOL_MODE): Document.
11641 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11642 (make_vector_bool_mode): New function.
11643 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11645 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11646 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11648 * stor-layout.c (int_mode_for_mode): Likewise.
11649 * tree.c (build_vector_type_for_mode): Likewise.
11650 * varasm.c (output_constant_pool_2): Likewise.
11651 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11652 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11653 for MODE_VECTOR_BOOL.
11654 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11655 of mode class checks.
11656 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11657 instead of a list of mode class checks.
11658 (expand_vector_scalar_condition): Likewise.
11659 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11662 Alan Hayward <alan.hayward@arm.com>
11663 David Sherwood <david.sherwood@arm.com>
11665 * machmode.h (mode_size): Change from unsigned short to
11667 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11668 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11669 or if measurement_type is not polynomial.
11670 (fixed_size_mode::includes_p): Check for constant-sized modes.
11671 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11672 return a poly_uint16 rather than an unsigned short.
11673 (emit_mode_size): Change the type of mode_size from unsigned short
11674 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11675 (emit_mode_adjustments): Cope with polynomial vector sizes.
11676 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11678 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11680 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11681 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11682 * caller-save.c (setup_save_areas): Likewise.
11683 (replace_reg_with_saved_mem): Likewise.
11684 * calls.c (emit_library_call_value_1): Likewise.
11685 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11686 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11687 (gen_lowpart_for_combine): Likewise.
11688 * convert.c (convert_to_integer_1): Likewise.
11689 * cse.c (equiv_constant, cse_insn): Likewise.
11690 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11691 (cselib_subst_to_values): Likewise.
11692 * dce.c (word_dce_process_block): Likewise.
11693 * df-problems.c (df_word_lr_mark_ref): Likewise.
11694 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11695 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11696 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11697 (rtl_for_decl_location): Likewise.
11698 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11699 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11700 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11701 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11702 (expand_expr_real_1): Likewise.
11703 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11704 (pad_below): Likewise.
11705 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11706 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11707 * ira.c (get_subreg_tracking_sizes): Likewise.
11708 * ira-build.c (ira_create_allocno_objects): Likewise.
11709 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11710 (ira_sort_regnos_for_alter_reg): Likewise.
11711 * ira-costs.c (record_operand_costs): Likewise.
11712 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11713 (resolve_simple_move): Likewise.
11714 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11715 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11716 (lra_constraints): Likewise.
11717 (CONST_POOL_OK_P): Reject variable-sized modes.
11718 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11719 (add_pseudo_to_slot, lra_spill): Likewise.
11720 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11721 * optabs-query.c (get_best_extraction_insn): Likewise.
11722 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11723 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11724 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11725 * recog.c (offsettable_address_addr_space_p): Likewise.
11726 * regcprop.c (maybe_mode_change): Likewise.
11727 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11728 * regrename.c (build_def_use): Likewise.
11729 * regstat.c (dump_reg_info): Likewise.
11730 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11731 (find_reloads, find_reloads_subreg_address): Likewise.
11732 * reload1.c (eliminate_regs_1): Likewise.
11733 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11734 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11735 (simplify_binary_operation_1, simplify_subreg): Likewise.
11736 * targhooks.c (default_function_arg_padding): Likewise.
11737 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11738 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11739 (verify_gimple_assign_ternary): Likewise.
11740 * tree-inline.c (estimate_move_cost): Likewise.
11741 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11742 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11743 (get_address_cost_ainc): Likewise.
11744 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11745 (vect_supportable_dr_alignment): Likewise.
11746 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11747 (vectorizable_reduction): Likewise.
11748 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11749 (vectorizable_operation, vectorizable_load): Likewise.
11750 * tree.c (build_same_sized_truth_vector_type): Likewise.
11751 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11752 * var-tracking.c (emit_note_insn_var_location): Likewise.
11753 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11754 (ADDR_VEC_ALIGN): Likewise.
11756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11757 Alan Hayward <alan.hayward@arm.com>
11758 David Sherwood <david.sherwood@arm.com>
11760 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11762 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11763 or if measurement_type is polynomial.
11764 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11765 * combine.c (make_extraction): Likewise.
11766 * dse.c (find_shift_sequence): Likewise.
11767 * dwarf2out.c (mem_loc_descriptor): Likewise.
11768 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11769 (extract_bit_field, extract_low_bits): Likewise.
11770 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11771 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11772 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11773 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11774 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11775 * reload.c (find_reloads): Likewise.
11776 * reload1.c (alter_reg): Likewise.
11777 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11778 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11779 * tree-if-conv.c (predicate_mem_writes): Likewise.
11780 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11781 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11782 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11783 * valtrack.c (dead_debug_insert_temp): Likewise.
11784 * varasm.c (mergeable_constant_section): Likewise.
11785 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11788 Alan Hayward <alan.hayward@arm.com>
11789 David Sherwood <david.sherwood@arm.com>
11791 * expr.c (expand_assignment): Cope with polynomial mode sizes
11792 when assigning to a CONCAT.
11794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11795 Alan Hayward <alan.hayward@arm.com>
11796 David Sherwood <david.sherwood@arm.com>
11798 * machmode.h (mode_precision): Change from unsigned short to
11800 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11802 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11803 or if measurement_type is not polynomial.
11804 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11805 in which the mode is already known to be a scalar_int_mode.
11806 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11807 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11809 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11810 for GET_MODE_PRECISION.
11811 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11812 for GET_MODE_PRECISION.
11813 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11815 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11816 (expand_field_assignment, make_extraction): Likewise.
11817 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11818 (get_last_value): Likewise.
11819 * convert.c (convert_to_integer_1): Likewise.
11820 * cse.c (cse_insn): Likewise.
11821 * expr.c (expand_expr_real_1): Likewise.
11822 * lra-constraints.c (simplify_operand_subreg): Likewise.
11823 * optabs-query.c (can_atomic_load_p): Likewise.
11824 * optabs.c (expand_atomic_load): Likewise.
11825 (expand_atomic_store): Likewise.
11826 * ree.c (combine_reaching_defs): Likewise.
11827 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11828 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11829 * tree.h (type_has_mode_precision_p): Likewise.
11830 * ubsan.c (instrument_si_overflow): Likewise.
11832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11833 Alan Hayward <alan.hayward@arm.com>
11834 David Sherwood <david.sherwood@arm.com>
11836 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11837 polynomial numbers of units.
11838 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11839 (valid_vector_subparts_p): New function.
11840 (build_vector_type): Remove temporary shim and take the number
11841 of units as a poly_uint64 rather than an int.
11842 (build_opaque_vector_type): Take the number of units as a
11843 poly_uint64 rather than an int.
11844 * tree.c (build_vector_from_ctor): Handle polynomial
11845 TYPE_VECTOR_SUBPARTS.
11846 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11847 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11848 (build_vector_from_val): If the number of units is variable,
11849 use build_vec_duplicate_cst for constant operands and
11850 VEC_DUPLICATE_EXPR otherwise.
11851 (make_vector_type): Remove temporary is_constant ().
11852 (build_vector_type, build_opaque_vector_type): Take the number of
11853 units as a poly_uint64 rather than an int.
11854 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11856 * cfgexpand.c (expand_debug_expr): Likewise.
11857 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11858 (store_constructor, expand_expr_real_1): Likewise.
11859 (const_scalar_mask_from_tree): Likewise.
11860 * fold-const-call.c (fold_const_reduction): Likewise.
11861 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11862 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11863 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11864 (fold_relational_const): Likewise.
11865 (native_interpret_vector): Likewise. Change the size from an
11866 int to an unsigned int.
11867 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11868 TYPE_VECTOR_SUBPARTS.
11869 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11870 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11871 duplicating a non-constant operand into a variable-length vector.
11872 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11873 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11874 * ipa-icf.c (sem_variable::equals): Likewise.
11875 * match.pd: Likewise.
11876 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11877 * print-tree.c (print_node): Likewise.
11878 * stor-layout.c (layout_type): Likewise.
11879 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11880 * tree-cfg.c (verify_gimple_comparison): Likewise.
11881 (verify_gimple_assign_binary): Likewise.
11882 (verify_gimple_assign_ternary): Likewise.
11883 (verify_gimple_assign_single): Likewise.
11884 * tree-pretty-print.c (dump_generic_node): Likewise.
11885 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11886 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11887 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11888 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11889 (vect_shift_permute_load_chain): Likewise.
11890 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11891 (expand_vector_condition, optimize_vector_constructor): Likewise.
11892 (lower_vec_perm, get_compute_type): Likewise.
11893 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11894 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11895 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11896 (vect_recog_mask_conversion_pattern): Likewise.
11897 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11898 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11899 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11900 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11901 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11902 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11903 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11904 (supportable_widening_operation): Likewise.
11905 (supportable_narrowing_operation): Likewise.
11906 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11908 * varasm.c (output_constant): Likewise.
11910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11911 Alan Hayward <alan.hayward@arm.com>
11912 David Sherwood <david.sherwood@arm.com>
11914 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11915 so that both the length == 3 and length != 3 cases set up their
11916 own permute vectors. Add comments explaining why we know the
11917 number of elements is constant.
11918 (vect_permute_load_chain): Likewise.
11920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11921 Alan Hayward <alan.hayward@arm.com>
11922 David Sherwood <david.sherwood@arm.com>
11924 * machmode.h (mode_nunits): Change from unsigned char to
11926 (ONLY_FIXED_SIZE_MODES): New macro.
11927 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11928 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11929 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11931 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11932 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11933 or if measurement_type is not polynomial.
11934 * genmodes.c (ZERO_COEFFS): New macro.
11935 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11937 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11938 Use ZERO_COEFFS when emitting initializers.
11939 * data-streamer.h (bp_pack_poly_value): New function.
11940 (bp_unpack_poly_value): Likewise.
11941 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11942 for GET_MODE_NUNITS.
11943 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11944 for GET_MODE_NUNITS.
11945 * tree.c (make_vector_type): Remove temporary shim and make
11946 the real function take the number of units as a poly_uint64
11947 rather than an int.
11948 (build_vector_type_for_mode): Handle polynomial nunits.
11949 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11950 * emit-rtl.c (const_vec_series_p_1): Likewise.
11951 (gen_rtx_CONST_VECTOR): Likewise.
11952 * fold-const.c (test_vec_duplicate_folding): Likewise.
11953 * genrecog.c (validate_pattern): Likewise.
11954 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11955 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11956 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11957 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11958 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11959 * rtlanal.c (subreg_get_info): Likewise.
11960 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11961 (vect_grouped_load_supported): Likewise.
11962 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11963 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11964 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11965 (simplify_const_unary_operation, simplify_binary_operation_1)
11966 (simplify_const_binary_operation, simplify_ternary_operation)
11967 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11968 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11969 instead of CONST_VECTOR_NUNITS.
11970 * varasm.c (output_constant_pool_2): Likewise.
11971 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11972 explicit-encoded elements in the XVEC for variable-length vectors.
11974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11976 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11979 Alan Hayward <alan.hayward@arm.com>
11980 David Sherwood <david.sherwood@arm.com>
11982 * coretypes.h (fixed_size_mode): Declare.
11983 (fixed_size_mode_pod): New typedef.
11984 * builtins.h (target_builtins::x_apply_args_mode)
11985 (target_builtins::x_apply_result_mode): Change type to
11986 fixed_size_mode_pod.
11987 * builtins.c (apply_args_size, apply_result_size, result_vector)
11988 (expand_builtin_apply_args_1, expand_builtin_apply)
11989 (expand_builtin_return): Update accordingly.
11991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11993 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11994 * cselib.c (cselib_hash_rtx): Likewise.
11995 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11996 CONST_VECTOR encoding.
11998 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11999 Jeff Law <law@redhat.com>
12002 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12003 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12004 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12005 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12008 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12009 explicitly probe *sp in a noreturn function if there were any callee
12010 register saves or frame pointer is needed.
12012 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12015 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12016 BLKmode for ternary, binary or unary expressions.
12019 * var-tracking.c (delete_vta_debug_insn): New inline function.
12020 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12021 insns from get_insns () to NULL instead of each bb separately.
12022 Use delete_vta_debug_insn. No longer static.
12023 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12024 delete_vta_debug_insns callers.
12025 * rtl.h (delete_vta_debug_insns): Declare.
12026 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12027 instead of variable_tracking_main.
12029 2018-01-03 Martin Sebor <msebor@redhat.com>
12031 PR tree-optimization/83603
12032 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12033 arguments past the endof the argument list in functions declared
12034 without a prototype.
12035 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12036 Avoid checking when arguments are null.
12038 2018-01-03 Martin Sebor <msebor@redhat.com>
12041 * doc/extend.texi (attribute const): Fix a typo.
12042 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12043 issuing -Wsuggest-attribute for void functions.
12045 2018-01-03 Martin Sebor <msebor@redhat.com>
12047 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12048 offset_int::from instead of wide_int::to_shwi.
12049 (maybe_diag_overlap): Remove assertion.
12050 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12051 * gimple-ssa-sprintf.c (format_directive): Same.
12052 (parse_directive): Same.
12053 (sprintf_dom_walker::compute_format_length): Same.
12054 (try_substitute_return_value): Same.
12056 2018-01-03 Jeff Law <law@redhat.com>
12058 PR middle-end/83654
12059 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12060 non-constant residual for zero at runtime and avoid probing in
12061 that case. Reorganize code for trailing problem to mirror handling
12064 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12066 PR tree-optimization/83501
12067 * tree-ssa-strlen.c (get_string_cst): New.
12068 (handle_char_store): Call get_string_cst.
12070 2018-01-03 Martin Liska <mliska@suse.cz>
12072 PR tree-optimization/83593
12073 * tree-ssa-strlen.c: Include tree-cfg.h.
12074 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12075 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12076 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12078 (strlen_dom_walker::before_dom_children): Call
12079 gimple_purge_dead_eh_edges. Dump tranformation with details
12081 (strlen_dom_walker::before_dom_children): Update call by adding
12082 new argument cleanup_eh.
12083 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12085 2018-01-03 Martin Liska <mliska@suse.cz>
12088 * cif-code.def (VARIADIC_THUNK): New enum value.
12089 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12092 2018-01-03 Jan Beulich <jbeulich@suse.com>
12094 * sse.md (mov<mode>_internal): Tighten condition for when to use
12095 vmovdqu<ssescalarsize> for TI and OI modes.
12097 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12099 Update copyright years.
12101 2018-01-03 Martin Liska <mliska@suse.cz>
12104 * ipa-visibility.c (function_and_variable_visibility): Skip
12105 functions with noipa attribure.
12107 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12109 * gcc.c (process_command): Update copyright notice dates.
12110 * gcov-dump.c (print_version): Ditto.
12111 * gcov.c (print_version): Ditto.
12112 * gcov-tool.c (print_version): Ditto.
12113 * gengtype.c (create_file): Ditto.
12114 * doc/cpp.texi: Bump @copying's copyright year.
12115 * doc/cppinternals.texi: Ditto.
12116 * doc/gcc.texi: Ditto.
12117 * doc/gccint.texi: Ditto.
12118 * doc/gcov.texi: Ditto.
12119 * doc/install.texi: Ditto.
12120 * doc/invoke.texi: Ditto.
12122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12124 * vector-builder.h (vector_builder::m_full_nelts): Change from
12125 unsigned int to poly_uint64.
12126 (vector_builder::full_nelts): Update prototype accordingly.
12127 (vector_builder::new_vector): Likewise.
12128 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12129 (vector_builder::operator ==): Likewise.
12130 (vector_builder::finalize): Likewise.
12131 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12132 Take the number of elements as a poly_uint64 rather than an
12134 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12135 from unsigned int to poly_uint64.
12136 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12137 (vec_perm_indices::new_vector): Likewise.
12138 (vec_perm_indices::length): Likewise.
12139 (vec_perm_indices::nelts_per_input): Likewise.
12140 (vec_perm_indices::input_nelts): Likewise.
12141 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12142 number of elements per input as a poly_uint64 rather than an
12143 unsigned int. Use the original encoding for variable-length
12144 vectors, rather than clamping each individual element.
12145 For the second and subsequent elements in each pattern,
12146 clamp the step and base before clamping their sum.
12147 (vec_perm_indices::series_p): Handle polynomial element counts.
12148 (vec_perm_indices::all_in_range_p): Likewise.
12149 (vec_perm_indices_to_tree): Likewise.
12150 (vec_perm_indices_to_rtx): Likewise.
12151 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12152 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12153 (tree_vector_builder::new_binary_operation): Handle polynomial
12154 element counts. Return false if we need to know the number
12155 of elements at compile time.
12156 * fold-const.c (fold_vec_perm): Punt if the number of elements
12157 isn't known at compile time.
12159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12161 * vec-perm-indices.h (vec_perm_builder): Change element type
12162 from HOST_WIDE_INT to poly_int64.
12163 (vec_perm_indices::element_type): Update accordingly.
12164 (vec_perm_indices::clamp): Handle polynomial element_types.
12165 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12166 (vec_perm_indices::all_in_range_p): Likewise.
12167 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12169 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12170 polynomial vec_perm_indices element types.
12171 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12172 * fold-const.c (fold_vec_perm): Likewise.
12173 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12174 * tree-vect-generic.c (lower_vec_perm): Likewise.
12175 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12176 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12177 element type to HOST_WIDE_INT.
12179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12180 Alan Hayward <alan.hayward@arm.com>
12181 David Sherwood <david.sherwood@arm.com>
12183 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12184 rather than an int. Use plus_constant.
12185 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12186 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12189 Alan Hayward <alan.hayward@arm.com>
12190 David Sherwood <david.sherwood@arm.com>
12192 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12193 a HOST_WIDE_INT to a poly_int64.
12195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12196 Alan Hayward <alan.hayward@arm.com>
12197 David Sherwood <david.sherwood@arm.com>
12199 * calls.c (load_register_parameters): Cope with polynomial
12200 mode sizes. Require a constant size for BLKmode parameters
12201 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12202 forces a parameter to be padded at the lsb end in order to
12203 fill a complete number of words, require the parameter size
12204 to be ordered wrt UNITS_PER_WORD.
12206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12207 Alan Hayward <alan.hayward@arm.com>
12208 David Sherwood <david.sherwood@arm.com>
12210 * reload1.c (spill_stack_slot_width): Change element type
12211 from unsigned int to poly_uint64_pod.
12212 (alter_reg): Treat mode sizes as polynomial.
12214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12215 Alan Hayward <alan.hayward@arm.com>
12216 David Sherwood <david.sherwood@arm.com>
12218 * reload.c (complex_word_subreg_p): New function.
12219 (reload_inner_reg_of_subreg, push_reload): Use it.
12221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12222 Alan Hayward <alan.hayward@arm.com>
12223 David Sherwood <david.sherwood@arm.com>
12225 * lra-constraints.c (process_alt_operands): Reject matched
12226 operands whose sizes aren't ordered.
12227 (match_reload): Refer to this check here.
12229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12230 Alan Hayward <alan.hayward@arm.com>
12231 David Sherwood <david.sherwood@arm.com>
12233 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12234 that the mode size is in the set {1, 2, 4, 8, 16}.
12236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12237 Alan Hayward <alan.hayward@arm.com>
12238 David Sherwood <david.sherwood@arm.com>
12240 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12241 Use plus_constant instead of gen_rtx_PLUS.
12243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12244 Alan Hayward <alan.hayward@arm.com>
12245 David Sherwood <david.sherwood@arm.com>
12247 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12248 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12249 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12250 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12251 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12252 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12253 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12254 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12255 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12256 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12258 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12259 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12260 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12261 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12262 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12263 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12264 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12265 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12266 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12267 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12269 * expr.c (emit_move_resolve_push): Treat the input and result
12270 of PUSH_ROUNDING as a poly_int64.
12271 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12272 (emit_push_insn): Likewise.
12273 * lra-eliminations.c (mark_not_eliminable): Likewise.
12274 * recog.c (push_operand): Likewise.
12275 * reload1.c (elimination_effects): Likewise.
12276 * rtlanal.c (nonzero_bits1): Likewise.
12277 * calls.c (store_one_arg): Likewise. Require the padding to be
12278 known at compile time.
12280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12281 Alan Hayward <alan.hayward@arm.com>
12282 David Sherwood <david.sherwood@arm.com>
12284 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12285 Use plus_constant instead of gen_rtx_PLUS.
12287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12288 Alan Hayward <alan.hayward@arm.com>
12289 David Sherwood <david.sherwood@arm.com>
12291 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12292 rather than an int.
12294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12295 Alan Hayward <alan.hayward@arm.com>
12296 David Sherwood <david.sherwood@arm.com>
12298 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12299 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12300 via stack temporaries. Treat the mode size as polynomial too.
12302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12303 Alan Hayward <alan.hayward@arm.com>
12304 David Sherwood <david.sherwood@arm.com>
12306 * expr.c (expand_expr_real_2): When handling conversions involving
12307 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12308 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12309 as a poly_uint64 too.
12311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12312 Alan Hayward <alan.hayward@arm.com>
12313 David Sherwood <david.sherwood@arm.com>
12315 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12318 Alan Hayward <alan.hayward@arm.com>
12319 David Sherwood <david.sherwood@arm.com>
12321 * combine.c (can_change_dest_mode): Handle polynomial
12322 REGMODE_NATURAL_SIZE.
12323 * expmed.c (store_bit_field_1): Likewise.
12324 * expr.c (store_constructor): Likewise.
12325 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12326 and polynomial REGMODE_NATURAL_SIZE.
12327 (gen_lowpart_common): Likewise.
12328 * reginfo.c (record_subregs_of_mode): Likewise.
12329 * rtlanal.c (read_modify_subreg_p): Likewise.
12331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12332 Alan Hayward <alan.hayward@arm.com>
12333 David Sherwood <david.sherwood@arm.com>
12335 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12336 numbers of elements.
12338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12339 Alan Hayward <alan.hayward@arm.com>
12340 David Sherwood <david.sherwood@arm.com>
12342 * match.pd: Cope with polynomial numbers of vector elements.
12344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12345 Alan Hayward <alan.hayward@arm.com>
12346 David Sherwood <david.sherwood@arm.com>
12348 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12349 in a POINTER_PLUS_EXPR.
12351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12352 Alan Hayward <alan.hayward@arm.com>
12353 David Sherwood <david.sherwood@arm.com>
12355 * omp-simd-clone.c (simd_clone_subparts): New function.
12356 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12357 (ipa_simd_modify_function_body): Likewise.
12359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12360 Alan Hayward <alan.hayward@arm.com>
12361 David Sherwood <david.sherwood@arm.com>
12363 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12364 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12365 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12366 (expand_vector_condition, vector_element): Likewise.
12367 (subparts_gt): New function.
12368 (get_compute_type): Use subparts_gt.
12369 (count_type_subparts): Delete.
12370 (expand_vector_operations_1): Use subparts_gt instead of
12371 count_type_subparts.
12373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12374 Alan Hayward <alan.hayward@arm.com>
12375 David Sherwood <david.sherwood@arm.com>
12377 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12378 (vect_compile_time_alias): ...this new function. Do the calculation
12379 on poly_ints rather than trees.
12380 (vect_prune_runtime_alias_test_list): Update call accordingly.
12382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12383 Alan Hayward <alan.hayward@arm.com>
12384 David Sherwood <david.sherwood@arm.com>
12386 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12388 (vect_schedule_slp_instance): Likewise.
12390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12391 Alan Hayward <alan.hayward@arm.com>
12392 David Sherwood <david.sherwood@arm.com>
12394 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12395 constant and extern definitions for variable-length vectors.
12396 (vect_get_constant_vectors): Note that the number of units
12397 is known to be constant.
12399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12400 Alan Hayward <alan.hayward@arm.com>
12401 David Sherwood <david.sherwood@arm.com>
12403 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12404 of units as polynomial. Choose between WIDE and NARROW based
12407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12408 Alan Hayward <alan.hayward@arm.com>
12409 David Sherwood <david.sherwood@arm.com>
12411 * tree-vect-stmts.c (simd_clone_subparts): New function.
12412 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12415 Alan Hayward <alan.hayward@arm.com>
12416 David Sherwood <david.sherwood@arm.com>
12418 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12419 vectors as polynomial. Use build_index_vector for
12420 IFN_GOMP_SIMD_LANE.
12422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12423 Alan Hayward <alan.hayward@arm.com>
12424 David Sherwood <david.sherwood@arm.com>
12426 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12427 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12428 for variable-length vectors.
12429 (vectorizable_mask_load_store): Treat the number of units as
12430 polynomial, asserting that it is constant if the condition has
12431 already been enforced.
12432 (vectorizable_store, vectorizable_load): Likewise.
12434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12435 Alan Hayward <alan.hayward@arm.com>
12436 David Sherwood <david.sherwood@arm.com>
12438 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12439 of units as polynomial. Punt if we can't tell at compile time
12440 which vector contains the final result.
12442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12443 Alan Hayward <alan.hayward@arm.com>
12444 David Sherwood <david.sherwood@arm.com>
12446 * tree-vect-loop.c (vectorizable_induction): Treat the number
12447 of units as polynomial. Punt on SLP inductions. Use an integer
12448 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12449 cast of such a series for variable-length floating-point
12452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12453 Alan Hayward <alan.hayward@arm.com>
12454 David Sherwood <david.sherwood@arm.com>
12456 * tree.h (build_index_vector): Declare.
12457 * tree.c (build_index_vector): New function.
12458 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12459 of units as polynomial, forcibly converting it to a constant if
12460 vectorizable_reduction has already enforced the condition.
12461 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12462 to create a {1,2,3,...} vector.
12463 (vectorizable_reduction): Treat the number of units as polynomial.
12464 Choose vectype_in based on the largest scalar element size rather
12465 than the smallest number of units. Enforce the restrictions
12468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12469 Alan Hayward <alan.hayward@arm.com>
12470 David Sherwood <david.sherwood@arm.com>
12472 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12473 number of units as polynomial.
12475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12476 Alan Hayward <alan.hayward@arm.com>
12477 David Sherwood <david.sherwood@arm.com>
12479 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12480 * target.def (autovectorize_vector_sizes): Return the vector sizes
12481 by pointer, using vector_sizes rather than a bitmask.
12482 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12483 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12484 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12486 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12487 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12488 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12489 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12490 * omp-general.c (omp_max_vf): Likewise.
12491 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12492 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12493 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12494 * tree-vect-slp.c (vect_slp_bb): Likewise.
12495 * doc/tm.texi: Regenerate.
12496 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12498 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12499 the vector size as a poly_uint64 rather than an unsigned int.
12500 (current_vector_size): Change from an unsigned int to a poly_uint64.
12501 (get_vectype_for_scalar_type): Update accordingly.
12502 * tree.h (build_truth_vector_type): Take the size and number of
12503 units as a poly_uint64 rather than an unsigned int.
12504 (build_vector_type): Add a temporary overload that takes
12505 the number of units as a poly_uint64 rather than an unsigned int.
12506 * tree.c (make_vector_type): Likewise.
12507 (build_truth_vector_type): Take the number of units as a poly_uint64
12508 rather than an unsigned int.
12510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12511 Alan Hayward <alan.hayward@arm.com>
12512 David Sherwood <david.sherwood@arm.com>
12514 * target.def (get_mask_mode): Take the number of units and length
12515 as poly_uint64s rather than unsigned ints.
12516 * targhooks.h (default_get_mask_mode): Update accordingly.
12517 * targhooks.c (default_get_mask_mode): Likewise.
12518 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12519 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12526 * omp-general.c (omp_max_vf): Likewise.
12527 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12528 (expand_omp_simd): Handle polynomial safelen.
12529 * omp-low.c (omplow_simd_context): Add a default constructor.
12530 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12531 (lower_rec_simd_input_clauses): Update accordingly.
12532 (lower_rec_input_clauses): Likewise.
12534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12535 Alan Hayward <alan.hayward@arm.com>
12536 David Sherwood <david.sherwood@arm.com>
12538 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12539 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12540 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12541 (vect_analyze_slp_cost): Likewise.
12542 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12543 (vect_model_load_cost): Likewise.
12545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12546 Alan Hayward <alan.hayward@arm.com>
12547 David Sherwood <david.sherwood@arm.com>
12549 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12550 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12551 from an unsigned int * to a poly_uint64_pod *.
12552 (calculate_unrolling_factor): New function.
12553 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12556 Alan Hayward <alan.hayward@arm.com>
12557 David Sherwood <david.sherwood@arm.com>
12559 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12560 from an unsigned int to a poly_uint64.
12561 (_loop_vec_info::slp_unrolling_factor): Likewise.
12562 (_loop_vec_info::vectorization_factor): Change from an int
12564 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12565 (vect_get_num_vectors): New function.
12566 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12567 (vect_get_num_copies): Use vect_get_num_vectors.
12568 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12569 to an unsigned int *.
12570 (vect_analyze_data_refs): Change min_vf from an int * to a
12572 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12573 than an unsigned HOST_WIDE_INT.
12574 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12575 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12576 to an unsigned int *.
12577 (vect_analyze_data_ref_dependences): Likewise.
12578 (vect_compute_data_ref_alignment): Handle polynomial vf.
12579 (vect_enhance_data_refs_alignment): Likewise.
12580 (vect_prune_runtime_alias_test_list): Likewise.
12581 (vect_shift_permute_load_chain): Likewise.
12582 (vect_supportable_dr_alignment): Likewise.
12583 (dependence_distance_ge_vf): Take the vectorization factor as a
12584 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12585 (vect_analyze_data_refs): Change min_vf from an int * to a
12587 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12588 vfm1 as a poly_uint64 rather than an int. Make the same change
12589 for the returned bound_scalar.
12590 (vect_gen_vector_loop_niters): Handle polynomial vf.
12591 (vect_do_peeling): Likewise. Update call to
12592 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12593 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12595 * tree-vect-loop.c (vect_determine_vectorization_factor)
12596 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12597 (vect_get_known_peeling_cost): Likewise.
12598 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12599 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12600 (vect_transform_loop): Likewise. Use the lowest possible VF when
12601 updating the upper bounds of the loop.
12602 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12603 rather than an int.
12604 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12605 polynomial unroll factors.
12606 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12607 (vect_make_slp_decision): Likewise.
12608 (vect_supported_load_permutation_p): Likewise, and polynomial
12610 (vect_analyze_slp_cost): Handle polynomial vf.
12611 (vect_slp_analyze_node_operations): Likewise.
12612 (vect_slp_analyze_bb_1): Likewise.
12613 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12614 than an unsigned HOST_WIDE_INT.
12615 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12616 (vectorizable_load): Handle polynomial vf.
12617 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12619 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12622 Alan Hayward <alan.hayward@arm.com>
12623 David Sherwood <david.sherwood@arm.com>
12625 * match.pd: Handle bit operations involving three constants
12626 and try to fold one pair.
12628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12630 * tree-vect-loop-manip.c: Include gimple-fold.h.
12631 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12632 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12633 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12634 Add a path that uses a step of VF instead of 1, but disable it
12636 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12637 and niters_no_overflow parameters. Update calls to
12638 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12639 Create a new SSA name if the latter choses to use a ste other
12640 than zero, and return it via niters_vector_mult_vf_var.
12641 * tree-vect-loop.c (vect_transform_loop): Update calls to
12642 vect_do_peeling, vect_gen_vector_loop_niters and
12643 slpeel_make_loop_iterate_ntimes.
12644 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12645 (vect_gen_vector_loop_niters): Update declarations after above changes.
12647 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12649 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12650 128-bit round to integer instructions.
12651 (ceil<mode>2): Likewise.
12652 (btrunc<mode>2): Likewise.
12653 (round<mode>2): Likewise.
12655 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12657 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12658 unaligned VSX load/store on P8/P9.
12659 (expand_block_clear): Allow the use of unaligned VSX
12660 load/store on P8/P9.
12662 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12664 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12666 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12667 swap associated with both a load and a store.
12669 2018-01-02 Andrew Waterman <andrew@sifive.com>
12671 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12672 * config/riscv/riscv.md (clear_cache): Use it.
12674 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12676 * web.c: Remove out-of-date comment.
12678 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12680 * expr.c (fixup_args_size_notes): Check that any existing
12681 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12682 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12683 (emit_single_push_insn): ...here.
12685 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12687 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12688 (const_vector_encoded_nelts): New function.
12689 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12690 (const_vector_int_elt, const_vector_elt): Declare.
12691 * emit-rtl.c (const_vector_int_elt_1): New function.
12692 (const_vector_elt): Likewise.
12693 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12694 of CONST_VECTOR_ELT.
12696 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12698 * expr.c: Include rtx-vector-builder.h.
12699 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12700 directly on the tree encoding.
12701 (const_vector_from_tree): Likewise.
12702 * optabs.c: Include rtx-vector-builder.h.
12703 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12704 sequence of "u" values.
12705 * vec-perm-indices.c: Include rtx-vector-builder.h.
12706 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12707 directly on the vec_perm_indices encoding.
12709 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12711 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12712 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12713 * rtx-vector-builder.h: New file.
12714 * rtx-vector-builder.c: Likewise.
12715 * rtl.h (rtx_def::u2): Add a const_vector field.
12716 (CONST_VECTOR_NPATTERNS): New macro.
12717 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12718 (CONST_VECTOR_DUPLICATE_P): Likewise.
12719 (CONST_VECTOR_STEPPED_P): Likewise.
12720 (CONST_VECTOR_ENCODED_ELT): Likewise.
12721 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12722 (unwrap_const_vec_duplicate): Likewise.
12723 (const_vec_series_p): Check for a non-duplicated vector encoding.
12724 Say that the function only returns true for integer vectors.
12725 * emit-rtl.c: Include rtx-vector-builder.h.
12726 (gen_const_vec_duplicate_1): Delete.
12727 (gen_const_vector): Call gen_const_vec_duplicate instead of
12728 gen_const_vec_duplicate_1.
12729 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12730 (gen_const_vec_duplicate): Use rtx_vector_builder.
12731 (gen_const_vec_series): Likewise.
12732 (gen_rtx_CONST_VECTOR): Likewise.
12733 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12734 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12735 Build a new vector rather than modifying a CONST_VECTOR in-place.
12736 (handle_special_swappables): Update call accordingly.
12737 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12738 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12739 Build a new vector rather than modifying a CONST_VECTOR in-place.
12740 (handle_special_swappables): Update call accordingly.
12742 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12744 * simplify-rtx.c (simplify_const_binary_operation): Use
12745 CONST_VECTOR_ELT instead of XVECEXP.
12747 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12749 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12750 the selector elements to be different from the data elements
12751 if the selector is a VECTOR_CST.
12752 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12753 ssizetype for the selector.
12755 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12757 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12758 before testing each element individually.
12759 * tree-vect-generic.c (lower_vec_perm): Likewise.
12761 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12763 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12764 * selftest-run-tests.c (selftest::run_tests): Call it.
12765 * vector-builder.h (vector_builder::operator ==): New function.
12766 (vector_builder::operator !=): Likewise.
12767 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12768 (vec_perm_indices::all_from_input_p): New function.
12769 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12770 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12771 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12772 instead of reading the VECTOR_CST directly. Detect whether both
12773 vector inputs are the same before constructing the vec_perm_indices,
12774 and update the number of inputs argument accordingly. Use the
12775 utility functions added above. Only construct sel2 if we need to.
12777 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12779 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12780 the broadcast of the low byte.
12781 (expand_mult_highpart): Use an explicit encoding for the permutes.
12782 * optabs-query.c (can_mult_highpart_p): Likewise.
12783 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12784 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12785 (vectorizable_bswap): Likewise.
12786 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12787 explicit encoding for the power-of-2 permutes.
12788 (vect_permute_store_chain): Likewise.
12789 (vect_grouped_load_supported): Likewise.
12790 (vect_permute_load_chain): Likewise.
12792 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12794 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12795 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12796 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12797 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12798 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12799 (vect_gen_perm_mask_any): Likewise.
12801 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12803 * int-vector-builder.h: New file.
12804 * vec-perm-indices.h: Include int-vector-builder.h.
12805 (vec_perm_indices): Redefine as an int_vector_builder.
12806 (auto_vec_perm_indices): Delete.
12807 (vec_perm_builder): Redefine as a stand-alone class.
12808 (vec_perm_indices::vec_perm_indices): New function.
12809 (vec_perm_indices::clamp): Likewise.
12810 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12811 (vec_perm_indices::new_vector): New function.
12812 (vec_perm_indices::new_expanded_vector): Update for new
12813 vec_perm_indices class.
12814 (vec_perm_indices::rotate_inputs): New function.
12815 (vec_perm_indices::all_in_range_p): Operate directly on the
12816 encoded form, without computing elided elements.
12817 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12818 encoding. Update for new vec_perm_indices class.
12819 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12820 the given vec_perm_builder.
12821 (expand_vec_perm_var): Update vec_perm_builder constructor.
12822 (expand_mult_highpart): Use vec_perm_builder instead of
12823 auto_vec_perm_indices.
12824 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12825 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12826 or double series encoding as appropriate.
12827 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12828 vec_perm_indices instead of auto_vec_perm_indices.
12829 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12830 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12831 (vect_permute_store_chain): Likewise.
12832 (vect_grouped_load_supported): Likewise.
12833 (vect_permute_load_chain): Likewise.
12834 (vect_shift_permute_load_chain): Likewise.
12835 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12836 (vect_transform_slp_perm_load): Likewise.
12837 (vect_schedule_slp_instance): Likewise.
12838 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12839 (vectorizable_mask_load_store): Likewise.
12840 (vectorizable_bswap): Likewise.
12841 (vectorizable_store): Likewise.
12842 (vectorizable_load): Likewise.
12843 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12844 vec_perm_indices instead of auto_vec_perm_indices. Use
12845 tree_to_vec_perm_builder to read the vector from a tree.
12846 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12847 vec_perm_builder instead of a vec_perm_indices.
12848 (have_whole_vector_shift): Use vec_perm_builder and
12849 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12850 truncation to calc_vec_perm_mask_for_shift.
12851 (vect_create_epilog_for_reduction): Likewise.
12852 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12853 from auto_vec_perm_indices to vec_perm_indices.
12854 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12855 instead of changing individual elements.
12856 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12857 the vector in d.perm.
12858 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12859 from auto_vec_perm_indices to vec_perm_indices.
12860 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12861 instead of changing individual elements.
12862 (arm_vectorize_vec_perm_const): Use new_vector to install
12863 the vector in d.perm.
12864 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12865 Update vec_perm_builder constructor.
12866 (rs6000_expand_interleave): Likewise.
12867 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12868 (rs6000_expand_interleave): Likewise.
12870 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12872 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12873 to qimode could truncate the indices.
12874 * optabs.c (expand_vec_perm_var): Likewise.
12876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12878 * Makefile.in (OBJS): Add vec-perm-indices.o.
12879 * vec-perm-indices.h: New file.
12880 * vec-perm-indices.c: Likewise.
12881 * target.h (vec_perm_indices): Replace with a forward class
12883 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12884 * optabs.h: Include vec-perm-indices.h.
12885 (expand_vec_perm): Delete.
12886 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12887 (expand_vec_perm_const): Declare.
12888 * target.def (vec_perm_const_ok): Replace with...
12889 (vec_perm_const): ...this new hook.
12890 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12891 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12892 * doc/tm.texi: Regenerate.
12893 * optabs.def (vec_perm_const): Delete.
12894 * doc/md.texi (vec_perm_const): Likewise.
12895 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12896 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12897 expand_vec_perm for constant permutation vectors. Assert that
12898 the mode of variable permutation vectors is the integer equivalent
12899 of the mode that is being permuted.
12900 * optabs-query.h (selector_fits_mode_p): Declare.
12901 * optabs-query.c: Include vec-perm-indices.h.
12902 (selector_fits_mode_p): New function.
12903 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12904 is defined, instead of checking whether the vec_perm_const_optab
12905 exists. Use targetm.vectorize.vec_perm_const instead of
12906 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12907 fit in the vector mode before using a variable permute.
12908 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12909 vec_perm_indices instead of an rtx.
12910 (expand_vec_perm): Replace with...
12911 (expand_vec_perm_const): ...this new function. Take the selector
12912 as a vec_perm_indices rather than an rtx. Also take the mode of
12913 the selector. Update call to shift_amt_for_vec_perm_mask.
12914 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12915 Use vec_perm_indices::new_expanded_vector to expand the original
12916 selector into bytes. Check whether the indices fit in the vector
12917 mode before using a variable permute.
12918 (expand_vec_perm_var): Make global.
12919 (expand_mult_highpart): Use expand_vec_perm_const.
12920 * fold-const.c: Includes vec-perm-indices.h.
12921 * tree-ssa-forwprop.c: Likewise.
12922 * tree-vect-data-refs.c: Likewise.
12923 * tree-vect-generic.c: Likewise.
12924 * tree-vect-loop.c: Likewise.
12925 * tree-vect-slp.c: Likewise.
12926 * tree-vect-stmts.c: Likewise.
12927 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12929 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12930 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12931 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12932 (aarch64_vectorize_vec_perm_const): ...this new function.
12933 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12934 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12935 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12936 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12937 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12938 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12939 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12941 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12942 check for NEON modes.
12943 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12944 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12945 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12946 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12948 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12949 the old VEC_PERM_CONST conditions.
12950 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12951 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12952 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12953 (ia64_vectorize_vec_perm_const_ok): Merge into...
12954 (ia64_vectorize_vec_perm_const): ...this new function.
12955 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12956 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12957 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12958 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12959 * config/mips/mips.c (mips_expand_vec_perm_const)
12960 (mips_vectorize_vec_perm_const_ok): Merge into...
12961 (mips_vectorize_vec_perm_const): ...this new function.
12962 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12963 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12964 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12965 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12966 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12967 (rs6000_expand_vec_perm_const): Delete.
12968 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12970 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12971 (altivec_expand_vec_perm_const_le): Take each operand individually.
12972 Operate on constant selectors rather than rtxes.
12973 (altivec_expand_vec_perm_const): Likewise. Update call to
12974 altivec_expand_vec_perm_const_le.
12975 (rs6000_expand_vec_perm_const): Delete.
12976 (rs6000_vectorize_vec_perm_const_ok): Delete.
12977 (rs6000_vectorize_vec_perm_const): New function.
12978 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12979 an element count and rtx array.
12980 (rs6000_expand_extract_even): Update call accordingly.
12981 (rs6000_expand_interleave): Likewise.
12982 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12983 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12984 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12985 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12986 (rs6000_expand_vec_perm_const): Delete.
12987 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12988 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12989 (altivec_expand_vec_perm_const_le): Take each operand individually.
12990 Operate on constant selectors rather than rtxes.
12991 (altivec_expand_vec_perm_const): Likewise. Update call to
12992 altivec_expand_vec_perm_const_le.
12993 (rs6000_expand_vec_perm_const): Delete.
12994 (rs6000_vectorize_vec_perm_const_ok): Delete.
12995 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12996 reference to the SPE evmerge intructions.
12997 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12998 an element count and rtx array.
12999 (rs6000_expand_extract_even): Update call accordingly.
13000 (rs6000_expand_interleave): Likewise.
13001 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13002 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13004 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13006 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13008 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13009 vector mode and that that mode matches the mode of the data
13011 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13012 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13013 directly using expand_vec_perm_1 when forcing selectors into
13015 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13017 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13019 * optabs-query.h (can_vec_perm_p): Delete.
13020 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13021 * optabs-query.c (can_vec_perm_p): Split into...
13022 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13023 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13024 particular selector is valid.
13025 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13026 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13027 (vect_grouped_load_supported): Likewise.
13028 (vect_shift_permute_load_chain): Likewise.
13029 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13030 (vect_transform_slp_perm_load): Likewise.
13031 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13032 (vectorizable_bswap): Likewise.
13033 (vect_gen_perm_mask_checked): Likewise.
13034 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13035 implementations of variable permutation vectors into account
13036 when deciding which selector to use.
13037 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13038 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13039 with a false third argument.
13040 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13041 to test whether the constant selector is valid and can_vec_perm_var_p
13042 to test whether a variable selector is valid.
13044 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13046 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13047 * optabs-query.c (can_vec_perm_p): Likewise.
13048 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13049 instead of vec_perm_indices.
13050 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13051 (vect_gen_perm_mask_checked): Likewise,
13052 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13053 (vect_gen_perm_mask_checked): Likewise,
13055 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13057 * optabs-query.h (qimode_for_vec_perm): Declare.
13058 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13059 (qimode_for_vec_perm): ...this new function.
13060 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13062 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13064 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13065 does not have a conditional at the top.
13067 2018-01-02 Richard Biener <rguenther@suse.de>
13069 * ipa-inline.c (big_speedup_p): Fix expression.
13071 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13074 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13077 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13081 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13082 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13083 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13084 cond_taken_branch_cost 3->4.
13086 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13088 PR tree-optimization/83581
13089 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13090 TODO_cleanup_cfg if any changes have been made.
13092 PR middle-end/83608
13093 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13094 convert_modes if target mode has the right side, but different mode
13097 PR middle-end/83609
13098 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13099 last argument when extracting from CONCAT. If either from_real or
13100 from_imag is NULL, use expansion through memory. If result is not
13101 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13102 the parts directly to inner mode, if even that fails, use expansion
13105 PR middle-end/83623
13106 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13107 check for bswap in mode rather than HImode and use that in expand_unop
13110 Copyright (C) 2018 Free Software Foundation, Inc.
13112 Copying and distribution of this file, with or without modification,
13113 are permitted in any medium without royalty provided the copyright
13114 notice and this notice are preserved.