1 2018-05-17 Martin Liska <mliska@suse.cz>
3 * gimple-ssa-sprintf.c (format_directive): Do not use
4 space in between 'G_' and '('.
6 2018-05-17 Jakub Jelinek <jakub@redhat.com>
9 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
10 even if the mask is not all ones.
13 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
15 (ix86_gimple_fold_builtin): Likewise.
18 * config/i386/i386.c: Include tree-vector-builder.h.
19 (ix86_vector_shift_count): New function.
20 (ix86_fold_builtin): Fold shift builtins by scalar count.
21 (ix86_gimple_fold_builtin): Likewise.
23 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
24 _mm512_setzero): New intrinsics.
26 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
30 code generation for cases where splatting a value is not useful.
31 * simplify-rtx.c (simplify_ternary_operation): Simplify
32 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
35 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
37 * config.gcc: Support "goldmont-plus".
38 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
40 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
41 PROCESSOR_GOLDMONT_PLUS.
42 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
43 (processor_target_table): Add "goldmont-plus".
44 (PTA_GOLDMONT_PLUS): Define.
45 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
46 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
47 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
48 (fold_builtin_cpu): Add "goldmont-plus".
49 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
50 (ix86_option_override_internal): Add "goldmont-plus".
51 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
52 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
53 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
54 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
56 2018-05-17 Richard Biener <rguenther@suse.de>
58 PR tree-optimization/85757
59 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
60 remove defs that only feed that PHI from further processing.
62 2018-05-16 Jim Wilson <jimw@sifive.com>
64 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
66 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
68 2018-05-16 Mark Wielaard <mark@klomp.org>
70 * dwarf2out.c (count_index_strings): New function.
71 (output_indirect_strings): Call count_index_strings and generate
72 header for dwarf_version >= 5.
74 2018-05-16 Mark Wielaard <mark@klomp.org>
76 * dwarf2out.c (dwarf_FORM): New function.
77 (set_indirect_string): Use dwarf_FORM.
78 (reset_indirect_string): Likewise.
79 (size_of_die): Likewise.
80 (value_format): Likewise.
81 (output_die): Likewise.
82 (add_skeleton_AT_string): Likewise.
83 (output_macinfo_op): Likewise.
84 (index_string): Likewise.
85 (output_index_string_offset): Likewise.
86 (output_index_string): Likewise.
87 (count_index_strings): Likewise.
89 2018-05-16 Carl Love <cel@us.ibm.com>
91 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
92 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
94 2018-05-16 Martin Jambor <mjambor@suse.cz>
96 * ipa-prop.c (ipa_free_all_edge_args): Remove.
97 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
99 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
101 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
102 (fnma<mode>4): Likewise.
103 (fms<mode>4): Likewise.
104 (fnms<mode>4): Likewise.
105 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
106 (aarch64_fnma<mode>4): Likewise.
107 (aarch64_fms<mode>4): Likewise.
108 (aarch64_fnms<mode>4): Likewise.
109 (aarch64_fnmadd<mode>4): Likewise.
111 2018-05-16 Jason Merrill <jason@redhat.com>
113 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
115 2018-05-16 Richard Biener <rguenther@suse.de>
117 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
118 (dump_stmt_cost): Declare.
119 (add_stmt_cost): Dump cost we add.
120 (add_stmt_costs): New function.
121 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
123 (vect_analyze_stmt): Adjust prototype.
124 (vectorizable_condition): Likewise.
125 (vectorizable_live_operation): Likewise.
126 (vectorizable_reduction): Likewise.
127 (vectorizable_induction): Likewise.
128 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
129 cost vector to pass to vectorizable_ and record afterwards.
130 (vect_model_reduction_cost): Take cost vector argument and adjust.
131 (vect_model_induction_cost): Likewise.
132 (vectorizable_reduction): Likewise.
133 (vectorizable_induction): Likewise.
134 (vectorizable_live_operation): Likewise.
135 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
136 SLP_TREE_NUMBER_OF_VEC_STMTS.
137 (vect_analyze_slp_cost_1): Remove.
138 (vect_analyze_slp_cost): Likewise.
139 (vect_slp_analyze_node_operations): Take visited args and
140 a target cost vector. Avoid processing already visited stmt sets.
141 (vect_slp_analyze_operations): Use a local cost vector to gather
142 costs and register those of non-discarded instances.
143 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
144 (vect_schedule_slp_instance): Remove copying of
145 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
147 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
148 adding cost. Record cost entry location.
149 (vect_prologue_cost_for_slp_op): Function to compute cost of
150 a constant or invariant generated for SLP vect in the prologue,
151 split out from vect_analyze_slp_cost_1.
152 (vect_model_simple_cost): Make static. Adjust for SLP costing.
153 (vect_model_promotion_demotion_cost): Likewise.
154 (vect_model_store_cost): Likewise, make static.
155 (vect_model_load_cost): Likewise.
156 (vectorizable_bswap): Add cost vector arg and adjust.
157 (vectorizable_call): Likewise.
158 (vectorizable_simd_clone_call): Likewise.
159 (vectorizable_conversion): Likewise.
160 (vectorizable_assignment): Likewise.
161 (vectorizable_shift): Likewise.
162 (vectorizable_operation): Likewise.
163 (vectorizable_store): Likewise.
164 (vectorizable_load): Likewise.
165 (vectorizable_condition): Likewise.
166 (vectorizable_comparison): Likewise.
167 (can_vectorize_live_stmts): Likewise.
168 (vect_analyze_stmt): Likewise.
169 (vect_transform_stmt): Adjust calls to vectorizable_*.
170 * tree-vectorizer.c: Include gimple-pretty-print.h.
171 (dump_stmt_cost): New function.
173 2018-05-16 Richard Biener <rguenther@suse.de>
175 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
176 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
177 * tree-ssa-dse.c: Include tree-ssa-loop.h.
178 (check_name): New callback.
179 (dse_classify_store): Track cycles via a visited bitmap of PHI
180 defs and simplify handling of in-loop and across loop dead stores
181 and properly fail for loop-variant refs. Handle byte-tracking with
182 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
185 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
187 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
188 (vect_get_mask_type_for_stmt): Likewise.
189 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
191 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
192 to determine the statement's vector type and the vector type that
193 should be used for calculating nunits. Deal with cases in which
194 the type has to be deferred.
195 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
196 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
197 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
198 (vect_determine_vf_for_stmt): New functions, split out from...
199 (vect_determine_vectorization_factor): ...here.
200 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
201 (vect_get_mask_type_for_stmt): New functions, split out from
202 vect_determine_vectorization_factor.
204 2018-05-16 Richard Biener <rguenther@suse.de>
206 * tree-cfg.c (verify_gimple_assign_ternary): Properly
207 verify the [VEC_]COND_EXPR embedded comparison.
209 2018-05-15 Martin Sebor <msebor@redhat.com>
211 PR tree-optimization/85753
212 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
213 RECORD_TYPE in addition to ARRAY_TYPE.
215 2018-05-15 Martin Sebor <msebor@redhat.com>
218 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
220 2018-05-15 Richard Biener <rguenther@suse.de>
222 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
223 add by_clobber_p one. Change algorithm to collect all defs
224 representing uses we need to walk and try reducing them to
225 a single one before failing.
226 (dse_dom_walker::dse_optimize_stmt): Adjust.
228 2018-05-13 Mark Wielaard <mark@klomp.org>
230 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
231 (size_of_loc_descr): Likewise.
232 (output_loc_operands): Likewise.
233 (output_loc_operands_raw): Likewise.
234 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
235 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
236 (hash_loc_operands): Likewise.
237 (compare_loc_operands): Likewise.
239 2018-05-14 Mark Wielaard <mark@klomp.org>
241 * dwarf2out.c (count_index_addrs): New function.
242 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
244 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
246 PR tree-optimization/83648
247 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
248 return value as malloc candidate.
250 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
253 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
254 as true in call to suggest_attribute.
256 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
258 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
261 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
263 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
265 (aarch64_crypto_aesd_fused): Likewise.
267 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
269 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
270 (movsi_aarch64): Likewise.
271 (load_pairsi): Likewise.
272 (load_pairdi): Likewise.
273 (store_pairsi): Likewise.
274 (store_pairdi): Likewise.
275 (load_pairsf): Likewise.
276 (load_pairdf): Likewise.
277 (store_pairsf): Likewise.
278 (store_pairdf): Likewise.
279 (zero_extend): Likewise.
280 (trunc): Swap alternatives.
281 (fcvt_target): Add '?' to prefer w over r.
283 2018-05-14 Jakub Jelinek <jakub@redhat.com>
286 * config/i386/i386.md: Disallow non-commutative arithmetics in
287 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
288 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
289 in the peephole2 before it.
291 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
293 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
294 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
295 (ix86_handle_option): Handle -mcldemote.
296 * config.gcc: New header.
297 * config/i386/cldemoteintrin.h: New file.
298 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
299 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
301 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
302 OPTION_MASK_ISA_CLDEMOTE.
303 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
304 (ix86_valid_target_attribute_inner_p): Ditto.
305 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
306 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
307 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
308 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
309 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
311 * config/i386/i386.opt: Add -mcldemote.
312 * config/i386/x86intrin.h: New header.
313 * doc/invoke.texi: Add -mcldemote.
315 2018-05-14 Richard Biener <rguenther@suse.de>
317 * doc/match-and-simplify.texi: Adjust :s documentation.
319 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
321 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
322 intended memcpy size.
323 (REORDER_45): Likewise.
325 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
328 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
329 * vec.c (qsort_chk): Use gcc_qsort.
330 * Makefile.in (OBJS-libcommon): Add sort.o.
331 (build/sort.o): New target. Use it...
332 (BUILD_RTL): ... here, and...
333 (build/gencfn-macros): ... here, and...
334 (build/genmatch): ... here.
336 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
337 Chung-Ju Wu <jasonwucj@gmail.com>
339 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
340 * config/nds32/nds32-graywolf.md: New file.
341 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
342 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
344 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
345 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
346 * config/nds32/nds32.md (pipeline_model): Add graywolf.
347 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
348 * config/nds32/pipelines.md: Include n15 settings.
350 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
351 Chung-Ju Wu <jasonwucj@gmail.com>
353 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
354 * config/nds32/nds32-n13.md: New file.
355 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
356 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
358 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
359 * config/nds32/nds32.md (pipeline_model): Add n13.
360 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
361 * config/nds32/pipelines.md: Include n13 settings.
363 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
364 Chung-Ju Wu <jasonwucj@gmail.com>
366 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
367 * config/nds32/nds32-n10.md: New file.
368 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
369 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
371 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
372 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
373 * config/nds32/nds32.md (pipeline_model): Add n10.
374 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
375 * config/nds32/pipelines.md: Include n10 settings.
377 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
378 Kito Cheng <kito.cheng@gmail.com>
379 Chung-Ju Wu <jasonwucj@gmail.com>
381 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
382 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
383 Add enum values for DSP extension instructions.
384 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
386 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
387 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
389 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
390 * config/nds32/nds32-dspext.md: New file for DSP implementation.
391 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
392 * config/nds32/nds32-intrinsic.md: Likewise.
393 * config/nds32/nds32_intrinsic.h: Likewise.
394 * config/nds32/nds32-md-auxiliary.c: Likewise.
395 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
396 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
397 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
398 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
399 * config/nds32/nds32-protos.h: New declarations for DSP extension.
400 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
401 TYPE_DMAC in switch statement.
402 * config/nds32/nds32.c: New checking and implementation for DSP
403 extension instructions.
404 * config/nds32/nds32.h: Likewise.
405 * config/nds32/nds32.md: Likewise.
406 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
407 * config/nds32/predicates.md: Implement new predicates for DSP
410 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
412 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
413 Reformat alternatives and attributes so it is easier to identify
414 which constraints/attributes go with which instruction.
415 (mov<mode>_hardfloat32, FMOVE64): Likewise.
416 (mov<mode>_softfloat32, FMOVE64): Likewise.
417 (mov<mode>_hardfloat64, FMOVE64): Likewise.
418 (mov<mode>_softfloat64, FMOVE64): Likewise.
420 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
422 * doc/extend.texi (PowerPC Built-in Functions): Rename this
424 (Basic PowerPC Built-in Functions): The new name of the
425 subsection previously known as "PowerPC Built-in Functions".
426 (Basic PowerPC Built-in Functions Available on all Configurations):
428 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
429 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
430 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
431 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
433 2018-05-11 Martin Jambor <mjambor@suse.cz>
436 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
439 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
442 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
444 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
446 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
447 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
448 (ix86_handle_option): Handle -mwaitpkg.
449 * config.gcc: New header.
450 * config/i386/cpuid.h (bit_WAITPKG): New bit.
451 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
452 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
454 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
455 OPTION_MASK_ISA_WAITPKG.
456 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
457 (ix86_option_override_internal): Add PTA_WAITPKG.
458 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
459 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
461 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
462 __builtin_ia32_umwait and __builtin_ia32_tpause.
463 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
464 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
465 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
466 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
467 UNSPECV_TPAUSE): New.
468 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
469 * config/i386/i386.opt: Add -mwaitpkg.
470 * config/i386/waitpkgintrin.h: New file.
471 * config/i386/x86intrin.h: New header.
472 * doc/invoke.texi: Add -mwaitpkg.
474 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
477 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
479 (cortex-m0): Use armv6s-m isa.
480 (cortex-m0plus): Likewise.
481 (cortex-m1): Likewise.
482 (cortex-m0.small-multiply): Likewise.
483 (cortex-m0plus.small-multiply): Likewise.
484 (cortex-m1.small-multiply): Likewise.
486 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
487 Jakub Jelinek <jakub@redhat.com>
489 PR tree-optimization/85692
490 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
491 source permute as well.
493 2018-05-11 Martin Liska <mliska@suse.cz>
496 * doc/extend.texi: Document LLVM style format for no_sanitize
499 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
501 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
502 mode_supports_vsx_dform_quad to mode_supports_dq_form.
503 (mode_supports_vsx_dform_quad): Likewise.
504 (mode_supports_vmx_dform): Move these functions to be next to the
505 other mode_supports functions.
506 (mode_supports_dq_form): Likewise.
507 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
508 mode_supports_dq_form.
509 (reg_offset_addressing_ok_p): Likewise.
510 (offsettable_ok_by_alignment): Likewise.
511 (rs6000_legitimate_offset_address_p): Likewise.
512 (legitimate_lo_sum_address_p): Likewise.
513 (rs6000_legitimize_address): Likewise.
514 (rs6000_legitimize_reload_address): Likewise.
515 (rs6000_secondary_reload_inner): Likewise.
516 (rs6000_preferred_reload_class): Likewise.
517 (rs6000_output_move_128bit): Likewise.
519 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
521 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
522 Generate SImode target register for null target.
523 <case IX86_BUILTIN_XGETBV>: Ditto.
524 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
525 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
527 2018-05-10 Carl Love <cel@us.ibm.com>
529 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
530 dcbtt and dcbtstt if operands[2] is 0.
532 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
535 * config/i386/sse.md (usadv64qi): New expander.
537 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
539 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
540 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
541 -maltivec=be support.
542 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
543 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
544 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
545 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
546 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
547 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
548 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
549 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
550 altivec_vsumsws): Adjust.
551 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
552 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
553 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
555 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
556 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
557 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
558 (altivec_lve<VI_char>x): Delete expand.
559 (*altivec_lve<VI_char>x_internal): Rename to...
560 (altivec_lve<VI_char>x): ... this.
561 (altivec_lvxl_<mode>): Delete expand.
562 (*altivec_lvxl_<mode>_internal): Rename to ...
563 (altivec_lvxl_<mode>): ... this.
564 (altivec_stvxl_<mode>): Delete expand.
565 (*altivec_stvxl_<mode>_internal): Rename to ...
566 (altivec_stvxl_<mode>): ... this.
567 (altivec_stve<VI_char>x): Delete expand.
568 (*altivec_stve<VI_char>x_internal): Rename to ...
569 (altivec_stve<VI_char>x): ... this.
570 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
571 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
572 reduc_plus_scal_<mode>): Adjust.
573 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
575 (rs6000_cpu_cpp_builtins): Adjust.
576 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
577 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
578 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
579 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
580 -maltivec=be support.
581 (rs6000_split_vec_extract_var): Adjust.
582 (rs6000_split_v4si_init): Adjust.
583 (swap_selector_for_mode): Delete.
584 (altivec_expand_lvx_be, altivec_expand_stvx_be,
585 altivec_expand_stvex_be): Delete.
586 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
587 -maltivec=be support.
588 (rs6000_gimple_fold_builtin): Ditto.
589 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
591 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
592 (TARGET_DIRECT_MOVE_64BIT): Adjust.
593 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
594 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
595 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
596 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
597 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
598 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
599 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
600 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
601 anonymous split): Adjust.
602 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
603 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
605 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
607 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
608 when --with-gxx-include-dir is also specified.
609 * configure: Regenerate.
611 2018-05-09 Jim Wilson <jimw@sifive.com>
614 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
615 * config/riscv/t-withmultilib: New.
616 * config/riscv/withmultilib.h: New.
617 * doc/install.texi: Document RISC-V --with-multilib-list support.
619 2018-05-09 Richard Biener <rguenther@suse.de>
621 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
623 (vect_bb_vectorization_profitable_p): Adjust. Compute
624 actual scalar cost using the cost vector and the add_stmt_cost
627 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
629 PR rtl-optimization/85645
630 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
631 in the REG_CFA_REGISTER note for LR, don't leave it empty.
633 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
635 PR rtl-optimization/85645
636 * shrink-wrap.c (spread_components): Return a boolean saying if
637 anything was changed.
638 (try_shrink_wrapping_separate): Iterate spread_components until
639 nothing changes anymore.
641 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
643 PR rtl-optimization/85645
644 * regrename.c (build_def_use): Also kill the chains that include the
645 destination of a REG_CFA_REGISTER note.
647 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
649 PR rtl-optimization/85645
650 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
651 insn that has a REG_CFA_REGISTER note.
653 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
655 * cfgexpand.c (expand_clobber): New function.
656 (expand_gimple_stmt_1): Use it.
657 * tree-vect-stmts.c (vect_clobber_variable): New function,
659 (vectorizable_simd_clone_call): ...here.
660 (vectorizable_store): Emit a clobber either side of an
661 IFN_STORE_LANES sequence.
662 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
664 2018-05-09 Tom de Vries <tom@codesourcery.com>
667 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
668 (define_insn "trap_if_false"): Add exit after trap.
670 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
672 PR rtl-optimization/85638
673 * bb-reorder.c: Include common/common-target.h.
674 (create_forwarder_block): New function extracted from...
675 (fix_up_crossing_landing_pad): ...here. Rename into...
676 (dw2_fix_up_crossing_landing_pad): ...this.
677 (sjlj_fix_up_crossing_landing_pad): New function.
678 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
679 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
680 from both partitions and exit the loop after one iteration.
682 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
685 * doc/extend.texi (PowerPC Built-in Functions): Rename this
687 (Basic PowerPC Built-in Functions): The new name of the
688 subsection previously known as "PowerPC Built-in Functions".
689 (Basic PowerPC Built-in Functions Available on all Configurations):
691 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
693 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
695 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
697 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
700 2018-05-08 Jim Wilson <jimw@sifive.com>
702 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
703 (LD_EMUL_SUFFIX): New.
706 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
708 * doc/extend.texi (PowerPC Built-in Functions): Rename this
710 (Basic PowerPC Built-in Functions): The new name of the
711 subsection previously known as "PowerPC Built-in Functions".
712 (Basic PowerPC Built-in Functions Available on all Configurations):
714 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
716 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
718 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
720 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
723 2018-05-08 Jakub Jelinek <jakub@redhat.com>
726 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
727 after cmpelim optimization.
729 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
731 * config.gcc: Support "goldmont".
732 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
733 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
735 * config/i386/i386.c (m_GOLDMONT): Define.
736 (processor_target_table): Add "goldmont".
737 (PTA_GOLDMONT): Define.
738 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
739 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
740 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
741 (fold_builtin_cpu): Add "goldmont".
742 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
743 (ix86_option_override_internal): Add "goldmont".
744 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
745 (processor_type): Add PROCESSOR_GOLDMONT.
746 * config/i386/i386.md: Add CPU "glm".
747 * config/i386/glm.md: New file.
748 * config/i386/x86-tune.def: Add m_GOLDMONT.
749 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
751 2018-05-08 Jakub Jelinek <jakub@redhat.com>
754 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
756 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
757 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
758 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
761 * config/i386/i386.c (ix86_fold_builtin): Handle
762 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
765 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
766 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
768 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
771 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
772 (check_arch): Likewise.
773 (check_fpu): Return the result rather than printing it.
774 (end arch): Fix operator precedence.
776 (END): Print the result from check_fpu.
778 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
779 Alan Hayward <alan.hayward@arm.com>
780 David Sherwood <david.sherwood@arm.com>
782 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
783 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
784 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
785 (*fcmuo<mode>_and): New patterns.
787 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
789 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
790 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
791 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
792 (cmp_op, sve_imm_con): New code attributes.
793 (SVE_COND_INT_CMP, imm_con): Delete.
794 (cmp_op): Remove above unspecs from int attribute.
795 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
797 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
798 comparison-specific unspecs.
799 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
800 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
801 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
802 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
803 (*vec_fcm<cmp_op><mode>): Rename to...
804 (*fcm<cmp_op><mode>): ...this and adjust likewise.
805 (*vec_fcmuo<mode>): Rename to...
806 (*fcmuo<mode>): ...this and adjust likewise.
807 (*pred_fcm<cmp_op><mode>): New pattern.
808 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
809 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
811 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
813 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
814 (aarch64_emit_sve_predicated_cond): New function.
815 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
816 (aarch64_emit_unspec_cond_or): Replace with...
817 (aarch64_emit_sve_or_conds): ...this new function. Use
818 aarch64_emit_sve_ptrue_op for the individual comparisons and
819 aarch64_emit_binop to OR them together.
820 (aarch64_emit_inverted_unspec_cond): Replace with...
821 (aarch64_emit_sve_inverted_cond): ...this new function. Use
822 aarch64_emit_sve_ptrue_op for the comparison and
823 aarch64_emit_unop to invert the result.
824 (aarch64_expand_sve_vec_cmp_float): Update after the above
825 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
827 2018-05-07 Nathan Sidwell <nathan@acm.org>
829 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
830 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
831 (Backwards Compatibility): Likewise.
833 2018-05-07 Luis Machado <luis.machado@linaro.org>
837 2018-05-07 Luis Machado <luis.machado@linaro.org>
839 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
840 <prefetch_dynamic_strides>: New const bool field.
841 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
842 prefetch_dynamic_strides.
843 (exynosm1_prefetch_tune): Likewise.
844 (thunderxt88_prefetch_tune): Likewise.
845 (thunderx_prefetch_tune): Likewise.
846 (thunderx2t99_prefetch_tune): Likewise.
847 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
848 (aarch64_override_options_internal): Update to set
849 PARAM_PREFETCH_DYNAMIC_STRIDES.
850 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
851 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
852 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
853 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
854 prefetch-dynamic-strides setting.
856 2018-05-07 Luis Machado <luis.machado@linaro.org>
858 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
859 <minimum_stride>: New const int field.
860 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
861 minimum_stride field.
862 (exynosm1_prefetch_tune): Likewise.
863 (thunderxt88_prefetch_tune): Likewise.
864 (thunderx_prefetch_tune): Likewise.
865 (thunderx2t99_prefetch_tune): Likewise.
866 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
867 (aarch64_override_options_internal): Update to set
868 PARAM_PREFETCH_MINIMUM_STRIDE.
869 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
870 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
871 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
872 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
873 stride is constant and is below the minimum stride threshold.
875 2018-05-07 Luis Machado <luis.machado@linaro.org>
877 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
880 2018-05-07 Luis Machado <luis.machado@linaro.org>
882 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
883 <prefetch_dynamic_strides>: New const bool field.
884 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
885 prefetch_dynamic_strides.
886 (exynosm1_prefetch_tune): Likewise.
887 (thunderxt88_prefetch_tune): Likewise.
888 (thunderx_prefetch_tune): Likewise.
889 (thunderx2t99_prefetch_tune): Likewise.
890 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
891 (aarch64_override_options_internal): Update to set
892 PARAM_PREFETCH_DYNAMIC_STRIDES.
893 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
894 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
895 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
896 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
897 prefetch-dynamic-strides setting.
899 2018-05-07 Luis Machado <luis.machado@linaro.org>
901 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
902 <minimum_stride>: New const int field.
903 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
904 minimum_stride field.
905 (exynosm1_prefetch_tune): Likewise.
906 (thunderxt88_prefetch_tune): Likewise.
907 (thunderx_prefetch_tune): Likewise.
908 (thunderx2t99_prefetch_tune): Likewise.
909 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
910 (aarch64_override_options_internal): Update to set
911 PARAM_PREFETCH_MINIMUM_STRIDE.
912 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
913 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
914 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
915 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
916 stride is constant and is below the minimum stride threshold.
918 2018-05-06 Jakub Jelinek <jakub@redhat.com>
921 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
922 the type is addressable. Don't force op into register if it has
925 2018-05-05 Roland McGrath <mcgrathr@google.com>
928 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
929 any section for which we don't know a specific type it should have,
930 regardless of name. Previously this was done only for the exact
931 names ".init_array", ".fini_array", and ".preinit_array".
932 (default_elf_asm_named_section): Add comment about
933 relationship with default_section_type_flags and SECTION_NOTYPE.
934 (get_section): Don't consider it a type conflict if one side has
935 SECTION_NOTYPE and the other doesn't, as long as neither has the
936 SECTION_BSS et al used in the default_section_type_flags logic.
938 2018-05-05 Tom de Vries <tom@codesourcery.com>
941 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
942 (workaround_barsyncs): New function.
943 (nvptx_reorg): Use workaround_barsyncs.
944 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
945 (define_expand "nvptx_membar_cta"): New define_expand.
946 (define_insn "*nvptx_membar_cta"): New insn.
948 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
950 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
951 To improve optimization opportunities.
952 * builtin-types.def: The new needed builtin types for the above.
954 2018-05-04 Richard Biener <rguenther@suse.de>
956 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
957 * gimple-ssa-store-merging.c
958 (imm_store_chain_info::output_merged_store): Remove redundant create,
959 release split_store vector contents on failure.
960 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
961 scalar stmt vector on cache hit.
963 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
965 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
967 * config.gcc (powerpc-xilinx-eabi*): Remove.
968 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
970 (fusion_addis_mem_combo_load): Ditto.
971 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
973 (rs6000_cpu_cpp_builtins): Ditto.
974 * config/rs6000/rs6000-linux.c
975 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
976 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
977 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
979 (rs6000_setup_reg_addr_masks): Ditto.
980 (rs6000_init_hard_regno_mode_ok): Ditto.
981 (rs6000_option_override_internal): Ditto.
982 (legitimate_lo_sum_address_p): Ditto.
983 (rs6000_legitimize_address): Ditto.
984 (rs6000_legitimize_reload_address): Ditto.
985 (rs6000_legitimate_address_p): Ditto.
986 (abi_v4_pass_in_fpr): Ditto.
987 (setup_incoming_varargs): Ditto.
988 (rs6000_gimplify_va_arg): Ditto.
989 (rs6000_split_multireg_move): Ditto.
990 (rs6000_savres_strategy): Ditto.
991 (rs6000_emit_prologue_components): Ditto.
992 (rs6000_emit_epilogue_components): Ditto.
993 (rs6000_emit_prologue): Ditto.
994 (rs6000_emit_epilogue): Ditto.
995 (rs6000_elf_file_end): Ditto.
996 (rs6000_function_value): Ditto.
997 (rs6000_libcall_value): Ditto.
998 * config/rs6000/rs6000.h: Ditto.
999 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1000 (TARGET_MINMAX): ... this. New.
1001 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1002 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1003 (*movsi_internal1_single): Delete.
1004 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1005 mfpu=, mxilinx-fpu): Delete.
1006 * config/rs6000/singlefp.h: Delete.
1007 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1008 * config/rs6000/t-rs6000: Ditto.
1009 * config/rs6000/t-xilinx: Delete.
1010 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1011 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1012 (VStype_simple): Delete.
1013 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1014 * config/rs6000/xfpu.h: Delete.
1015 * config/rs6000/xfpu.md: Delete.
1016 * config/rs6000/xilinx.h: Delete.
1017 * config/rs6000/xilinx.opt: Delete.
1018 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1019 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1021 2018-05-04 Tom de Vries <tom@codesourcery.com>
1024 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1027 2018-05-04 Richard Biener <rguenther@suse.de>
1030 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1031 (expand_complex_div_wide): Likewise.
1032 (expand_complex_operations_1): Likewise.
1033 (expand_complex_libcall): Preserve EH info of the original stmt.
1034 (tree_lower_complex): Handle removed blocks.
1035 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1036 on complex multiplication and division libcall builtins.
1038 2018-05-04 Richard Biener <rguenther@suse.de>
1041 * fold-const.c (negate_expr_p): Restrict negation of operand
1042 zero of a division to when we know that can happen without
1044 (fold_negate_expr_1): Likewise.
1046 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1049 * real.h (real_nextafter): Declare.
1050 * real.c (real_nextafter): New function.
1051 * fold-const-call.c (fold_const_nextafter): New function.
1052 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1053 CASE_CFN_NEXTTOWARD.
1054 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1055 even when arg1_mode is different from arg0_mode.
1057 2018-05-03 Nathan Sidwell <nathan@acm.org>
1059 * doc/extend.texi (Deprecated Features): Remove
1061 (Backwards Compatibility): Likewise.
1062 * doc/invoke.texi (C++ Language Options): Likewise.
1063 (C++ Dialect Options): Likewise.
1065 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1068 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1069 _mm512_mask_mullox_epi64): New intrinsics.
1071 2018-05-03 Tom de Vries <tom@codesourcery.com>
1074 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1075 dump files): Add offload-tree.
1077 2018-05-03 Richard Biener <rguenther@suse.de>
1079 PR tree-optimization/85615
1080 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1081 to loops not nested in BBs loop father to avoid creating multi-entry
1084 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1086 PR tree-optimization/70291
1087 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1088 arguments. Change return type to tree. Emit libcall as a new
1089 statement rather than replacing existing one when inplace_p is true.
1090 (expand_complex_multiplication_components): New function.
1091 (expand_complex_multiplication): Expand floating-point complex
1092 multiplication using the above.
1093 (expand_complex_division): Rename inner_type parameter to type.
1094 Update expand_complex_libcall call-site.
1095 (expand_complex_operations_1): Update expand_complex_multiplication
1096 and expand_complex_division call-sites.
1098 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1101 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1102 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1103 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1104 the highest significant bit of the shift count mask is clear. In
1105 check whether and[sq]i3 is needed verify that all significant bits
1106 of the shift count other than the highest are set.
1108 2018-05-02 Tom de Vries <tom@codesourcery.com>
1111 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1112 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1113 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1114 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1115 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1116 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1117 __builtin_goacc_parlevel_size.
1119 2018-05-02 Richard Biener <rguenther@suse.de>
1121 PR tree-optimization/85597
1122 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1123 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1126 2018-05-02 Tom de Vries <tom@codesourcery.com>
1129 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1130 dump files): Add ltrans-tree.
1132 2018-05-02 Tom de Vries <tom@codesourcery.com>
1135 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1136 dump files): Add wpa-ipa.
1138 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1140 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1141 powerpc*-*-linux*paired* target.
1142 * config/rs6000/750cl.h: Delete.
1143 * config/rs6000/paired.h: Delete.
1144 * config/rs6000/paired.md: Delete.
1145 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1147 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1148 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1149 comment. Remove paired float support.
1150 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1151 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1153 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1154 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1156 * config/rs6000/rs6000.c: Remove paired float support.
1157 (paired_expand_vector_init, paired_expand_vector_move,
1158 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1159 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1160 paired_expand_builtin, paired_expand_predicate_builtin,
1161 paired_init_builtins): Delete.
1162 * config/rs6000/rs6000.h: Remove paired float support.
1163 * config/rs6000/rs6000.md: Remove paired float support.
1164 (move_from_CR_ov_bit): Delete.
1165 * config/rs6000/rs6000.opt (mpaired): Delete.
1166 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1167 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1169 2018-05-02 Richard Biener <rguenther@suse.de>
1172 * gimplify.c (gimplify_save_expr): When in SSA form allow
1173 SAVE_EXPRs to compute to SSA vars.
1175 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1178 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1179 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1180 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1181 clobber operands[2], instead use a new pseudo. Formatting fixes.
1183 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1185 PR tree-optimization/85586
1186 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1187 exit early for statements in the same group if the accesses are
1190 2018-05-02 Tom de Vries <tom@codesourcery.com>
1193 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1196 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1198 PR tree-optimization/85143
1199 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1201 2018-05-01 Tom de Vries <tom@codesourcery.com>
1204 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1205 not found" error message.
1207 2018-05-01 Tom de Vries <tom@codesourcery.com>
1210 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1211 * vec.c (test_ordered_remove_if): New function.
1212 (vec_c_tests): Call test_ordered_remove_if.
1213 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1214 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1215 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1216 VEC_ORDERED_REMOVE_IF.
1218 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1220 PR tree-optimization/82665
1221 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1222 pointer subtraction where arguments come from a memchr call.
1224 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1226 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1227 --push-state --as-needed and --pop-state instead of --as-needed and
1228 --no-as-needed if ld supports it.
1229 * configure: Regenerated.
1232 * doc/install.texi2html: Replace _002d with - and _002a with * in
1233 generated html files using sed.
1235 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1238 * gcc-rich-location.c (blank_line_before_p): New function.
1239 (use_new_line): New function.
1240 (gcc_rich_location::add_fixit_insert_formatted): New function.
1241 * gcc-rich-location.h
1242 (gcc_rich_location::add_fixit_insert_formatted): New function.
1244 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1246 * selftest.c (assert_streq): Rename "expected" and "actual" to
1247 "val1" and "val2". Extend NULL-handling to cover both inputs
1248 symmetrically, while still requiring both to be non-NULL for a pass.
1249 * selftest.h (assert_streq): Rename "expected" and "actual" to
1251 (ASSERT_EQ): Likewise.
1252 (ASSERT_EQ_AT): Likewise.
1253 (ASSERT_KNOWN_EQ): Likewise.
1254 (ASSERT_KNOWN_EQ_AT): Likewise.
1255 (ASSERT_NE): Likewise.
1256 (ASSERT_MAYBE_NE): Likewise.
1257 (ASSERT_MAYBE_NE_AT): Likewise.
1258 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1259 the assertion to pass.
1260 (ASSERT_STREQ_AT): Likewise.
1262 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1264 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1265 interaction with -pie.
1267 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1269 * selftest.h: Fix alphabetization of per-source-file selftest
1272 2018-04-30 Jason Merrill <jason@redhat.com>
1274 PR c++/61982 - dead stores to destroyed objects.
1275 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1278 2018-04-30 Jason Merrill <jason@redhat.com>
1280 * tree.c (build_clobber): New.
1281 * tree.h: Declare it.
1282 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1284 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1286 * diagnostic-show-locus.c (layout::layout): Update for
1287 location_get_source_line returning a char_span.
1288 (struct char_span): Move to input.h.
1289 (struct correction): Update for fields in char_span becoming
1291 (struct source_line): Update for location_get_source_line
1292 returning a char_span.
1293 (layout::print_line): Likewise.
1294 * edit-context.c (edited_file::print_content): Likewise.
1295 (edited_file::print_diff_hunk): Likewise.
1296 (edited_file::print_run_of_changed_lines): Likewise.
1297 (edited_file::get_num_lines): Likewise.
1298 (edited_line::edited_line): Likewise.
1299 * final.c (asm_show_source): Likewise.
1300 * input.c (location_get_source_line): Convert return type
1301 from const char * to char_span, losing the final "line_len"
1303 (dump_location_info): Update for the above.
1304 (get_substring_ranges_for_loc): Likewise. Use a char_span
1305 when handling the literal within the line.
1306 (test_reading_source_line): Update for location_get_source_line
1307 returning a char_span.
1308 * input.h (class char_span): Move here from
1309 diagnostic-show-locus.c, converting from a struct to a class.
1310 Make data members private.
1311 (char_span::operator bool): New.
1312 (char_span::length): New.
1313 (char_span::get_buffer): New.
1314 (char_span::operator[]): New.
1315 (char_span::subspan): Make const.
1316 (char_span::xstrdup): New.
1317 (location_get_source_line): Convert return type from const char *
1318 to char_span, losing the final "line_size" param.
1320 2018-04-30 Jan Hubicka <jh@suse.cz>
1322 * lto-wrapper.c (ltrans_priorities): New static var.
1323 (cmp_priority): New.
1324 (run_gcc): Read priorities and if doing parallel build order
1325 the Makefile by them.
1327 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1329 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1331 2018-04-30 Richard Biener <rguenther@suse.de>
1333 * tree-cfg.c (verify_address): Remove base argument, add
1334 flag whether to check TREE_ADDRESSABLE and do that.
1335 (verify_expr): Remove.
1336 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1337 (verify_gimple_assign_single): Likewise.
1338 (verify_gimple_switch): Likewise.
1339 (verify_expr_location_1): Dereference tp once. Add (disabled)
1340 piece from verify_expr.
1341 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1343 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1345 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1347 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1349 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1350 (small_data_pattern): Likewise.
1351 (arc_rewrite_small_data): Likewise.
1352 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1353 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1354 (get_symbol_alignment): New function.
1355 (legitimate_small_data_address_p): Likewise.
1356 (legitimate_scaled_address): Update, call
1357 legitimate_small_data_address_p.
1358 (output_sdata): New static variable.
1359 (arc_print_operand): Update how we handle small data operands.
1360 (arc_print_operand_address): Likewise.
1361 (arc_legitimate_address_p): Update, use
1362 legitimate_small_data_address_p.
1363 (arc_rewrite_small_data_p): Remove.
1364 (arc_rewrite_small_data_1): Likewise.
1365 (arc_rewrite_small_data): Likewise.
1366 (small_data_pattern): Likewise.
1367 (compact_sda_memory_operand): Update to use
1368 legitimate_small_data_address_p and get_symbol_alignment.
1369 (prepare_move_operands): Don't rewite sdata pattern.
1370 (prepare_extend_operands): Remove.
1371 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1373 (zero_extendqisi2): Likewise.
1374 (zero_extendhisi2): Likewise.
1375 (extendqihi2): Likewise.
1376 (extendqisi2): Likewise.
1377 (extendhisi2): Likewise.
1381 * config/arc/constraints.md (Usd): Change it to memory constraint.
1383 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1385 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1386 as source of std instructions.
1387 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1388 allow 6-bit constants as source for store instructions.
1389 (movdi_insn): Update instruction pattern to allow 6-bit constants
1390 as source for store instructions.
1392 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1394 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1396 2018-04-30 Nathan Sidwell <nathan@acm.org>
1397 Sandra Loosemore <sandra@codesourcery.com>
1399 * dumpfile.c (dump_open): Allow '-' for stdout.
1400 * doc/invoke.texi (Developer Options): Document dump filename
1401 determination early. Document stdin/stdout selection.
1403 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1405 Microblaze Target: PIC data text relative
1407 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1408 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1410 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1411 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1412 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1413 New addressing mode for data-text relative position indepenedent code.
1414 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1415 'ADDRESS_SYMBOLIC_TXT_REL'.
1416 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1417 (microblaze_legitimate_pic_operand): Exclude function calls from
1418 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1419 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1421 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1422 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1423 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1424 for 'address + offset'.
1425 (microblaze_expand_prologue): Add new function prologue call for
1427 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1428 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1429 table in case of TARGET_PIC_DATA_TEXT_REL.
1430 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1431 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1432 Add new macros 'UNSPEC_TEXT',
1433 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1434 + exclude function calls from 'UNSPEC_PLT' in case of data text
1436 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1437 new target hook for generating address diff vector tables in case of
1439 * doc/tm.texi : Regenerate.
1440 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1441 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1442 of addr diff vector generation.
1443 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1444 target hook definition.
1445 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1446 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1447 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1448 option for data text relative.
1450 2018-04-30 Richard Biener <rguenther@suse.de>
1452 * tree-chrec.h (evolution_function_is_constant_p): Remove
1454 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1456 2018-04-30 Richard Biener <rguenther@suse.de>
1459 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1461 2018-04-30 Richard Biener <rguenther@suse.de>
1463 PR tree-optimization/28364
1464 PR tree-optimization/85275
1465 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1466 copying first exit test.
1468 2018-04-28 Mark Wielaard <mark@klomp.org>
1470 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1472 (dwarf_AT): Handle DW_AT_addr_base.
1473 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1475 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1478 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1479 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1480 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1481 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1483 2018-04-28 Richard Biener <rguenther@suse.de>
1485 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1486 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1487 to reflect use. Only add interesting stmts.
1489 2018-04-27 Martin Jambor <mjambor@suse.cz>
1492 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1493 the jump function allows for passing through aggregate values.
1495 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1497 * input.h (in_system_header_at): Convert from macro to inline
1499 (from_macro_expansion_at): Likewise.
1500 (from_macro_definition_at): Likewise.
1502 2018-04-27 Jeff Law <law@redhat.com>
1504 * config.gcc: Mark tile* targets as deprecated/obsolete.
1506 2018-04-27 Richard Biener <rguenther@suse.de>
1508 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1511 2018-04-27 Richard Biener <rguenther@suse.de>
1513 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1515 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1517 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1518 with Yd constraint. Set "preferred_for_speed" attribute from
1519 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1521 (*movdi_internal): Ditto.
1522 (movti_interunit splitters): Remove
1523 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1524 (movdi_interunit splitters): Ditto.
1525 * config/i386/constraints.md (Ye): Remove.
1526 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1528 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1531 * config/aarch64/constraints.md (Usg): Limit to 31.
1534 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1536 PR tree-optimization/85529
1537 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1538 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1539 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1540 zero extension or masking of the MSB bit.
1541 (optimize_range_tests): Add FIRST_BB argument, pass it through
1542 to optimize_range_tests_var_bound.
1543 (maybe_optimize_range_tests, reassociate_bb): Adjust
1544 optimize_range_tests callers.
1546 2018-04-26 Richard Biener <rguenther@suse.de>
1547 Jakub Jelinek <jakub@redhat.com>
1549 * cgraph.h (symbol_table): Just declare debug method here.
1550 * symtab.c (symbol_table::debug): Define.
1552 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1554 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1556 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1558 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1559 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1560 (*movdi_internal): Substitute Yi and Yj constraint with x
1561 and Ym and Yn constraint with y constraint. Update "isa"
1562 attribute and set "preferred_for_speed" attribute from
1563 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1564 (*movsi_internal): Ditto.
1565 (*movdf_internal): Ditto.
1566 (*movsf_internal): Ditto.
1567 (*zero_extendsidi2): Ditto.
1568 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1569 (sse2_loadld): Ditto.
1570 (*vec_extract<ssevecmodelower>_0): Ditto.
1571 (*vec_extractv4si_0_zext_sse4): Ditto.
1572 (vec_concatv2di): Ditto.
1573 (*vec_dup<mode>): Ditto.
1574 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1575 * config/i386/constraints.md (Yi): Remove.
1580 2018-04-26 Nathan Sidwell <nathan@acm.org>
1582 * dumpfile.c (dump_open): New.
1583 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1584 (dump_finish): Detect stdio/stderr by value not name.
1586 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1588 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1590 2018-04-26 Tom de Vries <tom@codesourcery.com>
1593 * config/nvptx/nvptx.c (verify_neutering_jumps)
1594 (verify_neutering_labels): New function
1595 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1597 2018-04-26 Tom de Vries <tom@codesourcery.com>
1600 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1601 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1604 2018-04-26 Richard Biener <rguenther@suse.de>
1605 Tom de Vries <tom@codesourcery.com>
1608 * lto-streamer-out.c (output_function): Fixup loops if required to match
1609 discovery done in the reader.
1611 2018-04-26 Richard Biener <rguenther@suse.de>
1613 PR tree-optimization/85116
1614 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1615 have a loop exit from the single latch predecessor. Remove
1616 case of header with just condition.
1617 (ch_base::copy_headers): Exclude infinite loops from any
1619 (pass_ch::execute): Record exits.
1621 2018-04-26 Richard Biener <rguenther@suse.de>
1623 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1624 prologue cost vector and pass it to vect_get_load_cost.
1625 (vect_get_peeling_costs_all_drs): Likewise.
1626 (vect_peeling_hash_get_lowest_cost): Likewise.
1627 (vect_enhance_data_refs_alignment): Likewise.
1629 2018-04-26 Richard Biener <rguenther@suse.de>
1632 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1633 checking of integer<->pointer conversions.
1634 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1635 sign-/zero-extending pointer types.
1636 (expand_omp_for_static_chunk): Likewise.
1638 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1639 Jean Lee <xiaoyur347@gmail.com>
1641 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1642 (TARGET_ASAN_SHADOW_OFFSET): Define.
1643 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1644 true for -fsanitize=address.
1646 2018-04-25 Mark Wielaard <mark@klomp.org>
1648 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1651 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1653 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1654 than "alu", remove explicit "memory" and "imm_disp" attributes.
1655 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1658 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1659 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1660 gen_lowpart_no_emit.
1662 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1665 * config/i386/i386.c (ix86_expand_builtin): Change memory
1666 operand to XI, extend p0 to Pmode.
1667 * config/i386/i386.md: Change unspec volatile and operand
1668 1 mode to XI, change operand 0 mode to P.
1670 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1672 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1673 GET_MODE_MASK before any checking.
1674 (nds32_can_use_bset_p): Likewise.
1675 (nds32_can_use_btgl_p): Likewise.
1677 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1679 * config/nds32/nds32-doubleword.md: New define_split pattern for
1680 illegal register number.
1682 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1684 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1686 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1688 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1690 2018-04-25 Richard Biener <rguenther@suse.de>
1692 * lto-streamer.h (LTO_major_version): Bump to 8.
1694 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1696 * BASE-VER: Set to 9.0.0.
1698 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1700 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1701 in __abskf2 and __powikf2.
1703 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1706 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1707 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1708 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1709 Use the above on operand 2. Reindent.
1710 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1712 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1715 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1717 * config/i386/i386.opt (mcet): Removed.
1718 * doc/install.texi: Remove -mcet documentation.
1719 * doc/invoke.texi: Likewise.
1721 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1724 * doc/install.texi: Remove -mcet from bootstrap-cet.
1726 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1729 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1730 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1734 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1735 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1736 containing a CONST_VECTOR.
1738 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1740 * doc/install.texi: Update newlib dependency for nvptx.
1742 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1745 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1746 instead of INTVAL when shifting x left.
1748 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1750 PR tree-optimization/85478
1751 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1752 vect_grouped_store_supported for single element vectors.
1754 2018-04-24 Richard Biener <rguenther@suse.de>
1757 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1758 load cost increase to the case of non-constant step.
1760 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1763 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1764 destination if any_malformed_asm.
1766 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1769 * expr.c (store_field): In the bitfield case, if the value comes from
1770 a function call and is returned in registers by means of a PARALLEL,
1771 do not change the mode of the temporary unless BLKmode and VOIDmode.
1773 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1775 PR rtl-optimization/85423
1776 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1777 dependencies to debug insns when the previous insn is non-debug.
1779 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1781 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1782 enums into a single definition.
1783 (fls): Fix predicates and printing.
1786 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1788 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1789 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1790 and short u6 immediate.
1791 (check_if_valid_sleep_operand): Remove.
1792 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1794 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1796 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1797 flag_always_save_lp condition.
1798 * config/nds32/nds32.opt (malways-save-lp): New option.
1800 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1802 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1803 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1804 * config/nds32/nds32.h
1805 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1806 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1808 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1810 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1811 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1813 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1814 Chung-Ju Wu <jasonwucj@gmail.com>
1816 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1818 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1819 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1821 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1823 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1825 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1827 * config/nds32/nds32-protos.h (nds32_data_alignment,
1828 nds32_local_alignment): Declare.
1829 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1830 nds32_local_alignment): New functions.
1831 (TARGET_CONSTANT_ALIGNMENT): Define.
1832 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1834 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1836 * config/nds32/nds32.c
1837 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1838 (TARGET_MODES_TIEABLE_P): Likewise.
1840 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1842 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1845 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1846 Chung-Ju Wu <jasonwucj@gmail.com>
1848 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1849 for unaligned access.
1850 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1852 * config/nds32/nds32-intrinsic.md: Likewise.
1853 * config/nds32/nds32_intrinsic.h: Likewise.
1854 * config/nds32/nds32.h (nds32_builtins): Likewise.
1855 * config/nds32/nds32.opt (munaligned-access): New option.
1856 * config/nds32/nds32.c (nds32_asm_file_start): Display
1857 flag_unaligned_access status.
1859 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1861 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1862 -mno-relax is present.
1863 * config/riscv/linux.h (LINK_SPEC): Ditto.
1865 2018-04-20 Martin Sebor <msebor@redhat.com>
1868 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1870 (gimple_fold_builtin_stxcpy_chk): Same.
1871 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1873 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1876 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1877 __powikf2 when long double is IEEE 128-bit.
1879 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1881 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1882 step to make sure stack always aligned.
1884 2018-04-20 Carl Love <cel@us.ibm.com>
1887 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1888 size check for arg0.
1890 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1891 Tom de Vries <tom@codesourcery.com>
1894 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1895 Emit insns for calls too.
1896 (nvptx_find_par): Always look for worker-level predecessor insn.
1897 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1899 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1900 (nvptx_process_pars): Propagate frames for calls.
1902 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1905 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1907 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1908 (ix86_handle_option): Don't handle OPT_mibt.
1909 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1911 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1913 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1914 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1915 (ix86_target_macros): Define __CET__ with flag_cf_protection
1916 for -fcf-protection.
1917 * config/i386/i386.c (isa2_opts): Remove -mibt.
1918 * config/i386/i386.h (TARGET_IBT): Removed.
1919 (TARGET_IBT_P): Likewise.
1920 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1921 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1922 * config/i386/i386.opt (mcet): Update help message.
1925 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1926 -mcet as an alias for -mshstk.
1928 2018-04-20 Richard Biener <rguenther@suse.de>
1931 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1932 complexity by forcing a single use of the multiply operand.
1934 2018-04-20 Martin Jambor <mjambor@suse.cz>
1937 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1938 recursion dependency to only apply to non-clones.
1940 2018-04-20 Martin Jambor <mjambor@suse.cz>
1943 * ipa-cp.c (create_specialized_node): Check that clones of
1944 self-recursive edges exist during IPA-CP.
1946 2018-04-19 Toon Moene <toon@moene.org>
1948 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1951 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1953 PR tree-optimization/85467
1954 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1955 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1956 VECTOR_CST element to type.
1958 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1961 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1962 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1963 (builtin_longjmp): Likewise.
1964 (save_stack_nonlocal): New pattern.
1965 (restore_stack_nonlocal): Likewise.
1967 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1970 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1971 Replace ASM_OUTPUT_LABEL with fprintf.
1973 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1976 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1977 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1978 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1979 define __IBT__ and __SHSTK__ for -fcf-protection.
1980 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1982 (ix86_trampoline_init): Likewise.
1983 (x86_output_mi_thunk): Likewise.
1984 (ix86_notrack_prefixed_insn_p): Likewise.
1985 (ix86_option_override_internal): Don't disallow -fcf-protection.
1986 * config/i386/i386.md (rdssp<mode>): Also enable for
1988 (incssp<mode>): Likewise.
1989 (nop_endbr): Likewise.
1990 * config/i386/i386.opt (mcet): Change help message to built-in
1994 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1995 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1996 enable CET built-in functions.
1998 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2000 * common/config/i386/i386-common.c
2001 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2002 OPTION_MASK_ISA_MOVDIRI_UNSET,
2003 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2004 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2005 * config.gcc (movdirintrin.h): New header.
2006 * config/i386/cpuid.h (bit_MOVDIRI,
2007 bit_MOVDIR64B): New bits.
2008 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2010 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2011 (VOID, PVOID, PCVOID)): New function types.
2012 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2013 __builtin_ia32_directstoreu_u64,
2014 __builtin_ia32_movdir64b): New builtins.
2015 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2016 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2018 (ix86_valid_target_attribute_inner_p): Ditto.
2019 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2020 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2021 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2022 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2023 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2024 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2025 (movdiri<mode>, movdir64b_<mode>): New.
2026 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2027 * config/i386/immintrin.h: Include movdirintrin.h.
2028 * config/i386/movdirintrin.h: New file.
2029 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2031 2018-04-19 Richard Biener <rguenther@suse.de>
2034 * cfg.c (clear_bb_flags): When loop state says we have
2035 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2037 2018-04-19 Richard Biener <rguenther@suse.de>
2039 PR tree-optimization/84737
2040 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2041 copying restrict info.
2042 (vect_setup_realignment): Use it.
2043 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2044 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2045 the first DR to all generated stores.
2046 (vectorizable_load): Likewise for loads.
2048 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2050 PR tree-optimization/85446
2051 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2052 the integral and pointer types to have the same precision.
2054 * doc/install.texi: Document --disable-cet being the default and
2057 2018-04-18 Martin Liska <mliska@suse.cz>
2059 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2062 2018-04-18 Martin Liska <mliska@suse.cz>
2065 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2068 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2069 arguments if they are comparable.
2071 2018-04-18 Martin Liska <mliska@suse.cz>
2074 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2077 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2080 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2083 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2084 ENDBR after calling __morestack.
2086 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2089 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2090 by using gcc_base_ver to generate a gcc_driver_version, and use
2091 it when generating GCC_DRIVER_NAME.
2092 * configure: Regenerate.
2094 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2097 * config.gcc: Obsolete powerpc*-*-*spe*.
2099 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2102 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2103 (stabstr_D): Change type of unum from unsigned int to
2104 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2107 2018-04-17 Jim Wilson <jimw@sifive.com>
2110 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2111 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2112 Set arg_pointer_offset after using pretend_args_size.
2114 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2116 PR rtl-optimization/85431
2117 * dse.c (record_store): Ignore zero width stores.
2120 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2121 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2122 __builtin_stack_restore rather than after it.
2123 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2124 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2125 argument instead of virtual_dynamic_stack_rtx.
2127 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2129 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2131 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2132 Add note to error message to explain internal mapping of overloaded
2133 built-in function name to non-overloaded built-in function name.
2134 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2137 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2140 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2141 where the inputs overlap with the output.
2143 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2145 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2146 (=v, v) alternative and explicit "memory" attribute.
2147 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2148 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2150 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2151 "sselog1" type instead of "sselog".
2152 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2153 "sselog". Remove explicit "memory" attribute.
2154 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2155 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2157 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2158 "sselog1" type instead of "sselog". Remove explicit "memory"
2160 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2161 use "sselog1" type instead of "sselog". Remove explicit "memory"
2163 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2164 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2166 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2167 "sselog1" type instead of "sselog". Remove explicit "memory"
2169 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2170 use "sselog1" type instead of "sselog". Remove explicit "memory"
2174 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2177 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2180 2018-04-17 Martin Jambor <mjambor@suse.cz>
2183 * ipa-cp.c (create_specialized_node): Call
2184 expand_all_artificial_thunks if necessary.
2186 2018-04-17 Martin Liska <mliska@suse.cz>
2189 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2190 in message, remote space in between '_G' and '('.
2192 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2195 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2196 avx512f_vmcmp<mode>3<round_saeonly_name>,
2197 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2198 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2199 avx512f_rndscale<mode><round_saeonly_name>,
2200 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2201 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2202 Use %<iptr>2 instead of %2 for -masm=intel.
2203 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2204 avx512f_vcvttss2usi<round_saeonly_name>,
2205 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2207 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2208 avx512f_vcvttsd2usi<round_saeonly_name>,
2209 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2210 Use %q1 instead of %1 for -masm=intel.
2211 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2212 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2213 of %3 for -masm=intel.
2214 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2216 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2218 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2220 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2221 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2223 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2225 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2226 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2227 %g1 and one with %0 and %1.
2228 (avx512er_vmrcp28<mode><round_saeonly_name>,
2229 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2231 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2232 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2233 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2234 of %0 and %{%4%} for -masm=intel.
2235 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2236 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2237 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2238 order of %0 and %{%5%}%{z%} for -masm=intel.
2240 2018-04-17 Jan Hubicka <jh@suse.cz>
2243 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2245 2018-04-17 Martin Liska <mliska@suse.cz>
2248 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2249 for target_clone error message. Make default implementation
2250 clone to be a local declaration.
2251 (separate_attrs): Add new argument and check for an empty
2253 (expand_target_clones): Handle it.
2254 (ipa_target_clone): Make redirection just for target_clones
2257 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2258 Tom de Vries <tom@codesourcery.com>
2261 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2262 tiled basic blocks without omp continue statements.
2264 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2267 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2268 vec_extract expression as having side effects to make sure it gets
2271 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2274 * config/i386/i386.c (get_builtin_code_for_version): Check
2277 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2280 * gcc/config.gcc: Support "skylake".
2281 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2283 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2284 (processor_target_table): Add "skylake".
2285 (ix86_option_override_internal): Add "skylake".
2286 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2287 PROCESSOR_CANNONLAKE.
2288 (get_builtin_code_for_version): Fix priority for
2289 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2290 PROCESSOR_SKYLAKE-AVX512.
2291 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2292 (processor_type): Add PROCESSOR_SKYLAKE.
2294 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2295 Jason Merrill <jason@redhat.com>
2298 * convert.c (convert_to_integer_1): Use direct recursion for
2299 enumeral types and types with a precision less than the number
2300 of bits in their mode.
2302 2018-04-16 Julia Koval <julia.koval@intel.com>
2305 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2306 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2308 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2311 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2312 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2313 and -mno-direct-move.
2315 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2318 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2319 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2320 Ensure negative shifts result in {0}.
2322 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2324 PR rtl-optimization/79916
2325 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2326 regs (if any) to define how to gnerate SD moves when LRA is in
2329 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2331 PR rtl-optimization/85393
2332 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2333 * except.c (expand_dw2_landing_pad_for_region): Make static.
2334 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2335 a label and unconditional jump to old_bb, rather than
2336 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2339 PR rtl-optimization/85376
2340 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2341 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2342 instead of a specific value.
2344 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2345 Bin Cheng <bin.cheng@arm.com>
2347 PR tree-optimization/82965
2348 PR tree-optimization/83991
2349 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2350 by_profile_only parameter.
2351 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2352 information if the loop was predicted to iterate too many times.
2353 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2355 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2358 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2361 2018-04-13 Martin Liska <mliska@suse.cz>
2362 Jakub Jelinek <jakub@redhat.com>
2365 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2366 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2367 * builtins.c (expand_builtin_memory_copy_args): Use
2368 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2369 handle dest_addr == pc_rtx.
2371 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2374 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2375 asked to not generate direct moves.
2376 (fix_trunc<mode>si2_stfiwx): Similar.
2377 (fix_trunc<mode>si2_internal): Similar.
2379 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2382 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2383 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2384 lookup if dest in some wider mode is known to be const0_rtx and
2385 if so, record permanent equivalence for it to be ZERO_EXTEND of
2386 the narrower mode destination.
2388 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2390 * lto-streamer-out.c (output_function): Revert 259346.
2391 * omp-expand.c (expand_oacc_for): Likewise.
2393 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2395 PR rtl-optimization/85354
2396 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2397 * sel-sched.c (sel_global_init): ... here.
2399 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2402 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2403 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2404 mode for PE-COFF targets.
2405 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2406 (i386_pe_asm_lto_end): Likewise.
2407 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2408 (TARGET_ASM_LTO_END): Likewise.
2409 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2410 (i386_pe_asm_lto_start): New function.
2411 (i386_pe_asm_lto_end): Likewise.
2413 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2414 Richard Biener <rguenther@suse.de>
2417 * lto-streamer-out.c (output_function): Fix CFG loop state before
2419 * omp-expand.c (expand_oacc_for): Handle calls to internal
2420 functions like regular functions.
2422 2018-04-12 Richard Biener <rguenther@suse.de>
2425 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2426 for the early LTO debug to properly generate references to it
2427 during DIE emission. Do not re-use that for the skeleton for
2429 (dwarf2out_early_finish): Likewise.
2431 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2434 * config/i386/sse.md
2435 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2436 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2437 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2438 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2439 and output is a reg, avoid creating invalid lowpart subreg, but
2440 instead split into a 512-bit move. Don't split if not AVX512VL,
2441 input is xmm16+ reg and output is a mem.
2442 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2443 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2444 xmm16+ reg and output is a mem.
2446 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2448 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2449 also for flag_dwarf2_cfi_asm.
2451 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2453 PR rtl-optimization/85342
2454 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2455 a bool scalar var inside of the loop instead. Don't try to update
2456 recog_data.operand after failed apply_change_group.
2458 2018-04-12 Tom de Vries <tom@codesourcery.com>
2461 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2462 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2463 array with flexible array member as array without given dimension.
2464 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2465 argument for undefined param to true.
2467 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2470 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2471 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2472 from PowerPC section.
2473 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2474 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2475 help text that is too long.
2476 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2477 help text that is too long.
2478 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2479 help text that is too long.
2481 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2483 * config/alpha/alpha.md (stack_probe_internal): Rename
2484 from "probe_stack". Update all callers.
2486 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2488 PR rtl-optimization/84566
2489 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2490 sched_macro_fuse_insns.
2492 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2495 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2496 (compute_block_dependences): ... from here.
2498 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2500 PR tree-optimization/85331
2501 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2502 from int to HOST_WIDE_INT.
2504 2018-04-11 Martin Jambor <mjambor@suse.cz>
2507 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2508 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2509 not the same as the source val.
2510 (cgraph_edge_brings_value_p): New parameter.
2511 (gather_edges_for_value): Pass destination value to
2512 cgraph_edge_brings_value_p.
2513 (perhaps_add_new_callers): Likewise.
2514 (get_info_about_necessary_edges): Likewise and exclude values brought
2515 only by self-recursive edges.
2516 (create_specialized_node): Redirect only clones of self-calling edges.
2517 (+self_recursive_pass_through_p): New function.
2518 (find_more_scalar_values_for_callers_subset): Use it.
2519 (find_aggregate_values_for_callers_subset): Likewise.
2520 (known_aggs_to_agg_replacement_list): Removed.
2521 (decide_whether_version_node): Re-calculate known constants for all
2522 remaining context clones.
2524 2018-04-11 Richard Biener <rguenther@suse.de>
2527 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2528 from early DWARF output.
2529 (dwarf2out_early_finish): Output line info unconditionally into
2530 early DWARF and add reference to it.
2532 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2535 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2536 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2537 other than V2DFmode using iptr mode attribute.
2538 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2540 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2542 PR rtl-optimization/84659
2543 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2545 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2548 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2550 (output_loc_list): Pass address of a dummy size variable even in the
2551 locview handling loop.
2552 (index_location_lists): Add comment on why skip_loc_list_entry can't
2555 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2558 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2561 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2564 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2565 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2566 and -mstring-compare-inline-limit.
2568 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2571 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2572 for stack clash protection in a register whenever we need it to be in
2575 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2577 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2578 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2580 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2583 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2585 (mlong-double-): Ditto.
2586 * config/rs6000/sysv4.opt (msdata=): Ditto.
2587 (mtls-size=): Ditto.
2589 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2591 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2592 erroneous entries for
2593 "vector int vec_ldl (int, long int *)", and
2594 "vector unsigned int vec_ldl (int, unsigned long int *)".
2595 Add comments and entries for
2596 "vector bool char vec_ldl (int, bool char *)",
2597 "vector bool short vec_ldl (int, bool short *)",
2598 "vector bool int vec_ldl (int, bool int *)",
2599 "vector bool long long vec_ldl (int, bool long long *)",
2600 "vector pixel vec_ldl (int, pixel *)",
2601 "vector long long vec_ldl (int, long long *)",
2602 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2603 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2604 type tree bool_long_long_type_node and correct definition of
2605 bool_V2DI_type_node to make reference to this new type tree.
2606 (rs6000_mangle_type): Replace erroneous reference to
2607 bool_long_type_node with bool_long_long_type_node.
2608 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2609 comments to emphasize sign distinctions for char and int types and
2610 replace RS6000_BTI_bool_long constant with
2611 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2612 use of RS6000_BTI_pixel.
2613 (bool_long_type_node): Remove this macro definition.
2614 (bool_long_long_type_node): New macro definition
2616 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2618 PR rtl-optimization/85300
2619 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2620 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2621 simplify_unary_operation fails.
2623 2018-04-10 Martin Liska <mliska@suse.cz>
2625 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2626 cgraph_edge and ipa_ref.
2628 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2632 * config/i386/sse.md
2633 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2634 computation of the VEC_MERGE selector from mask.
2635 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2636 Fix decoding of the VEC_MERGE selector into mask.
2638 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2640 PR tree-optimization/85286
2641 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2643 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2645 * final.c (final_1): Set insn_last_address as well as
2646 insn_current_address.
2648 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2651 * explow.c (emit_stack_probe): Call validize_mem on memory location
2652 before passing it to gen_probe_stack. Create address operand and
2653 legitimize it for the probe_stack_address case.
2655 2018-04-09 Jan Hubicka <jh@suse.cz>
2658 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2659 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2660 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2662 (free_lang_data): Rebuild type inheritance graph.
2664 2018-04-09 Martin Sebor <msebor@redhat.com>
2666 * invoke.texi (-finline-small-functions): Mention other optimization
2668 (-findirect-inlining, -fpartial-inlining): Same.
2669 (-finline-functions-called-once): Same.
2670 (-freorder-blocks-and-partition): Same.
2672 2018-04-09 Jan Hubicka <jh@suse.cz>
2675 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2676 jumps; choose last target that matches the criteria (i.e.
2677 no partition changes for non-crossing jumps).
2678 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2679 support for redirecting crossing jumps to non-crossing.
2681 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2683 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2684 also for naked functions.
2686 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2688 * config/arc/arc.md (add_shift): New pattern.
2689 (add_shift2): Likewise.
2690 (sub_shift): Likewise.
2691 (sub_shift_cmp0_noout): Likewise.
2692 (compare_si_ashiftsi): Likewise.
2693 (xbfu_cmp0_noout): New combine pattern.
2694 (xbfu_cmp0"): Likewise.
2695 (movsi_set_cc_insn): Place the predicable variant first.
2696 (commutative_binary_cmp0_noout): Remove clobber.
2697 (commutative_binary_cmp0): New pattern.
2698 (noncommutative_binary_cmp0): Likewise.
2699 (noncommutative_binary_cmp0_noout): Likewise.
2700 (noncommutative_binary_comparison_result_used): Removed.
2701 (rsub_cmp0): New pattern.
2702 (rsub_cmp0_noout): Likewise.
2703 (extzvsi): Changed, keep only meaningful variants.
2704 (SQH, SEZ): New iterators.
2705 (SQH_postfix): New mode attribute.
2706 (SEZ_prefix): New code attribute.
2707 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2708 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2709 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2711 (noncommutative_operator): Check the availability of barrel
2714 2018-04-09 Richard Biener <rguenther@suse.de>
2716 PR tree-optimization/85284
2717 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2718 Only use the niter constraining form of simple_iv when the exit
2721 2018-04-09 Tom de Vries <tom@codesourcery.com>
2724 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2725 (define_expand "*memory_barrier"): New define_expand.
2726 (define_insn "memory_barrier"): New insn.
2728 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2730 PR rtl-optimization/80463
2731 PR rtl-optimization/83972
2732 PR rtl-optimization/83480
2734 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2735 correct producer for the insn.
2736 (tidy_control_flow): Fixup seqnos in case of debug insns.
2738 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2740 PR rtl-optimization/83913
2742 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2743 different sched-times when merging exprs.
2745 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2747 PR rtl-optimization/83962
2749 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2750 tidy_fallthru_edge and tidy_control_flow.
2752 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2754 PR rtl-optimization/83530
2756 * sel-sched.c (force_next_insn): New global variable.
2757 (remove_insn_for_debug): When force_next_insn is true, also leave only
2758 next insn in the ready list.
2759 (sel_sched_region): When the region wasn't scheduled, make another pass
2760 over it with force_next_insn set to 1.
2762 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2764 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2766 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2767 for interrupt control.
2768 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2769 functions for interrupt control.
2770 * config/nds32/nds32-intrinsic.md: Likewise.
2771 * config/nds32/nds32_intrinsic.h: Likewise.
2772 * config/nds32/nds32.h (nds32_builtins): Likewise.
2774 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2776 * config/nds32/nds32.c (nds32_init_machine_status,
2777 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2778 strict_aligned_p field.
2779 (nds32_expand_to_rtl_hook): New function.
2780 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2781 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2783 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2784 Chung-Ju Wu <jasonwucj@gmail.com>
2786 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2787 * config/nds32/nds32-n7.md: New file.
2788 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2789 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2791 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2792 * config/nds32/nds32.md (pipeline_model): Add n7.
2793 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2794 * config/nds32/pipelines.md: Include n7 settings.
2796 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2797 Chung-Ju Wu <jasonwucj@gmail.com>
2799 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2800 * config/nds32/nds32-e8.md: New file.
2801 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2802 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2804 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2805 * config/nds32/nds32.md (pipeline_model): Add e8.
2806 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2807 * config/nds32/pipelines.md: Include e8 settings.
2809 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2810 Chung-Ju Wu <jasonwucj@gmail.com>
2812 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2813 * config/nds32/nds32-n8.md: New file.
2814 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2815 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2817 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2818 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2819 * config/nds32/nds32.md (pipeline_model): Add n8.
2820 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2821 * config/nds32/pipelines.md: Include n8 settings.
2823 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2824 Chung-Ju Wu <jasonwucj@gmail.com>
2826 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2827 * config/nds32/nds32-n9-2r1w.md: New file.
2828 * config/nds32/nds32-n9-3r2w.md: New file.
2829 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2830 nds32_register_ports): New or modify for cpu n9.
2831 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2833 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2834 * config/nds32/nds32-utils.c: New file.
2835 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2836 TARGET_MUL_SLOW): Define.
2837 * config/nds32/nds32.md (pipeline_model): New attribute.
2838 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2839 New options that support cpu n9.
2840 * config/nds32/pipelines.md: Include n9 settings.
2841 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2843 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2845 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2846 information if necessary.
2847 (output_cond_branch_compare_zero): Likewise.
2848 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2849 (nds32_target_alignment): Refine for alignment.
2850 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2851 (FUNCTION_BOUNDARY): Modify.
2852 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2854 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2856 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2858 * config/nds32/constants.md (unspec_volatile_element): Add values for
2859 TLB operation and data prefetch.
2860 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2861 functions for TLB operation and data prefetch.
2862 * config/nds32/nds32-intrinsic.md: Likewise.
2863 * config/nds32/nds32_intrinsic.h: Likewise.
2864 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2865 (nds32_print_operand): Likewise.
2866 * config/nds32/nds32.h (nds32_builtins): Likewise.
2868 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2869 Andrew Pinski <pinsika@gcc.gnu.org>
2872 * match.pd: Use constant_boolean_node of correct type instead of
2873 boolean_true_node or boolean_false_node for simplifying
2874 pointer comparisons to zero.
2876 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2878 PR tree-optimization/80021
2879 * tree.c (verify_type_variant): Make error call in verify_variant_match
2880 translatable and remove final full stop.
2882 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2884 * config/nds32/constants.md (unspec_volatile_element): Add
2885 UNSPEC_VOLATILE_EH_RETURN.
2886 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2887 nds32_output_stack_pop): Support dwarf exception handling process.
2888 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2889 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2890 exception handling process.
2891 (nds32_compute_stack_frame): Likewise.
2892 (nds32_return_addr_rtx): Likewise.
2893 (nds32_initial_elimination_offset): Likewise.
2894 (nds32_expand_prologue): Likewise.
2895 (nds32_expand_epilogue): Likewise.
2896 (nds32_dynamic_chain_address): New function.
2897 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2899 (DYNAMIC_CHAIN_ADDRESS): Define.
2900 (EH_RETURN_DATA_REGNO): Define.
2901 (EH_RETURN_STACKADJ_RTX): Define.
2902 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2903 patterns for dwarf exception handling.
2905 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2907 * config/nds32/nds32.h: Clean up obsolete macros.
2909 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2911 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2912 Add enum values for particular instructions.
2913 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2914 particular intrinsic functions.
2915 * config/nds32/nds32-intrinsic.md: Likewise.
2916 * config/nds32/nds32_intrinsic.h: Likewise.
2917 * config/nds32/nds32.h (nds32_builtins): Likewise.
2918 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2919 (btst, ave): New patterns for particular instructions.
2921 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2923 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2924 Add enum values for atomic load/store and memory sync.
2925 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2927 * config/nds32/nds32-intrinsic.md: Likewise.
2928 * config/nds32/nds32_intrinsic.h: Likewise.
2929 * config/nds32/nds32.h (nds32_builtins): Likewise.
2931 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2933 PR tree-optimization/85257
2934 * fold-const.c (native_encode_vector): If not all elts could fit
2935 and off is -1, return 0 rather than offset.
2936 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2937 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2938 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2939 adjust buffer in native_interpret_expr call.
2941 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2943 * config/nds32/constants.md (unspec_volatile_element): Add cache
2944 control enum values.
2945 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2946 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2947 * config/nds32/nds32.c (nds32_cctl_names): New.
2948 (nds32_print_operand): Handle cache control register names.
2949 * config/nds32/nds32.h (nds32_builtins): New enum values.
2950 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2952 * config/nds32/nds32.md (type): Add mmu.
2953 * config/nds32/pipelines.md (simple_insn): Add mmu.
2955 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2957 * config/nds32/nds32.md (type): Remove call.
2958 * config/nds32/pipelines.md (simple_insn): Likewise.
2960 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2962 * config/nds32/constants.md (unspec_volatile_element): Add
2963 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2964 UNSPEC_VOLATILE_FMFCFG.
2965 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2966 description for fmfcfg and fmfcsr.
2967 (bdesc_1arg): Add fmtcsr.
2968 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2969 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2970 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2971 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2972 unspec_fmfcfg): New patterns.
2973 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2974 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2975 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2976 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2977 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2978 __nds32__fmfcfg): Define.
2980 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2982 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2983 intrinsic register names.
2984 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2985 intrinsic register enum values and macros.
2987 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2989 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2990 for load/store addressing form.
2991 (nds32_print_operand_address): Likewise.
2993 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2996 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2997 based on LABEL_REF. Remove useless assertion.
2998 (pic_address_needs_scratch): Fix formatting.
2999 (sparc_legitimize_pic_address): Minor tweaks.
3000 (sparc_delegitimize_address): Adjust assertion accordingly.
3001 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3002 into symbolic_operand.
3003 (movsi_high_pic_label_ref): Likewise.
3004 (movsi_lo_sum_pic_label_ref): Likewise.
3005 (movdi_pic_label_ref): Likewise.
3006 (movdi_high_pic_label_ref): Likewise.
3007 (movdi_lo_sum_pic_label_ref): Likewise.
3009 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3011 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3012 custom LIB_SPEC setup.
3014 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3015 Kito Cheng <kito.cheng@gmail.com>
3017 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3018 * config/riscv/freebsd.h: New.
3020 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3022 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3023 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3026 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3027 Kito Cheng <kito.cheng@gmail.com>
3029 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3030 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3031 * config/nds32/nds32-protos.h (nds32_output_call,
3032 nds32_output_return): Declare.
3033 * config/nds32/nds32.md: Refine all the call and return patterns.
3035 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3038 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3039 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3041 PR rtl-optimization/84872
3042 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3043 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3046 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3048 * expr.c (copy_blkmode_to_reg): Revert 254862.
3049 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3051 2018-04-06 Richard Biener <rguenther@suse.de>
3054 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3055 after seeing a component reference with an adjacent field. Treat
3056 refs to arrays at struct end of external decls similar to
3057 refs to unconstrained commons.
3059 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3062 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3063 look through SAVE_EXPRs with non-side-effects argument. Adjust
3065 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3068 2018-04-06 Richard Biener <rguenther@suse.de>
3071 * alias.c (find_base_term): New wrapper around find_base_term
3072 unwinding CSELIB_VAL_PTR changes.
3073 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3076 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3078 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3080 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3081 constant definitions.
3082 ("nop"): lr 0,0 -> nopr r0
3083 ("nop_lr0", "nop_lr1"): New insn definitions.
3085 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3087 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3088 NDS32_V3PUSH_AVAILABLE_P macro.
3090 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3091 Chung-Ju Wu <jasonwucj@gmail.com>
3093 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3094 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3095 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3096 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3097 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3098 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3099 * config/nds32/constraints.md: New constraints and checking for hard
3100 float configuration.
3101 * config/nds32/iterators.md: New mode iterator and attribute for hard
3102 float configuration.
3103 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3105 * config/nds32/nds32-fpu.md: New file.
3106 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3107 deal with hard float code generation.
3108 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3110 (abi_type, float_reg_number): New enum type.
3111 * config/nds32/nds32-predicates.c: New predicates for hard float.
3112 * config/nds32/nds32-protos.h: Declare functions for hard float.
3113 * config/nds32/nds32.c: Implementation for hard float configuration.
3114 * config/nds32/nds32.h: Definitions for hard float configuration.
3115 * config/nds32/nds32.md: Include hard float machine description and
3116 modify patterns for hard float configuration.
3117 * config/nds32/nds32.opt: New options for hard float configuration.
3118 * config/nds32/predicates.md: New predicates for hard float
3121 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3123 * common/config/nds32/nds32-common.c
3124 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3126 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3129 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3130 CONSTRUCTOR_ELT (ctor, ...)->value.
3132 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3135 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3137 2018-04-05 Tom de Vries <tom@codesourcery.com>
3140 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3143 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3144 Kito Cheng <kito.cheng@gmail.com>
3146 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3147 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3148 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3149 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3151 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3152 Kito Cheng <kito.cheng@gmail.com>
3154 * config/nds32/constraints.md (Ufe): New memory constraint.
3155 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3156 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3157 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3159 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3160 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3162 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3164 * config/nds32/nds32.md: Use optimize_size in the condition for
3165 alu-shift instructions.
3167 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3169 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3171 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3173 * config/nds32/nds32.md (negsi2): Refine pattern.
3175 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3176 Chung-Ju Wu <jasonwucj@gmail.com>
3178 * config/nds32/iterators.md (shift_rotate): New code iterator.
3179 (shift): New code attribute.
3180 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3181 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3182 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3183 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3184 bit-wise operations.
3185 (andsi3, *andsi3): Ditto.
3186 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3187 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3188 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3189 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3190 nds32_ior_operand, nds32_xor_operand): New predicates.
3192 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3194 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3195 (addsi3, subsi3): ... this.
3197 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3199 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3201 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3203 * config/nds32/nds32.md: Adjust indention.
3205 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3207 * config/nds32/nds32.md (feature): New attribute.
3209 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3211 * config/nds32/nds32.md (subtype): New attribute.
3213 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3216 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3217 expansion to perform a bitwise AND of the argument followed by a
3218 boolean negation of the result.
3220 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3222 PR rtl-optimization/84878
3223 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3224 the basic block. Assert the use reference is not artificial and that
3225 it has an associated insn.
3227 2018-04-04 Michael Matz <matz@suse.de>
3229 * builtins.c (compute_objsize): Pass correct operand
3230 to array_at_struct_end_p.
3232 2018-04-04 Richard Biener <rguenther@suse.de>
3235 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3236 from contexts for DINFO_LEVEL_TERSE and below.
3238 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3240 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3241 resiter_operand condition.
3242 * config/nds32/nds32.md (*move<mode>): Ditto.
3244 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3245 Monk Chiang <sh.chiang04@gmail.com>
3247 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3249 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3251 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3253 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3254 Kito Cheng <kito.cheng@gmail.com>
3256 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3257 nds32_cond_code_str, output_cond_branch,
3258 output_cond_branch_compare_zero, nds32_expand_cbranch,
3259 nds32_expand_cstore, nds32_expand_movcc,
3260 nds32_output_cbranchsi4_equality_zero,
3261 nds32_output_cbranchsi4_equality_reg,
3262 nds32_output_cbranchsi4_equality_reg_or_const_int,
3263 nds32_output_cbranchsi4_greater_less_zero: New functions.
3264 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3265 nds32_expand_cstore, nds32_expand_movcc,
3266 nds32_output_cbranchsi4_equality_zero,
3267 nds32_output_cbranchsi4_equality_reg,
3268 nds32_output_cbranchsi4_equality_reg_or_const_int,
3269 nds32_output_cbranchsi4_greater_less_zero): Declare.
3270 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3271 nds32_rimm11s_operand): New predicates.
3272 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3273 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3276 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3278 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3279 * config/nds32/nds32.md: Ditto.
3280 * config/nds32/pipelines.md: Ditto.
3282 2018-04-04 Richard Biener <rguenther@suse.de>
3284 PR tree-optimization/85168
3285 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3286 propagating abnormals.
3288 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3290 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3292 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3293 Kito Cheng <kito.cheng@gmail.com>
3295 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3296 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3297 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3298 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3299 * config/nds32/nds32.md (sibcall_internal): New.
3300 (sibcall_register): Remove.
3301 (sibcall_immediate): Remove.
3302 (sibcall_value_internal): New.
3303 (sibcall_value_register): Remove.
3304 (sibcall_value_immediate): Remove.
3305 * config/nds32/predicates.md (nds32_general_register_operand): New.
3306 (nds32_call_address_operand): New.
3308 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3310 PR rtl-optimization/85167
3311 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3312 bb_defs if *split_p, instead preinitialize it to NULL.
3314 PR tree-optimization/85156
3315 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3316 evaluating the argument multiple times.
3318 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3320 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3322 (_mm_cvtpd_ps): Likewise.
3323 (_mm_cvttpd_epi32): Likewise.
3324 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3325 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3326 vector, pixel, and bool following altivec.h include.
3328 2018-04-03 Martin Sebor <msebor@redhat.com>
3330 * doc/extend.texi (Common Function Attributes): Clarify.
3331 (const attribute): Likewise.
3332 (pure attribute): Likewise.
3334 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3337 * config/i386/i386.c (ix86_expand_vector_set): Use
3338 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3340 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3342 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3343 instructions when changing rounding bits to preserve precision bits
3344 in the x87 control word.
3346 2018-04-03 Martin Liska <mliska@suse.cz>
3348 PR tree-optimization/82491
3349 * rtl.h (strip_offset_and_add): Replace += suboffset with
3350 poly_uint64 () + suboffset.
3352 2018-03-29 Martin Liska <mliska@suse.cz>
3353 Martin Jambor <mjambor@suse.cz>
3356 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3357 param_type is not an integral or pointer type.
3359 2018-04-03 Richard Biener <rguenther@suse.de>
3361 * sese.h (recompute_all_dominators): Remove.
3363 2018-04-02 Martin Sebor <msebor@redhat.com>
3365 * doc/invoke.texi (-Wrestrict): Fix typos.
3367 2018-04-02 Jim Wilson <jimw@sifive.com>
3369 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3370 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3371 (<optab>di3, <optab>si3_extend): Likewise.
3372 (<optab>si3_mask, <optab>si3_mask_1): New.
3373 (<optab>di3_mask, <optab>di3_mask_1): New.
3374 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3375 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3376 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3378 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3380 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3383 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3385 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3386 (nds32_canonicalize_comparison): New function.
3388 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3389 Kito Cheng <kito.cheng@gmail.com>
3390 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3392 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3393 * config/nds32/constants.md (unspec_volatile_element): Add
3394 UNSPEC_VOLATILE_RELAX_GROUP.
3395 * config/nds32/nds32-relax-opt.c: New file.
3396 * config/nds32/nds32-predicates.c
3397 (nds32_symbol_load_store_p): New function.
3398 * config/nds32/nds32-protos.h
3399 (nds32_symbol_load_store_p): Declare function.
3400 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3401 * config/nds32/nds32.c
3402 (nds32_register_pass): New function to register pass.
3403 (nds32_register_passes): New function to register passes.
3404 * config/nds32/nds32.md (relax_group): New pattern.
3405 * config/nds32/nds32.opt (mrelax-hint): New option.
3406 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3408 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3410 * config/nds32/t-nds32: Modify files dependency.
3412 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3414 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3415 (PROFILE_HOOK): Define its implementation.
3417 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3419 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3420 type and 32-bit size.
3422 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3425 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3426 (V_128_256): New mode iterator.
3427 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3428 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3429 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3431 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3432 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3434 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3437 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3438 NaN inputs correctly.
3440 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3443 * config/rs6000/vsx.md (??r): New mode attribute.
3444 (*vsx_mov<mode>_64bit): Use it.
3445 (*vsx_mov<mode>_32bit): Likewise.
3447 2018-03-30 Martin Sebor <msebor@redhat.com>
3449 PR tree-optimization/84818
3450 * builtins.c (check_access): Use warning_n.
3452 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3455 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3457 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3460 2018-03-30 Julia Koval <julia.koval@intel.com>
3463 * x86-tune.def (movx, partial_reg_dependency): Enable for
3466 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3469 * lra-constraints.c (process_alt_operands): Move setting
3470 this_alternative_matches below.
3472 2018-03-29 Martin Liska <mliska@suse.cz>
3475 * doc/invoke.texi: Document how LTO works with debug info.
3476 Describe auto-load support of binutils. Mention 'x86-64'
3477 as valid option value of -march option.
3479 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3481 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3484 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3485 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3486 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3489 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3492 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3493 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3494 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3495 for RS6000_BTM_POWERPC64.
3496 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3497 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3498 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3503 2018-03-28 Carl Love <cel@us.ibm.com>
3506 2017-09-27 Carl Love <cel@us.ibm.com>
3508 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3509 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3510 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3513 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3515 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3516 instead of __vector bool.
3517 (_mm_max_pu8): Likewise.
3518 (_mm_min_pi16): Likewise.
3520 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3523 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3524 (DIVWEUO): Likewise.
3526 (DIVDEUO): Likewise.
3527 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3528 DIVWEUO and DIVDEUO.
3529 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3530 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3531 (div_extend): Likewise.
3532 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3534 (__builtin_divweuo): Likewise.
3535 (__builtin_divdeo): Likewise.
3536 (__builtin_divdeuo): Likewise.
3538 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3541 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3542 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3544 PR tree-optimization/82004
3545 * gimple-match-head.c (optimize_pow_to_exp): New function.
3546 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3547 Don't fold to exp if optimize_pow_to_exp is false.
3549 2018-03-28 Martin Liska <mliska@suse.cz>
3552 * calls.c (initialize_argument_information): Fix trailing space.
3553 * common.opt: Fix typo and provide better explanation for
3554 -fsanitize-coverage option.
3555 * config/i386/i386.opt: Fix typo.
3557 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3558 Martin Liska <mliska@suse.cz>
3561 * gimplify.c (asan_poison_variable): Don't do the check for
3562 gimplify_omp_ctxp here.
3563 (gimplify_decl_expr): Do it here.
3564 (gimplify_target_expr): Likewise.
3566 2018-03-28 Martin Liska <mliska@suse.cz>
3569 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3570 chkp_type_bounds_count if MPX is not enabled.
3572 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3574 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3576 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3579 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3580 function to create the function decl for complex long double
3581 multiply and divide for -mabi=ieeelongdouble.
3582 (init_float128_ieee): Call it.
3584 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3587 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3588 -fcf-protection=branch -mibt.
3589 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3591 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3594 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3596 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3599 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3600 extern array declarations.
3602 2018-03-27 Richard Biener <rguenther@suse.de>
3605 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3606 explicit single_use checks.
3608 2018-03-27 Richard Biener <rguenther@suse.de>
3610 PR tree-optimization/85082
3611 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3614 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3616 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3617 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3618 Turn on fasynchronous-unwind-tables and funwind-tables.
3620 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3623 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3624 (*bmi_blsr_<mode>_ccz): Ditto.
3626 2018-03-26 Tom de Vries <tom@codesourcery.com>
3628 PR tree-optimization/85063
3629 * omp-general.c (offloading_function_p): New function. Factor out
3631 * omp-offload.c (pass_omp_target_link::gate): ... here.
3632 * omp-general.h (offloading_function_p): Declare.
3633 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3634 with attribute omp declare target for offloading functions.
3636 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3638 PR tree-optimization/84005
3639 * tree-data-ref.h (get_base_for_alignment): Declare.
3640 * tree-data-ref.c (get_base_for_alignment_1): New function.
3641 (get_base_for_alignment): Likewise.
3642 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3643 get_base_for_alignment to find a suitable base object, instead
3644 of always using drb->base_address.
3646 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3649 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3650 known size by default.
3652 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3655 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3656 and non BLKmode operands.
3658 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3661 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3662 Clean up attributes.
3664 2018-03-23 Richard Biener <rguenther@suse.de>
3667 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3668 we are going to emit early debug for LTO.
3670 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3673 * function.c (match_asm_constraints_1): Don't optimize if input
3674 doesn't satisfy general_operand predicate for output's mode.
3677 * alias.c (write_dependence_p): Don't require for x_canonicalized
3678 non-VOIDmode if x has VOIDmode.
3681 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3682 just don't try to optimize it rather than assert it never happens.
3684 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3686 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3687 macro expansions for definition of ST_INTERNAL_<mode> and
3688 LD_INTERNAL_<mode> builtins.
3689 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3691 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3693 (altivec_expand_st_builtin): Likewise.
3694 (altivec_expand_builtin): Remove calls to deleted functions.
3695 (rs6000_address_for_altivec): Delete this function.
3696 * config/rs6000/vector.md: Remove expands for
3697 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3699 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3702 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3703 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3704 re-computing once computed.
3705 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3706 (arm_init_machine_status): Initialize
3707 machine->static_chain_stack_bytes.
3709 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3712 * doc/extend.texi: Add four new prototypes for vec_ld.
3713 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3714 definitions for more logical presentation.
3715 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3716 entries for V1TI variants of __builtin_altivec_ld builtin.
3717 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3718 handling of V1TI variant of LVX icode pattern.
3719 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3720 (rs6000_gimple_fold_builtin): Likewise.
3721 (altivec_init_builtins): Add code to define
3722 __builtin_altivec_lvx_v1ti function.
3724 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3727 * function.c (match_asm_constraints_1): Don't do the optimization
3728 if input isn't a REG, SUBREG, MEM or constant.
3730 2018-03-22 Tom de Vries <tom@codesourcery.com>
3732 PR tree-optimization/84956
3733 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3734 bb_has_abnormal_pred.
3736 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3739 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3740 DECL_INITIAL (decl) to decl at the end.
3741 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3744 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3746 * doc/extend.texi (__builtin_tgmath): Document when complex
3747 integer types are treated as _Complex _Float64.
3749 2018-03-21 Tom de Vries <tom@codesourcery.com>
3751 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3753 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3755 PR tree-optimization/84960
3756 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3757 if it is ENTRY block, move them into single succ of ENTRY in that case.
3759 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3761 PR tree-optimization/84811
3762 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3763 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3764 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3765 (known_subrange_p): Update accordingly. Cast each value involved
3766 in the size comparison, rather than casting the result of the
3769 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3771 PR tree-optimization/84982
3772 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3773 by flipping the least significant bit rather than all bits from
3774 bitpos to bitpos + bitsize - 1.
3776 2018-03-21 Nathan Sidwell <nathan@acm.org>
3778 * doc/extend.texi (Deprecated Features): Remove mention of
3779 long-deleted deprecations.
3781 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3784 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3785 * configure: Regenerate.
3787 2018-03-21 Tom de Vries <tom@codesourcery.com>
3789 PR tree-optimization/83126
3790 * tree-parloops.c (num_phis): New function.
3791 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3793 2018-03-21 Nathan Sidwell <nathan@acm.org>
3795 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3796 mention anon-struct/union members and trailing attributes.
3798 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3800 PR tree-optimization/84969
3801 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3802 builtin memset partitions if they set different rhs values.
3804 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3806 PR rtl-optimization/84989
3807 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3808 VEC_DUPLICATE with scalar result mode.
3810 2018-03-21 Martin Liska <mliska@suse.cz>
3813 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3814 not intended return statement.
3816 2018-03-21 Martin Liska <mliska@suse.cz>
3819 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3820 (chkp_find_bound_slots_1): Limit number of iterations.
3822 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3825 * Minor grammar fixes for x86 options.
3827 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3830 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3831 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3833 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3836 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3837 (altivec_lvx_<mode>_1op): Likewise.
3838 (altivec_stvx_<mode>_2op): Likewise.
3839 (altivec_stvx_<mode>_1op): Likewise.
3840 (altivec_lvx_<VM2:mode>): New define_expand.
3841 (altivec_stvx_<VM2:mode>): Likewise.
3842 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3843 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3844 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3845 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3846 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3847 (rs6000_gen_lvx): Likewise.
3848 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3849 (altivec_expand_stv_builtin): Likewise.
3850 (altivec_expand_builtin): Likewise.
3851 * config/rs6000/vector.md: Likewise.
3853 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3856 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3859 2018-03-20 Richard Biener <rguenther@suse.de>
3862 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3863 sign-conversions as zero, fall back to standard scalar_stmt
3866 2018-03-20 Martin Liska <mliska@suse.cz>
3869 * predict.c (rebuild_frequencies): Handle case when we have
3870 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3872 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3875 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3876 flag_section_anchors.
3877 * varasm.c (use_blocks_for_decl_p): Remove hack for
3878 dw2_force_const_mem.
3881 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3883 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3884 be created, use lowpart_subreg of operands[0] rather than operands[0]
3886 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3887 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3888 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3889 and n constraint instead of aarch64_shift_imm_di and Usd.
3890 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3891 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3893 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3896 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3897 to favor GPR over NEON registers.
3898 (<shift>di3_neon): Likewise.
3900 2018-03-20 Tom de Vries <tom@codesourcery.com>
3903 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3904 (nvptx_process_pars): Emit bar.sync asap and alap.
3906 2018-03-20 Tom de Vries <tom@codesourcery.com>
3909 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3910 seen_label if seen_label is already set.
3912 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3915 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3916 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3917 Use 1U instead of 1. Formatting fixes.
3920 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3921 instead of TREE_TYPE (s1) for the return value.
3923 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3925 PR tree-optimization/84946
3926 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3927 bitsize + bitsize in poly_uint64 rather than poly_int64.
3930 * dwarf2asm.c: Include fold-const.c.
3931 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3932 of decl rather than decl itself.
3934 PR rtl-optimization/84643
3935 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3937 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3940 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3941 calling assemble_variable.
3943 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3946 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3947 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3949 2018-03-19 Jim Wilson <jimw@sifive.com>
3952 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3953 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3954 (riscv_first_stack_step): Likewise.
3955 (riscv_option_override): Use STACK_BOUNDARY instead of
3957 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3959 (BIGGEST_ALIGNMENT): Set to 128.
3960 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3961 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3964 2018-03-19 Richard Biener <rguenther@suse.de>
3966 PR tree-optimization/84933
3967 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3968 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3970 2018-03-19 Richard Biener <rguenther@suse.de>
3972 PR tree-optimization/84859
3973 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3974 (cond_if_else_store_replacement): Perform sinking operation on
3975 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3976 Generalize what a BB with a single eligible store is.
3978 2018-03-19 Richard Biener <rguenther@suse.de>
3980 PR tree-optimization/84929
3981 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3982 chrec_is_positive against non-chrec arg.
3984 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3987 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3989 2018-03-18 Martin Liska <mliska@suse.cz>
3991 PR rtl-optimization/84635
3992 * regrename.c (build_def_use): Use matches_mode only when
3995 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3997 PR tree-optimization/84913
3998 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3999 vectorize chains of COND_EXPRs.
4001 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4003 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4005 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4007 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4009 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4011 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4013 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4014 Kito Cheng <kito.cheng@gmail.com>
4016 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4017 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4018 (nds32_adjust_reg_alloc_order): New function.
4019 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4021 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4023 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4024 nds32_print_operand, nds32_print_operand_address): Use
4025 HOST_WIDE_INT_PRINT_DEC instead.
4027 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4029 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4031 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4034 * config/i386/i386.c (initial_ix86_tune_features,
4035 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4037 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4038 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4039 rather than 1u << ix86_tune. Formatting fix.
4040 (ix86_option_override_internal): Change ix86_arch_mask from
4041 unsigned int to unsigned HOST_WIDE_INT, initialize to
4042 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4043 (ix86_function_specific_restore): Likewise.
4045 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4048 * postreload.c (reload_combine_recognize_pattern): Perform
4049 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4050 truncate_int_for_mode the result for the destination's mode.
4053 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4054 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4056 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4059 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4062 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4063 regno_allocno_class_array and sorted_pseudos.
4064 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4065 insns where regno is used.
4067 2018-03-16 Martin Liska <mliska@suse.cz>
4070 * multiple_target.c (create_dispatcher_calls): Redirect
4071 reference in the symbol table.
4073 2018-03-16 Martin Liska <mliska@suse.cz>
4076 * multiple_target.c (create_dispatcher_calls): Redirect also
4079 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4083 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4084 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4085 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4087 2018-03-16 Julia Koval <julia.koval@intel.com>
4089 * doc/invoke.texi (Skylake Server): Add CLWB.
4090 Cannonlake): Remove CLWB.
4092 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4094 PR tree-optimization/84841
4095 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4097 (FLOAT_ONE_CONST_TYPE): Define.
4098 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4099 (sort_by_operand_rank): Put entries with higher constant_type last
4100 rather than first to match comments.
4102 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4104 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4107 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4110 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4111 check and use address of target always.
4113 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4116 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4117 (indirect_thunk_bnd_needed): Likewise.
4118 (indirect_thunks_used): Likewise.
4119 (indirect_thunks_bnd_used): Likewise.
4120 (indirect_return_needed): New.
4121 (indirect_return_bnd_needed): Likewise.
4122 (output_indirect_thunk_function): Add a bool argument for
4124 (output_indirect_thunk_function): Don't generate alias for
4125 function return thunk.
4126 (ix86_code_end): Call output_indirect_thunk_function to generate
4127 function return thunks.
4128 (ix86_output_function_return): Set indirect_return_bnd_needed
4129 and indirect_return_needed instead of indirect_thunk_bnd_needed
4130 and indirect_thunk_needed.
4132 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4134 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4135 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4136 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4138 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4139 Paul Hua <paul.hua.gm@gmail.com>
4142 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4144 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4146 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4147 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4150 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4153 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4154 instead of GET_MODE_SIZE when comparing Units.
4156 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4159 * varasm.c (hash_section): Return an unchangeble hash value
4160 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4161 Return !aarch64_can_use_per_function_literal_pools_p ().
4163 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4166 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4167 rather than address of cmode as last argument to prepare_cmp_insn.
4169 2018-03-15 Julia Koval <julia.koval@intel.com>
4171 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4172 F_AVX512VNNI, F_AVX512BITALG): New.
4174 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4177 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4178 insn for floating-point loads and stores.
4180 2018-03-14 Carl Love <cel@us.ibm.com>
4182 * config/rs6000/rs6000-c.c: Add macro definitions for
4183 ALTIVEC_BUILTIN_VEC_PERMXOR.
4184 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4185 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4186 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4187 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4189 * config/doc/extend.texi: Add prototypes for vec_permxor.
4191 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4194 * diagnostic-show-locus.c (class layout_point): Convert m_line
4195 from int to linenum_type.
4196 (line_span::comparator): Use linenum "compare" function when
4197 comparing line numbers.
4198 (test_line_span): New function.
4199 (layout_range::contains_point): Convert param "row" from int to
4201 (layout_range::intersects_line_p): Likewise.
4202 (layout::will_show_line_p): Likewise.
4203 (layout::print_source_line): Likewise.
4204 (layout::should_print_annotation_line_p): Likewise.
4205 (layout::print_annotation_line): Likewise.
4206 (layout::print_leading_fixits): Likewise.
4207 (layout::annotation_line_showed_range_p): Likewise.
4208 (struct line_corrections): Likewise for field m_row.
4209 (line_corrections::line_corrections): Likewise for param "row".
4210 (layout::print_trailing_fixits): Likewise.
4211 (layout::get_state_at_point): Likewise.
4212 (layout::get_x_bound_for_row): Likewise.
4213 (layout::print_line): Likewise.
4214 (diagnostic_show_locus): Likewise for locals "last_line" and
4216 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4217 * input.c (selftest::test_linenum_comparisons): New function.
4218 (selftest::input_c_tests): Call it.
4219 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4220 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4221 * selftest.h (ASSERT_GT): New macro.
4222 (ASSERT_GT_AT): New macro.
4223 (ASSERT_LT): New macro.
4224 (ASSERT_LT_AT): New macro.
4226 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4228 PR rtl-optimization/84780
4229 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4231 2018-03-14 Martin Liska <mliska@suse.cz>
4233 * tree.c (record_node_allocation_statistics): Use
4234 get_stats_node_kind.
4235 (get_stats_node_kind): New function extracted from
4236 record_node_allocation_statistics.
4237 (free_node): Use get_stats_node_kind.
4239 2018-03-14 Richard Biener <rguenther@suse.de>
4241 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4242 that the value-set of ANTIC_IN doesn't grow.
4245 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4247 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4248 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4250 2018-03-14 Julia Koval <julia.koval@intel.com>
4252 * config.gcc (icelake-client, icelake-server): New.
4254 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4255 (initial_ix86_arch_features): Ditto.
4256 (PTA_SKYLAKE): Add SGX.
4257 (PTA_ICELAKE): Remove.
4258 (PTA_ICELAKE_CLIENT): New.
4259 (PTA_ICELAKE_SERVER): New.
4260 (ix86_option_override_internal): Split up icelake on icelake client and
4262 (get_builtin_code_for_version): Ditto.
4263 (fold_builtin_cpu): Ditto.
4264 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4265 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4266 * config/i386/i386.h (processor_type): Ditto.
4267 * doc/invoke.texi: Ditto.
4269 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4272 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4273 INTEGER_CST offset, add it together with bitpos / 8 and
4274 sign extend based on POINTER_SIZE.
4278 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4281 * config/i386/constraints.md (Yc): New register constraint.
4282 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4283 Use Yc constraint for alternative 2 of operand 0. Remove
4284 preferred_for_speed attribute.
4286 2018-03-14 Richard Biener <rguenther@suse.de>
4288 PR tree-optimization/84830
4289 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4290 with the old one to avoid oscillations.
4292 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4295 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4297 (assign_by_spills): Return a flag of reload assignment failure.
4298 Do not process the reload assignment failures. Do not spill other
4299 reload pseudos if they has the same reg class. Update n if
4301 (lra_assign): Add a return arg. Set up from the result of
4302 assign_by_spills call.
4303 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4304 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4305 usage_insns if it is not NULL.
4306 (spill_hard_reg_in_range): New function.
4307 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4308 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4309 function prototypes.
4310 (lra_assign): Change prototype.
4311 * lra.c (lra): Add code to deal with fails by splitting hard reg
4314 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4316 * config/riscv/riscv.opt (mrelax): New option.
4317 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4318 "norelax" when riscv_mrelax is disabled.
4319 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4321 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4324 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4325 reassociation for int modes.
4327 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4329 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4330 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4332 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4333 * config/aarch64/aarch64-sve.md
4334 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4335 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4336 (*extend<mode><Vwide>2): Rename to...
4337 (aarch64_sve_extend<mode><Vwide>2): ...this.
4338 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4339 renaming the old pattern to...
4340 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4342 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4343 define_expand, renaming the old pattern to...
4344 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4345 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4346 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4347 account when deciding which SVE instruction the optab should use.
4348 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4350 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4352 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4353 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4354 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4355 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4356 (tlsdesc_small_<mode>): Turn a define_expand and use
4357 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4358 (tlsdesc_small_advsimd_<mode>): ...this.
4359 (tlsdesc_small_sve_<mode>): New pattern.
4361 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4363 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4364 (UNSPEC_UMUL_HIGHPART): New constants.
4365 (MUL_HIGHPART): New int iteraor.
4366 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4367 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4369 (*<su>mul<mode>3_highpart): New define_insn.
4371 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4374 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4377 2018-03-13 Martin Liska <mliska@suse.cz>
4380 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4382 (sem_item_optimizer::~sem_item_optimizer): Release it.
4383 (sem_item_optimizer::merge_classes): Register variable aliases.
4384 (sem_item_optimizer::fixup_pt_set): New function.
4385 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4386 * ipa-icf.h: Declare new variables and functions.
4388 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4391 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4392 integer_pow2p@2 and test integer_pow2p in condition.
4393 (A < 0 ? C : 0): Similarly for @1.
4396 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4397 characters starting at p contain '\0' character, don't look beyond
4401 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4402 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4405 * reg-stack.c (change_stack): Change update_end var from int to
4406 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4407 also call set_block_for_insn on the newly added insns and rescan.
4410 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4411 on the last operand.
4414 * tree.c (stabilize_reference_1): Return save_expr (e) for
4415 STATEMENT_LIST even if it doesn't have side-effects.
4417 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4419 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4421 2018-03-12 Renlin Li <renlin.li@arm.com>
4423 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4424 aarch64_output_scalar_simd_mov_immediate.
4426 2018-03-12 Martin Sebor <msebor@redhat.com>
4428 PR tree-optimization/83456
4429 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4430 for perfectly overlapping calls to memcpy.
4431 (gimple_fold_builtin_memory_chk): Same.
4432 (gimple_fold_builtin_strcpy): Handle no-warning.
4433 (gimple_fold_builtin_stxcpy_chk): Same.
4434 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4436 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4438 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4439 parameter. Use it for SFmode.
4440 (rs6000_function_arg_advance_1): Adjust.
4441 (rs6000_function_arg): Adjust.
4442 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4444 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4446 PR rtl-optimization/84169
4447 PR rtl-optimization/84780
4448 * combine.c (can_combine_p): Check for a 2-insn combination whether
4449 the destination register is used between the two insns, too.
4451 2018-03-12 Richard Biener <rguenther@suse.de>
4453 PR tree-optimization/84803
4454 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4455 for refs DR analysis didn't process.
4457 2018-03-12 Richard Biener <rguenther@suse.de>
4459 PR tree-optimization/84777
4460 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4461 force-vectorize loops ignore whether we are optimizing for size.
4463 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4465 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4466 (TARGET_MD_ASM_ADJUST): Define.
4468 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4469 Kito Cheng <kito.cheng@gmail.com>
4470 Chung-Ju Wu <jasonwucj@gmail.com>
4472 * config/nds32/nds32.c (nds32_compute_stack_frame,
4473 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4474 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4475 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4476 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4477 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4478 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4479 * config/nds32/nds32.md (prologue, epilogue): Use macro
4480 NDS32_V3PUSH_AVAILABLE_P to do checking.
4482 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4485 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4486 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4487 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4488 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4489 addition of most attributes on !orig_type_die or the attribute not
4490 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4492 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4493 Chung-Ju Wu <jasonwucj@gmail.com>
4495 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4497 * config/nds32/nds32.opt (mvh): New option.
4499 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4500 Chung-Ju Wu <jasonwucj@gmail.com>
4502 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4504 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4505 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4508 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4509 Chung-Ju Wu <jasonwucj@gmail.com>
4511 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4513 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4514 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4516 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4517 Kito Cheng <kito.cheng@gmail.com>
4518 Chung-Ju Wu <jasonwucj@gmail.com>
4520 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4521 UNSPEC_FFMISM and UNSPEC_FLMISM.
4522 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4523 for ffb, ffmism and flmism.
4524 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4525 (unspec_ffmism): Ditto.
4526 (unspec_flmism): Ditto.
4527 (nds32_expand_builtin_impl): Check if string extension is available.
4528 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4529 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4531 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4534 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4537 * lra-assigns.c (assign_by_spills): Return a flag of reload
4538 assignment failure. Do not process the reload assignment
4539 failures. Do not spill other reload pseudos if they has the same
4541 (lra_assign): Add a return arg. Set up from the result of
4542 assign_by_spills call.
4543 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4544 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4545 usage_insns if it is not NULL.
4546 (spill_hard_reg_in_range): New function.
4547 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4548 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4549 function prototypes.
4550 (lra_assign): Change prototype.
4551 * lra.c (lra): Add code to deal with fails by splitting hard reg
4554 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4557 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4559 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4562 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4563 (dw_val_node): Add val_symbolic_view.
4564 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4565 (symview_upper_bound): New.
4566 (new_line_info_table): Initialize symviews_since_reset.
4567 (dwarf2out_source_line): Count symviews_since_reset and set
4568 symview_upper_bound.
4569 (dw_val_equal_p): Handle symview.
4570 (add_AT_symview): New.
4571 (print_dw_val): Handle symview.
4572 (attr_checksum, attr_checksum_ordered): Likewise.
4573 (same_dw_val_p, size_of_die): Likewise.
4574 (value_format, output_die): Likewise.
4575 (add_high_low_attributes): Use add_AT_symview for entry_view.
4576 (dwarf2out_finish): Reset symview_upper_bound, clear
4579 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4582 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4583 Add strict argument and use it.
4584 (rs6000_split_multireg_move): Update for new strict argument.
4585 (mem_operand_gpr): Disallow all non-offsettable addresses.
4586 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4588 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4591 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4592 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4593 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4596 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4597 decl, use remap_type if we want to use the type.
4599 2018-03-09 Martin Sebor <msebor@redhat.com>
4601 PR tree-optimization/84526
4602 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4604 (builtin_access::generic_overlap): Be prepared to handle non-array
4607 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4609 PR rtl-optimization/84682
4610 * lra-constraints.c (process_address_1): Check is_address flag
4611 for address constraints.
4612 (process_alt_operands): Likewise.
4613 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4614 preprocess_constraints.
4615 * recog.h (preprocess_constraints): Add oploc parameter.
4617 * recog.c (preprocess_constraints): Test address_operand for
4618 CT_ADDRESS constraints.
4620 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4623 * lra-assigns.c (assign_by_spills): Return a flag of reload
4624 assignment failure. Do not process the reload assignment
4625 failures. Do not spill other reload pseudos if they has the same
4627 (lra_assign): Add a return arg. Set up from the result of
4628 assign_by_spills call.
4629 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4630 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4631 usage_insns if it is not NULL.
4632 (spill_hard_reg_in_range): New function.
4633 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4634 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4635 function prototypes.
4636 (lra_assign): Change prototype.
4637 * lra.c (lra): Add code to deal with fails by splitting hard reg
4640 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4643 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4644 Accept complain bool parameter. Only emit errors if it is true.
4645 (arm_parse_cpu_option_name): Likewise.
4646 (arm_target_thumb_only): Adjust callers of the above.
4647 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4648 prototype to take a default true bool parameter.
4649 (arm_parse_arch_option_name): Likewise.
4651 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4652 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4656 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4657 * configure: Regenerate.
4658 * configure.ac ("linker --version-script option"): New.
4659 ("linker soname option"): New.
4661 2018-03-09 Richard Biener <rguenther@suse.de>
4663 PR tree-optimization/84775
4664 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4665 immediate uses of predicate stmts and mark them modified.
4668 PR tree-optimization/84178
4669 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4671 (version_loop_for_if_conversion): Delay update_ssa call.
4672 (tree_if_conversion): Delay update_ssa until after predicate
4675 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4678 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4679 when the function accesses prior frames.
4681 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4684 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4685 gen_llsym, otherwise call maybe_gen_llsym.
4688 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4689 has ',' character inside of it.
4691 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4694 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4695 as clobbering CC_REGNUM.
4697 2018-03-08 Richard Biener <rguenther@suse.de>
4700 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4701 (follow_copies_to_constant): Do not follow SSA names registered
4704 2018-03-08 Richard Biener <rguenther@suse.de>
4706 PR tree-optimization/84178
4707 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4709 (version_loop_for_if_conversion): Delay update_ssa call.
4710 (tree_if_conversion): Delay update_ssa until after predicate
4713 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4715 PR tree-optimization/84178
4716 * tree-if-conv.c (release_bb_predicate): Remove the
4717 the assertion that the stmts have NULL use_ops.
4718 Discard the statements, asserting that they haven't
4719 yet been added to a BB.
4721 2018-03-08 Richard Biener <rguenther@suse.de>
4723 PR tree-optimization/84746
4724 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4725 (phi_translate): Pass in destination ANTIC_OUT set.
4726 (phi_translate_1): Likewise. For a simplified result lookup
4727 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4728 (phi_translate_set): Adjust.
4729 (do_pre_regular_insertion): Likewise.
4730 (do_pre_partial_partial_insertion): Likewise.
4732 2018-03-08 Martin Liska <mliska@suse.cz>
4734 PR gcov-profile/84735
4735 * doc/gcov.texi: Document usage of profile files.
4736 * gcov-io.h: Document changes in the format.
4738 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4742 * dwarf2out.c (struct dw_line_info_table): Update comments for
4744 (FORCE_RESET_NEXT_VIEW): New.
4745 (FORCE_RESETTING_VIEW_P): New.
4746 (RESETTING_VIEW_P): Check for -1 too.
4747 (ZERO_VIEW_P): Likewise.
4748 (new_line_info_table): Force-reset next view.
4749 (dwarf2out_begin_function): Likewise.
4750 (dwarf2out_source_line): Simplify zero_view_p initialization.
4751 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4752 view directly. Omit view when omitting .loc at line 0.
4754 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4756 PR tree-optimization/84740
4757 * tree-switch-conversion.c (process_switch): Call build_constructors
4758 only if info.phi_count is non-zero.
4760 PR tree-optimization/84739
4761 * tree-tailcall.c (find_tail_calls): Check call arguments against
4762 DECL_ARGUMENTS (current_function_decl) rather than
4763 DECL_ARGUMENTS (func) when checking for tail recursion.
4765 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4767 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4768 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4769 Volker Reichelt's entry and add entries for people that perform
4770 GCC fuzzy testing and report numerous bugs.
4772 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4775 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4776 readonly data in sdata, if that is disabled.
4777 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4778 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4779 -mreadonly-in-sdata option.
4781 2018-03-07 Martin Sebor <msebor@redhat.com>
4783 PR tree-optimization/84468
4784 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4785 basic block when looking for nul assignment.
4787 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4790 * except.h (output_function_exception_table): Adjust prototype.
4791 * except.c (output_function_exception_table): Remove FNNAME parameter
4792 and add SECTION parameter. Ouput one part of the table at a time.
4793 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4794 the first part of the exception table and emit unwind directives.
4795 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4796 (i386_pe_seh_cold_init): Likewise.
4797 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4798 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4799 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4800 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4801 * config/i386/winnt.c: Include except.h.
4802 (struct seh_frame_state): Add reg_offset, after_prologue and
4803 in_cold_section fields.
4804 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4805 (i386_pe_seh_cold_init): New function.
4806 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4807 to seh->in_cold_section.
4808 (seh_emit_push): Record the offset of the push.
4809 (seh_emit_save): Record the offet of the save.
4810 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4811 Test seh->after_prologue to disregard the epilogue.
4812 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4813 (i386_pe_end_cold_function): New function.
4815 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4818 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4819 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4822 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4824 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4825 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4828 * multiple_target.c: Include tree-inline.h and intl.h.
4829 (expand_target_clones): Diagnose and fail if node->definition and
4830 !tree_versionable_function_p (node->decl).
4832 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4834 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4836 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4837 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4838 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4840 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4843 * combine.c (try_combine): Use reg_or_subregno instead of handling
4844 just paradoxical SUBREGs and REGs.
4846 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4848 * config/arc/arc.c (arc_finalize_pic): Remove function.
4849 (arc_must_save_register): We use single base PIC register, remove
4850 checks to save/restore the PIC register.
4851 (arc_expand_prologue): Likewise.
4852 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4854 (arc_verify_short): Likewise.
4855 (arc_attr_type): Likewise.
4856 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4857 (walk_stores): Likewise.
4858 (arc_address_cost): Make it static.
4859 (arc_verify_short): Likewise.
4860 (branch_dest): Likewise.
4861 (arc_attr_type): Likewise.
4862 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4863 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4864 (arc_final_prescan_insn): Remove inserting the nops due to
4865 hardware hazards. It is done in reorg step.
4866 (insn_length_variant_t): Remove.
4867 (insn_length_parameters_t): Likewise.
4868 (arc_insn_length_parameters): Likewise.
4869 (arc_get_insn_variants): Likewise.
4870 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4872 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4875 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4878 PR tree-optimization/84687
4879 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4881 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4883 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4885 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4886 Rename to ppc_speculation_barrier.
4887 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4888 __builtin_ppc_speculation_barrier.
4890 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4893 * combine.c (combine_simplify_rtx): Don't try to simplify if
4894 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4897 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4899 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4900 to 32 bytes when compiling for POWER9.
4902 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4905 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4906 regparm >= 3 with no arg reg available also for calls with
4907 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4910 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4912 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4914 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4917 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4919 2018-03-05 Richard Biener <rguenther@suse.de>
4921 PR tree-optimization/84486
4922 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4923 When inserting a __builtin_assume_aligned call set the LHS
4924 SSA name alignment info accordingly.
4926 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4928 PR tree-optimization/84114
4929 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4930 Avoid reassociation of FLOAT_MODE addition.
4932 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4934 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4935 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4936 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4937 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4938 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4939 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4940 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4942 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4943 __builtin_ia32_wbinvd): New builtins.
4944 (SPECIAL_ARGS2): New.
4945 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4946 (SPECIAL_ARGS2): New.
4947 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4948 (ix86_valid_target_attribute_inner_p): Ditto.
4949 (ix86_init_mmx_sse_builtins): Add special_args2.
4950 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4951 TARGET_WBNOINVD_P): New.
4952 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4953 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4954 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4955 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4956 * config/i386/pconfigintrin.h: New file.
4957 * config/i386/wbnoinvdintrin.h: Ditto.
4958 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4959 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4961 2018-03-05 Richard Biener <rguenther@suse.de>
4963 PR tree-optimization/84670
4964 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4966 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4967 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4968 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4969 doesn't grow if all successors (recursively) were visited at least
4972 2018-03-05 Richard Biener <rguenther@suse.de>
4974 PR tree-optimization/84650
4975 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4976 if executed in the loop pipeline.
4978 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4980 * doc/configfiles.texi (Configuration Files): Move info about
4981 conditionalizing $target-protos.h to...
4982 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4983 differs from $target-protos.h.
4985 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4986 Chung-Ju Wu <jasonwucj@gmail.com>
4988 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4989 * config/nds32/nds32-multiple.md (setmemsi): Define.
4990 * config/nds32/nds32-memory-manipulation.c
4991 (nds32_gen_dup_4_byte_to_word_value): New.
4992 (emit_setmem_word_loop): New.
4993 (emit_setmem_byte_loop): New.
4994 (nds32_expand_setmem_loop): New.
4995 (nds32_expand_setmem_loop_v3m): New.
4996 (nds32_expand_setmem_unroll): New.
4997 (nds32_expand_setmem): New.
4999 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5000 Chung-Ju Wu <jasonwucj@gmail.com>
5002 * config/nds32/nds32-memory-manipulation.c
5003 (nds32_emit_load_store): New.
5004 (nds32_emit_post_inc_load_store): New.
5005 (nds32_emit_mem_move): New.
5006 (nds32_emit_mem_move_block): New.
5007 (nds32_expand_movmemsi_loop_unknown_size): New.
5008 (nds32_expand_movmemsi_loop_known_size): New.
5009 (nds32_expand_movmemsi_loop): New.
5010 (nds32_expand_movmemsi_unroll): New.
5011 (nds32_expand_movmemqi): Rename ...
5012 (nds32_expand_movmemsi): ... to this.
5013 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5014 (movmemsi): ... to this.
5015 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5016 (nds32_expand_movmemsi): ... to this.
5018 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5019 Monk Chiang <sh.chiang04@gmail.com>
5020 Chung-Ju Wu <jasonwucj@gmail.com>
5022 * config/nds32/nds32-protos.h
5023 (nds32_expand_load_multiple): New arguments.
5024 (nds32_expand_store_multiple): Ditto.
5025 (nds32_valid_multiple_load_store): Rename ...
5026 (nds32_valid_multiple_load_store_p): ... to this.
5027 * config/nds32/nds32-memory-manipulation.c
5028 (nds32_expand_load_multiple): Refine implementation.
5029 (nds32_expand_store_multiple): Ditto.
5030 * config/nds32/nds32-multiple.md
5031 (load_multiple): Update nds32_expand_load_multiple interface.
5032 (store_multiple): Update nds32_expand_store_multiple interface.
5033 * config/nds32/nds32-predicates.c
5034 (nds32_valid_multiple_load_store): Rename ...
5035 (nds32_valid_multiple_load_store_p): ... to this and refine
5037 * config/nds32/predicates.md
5038 (nds32_load_multiple_and_update_address_operation): New predicate.
5039 (nds32_store_multiple_and_update_address_operation): New predicate.
5041 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5042 Chung-Ju Wu <jasonwucj@gmail.com>
5044 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5045 (combo): New attribute.
5046 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5048 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5050 * config/nds32/nds32.opt: Change -mcmodel= default value.
5052 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5053 Monk Chiang <sh.chiang04@gmail.com>
5054 Chung-Ju Wu <jasonwucj@gmail.com>
5056 * config/nds32/constants.md (unspec_element): New enum.
5057 * config/nds32/constraints.md (Umw): New constraint.
5058 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5059 * config/nds32/nds32-intrinsic.md: Likewise.
5060 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5061 (nds32_valid_smw_lwm_base_p): New.
5062 (nds32_output_smw_single_word): New.
5063 (nds32_output_lmw_single_word): New.
5064 (nds32_expand_unaligned_load): New.
5065 (nds32_expand_unaligned_store): New.
5066 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5067 (nds32_output_smw_single_word): Declare.
5068 (nds32_output_lmw_single_word): Declare.
5069 (nds32_expand_unaligned_load): Declare.
5070 (nds32_expand_unaligned_store): Declare.
5071 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5072 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5073 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5074 NDS32_BUILTIN_UASTORE_DW.
5075 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5078 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5079 Kito Cheng <kito.cheng@gmail.com>
5080 Chung-Ju Wu <jasonwucj@gmail.com>
5082 * config/nds32/nds32-intrinsic.c
5083 (nds32_expand_builtin_null_ftype_reg): Delete.
5084 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5085 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5086 (nds32_read_argument): New.
5087 (nds32_legitimize_target): Ditto.
5088 (nds32_legitimize_argument): Ditto.
5089 (nds32_check_constant_argument): Ditto.
5090 (nds32_expand_unop_builtin): Ditto.
5091 (nds32_expand_unopimm_builtin): Ditto.
5092 (nds32_expand_binop_builtin): Ditto.
5093 (nds32_builtin_decl_impl): Ditto.
5094 (builtin_description): Ditto.
5095 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5096 (nds32_init_builtins_impl): Ditto.
5097 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5098 (nds32_builtin_decl): New.
5099 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5100 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5102 2018-03-02 Jeff Law <law@redhat.com>
5104 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5105 (redundant_insn, fill_simple_delay_slots): Likewise.
5106 (fill_slots_from_thread): Likewise.
5107 * resource.c (mark_referenced_resources): Likewise.
5108 (mark_set_resources, find_dead_or_set_registers): Likewise.
5110 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5112 * substring-locations.h (format_warning_va): Formatting fix for
5114 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5116 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5117 * substring-locations.c: Include intl.h.
5118 (format_warning_va): Turned into small wrapper around
5119 format_warning_n_va, renamed to ...
5120 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5121 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5123 (format_warning_at_substring_n): New function.
5124 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5125 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5126 format_warning_at_substring with just a shorter name instead of
5127 const function pointer.
5128 (fmtwarn_n): New function.
5129 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5130 appropriate, get rid of all the fmtstr temporaries, move conditionals
5131 with G_() wrapped string literals directly into fmtwarn arguments,
5132 cast dir.len to (int), formatting fixes.
5134 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5136 * doc/invoke.texi: Remove "Cilk Plus" references.
5138 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5139 Richard Biener <rguenther@suse.de>
5142 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5143 for error or warning attributes if CALL_FROM_THUNK_P is set.
5146 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5149 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5150 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5153 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5154 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5156 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5157 instead of -1U in last predictors element's probability member.
5159 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5162 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5163 arguments if they are comparable.
5165 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5167 PR tree-optimization/84634
5168 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5169 masks and masked_loop_p with a single loop_masks, making sure it's
5170 null for bb vectorization.
5172 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5174 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5175 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5176 loop->force_vectorize to check whether there is no alias.
5178 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5181 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5183 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5185 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5186 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5187 instead of a loop around prev_real_insn.
5188 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5192 * config/i386/i386.c (ix86_print_operand): Use conditional
5193 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5196 2018-03-02 Richard Biener <rguenther@suse.de>
5198 PR tree-optimization/84427
5199 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5200 (bitmap_set_subtract_values): Rewrite to handle multiple
5203 (prune_clobbered_mems): Likewise.
5204 (phi_translate): Take edge instead of pred/phiblock.
5205 (phi_translate_1): Likewise.
5206 (phi_translate_set): Likewise. Insert all translated
5207 exprs for a value into the set, keeping possibly multiple
5208 expressions per value.
5209 (compute_antic_aux): Adjust for phi_translate changes.
5210 When intersecting union the expressions and prune those
5211 not in the final value set, keeping possibly multiple
5212 expressions per value. Do not use value-insertion
5213 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5214 all expressions. Add verification that the value-sets
5215 only shrink during iteration.
5216 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5217 (do_pre_regular_insertion): Likewise.
5218 (do_pre_partial_partial_insertion): Likewise.
5220 2018-03-02 Richard Biener <rguenther@suse.de>
5223 * config/darwin.c (saved_debug_info_level): New static global.
5224 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5225 (darwin_asm_lto_end): Restore debug info generation settings.
5227 2018-03-01 Martin Liska <mliska@suse.cz>
5230 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5233 2018-03-01 Richard Biener <rguenther@suse.de>
5236 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5237 type annotation with LTO when debug was disabled at compile-time.
5239 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5241 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5243 (mips_final_postscan_insn): Likewise.
5245 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5247 PR rtl-optimization/84528
5248 * alias.c (init_alias_target): Add commentary.
5249 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5250 a unique base value if the frame pointer is not eliminated
5251 to the stack pointer.
5253 2018-03-01 Tom de Vries <tom@codesourcery.com>
5255 PR rtl-optimization/83327
5256 * lra-int.h (hard_regs_spilled_into): Declare.
5257 * lra.c (hard_regs_spilled_into): Define.
5258 (init_reg_info): Init hard_regs_spilled_into.
5259 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5260 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5261 (process_bb_lives): Handle hard_regs_spilled_into.
5262 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5263 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5265 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5267 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5268 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5269 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5270 * config/rs6000/aix72.h: New file.
5272 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5274 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5275 instead of warning_at with conditional singular and plural messages
5279 * stor-layout.c (update_alignment_for_field): For
5280 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5281 && !DECL_PACKED (field), do the alignment update, just use
5282 only desired_align instead of MAX (type_align, desired_align)
5284 (place_field): Don't do known_align < desired_align handling
5285 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5286 is non-NULL, instead do it after rli->prev_field handling and
5287 only if not within a bitfield word. For DECL_PACKED (field)
5288 use type_align of BITS_PER_UNIT.
5290 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5292 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5293 superfluous parentheses and trailing spaces.
5295 2018-02-28 Richard Biener <rguenther@suse.de>
5297 PR tree-optimization/84584
5298 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5299 SCoPs with fake exit edge.
5301 2018-02-28 Martin Liska <mliska@suse.cz>
5304 * timevar.c (timer::print): Fix format to properly print 100%
5307 2018-02-28 Richard Biener <rguenther@suse.de>
5310 * genmatch.c (capture_info::walk_match): Do not mark
5311 captured expressions without operands as expr_p given
5312 they act more like predicates and should be subject to
5313 "lost tail" side-effect preserving.
5315 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5317 PR rtl-optimization/81611
5318 * auto-inc-dec.c (attempt_change): Move dead note from
5319 mem_insn if it's the next use of regno
5320 (find_address): Take address use of reg holding
5321 non-incremented value. Add parm to limit search to the named
5323 (merge_in_block): Attempt to use a mem insn that is the next
5324 use of the original regno.
5326 2018-02-27 Martin Sebor <msebor@redhat.com>
5329 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5330 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5332 2018-02-27 Martin Sebor <msebor@redhat.com>
5334 PR translation/84207
5335 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5336 n argument to unsigned HOST_WIDE_INT.
5337 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5338 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5339 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5340 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5342 2018-02-27 Richard Biener <rguenther@suse.de>
5344 PR tree-optimization/84512
5345 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5346 Do not use the estimate returned from record_stmt_cost for
5347 the scalar iteration cost but sum properly using add_stmt_cost.
5349 2018-02-27 Richard Biener <rguenther@suse.de>
5351 PR tree-optimization/84466
5352 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5353 Adjust last change to less strictly validate use operands.
5355 2018-02-27 Martin Liska <mliska@suse.cz>
5357 PR gcov-profile/84548
5358 * gcov.c (process_file): Allow partial overlap and consider it
5359 also as group functions.
5360 (output_lines): Properly calculate range of lines for a group.
5362 2018-02-27 Martin Liska <mliska@suse.cz>
5364 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5365 'ggc' suffixes. Change first column width.
5366 (timer::print): Fix formatting of the column.
5368 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5370 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5371 preserve inline entry blocks for the sake of debug inline
5372 entry point markers alone.
5373 (remove_unused_locals): Suggest in comments a better place to
5374 force the preservation of inline entry blocks that are
5375 otherwise unused, but do not preserve them.
5377 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5379 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5381 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5384 * config/i386/constraints.md (Bs): Replace
5385 ix86_indirect_branch_register with
5386 TARGET_INDIRECT_BRANCH_REGISTER.
5388 * config/i386/i386.md (indirect_jump): Likewise.
5389 (tablejump): Likewise.
5390 (*sibcall_memory): Likewise.
5391 (*sibcall_value_memory): Likewise.
5392 Peepholes of indirect call and jump via memory: Likewise.
5393 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5394 (*sibcall_value_GOT_32): Likewise.
5395 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5396 (GOT_memory_operand): Likewise.
5397 (call_insn_operand): Likewise.
5398 (sibcall_insn_operand): Likewise.
5399 (GOT32_symbol_operand): Likewise.
5400 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5402 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5404 PR rtl-optimization/83496
5405 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5406 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5407 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5408 redundant insn, if any.
5409 (relax_delay_slots): Likewise.
5410 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5412 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5414 PR tree-optimization/83965
5415 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5416 that grouped statements are part of a reduction chain. Return
5417 true if the statement is not marked as a reduction itself but
5419 (vect_recog_dot_prod_pattern): Don't check whether the statement
5420 is part of a group here.
5421 (vect_recog_sad_pattern): Likewise.
5422 (vect_recog_widen_sum_pattern): Likewise.
5424 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5427 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5429 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5432 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5434 (ix86_output_indirect_function_return): New prototype.
5435 (ix86_split_simple_return_pop_internal): Likewise.
5436 * config/i386/i386.c (indirect_return_via_cx): New.
5437 (indirect_return_via_cx_bnd): Likewise.
5438 (indirect_thunk_name): Handle return va CX_REG.
5439 (output_indirect_thunk_function): Create alias for
5440 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5441 (ix86_output_indirect_jmp): Remove the bool argument.
5442 (ix86_output_indirect_function_return): New function.
5443 (ix86_split_simple_return_pop_internal): Likewise.
5444 * config/i386/i386.md (*indirect_jump): Don't pass false
5445 to ix86_output_indirect_jmp.
5446 (*tablejump_1): Likewise.
5447 (simple_return_pop_internal): Change it to define_insn_and_split.
5448 Call ix86_split_simple_return_pop_internal to split it for
5450 (simple_return_indirect_internal): Call
5451 ix86_output_indirect_function_return instead of
5452 ix86_output_indirect_jmp.
5454 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5457 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5458 memset and value initialization afterwards.
5460 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5462 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5464 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5467 * common/config/aarch64/aarch64-common.c
5468 (aarch_option_optimization_table[]): Switch
5469 off fomit-frame-pointer
5471 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5472 Chung-Ju Wu <jasonwucj@gmail.com>
5474 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5476 (store_multiple): Ditto.
5478 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5480 * config.gcc: Add --with-cpu support for nds32 target.
5481 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5482 * config/nds32/nds32.opt: Add -mcpu= option.
5484 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5486 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5487 isel=yes): Warn for these deprecated options.
5489 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5491 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5492 ISA_2_5_MASKS_EMBEDDED.
5494 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5496 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5497 p->max as pointers rather than using iterative_hash_expr.
5499 2018-02-23 Carl Love <cel@us.ibm.com>
5501 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5502 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5504 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5505 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5506 P8V_BUILTIN_VEC_VUNSIGNED2.
5508 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5511 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5512 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5514 * lra-constraints.c (curr_insn_transform): Set up
5515 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5517 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5518 * lra-eliminations.c (spill_pseudos): Ditto.
5519 (process_insn_for_elimination): Ditto.
5520 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5521 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5522 LRA_NON_CLOBBERED_ALT.
5524 2018-02-22 Martin Sebor <msebor@redhat.com>
5526 PR tree-optimization/84480
5527 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5528 to maybe_diag_stxncpy_trunc. Call it.
5529 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5530 from gimple_fold_builtin_strcpy. Print inlining stack.
5531 (handle_builtin_stxncpy): Print inlining stack.
5532 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5534 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5537 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5538 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5539 and -fcheck-pointer-bounds are used together.
5540 (indirect_thunk_prefix): New enum.
5541 (indirect_thunk_need_prefix): New function.
5542 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5543 "_nt" instead of "_bnd" for NOTRACK prefix.
5544 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5545 (output_indirect_thunk_function): Likewise.
5547 (ix86_code_end): Update output_indirect_thunk_function calls.
5548 (ix86_output_indirect_branch_via_reg): Replace
5549 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5550 (ix86_output_indirect_branch_via_push): Likewise.
5551 (ix86_output_function_return): Likewise.
5552 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5553 incompatible with -fcf-protection=branch and
5554 -fcheck-pointer-bounds.
5556 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5559 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5560 Change gcc_assert call to output_operand_lossage.
5562 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5564 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5566 2018-02-22 DJ Delorie <dj@redhat.com>
5567 Sebastian Perta <sebastian.perta@renesas.com>
5568 Oleg Endo <olegendo@gcc.gnu.org>
5570 * config/rx/rx.c (rx_rtx_costs): New function.
5571 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5573 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5575 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5577 2018-02-22 Martin Liska <mliska@suse.cz>
5580 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5581 Add "native" as a possible value.
5583 2018-02-22 Martin Liska <mliska@suse.cz>
5586 * config/i386/i386.c (ix86_option_override_internal):
5587 Add "native" as a possible value for -march and -mtune.
5589 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5592 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5593 to all type variants.
5595 PR tree-optimization/84503
5596 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5597 width as info->bitpos + info->bitsize - start.
5598 (merged_store_group::merge_overlapping): Simplify width computation.
5599 (check_no_overlap): New function.
5600 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5601 start + width and last_order of the group, fail if check_no_overlap
5603 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5604 to group if check_no_overlap fails.
5606 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5608 * config/rs6000/altivec.md: Delete contraint arguments to
5609 define_expand, define_split, and define_peephole2, and in
5610 define_insn_and_split if always unused.
5611 * config/rs6000/darwin.md: Ditto.
5612 * config/rs6000/dfp.md: Ditto.
5613 * config/rs6000/rs6000.md: Ditto.
5614 * config/rs6000/sync.md: Ditto.
5615 * config/rs6000/vector.md: Ditto.
5616 * config/rs6000/vsx.md: Ditto.
5618 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5620 * config/rs6000/altivec.md: Write output control strings as braced
5621 blocks instead of double-quoted strings.
5622 * config/rs6000/darwin.md: Ditto.
5623 * config/rs6000/rs6000.md: Ditto.
5624 * config/rs6000/vector.md: Ditto.
5625 * config/rs6000/vsx.md: Ditto.
5627 2018-02-21 Jason Merrill <jason@redhat.com>
5629 PR c++/84314 - ICE with templates and fastcall attribute.
5630 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5632 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5634 * ipa-cp.c (determine_versionability): Fix comment typos.
5636 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5639 * ipa-cp.c (determine_versionability): Do not version functions caling
5642 2018-02-21 Martin Liska <mliska@suse.cz>
5645 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5646 Add "native" as a possible value.
5647 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5648 the macro when native cpu detection is available.
5650 2018-02-21 Martin Liska <mliska@suse.cz>
5653 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5654 Add "native" as a possible value.
5655 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5656 when native cpu detection is available.
5658 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5659 Martin Sebor <msebor@redhat.com>
5661 PR tree-optimization/84478
5662 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5664 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5665 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5666 support which is conservatively correct, for 2 only stay conservative
5667 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5668 argument to the 2 argument get_range_strlen, adjust 6 arg
5669 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5671 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5672 (gimple_fold_builtin_strlen): Pass true as last argument to
5675 2018-02-20 Martin Sebor <msebor@redhat.com>
5678 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5679 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5680 (builtin_memref::builtin_memref): Factor out parts into
5681 set_base_and_offset and call it.
5683 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5686 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5687 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5688 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5689 search at the associated MODE_INT.
5691 2018-02-20 Jeff Law <law@redhat.com>
5694 PR tree-optimization/81592
5696 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5697 for range data rather than using global data.
5698 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5699 range data rather than using global data.
5700 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5701 pass it to children as needed.
5702 (struct directive::fmtresult): Similarly.
5703 (struct directive::set_width): Similarly.
5704 (struct directive::set_precision): Similarly.
5705 (format_integer, format_directive, parse_directive): Similarly.
5706 (format_none): Accept unnamed vr_values parameter.
5707 (format_percent, format_floating, format_character): Similarly.
5708 (format_string, format_plain): Similarly.
5709 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5710 the EVRP range analyzer for range data rather than using global data.
5711 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5712 gimple-ssa-evrp-analyze.h
5713 (class sprintf_dom_walker): Add after_dom_children member function.
5714 Add evrp_range_analyzer member.
5715 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5716 range analyzer as needed.
5717 (sprintf_dom_walker::after_dom_children): New member function.
5718 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5720 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5721 (evrp_range_analyzer::pop_to_marker): Likewise.
5723 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5725 PR tree-optimization/84419
5726 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5727 with the required type if its current type is compatible but
5730 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5733 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5734 after vectorization.
5736 2018-02-20 Martin Liska <mliska@suse.cz>
5739 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5740 possible values if we don't have a hint.
5742 2018-02-20 Martin Liska <mliska@suse.cz>
5746 * final.c (shorten_branches): Build align_tab array with one
5748 * opts.c (finish_options): Add alignment option limit check.
5749 (MAX_CODE_ALIGN): Likewise.
5750 (MAX_CODE_ALIGN_VALUE): Likewise.
5751 * doc/invoke.texi: Document maximum allowed option value for
5752 all -falign-* options.
5754 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5757 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5758 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5759 * var-tracking.c (emit_note_insn_var_location): Remove all references
5760 to NOTE_INSN_CALL_ARG_LOCATION.
5761 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5762 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5763 Use copy_rtx_if_shared.
5764 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5765 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5766 (dwarf2out_var_location): Remove handling of
5767 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5769 * final.c (final_scan_insn): Remove all references to
5770 NOTE_INSN_CALL_ARG_LOCATION.
5771 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5772 before dumping final insns.
5773 * except.c (emit_note_eh_region_end): Remove all references to
5774 NOTE_INSN_CALL_ARG_LOCATION.
5775 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5776 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5777 * config/arc/arc.c (hwloop_optimize): Likewise.
5778 * config/arm/arm.c (create_fix_barrier): Likewise.
5779 * config/s390/s390.c (s390_chunkify_start): Likewise.
5780 * config/sh/sh.c (find_barrier): Likewise.
5781 * config/i386/i386.c (rest_of_insert_endbranch,
5782 ix86_seh_fixup_eh_fallthru): Likewise.
5783 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5784 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5785 * config/frv/frv.c (frv_function_prologue): Likewise.
5786 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5788 (note_outside_basic_block_p): Remove all references to
5789 NOTE_INSN_CALL_ARG_LOCATION.
5790 * gengtype.c (adjust_field_rtx_def): Likewise.
5791 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5793 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5794 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5797 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5800 PR tree-optimization/84452
5801 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5802 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5805 2018-02-19 Martin Liska <mliska@suse.cz>
5808 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5810 2018-02-19 Martin Liska <mliska@suse.cz>
5811 Richard Sandiford <richard.sandiford@linaro.org>
5813 PR tree-optimization/82491
5814 * gimple-fold.c (get_base_constructor): Make earlier bail out
5817 2018-02-19 Carl Love <cel@us.ibm.com>
5819 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5820 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5822 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5823 P8V_BUILTIN_VEC_NEG.
5825 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5827 * config/rl78/rl78.md (movdf): New define expand.
5829 2018-02-19 Martin Liska <mliska@suse.cz>
5832 * doc/invoke.texi: Fix typo.
5833 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5835 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5837 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5838 handle rs6000_single_float and rs6000_double_float specially for
5841 2018-02-16 Jeff Law <law@redhat.com>
5843 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5846 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5849 * ipa-inline.c (inline_small_functions): Fix a typo.
5851 2018-02-16 Nathan Sidwell <nathan@acm.org>
5853 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5855 2018-02-16 Carl Love <cel@us.ibm.com>
5857 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5858 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5859 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5860 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5861 expansion to P8V_BUILTIN_VEC_FLOAT2.
5863 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5865 PR rtl-optimization/70023
5866 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5867 src_regno into account.
5869 2018-02-16 Carl Love <cel@us.ibm.com>
5871 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5872 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5873 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5874 * config/rs6000/rs6000.c: Remove case statements for
5875 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5876 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5877 and P9V_BUILTIN_VEC_VINSERT4B.
5878 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5879 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5880 * config/rs6000/vsx.md:
5881 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5884 2018-02-16 Carl Love <cel@us.ibm.com>
5886 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5888 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5890 * config/rs6000/rs6000-c.c: Add the definitions for
5891 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5892 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5893 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5894 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5895 definition for insert4b and define insn *insert3b_internal.
5896 * doc/extend.texi: Add documentation for vec_extract4b.
5898 2018-02-16 Nathan Sidwell <nathan@acm.org>
5900 * doc/extend.texi (Backwards Compatibility): Mention friend
5901 injection. Note for-scope is deprecated.
5902 * doc/invoke.texi (-ffriend-injection): Deprecate.
5904 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5906 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5907 that moved to I2, also allow destinations that are a paradoxical
5908 subreg (instead of a normal reg).
5910 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5913 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5916 2018-02-16 Richard Biener <rguenther@suse.de>
5918 PR tree-optimization/84037
5919 PR tree-optimization/84016
5921 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5922 Adjust vec_construct for the fact we need additional higher latency
5923 128bit inserts for AVX256 and AVX512 vector builds.
5924 (ix86_add_stmt_cost): Scale vector construction cost for
5927 2018-02-16 Richard Biener <rguenther@suse.de>
5929 PR tree-optimization/84417
5930 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5931 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5932 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5934 2018-02-16 Martin Liska <mliska@suse.cz>
5937 * internal-fn.def (ASAN_CHECK): Set proper flags.
5938 (ASAN_MARK): Likewise.
5940 2018-02-16 Julia Koval <julia.koval@intel.com>
5942 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5943 from PTA_CANNONLAKE.
5945 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5948 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5949 Use ++iter rather than iter++ for std::list iterators.
5950 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5951 defer deleting them until all nodes in the forest are processed. Do
5952 free even leaf nodes. Change to_process into auto_vec.
5955 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5956 * vec.h (vec_default_construct): Use memset instead of placement new
5957 if BROKEN_VALUE_INITIALIZATION is defined.
5958 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5959 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5962 PR rtl-optimization/83723
5963 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5964 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5965 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5967 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5969 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5971 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5973 PR rtl-optimization/81443
5974 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5975 from inner REGs to paradoxical SUBREGs.
5977 2018-02-16 Richard Biener <rguenther@suse.de>
5979 PR tree-optimization/84399
5980 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5981 For operands we can analyze at their definition make sure we can
5982 analyze them at each use as well.
5984 2018-02-16 Richard Biener <rguenther@suse.de>
5986 PR tree-optimization/84190
5987 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5988 volatile accesses if the decl isn't volatile.
5990 2018-02-15 Jason Merrill <jason@redhat.com>
5992 PR c++/84314 - ICE with templates and fastcall attribute.
5993 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5994 TYPE_CANONICAL on an existing type.
5996 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5998 PR tree-optimization/84383
5999 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6000 dstoff nor call operand_equal_p if dstbase is NULL.
6002 PR tree-optimization/84334
6003 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6004 also a CONSTANT_CLASS_P, punt.
6006 2018-02-14 Jim Wilson <jimw@sifive.com>
6008 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6009 first SMALL_OPERAND check. New local min_second_step. Move assert
6010 to where locals are set. Add TARGET_RVC support.
6011 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6013 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6015 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6017 2018-02-14 Martin Sebor <msebor@redhat.com>
6019 PR tree-optimization/83698
6020 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6021 arrays constrain the offset range to their bounds.
6022 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6023 (builtin_access::overlap): Avoid setting the size of overlap if it's
6025 (maybe_diag_overlap): Also consider arrays when deciding what values
6026 of offsets to include in diagnostics.
6028 2018-02-14 Martin Sebor <msebor@redhat.com>
6031 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6032 that correspond to the kind of a declaration.
6034 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6037 * config/pa/pa.md: Load address of PIC label using the linkage table
6038 if the label is nonlocal.
6040 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6042 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6043 warning message if user requests -maltivec=be.
6044 * doc/invoke.texi: Document deprecation of -maltivec=be.
6046 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6049 * config/rs6000/rs6000-c.c: Update definitions for
6050 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6051 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6053 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6056 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6057 add _get_ssp intrinsics. Remove argument from
6058 __builtin_ia32_rdssp[d|q].
6059 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6060 * config/i386/i386-builtin.def: Remove argument from
6061 __builtin_ia32_rdssp[d|q].
6062 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6063 ix86_expand_special_args_builtin for _rdssp[d|q].
6064 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6065 Clear register before usage.
6066 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6067 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6069 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6071 PR tree-optimization/84357
6072 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6073 operand 1 of an ARRAY_REF too.
6075 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6078 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6079 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6081 (set_of_reg): New struct.
6082 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6083 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6084 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6086 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6087 Split into bitclr, bitset, bitinvert patterns if appropriate.
6088 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6089 use rx_fuse_in_memory_bitop.
6090 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6091 to named insn, correct maximum insn length.
6093 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6096 * machmode.def: Define a complex mode for PARTIAL_INT.
6097 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6099 * doc/rtl.texi: Document CSPImode.
6100 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6102 (msp430_hard_regno_nregs_with_padding): Likewise.
6104 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6107 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6109 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6111 PR rtl-optimization/84169
6112 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6113 we generated a parallel as new i3 and we split that to new i2 and i3
6114 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6115 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6116 those to i2, not i1. Partially rewrite this scan code.
6118 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6121 * stor-layout.c (place_field): For variable length fields, adjust
6122 offset_align afterwards not just based on the field's alignment,
6123 but also on the size.
6126 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6127 of exps and logs in the use_exp2 case.
6129 2018-02-13 Jeff Law <law@redhat.com>
6131 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6134 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6137 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6141 * common.opt (gas-loc-support, gas-locview-support): New.
6142 (ginline-points, ginternal-reset-location-views): New.
6143 * doc/invoke.texi: Document them. Use @itemx where intended.
6144 (gvariable-location-views): Adjust.
6145 * target.def (reset_location_view): New.
6146 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6147 (TARGET_RESET_LOCATION_VIEW): New.
6148 * doc/tm.texi: Rebuilt.
6149 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6150 (dwarf2out_default_as_locview_support): New.
6151 (output_asm_line_debug_info): Use option variables.
6152 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6153 (output_loc_list): Likewise.
6154 (add_high_low_attributes): Check option variables.
6155 Don't output entry view attribute in strict mode.
6156 (gen_inlined_subroutine_die): Check option variables.
6157 (dwarf2out_inline_entry): Likewise.
6158 (init_sections_and_labels): Likewise.
6159 (dwarf2out_early_finish): Likewise.
6160 (maybe_reset_location_view): New, from...
6161 (dwarf2out_var_location): ... here. Call it.
6162 * debug.h (dwarf2out_default_as_loc_support): Declare.
6163 (dwarf2out_default_as_locview_support): Declare.
6164 * hooks.c (hook_int_rtx_insn_0): New.
6165 * hooks.h (hook_int_rtx_insn_0): Declare.
6166 * toplev.c (process_options): Take -gas-loc-support and
6167 -gas-locview-support from dwarf2out. Enable
6168 -gvariable-location-views by default only with locview
6169 assembler support. Enable -ginternal-reset-location-views by
6170 default only if the target defines the corresponding hook.
6171 Enable -ginline-points by default if location views are
6172 enabled; force it disabled if statement frontiers are
6174 * tree-inline.c (expand_call_inline): Check option variables.
6175 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6177 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6179 PR tree-optimization/84321
6180 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6181 handling. Also check whether the anti-range contains any values
6182 that satisfy the mask; switch to a VR_RANGE if not.
6184 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6187 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6189 2018-02-13 Martin Jambor <mjambor@suse.cz>
6192 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6193 of call statements, also set location of a load to a temporary.
6195 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6197 * config/rl78/rl78.c (add_vector_labels): New function.
6198 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6199 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6200 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6201 which checks that no arguments are passed.
6202 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6203 * doc/extend.texi: Documentation for the new attribute.
6205 2018-02-13 Andreas Schwab <schwab@suse.de>
6207 * config/riscv/linux.h (CPP_SPEC): Define.
6209 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6212 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6213 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6214 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6215 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6216 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6217 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6218 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6219 temporarily for AES and PCLMUL builtins.
6221 PR tree-optimization/84339
6222 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6223 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6227 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6228 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6229 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6231 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6233 * omp-simd-clone.h: New file.
6234 * omp-simd-clone.c: Include omp-simd-clone.h.
6235 (expand_simd_clones): No longer static.
6236 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6237 cgraph.h and omp-simd-clone.h.
6238 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6239 (vect_recog_widen_shift_pattern): Formatting fix.
6240 (vect_pattern_recog_1): Don't check optab for calls.
6243 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6244 operands[2] into a REG before using gen_lowpart on it.
6246 2018-02-12 Jeff Law <law@redhat.com>
6249 * config/sh/sh.c (find_barrier): Consider a sibling call
6252 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6253 successfully back substituting a reg.
6255 2018-02-12 Richard Biener <rguenther@suse.de>
6257 PR tree-optimization/84037
6258 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6259 parameter, move visited init to caller.
6260 (vect_slp_analyze_operations): Separate cost from validity
6261 check, initialize visited once for all instances.
6262 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6264 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6266 (vect_model_promotion_demotion_cost): Likewise.
6267 (vectorizable_bswap): Guard cost modeling with !slp_node
6268 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6270 (vectorizable_call): Likewise.
6271 (vectorizable_conversion): Likewise.
6272 (vectorizable_assignment): Likewise.
6273 (vectorizable_shift): Likewise.
6274 (vectorizable_operation): Likewise.
6275 (vectorizable_store): Likewise.
6276 (vectorizable_load): Likewise.
6277 (vectorizable_condition): Likewise.
6278 (vectorizable_comparison): Likewise.
6280 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6283 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6284 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6285 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6287 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6290 * params.def (inline-min-speedup): Increase from 8 to 15.
6291 (max-inline-insns-auto): Decrease from 40 to 30.
6292 * ipa-split.c (consider_split): Add some buffer for function to
6293 be considered inlining candidate.
6294 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6297 2018-02-12 Richard Biener <rguenther@suse.de>
6299 PR tree-optimization/84037
6300 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6301 matched stmts if we cannot swap the non-matched ones.
6303 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6305 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6306 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6307 _mm_maskz_scalef_round_ss): New intrinsics.
6308 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6309 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6310 __builtin_ia32_scalefss_round): Remove.
6311 (__builtin_ia32_scalefsd_mask_round,
6312 __builtin_ia32_scalefss_mask_round): New intrinsics.
6313 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6314 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6315 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6316 "<round_constraint>")): Changed to ...
6317 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6318 "<round_scalar_constraint>")): ... this.
6319 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6320 %0, %1, %2<round_op3>}"): Changed to ...
6321 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6322 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6323 %2<round_scalar_mask_op3>}"): ... this.
6324 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6326 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6328 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6329 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6330 (_mm_maskz_sqrt_round_ss): New intrinsics.
6331 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6332 (__builtin_ia32_sqrtsd_mask_round)
6333 (__builtin_ia32_sqrtss_mask_round): New builtins.
6334 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6335 (__builtin_ia32_sqrtss_round): Remove.
6336 (__builtin_ia32_sqrtsd_mask_round)
6337 (__builtin_ia32_sqrtss_mask_round): New builtins.
6338 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6339 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6340 ((match_operand:VF_128 1 "vector_operand"
6341 "xBm,<round_constraint>")): Changed to ...
6342 ((match_operand:VF_128 1 "vector_operand"
6343 "xBm,<round_scalar_constraint>")): ... this.
6344 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6345 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6346 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6347 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6348 %<iptr>1<round_scalar_mask_op3>}): ... this.
6349 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6350 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6352 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6355 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6356 Cast vec_cmpeq result to correct type.
6357 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6358 Cast vec_cmpgt result to correct type.
6360 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6362 * final.c (final_scan_insn_1): Renamed from...
6363 (final_scan_insn): ... this. New wrapper, to recover
6364 seen from the outermost call in recursive ones.
6365 * config/sparc/sparc.c (output_return): Drop seen from call.
6366 (output_sibcall): Likewise.
6367 * config/visium/visium.c (output_branch): Likewise.
6369 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6371 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6374 2018-02-10 Alan Modra <amodra@gmail.com>
6377 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6378 Specify LR as an input.
6380 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6383 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6384 remove_member_access_dummy_vars): New functions.
6385 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6386 lower_omp_1, execute_lower_omp): Use them.
6388 PR rtl-optimization/84308
6389 * shrink-wrap.c (spread_components): Release todo vector.
6391 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6393 PR rtl-optimization/57193
6394 * ira-color.c (struct allocno_color_data): Add member
6395 conflict_allocno_hard_prefs.
6396 (update_conflict_allocno_hard_prefs): New.
6397 (bucket_allocno_compare_func): Add a preference based on
6398 conflict_allocno_hard_prefs.
6399 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6400 (color_allocnos): Remove a dead code. Initiate
6401 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6403 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6406 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6407 constraint from =wa to wa. Avoid a subreg on the output operand,
6408 instead use a pseudo and subreg it in a move.
6409 (p9_xxbrd_<mode>): Changed to ...
6410 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6411 (p9_xxbrd_v2df): New expander.
6412 (p9_xxbrw_<mode>): Changed to ...
6413 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6414 (p9_xxbrw_v4sf): New expander.
6416 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6418 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6420 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6423 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6424 multiply in 32-bit mode.
6425 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6426 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6429 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6431 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6432 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6433 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6434 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6436 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6439 * dwarf2out.c (is_trivial_indirect_ref): New function.
6440 (dwarf2out_late_global_decl): Do not generate a location
6441 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6442 and that are not defined in the current unit.
6444 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6446 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6447 instead of a libcall for UNORDERED.
6449 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6452 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6453 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6455 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6458 * config/s390/s390.c (s390_set_current_function): Invoke
6459 s390_indirect_branch_settings also if fndecl didn't change.
6461 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6463 * config/rs6000/rs6000.md (blockage): Set length to zero.
6465 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6467 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6469 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6472 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6473 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6477 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6478 PARALLEL incoming that failed vt_get_decl_and_offset check.
6481 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6482 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6484 (get_variable_section): For decls in named .bss* sections pass true as
6485 second argument to bss_initializer_p.
6487 2018-02-09 Marek Polacek <polacek@redhat.com>
6488 Jakub Jelinek <jakub@redhat.com>
6491 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6492 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6493 Sync some changes from cxx_fold_indirect_ref.
6495 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6497 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6499 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6500 (BLOCK_INLINE_ENTRY_LABEL): New.
6501 (dwarf2out_var_location): Disregard inline entry markers.
6502 (inline_entry_data): New struct.
6503 (inline_entry_data_hasher): New hashtable type.
6504 (inline_entry_data_hasher::hash): New.
6505 (inline_entry_data_hasher::equal): New.
6506 (inline_entry_data_table): New variable.
6507 (add_high_low_attributes): Add DW_AT_entry_pc and
6508 DW_AT_GNU_entry_view attributes if a pending entry is found
6509 in inline_entry_data_table. Add old entry_pc attribute only
6510 if debug nonbinding markers are disabled.
6511 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6512 markers are enabled.
6513 (block_within_block_p, dwarf2out_inline_entry): New.
6514 (dwarf2out_finish): Check that no entries remained in
6515 inline_entry_data_table.
6516 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6517 (final_scan_insn, notice_source_line): Likewise.
6518 (rest_of_clean_state): Skip inline entry markers.
6519 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6521 * gimple.c (gimple_build_debug_inline_entry): New.
6522 * gimple.h (enum gimple_debug_subcode): Add
6523 GIMPLE_DEBUG_INLINE_ENTRY.
6524 (gimple_build_debug_inline_entry): Declare.
6525 (gimple_debug_inline_entry_p): New.
6526 (gimple_debug_nonbind_marker_p): Adjust.
6527 * insn-notes.def (INLINE_ENTRY): New.
6528 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6529 inline entry marker notes.
6530 (print_insn): Likewise.
6531 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6532 (INSN_DEBUG_MARKER_KIND): Likewise.
6533 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6534 * tree-inline.c (expand_call_inline): Build and insert
6535 debug_inline_entry stmt.
6536 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6537 inline entry blocks early, if nonbind markers are enabled.
6538 (dump_scope_block): Dump fragment info.
6539 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6540 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6541 (gimple_build_debug_inline_entry): New.
6542 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6543 Enable/disable inline entry points too.
6544 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6545 (DEBUG_INSN): Describe inline entry markers.
6547 * common.opt (gvariable-location-views): New.
6548 (gvariable-location-views=incompat5): New.
6549 * config.in: Rebuilt.
6550 * configure: Rebuilt.
6551 * configure.ac: Test assembler for view support.
6552 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6553 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6554 * dwarf2out.c (var_loc_view): New typedef.
6555 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6556 (dwarf2out_locviews_in_attribute): New.
6557 (dwarf2out_locviews_in_loclist): New.
6558 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6559 (enum dw_line_info_opcode): Add LI_adv_address.
6560 (struct dw_line_info_table): Add view.
6561 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6562 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6563 (zero_view_p): New variable.
6564 (ZERO_VIEW_P): New macro.
6565 (output_asm_line_debug_info): New.
6566 (struct var_loc_node): Add view.
6567 (add_AT_view_list, AT_loc_list): New.
6568 (add_var_loc_to_decl): Add view param. Test it against last.
6569 (new_loc_list): Add view params. Record them.
6570 (AT_loc_list_ptr): Handle loc and view lists.
6571 (view_list_to_loc_list_val_node): New.
6572 (print_dw_val): Handle dw_val_class_view_list.
6573 (size_of_die): Likewise.
6574 (value_format): Likewise.
6575 (loc_list_has_views): New.
6576 (gen_llsym): Set vl_symbol too.
6577 (maybe_gen_llsym, skip_loc_list_entry): New.
6578 (dwarf2out_maybe_output_loclist_view_pair): New.
6579 (output_loc_list): Output view list or entries too.
6580 (output_view_list_offset): New.
6581 (output_die): Handle dw_val_class_view_list.
6582 (output_dwarf_version): New.
6583 (output_compilation_unit_header): Use it.
6584 (output_skeleton_debug_sections): Likewise.
6585 (output_rnglists, output_line_info): Likewise.
6586 (output_pubnames, output_aranges): Update version comments.
6587 (output_one_line_info_table): Output view numbers in asm comments.
6588 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6589 Call maybe_gen_llsym.
6590 (loc_list_from_tree_1): Adjust.
6591 (add_AT_location_description): Create view list attribute if
6592 needed, check it's absent otherwise.
6593 (convert_cfa_to_fb_loc_list): Adjust.
6594 (maybe_emit_file): Call output_asm_line_debug_info for test.
6595 (dwarf2out_var_location): Reset views as needed. Precompute
6596 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6597 attribute. Set view.
6598 (new_line_info_table): Reset next view.
6599 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6600 (dwarf2out_source_line): Likewise. Output view resets and labels to
6601 the assembler, or select appropriate line info opcodes.
6602 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6603 (optimize_string_length): Catch it. Adjust.
6604 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6605 dw_val_class_view_list, and remove it if no longer needed.
6606 (hash_loc_list): Hash view numbers.
6607 (loc_list_hasher::equal): Compare them.
6608 (optimize_location_lists): Check whether a view list symbol is
6609 needed, and whether the locview attribute is present, and
6610 whether they match. Remove the locview attribute if no longer
6612 (index_location_lists): Call skip_loc_list_entry for test.
6613 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6614 Use output_dwarf_version.
6615 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6616 (struct dw_val_node): Add val_view_list.
6617 * final.c (SEEN_NEXT_VIEW): New.
6618 (set_next_view_needed): New.
6619 (clear_next_view_needed): New.
6620 (maybe_output_next_view): New.
6621 (final_start_function): Rename to...
6622 (final_start_function_1): ... this. Take pointer to FIRST,
6623 add SEEN parameter. Emit param bindings in the initial view.
6624 (final_start_function): Reintroduce SEEN-less interface.
6625 (final): Rename to...
6626 (final_1): ... this. Take SEEN parameter. Output final pending
6627 next view at the end.
6628 (final): Reintroduce seen-less interface.
6629 (final_scan_insn): Output pending next view before switching
6630 sections or ending a block. Mark the next view as needed when
6631 outputting variable locations. Notify debug backend of section
6632 changes, and of location view changes.
6633 (rest_of_handle_final): Adjust.
6634 * toplev.c (process_options): Autodetect value for debug variable
6635 location views option. Warn on incompat5 without -gdwarf-5.
6636 * doc/invoke.texi (gvariable-location-views): New.
6637 (gvariable-location-views=incompat5): New.
6638 (gno-variable-location-views): New.
6640 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6642 PR tree-optimization/84136
6643 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6644 that the result of find_edge is non-NULL.
6646 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6649 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6650 storing integer register in SImode. Fix cost of 256 and 512
6651 byte aligned SSE register store.
6653 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6655 * config/i386/i386.c (ix86_multiplication_cost): Fix
6656 multiplication cost for TARGET_AVX512DQ.
6658 2018-02-08 Marek Polacek <polacek@redhat.com>
6660 PR tree-optimization/84238
6661 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6664 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6666 PR tree-optimization/84265
6667 * tree-vect-stmts.c (vectorizable_store): Don't treat
6668 VMAT_CONTIGUOUS accesses as grouped.
6669 (vectorizable_load): Likewise.
6671 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6673 PR tree-optimization/81635
6674 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6675 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6676 (test_round_for_mask): New functions.
6677 (wide_int_cc_tests): Call test_round_for_mask.
6678 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6679 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6680 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6681 range returned by get_range_info.
6683 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6686 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6687 * symtab.c: Include builtins.h
6688 (symtab_node::output_to_lto_symbol_table_p): Move here
6689 from lto-streamer-out.c:output_symbol_p.
6690 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6691 (output_symbol_p): Move all logic to symtab.c
6692 (produce_symtab): Update.
6694 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6696 * config/s390/s390-opts.h (enum indirect_branch): Define.
6697 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6698 (s390_indirect_branch_via_thunk)
6699 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6700 (enum s390_indirect_branch_type): Define.
6701 * config/s390/s390.c (struct s390_frame_layout, struct
6702 machine_function): Remove.
6703 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6704 (indirect_branch_table_label_no, indirect_branch_table_name):
6706 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6707 (enum s390_indirect_branch_option): Define.
6708 (s390_return_addr_from_memory): New function.
6709 (s390_handle_string_attribute): New function.
6710 (s390_attribute_table): Add new attribute handler.
6711 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6712 (s390_indirect_branch_via_thunk): New function.
6713 (s390_indirect_branch_via_inline_thunk): New function.
6714 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6715 sibling call optimization for non z10 compiles.
6716 (s390_emit_call): Force indirect branch target to be a single
6717 register. Add r1 clobber for non-z10 compiles.
6718 (s390_emit_epilogue): Emit return jump via return_use expander.
6719 (s390_reorg): Handle JUMP_INSNs as execute targets.
6720 (s390_option_override_internal): Perform validity checks for the
6721 new command line options.
6722 (s390_indirect_branch_attrvalue): New function.
6723 (s390_indirect_branch_settings): New function.
6724 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6725 (s390_output_indirect_thunk_function): New function.
6726 (s390_code_end): Implement target hook.
6727 (s390_case_values_threshold): Implement target hook.
6728 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6730 * config/s390/s390.h (struct s390_frame_layout)
6731 (struct machine_function): Move here from s390.c.
6732 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6733 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6734 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6735 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6736 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6737 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6738 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6739 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6740 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6741 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6742 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6743 (mnemonic attribute): Add values which aren't recognized
6745 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6746 pattern for branch conversion. Fix mnemonic attribute.
6747 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6748 indirect branch via thunk if requested.
6749 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6750 ("*indirect_jump"): Disable for branch conversion using out of
6752 ("indirect_jump_via_thunk<mode>_z10")
6753 ("indirect_jump_via_thunk<mode>")
6754 ("indirect_jump_via_inlinethunk<mode>_z10")
6755 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6756 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6757 ("casesi_jump_via_inlinethunk<mode>_z10")
6758 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6759 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6760 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6761 ("*indirect2_jump"): Disable for branch conversion.
6762 ("casesi_jump"): Turn into expander and expand patterns for branch
6764 ("return_use"): New expander.
6765 ("*return"): Emit return via thunk and rename it to ...
6766 ("*return<mode>"): ... this one.
6767 * config/s390/s390.opt: Add new options and and enum for the
6770 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6772 * lra-constraints.c (match_reload): Unconditionally use
6773 gen_lowpart_SUBREG, rather than selecting between that
6774 and equivalent gen_rtx_SUBREG code.
6776 2018-02-08 Richard Biener <rguenther@suse.de>
6778 PR tree-optimization/84233
6779 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6780 changed flag instead of boguously re-using phi_inserted.
6782 2018-02-08 Martin Jambor <mjambor@suse.cz>
6784 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6785 static local variables.
6787 2018-02-08 Richard Biener <rguenther@suse.de>
6789 PR tree-optimization/84278
6790 * tree-vect-stmts.c (vectorizable_store): When looking for
6791 smaller vector types to perform grouped strided loads/stores
6792 make sure the mode is supported by the target.
6793 (vectorizable_load): Likewise.
6795 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6797 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6798 Increase LDP/STP opportunities by adding adjacent callee-saves.
6800 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6802 PR rtl-optimization/84068
6803 PR rtl-optimization/83459
6804 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6806 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6808 PR tree-optimization/84224
6809 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6810 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6813 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6816 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6817 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6818 count, remove one USE.
6820 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6822 * doc/install.texi (Configuration): Document the
6823 --with-long-double-format={ibm,ieee} PowerPC configuration
6827 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6828 Convert from define_expand to be define_insn_and_split. Rework
6829 float/double/_Float128 conversions to QI/HI/SImode to work with
6830 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6831 conversions to QI/HImode types did a store and then a load to
6832 truncate the value. For conversions to VSX registers, don't split
6833 the insn, instead emit the code directly. Use the code iterator
6834 any_fix to combine signed and unsigned conversions.
6835 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6836 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6837 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6838 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6839 (fix_<mode>di2_hw): Likewise.
6840 (fixuns_<mode>di2_hw): Likewise.
6841 (fix_<mode>si2_hw): Likewise.
6842 (fixuns_<mode>si2_hw): Likewise.
6843 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6844 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6845 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6846 fix<uns>_trunc<SFDF:mode>si2_p8.
6847 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6849 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6850 (fix<uns>_<mode>_mem): Likewise.
6851 (fctiw<u>z_<mode>_mem): Likewise.
6852 (fix<uns>_<mode>_mem): Likewise.
6853 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6854 the register allocator from doing a direct move to the GPRs to do
6855 a store, and instead use the ISA 3.0 store byte/half-word from
6856 vector register instruction. For IEEE 128-bit floating point,
6857 also optimize stores of 32-bit ints.
6858 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6860 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6862 * genextract.c (push_pathstr_operand): New function to support
6864 (walk_rtx): Call push_pathstr_operand.
6865 (print_path): Support [a-zA-Z].
6867 2018-02-07 Richard Biener <rguenther@suse.de>
6869 PR tree-optimization/84037
6870 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6871 (cse_and_gimplify_to_preheader): Declare.
6872 (vect_get_place_in_interleaving_chain): Likewise.
6873 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6875 (_loop_vec_info::~_loop_vec_info): Delete it.
6876 (cse_and_gimplify_to_preheader): New function.
6877 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6878 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6879 (vectorizable_load): Likewise. For grouped stores always base
6880 the IV on the first element.
6881 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6882 condition before gimplifying.
6884 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6886 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6887 *DIV_EXPR and *MOD_EXPR.
6889 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6892 * config/i386/i386.c (ix86_option_override_internal): Mask out
6893 the CF_SET bit when checking -fcf-protection.
6895 2018-02-07 Tom de Vries <tom@codesourcery.com>
6898 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6901 2018-02-07 Richard Biener <rguenther@suse.de>
6903 PR tree-optimization/84204
6904 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6907 PR tree-optimization/84205
6908 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6909 special-case isl_ast_op_zdiv_r.
6911 PR tree-optimization/84223
6912 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6913 Only add conditions from within the region.
6914 (gather_bbs::after_dom_children): Adjust.
6916 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6919 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6920 * config/avr/avr.md: Only post-reload split REG-REG moves if
6921 either register is GENERAL_REG_P.
6923 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6925 PR tree-optimization/84235
6926 * tree-ssa-scopedtables.c
6927 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6928 if the subtraction is performed in floating point type where NaNs are
6929 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6930 build 1. Formatting fix.
6932 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6935 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6936 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6937 and skip it regardless of bb boundaries. Use CALL_P macro,
6938 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6939 unnecessarily, formatting fix.
6941 2018-02-06 Michael Collison <michael.collison@arm.com>
6943 * config/arm/thumb2.md:
6944 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6945 (*thumb_mov_notscc): Ditto.
6947 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6950 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6951 unsigned_fix, not "s".
6953 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6955 * configure.ac (gcc_fn_eh_frame_ro): New function.
6956 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6957 correct .eh_frame permissions.
6958 * configure: Regenerate.
6960 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6962 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6965 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6967 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6968 Display warning message for -mno-speculate-indirect-jumps.
6970 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6972 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6974 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6976 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6978 PR tree-optimization/84225
6979 * tree-eh.c (find_trapping_overflow): Only call
6980 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6982 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6985 * config/i386/i386.c: Reimplement the check of possible options
6986 -mibt/-mshstk conbination. Change error messages.
6987 * doc/invoke.texi: Fix a typo: remove extra '='.
6989 2018-02-06 Marek Polacek <polacek@redhat.com>
6991 PR tree-optimization/84228
6992 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6994 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6997 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6998 emitted arch directives.
6999 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7000 __ARM_FEATURE_COPROC before changing architectures.
7002 2018-02-06 Richard Biener <rguenther@suse.de>
7004 * config/i386/i386.c (print_reg): Fix typo.
7005 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7007 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7009 * configure: Regenerate.
7011 2018-02-05 Martin Sebor <msebor@redhat.com>
7013 PR tree-optimization/83369
7014 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7017 2018-02-05 Martin Liska <mliska@suse.cz>
7019 * doc/invoke.texi: Cherry-pick upstream r323995.
7021 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7023 * ira.c (ira_init_register_move_cost): Adjust comment.
7025 2018-02-05 Martin Liska <mliska@suse.cz>
7027 PR gcov-profile/84137
7028 * doc/gcov.texi: Fix typo in documentation.
7030 2018-02-05 Martin Liska <mliska@suse.cz>
7032 PR gcov-profile/83879
7033 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7034 using dlopen functionality.
7036 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7038 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7039 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7040 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7041 _mm_maskz_range_round_ss): New intrinsics.
7042 (__builtin_ia32_rangesd128_round)
7043 (__builtin_ia32_rangess128_round): Remove.
7044 (__builtin_ia32_rangesd128_mask_round,
7045 __builtin_ia32_rangess128_mask_round): New builtins.
7046 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7047 __builtin_ia32_rangess128_round): Remove.
7048 (__builtin_ia32_rangesd128_mask_round,
7049 __builtin_ia32_rangess128_mask_round): New builtins.
7050 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7051 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7052 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7053 "<round_saeonly_constraint>")): Changed to ...
7054 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7055 "<round_saeonly_scalar_constraint>")): ... this.
7056 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7057 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7058 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7059 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7060 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7062 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7064 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7066 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7067 Remove all values except native, 8540 and 8548.
7069 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7071 * config/i386/i386.c (ix86_output_function_return): Pass
7072 INVALID_REGNUM, instead of -1, as invalid register number to
7073 indirect_thunk_name and output_indirect_thunk.
7075 2018-02-02 Julia Koval <julia.koval@intel.com>
7077 * config.gcc: Add -march=icelake.
7078 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7079 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7080 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7081 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7082 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7083 (processor_target_table): Add icelake.
7084 (ix86_option_override_internal): Handle new PTAs.
7085 (get_builtin_code_for_version): Handle icelake.
7086 (M_INTEL_COREI7_ICELAKE): New.
7087 (fold_builtin_cpu): Handle icelake.
7088 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7089 * doc/invoke.texi: Add -march=icelake.
7091 2018-02-02 Julia Koval <julia.koval@intel.com>
7093 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7094 to wide_int_bitmask.
7095 * wide-int-bitmask.h: New.
7097 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7100 * config/i386/i386.md: Replace Pmode with word_mode in
7101 builtin_setjmp_setup and builtin_longjmp to support x32.
7103 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7107 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7109 (rs6000_supported_cpu_names): New static variable.
7110 (linux_cpu_translation_table): Likewise.
7111 (elf_platform) <cpu>: Define new static variable and use it.
7112 Translate kernel AT_PLATFORM name to canonical name if needed.
7113 Error if platform name is unknown.
7115 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7118 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7120 2018-02-01 Jeff Law <law@redhat.com>
7123 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7124 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7125 the scratch if RELEASE_VIA_POP is false.
7126 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7127 If we have to save a temporary register, decrement SIZE appropriately.
7128 Pass new arguments to release_scratch_register_on_entry.
7129 (ix86_adjust_stack_and_probe): Likewise.
7130 (ix86_emit_probe_stack_range): Pass new arguments to
7131 release_scratch_register_on_entry.
7133 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7135 PR rtl-optimization/84157
7136 * combine.c (change_zero_ext): Use REG_P predicate in
7137 front of HARD_REGISTER_P predicate.
7139 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7141 * config/avr/avr.c (avr_option_override): Move disabling of
7142 -fdelete-null-pointer-checks to...
7143 * common/config/avr/avr-common.c (avr_option_optimization_table):
7146 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7148 PR tree-optimization/81635
7149 * tree-data-ref.c (split_constant_offset_1): For types that
7150 wrap on overflow, try to use range info to prove that wrapping
7153 2018-02-01 Renlin Li <renlin.li@arm.com>
7156 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7158 (aarch64_register_move_cost): Likewise.
7159 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7161 (REG_CLASS_NAMES): Likewise.
7162 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7163 TAILCALL_ADDR_REGS. Remove IP registers.
7164 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7166 2018-02-01 Richard Biener <rguenther@suse.de>
7168 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7169 for specifying RPO order and allow NULL for that.
7170 * domwalk.c (dom_walker::dom_walker): Likewise.
7171 (dom_walker::walk): Handle NULL RPO order.
7172 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7174 (rewrite_update_dom_walker): Likewise.
7175 (mark_def_dom_walker): Likewise.
7177 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7179 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7180 (aarch64_maybe_expand_sve_subreg_move): Declare.
7181 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7182 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7184 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7185 that are semantically a reverse operation.
7186 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7187 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7188 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7190 (aarch64_can_change_mode_class): For big-endian, forbid changes
7191 between two SVE modes if they have different element sizes.
7193 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7195 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7196 the TImode handling for big-endian targets.
7198 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7200 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7201 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7203 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7204 Remove BSWAP handing for big-endian targets and use the form of
7205 LD1RQ appropariate for the mode.
7207 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7209 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7210 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7213 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7216 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7217 check for operands that need to go through aarch64_sve_reload_be.
7219 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7221 PR tree-optimization/81661
7222 PR tree-optimization/84117
7223 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7224 * tree-eh.c: Include gimplify.h.
7225 (find_trapping_overflow, replace_trapping_overflow,
7226 rewrite_to_non_trapping_overflow): New functions.
7227 * tree-vect-loop.c: Include tree-eh.h.
7228 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7229 * tree-data-ref.c: Include tree-eh.h.
7230 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7232 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7234 PR rtl-optimization/84123
7235 * combine.c (change_zero_ext): Check if hard register satisfies
7236 can_change_dest_mode before calling gen_lowpart_SUBREG.
7238 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7241 * ira.c (ira_init_register_move_cost): Remove assert.
7243 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7245 PR rtl-optimization/84071
7246 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7247 * doc/tm.texi: Regenerate.
7249 2018-01-31 Richard Biener <rguenther@suse.de>
7251 PR tree-optimization/84132
7252 * tree-data-ref.c (analyze_miv_subscript): Properly
7253 check whether evolution_function_is_affine_multivariate_p
7254 before calling gcd_of_steps_may_divide_p.
7256 2018-01-31 Julia Koval <julia.koval@intel.com>
7259 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7260 * config/i386/i386.md (rdpid_rex64) New.
7261 (rdpid): Make 32bit only.
7263 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7266 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7267 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7269 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7272 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7274 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7276 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7278 PR rtl-optimization/84071
7279 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7280 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7282 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7284 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7285 (arc_attribute_table): Add 'aux' attribute.
7286 (arc_in_small_data_p): Consider aux like variables.
7287 (arc_is_aux_reg_p): New function.
7288 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7289 (arc_get_aux_arg): New function.
7290 (prepare_move_operands): Handle aux-register access.
7291 (arc_handle_aux_attribute): New function.
7292 * doc/extend.texi (ARC Variable attributes): Add subsection.
7294 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7296 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7297 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7298 (arc_attribute_table): Add 'uncached' attribute.
7299 (arc_print_operand): Print '.di' flag for uncached memory
7301 (arc_in_small_data_p): Do not consider for small data the uncached
7303 (arc_is_uncached_mem_p): New function.
7304 * config/arc/predicates.md (compact_store_memory_operand): Check
7305 for uncached memory accesses.
7306 (nonvol_nonimm_operand): Likewise.
7307 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7309 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7312 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7313 falign-loops=): Add Optimization flag.
7315 2018-01-30 Jeff Law <law@redhat.com>
7318 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7319 INT_REGISTERS_SAVED. Check it prior to calling
7320 get_scratch_register_on_entry.
7321 (ix86_adjust_stack_and_probe): Similarly.
7322 (ix86_emit_probe_stack_range): Similarly.
7323 (ix86_expand_prologue): Corresponding changes.
7325 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7328 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7329 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7331 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7334 * lra-constraints.c (curr_insn_transform): Process AND in the
7337 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7339 PR rtl-optimization/83986
7340 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7341 dependence against last_pending_memory_flush in addition to
7344 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7346 PR tree-optimization/81611
7347 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7350 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7353 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7356 2018-01-30 Richard Biener <rguenther@suse.de>
7357 Jakub Jelinek <jakub@redhat.com>
7359 PR tree-optimization/84111
7360 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7361 inner loops added during recursion, as they don't have up-to-date
7364 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7367 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7368 (can_inline_edge_by_limits_p): ... here.
7369 (can_early_inline_edge_p, check_callers,
7370 update_caller_keys, update_callee_keys, recursive_inlining,
7371 add_new_edges_to_heap, speculation_useful_p,
7372 inline_small_functions,
7373 inline_small_functions, flatten_function,
7374 inline_to_all_callers_1): Update.
7376 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7378 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7381 2018-01-30 Richard Biener <rguenther@suse.de>
7383 PR tree-optimization/83008
7384 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7385 invariant and constant vector uses in stmts when they need
7388 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7391 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7392 * configure: Regenerate.
7394 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7396 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7398 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7399 Use gen_rtx_REG rather than gen_lowpart.
7401 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7403 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7404 rather than 0 when creating partial subregs.
7406 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7408 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7411 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7414 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7415 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7416 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7417 flags. This restores the settings used before the 2017-07-24.
7418 Turning off pre increment/decrement/modify allows IVOPTS to
7419 optimize DF/SF loops where the index is an int.
7421 2018-01-29 Richard Biener <rguenther@suse.de>
7422 Kelvin Nilsen <kelvin@gcc.gnu.org>
7425 * tree-vect-stmts.c (vectorizable_call): Don't call
7426 targetm.vectorize_builtin_md_vectorized_function if callee is
7429 2018-01-22 Carl Love <cel@us.ibm.com>
7431 * doc/extend.tex: Fix typo in second arg in
7432 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7434 2018-01-29 Richard Biener <rguenther@suse.de>
7436 PR tree-optimization/84086
7437 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7438 (flush_ssaname_freelist): When SSA names were released reset
7439 the SCEV hash table.
7441 2018-01-29 Richard Biener <rguenther@suse.de>
7443 PR tree-optimization/84057
7444 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7445 removed paths when removing edges.
7447 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7449 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7450 -mfunction-return=@var{choice}.
7452 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7455 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7457 (layout::print_source_line): Likewise.
7458 (test_get_line_width_without_trailing_whitespace): Add test cases.
7460 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7463 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7466 2018-01-26 Jim Wilson <jimw@sifive.com>
7468 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7470 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7473 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7475 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7476 and CMP + SUB-immediate -> SUBS.
7478 2018-01-26 Martin Sebor <msebor@redhat.com>
7480 PR tree-optimization/83896
7481 * tree-ssa-strlen.c (get_string_len): Rename...
7482 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7483 Avoid assuming length is constant.
7484 (handle_char_store): Use HOST_WIDE_INT for string length.
7486 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7489 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7490 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7492 2018-01-26 Richard Biener <rguenther@suse.de>
7494 PR rtl-optimization/84003
7495 * dse.c (record_store): Only record redundant stores when
7496 the earlier store aliases at least all accesses the later one does.
7498 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7500 PR rtl-optimization/83985
7501 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7502 REG_CFA_RESTORE insns.
7503 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7504 REG_CFA_RESTORE insns here.
7507 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7508 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7510 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7512 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7513 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7514 (arc_init): Likewise.
7515 (arc_override_options): Likewise.
7516 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7518 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7520 * config/arc/arc.h (TARGET_DBNZ): Define.
7521 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7522 properly set the tune attribute.
7523 (dbnz): Use TARGET_DBNZ guard.
7524 * config/arc/arc.opt (mtune): Add core3 option.
7526 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7528 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7529 recognize new pic like addresses.
7530 (arc_delegitimize_address): Clean up.
7532 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7534 * config/arc/arc-arches.def: Option mrf16 valid for all
7536 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7537 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7538 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7539 * config/arc/arc-tables.opt: Regenerate.
7540 * config/arc/arc.c (arc_conditional_register_usage): Handle
7541 reduced register file case.
7542 (arc_file_start): Set must have build attributes.
7543 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7545 * config/arc/arc.opt (mrf16): Add new option.
7546 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7547 * config/arc/genmultilib.awk: Handle new mrf16 option.
7548 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7549 * config/arc/t-multilib: Regenerate.
7550 * doc/invoke.texi (ARC Options): Document mrf16 option.
7552 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7554 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7555 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7556 (arc_attribute_table): Add 'secure_call' attribute.
7557 (arc_print_operand): Print secure call operand.
7558 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7560 (arc_is_secure_call_p): New function. * config/arc/arc.md
7561 (call_i): Add support for sjli instruction.
7562 (call_value_i): Likewise.
7563 * config/arc/constraints.md (Csc): New constraint.
7565 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7566 John Eric Martin <John.Martin@emmicro-us.com>
7568 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7569 * config/arc/arc.c (_arc_jli_section): New struct.
7570 (arc_jli_section): New type.
7571 (rc_jli_sections): New static variable.
7572 (arc_handle_jli_attribute): New function.
7573 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7574 (arc_file_end): New function.
7575 (TARGET_ASM_FILE_END): Define.
7576 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7577 (arc_add_jli_section): New function.
7578 (jli_call_scan): Likewise.
7579 (arc_reorg): Call jli_call_scan.
7580 (arc_output_addsi): Remove 'S' from printing asm operand.
7581 (arc_is_jli_call_p): New function.
7582 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7584 (movhi_insn): Likewise.
7585 (movsi_insn): Likewise.
7586 (movsi_set_cc_insn): Likewise.
7587 (loadqi_update): Likewise.
7588 (load_zeroextendqisi_update): Likewise.
7589 (load_signextendqisi_update): Likewise.
7590 (loadhi_update): Likewise.
7591 (load_zeroextendhisi_update): Likewise.
7592 (load_signextendhisi_update): Likewise.
7593 (loadsi_update): Likewise.
7594 (loadsf_update): Likewise.
7595 (movsicc_insn): Likewise.
7596 (bset_insn): Likewise.
7597 (bxor_insn): Likewise.
7598 (bclr_insn): Likewise.
7599 (bmsk_insn): Likewise.
7600 (bicsi3_insn): Likewise.
7601 (cmpsi_cc_c_insn): Likewise.
7602 (movsi_ne): Likewise.
7603 (movsi_cond_exec): Likewise.
7604 (clrsbsi2): Likewise.
7615 (call_i): Remove 'S' asm letter, add jli instruction.
7616 (call_value_i): Likewise.
7617 * config/arc/arc.op (mjli-always): New option.
7618 * config/arc/constraints.md (Cji): New constraint.
7619 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7621 (subsf3_fpx): Likewise.
7622 (mulsf3_fpx): Likewise.
7623 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7625 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7627 * doc/invoke.texi (ARC): Document mjli-always option.
7629 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7631 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7632 avoid addition with 0 and use incw and decw where possible.
7634 2018-01-26 Richard Biener <rguenther@suse.de>
7636 PR tree-optimization/81082
7637 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7638 association if it requires casting to unsigned.
7639 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7640 from fold_plusminus_mult_expr to catch important cases late when
7641 range info is available.
7643 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7645 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7646 * configure.ac (hidden_linkonce): New test.
7647 * configure: Regenerate.
7648 * config.in: Regenerate.
7650 2018-01-26 Julia Koval <julia.koval@intel.com>
7652 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7653 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7654 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7655 _mm_mask_bitshuffle_epi64_mask): Fix type.
7656 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7657 USI_FTYPE_V4DI_V4DI_USI): Remove.
7658 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7659 __builtin_ia32_vpshufbitqmb256_mask,
7660 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7661 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7662 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7664 2018-01-26 Alan Modra <amodra@gmail.com>
7667 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7668 UNSPEC_VBPERMQ. Sort other unspecs.
7670 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7672 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7674 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7677 * predict.c (drop_profile): Do not push/pop cfun; update also
7679 (handle_missing_profiles): Fix logic looking for zero profiles.
7681 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7684 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7685 on functions with #pragma omp declare simd or functions with simd
7687 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7688 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7689 Remove trailing \n from warning_at calls.
7691 2018-01-25 Tom de Vries <tom@codesourcery.com>
7694 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7695 for neutered workers.
7697 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7700 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7701 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7703 2018-01-24 Jeff Law <law@redhat.com>
7706 * i386.c (get_probe_interval): Move to earlier point.
7707 (ix86_compute_frame_layout): If -fstack-clash-protection and
7708 the frame is larger than the probe interval, then use pushes
7709 to save registers rather than reg->mem moves.
7710 (ix86_expand_prologue): Remove conditional for int_registers_saved
7713 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7716 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7717 min/max for never referenced object.
7719 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7722 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7724 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7725 attributes from DECL_ATTRIBUTES (decl) without affecting
7726 DECL_ATTRIBUTES (current_function_decl).
7727 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7728 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7730 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7732 PR tree-optimization/83979
7733 * fold-const.c (fold_comparison): Use constant_boolean_node
7734 instead of boolean_{true,false}_node.
7736 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7738 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7741 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7743 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7744 Simplify the clause that sets the length attribute.
7745 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7746 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7747 clause that sets the length attribute.
7748 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7750 2018-01-24 Tom de Vries <tom@codesourcery.com>
7753 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7754 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7755 Add strict parameter.
7756 (prevent_branch_around_nothing): Insert dummy insn between branch to
7757 label and label with no ptx insn inbetween.
7758 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7760 2018-01-24 Tom de Vries <tom@codesourcery.com>
7763 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7764 for neutered threads in warp.
7765 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7767 2018-01-24 Richard Biener <rguenther@suse.de>
7769 PR tree-optimization/83176
7770 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7773 2018-01-24 Richard Biener <rguenther@suse.de>
7775 PR tree-optimization/82819
7776 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7777 code generating pluses that are no-ops in the target precision.
7779 2018-01-24 Richard Biener <rguenther@suse.de>
7782 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7784 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7786 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7787 to merge probabilities.
7788 * predict.c (probably_never_executed): Also mark as cold functions
7789 with global 0 profile and guessed local profile.
7790 * profile-count.c (profile_probability::combine_with_count): New
7792 * profile-count.h (profile_probability::operator*,
7793 profile_probability::operator*=, profile_probability::operator/,
7794 profile_probability::operator/=): Reduce precision to adjusted
7795 and set value to guessed on contradictory divisions.
7796 (profile_probability::combine_with_freq): Remove.
7797 (profile_probability::combine_wiht_count): Declare.
7798 (profile_count::force_nonzero):: Set to adjusted.
7799 (profile_count::probability_in):: Set quality to adjusted.
7800 * tree-ssa-tail-merge.c (replace_block_by): Use
7803 2018-01-23 Andrew Waterman <andrew@sifive.com>
7804 Jim Wilson <jimw@sifive.com>
7806 * config/riscv/riscv.c (riscv_stack_boundary): New.
7807 (riscv_option_override): Set riscv_stack_boundary. Handle
7808 riscv_preferred_stack_boundary_arg.
7809 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7810 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7811 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7812 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7813 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7814 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7816 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7819 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7820 of struct ix86_frame.
7821 (ix86_expand_epilogue): Likewise. Add a local variable for
7822 the reg_save_offset field in struct ix86_frame.
7824 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7826 PR tree-optimization/82604
7827 * tree-loop-distribution.c (enum partition_kind): New enum item
7828 PKIND_PARTIAL_MEMSET.
7829 (partition_builtin_p): Support above new enum item.
7830 (generate_code_for_partition): Ditto.
7831 (compute_access_range): Differentiate cases that equality can be
7832 proven at all loops, the innermost loops or no loops.
7833 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7834 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7835 (finalize_partitions, distribute_loop): Don't fuse partition of
7836 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7837 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7840 2018-01-23 Martin Liska <mliska@suse.cz>
7842 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7843 order to ignore the predictor.
7844 (PRED_POLYMORPHIC_CALL): Likewise.
7845 (PRED_RECURSIVE_CALL): Likewise.
7847 2018-01-23 Martin Liska <mliska@suse.cz>
7849 * tree-profile.c (tree_profiling): Print function header to
7850 aware reader which function we are working on.
7851 * value-prof.c (gimple_find_values_to_profile): Do not print
7852 not interesting value histograms.
7854 2018-01-23 Martin Liska <mliska@suse.cz>
7856 * profile-count.h (enum profile_quality): Add
7857 profile_uninitialized as the first value. Do not number values
7858 as they are zero based.
7859 (profile_count::verify): Update sanity check.
7860 (profile_probability::verify): Likewise.
7862 2018-01-23 Nathan Sidwell <nathan@acm.org>
7864 * doc/invoke.texi (ffor-scope): Deprecate.
7866 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7868 PR tree-optimization/83510
7869 * domwalk.c (set_all_edges_as_executable): New function.
7870 (dom_walker::dom_walker): Convert bool param
7871 "skip_unreachable_blocks" to enum reachability. Move setup of
7872 edge flags to set_all_edges_as_executable and only do it when
7873 reachability is REACHABLE_BLOCKS.
7874 * domwalk.h (enum dom_walker::reachability): New enum.
7875 (dom_walker::dom_walker): Convert bool param
7876 "skip_unreachable_blocks" to enum reachability.
7877 (set_all_edges_as_executable): New decl.
7878 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7879 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7881 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7882 but converting true to REACHABLE_BLOCKS.
7883 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7885 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7886 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7887 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7889 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7890 if check_all_array_refs will be called.
7892 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7894 * tree.c (selftest::test_location_wrappers): Add more test
7897 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7899 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7900 (selftest::test_bit_in_range): Likewise.
7902 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7905 * doc/sourcebuild.texi (vect_float): Say that the selector
7906 only describes the situation when -funsafe-math-optimizations is on.
7907 (vect_float_strict): Document.
7909 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7911 PR tree-optimization/83965
7912 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7913 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7914 instead of checking only for a reduction.
7915 (vect_recog_widen_sum_pattern): Likewise.
7917 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7919 * predict.c (probably_never_executed): Only use precise profile info.
7920 (compute_function_frequency): Skip after inlining hack since we now
7921 have quality checking.
7923 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7925 * profile-count.h (profile_probability::very_unlikely,
7926 profile_probability::unlikely, profile_probability::even): Set
7927 precision to guessed.
7929 2018-01-23 Richard Biener <rguenther@suse.de>
7931 PR tree-optimization/83963
7932 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7933 Properly terminate dominator walk when crossing the exit edge not
7934 when visiting its source block.
7936 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7939 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7940 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7942 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7944 PR tree-optimization/83957
7945 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7946 semicolon after for body surrounded by braces.
7948 PR tree-optimization/83081
7949 * profile-count.h (profile_probability::split): New method.
7950 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7951 Use profile_probability::split.
7952 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7953 when splitting a single conditional jump into 2.
7955 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7957 PR tree-optimization/69452
7958 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7961 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7963 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7964 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7965 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7967 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7969 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7970 * config/rl78/rl78.md (movdi): New define_expand.
7971 * config/rl78/rl78.c (rl78_split_movdi): New function.
7973 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7976 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7978 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7979 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7980 128-bit to produce an UNSPEC move to get the double word with the
7981 signbit and then a shift directly to do signbit.
7982 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7983 implementation with a new version that just does either a direct
7984 move or a regular move. Move memory interface to separate insns.
7985 Move insns so they are next to the expander.
7986 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7987 with signbit move. Split big and little endian case.
7988 (signbit<mode>2_dm_mem_le): Likewise.
7989 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7990 (signbit<mode>2_dm2): Likewise.
7992 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7994 * config/rl78/rl78.md (anddi3): New define_expand.
7996 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7998 * config/rl78/rl78.md (umindi3): New define_expand.
8000 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8002 * config/rl78/rl78.md (smindi3): New define_expand.
8004 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8006 * config/rl78/rl78.md (smaxdi3): New define_expand.
8008 2018-01-22 Carl Love <cel@us.ibm.com>
8010 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8011 LVX_V1TI): Add macro expansion.
8012 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8013 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8014 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8015 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8016 Change check to determine if the instruction is a byte reversing
8017 entry. Fix typo in comment.
8018 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8019 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8020 Add def_builtin calls for new builtins.
8021 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8022 Add define_insn expansion.
8024 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8026 * config/rl78/rl78.md (umaxdi3): New define_expand.
8028 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8030 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8031 for non-QImode registers.
8033 2018-01-22 Richard Biener <rguenther@suse.de>
8035 PR tree-optimization/83963
8036 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8037 including the loop exit block.
8038 (scop_detection::merge_sese): Likewise.
8039 (scop_detection::add_scop): Do it here instead.
8041 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8043 * doc/sourcebuild.texi (arm_softfloat): Document.
8045 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8048 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8049 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8050 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8052 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8053 David Edelsohn <dje.gcc@gmail.com>
8056 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8057 Change "crset eq" to "crset 2".
8058 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8059 (*call_indirect_aix<mode>_nospec): Likewise.
8060 (*call_value_indirect_aix<mode>_nospec): Likewise.
8061 (*call_indirect_elfv2<mode>_nospec): Likewise.
8062 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8063 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8064 change assembly output from . to $.
8065 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8066 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8067 (*tablejump<mode>_internal1_nospec): Likewise.
8069 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8072 * config/sh/sh_optimize_sett_clrt.cc:
8073 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8075 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8077 PR tree-optimization/83940
8078 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8079 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8080 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8081 use it to pass back the definition type.
8082 (vect_check_store_rhs): Likewise rhs_dt_out.
8083 (vect_build_gather_load_calls): Add a mask_dt argument and use
8084 it instead of a call to vect_is_simple_use.
8085 (vectorizable_store): Update calls to vect_check_load_store_mask
8086 and vect_check_store_rhs. Use the dt returned by the latter instead
8087 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8088 instead of calls to vect_is_simple_use. Pass the scalar rather
8089 than the vector operand to vect_is_simple_use when handling
8090 second and subsequent copies of an rhs value.
8091 (vectorizable_load): Update calls to vect_check_load_store_mask
8092 and vect_build_gather_load_calls. Use the cached mask_dt and
8093 gs_info.offset_dt instead of calls to vect_is_simple_use.
8095 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8098 * tree-emutls.c: Include gimplify.h.
8099 (lower_emutls_2): New function.
8100 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8101 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8102 it before further processing.
8105 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8106 UINTVAL (trueop1) instead of INTVAL (op1).
8108 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8112 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8113 INCOMING_FRAME_SP_OFFSET if not defined.
8114 (scan_trace): Add ENTRY argument. If true and
8115 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8116 emit a note to adjust the CFA offset.
8117 (create_cfi_notes): Adjust scan_trace callers.
8118 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8119 INCOMING_FRAME_SP_OFFSET in the CIE.
8120 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8121 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8123 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8124 * doc/tm.texi: Regenerated.
8126 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8128 PR rtl-optimization/83147
8129 * lra-constraints.c (remove_inheritance_pseudos): Use
8130 lra_substitute_pseudo_within_insn.
8132 2018-01-19 Tom de Vries <tom@codesourcery.com>
8133 Cesar Philippidis <cesar@codesourcery.com>
8136 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8138 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8141 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8142 spaces for function labels.
8144 2018-01-19 Martin Liska <mliska@suse.cz>
8146 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8147 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8148 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8149 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8150 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8151 (PRED_CONST_RETURN): Change from 69 to 65.
8152 (PRED_NULL_RETURN): Change from 91 to 71.
8153 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8154 (PRED_LOOP_GUARD): Change from 66 to 73.
8156 2018-01-19 Martin Liska <mliska@suse.cz>
8158 * predict.c (predict_insn_def): Add new assert.
8159 (struct branch_predictor): Change type to signed integer.
8160 (test_prediction_value_range): Amend test to cover
8162 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8163 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8164 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8165 (PRED_LOOP_IV_COMPARE): Likewise.
8166 * predict.h (PROB_UNINITIALIZED): Define new constant.
8168 2018-01-19 Martin Liska <mliska@suse.cz>
8170 * predict.c (dump_prediction): Add new format for
8171 analyze_brprob.py script which is enabled with -details
8173 * profile-count.h (precise_p): New function.
8175 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8177 PR tree-optimization/83922
8178 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8179 there are no statements that need masking.
8180 (vect_active_double_reduction_p): New function.
8181 (vect_analyze_loop_operations): Use it when handling phis that
8182 are not in the loop header.
8184 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8186 PR tree-optimization/83914
8187 * tree-vect-loop.c (vectorizable_induction): Don't convert
8188 init_expr or apply the peeling adjustment for inductions
8189 that are nested within the vectorized loop.
8191 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8193 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8196 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8200 * function.h (gimplify_parameters): Add gimple_seq * argument.
8201 * function.c: Include gimple.h and options.h.
8202 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8203 for the added local temporaries if needed.
8204 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8205 if there are any parameter cleanups, wrap whole body into a
8206 try/finally with the cleanups.
8208 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8211 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8212 Use GET_MODE_CLASS for scalar floating point.
8214 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8218 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8219 Fix call of call_cgraph_insertion_hooks.
8221 2018-01-18 Martin Sebor <msebor@redhat.com>
8223 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8225 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8228 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8231 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8234 * common.opt: (-ffile-prefix-map): New option.
8235 * opts.c (common_handle_option): Defer it.
8236 * opts-global.c (handle_common_deferred_options): Handle it.
8237 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8238 * file-prefix-map.h: New file.
8239 (remap_debug_filename, add_debug_prefix_map): ...here.
8240 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8241 * final.c (debug_prefix_map, add_debug_prefix_map
8242 remap_debug_filename): Move to...
8243 * file-prefix-map.c: New file.
8244 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8245 generalize, get rid of alloca(), use strrchr() instead of strchr().
8246 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8247 Implement in terms of add_prefix_map().
8248 (remap_macro_filename, remap_debug_filename): Implement in term of
8250 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8251 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8252 * dbxout.c: Include file-prefix-map.h.
8253 * varasm.c: Likewise.
8254 * vmsdbgout.c: Likewise.
8255 * xcoffout.c: Likewise.
8256 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8257 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8258 * doc/invoke.texi (-ffile-prefix-map): Document.
8259 (-fdebug-prefix-map): Update description.
8261 2018-01-18 Martin Liska <mliska@suse.cz>
8263 * config/i386/i386.c (indirect_thunk_name): Document that also
8265 (output_indirect_thunk): Document why both instructions
8266 (pause and lfence) are generated.
8268 2018-01-18 Richard Biener <rguenther@suse.de>
8270 PR tree-optimization/83887
8271 * graphite-scop-detection.c
8272 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8273 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8274 (scop_detection::merge_sese): Re-implement with a flood-fill
8275 algorithm that properly finds a SESE region if it exists.
8277 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8280 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8281 pointer_diff optimizations use view_convert instead of convert.
8283 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8285 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8286 Generate different code for -mno-speculate-indirect-jumps.
8287 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8288 (*call_indirect_aix<mode>): Disable for
8289 -mno-speculate-indirect-jumps.
8290 (*call_indirect_aix<mode>_nospec): New define_insn.
8291 (*call_value_indirect_aix<mode>): Disable for
8292 -mno-speculate-indirect-jumps.
8293 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8294 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8295 -mno-speculate-indirect-jumps.
8296 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8298 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8300 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8301 long double type, set the flags for noting the default long double
8302 type, even if we don't pass or return a long double type.
8304 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8307 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8310 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8312 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8313 support for merge[hl].
8314 (fold_mergehl_helper): New helper function.
8315 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8316 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8317 (altivec_vmrglw_direct): Add xxmrglw insn.
8319 2018-01-17 Andrew Waterman <andrew@sifive.com>
8321 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8322 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8324 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8327 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8328 call the lto_location_cache before reading the
8329 DECL_SOURCE_LOCATION of the types.
8331 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8332 Richard Sandiford <richard.sandiford@linaro.org>
8334 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8335 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8336 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8337 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8338 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8340 * config/aarch64/constraints.md (aarch64_movti_operand):
8342 * config/aarch64/predicates.md (Uti): Add new constraint.
8344 2018-01-17 Carl Love <cel@us.ibm.com>
8346 * config/rs6000/vsx.md (define_expand xl_len_r,
8347 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8348 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8350 (define_expand, define_insn): Move the shift left from the
8351 define_insn to the define_expand for lxvl and stxvl instructions.
8352 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8353 and XL_LEN_R definitions to PURE.
8355 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8357 * config/i386/i386.c (indirect_thunk_name): Declare regno
8358 as unsigned int. Compare regno with INVALID_REGNUM.
8359 (output_indirect_thunk): Ditto.
8360 (output_indirect_thunk_function): Ditto.
8361 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8362 in the call to output_indirect_thunk_function.
8364 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8367 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8368 rather than the size of inner_type to determine the stack slot size
8369 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8371 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8374 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8377 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8379 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8380 endian Linux systems to optionally enable multilibs for selecting
8381 the long double type if the user configured an explicit type.
8382 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8383 have no long double multilibs if not defined.
8384 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8385 warn if the user used -mabi={ieee,ibm}longdouble and we built
8386 multilibs for long double.
8387 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8388 appropriate multilib option.
8389 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8391 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8392 for building long double multilibs.
8393 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8395 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8397 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8400 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8402 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8405 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8408 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8411 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8413 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8414 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8416 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8418 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8419 different rtl trees depending on TARGET_64BIT.
8420 (rs6000_gen_lvx): Likewise.
8422 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8424 * config/visium/visium.md (nop): Tweak comment.
8425 (hazard_nop): Likewise.
8427 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8429 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8430 -mspeculate-indirect-jumps.
8431 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8432 for -mno-speculate-indirect-jumps.
8433 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8434 (*call_value_indirect_elfv2<mode>): Disable for
8435 -mno-speculate-indirect-jumps.
8436 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8437 (indirect_jump): Emit different RTL for
8438 -mno-speculate-indirect-jumps.
8439 (*indirect_jump<mode>): Disable for
8440 -mno-speculate-indirect-jumps.
8441 (*indirect_jump<mode>_nospec): New define_insn.
8442 (tablejump): Emit different RTL for
8443 -mno-speculate-indirect-jumps.
8444 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8445 (tablejumpsi_nospec): New define_expand.
8446 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8447 (tablejumpdi_nospec): New define_expand.
8448 (*tablejump<mode>_internal1): Disable for
8449 -mno-speculate-indirect-jumps.
8450 (*tablejump<mode>_internal1_nospec): New define_insn.
8451 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8454 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8456 * caller-save.c (insert_save): Drop unnecessary parameter. All
8459 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8460 Richard Biener <rguenth@suse.de>
8463 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8464 return early, inline manually is_gimple_sizepos. Make sure if we
8465 call gimplify_expr we don't end up with a gimple constant.
8466 * tree.c (variably_modified_type_p): Don't return true for
8467 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8468 * gimplify.h (is_gimple_sizepos): Remove.
8470 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8472 PR tree-optimization/83857
8473 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8474 vectorizable_live_operation for pure SLP statements.
8475 (vectorizable_live_operation): Handle PHIs.
8477 2018-01-16 Richard Biener <rguenther@suse.de>
8479 PR tree-optimization/83867
8480 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8481 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8483 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8486 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8487 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8488 If off is not INTEGER_CST, issue a may not be aligned warning
8489 rather than isn't aligned. Use isn%'t rather than isn't.
8490 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8492 <case MULT_EXPR>: Improve the case when bottom and one of the
8493 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8494 operand, in that case check if the other operand is multiple of
8495 bottom divided by the INTEGER_CST operand.
8497 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8500 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8501 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8502 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8503 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8504 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8505 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8506 (pa_function_arg_size): New function.
8508 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8510 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8511 in a separate statement.
8513 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8515 PR tree-optimization/83847
8516 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8517 group gathers and scatters.
8519 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8521 PR rtl-optimization/86620
8522 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8524 PR rtl-optimization/83213
8525 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8526 to last if both are JUMP_INSNs.
8528 PR tree-optimization/83843
8529 * gimple-ssa-store-merging.c
8530 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8531 store_immediate_info for bswap/nop orig_stores.
8533 2018-01-15 Andrew Waterman <andrew@sifive.com>
8535 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8537 <UDIV>: Increase cost if !TARGET_DIV.
8539 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8541 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8542 (define_attr "cr_logical_3op"): New.
8543 (cceq_ior_compare): Adjust.
8544 (cceq_ior_compare_complement): Adjust.
8545 (*cceq_rev_compare): Adjust.
8546 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8547 (is_cracked_insn): Adjust.
8548 (insn_must_be_first_in_group): Adjust.
8549 * config/rs6000/40x.md: Adjust.
8550 * config/rs6000/440.md: Adjust.
8551 * config/rs6000/476.md: Adjust.
8552 * config/rs6000/601.md: Adjust.
8553 * config/rs6000/603.md: Adjust.
8554 * config/rs6000/6xx.md: Adjust.
8555 * config/rs6000/7450.md: Adjust.
8556 * config/rs6000/7xx.md: Adjust.
8557 * config/rs6000/8540.md: Adjust.
8558 * config/rs6000/cell.md: Adjust.
8559 * config/rs6000/e300c2c3.md: Adjust.
8560 * config/rs6000/e500mc.md: Adjust.
8561 * config/rs6000/e500mc64.md: Adjust.
8562 * config/rs6000/e5500.md: Adjust.
8563 * config/rs6000/e6500.md: Adjust.
8564 * config/rs6000/mpc.md: Adjust.
8565 * config/rs6000/power4.md: Adjust.
8566 * config/rs6000/power5.md: Adjust.
8567 * config/rs6000/power6.md: Adjust.
8568 * config/rs6000/power7.md: Adjust.
8569 * config/rs6000/power8.md: Adjust.
8570 * config/rs6000/power9.md: Adjust.
8571 * config/rs6000/rs64.md: Adjust.
8572 * config/rs6000/titan.md: Adjust.
8574 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8576 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8577 ix86_indirect_branch_register logic.
8579 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8581 * config/i386/constraints.md (Bs): Update
8582 ix86_indirect_branch_register check. Don't check
8583 ix86_indirect_branch_register with GOT_memory_operand.
8585 * config/i386/predicates.md (GOT_memory_operand): Don't check
8586 ix86_indirect_branch_register here.
8587 (GOT32_symbol_operand): Likewise.
8589 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8591 * config/i386/predicates.md (constant_call_address_operand):
8592 Rewrite ix86_indirect_branch_register logic.
8593 (sibcall_insn_operand): Likewise.
8595 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8597 * config/i386/constraints.md (Bs): Replace
8598 ix86_indirect_branch_thunk_register with
8599 ix86_indirect_branch_register.
8601 * config/i386/i386.md (indirect_jump): Likewise.
8602 (tablejump): Likewise.
8603 (*sibcall_memory): Likewise.
8604 (*sibcall_value_memory): Likewise.
8605 Peepholes of indirect call and jump via memory: Likewise.
8606 * config/i386/i386.opt: Likewise.
8607 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8608 (GOT_memory_operand): Likewise.
8609 (call_insn_operand): Likewise.
8610 (sibcall_insn_operand): Likewise.
8611 (GOT32_symbol_operand): Likewise.
8613 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8616 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8617 type rather than type addr's type points to.
8618 (expand_omp_atomic_mutex): Likewise.
8619 (expand_omp_atomic): Likewise.
8621 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8624 * config/i386/i386.c (output_indirect_thunk_function): Use
8625 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8626 for __x86_return_thunk.
8628 2018-01-15 Richard Biener <rguenther@suse.de>
8631 * expmed.c (extract_bit_field_1): Fix typo.
8633 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8636 * config/arm/iterators.md (VF): New mode iterator.
8637 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8638 Remove integer-related logic from pattern.
8639 (neon_vabd<mode>_3): Likewise.
8641 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8644 * common.opt (fstrict-overflow): No longer an alias.
8645 (fwrapv-pointer): New option.
8646 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8647 also for pointer types based on flag_wrapv_pointer.
8648 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8649 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8650 opts->x_flag_wrapv got set.
8651 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8652 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8653 POINTER_TYPE_OVERFLOW_UNDEFINED.
8654 * match.pd: Likewise in address comparison pattern.
8655 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8657 2018-01-15 Richard Biener <rguenther@suse.de>
8660 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8661 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8662 Reset type names to their identifier if their TYPE_DECL doesn't
8663 have linkage (and thus is used for ODR and devirt).
8664 (save_debug_info_for_decl): Remove.
8665 (save_debug_info_for_type): Likewise.
8666 (add_tree_to_fld_list): Adjust.
8667 * tree-pretty-print.c (dump_generic_node): Make dumping of
8668 type names more robust.
8670 2018-01-15 Richard Biener <rguenther@suse.de>
8672 * BASE-VER: Bump to 8.0.1.
8674 2018-01-14 Martin Sebor <msebor@redhat.com>
8677 * builtins.c (check_access): Avoid warning when the no-warning bit
8680 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8682 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8683 * ira-color (allocno_hard_regs_compare): Likewise.
8685 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8688 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8689 Use .pushsection/.popsection.
8691 2018-01-14 Martin Sebor <msebor@redhat.com>
8694 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8696 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8698 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8699 entry from extra_headers.
8700 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8701 extra_headers, make the list bitwise identical to the i?86-*-* one.
8703 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8705 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8706 -mcmodel=large with -mindirect-branch=thunk,
8707 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8708 -mfunction-return=thunk-extern.
8709 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8710 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8711 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8713 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8715 * config/i386/i386.c (print_reg): Print the name of the full
8716 integer register without '%'.
8717 (ix86_print_operand): Handle 'V'.
8718 * doc/extend.texi: Document 'V' modifier.
8720 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8722 * config/i386/constraints.md (Bs): Disallow memory operand for
8723 -mindirect-branch-register.
8725 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8726 (GOT_memory_operand): Likewise.
8727 (call_insn_operand): Likewise.
8728 (sibcall_insn_operand): Likewise.
8729 (GOT32_symbol_operand): Likewise.
8730 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8731 for -mindirect-branch-register.
8732 (tablejump): Likewise.
8733 (*sibcall_memory): Likewise.
8734 (*sibcall_value_memory): Likewise.
8735 Disallow peepholes of indirect call and jump via memory for
8736 -mindirect-branch-register.
8737 (*call_pop): Replace m with Bw.
8738 (*call_value_pop): Likewise.
8739 (*sibcall_pop_memory): Replace m with Bs.
8740 * config/i386/i386.opt (mindirect-branch-register): New option.
8741 * doc/invoke.texi: Document -mindirect-branch-register option.
8743 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8745 * config/i386/i386-protos.h (ix86_output_function_return): New.
8746 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8747 set function_return_type.
8748 (indirect_thunk_name): Add ret_p to indicate thunk for function
8750 (output_indirect_thunk_function): Pass false to
8751 indirect_thunk_name.
8752 (ix86_output_indirect_branch_via_reg): Likewise.
8753 (ix86_output_indirect_branch_via_push): Likewise.
8754 (output_indirect_thunk_function): Create alias for function
8755 return thunk if regno < 0.
8756 (ix86_output_function_return): New function.
8757 (ix86_handle_fndecl_attribute): Handle function_return.
8758 (ix86_attribute_table): Add function_return.
8759 * config/i386/i386.h (machine_function): Add
8760 function_return_type.
8761 * config/i386/i386.md (simple_return_internal): Use
8762 ix86_output_function_return.
8763 (simple_return_internal_long): Likewise.
8764 * config/i386/i386.opt (mfunction-return=): New option.
8765 (indirect_branch): Mention -mfunction-return=.
8766 * doc/extend.texi: Document function_return function attribute.
8767 * doc/invoke.texi: Document -mfunction-return= option.
8769 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8771 * config/i386/i386-opts.h (indirect_branch): New.
8772 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8773 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8774 with local indirect jump when converting indirect call and jump.
8775 (ix86_set_indirect_branch_type): New.
8776 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8777 (indirectlabelno): New.
8778 (indirect_thunk_needed): Likewise.
8779 (indirect_thunk_bnd_needed): Likewise.
8780 (indirect_thunks_used): Likewise.
8781 (indirect_thunks_bnd_used): Likewise.
8782 (INDIRECT_LABEL): Likewise.
8783 (indirect_thunk_name): Likewise.
8784 (output_indirect_thunk): Likewise.
8785 (output_indirect_thunk_function): Likewise.
8786 (ix86_output_indirect_branch_via_reg): Likewise.
8787 (ix86_output_indirect_branch_via_push): Likewise.
8788 (ix86_output_indirect_branch): Likewise.
8789 (ix86_output_indirect_jmp): Likewise.
8790 (ix86_code_end): Call output_indirect_thunk_function if needed.
8791 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8793 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8794 (ix86_attribute_table): Add indirect_branch.
8795 * config/i386/i386.h (machine_function): Add indirect_branch_type
8796 and has_local_indirect_jump.
8797 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8799 (tablejump): Likewise.
8800 (*indirect_jump): Use ix86_output_indirect_jmp.
8801 (*tablejump_1): Likewise.
8802 (simple_return_indirect_internal): Likewise.
8803 * config/i386/i386.opt (mindirect-branch=): New option.
8804 (indirect_branch): New.
8807 (thunk-inline): Likewise.
8808 (thunk-extern): Likewise.
8809 * doc/extend.texi: Document indirect_branch function attribute.
8810 * doc/invoke.texi: Document -mindirect-branch= option.
8812 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8815 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8817 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8819 * ipa-inline.c (want_inline_small_function_p): Return false if
8820 inlining has already failed with CIF_FINAL_ERROR.
8821 (update_caller_keys): Call want_inline_small_function_p before
8823 (update_callee_keys): Likewise.
8825 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8827 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8829 (rs6000_quadword_masked_address_p): Likewise.
8830 (quad_aligned_load_p): Likewise.
8831 (quad_aligned_store_p): Likewise.
8832 (const_load_sequence_p): Add comment to describe the outer-most loop.
8833 (mimic_memory_attributes_and_flags): New function.
8834 (rs6000_gen_stvx): Likewise.
8835 (replace_swapped_aligned_store): Likewise.
8836 (rs6000_gen_lvx): Likewise.
8837 (replace_swapped_aligned_load): Likewise.
8838 (replace_swapped_load_constant): Capitalize argument name in
8839 comment describing this function.
8840 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8841 and stores that access quad-word aligned addresses and replace
8842 with stvx or lvx instructions when appropriate.
8843 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8844 New function prototype.
8845 (rs6000_quadword_masked_address_p): Likewise.
8846 (rs6000_gen_lvx): Likewise.
8847 (rs6000_gen_stvx): Likewise.
8848 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8849 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8850 when memory address is aligned.
8851 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8852 this split to select lvx instruction when memory address is aligned.
8853 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8854 instruction when memory address is aligned.
8855 (*vsx_le_perm_load_v16qi): Likewise.
8856 (four unnamed splitters): Modify to select the stvx instruction
8857 when memory is aligned.
8859 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8861 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8862 which appears in the queue multiple times.
8864 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8865 Alan Hayward <alan.hayward@arm.com>
8866 David Sherwood <david.sherwood@arm.com>
8868 * tree-vectorizer.h (vec_lower_bound): New structure.
8869 (_loop_vec_info): Add check_nonzero and lower_bounds.
8870 (LOOP_VINFO_CHECK_NONZERO): New macro.
8871 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8872 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8873 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8874 fields. Make seg_len the distance travelled, not including the
8876 (dr_direction_indicator): Declare.
8877 (dr_zero_step_indicator): Likewise.
8878 (dr_known_forward_stride_p): Likewise.
8879 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8881 (runtime_alias_check_p): Allow runtime alias checks with
8883 (operator ==): Compare access_size and align.
8884 (prune_runtime_alias_test_list): Rework for new distinction between
8885 the access_size and seg_len.
8886 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8888 (get_segment_min_max): New function.
8889 (create_intersect_range_checks): Use it.
8890 (dr_step_indicator): New function.
8891 (dr_direction_indicator): Likewise.
8892 (dr_zero_step_indicator): Likewise.
8893 (dr_known_forward_stride_p): Likewise.
8894 * tree-loop-distribution.c (data_ref_segment_size): Return
8895 DR_STEP * (niters - 1).
8896 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8898 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8899 (vect_preserves_scalar_order_p): New function, split out from...
8900 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8901 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8902 (vect_vfa_access_size): New function.
8903 (vect_vfa_align): Likewise.
8904 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8905 (dump_lower_bound): New function.
8906 (vect_check_lower_bound): Likewise.
8907 (vect_small_gap_p): Likewise.
8908 (vectorizable_with_step_bound_p): Likewise.
8909 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8910 depencies if the vectorization factor is 1. Convert the checks
8911 for nonzero steps into checks on the bounds of DR_STEP. Try using
8912 a bunds check for variable steps if the minimum required step is
8913 relatively small. Update calls to the dr_with_seg_len
8914 constructor and to vect_compile_time_alias.
8915 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8917 (vect_loop_versioning): Call it.
8918 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8920 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8922 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8923 Alan Hayward <alan.hayward@arm.com>
8924 David Sherwood <david.sherwood@arm.com>
8926 * doc/sourcebuild.texi (vect_scatter_store): Document.
8927 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8929 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8931 * genopinit.c (main): Add supports_vec_scatter_store and
8932 supports_vec_scatter_store_cached to target_optabs.
8933 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8934 IFN_MASK_SCATTER_STORE.
8935 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8937 * internal-fn.h (internal_store_fn_p): Declare.
8938 (internal_fn_stored_value_index): Likewise.
8939 * internal-fn.c (scatter_store_direct): New macro.
8940 (expand_scatter_store_optab_fn): New function.
8941 (direct_scatter_store_optab_supported_p): New macro.
8942 (internal_store_fn_p): New function.
8943 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8944 IFN_MASK_SCATTER_STORE.
8945 (internal_fn_mask_index): Likewise.
8946 (internal_fn_stored_value_index): New function.
8947 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8949 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8950 * optabs-query.c (supports_vec_scatter_store_p): New function.
8951 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8952 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8953 true for scatter stores.
8954 (vect_gather_scatter_fn_p): Handle scatter stores too.
8955 (vect_check_gather_scatter): Consider using scatter stores if
8956 supports_vec_scatter_store_p.
8957 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8959 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8960 internal_fn_stored_value_index.
8961 (check_load_store_masking): Handle scatter stores too.
8962 (vect_get_store_rhs): Make public.
8963 (vectorizable_call): Use internal_store_fn_p.
8964 (vectorizable_store): Handle scatter store internal functions.
8965 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8966 when deciding whether the end of the group has been reached.
8967 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8968 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8969 (mask_scatter_store<mode>): New insns.
8971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8972 Alan Hayward <alan.hayward@arm.com>
8973 David Sherwood <david.sherwood@arm.com>
8975 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8976 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8977 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8979 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8980 Use vect_truncate_gather_scatter_offset if we can't treat the
8981 operation as a normal gather load or scatter store.
8982 (get_group_load_store_type): Take the gather_scatter_info
8983 as argument. Try using a gather load or scatter store for
8984 single-element groups.
8985 (get_load_store_type): Update calls to get_group_load_store_type
8986 and vect_use_strided_gather_scatters_p.
8988 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8989 Alan Hayward <alan.hayward@arm.com>
8990 David Sherwood <david.sherwood@arm.com>
8992 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8993 optional tree argument.
8994 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8996 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8997 but continue to use the current value as a fallback.
8998 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8999 to compare the updates.
9000 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9001 (get_load_store_type): Use it when handling a strided access.
9002 (vect_get_strided_load_store_ops): New function.
9003 (vect_get_data_ptr_increment): Likewise.
9004 (vectorizable_load): Handle strided gather loads. Always pass
9005 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9008 Alan Hayward <alan.hayward@arm.com>
9009 David Sherwood <david.sherwood@arm.com>
9011 * doc/md.texi (gather_load@var{m}): Document.
9012 (mask_gather_load@var{m}): Likewise.
9013 * genopinit.c (main): Add supports_vec_gather_load and
9014 supports_vec_gather_load_cached to target_optabs.
9015 * optabs-tree.c (init_tree_optimization_optabs): Use
9016 ggc_cleared_alloc to allocate target_optabs.
9017 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9018 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9020 * internal-fn.h (internal_load_fn_p): Declare.
9021 (internal_gather_scatter_fn_p): Likewise.
9022 (internal_fn_mask_index): Likewise.
9023 (internal_gather_scatter_fn_supported_p): Likewise.
9024 * internal-fn.c (gather_load_direct): New macro.
9025 (expand_gather_load_optab_fn): New function.
9026 (direct_gather_load_optab_supported_p): New macro.
9027 (direct_internal_fn_optab): New function.
9028 (internal_load_fn_p): Likewise.
9029 (internal_gather_scatter_fn_p): Likewise.
9030 (internal_fn_mask_index): Likewise.
9031 (internal_gather_scatter_fn_supported_p): Likewise.
9032 * optabs-query.c (supports_at_least_one_mode_p): New function.
9033 (supports_vec_gather_load_p): Likewise.
9034 * optabs-query.h (supports_vec_gather_load_p): Declare.
9035 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9036 and memory_type field.
9037 (NUM_PATTERNS): Bump to 15.
9038 * tree-vect-data-refs.c: Include internal-fn.h.
9039 (vect_gather_scatter_fn_p): New function.
9040 (vect_describe_gather_scatter_call): Likewise.
9041 (vect_check_gather_scatter): Try using internal functions for
9042 gather loads. Recognize existing calls to a gather load function.
9043 (vect_analyze_data_refs): Consider using gather loads if
9044 supports_vec_gather_load_p.
9045 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9046 (vect_get_gather_scatter_offset_type): Likewise.
9047 (vect_convert_mask_for_vectype): Likewise.
9048 (vect_add_conversion_to_patterm): Likewise.
9049 (vect_try_gather_scatter_pattern): Likewise.
9050 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9051 (vect_vect_recog_func_ptrs): Add it.
9052 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9053 internal_fn_mask_index and internal_gather_scatter_fn_p.
9054 (check_load_store_masking): Take the gather_scatter_info as an
9055 argument and handle gather loads.
9056 (vect_get_gather_scatter_ops): New function.
9057 (vectorizable_call): Check internal_load_fn_p.
9058 (vectorizable_load): Likewise. Handle gather load internal
9060 (vectorizable_store): Update call to check_load_store_masking.
9061 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9062 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9063 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9064 (aarch64_gather_scale_operand_d): New predicates.
9065 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9066 (mask_gather_load<mode>): New insns.
9068 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9069 Alan Hayward <alan.hayward@arm.com>
9070 David Sherwood <david.sherwood@arm.com>
9072 * optabs.def (fold_left_plus_optab): New optab.
9073 * doc/md.texi (fold_left_plus_@var{m}): Document.
9074 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9075 * internal-fn.c (fold_left_direct): Define.
9076 (expand_fold_left_optab_fn): Likewise.
9077 (direct_fold_left_optab_supported_p): Likewise.
9078 * fold-const-call.c (fold_const_fold_left): New function.
9079 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9080 * tree-parloops.c (valid_reduction_p): New function.
9081 (gather_scalar_reductions): Use it.
9082 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9083 (vect_finish_replace_stmt): Declare.
9084 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9085 (needs_fold_left_reduction_p): New function, split out from...
9086 (vect_is_simple_reduction): ...here. Accept reductions that
9087 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9088 (vect_force_simple_reduction): Also store the reduction type in
9089 the assignment's STMT_VINFO_REDUC_TYPE.
9090 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9091 (merge_with_identity): New function.
9092 (vect_expand_fold_left): Likewise.
9093 (vectorize_fold_left_reduction): Likewise.
9094 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9095 scalar phi in place for it. Check for target support and reject
9096 cases that would reassociate the operation. Defer the transform
9097 phase to vectorize_fold_left_reduction.
9098 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9099 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9100 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9104 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9105 call to ifc_temp_var.
9107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9108 Alan Hayward <alan.hayward@arm.com>
9109 David Sherwood <david.sherwood@arm.com>
9111 * target.def (legitimize_address_displacement): Take the original
9112 offset as a poly_int.
9113 * targhooks.h (default_legitimize_address_displacement): Update
9115 * targhooks.c (default_legitimize_address_displacement): Likewise.
9116 * doc/tm.texi: Regenerate.
9117 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9118 as an argument, moving assert of ad->disp == ad->disp_term to...
9119 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9120 Try calling targetm.legitimize_address_displacement before expanding
9121 the address rather than afterwards, and adjust for the new interface.
9122 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9123 Match the new hook interface. Handle SVE addresses.
9124 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9127 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9129 * Makefile.in (OBJS): Add early-remat.o.
9130 * target.def (select_early_remat_modes): New hook.
9131 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9132 * doc/tm.texi: Regenerate.
9133 * targhooks.h (default_select_early_remat_modes): Declare.
9134 * targhooks.c (default_select_early_remat_modes): New function.
9135 * timevar.def (TV_EARLY_REMAT): New timevar.
9136 * passes.def (pass_early_remat): New pass.
9137 * tree-pass.h (make_pass_early_remat): Declare.
9138 * early-remat.c: New file.
9139 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9141 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9143 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9144 Alan Hayward <alan.hayward@arm.com>
9145 David Sherwood <david.sherwood@arm.com>
9147 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9148 vfm1 with a bound_epilog parameter.
9149 (vect_do_peeling): Update calls accordingly, and move the prologue
9150 call earlier in the function. Treat the base bound_epilog as 0 for
9151 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9152 this base when peeling for gaps.
9153 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9154 with fully-masked loops.
9155 (vect_estimate_min_profitable_iters): Handle the single peeled
9156 iteration in that case.
9158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9159 Alan Hayward <alan.hayward@arm.com>
9160 David Sherwood <david.sherwood@arm.com>
9162 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9163 single-element interleaving even if the size is not a power of 2.
9164 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9165 accesses for single-element interleaving if the group size is
9168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9169 Alan Hayward <alan.hayward@arm.com>
9170 David Sherwood <david.sherwood@arm.com>
9172 * doc/md.texi (fold_extract_last_@var{m}): Document.
9173 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9174 * optabs.def (fold_extract_last_optab): New optab.
9175 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9176 * internal-fn.c (fold_extract_direct): New macro.
9177 (expand_fold_extract_optab_fn): Likewise.
9178 (direct_fold_extract_optab_supported_p): Likewise.
9179 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9180 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9181 EXTRACT_LAST_REDUCTION.
9182 (get_initial_def_for_reduction): Do not create an initial vector
9183 for EXTRACT_LAST_REDUCTION reductions.
9184 (vectorizable_reduction): Leave the scalar phi in place for
9185 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9186 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9187 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9188 transform phase to vectorizable_condition.
9189 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9191 (vect_finish_stmt_generation): ...here.
9192 (vect_finish_replace_stmt): New function.
9193 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9194 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9196 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9198 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9199 Alan Hayward <alan.hayward@arm.com>
9200 David Sherwood <david.sherwood@arm.com>
9202 * doc/md.texi (extract_last_@var{m}): Document.
9203 * optabs.def (extract_last_optab): New optab.
9204 * internal-fn.def (EXTRACT_LAST): New internal function.
9205 * internal-fn.c (cond_unary_direct): New macro.
9206 (expand_cond_unary_optab_fn): Likewise.
9207 (direct_cond_unary_optab_supported_p): Likewise.
9208 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9209 loops using EXTRACT_LAST.
9210 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9211 (extract_last_<mode>): ...this optab.
9212 (vec_extract<mode><Vel>): Update accordingly.
9214 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9215 Alan Hayward <alan.hayward@arm.com>
9216 David Sherwood <david.sherwood@arm.com>
9218 * target.def (empty_mask_is_expensive): New hook.
9219 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9220 * doc/tm.texi: Regenerate.
9221 * targhooks.h (default_empty_mask_is_expensive): Declare.
9222 * targhooks.c (default_empty_mask_is_expensive): New function.
9223 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9224 if the target says that empty masks are expensive.
9225 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9227 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9230 Alan Hayward <alan.hayward@arm.com>
9231 David Sherwood <david.sherwood@arm.com>
9233 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9234 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9235 (vect_use_loop_mask_for_alignment_p): New function.
9236 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9237 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9238 niters_skip argument. Make sure that the first niters_skip elements
9239 of the first iteration are inactive.
9240 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9241 Update call to vect_set_loop_masks_directly.
9242 (get_misalign_in_elems): New function, split out from...
9243 (vect_gen_prolog_loop_niters): ...here.
9244 (vect_update_init_of_dr): Take a code argument that specifies whether
9245 the adjustment should be added or subtracted.
9246 (vect_update_init_of_drs): Likewise.
9247 (vect_prepare_for_masked_peels): New function.
9248 (vect_do_peeling): Skip prologue peeling if we're using a mask
9249 instead. Update call to vect_update_inits_of_drs.
9250 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9252 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9253 alignment. Do not include the number of peeled iterations in
9254 the minimum threshold in that case.
9255 (vectorizable_induction): Adjust the start value down by
9256 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9257 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9258 Take the number of skipped iterations into account when calculating
9260 * tree-vect-stmts.c (vect_gen_while_not): New function.
9262 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9263 Alan Hayward <alan.hayward@arm.com>
9264 David Sherwood <david.sherwood@arm.com>
9266 * doc/sourcebuild.texi (vect_fully_masked): Document.
9267 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9269 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9271 (vect_analyze_loop_2): ...here. Don't check the vectorization
9272 factor against the number of loop iterations if the loop is
9275 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9276 Alan Hayward <alan.hayward@arm.com>
9277 David Sherwood <david.sherwood@arm.com>
9279 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9280 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9281 (dump_groups): Update accordingly.
9282 (iv_use::mem_type): New member variable.
9283 (address_p): New function.
9284 (record_use): Add a mem_type argument and initialize the new
9286 (record_group_use): Add a mem_type argument. Use address_p.
9287 Remove obsolete null checks of base_object. Update call to record_use.
9288 (find_interesting_uses_op): Update call to record_group_use.
9289 (find_interesting_uses_cond): Likewise.
9290 (find_interesting_uses_address): Likewise.
9291 (get_mem_type_for_internal_fn): New function.
9292 (find_address_like_use): Likewise.
9293 (find_interesting_uses_stmt): Try find_address_like_use before
9294 calling find_interesting_uses_op.
9295 (addr_offset_valid_p): Use the iv mem_type field as the type
9296 of the addressed memory.
9297 (add_autoinc_candidates): Likewise.
9298 (get_address_cost): Likewise.
9299 (split_small_address_groups_p): Use address_p.
9300 (split_address_groups): Likewise.
9301 (add_iv_candidate_for_use): Likewise.
9302 (autoinc_possible_for_pair): Likewise.
9303 (rewrite_groups): Likewise.
9304 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9305 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9306 (get_alias_ptr_type_for_ptr_address): New function.
9307 (rewrite_use_address): Rewrite address uses in calls that were
9308 identified by find_address_like_use.
9310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9311 Alan Hayward <alan.hayward@arm.com>
9312 David Sherwood <david.sherwood@arm.com>
9314 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9316 * gimple-expr.h (is_gimple_addressable: Likewise.
9317 * gimple-expr.c (is_gimple_address): Likewise.
9318 * internal-fn.c (expand_call_mem_ref): New function.
9319 (expand_mask_load_optab_fn): Use it.
9320 (expand_mask_store_optab_fn): Likewise.
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 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9327 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9328 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9329 (cond_umax@var{mode}): Document.
9330 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9331 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9332 (cond_umin_optab, cond_umax_optab): New optabs.
9333 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9334 (COND_IOR, COND_XOR): New internal functions.
9335 * internal-fn.h (get_conditional_internal_fn): Declare.
9336 * internal-fn.c (cond_binary_direct): New macro.
9337 (expand_cond_binary_optab_fn): Likewise.
9338 (direct_cond_binary_optab_supported_p): Likewise.
9339 (get_conditional_internal_fn): New function.
9340 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9341 Cope with reduction statements that are vectorized as calls rather
9343 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9344 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9345 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9346 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9347 (UNSPEC_COND_EOR): New unspecs.
9348 (optab): Add mappings for them.
9349 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9350 (sve_int_op, sve_fp_op): New int attributes.
9352 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9353 Alan Hayward <alan.hayward@arm.com>
9354 David Sherwood <david.sherwood@arm.com>
9356 * optabs.def (while_ult_optab): New optab.
9357 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9358 * internal-fn.def (WHILE_ULT): New internal function.
9359 * internal-fn.h (direct_internal_fn_supported_p): New override
9360 that takes two types as argument.
9361 * internal-fn.c (while_direct): New macro.
9362 (expand_while_optab_fn): New function.
9363 (convert_optab_supported_p): Likewise.
9364 (direct_while_optab_supported_p): New macro.
9365 * wide-int.h (wi::udiv_ceil): New function.
9366 * tree-vectorizer.h (rgroup_masks): New structure.
9367 (vec_loop_masks): New typedef.
9368 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9370 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9371 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9372 (vect_max_vf): New function.
9373 (slpeel_make_loop_iterate_ntimes): Delete.
9374 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9375 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9376 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9377 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9378 internal-fn.h, stor-layout.h and optabs-query.h.
9379 (vect_set_loop_mask): New function.
9380 (add_preheader_seq): Likewise.
9381 (add_header_seq): Likewise.
9382 (interleave_supported_p): Likewise.
9383 (vect_maybe_permute_loop_masks): Likewise.
9384 (vect_set_loop_masks_directly): Likewise.
9385 (vect_set_loop_condition_masked): Likewise.
9386 (vect_set_loop_condition_unmasked): New function, split out from
9387 slpeel_make_loop_iterate_ntimes.
9388 (slpeel_make_loop_iterate_ntimes): Rename to..
9389 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9390 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9391 (vect_do_peeling): Update call accordingly.
9392 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9394 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9395 mask_compare_type, can_fully_mask_p and fully_masked_p.
9396 (release_vec_loop_masks): New function.
9397 (_loop_vec_info): Use it to free the loop masks.
9398 (can_produce_all_loop_masks_p): New function.
9399 (vect_get_max_nscalars_per_iter): Likewise.
9400 (vect_verify_full_masking): Likewise.
9401 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9402 retries, and free the mask rgroups before retrying. Check loop-wide
9403 reasons for disallowing fully-masked loops. Make the final decision
9404 about whether use a fully-masked loop or not.
9405 (vect_estimate_min_profitable_iters): Do not assume that peeling
9406 for the number of iterations will be needed for fully-masked loops.
9407 (vectorizable_reduction): Disable fully-masked loops.
9408 (vectorizable_live_operation): Likewise.
9409 (vect_halve_mask_nunits): New function.
9410 (vect_double_mask_nunits): Likewise.
9411 (vect_record_loop_mask): Likewise.
9412 (vect_get_loop_mask): Likewise.
9413 (vect_transform_loop): Handle the case in which the final loop
9414 iteration might handle a partial vector. Call vect_set_loop_condition
9415 instead of slpeel_make_loop_iterate_ntimes.
9416 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9417 (check_load_store_masking): New function.
9418 (prepare_load_store_mask): Likewise.
9419 (vectorizable_store): Handle fully-masked loops.
9420 (vectorizable_load): Likewise.
9421 (supportable_widening_operation): Use vect_halve_mask_nunits for
9423 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9424 (vect_gen_while): New function.
9425 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9426 (aarch64_uqdec<mode>): New insn.
9428 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9429 Alan Hayward <alan.hayward@arm.com>
9430 David Sherwood <david.sherwood@arm.com>
9432 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9433 (reduc_xor_scal_optab): New optabs.
9434 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9435 (reduc_xor_scal_@var{m}): Document.
9436 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9437 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9439 * fold-const-call.c (fold_const_call): Handle them.
9440 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9441 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9442 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9443 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9444 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9445 (UNSPEC_XORV): New unspecs.
9446 (optab): Add entries for them.
9447 (BITWISEV): New int iterator.
9448 (bit_reduc_op): New int attributes.
9450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9451 Alan Hayward <alan.hayward@arm.com>
9452 David Sherwood <david.sherwood@arm.com>
9454 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9455 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9456 * optabs.def (vec_shl_insert_optab): New optab.
9457 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9458 (duplicate_and_interleave): Likewise.
9459 * tree-vect-loop.c: Include internal-fn.h.
9460 (neutral_op_for_slp_reduction): New function, split out from
9461 get_initial_defs_for_reduction.
9462 (get_initial_def_for_reduction): Handle option 2 for variable-length
9463 vectors by loading the neutral value into a vector and then shifting
9464 the initial value into element 0.
9465 (get_initial_defs_for_reduction): Replace the code argument with
9466 the neutral value calculated by neutral_op_for_slp_reduction.
9467 Use gimple_build_vector for constant-length vectors.
9468 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9469 but the first group_size elements have a neutral value.
9470 Use duplicate_and_interleave otherwise.
9471 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9472 Update call to get_initial_defs_for_reduction. Handle SLP
9473 reductions for variable-length vectors by creating one vector
9474 result for each scalar result, with the elements associated
9475 with other scalar results stubbed out with the neutral value.
9476 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9477 Require IFN_VEC_SHL_INSERT for double reductions on
9478 variable-length vectors, or SLP reductions that have
9479 a neutral value. Require can_duplicate_and_interleave_p
9480 support for variable-length unchained SLP reductions if there
9481 is no neutral value, such as for MIN/MAX reductions. Also require
9482 the number of vector elements to be a multiple of the number of
9483 SLP statements when doing variable-length unchained SLP reductions.
9484 Update call to vect_create_epilog_for_reduction.
9485 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9486 and remove initial values.
9487 (duplicate_and_interleave): Make public.
9488 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9489 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9491 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9492 Alan Hayward <alan.hayward@arm.com>
9493 David Sherwood <david.sherwood@arm.com>
9495 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9496 (can_duplicate_and_interleave_p): New function.
9497 (vect_get_and_check_slp_defs): Take the vector of statements
9498 rather than just the current one. Remove excess parentheses.
9499 Restriction rejectinon of vect_constant_def and vect_external_def
9500 for variable-length vectors to boolean types, or types for which
9501 can_duplicate_and_interleave_p is false.
9502 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9503 (duplicate_and_interleave): New function.
9504 (vect_get_constant_vectors): Use gimple_build_vector for
9505 constant-length vectors and suitable variable-length constant
9506 vectors. Use duplicate_and_interleave for other variable-length
9507 vectors. Don't defer the update when inserting new statements.
9509 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9510 Alan Hayward <alan.hayward@arm.com>
9511 David Sherwood <david.sherwood@arm.com>
9513 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9514 min_profitable_iters doesn't go negative.
9516 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9517 Alan Hayward <alan.hayward@arm.com>
9518 David Sherwood <david.sherwood@arm.com>
9520 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9521 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9522 * optabs.def (vec_mask_load_lanes_optab): New optab.
9523 (vec_mask_store_lanes_optab): Likewise.
9524 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9525 (MASK_STORE_LANES): Likewise.
9526 * internal-fn.c (mask_load_lanes_direct): New macro.
9527 (mask_store_lanes_direct): Likewise.
9528 (expand_mask_load_optab_fn): Handle masked operations.
9529 (expand_mask_load_lanes_optab_fn): New macro.
9530 (expand_mask_store_optab_fn): Handle masked operations.
9531 (expand_mask_store_lanes_optab_fn): New macro.
9532 (direct_mask_load_lanes_optab_supported_p): Likewise.
9533 (direct_mask_store_lanes_optab_supported_p): Likewise.
9534 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9536 (vect_load_lanes_supported): Likewise.
9537 * tree-vect-data-refs.c (strip_conversion): New function.
9538 (can_group_stmts_p): Likewise.
9539 (vect_analyze_data_ref_accesses): Use it instead of checking
9540 for a pair of assignments.
9541 (vect_store_lanes_supported): Take a masked_p parameter.
9542 (vect_load_lanes_supported): Likewise.
9543 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9544 vect_store_lanes_supported and vect_load_lanes_supported.
9545 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9546 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9547 parameter. Don't allow gaps for masked accesses.
9548 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9549 and vect_load_lanes_supported.
9550 (get_load_store_type): Take a masked_p parameter and update
9551 call to get_group_load_store_type.
9552 (vectorizable_store): Update call to get_load_store_type.
9553 Handle IFN_MASK_STORE_LANES.
9554 (vectorizable_load): Update call to get_load_store_type.
9555 Handle IFN_MASK_LOAD_LANES.
9557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9558 Alan Hayward <alan.hayward@arm.com>
9559 David Sherwood <david.sherwood@arm.com>
9561 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9563 * config/aarch64/aarch64-protos.h
9564 (aarch64_sve_struct_memory_operand_p): Declare.
9565 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9566 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9567 (VPRED, vpred): Handle SVE structure modes.
9568 * config/aarch64/constraints.md (Utx): New constraint.
9569 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9570 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9571 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9572 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9573 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9574 structure modes. Split into pieces after RA.
9575 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9576 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9578 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9579 SVE structure modes.
9580 (aarch64_classify_address): Likewise.
9581 (sizetochar): Move earlier in file.
9582 (aarch64_print_operand): Handle SVE register lists.
9583 (aarch64_array_mode): New function.
9584 (aarch64_sve_struct_memory_operand_p): Likewise.
9585 (TARGET_ARRAY_MODE): Redefine.
9587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9588 Alan Hayward <alan.hayward@arm.com>
9589 David Sherwood <david.sherwood@arm.com>
9591 * target.def (array_mode): New target hook.
9592 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9593 * doc/tm.texi: Regenerate.
9594 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9595 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9596 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9598 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9601 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9602 Alan Hayward <alan.hayward@arm.com>
9603 David Sherwood <david.sherwood@arm.com>
9605 * fold-const.c (fold_binary_loc): Check the argument types
9606 rather than the result type when testing for a vector operation.
9608 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9610 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9611 * doc/tm.texi: Regenerate.
9613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9614 Alan Hayward <alan.hayward@arm.com>
9615 David Sherwood <david.sherwood@arm.com>
9617 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9618 (sve): Document new AArch64 extension.
9619 * doc/md.texi (w): Extend the description of the AArch64
9620 constraint to include SVE vectors.
9621 (Upl, Upa): Document new AArch64 predicate constraints.
9622 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9624 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9625 (msve-vector-bits=): New option.
9626 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9627 SVE when these are disabled.
9628 (sve): New extension.
9629 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9630 modes. Adjust their number of units based on aarch64_sve_vg.
9631 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9632 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9633 aarch64_addr_query_type.
9634 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9635 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9636 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9637 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9638 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9639 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9640 (aarch64_simd_imm_zero_p): Delete.
9641 (aarch64_check_zero_based_sve_index_immediate): Declare.
9642 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9643 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9644 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9645 (aarch64_sve_float_mul_immediate_p): Likewise.
9646 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9648 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9649 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9650 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9651 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9652 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9653 (aarch64_regmode_natural_size): Likewise.
9654 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9655 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9657 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9658 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9659 for VG and the SVE predicate registers.
9660 (V_ALIASES): Add a "z"-prefixed alias.
9661 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9662 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9663 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9664 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9665 (REG_CLASS_NAMES): Add entries for them.
9666 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9667 and the predicate registers.
9668 (aarch64_sve_vg): Declare.
9669 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9670 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9671 (REGMODE_NATURAL_SIZE): Define.
9672 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9674 * config/aarch64/aarch64.c: Include cfgrtl.h.
9675 (simd_immediate_info): Add a constructor for series vectors,
9676 and an associated step field.
9677 (aarch64_sve_vg): New variable.
9678 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9679 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9680 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9681 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9682 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9683 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9684 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9685 (aarch64_get_mask_mode): New functions.
9686 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9687 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9688 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9689 predicate modes and predicate registers. Explicitly restrict
9690 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9691 to store a vector mode if it is recognized by
9692 aarch64_classify_vector_mode.
9693 (aarch64_regmode_natural_size): New function.
9694 (aarch64_hard_regno_caller_save_mode): Return the original mode
9696 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9697 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9698 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9699 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9701 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9702 does not overlap dest if the function is frame-related. Handle
9704 (aarch64_split_add_offset): New function.
9705 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9706 them aarch64_add_offset.
9707 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9708 and update call to aarch64_sub_sp.
9709 (aarch64_add_cfa_expression): New function.
9710 (aarch64_expand_prologue): Pass extra temporary registers to the
9711 functions above. Handle the case in which we need to emit new
9712 DW_CFA_expressions for registers that were originally saved
9713 relative to the stack pointer, but now have to be expressed
9714 relative to the frame pointer.
9715 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9717 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9718 IP0 and IP1 values for SVE frames.
9719 (aarch64_expand_vec_series): New function.
9720 (aarch64_expand_sve_widened_duplicate): Likewise.
9721 (aarch64_expand_sve_const_vector): Likewise.
9722 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9723 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9724 into the register, rather than emitting a SET directly.
9725 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9726 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9727 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9728 (offset_9bit_signed_scaled_p): New functions.
9729 (aarch64_replicate_bitmask_imm): New function.
9730 (aarch64_bitmask_imm): Use it.
9731 (aarch64_cannot_force_const_mem): Reject expressions involving
9732 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9733 (aarch64_classify_index): Handle SVE indices, by requiring
9734 a plain register index with a scale that matches the element size.
9735 (aarch64_classify_address): Handle SVE addresses. Assert that
9736 the mode of the address is VOIDmode or an integer mode.
9737 Update call to aarch64_classify_symbol.
9738 (aarch64_classify_symbolic_expression): Update call to
9739 aarch64_classify_symbol.
9740 (aarch64_const_vec_all_in_range_p): New function.
9741 (aarch64_print_vector_float_operand): Likewise.
9742 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9743 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9744 and the FP immediates 1.0 and 0.5.
9745 (aarch64_print_address_internal): Handle SVE addresses.
9746 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9747 (aarch64_regno_regclass): Handle predicate registers.
9748 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9750 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9751 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9752 (aarch64_convert_sve_vector_bits): New function.
9753 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9754 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9756 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9757 Handle SVE vector and predicate modes. Accept VL-based constants
9758 that need only one temporary register, and VL offsets that require
9759 no temporary registers.
9760 (aarch64_conditional_register_usage): Mark the predicate registers
9761 as fixed if SVE isn't available.
9762 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9763 Return true for SVE vector and predicate modes.
9764 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9765 rather than an unsigned int. Handle SVE modes.
9766 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9768 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9770 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9771 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9772 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9773 (aarch64_sve_float_mul_immediate_p): New functions.
9774 (aarch64_sve_valid_immediate): New function.
9775 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9776 Explicitly reject structure modes. Check for INDEX constants.
9777 Handle PTRUE and PFALSE constants.
9778 (aarch64_check_zero_based_sve_index_immediate): New function.
9779 (aarch64_simd_imm_zero_p): Delete.
9780 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9781 vector modes. Accept constants in the range of CNT[BHWD].
9782 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9783 ask for an Advanced SIMD mode.
9784 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9785 (aarch64_simd_vector_alignment): Handle SVE predicates.
9786 (aarch64_vectorize_preferred_vector_alignment): New function.
9787 (aarch64_simd_vector_alignment_reachable): Use it instead of
9789 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9790 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9792 (MAX_VECT_LEN): Delete.
9793 (expand_vec_perm_d): Add a vec_flags field.
9794 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9795 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9796 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9798 (aarch64_evpc_rev): Rename to...
9799 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9800 (aarch64_evpc_rev_global): New function.
9801 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9802 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9804 (aarch64_evpc_sve_tbl): New function.
9805 (aarch64_expand_vec_perm_const_1): Update after rename of
9806 aarch64_evpc_rev. Handle SVE permutes too, trying
9807 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9808 than aarch64_evpc_tbl.
9809 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9810 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9811 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9812 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9813 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9814 (aarch64_expand_sve_vcond): New functions.
9815 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9816 of aarch64_vector_mode_p.
9817 (aarch64_dwarf_poly_indeterminate_value): New function.
9818 (aarch64_compute_pressure_classes): Likewise.
9819 (aarch64_can_change_mode_class): Likewise.
9820 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9821 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9822 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9823 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9824 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9825 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9826 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9827 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9829 (Dn, Dl, Dr): Accept const as well as const_vector.
9830 (Dz): Likewise. Compare against CONST0_RTX.
9831 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9832 of "vector" where appropriate.
9833 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9834 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9835 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9836 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9837 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9838 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9839 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9840 (v_int_equiv): Extend to SVE modes.
9841 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9843 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9844 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9845 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9846 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9847 (SVE_COND_FP_CMP): New int iterators.
9848 (perm_hilo): Handle the new unpack unspecs.
9849 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9851 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9852 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9853 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9854 (aarch64_equality_operator, aarch64_constant_vector_operand)
9855 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9856 (aarch64_sve_nonimmediate_operand): Likewise.
9857 (aarch64_sve_general_operand): Likewise.
9858 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9859 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9860 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9861 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9862 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9863 (aarch64_sve_float_arith_immediate): Likewise.
9864 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9865 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9866 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9867 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9868 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9869 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9870 (aarch64_sve_float_arith_operand): Likewise.
9871 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9872 (aarch64_sve_float_mul_operand): Likewise.
9873 (aarch64_sve_vec_perm_operand): Likewise.
9874 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9875 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9876 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9877 as well as const_vector.
9878 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9879 in file. Use CONST0_RTX and CONSTM1_RTX.
9880 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9881 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9882 Use aarch64_simd_imm_zero.
9883 * config/aarch64/aarch64-sve.md: New file.
9884 * config/aarch64/aarch64.md: Include it.
9885 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9886 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9887 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9888 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9889 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9890 (sve): New attribute.
9891 (enabled): Disable instructions with the sve attribute unless
9893 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9894 aarch64_expand_mov_immediate.
9895 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9896 CNT[BHSD] immediates.
9897 (movti): Split CONST_POLY_INT moves into two halves.
9898 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9899 Split additions that need a temporary here if the destination
9900 is the stack pointer.
9901 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9902 (*add<mode>3_poly_1): New instruction.
9903 (set_clobber_cc): New expander.
9905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9907 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9908 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9909 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9910 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9911 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9912 Change innermode from fixed_mode_size to machine_mode.
9913 (simplify_subreg): Update call accordingly. Handle a constant-sized
9914 subreg of a variable-length CONST_VECTOR.
9916 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9917 Alan Hayward <alan.hayward@arm.com>
9918 David Sherwood <david.sherwood@arm.com>
9920 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9921 (add_offset_to_base): New function, split out from...
9922 (create_mem_ref): ...here. When handling a scale other than 1,
9923 check first whether the address is valid without the offset.
9924 Add it into the base if so, leaving the index and scale as-is.
9926 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9929 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9930 fold_for_warn before checking if arg2 is INTEGER_CST.
9932 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9934 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9935 (store_multiple_operation): Delete.
9936 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9937 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9938 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9939 guarded by TARGET_STRING.
9940 (rs6000_output_load_multiple): Delete.
9941 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9942 OPTION_MASK_STRING / TARGET_STRING handling.
9943 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9944 (const rs6000_opt_masks) <"string">: Change mask to 0.
9945 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9946 (MASK_STRING): Delete.
9947 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9949 (load_multiple): Delete.
9956 (store_multiple): Delete.
9963 (movmemsi_8reg): Delete.
9964 (corresponding unnamed define_insn): Delete.
9965 (movmemsi_6reg): Delete.
9966 (corresponding unnamed define_insn): Delete.
9967 (movmemsi_4reg): Delete.
9968 (corresponding unnamed define_insn): Delete.
9969 (movmemsi_2reg): Delete.
9970 (corresponding unnamed define_insn): Delete.
9971 (movmemsi_1reg): Delete.
9972 (corresponding unnamed define_insn): Delete.
9973 * config/rs6000/rs6000.opt (mno-string): New.
9974 (mstring): Replace by deprecation warning stub.
9975 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9977 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9979 * regrename.c (regrename_do_replace): If replacing the same
9980 reg multiple times, try to reuse last created gen_raw_REG.
9983 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9984 main to workaround a bug in GDB.
9986 2018-01-12 Tom de Vries <tom@codesourcery.com>
9989 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9991 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9993 PR rtl-optimization/80481
9994 * ira-color.c (get_cap_member): New function.
9995 (allocnos_conflict_by_live_ranges_p): Use it.
9996 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9997 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9999 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10002 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10003 (*saddl_se_1): Ditto.
10004 (*ssubsi_1): Ditto.
10005 (*ssubl_se_1): Ditto.
10007 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10009 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10010 rather than wi::to_widest for DR_INITs.
10011 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10012 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10013 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10015 (vect_analyze_group_access_1): Note that here.
10017 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10019 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10020 polynomial type sizes.
10022 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10024 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10025 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10026 (gimple_add_tmp_var): Likewise.
10028 2018-01-12 Martin Liska <mliska@suse.cz>
10030 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10031 (gimple_alloc_sizes): Likewise.
10032 (dump_gimple_statistics): Use PRIu64 in printf format.
10033 * gimple.h: Change uint64_t to int.
10035 2018-01-12 Martin Liska <mliska@suse.cz>
10037 * tree-core.h: Use uint64_t instead of int.
10038 * tree.c (tree_node_counts): Likewise.
10039 (tree_node_sizes): Likewise.
10040 (dump_tree_statistics): Use PRIu64 in printf format.
10042 2018-01-12 Martin Liska <mliska@suse.cz>
10044 * Makefile.in: As qsort_chk is implemented in vec.c, add
10045 vec.o to linkage of gencfn-macros.
10046 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10047 passing the info to record_node_allocation_statistics.
10048 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10050 * ggc-common.c (struct ggc_usage): Add operator== and use
10051 it in operator< and compare function.
10052 * mem-stats.h (struct mem_usage): Likewise.
10053 * vec.c (struct vec_usage): Remove operator< and compare
10054 function. Can be simply inherited.
10056 2018-01-12 Martin Jambor <mjambor@suse.cz>
10059 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10060 * tree-ssa-math-opts.c: Include domwalk.h.
10061 (convert_mult_to_fma_1): New function.
10062 (fma_transformation_info): New type.
10063 (fma_deferring_state): Likewise.
10064 (cancel_fma_deferring): New function.
10065 (result_of_phi): Likewise.
10066 (last_fma_candidate_feeds_initial_phi): Likewise.
10067 (convert_mult_to_fma): Added deferring logic, split actual
10068 transformation to convert_mult_to_fma_1.
10069 (math_opts_dom_walker): New type.
10070 (math_opts_dom_walker::after_dom_children): New method, body moved
10071 here from pass_optimize_widening_mul::execute, added deferring logic
10073 (pass_optimize_widening_mul::execute): Moved most of code to
10074 math_opts_dom_walker::after_dom_children.
10075 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10076 * config/i386/i386.c (ix86_option_override_internal): Added
10077 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10079 2018-01-12 Richard Biener <rguenther@suse.de>
10082 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10083 inline instance vars.
10085 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10088 * config/rx/rx.c (rx_is_restricted_memory_address):
10089 Handle SUBREG case.
10091 2018-01-12 Richard Biener <rguenther@suse.de>
10093 PR tree-optimization/80846
10094 * target.def (split_reduction): New target hook.
10095 * targhooks.c (default_split_reduction): New function.
10096 * targhooks.h (default_split_reduction): Declare.
10097 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10098 target requests first reduce vectors by combining low and high
10100 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10101 (get_vectype_for_scalar_type_and_size): Export.
10102 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10103 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10104 * doc/tm.texi: Regenerate.
10105 * config/i386/i386.c (ix86_split_reduction): Implement
10106 TARGET_VECTORIZE_SPLIT_REDUCTION.
10108 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10111 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10112 in PIC mode except for TARGET_VXWORKS_RTP.
10113 * config/sparc/sparc.c: Include cfgrtl.h.
10114 (TARGET_INIT_PIC_REG): Define.
10115 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10116 (sparc_pic_register_p): New predicate.
10117 (sparc_legitimate_address_p): Use it.
10118 (sparc_legitimize_pic_address): Likewise.
10119 (sparc_delegitimize_address): Likewise.
10120 (sparc_mode_dependent_address_p): Likewise.
10121 (gen_load_pcrel_sym): Remove 4th parameter.
10122 (load_got_register): Adjust call to above. Remove obsolete stuff.
10123 (sparc_expand_prologue): Do not call load_got_register here.
10124 (sparc_flat_expand_prologue): Likewise.
10125 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10126 (sparc_use_pseudo_pic_reg): New function.
10127 (sparc_init_pic_reg): Likewise.
10128 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10129 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10131 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10133 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10134 Add item for branch_cost.
10136 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10138 PR rtl-optimization/83565
10139 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10140 not extend the result to a larger mode for rotate operations.
10141 (num_sign_bit_copies1): Likewise.
10143 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10146 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10148 Use values-Xc.o for -pedantic.
10149 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10151 2018-01-12 Martin Liska <mliska@suse.cz>
10154 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10156 (possible_polymorphic_call_targets): Use it.
10157 (ipa_devirt): Likewise.
10159 2018-01-12 Martin Liska <mliska@suse.cz>
10161 * profile-count.h (enum profile_quality): Use 0 as invalid
10162 enum value of profile_quality.
10164 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10166 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10167 -mext-string options.
10169 2018-01-12 Richard Biener <rguenther@suse.de>
10171 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10172 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10173 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10175 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10177 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10179 * configure.ac (--with-long-double-format): Add support for the
10180 configuration option to change the default long double format on
10182 * config.gcc (powerpc*-linux*-*): Likewise.
10183 * configure: Regenerate.
10184 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10185 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10186 used without modification.
10188 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10190 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10191 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10192 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10193 MISC_BUILTIN_SPEC_BARRIER.
10194 (rs6000_init_builtins): Likewise.
10195 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10197 (speculation_barrier): New define_insn.
10198 * doc/extend.texi: Document __builtin_speculation_barrier.
10200 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10203 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10204 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10205 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10207 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10208 integral modes instead of "ss" and "sd".
10209 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10210 vectors with 32-bit and 64-bit elements.
10211 (vecdupssescalarmodesuffix): New mode attribute.
10212 (vec_dup<mode>): Use it.
10214 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10217 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10218 frame if argument is passed on stack.
10220 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10223 * ree.c (combine_reaching_defs): Optimize also
10224 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10225 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10227 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10229 PR middle-end/83189
10230 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10232 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10234 PR middle-end/83718
10235 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10236 after they are computed.
10238 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10240 PR tree-optimization/83695
10241 * gimple-loop-linterchange.cc
10242 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10243 reset cached scev information after interchange.
10244 (pass_linterchange::execute): Remove call to scev_reset_htab.
10246 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10248 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10249 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10250 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10251 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10252 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10253 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10254 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10255 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10256 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10257 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10258 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10259 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10260 (V_lane_reg): Likewise.
10261 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10263 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10264 (vfmal_lane_low<mode>_intrinsic,
10265 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10266 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10267 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10268 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10269 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10270 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10272 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10274 * config/arm/arm-cpus.in (fp16fml): New feature.
10275 (ALL_SIMD): Add fp16fml.
10276 (armv8.2-a): Add fp16fml as an option.
10277 (armv8.3-a): Likewise.
10278 (armv8.4-a): Add fp16fml as part of fp16.
10279 * config/arm/arm.h (TARGET_FP16FML): Define.
10280 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10282 * config/arm/arm-modes.def (V2HF): Define.
10283 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10284 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10285 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10286 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10287 vfmsl_low, vfmsl_high): New set of builtins.
10288 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10289 (vfml_op): New code attribute.
10290 (VFMLHALVES): New int iterator.
10291 (VFML, VFMLSEL): New mode attributes.
10292 (V_reg): Define mapping for V2HF.
10293 (V_hi, V_lo): New mode attributes.
10294 (VF_constraint): Likewise.
10295 (vfml_half, vfml_half_selector): New int attributes.
10296 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10298 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10299 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10301 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10302 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10303 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10304 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10306 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10307 Document new effective target and option set.
10309 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10311 * config/arm/arm-cpus.in (armv8_4): New feature.
10312 (ARMv8_4a): New fgroup.
10313 (armv8.4-a): New arch.
10314 * config/arm/arm-tables.opt: Regenerate.
10315 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10316 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10317 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10318 Add matching rules for -march=armv8.4-a and extensions.
10319 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10321 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10324 * config/rx/rx.md (BW): New mode attribute.
10325 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10327 2018-01-11 Richard Biener <rguenther@suse.de>
10329 PR tree-optimization/83435
10330 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10331 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10332 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10334 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10335 Alan Hayward <alan.hayward@arm.com>
10336 David Sherwood <david.sherwood@arm.com>
10338 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10340 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10341 (aarch64_print_address_internal): Use it to check for a zero offset.
10343 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10344 Alan Hayward <alan.hayward@arm.com>
10345 David Sherwood <david.sherwood@arm.com>
10347 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10348 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10349 Return a poly_int64 rather than a HOST_WIDE_INT.
10350 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10351 rather than a HOST_WIDE_INT.
10352 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10353 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10354 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10355 final_offset from HOST_WIDE_INT to poly_int64.
10356 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10357 to_constant when getting the number of units in an Advanced SIMD
10359 (aarch64_builtin_vectorized_function): Check for a constant number
10361 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10363 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10364 attribute instead of GET_MODE_NUNITS.
10365 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10366 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10367 GET_MODE_SIZE for fixed-size registers.
10368 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10369 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10370 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10371 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10372 (aarch64_print_operand, aarch64_print_address_internal)
10373 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10374 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10375 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10376 Handle polynomial GET_MODE_SIZE.
10377 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10378 wider than SImode without modification.
10379 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10380 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10381 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10382 passing and returning SVE modes.
10383 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10384 rather than GEN_INT.
10385 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10386 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10387 (aarch64_allocate_and_probe_stack_space): Likewise.
10388 (aarch64_layout_frame): Cope with polynomial offsets.
10389 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10390 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10391 polynomial offsets.
10392 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10393 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10394 poly_int64 rather than a HOST_WIDE_INT.
10395 (aarch64_get_separate_components, aarch64_process_components)
10396 (aarch64_expand_prologue, aarch64_expand_epilogue)
10397 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10398 (aarch64_anchor_offset): New function, split out from...
10399 (aarch64_legitimize_address): ...here.
10400 (aarch64_builtin_vectorization_cost): Handle polynomial
10401 TYPE_VECTOR_SUBPARTS.
10402 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10404 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10405 number of elements from the PARALLEL rather than the mode.
10406 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10407 rather than GET_MODE_BITSIZE.
10408 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10409 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10410 (aarch64_expand_vec_perm_const_1): Handle polynomial
10411 d->perm.length () and d->perm elements.
10412 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10413 Apply to_constant to d->perm elements.
10414 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10415 polynomial CONST_VECTOR_NUNITS.
10416 (aarch64_move_pointer): Take amount as a poly_int64 rather
10418 (aarch64_progress_pointer): Avoid temporary variable.
10419 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10420 the mode attribute instead of GET_MODE.
10422 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10423 Alan Hayward <alan.hayward@arm.com>
10424 David Sherwood <david.sherwood@arm.com>
10426 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10427 x exists before using it.
10428 (aarch64_add_constant_internal): Rename to...
10429 (aarch64_add_offset_1): ...this. Replace regnum with separate
10430 src and dest rtxes. Handle the case in which they're different,
10431 including when the offset is zero. Replace scratchreg with an rtx.
10432 Use 2 additions if there is no spare register into which we can
10433 move a 16-bit constant.
10434 (aarch64_add_constant): Delete.
10435 (aarch64_add_offset): Replace reg with separate src and dest
10436 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10437 Use aarch64_add_offset_1.
10438 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10439 an rtx rather than an int. Take the delta as a poly_int64
10440 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10441 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10442 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10443 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10444 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10445 and aarch64_add_sp.
10446 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10447 aarch64_add_constant.
10449 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10451 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10452 Use scalar_float_mode.
10454 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10456 * config/aarch64/aarch64-simd.md
10457 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10458 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10459 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10460 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10461 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10462 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10463 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10464 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10465 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10466 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10468 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10471 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10472 targ_options->x_arm_arch_string is non NULL.
10474 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10476 * config/aarch64/aarch64.h
10477 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10479 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10482 * expmed.c (emit_store_flag_force): Swap if const op0
10483 and change VOIDmode to mode of op0.
10485 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10487 PR rtl-optimization/83761
10488 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10489 than bytes to mode_for_size.
10491 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10493 PR middle-end/83189
10494 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10495 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10498 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10500 PR middle-end/83575
10501 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10502 when in layout mode.
10503 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10504 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10507 2018-01-10 Michael Collison <michael.collison@arm.com>
10509 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10510 * config/aarch64/aarch64-option-extension.def: Add
10511 AARCH64_OPT_EXTENSION of 'fp16fml'.
10512 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10513 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10514 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10515 * config/aarch64/constraints.md (Ui7): New constraint.
10516 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10517 (VFMLA_SEL_W): Ditto.
10520 (VFMLA16_LOW): New int iterator.
10521 (VFMLA16_HIGH): Ditto.
10522 (UNSPEC_FMLAL): New unspec.
10523 (UNSPEC_FMLSL): Ditto.
10524 (UNSPEC_FMLAL2): Ditto.
10525 (UNSPEC_FMLSL2): Ditto.
10526 (f16mac): New code attribute.
10527 * config/aarch64/aarch64-simd-builtins.def
10528 (aarch64_fmlal_lowv2sf): Ditto.
10529 (aarch64_fmlsl_lowv2sf): Ditto.
10530 (aarch64_fmlalq_lowv4sf): Ditto.
10531 (aarch64_fmlslq_lowv4sf): Ditto.
10532 (aarch64_fmlal_highv2sf): Ditto.
10533 (aarch64_fmlsl_highv2sf): Ditto.
10534 (aarch64_fmlalq_highv4sf): Ditto.
10535 (aarch64_fmlslq_highv4sf): Ditto.
10536 (aarch64_fmlal_lane_lowv2sf): Ditto.
10537 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10538 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10539 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10540 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10541 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10542 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10543 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10544 (aarch64_fmlal_lane_highv2sf): Ditto.
10545 (aarch64_fmlsl_lane_highv2sf): Ditto.
10546 (aarch64_fmlal_laneq_highv2sf): Ditto.
10547 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10548 (aarch64_fmlalq_lane_highv4sf): Ditto.
10549 (aarch64_fmlsl_lane_highv4sf): Ditto.
10550 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10551 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10552 * config/aarch64/aarch64-simd.md:
10553 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10554 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10555 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10556 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10557 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10558 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10559 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10560 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10561 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10562 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10563 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10564 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10565 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10566 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10567 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10568 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10569 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10570 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10571 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10572 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10573 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10574 (vfmlsl_low_u32): Ditto.
10575 (vfmlalq_low_u32): Ditto.
10576 (vfmlslq_low_u32): Ditto.
10577 (vfmlal_high_u32): Ditto.
10578 (vfmlsl_high_u32): Ditto.
10579 (vfmlalq_high_u32): Ditto.
10580 (vfmlslq_high_u32): Ditto.
10581 (vfmlal_lane_low_u32): Ditto.
10582 (vfmlsl_lane_low_u32): Ditto.
10583 (vfmlal_laneq_low_u32): Ditto.
10584 (vfmlsl_laneq_low_u32): Ditto.
10585 (vfmlalq_lane_low_u32): Ditto.
10586 (vfmlslq_lane_low_u32): Ditto.
10587 (vfmlalq_laneq_low_u32): Ditto.
10588 (vfmlslq_laneq_low_u32): Ditto.
10589 (vfmlal_lane_high_u32): Ditto.
10590 (vfmlsl_lane_high_u32): Ditto.
10591 (vfmlal_laneq_high_u32): Ditto.
10592 (vfmlsl_laneq_high_u32): Ditto.
10593 (vfmlalq_lane_high_u32): Ditto.
10594 (vfmlslq_lane_high_u32): Ditto.
10595 (vfmlalq_laneq_high_u32): Ditto.
10596 (vfmlslq_laneq_high_u32): Ditto.
10597 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10598 (AARCH64_FL_FOR_ARCH8_4): New.
10599 (AARCH64_ISA_F16FML): New ISA flag.
10600 (TARGET_F16FML): New feature flag for fp16fml.
10601 (doc/invoke.texi): Document new fp16fml option.
10603 2018-01-10 Michael Collison <michael.collison@arm.com>
10605 * config/aarch64/aarch64-builtins.c:
10606 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10607 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10608 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10609 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10610 (AARCH64_ISA_SHA3): New ISA flag.
10611 (TARGET_SHA3): New feature flag for sha3.
10612 * config/aarch64/iterators.md (sha512_op): New int attribute.
10613 (CRYPTO_SHA512): New int iterator.
10614 (UNSPEC_SHA512H): New unspec.
10615 (UNSPEC_SHA512H2): Ditto.
10616 (UNSPEC_SHA512SU0): Ditto.
10617 (UNSPEC_SHA512SU1): Ditto.
10618 * config/aarch64/aarch64-simd-builtins.def
10619 (aarch64_crypto_sha512hqv2di): New builtin.
10620 (aarch64_crypto_sha512h2qv2di): Ditto.
10621 (aarch64_crypto_sha512su0qv2di): Ditto.
10622 (aarch64_crypto_sha512su1qv2di): Ditto.
10623 (aarch64_eor3qv8hi): Ditto.
10624 (aarch64_rax1qv2di): Ditto.
10625 (aarch64_xarqv2di): Ditto.
10626 (aarch64_bcaxqv8hi): Ditto.
10627 * config/aarch64/aarch64-simd.md:
10628 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10629 (aarch64_crypto_sha512su0qv2di): Ditto.
10630 (aarch64_crypto_sha512su1qv2di): Ditto.
10631 (aarch64_eor3qv8hi): Ditto.
10632 (aarch64_rax1qv2di): Ditto.
10633 (aarch64_xarqv2di): Ditto.
10634 (aarch64_bcaxqv8hi): Ditto.
10635 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10636 (vsha512h2q_u64): Ditto.
10637 (vsha512su0q_u64): Ditto.
10638 (vsha512su1q_u64): Ditto.
10639 (veor3q_u16): Ditto.
10640 (vrax1q_u64): Ditto.
10641 (vxarq_u64): Ditto.
10642 (vbcaxq_u16): Ditto.
10643 * config/arm/types.md (crypto_sha512): New type attribute.
10644 (crypto_sha3): Ditto.
10645 (doc/invoke.texi): Document new sha3 option.
10647 2018-01-10 Michael Collison <michael.collison@arm.com>
10649 * config/aarch64/aarch64-builtins.c:
10650 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10651 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10652 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10653 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10654 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10655 (AARCH64_ISA_SM4): New ISA flag.
10656 (TARGET_SM4): New feature flag for sm4.
10657 * config/aarch64/aarch64-simd-builtins.def
10658 (aarch64_sm3ss1qv4si): Ditto.
10659 (aarch64_sm3tt1aq4si): Ditto.
10660 (aarch64_sm3tt1bq4si): Ditto.
10661 (aarch64_sm3tt2aq4si): Ditto.
10662 (aarch64_sm3tt2bq4si): Ditto.
10663 (aarch64_sm3partw1qv4si): Ditto.
10664 (aarch64_sm3partw2qv4si): Ditto.
10665 (aarch64_sm4eqv4si): Ditto.
10666 (aarch64_sm4ekeyqv4si): Ditto.
10667 * config/aarch64/aarch64-simd.md:
10668 (aarch64_sm3ss1qv4si): Ditto.
10669 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10670 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10671 (aarch64_sm4eqv4si): Ditto.
10672 (aarch64_sm4ekeyqv4si): Ditto.
10673 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10674 (sm3part_op): Ditto.
10675 (CRYPTO_SM3TT): Ditto.
10676 (CRYPTO_SM3PART): Ditto.
10677 (UNSPEC_SM3SS1): New unspec.
10678 (UNSPEC_SM3TT1A): Ditto.
10679 (UNSPEC_SM3TT1B): Ditto.
10680 (UNSPEC_SM3TT2A): Ditto.
10681 (UNSPEC_SM3TT2B): Ditto.
10682 (UNSPEC_SM3PARTW1): Ditto.
10683 (UNSPEC_SM3PARTW2): Ditto.
10684 (UNSPEC_SM4E): Ditto.
10685 (UNSPEC_SM4EKEY): Ditto.
10686 * config/aarch64/constraints.md (Ui2): New constraint.
10687 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10688 * config/arm/types.md (crypto_sm3): New type attribute.
10689 (crypto_sm4): Ditto.
10690 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10691 (vsm3tt1aq_u32): Ditto.
10692 (vsm3tt1bq_u32): Ditto.
10693 (vsm3tt2aq_u32): Ditto.
10694 (vsm3tt2bq_u32): Ditto.
10695 (vsm3partw1q_u32): Ditto.
10696 (vsm3partw2q_u32): Ditto.
10697 (vsm4eq_u32): Ditto.
10698 (vsm4ekeyq_u32): Ditto.
10699 (doc/invoke.texi): Document new sm4 option.
10701 2018-01-10 Michael Collison <michael.collison@arm.com>
10703 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10704 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10705 (AARCH64_FL_FOR_ARCH8_4): New.
10706 (AARCH64_FL_V8_4): New flag.
10707 (doc/invoke.texi): Document new armv8.4-a option.
10709 2018-01-10 Michael Collison <michael.collison@arm.com>
10711 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10712 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10713 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10714 * config/aarch64/aarch64-option-extension.def: Add
10715 AARCH64_OPT_EXTENSION of 'sha2'.
10716 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10717 (crypto): Disable sha2 and aes if crypto disabled.
10718 (crypto): Enable aes and sha2 if enabled.
10719 (simd): Disable sha2 and aes if simd disabled.
10720 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10722 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10723 (TARGET_SHA2): New feature flag for sha2.
10724 (TARGET_AES): New feature flag for aes.
10725 * config/aarch64/aarch64-simd.md:
10726 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10727 conditional on TARGET_AES.
10728 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10729 (aarch64_crypto_sha1hsi): Make pattern conditional
10731 (aarch64_crypto_sha1hv4si): Ditto.
10732 (aarch64_be_crypto_sha1hv4si): Ditto.
10733 (aarch64_crypto_sha1su1v4si): Ditto.
10734 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10735 (aarch64_crypto_sha1su0v4si): Ditto.
10736 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10737 (aarch64_crypto_sha256su0v4si): Ditto.
10738 (aarch64_crypto_sha256su1v4si): Ditto.
10739 (doc/invoke.texi): Document new aes and sha2 options.
10741 2018-01-10 Martin Sebor <msebor@redhat.com>
10743 PR tree-optimization/83781
10744 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10747 2018-01-11 Martin Sebor <msebor@gmail.com>
10748 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10750 PR tree-optimization/83501
10751 PR tree-optimization/81703
10753 * tree-ssa-strlen.c (get_string_cst): Rename...
10754 (get_string_len): ...to this. Handle global constants.
10755 (handle_char_store): Adjust.
10757 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10758 Jim Wilson <jimw@sifive.com>
10760 * config/riscv/riscv-protos.h (riscv_output_return): New.
10761 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10762 (riscv_attribute_table, riscv_output_return),
10763 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10764 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10765 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10767 (riscv_expand_prologue): Add early return for naked function.
10768 (riscv_expand_epilogue): Likewise.
10769 (riscv_function_ok_for_sibcall): Return false for naked function.
10770 (riscv_set_current_function): New.
10771 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10772 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10773 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10774 * doc/extend.texi (RISC-V Function Attributes): New.
10776 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10778 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10779 check for 128-bit long double before checking TCmode.
10780 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10781 128-bit long doubles before checking TFmode or TCmode.
10782 (FLOAT128_IBM_P): Likewise.
10784 2018-01-10 Martin Sebor <msebor@redhat.com>
10786 PR tree-optimization/83671
10787 * builtins.c (c_strlen): Unconditionally return zero for the empty
10789 Use -Warray-bounds for warnings.
10790 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10791 for non-constant array indices with COMPONENT_REF, arrays of
10792 arrays, and pointers to arrays.
10793 (gimple_fold_builtin_strlen): Determine and set length range for
10794 non-constant character arrays.
10796 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10798 PR middle-end/81897
10799 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10802 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10804 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10806 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10809 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10810 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10811 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10812 indexed_or_indirect_operand predicate.
10813 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10814 (*vsx_le_perm_load_v8hi): Likewise.
10815 (*vsx_le_perm_load_v16qi): Likewise.
10816 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10817 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10818 (*vsx_le_perm_store_v8hi): Likewise.
10819 (*vsx_le_perm_store_v16qi): Likewise.
10820 (eight unnamed splitters): Likewise.
10822 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10824 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10825 * config/rs6000/emmintrin.h: Likewise.
10826 * config/rs6000/mmintrin.h: Likewise.
10827 * config/rs6000/xmmintrin.h: Likewise.
10829 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10832 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10834 * tree.c (tree_nop_conversion): Return true for location wrapper
10836 (maybe_wrap_with_location): New function.
10837 (selftest::check_strip_nops): New function.
10838 (selftest::test_location_wrappers): New function.
10839 (selftest::tree_c_tests): Call it.
10840 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10841 (maybe_wrap_with_location): New decl.
10842 (EXPR_LOCATION_WRAPPER_P): New macro.
10843 (location_wrapper_p): New inline function.
10844 (tree_strip_any_location_wrapper): New inline function.
10846 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10849 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10850 stack_realign_offset for the largest alignment of stack slot
10852 (ix86_find_max_used_stack_alignment): New function.
10853 (ix86_finalize_stack_frame_flags): Use it. Set
10854 max_used_stack_alignment if we don't realign stack.
10855 * config/i386/i386.h (machine_function): Add
10856 max_used_stack_alignment.
10858 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10860 * config/arm/arm.opt (-mbranch-cost): New option.
10861 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10864 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10867 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10868 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10870 2018-01-10 Richard Biener <rguenther@suse.de>
10873 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10874 early out so it also covers the case where we have a non-NULL
10877 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10879 PR tree-optimization/83753
10880 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10881 for non-strided grouped accesses if the number of elements is 1.
10883 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10886 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10887 * i386.h (TARGET_USE_GATHER): Define.
10888 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10890 2018-01-10 Martin Liska <mliska@suse.cz>
10893 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10894 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10896 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10897 CLEANUP_NO_PARTITIONING is not set.
10899 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10901 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10902 for vectors, as a partial revert of r254296.
10903 * rtl.h (const_vec_p): Delete.
10904 (const_vec_duplicate_p): Don't test for vector CONSTs.
10905 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10906 * expmed.c (make_tree): Likewise.
10909 * common.md (E, F): Use CONSTANT_P instead of checking for
10911 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10912 checking for CONST_VECTOR.
10914 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10916 PR middle-end/83575
10917 * predict.c (force_edge_cold): Handle in more sane way edges
10918 with no prediction.
10920 2018-01-09 Carl Love <cel@us.ibm.com>
10922 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10924 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10925 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10926 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10927 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10928 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10929 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10930 * config/rs6000/rs6000-protos.h: Add extern defition for
10931 rs6000_generate_float2_double_code.
10932 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10934 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10935 (float2_v2df): Add define_expand.
10937 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10940 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10941 op_mode in the force_to_mode call.
10943 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10945 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10946 instead of checking each element individually.
10947 (aarch64_evpc_uzp): Likewise.
10948 (aarch64_evpc_zip): Likewise.
10949 (aarch64_evpc_ext): Likewise.
10950 (aarch64_evpc_rev): Likewise.
10951 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10952 instead of checking each element individually. Return true without
10954 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10955 whether all selected elements come from the same input, instead of
10956 checking each element individually. Remove calls to gen_rtx_REG,
10957 start_sequence and end_sequence and instead assert that no rtl is
10960 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10962 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10963 order of HIGH and CONST checks.
10965 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10967 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10968 if the destination isn't an SSA_NAME.
10970 2018-01-09 Richard Biener <rguenther@suse.de>
10972 PR tree-optimization/83668
10973 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10975 (canonicalize_loop_form): ... here, renamed from ...
10976 (canonicalize_loop_closed_ssa_form): ... this and amended to
10977 swap successor edges for loop exit blocks to make us use
10978 the RPO order we need for initial schedule generation.
10980 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10982 PR tree-optimization/64811
10983 * match.pd: When optimizing comparisons with Inf, avoid
10984 introducing or losing exceptions from comparisons with NaN.
10986 2018-01-09 Martin Liska <mliska@suse.cz>
10989 * asan.c (shadow_mem_size): Add gcc_assert.
10991 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10993 Don't save registers in main().
10996 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10997 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10998 * config/avr/avr.c (avr_set_current_function): Don't error if
10999 naked, OS_task or OS_main are specified at the same time.
11000 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11002 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11004 * common/config/avr/avr-common.c (avr_option_optimization_table):
11005 Switch on -mmain-is-OS_task for optimizing compilations.
11007 2018-01-09 Richard Biener <rguenther@suse.de>
11009 PR tree-optimization/83572
11010 * graphite.c: Include cfganal.h.
11011 (graphite_transform_loops): Connect infinite loops to exit
11012 and remove fake edges at the end.
11014 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11016 * ipa-inline.c (edge_badness): Revert accidental checkin.
11018 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11021 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11022 symbols; not inline clones.
11024 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11027 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11028 hard registers. Formatting fixes.
11030 PR preprocessor/83722
11031 * gcc.c (try_generate_repro): Pass
11032 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11033 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11036 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11037 Kito Cheng <kito.cheng@gmail.com>
11039 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11040 (riscv_leaf_function_p): Delete.
11041 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11043 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11045 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11047 (do_ifelse): New function.
11048 (do_isel): New function.
11049 (do_sub3): New function.
11050 (do_add3): New function.
11051 (do_load_mask_compare): New function.
11052 (do_overlap_load_compare): New function.
11053 (expand_compare_loop): New function.
11054 (expand_block_compare): Call expand_compare_loop() when appropriate.
11055 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11056 option description.
11057 (-mblock-compare-inline-loop-limit): New option.
11059 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11062 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11063 Reverse order of second and third operands in first alternative.
11064 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11065 of first and second elements in UNSPEC_VPERMR vector.
11066 (altivec_expand_vec_perm_le): Likewise.
11068 2018-01-08 Jeff Law <law@redhat.com>
11070 PR rtl-optimizatin/81308
11071 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11072 (process_switch): If group_case_labels makes a change, then set
11074 (pass_convert_switch::execute): If a switch is converted, then
11075 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11077 PR rtl-optimization/81308
11078 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11081 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11083 PR target/83663 - Revert r255946
11084 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11085 generation for cases where splatting a value is not useful.
11086 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11087 across a vec_duplicate and a paradoxical subreg forming a vector
11088 mode to a vec_concat.
11090 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11092 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11093 -march=armv8.3-a variants.
11094 * config/arm/t-multilib: Likewise.
11095 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11097 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11099 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11101 (cceq_ior_compare_complement): Give it a name so I can use it, and
11102 change boolean_or_operator predicate to boolean_operator so it can
11103 be used to generate a crand.
11104 (eqne): New code iterator.
11105 (bd/bd_neg): New code_attrs.
11106 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11107 a single define_insn.
11108 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11109 decrement (bdnzt/bdnzf/bdzt/bdzf).
11110 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11111 with the new names of the branch decrement patterns, and added the
11112 names of the branch decrement conditional patterns.
11114 2018-01-08 Richard Biener <rguenther@suse.de>
11116 PR tree-optimization/83563
11117 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11120 2018-01-08 Richard Biener <rguenther@suse.de>
11122 PR middle-end/83713
11123 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11125 2018-01-08 Richard Biener <rguenther@suse.de>
11127 PR tree-optimization/83685
11128 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11129 references to abnormals.
11131 2018-01-08 Richard Biener <rguenther@suse.de>
11134 * dwarf2out.c (output_indirect_strings): Handle empty
11135 skeleton_debug_str_hash.
11136 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11138 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11140 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11141 (emit_store_direct): Likewise.
11142 (arc_trampoline_adjust_address): Likewise.
11143 (arc_asm_trampoline_template): New function.
11144 (arc_initialize_trampoline): Use asm_trampoline_template.
11145 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11146 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11147 * config/arc/arc.md (flush_icache): Delete pattern.
11149 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11151 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11152 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11155 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11158 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11159 by not USED_FOR_TARGET.
11160 (make_pass_resolve_sw_modes): Likewise.
11162 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11164 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11167 2018-01-08 Richard Biener <rguenther@suse.de>
11169 PR middle-end/83580
11170 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11172 2018-01-08 Richard Biener <rguenther@suse.de>
11174 PR middle-end/83517
11175 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11177 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11179 PR middle-end/81897
11180 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11181 basic blocks with a small number of successors.
11182 (convert_control_dep_chain_into_preds): Improve handling of
11184 (dump_predicates): Split apart into...
11185 (dump_pred_chain): ...here...
11186 (dump_pred_info): ...and here.
11187 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11188 (can_chain_union_be_invalidated_p): Improve check for invalidation
11190 (uninit_uses_cannot_happen): Avoid unnecessary if
11191 convert_control_dep_chain_into_preds yielded nothing.
11193 2018-01-06 Martin Sebor <msebor@redhat.com>
11195 PR tree-optimization/83640
11196 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11197 subtracting negative offset from size.
11198 (builtin_access::overlap): Adjust offset bounds of the access to fall
11199 within the size of the object if possible.
11201 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11203 PR rtl-optimization/83699
11204 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11205 extract_bit_field_as_subreg to cases in which the extracted
11206 value is also a vector.
11208 * lra-constraints.c (process_alt_operands): Test for the equivalence
11209 substitutions when detecting a possible reload cycle.
11211 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11214 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11215 by default if flag_selective_schedling{,2}. Formatting fixes.
11217 PR rtl-optimization/83682
11218 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11219 if it has non-VECTOR_MODE element mode.
11220 (vec_duplicate_p): Likewise.
11222 PR middle-end/83694
11223 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11224 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11226 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11229 * config/i386/i386-builtin.def
11230 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11231 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11232 Require also OPTION_MASK_ISA_AVX512F in addition to
11233 OPTION_MASK_ISA_GFNI.
11234 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11235 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11236 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11237 to OPTION_MASK_ISA_GFNI.
11238 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11239 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11240 OPTION_MASK_ISA_AVX512BW.
11241 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11242 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11243 addition to OPTION_MASK_ISA_GFNI.
11244 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11245 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11246 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11247 to OPTION_MASK_ISA_GFNI.
11248 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11249 a requirement for all ISAs rather than any of them with a few
11251 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11253 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11254 bitmasks to be enabled with 3 exceptions, instead of requiring any
11255 enabled ISA with lots of exceptions.
11256 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11257 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11258 Change avx512bw in isa attribute to avx512f.
11259 * config/i386/sgxintrin.h: Add license boilerplate.
11260 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11261 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11262 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11263 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11265 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11266 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11267 temporarily sse2 rather than sse if not enabled already.
11270 * config/i386/sse.md (VI248_VLBW): Rename to ...
11271 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11272 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11273 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11274 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11275 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11276 mode iterator instead of VI248_VLBW.
11278 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11280 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11281 (record_modified): Skip clobbers; add debug output.
11282 (param_change_prob): Use sreal frequencies.
11284 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11286 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11287 punt for user-aligned variables.
11289 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11291 * tree-chrec.c (chrec_contains_symbols): Return true for
11294 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11297 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11298 of (x|y) == x for BICS pattern.
11300 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11302 PR tree-optimization/83605
11303 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11304 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11307 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11309 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11310 * config/epiphany/rtems.h: New file.
11312 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11313 Uros Bizjak <ubizjak@gmail.com>
11316 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11317 QIreg_operand instead of register_operand predicate.
11318 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11319 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11320 comments instead of -fmitigate[-_]rop.
11322 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11325 * cgraphunit.c (symbol_table::compile): Switch to text_section
11326 before calling assembly_start debug hook.
11327 * run-rtl-passes.c (run_rtl_passes): Likewise.
11330 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11332 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11333 range_int_cst_p rather than !symbolic_range_p before calling
11334 extract_range_from_multiplicative_op_1.
11336 2018-01-04 Jeff Law <law@redhat.com>
11338 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11339 redundant test in assertion.
11341 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11343 * doc/rtl.texi: Document machine_mode wrapper classes.
11345 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11347 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11348 using tree_to_uhwi.
11350 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11352 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11353 the VEC_PERM_EXPR fold to fail.
11355 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11358 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11359 to switched_sections.
11361 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11364 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11365 test for d.testing.
11367 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11370 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11371 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11373 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11376 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11377 is BLKmode and bitpos not zero or mode change is needed.
11379 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11382 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11385 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11388 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11389 instead of MULT rtx. Update all corresponding splitters.
11390 (*saddl_se): Ditto.
11391 (*ssub<modesuffix>): Ditto.
11392 (*ssubl_se): Ditto.
11393 (*cmp_sadd_di): Update split patterns.
11394 (*cmp_sadd_si): Ditto.
11395 (*cmp_sadd_sidi): Ditto.
11396 (*cmp_ssub_di): Ditto.
11397 (*cmp_ssub_si): Ditto.
11398 (*cmp_ssub_sidi): Ditto.
11399 * config/alpha/predicates.md (const23_operand): New predicate.
11400 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11401 Look for ASHIFT, not MULT inner operand.
11402 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11404 2018-01-04 Martin Liska <mliska@suse.cz>
11406 PR gcov-profile/83669
11407 * gcov.c (output_intermediate_file): Add version to intermediate
11409 * doc/gcov.texi: Document new field 'version' in intermediate
11410 file format. Fix location of '-k' option of gcov command.
11412 2018-01-04 Martin Liska <mliska@suse.cz>
11415 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11417 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11419 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11421 2018-01-03 Martin Sebor <msebor@redhat.com>
11423 PR tree-optimization/83655
11424 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11425 checking calls with invalid arguments.
11427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11429 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11430 (vectorizable_mask_load_store): Delete.
11431 (vectorizable_call): Return false for masked loads and stores.
11432 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11433 instead of gimple_assign_rhs1.
11434 (vectorizable_load): Handle IFN_MASK_LOAD.
11435 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11439 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11441 (vectorizable_mask_load_store): ...here.
11442 (vectorizable_load): ...and here.
11444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11446 * tree-vect-stmts.c (vect_build_all_ones_mask)
11447 (vect_build_zero_merge_argument): New functions, split out from...
11448 (vectorizable_load): ...here.
11450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11452 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11454 (vectorizable_mask_load_store): ...here.
11455 (vectorizable_store): ...and here.
11457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11459 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11461 (vectorizable_mask_load_store): ...here.
11463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11465 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11466 (vect_model_store_cost): Take a vec_load_store_type instead of a
11468 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11469 (vect_model_store_cost): Take a vec_load_store_type instead of a
11471 (vectorizable_mask_load_store): Update accordingly.
11472 (vectorizable_store): Likewise.
11473 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11477 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11478 IFN_MASK_LOAD calls here rather than...
11479 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11482 Alan Hayward <alan.hayward@arm.com>
11483 David Sherwood <david.sherwood@arm.com>
11485 * expmed.c (extract_bit_field_1): For vector extracts,
11486 fall back to extract_bit_field_as_subreg if vec_extract
11489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11490 Alan Hayward <alan.hayward@arm.com>
11491 David Sherwood <david.sherwood@arm.com>
11493 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11494 they are variable or constant sized.
11495 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11496 slots for constant-sized data.
11498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11499 Alan Hayward <alan.hayward@arm.com>
11500 David Sherwood <david.sherwood@arm.com>
11502 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11503 handling COND_EXPRs with boolean comparisons, try to find a better
11504 basis for the mask type than the boolean itself.
11506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11508 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11509 is calculated and how it can be overridden.
11510 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11511 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11513 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11517 Alan Hayward <alan.hayward@arm.com>
11518 David Sherwood <david.sherwood@arm.com>
11520 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11521 Remove the mode argument.
11522 (aarch64_simd_valid_immediate): Remove the mode and inverse
11524 * config/aarch64/iterators.md (bitsize): New iterator.
11525 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11526 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11527 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11528 aarch64_simd_valid_immediate.
11529 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11530 (aarch64_reg_or_bic_imm): Likewise.
11531 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11532 with an insn_type enum and msl with a modifier_type enum.
11533 Replace element_width with a scalar_mode. Change the shift
11534 to unsigned int. Add constructors for scalar_float_mode and
11535 scalar_int_mode elements.
11536 (aarch64_vect_float_const_representable_p): Delete.
11537 (aarch64_can_const_movi_rtx_p)
11538 (aarch64_simd_scalar_immediate_valid_for_move)
11539 (aarch64_simd_make_constant): Update call to
11540 aarch64_simd_valid_immediate.
11541 (aarch64_advsimd_valid_immediate_hs): New function.
11542 (aarch64_advsimd_valid_immediate): Likewise.
11543 (aarch64_simd_valid_immediate): Remove mode and inverse
11544 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11545 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11546 and aarch64_float_const_representable_p on the result.
11547 (aarch64_output_simd_mov_immediate): Remove mode argument.
11548 Update call to aarch64_simd_valid_immediate and use of
11549 simd_immediate_info.
11550 (aarch64_output_scalar_simd_mov_immediate): Update call
11553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11554 Alan Hayward <alan.hayward@arm.com>
11555 David Sherwood <david.sherwood@arm.com>
11557 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11558 (mode_nunits): Likewise CONST_MODE_NUNITS.
11559 * machmode.def (ADJUST_NUNITS): Document.
11560 * genmodes.c (mode_data::need_nunits_adj): New field.
11561 (blank_mode): Update accordingly.
11562 (adj_nunits): New variable.
11563 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11565 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11566 listed in adj_nunits.
11567 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11568 listed in adj_nunits. Don't emit case statements for such modes.
11569 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11570 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11571 nothing if adj_nunits is nonnull.
11572 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11573 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11574 (emit_mode_fbit): Update use of print_maybe_const_decl.
11575 (emit_move_size): Likewise. Treat the array as non-const
11577 (emit_mode_adjustments): Handle adj_nunits.
11579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11581 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11582 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11583 (VECTOR_MODES): Use it.
11584 (make_vector_modes): Take the prefix as an argument.
11586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11587 Alan Hayward <alan.hayward@arm.com>
11588 David Sherwood <david.sherwood@arm.com>
11590 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11591 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11592 for MODE_VECTOR_BOOL.
11593 * machmode.def (VECTOR_BOOL_MODE): Document.
11594 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11595 (make_vector_bool_mode): New function.
11596 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11598 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11599 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11601 * stor-layout.c (int_mode_for_mode): Likewise.
11602 * tree.c (build_vector_type_for_mode): Likewise.
11603 * varasm.c (output_constant_pool_2): Likewise.
11604 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11605 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11606 for MODE_VECTOR_BOOL.
11607 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11608 of mode class checks.
11609 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11610 instead of a list of mode class checks.
11611 (expand_vector_scalar_condition): Likewise.
11612 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11615 Alan Hayward <alan.hayward@arm.com>
11616 David Sherwood <david.sherwood@arm.com>
11618 * machmode.h (mode_size): Change from unsigned short to
11620 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11621 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11622 or if measurement_type is not polynomial.
11623 (fixed_size_mode::includes_p): Check for constant-sized modes.
11624 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11625 return a poly_uint16 rather than an unsigned short.
11626 (emit_mode_size): Change the type of mode_size from unsigned short
11627 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11628 (emit_mode_adjustments): Cope with polynomial vector sizes.
11629 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11631 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11633 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11634 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11635 * caller-save.c (setup_save_areas): Likewise.
11636 (replace_reg_with_saved_mem): Likewise.
11637 * calls.c (emit_library_call_value_1): Likewise.
11638 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11639 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11640 (gen_lowpart_for_combine): Likewise.
11641 * convert.c (convert_to_integer_1): Likewise.
11642 * cse.c (equiv_constant, cse_insn): Likewise.
11643 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11644 (cselib_subst_to_values): Likewise.
11645 * dce.c (word_dce_process_block): Likewise.
11646 * df-problems.c (df_word_lr_mark_ref): Likewise.
11647 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11648 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11649 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11650 (rtl_for_decl_location): Likewise.
11651 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11652 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11653 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11654 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11655 (expand_expr_real_1): Likewise.
11656 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11657 (pad_below): Likewise.
11658 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11659 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11660 * ira.c (get_subreg_tracking_sizes): Likewise.
11661 * ira-build.c (ira_create_allocno_objects): Likewise.
11662 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11663 (ira_sort_regnos_for_alter_reg): Likewise.
11664 * ira-costs.c (record_operand_costs): Likewise.
11665 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11666 (resolve_simple_move): Likewise.
11667 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11668 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11669 (lra_constraints): Likewise.
11670 (CONST_POOL_OK_P): Reject variable-sized modes.
11671 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11672 (add_pseudo_to_slot, lra_spill): Likewise.
11673 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11674 * optabs-query.c (get_best_extraction_insn): Likewise.
11675 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11676 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11677 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11678 * recog.c (offsettable_address_addr_space_p): Likewise.
11679 * regcprop.c (maybe_mode_change): Likewise.
11680 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11681 * regrename.c (build_def_use): Likewise.
11682 * regstat.c (dump_reg_info): Likewise.
11683 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11684 (find_reloads, find_reloads_subreg_address): Likewise.
11685 * reload1.c (eliminate_regs_1): Likewise.
11686 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11687 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11688 (simplify_binary_operation_1, simplify_subreg): Likewise.
11689 * targhooks.c (default_function_arg_padding): Likewise.
11690 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11691 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11692 (verify_gimple_assign_ternary): Likewise.
11693 * tree-inline.c (estimate_move_cost): Likewise.
11694 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11695 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11696 (get_address_cost_ainc): Likewise.
11697 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11698 (vect_supportable_dr_alignment): Likewise.
11699 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11700 (vectorizable_reduction): Likewise.
11701 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11702 (vectorizable_operation, vectorizable_load): Likewise.
11703 * tree.c (build_same_sized_truth_vector_type): Likewise.
11704 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11705 * var-tracking.c (emit_note_insn_var_location): Likewise.
11706 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11707 (ADDR_VEC_ALIGN): Likewise.
11709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11710 Alan Hayward <alan.hayward@arm.com>
11711 David Sherwood <david.sherwood@arm.com>
11713 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11715 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11716 or if measurement_type is polynomial.
11717 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11718 * combine.c (make_extraction): Likewise.
11719 * dse.c (find_shift_sequence): Likewise.
11720 * dwarf2out.c (mem_loc_descriptor): Likewise.
11721 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11722 (extract_bit_field, extract_low_bits): Likewise.
11723 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11724 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11725 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11726 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11727 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11728 * reload.c (find_reloads): Likewise.
11729 * reload1.c (alter_reg): Likewise.
11730 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11731 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11732 * tree-if-conv.c (predicate_mem_writes): Likewise.
11733 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11734 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11735 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11736 * valtrack.c (dead_debug_insert_temp): Likewise.
11737 * varasm.c (mergeable_constant_section): Likewise.
11738 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11741 Alan Hayward <alan.hayward@arm.com>
11742 David Sherwood <david.sherwood@arm.com>
11744 * expr.c (expand_assignment): Cope with polynomial mode sizes
11745 when assigning to a CONCAT.
11747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11748 Alan Hayward <alan.hayward@arm.com>
11749 David Sherwood <david.sherwood@arm.com>
11751 * machmode.h (mode_precision): Change from unsigned short to
11753 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11755 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11756 or if measurement_type is not polynomial.
11757 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11758 in which the mode is already known to be a scalar_int_mode.
11759 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11760 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11762 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11763 for GET_MODE_PRECISION.
11764 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11765 for GET_MODE_PRECISION.
11766 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11768 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11769 (expand_field_assignment, make_extraction): Likewise.
11770 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11771 (get_last_value): Likewise.
11772 * convert.c (convert_to_integer_1): Likewise.
11773 * cse.c (cse_insn): Likewise.
11774 * expr.c (expand_expr_real_1): Likewise.
11775 * lra-constraints.c (simplify_operand_subreg): Likewise.
11776 * optabs-query.c (can_atomic_load_p): Likewise.
11777 * optabs.c (expand_atomic_load): Likewise.
11778 (expand_atomic_store): Likewise.
11779 * ree.c (combine_reaching_defs): Likewise.
11780 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11781 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11782 * tree.h (type_has_mode_precision_p): Likewise.
11783 * ubsan.c (instrument_si_overflow): Likewise.
11785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11786 Alan Hayward <alan.hayward@arm.com>
11787 David Sherwood <david.sherwood@arm.com>
11789 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11790 polynomial numbers of units.
11791 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11792 (valid_vector_subparts_p): New function.
11793 (build_vector_type): Remove temporary shim and take the number
11794 of units as a poly_uint64 rather than an int.
11795 (build_opaque_vector_type): Take the number of units as a
11796 poly_uint64 rather than an int.
11797 * tree.c (build_vector_from_ctor): Handle polynomial
11798 TYPE_VECTOR_SUBPARTS.
11799 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11800 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11801 (build_vector_from_val): If the number of units is variable,
11802 use build_vec_duplicate_cst for constant operands and
11803 VEC_DUPLICATE_EXPR otherwise.
11804 (make_vector_type): Remove temporary is_constant ().
11805 (build_vector_type, build_opaque_vector_type): Take the number of
11806 units as a poly_uint64 rather than an int.
11807 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11809 * cfgexpand.c (expand_debug_expr): Likewise.
11810 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11811 (store_constructor, expand_expr_real_1): Likewise.
11812 (const_scalar_mask_from_tree): Likewise.
11813 * fold-const-call.c (fold_const_reduction): Likewise.
11814 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11815 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11816 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11817 (fold_relational_const): Likewise.
11818 (native_interpret_vector): Likewise. Change the size from an
11819 int to an unsigned int.
11820 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11821 TYPE_VECTOR_SUBPARTS.
11822 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11823 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11824 duplicating a non-constant operand into a variable-length vector.
11825 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11826 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11827 * ipa-icf.c (sem_variable::equals): Likewise.
11828 * match.pd: Likewise.
11829 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11830 * print-tree.c (print_node): Likewise.
11831 * stor-layout.c (layout_type): Likewise.
11832 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11833 * tree-cfg.c (verify_gimple_comparison): Likewise.
11834 (verify_gimple_assign_binary): Likewise.
11835 (verify_gimple_assign_ternary): Likewise.
11836 (verify_gimple_assign_single): Likewise.
11837 * tree-pretty-print.c (dump_generic_node): Likewise.
11838 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11839 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11840 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11841 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11842 (vect_shift_permute_load_chain): Likewise.
11843 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11844 (expand_vector_condition, optimize_vector_constructor): Likewise.
11845 (lower_vec_perm, get_compute_type): Likewise.
11846 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11847 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11848 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11849 (vect_recog_mask_conversion_pattern): Likewise.
11850 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11851 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11852 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11853 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11854 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11855 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11856 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11857 (supportable_widening_operation): Likewise.
11858 (supportable_narrowing_operation): Likewise.
11859 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11861 * varasm.c (output_constant): Likewise.
11863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11864 Alan Hayward <alan.hayward@arm.com>
11865 David Sherwood <david.sherwood@arm.com>
11867 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11868 so that both the length == 3 and length != 3 cases set up their
11869 own permute vectors. Add comments explaining why we know the
11870 number of elements is constant.
11871 (vect_permute_load_chain): Likewise.
11873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11874 Alan Hayward <alan.hayward@arm.com>
11875 David Sherwood <david.sherwood@arm.com>
11877 * machmode.h (mode_nunits): Change from unsigned char to
11879 (ONLY_FIXED_SIZE_MODES): New macro.
11880 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11881 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11882 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11884 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11885 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11886 or if measurement_type is not polynomial.
11887 * genmodes.c (ZERO_COEFFS): New macro.
11888 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11890 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11891 Use ZERO_COEFFS when emitting initializers.
11892 * data-streamer.h (bp_pack_poly_value): New function.
11893 (bp_unpack_poly_value): Likewise.
11894 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11895 for GET_MODE_NUNITS.
11896 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11897 for GET_MODE_NUNITS.
11898 * tree.c (make_vector_type): Remove temporary shim and make
11899 the real function take the number of units as a poly_uint64
11900 rather than an int.
11901 (build_vector_type_for_mode): Handle polynomial nunits.
11902 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11903 * emit-rtl.c (const_vec_series_p_1): Likewise.
11904 (gen_rtx_CONST_VECTOR): Likewise.
11905 * fold-const.c (test_vec_duplicate_folding): Likewise.
11906 * genrecog.c (validate_pattern): Likewise.
11907 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11908 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11909 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11910 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11911 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11912 * rtlanal.c (subreg_get_info): Likewise.
11913 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11914 (vect_grouped_load_supported): Likewise.
11915 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11916 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11917 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11918 (simplify_const_unary_operation, simplify_binary_operation_1)
11919 (simplify_const_binary_operation, simplify_ternary_operation)
11920 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11921 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11922 instead of CONST_VECTOR_NUNITS.
11923 * varasm.c (output_constant_pool_2): Likewise.
11924 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11925 explicit-encoded elements in the XVEC for variable-length vectors.
11927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11929 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11932 Alan Hayward <alan.hayward@arm.com>
11933 David Sherwood <david.sherwood@arm.com>
11935 * coretypes.h (fixed_size_mode): Declare.
11936 (fixed_size_mode_pod): New typedef.
11937 * builtins.h (target_builtins::x_apply_args_mode)
11938 (target_builtins::x_apply_result_mode): Change type to
11939 fixed_size_mode_pod.
11940 * builtins.c (apply_args_size, apply_result_size, result_vector)
11941 (expand_builtin_apply_args_1, expand_builtin_apply)
11942 (expand_builtin_return): Update accordingly.
11944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11946 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11947 * cselib.c (cselib_hash_rtx): Likewise.
11948 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11949 CONST_VECTOR encoding.
11951 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11952 Jeff Law <law@redhat.com>
11955 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11956 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11957 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11958 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11961 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11962 explicitly probe *sp in a noreturn function if there were any callee
11963 register saves or frame pointer is needed.
11965 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11968 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11969 BLKmode for ternary, binary or unary expressions.
11972 * var-tracking.c (delete_vta_debug_insn): New inline function.
11973 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11974 insns from get_insns () to NULL instead of each bb separately.
11975 Use delete_vta_debug_insn. No longer static.
11976 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11977 delete_vta_debug_insns callers.
11978 * rtl.h (delete_vta_debug_insns): Declare.
11979 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11980 instead of variable_tracking_main.
11982 2018-01-03 Martin Sebor <msebor@redhat.com>
11984 PR tree-optimization/83603
11985 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11986 arguments past the endof the argument list in functions declared
11987 without a prototype.
11988 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11989 Avoid checking when arguments are null.
11991 2018-01-03 Martin Sebor <msebor@redhat.com>
11994 * doc/extend.texi (attribute const): Fix a typo.
11995 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11996 issuing -Wsuggest-attribute for void functions.
11998 2018-01-03 Martin Sebor <msebor@redhat.com>
12000 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12001 offset_int::from instead of wide_int::to_shwi.
12002 (maybe_diag_overlap): Remove assertion.
12003 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12004 * gimple-ssa-sprintf.c (format_directive): Same.
12005 (parse_directive): Same.
12006 (sprintf_dom_walker::compute_format_length): Same.
12007 (try_substitute_return_value): Same.
12009 2018-01-03 Jeff Law <law@redhat.com>
12011 PR middle-end/83654
12012 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12013 non-constant residual for zero at runtime and avoid probing in
12014 that case. Reorganize code for trailing problem to mirror handling
12017 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12019 PR tree-optimization/83501
12020 * tree-ssa-strlen.c (get_string_cst): New.
12021 (handle_char_store): Call get_string_cst.
12023 2018-01-03 Martin Liska <mliska@suse.cz>
12025 PR tree-optimization/83593
12026 * tree-ssa-strlen.c: Include tree-cfg.h.
12027 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12028 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12029 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12031 (strlen_dom_walker::before_dom_children): Call
12032 gimple_purge_dead_eh_edges. Dump tranformation with details
12034 (strlen_dom_walker::before_dom_children): Update call by adding
12035 new argument cleanup_eh.
12036 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12038 2018-01-03 Martin Liska <mliska@suse.cz>
12041 * cif-code.def (VARIADIC_THUNK): New enum value.
12042 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12045 2018-01-03 Jan Beulich <jbeulich@suse.com>
12047 * sse.md (mov<mode>_internal): Tighten condition for when to use
12048 vmovdqu<ssescalarsize> for TI and OI modes.
12050 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12052 Update copyright years.
12054 2018-01-03 Martin Liska <mliska@suse.cz>
12057 * ipa-visibility.c (function_and_variable_visibility): Skip
12058 functions with noipa attribure.
12060 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12062 * gcc.c (process_command): Update copyright notice dates.
12063 * gcov-dump.c (print_version): Ditto.
12064 * gcov.c (print_version): Ditto.
12065 * gcov-tool.c (print_version): Ditto.
12066 * gengtype.c (create_file): Ditto.
12067 * doc/cpp.texi: Bump @copying's copyright year.
12068 * doc/cppinternals.texi: Ditto.
12069 * doc/gcc.texi: Ditto.
12070 * doc/gccint.texi: Ditto.
12071 * doc/gcov.texi: Ditto.
12072 * doc/install.texi: Ditto.
12073 * doc/invoke.texi: Ditto.
12075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12077 * vector-builder.h (vector_builder::m_full_nelts): Change from
12078 unsigned int to poly_uint64.
12079 (vector_builder::full_nelts): Update prototype accordingly.
12080 (vector_builder::new_vector): Likewise.
12081 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12082 (vector_builder::operator ==): Likewise.
12083 (vector_builder::finalize): Likewise.
12084 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12085 Take the number of elements as a poly_uint64 rather than an
12087 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12088 from unsigned int to poly_uint64.
12089 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12090 (vec_perm_indices::new_vector): Likewise.
12091 (vec_perm_indices::length): Likewise.
12092 (vec_perm_indices::nelts_per_input): Likewise.
12093 (vec_perm_indices::input_nelts): Likewise.
12094 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12095 number of elements per input as a poly_uint64 rather than an
12096 unsigned int. Use the original encoding for variable-length
12097 vectors, rather than clamping each individual element.
12098 For the second and subsequent elements in each pattern,
12099 clamp the step and base before clamping their sum.
12100 (vec_perm_indices::series_p): Handle polynomial element counts.
12101 (vec_perm_indices::all_in_range_p): Likewise.
12102 (vec_perm_indices_to_tree): Likewise.
12103 (vec_perm_indices_to_rtx): Likewise.
12104 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12105 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12106 (tree_vector_builder::new_binary_operation): Handle polynomial
12107 element counts. Return false if we need to know the number
12108 of elements at compile time.
12109 * fold-const.c (fold_vec_perm): Punt if the number of elements
12110 isn't known at compile time.
12112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12114 * vec-perm-indices.h (vec_perm_builder): Change element type
12115 from HOST_WIDE_INT to poly_int64.
12116 (vec_perm_indices::element_type): Update accordingly.
12117 (vec_perm_indices::clamp): Handle polynomial element_types.
12118 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12119 (vec_perm_indices::all_in_range_p): Likewise.
12120 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12122 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12123 polynomial vec_perm_indices element types.
12124 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12125 * fold-const.c (fold_vec_perm): Likewise.
12126 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12127 * tree-vect-generic.c (lower_vec_perm): Likewise.
12128 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12129 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12130 element type to HOST_WIDE_INT.
12132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12133 Alan Hayward <alan.hayward@arm.com>
12134 David Sherwood <david.sherwood@arm.com>
12136 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12137 rather than an int. Use plus_constant.
12138 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12139 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12142 Alan Hayward <alan.hayward@arm.com>
12143 David Sherwood <david.sherwood@arm.com>
12145 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12146 a HOST_WIDE_INT to a poly_int64.
12148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12149 Alan Hayward <alan.hayward@arm.com>
12150 David Sherwood <david.sherwood@arm.com>
12152 * calls.c (load_register_parameters): Cope with polynomial
12153 mode sizes. Require a constant size for BLKmode parameters
12154 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12155 forces a parameter to be padded at the lsb end in order to
12156 fill a complete number of words, require the parameter size
12157 to be ordered wrt UNITS_PER_WORD.
12159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12160 Alan Hayward <alan.hayward@arm.com>
12161 David Sherwood <david.sherwood@arm.com>
12163 * reload1.c (spill_stack_slot_width): Change element type
12164 from unsigned int to poly_uint64_pod.
12165 (alter_reg): Treat mode sizes as polynomial.
12167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12168 Alan Hayward <alan.hayward@arm.com>
12169 David Sherwood <david.sherwood@arm.com>
12171 * reload.c (complex_word_subreg_p): New function.
12172 (reload_inner_reg_of_subreg, push_reload): Use it.
12174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12175 Alan Hayward <alan.hayward@arm.com>
12176 David Sherwood <david.sherwood@arm.com>
12178 * lra-constraints.c (process_alt_operands): Reject matched
12179 operands whose sizes aren't ordered.
12180 (match_reload): Refer to this check here.
12182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12183 Alan Hayward <alan.hayward@arm.com>
12184 David Sherwood <david.sherwood@arm.com>
12186 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12187 that the mode size is in the set {1, 2, 4, 8, 16}.
12189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12190 Alan Hayward <alan.hayward@arm.com>
12191 David Sherwood <david.sherwood@arm.com>
12193 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12194 Use plus_constant instead of gen_rtx_PLUS.
12196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12197 Alan Hayward <alan.hayward@arm.com>
12198 David Sherwood <david.sherwood@arm.com>
12200 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12201 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12202 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12203 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12204 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12205 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12206 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12207 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12208 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12209 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12211 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12212 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12213 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12214 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12215 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12216 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12217 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12218 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12219 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12220 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12222 * expr.c (emit_move_resolve_push): Treat the input and result
12223 of PUSH_ROUNDING as a poly_int64.
12224 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12225 (emit_push_insn): Likewise.
12226 * lra-eliminations.c (mark_not_eliminable): Likewise.
12227 * recog.c (push_operand): Likewise.
12228 * reload1.c (elimination_effects): Likewise.
12229 * rtlanal.c (nonzero_bits1): Likewise.
12230 * calls.c (store_one_arg): Likewise. Require the padding to be
12231 known at compile time.
12233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12234 Alan Hayward <alan.hayward@arm.com>
12235 David Sherwood <david.sherwood@arm.com>
12237 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12238 Use plus_constant instead of gen_rtx_PLUS.
12240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12241 Alan Hayward <alan.hayward@arm.com>
12242 David Sherwood <david.sherwood@arm.com>
12244 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12245 rather than an int.
12247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12248 Alan Hayward <alan.hayward@arm.com>
12249 David Sherwood <david.sherwood@arm.com>
12251 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12252 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12253 via stack temporaries. Treat the mode size as polynomial too.
12255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12256 Alan Hayward <alan.hayward@arm.com>
12257 David Sherwood <david.sherwood@arm.com>
12259 * expr.c (expand_expr_real_2): When handling conversions involving
12260 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12261 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12262 as a poly_uint64 too.
12264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12265 Alan Hayward <alan.hayward@arm.com>
12266 David Sherwood <david.sherwood@arm.com>
12268 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12271 Alan Hayward <alan.hayward@arm.com>
12272 David Sherwood <david.sherwood@arm.com>
12274 * combine.c (can_change_dest_mode): Handle polynomial
12275 REGMODE_NATURAL_SIZE.
12276 * expmed.c (store_bit_field_1): Likewise.
12277 * expr.c (store_constructor): Likewise.
12278 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12279 and polynomial REGMODE_NATURAL_SIZE.
12280 (gen_lowpart_common): Likewise.
12281 * reginfo.c (record_subregs_of_mode): Likewise.
12282 * rtlanal.c (read_modify_subreg_p): Likewise.
12284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12285 Alan Hayward <alan.hayward@arm.com>
12286 David Sherwood <david.sherwood@arm.com>
12288 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12289 numbers of elements.
12291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12292 Alan Hayward <alan.hayward@arm.com>
12293 David Sherwood <david.sherwood@arm.com>
12295 * match.pd: Cope with polynomial numbers of vector elements.
12297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12298 Alan Hayward <alan.hayward@arm.com>
12299 David Sherwood <david.sherwood@arm.com>
12301 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12302 in a POINTER_PLUS_EXPR.
12304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12305 Alan Hayward <alan.hayward@arm.com>
12306 David Sherwood <david.sherwood@arm.com>
12308 * omp-simd-clone.c (simd_clone_subparts): New function.
12309 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12310 (ipa_simd_modify_function_body): Likewise.
12312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12313 Alan Hayward <alan.hayward@arm.com>
12314 David Sherwood <david.sherwood@arm.com>
12316 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12317 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12318 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12319 (expand_vector_condition, vector_element): Likewise.
12320 (subparts_gt): New function.
12321 (get_compute_type): Use subparts_gt.
12322 (count_type_subparts): Delete.
12323 (expand_vector_operations_1): Use subparts_gt instead of
12324 count_type_subparts.
12326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12327 Alan Hayward <alan.hayward@arm.com>
12328 David Sherwood <david.sherwood@arm.com>
12330 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12331 (vect_compile_time_alias): ...this new function. Do the calculation
12332 on poly_ints rather than trees.
12333 (vect_prune_runtime_alias_test_list): Update call accordingly.
12335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12336 Alan Hayward <alan.hayward@arm.com>
12337 David Sherwood <david.sherwood@arm.com>
12339 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12341 (vect_schedule_slp_instance): Likewise.
12343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12344 Alan Hayward <alan.hayward@arm.com>
12345 David Sherwood <david.sherwood@arm.com>
12347 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12348 constant and extern definitions for variable-length vectors.
12349 (vect_get_constant_vectors): Note that the number of units
12350 is known to be constant.
12352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12353 Alan Hayward <alan.hayward@arm.com>
12354 David Sherwood <david.sherwood@arm.com>
12356 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12357 of units as polynomial. Choose between WIDE and NARROW based
12360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12361 Alan Hayward <alan.hayward@arm.com>
12362 David Sherwood <david.sherwood@arm.com>
12364 * tree-vect-stmts.c (simd_clone_subparts): New function.
12365 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12368 Alan Hayward <alan.hayward@arm.com>
12369 David Sherwood <david.sherwood@arm.com>
12371 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12372 vectors as polynomial. Use build_index_vector for
12373 IFN_GOMP_SIMD_LANE.
12375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12376 Alan Hayward <alan.hayward@arm.com>
12377 David Sherwood <david.sherwood@arm.com>
12379 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12380 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12381 for variable-length vectors.
12382 (vectorizable_mask_load_store): Treat the number of units as
12383 polynomial, asserting that it is constant if the condition has
12384 already been enforced.
12385 (vectorizable_store, vectorizable_load): Likewise.
12387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12388 Alan Hayward <alan.hayward@arm.com>
12389 David Sherwood <david.sherwood@arm.com>
12391 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12392 of units as polynomial. Punt if we can't tell at compile time
12393 which vector contains the final result.
12395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12396 Alan Hayward <alan.hayward@arm.com>
12397 David Sherwood <david.sherwood@arm.com>
12399 * tree-vect-loop.c (vectorizable_induction): Treat the number
12400 of units as polynomial. Punt on SLP inductions. Use an integer
12401 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12402 cast of such a series for variable-length floating-point
12405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12406 Alan Hayward <alan.hayward@arm.com>
12407 David Sherwood <david.sherwood@arm.com>
12409 * tree.h (build_index_vector): Declare.
12410 * tree.c (build_index_vector): New function.
12411 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12412 of units as polynomial, forcibly converting it to a constant if
12413 vectorizable_reduction has already enforced the condition.
12414 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12415 to create a {1,2,3,...} vector.
12416 (vectorizable_reduction): Treat the number of units as polynomial.
12417 Choose vectype_in based on the largest scalar element size rather
12418 than the smallest number of units. Enforce the restrictions
12421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12422 Alan Hayward <alan.hayward@arm.com>
12423 David Sherwood <david.sherwood@arm.com>
12425 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12426 number of units as polynomial.
12428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12429 Alan Hayward <alan.hayward@arm.com>
12430 David Sherwood <david.sherwood@arm.com>
12432 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12433 * target.def (autovectorize_vector_sizes): Return the vector sizes
12434 by pointer, using vector_sizes rather than a bitmask.
12435 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12436 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12437 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12439 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12440 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12441 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12442 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12443 * omp-general.c (omp_max_vf): Likewise.
12444 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12445 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12446 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12447 * tree-vect-slp.c (vect_slp_bb): Likewise.
12448 * doc/tm.texi: Regenerate.
12449 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12451 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12452 the vector size as a poly_uint64 rather than an unsigned int.
12453 (current_vector_size): Change from an unsigned int to a poly_uint64.
12454 (get_vectype_for_scalar_type): Update accordingly.
12455 * tree.h (build_truth_vector_type): Take the size and number of
12456 units as a poly_uint64 rather than an unsigned int.
12457 (build_vector_type): Add a temporary overload that takes
12458 the number of units as a poly_uint64 rather than an unsigned int.
12459 * tree.c (make_vector_type): Likewise.
12460 (build_truth_vector_type): Take the number of units as a poly_uint64
12461 rather than an unsigned int.
12463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12464 Alan Hayward <alan.hayward@arm.com>
12465 David Sherwood <david.sherwood@arm.com>
12467 * target.def (get_mask_mode): Take the number of units and length
12468 as poly_uint64s rather than unsigned ints.
12469 * targhooks.h (default_get_mask_mode): Update accordingly.
12470 * targhooks.c (default_get_mask_mode): Likewise.
12471 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12472 * doc/tm.texi: Regenerate.
12474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12475 Alan Hayward <alan.hayward@arm.com>
12476 David Sherwood <david.sherwood@arm.com>
12478 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12479 * omp-general.c (omp_max_vf): Likewise.
12480 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12481 (expand_omp_simd): Handle polynomial safelen.
12482 * omp-low.c (omplow_simd_context): Add a default constructor.
12483 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12484 (lower_rec_simd_input_clauses): Update accordingly.
12485 (lower_rec_input_clauses): Likewise.
12487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12488 Alan Hayward <alan.hayward@arm.com>
12489 David Sherwood <david.sherwood@arm.com>
12491 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12492 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12493 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12494 (vect_analyze_slp_cost): Likewise.
12495 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12496 (vect_model_load_cost): Likewise.
12498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12499 Alan Hayward <alan.hayward@arm.com>
12500 David Sherwood <david.sherwood@arm.com>
12502 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12503 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12504 from an unsigned int * to a poly_uint64_pod *.
12505 (calculate_unrolling_factor): New function.
12506 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12509 Alan Hayward <alan.hayward@arm.com>
12510 David Sherwood <david.sherwood@arm.com>
12512 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12513 from an unsigned int to a poly_uint64.
12514 (_loop_vec_info::slp_unrolling_factor): Likewise.
12515 (_loop_vec_info::vectorization_factor): Change from an int
12517 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12518 (vect_get_num_vectors): New function.
12519 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12520 (vect_get_num_copies): Use vect_get_num_vectors.
12521 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12522 to an unsigned int *.
12523 (vect_analyze_data_refs): Change min_vf from an int * to a
12525 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12526 than an unsigned HOST_WIDE_INT.
12527 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12528 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12529 to an unsigned int *.
12530 (vect_analyze_data_ref_dependences): Likewise.
12531 (vect_compute_data_ref_alignment): Handle polynomial vf.
12532 (vect_enhance_data_refs_alignment): Likewise.
12533 (vect_prune_runtime_alias_test_list): Likewise.
12534 (vect_shift_permute_load_chain): Likewise.
12535 (vect_supportable_dr_alignment): Likewise.
12536 (dependence_distance_ge_vf): Take the vectorization factor as a
12537 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12538 (vect_analyze_data_refs): Change min_vf from an int * to a
12540 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12541 vfm1 as a poly_uint64 rather than an int. Make the same change
12542 for the returned bound_scalar.
12543 (vect_gen_vector_loop_niters): Handle polynomial vf.
12544 (vect_do_peeling): Likewise. Update call to
12545 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12546 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12548 * tree-vect-loop.c (vect_determine_vectorization_factor)
12549 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12550 (vect_get_known_peeling_cost): Likewise.
12551 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12552 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12553 (vect_transform_loop): Likewise. Use the lowest possible VF when
12554 updating the upper bounds of the loop.
12555 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12556 rather than an int.
12557 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12558 polynomial unroll factors.
12559 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12560 (vect_make_slp_decision): Likewise.
12561 (vect_supported_load_permutation_p): Likewise, and polynomial
12563 (vect_analyze_slp_cost): Handle polynomial vf.
12564 (vect_slp_analyze_node_operations): Likewise.
12565 (vect_slp_analyze_bb_1): Likewise.
12566 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12567 than an unsigned HOST_WIDE_INT.
12568 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12569 (vectorizable_load): Handle polynomial vf.
12570 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12572 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12575 Alan Hayward <alan.hayward@arm.com>
12576 David Sherwood <david.sherwood@arm.com>
12578 * match.pd: Handle bit operations involving three constants
12579 and try to fold one pair.
12581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12583 * tree-vect-loop-manip.c: Include gimple-fold.h.
12584 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12585 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12586 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12587 Add a path that uses a step of VF instead of 1, but disable it
12589 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12590 and niters_no_overflow parameters. Update calls to
12591 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12592 Create a new SSA name if the latter choses to use a ste other
12593 than zero, and return it via niters_vector_mult_vf_var.
12594 * tree-vect-loop.c (vect_transform_loop): Update calls to
12595 vect_do_peeling, vect_gen_vector_loop_niters and
12596 slpeel_make_loop_iterate_ntimes.
12597 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12598 (vect_gen_vector_loop_niters): Update declarations after above changes.
12600 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12602 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12603 128-bit round to integer instructions.
12604 (ceil<mode>2): Likewise.
12605 (btrunc<mode>2): Likewise.
12606 (round<mode>2): Likewise.
12608 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12610 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12611 unaligned VSX load/store on P8/P9.
12612 (expand_block_clear): Allow the use of unaligned VSX
12613 load/store on P8/P9.
12615 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12617 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12619 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12620 swap associated with both a load and a store.
12622 2018-01-02 Andrew Waterman <andrew@sifive.com>
12624 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12625 * config/riscv/riscv.md (clear_cache): Use it.
12627 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12629 * web.c: Remove out-of-date comment.
12631 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12633 * expr.c (fixup_args_size_notes): Check that any existing
12634 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12635 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12636 (emit_single_push_insn): ...here.
12638 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12640 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12641 (const_vector_encoded_nelts): New function.
12642 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12643 (const_vector_int_elt, const_vector_elt): Declare.
12644 * emit-rtl.c (const_vector_int_elt_1): New function.
12645 (const_vector_elt): Likewise.
12646 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12647 of CONST_VECTOR_ELT.
12649 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12651 * expr.c: Include rtx-vector-builder.h.
12652 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12653 directly on the tree encoding.
12654 (const_vector_from_tree): Likewise.
12655 * optabs.c: Include rtx-vector-builder.h.
12656 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12657 sequence of "u" values.
12658 * vec-perm-indices.c: Include rtx-vector-builder.h.
12659 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12660 directly on the vec_perm_indices encoding.
12662 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12664 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12665 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12666 * rtx-vector-builder.h: New file.
12667 * rtx-vector-builder.c: Likewise.
12668 * rtl.h (rtx_def::u2): Add a const_vector field.
12669 (CONST_VECTOR_NPATTERNS): New macro.
12670 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12671 (CONST_VECTOR_DUPLICATE_P): Likewise.
12672 (CONST_VECTOR_STEPPED_P): Likewise.
12673 (CONST_VECTOR_ENCODED_ELT): Likewise.
12674 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12675 (unwrap_const_vec_duplicate): Likewise.
12676 (const_vec_series_p): Check for a non-duplicated vector encoding.
12677 Say that the function only returns true for integer vectors.
12678 * emit-rtl.c: Include rtx-vector-builder.h.
12679 (gen_const_vec_duplicate_1): Delete.
12680 (gen_const_vector): Call gen_const_vec_duplicate instead of
12681 gen_const_vec_duplicate_1.
12682 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12683 (gen_const_vec_duplicate): Use rtx_vector_builder.
12684 (gen_const_vec_series): Likewise.
12685 (gen_rtx_CONST_VECTOR): Likewise.
12686 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12687 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12688 Build a new vector rather than modifying a CONST_VECTOR in-place.
12689 (handle_special_swappables): Update call accordingly.
12690 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12691 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12692 Build a new vector rather than modifying a CONST_VECTOR in-place.
12693 (handle_special_swappables): Update call accordingly.
12695 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12697 * simplify-rtx.c (simplify_const_binary_operation): Use
12698 CONST_VECTOR_ELT instead of XVECEXP.
12700 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12702 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12703 the selector elements to be different from the data elements
12704 if the selector is a VECTOR_CST.
12705 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12706 ssizetype for the selector.
12708 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12710 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12711 before testing each element individually.
12712 * tree-vect-generic.c (lower_vec_perm): Likewise.
12714 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12716 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12717 * selftest-run-tests.c (selftest::run_tests): Call it.
12718 * vector-builder.h (vector_builder::operator ==): New function.
12719 (vector_builder::operator !=): Likewise.
12720 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12721 (vec_perm_indices::all_from_input_p): New function.
12722 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12723 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12724 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12725 instead of reading the VECTOR_CST directly. Detect whether both
12726 vector inputs are the same before constructing the vec_perm_indices,
12727 and update the number of inputs argument accordingly. Use the
12728 utility functions added above. Only construct sel2 if we need to.
12730 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12732 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12733 the broadcast of the low byte.
12734 (expand_mult_highpart): Use an explicit encoding for the permutes.
12735 * optabs-query.c (can_mult_highpart_p): Likewise.
12736 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12737 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12738 (vectorizable_bswap): Likewise.
12739 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12740 explicit encoding for the power-of-2 permutes.
12741 (vect_permute_store_chain): Likewise.
12742 (vect_grouped_load_supported): Likewise.
12743 (vect_permute_load_chain): Likewise.
12745 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12747 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12748 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12749 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12750 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12751 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12752 (vect_gen_perm_mask_any): Likewise.
12754 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12756 * int-vector-builder.h: New file.
12757 * vec-perm-indices.h: Include int-vector-builder.h.
12758 (vec_perm_indices): Redefine as an int_vector_builder.
12759 (auto_vec_perm_indices): Delete.
12760 (vec_perm_builder): Redefine as a stand-alone class.
12761 (vec_perm_indices::vec_perm_indices): New function.
12762 (vec_perm_indices::clamp): Likewise.
12763 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12764 (vec_perm_indices::new_vector): New function.
12765 (vec_perm_indices::new_expanded_vector): Update for new
12766 vec_perm_indices class.
12767 (vec_perm_indices::rotate_inputs): New function.
12768 (vec_perm_indices::all_in_range_p): Operate directly on the
12769 encoded form, without computing elided elements.
12770 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12771 encoding. Update for new vec_perm_indices class.
12772 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12773 the given vec_perm_builder.
12774 (expand_vec_perm_var): Update vec_perm_builder constructor.
12775 (expand_mult_highpart): Use vec_perm_builder instead of
12776 auto_vec_perm_indices.
12777 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12778 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12779 or double series encoding as appropriate.
12780 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12781 vec_perm_indices instead of auto_vec_perm_indices.
12782 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12783 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12784 (vect_permute_store_chain): Likewise.
12785 (vect_grouped_load_supported): Likewise.
12786 (vect_permute_load_chain): Likewise.
12787 (vect_shift_permute_load_chain): Likewise.
12788 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12789 (vect_transform_slp_perm_load): Likewise.
12790 (vect_schedule_slp_instance): Likewise.
12791 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12792 (vectorizable_mask_load_store): Likewise.
12793 (vectorizable_bswap): Likewise.
12794 (vectorizable_store): Likewise.
12795 (vectorizable_load): Likewise.
12796 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12797 vec_perm_indices instead of auto_vec_perm_indices. Use
12798 tree_to_vec_perm_builder to read the vector from a tree.
12799 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12800 vec_perm_builder instead of a vec_perm_indices.
12801 (have_whole_vector_shift): Use vec_perm_builder and
12802 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12803 truncation to calc_vec_perm_mask_for_shift.
12804 (vect_create_epilog_for_reduction): Likewise.
12805 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12806 from auto_vec_perm_indices to vec_perm_indices.
12807 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12808 instead of changing individual elements.
12809 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12810 the vector in d.perm.
12811 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12812 from auto_vec_perm_indices to vec_perm_indices.
12813 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12814 instead of changing individual elements.
12815 (arm_vectorize_vec_perm_const): Use new_vector to install
12816 the vector in d.perm.
12817 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12818 Update vec_perm_builder constructor.
12819 (rs6000_expand_interleave): Likewise.
12820 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12821 (rs6000_expand_interleave): Likewise.
12823 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12825 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12826 to qimode could truncate the indices.
12827 * optabs.c (expand_vec_perm_var): Likewise.
12829 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12831 * Makefile.in (OBJS): Add vec-perm-indices.o.
12832 * vec-perm-indices.h: New file.
12833 * vec-perm-indices.c: Likewise.
12834 * target.h (vec_perm_indices): Replace with a forward class
12836 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12837 * optabs.h: Include vec-perm-indices.h.
12838 (expand_vec_perm): Delete.
12839 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12840 (expand_vec_perm_const): Declare.
12841 * target.def (vec_perm_const_ok): Replace with...
12842 (vec_perm_const): ...this new hook.
12843 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12844 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12845 * doc/tm.texi: Regenerate.
12846 * optabs.def (vec_perm_const): Delete.
12847 * doc/md.texi (vec_perm_const): Likewise.
12848 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12849 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12850 expand_vec_perm for constant permutation vectors. Assert that
12851 the mode of variable permutation vectors is the integer equivalent
12852 of the mode that is being permuted.
12853 * optabs-query.h (selector_fits_mode_p): Declare.
12854 * optabs-query.c: Include vec-perm-indices.h.
12855 (selector_fits_mode_p): New function.
12856 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12857 is defined, instead of checking whether the vec_perm_const_optab
12858 exists. Use targetm.vectorize.vec_perm_const instead of
12859 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12860 fit in the vector mode before using a variable permute.
12861 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12862 vec_perm_indices instead of an rtx.
12863 (expand_vec_perm): Replace with...
12864 (expand_vec_perm_const): ...this new function. Take the selector
12865 as a vec_perm_indices rather than an rtx. Also take the mode of
12866 the selector. Update call to shift_amt_for_vec_perm_mask.
12867 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12868 Use vec_perm_indices::new_expanded_vector to expand the original
12869 selector into bytes. Check whether the indices fit in the vector
12870 mode before using a variable permute.
12871 (expand_vec_perm_var): Make global.
12872 (expand_mult_highpart): Use expand_vec_perm_const.
12873 * fold-const.c: Includes vec-perm-indices.h.
12874 * tree-ssa-forwprop.c: Likewise.
12875 * tree-vect-data-refs.c: Likewise.
12876 * tree-vect-generic.c: Likewise.
12877 * tree-vect-loop.c: Likewise.
12878 * tree-vect-slp.c: Likewise.
12879 * tree-vect-stmts.c: Likewise.
12880 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12882 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12883 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12884 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12885 (aarch64_vectorize_vec_perm_const): ...this new function.
12886 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12887 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12888 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12889 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12890 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12891 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12892 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12894 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12895 check for NEON modes.
12896 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12897 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12898 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12899 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12901 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12902 the old VEC_PERM_CONST conditions.
12903 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12904 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12905 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12906 (ia64_vectorize_vec_perm_const_ok): Merge into...
12907 (ia64_vectorize_vec_perm_const): ...this new function.
12908 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12909 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12910 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12911 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12912 * config/mips/mips.c (mips_expand_vec_perm_const)
12913 (mips_vectorize_vec_perm_const_ok): Merge into...
12914 (mips_vectorize_vec_perm_const): ...this new function.
12915 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12916 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12917 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12918 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12919 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12920 (rs6000_expand_vec_perm_const): Delete.
12921 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12923 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12924 (altivec_expand_vec_perm_const_le): Take each operand individually.
12925 Operate on constant selectors rather than rtxes.
12926 (altivec_expand_vec_perm_const): Likewise. Update call to
12927 altivec_expand_vec_perm_const_le.
12928 (rs6000_expand_vec_perm_const): Delete.
12929 (rs6000_vectorize_vec_perm_const_ok): Delete.
12930 (rs6000_vectorize_vec_perm_const): New function.
12931 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12932 an element count and rtx array.
12933 (rs6000_expand_extract_even): Update call accordingly.
12934 (rs6000_expand_interleave): Likewise.
12935 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12936 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12937 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12938 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12939 (rs6000_expand_vec_perm_const): Delete.
12940 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12941 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12942 (altivec_expand_vec_perm_const_le): Take each operand individually.
12943 Operate on constant selectors rather than rtxes.
12944 (altivec_expand_vec_perm_const): Likewise. Update call to
12945 altivec_expand_vec_perm_const_le.
12946 (rs6000_expand_vec_perm_const): Delete.
12947 (rs6000_vectorize_vec_perm_const_ok): Delete.
12948 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12949 reference to the SPE evmerge intructions.
12950 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12951 an element count and rtx array.
12952 (rs6000_expand_extract_even): Update call accordingly.
12953 (rs6000_expand_interleave): Likewise.
12954 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12955 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12957 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12959 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12961 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12962 vector mode and that that mode matches the mode of the data
12964 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12965 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12966 directly using expand_vec_perm_1 when forcing selectors into
12968 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12970 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12972 * optabs-query.h (can_vec_perm_p): Delete.
12973 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12974 * optabs-query.c (can_vec_perm_p): Split into...
12975 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12976 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12977 particular selector is valid.
12978 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12979 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12980 (vect_grouped_load_supported): Likewise.
12981 (vect_shift_permute_load_chain): Likewise.
12982 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12983 (vect_transform_slp_perm_load): Likewise.
12984 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12985 (vectorizable_bswap): Likewise.
12986 (vect_gen_perm_mask_checked): Likewise.
12987 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12988 implementations of variable permutation vectors into account
12989 when deciding which selector to use.
12990 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12991 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12992 with a false third argument.
12993 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12994 to test whether the constant selector is valid and can_vec_perm_var_p
12995 to test whether a variable selector is valid.
12997 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12999 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13000 * optabs-query.c (can_vec_perm_p): Likewise.
13001 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13002 instead of vec_perm_indices.
13003 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13004 (vect_gen_perm_mask_checked): Likewise,
13005 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13006 (vect_gen_perm_mask_checked): Likewise,
13008 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13010 * optabs-query.h (qimode_for_vec_perm): Declare.
13011 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13012 (qimode_for_vec_perm): ...this new function.
13013 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13015 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13017 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13018 does not have a conditional at the top.
13020 2018-01-02 Richard Biener <rguenther@suse.de>
13022 * ipa-inline.c (big_speedup_p): Fix expression.
13024 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13027 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13030 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13034 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13035 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13036 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13037 cond_taken_branch_cost 3->4.
13039 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13041 PR tree-optimization/83581
13042 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13043 TODO_cleanup_cfg if any changes have been made.
13045 PR middle-end/83608
13046 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13047 convert_modes if target mode has the right side, but different mode
13050 PR middle-end/83609
13051 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13052 last argument when extracting from CONCAT. If either from_real or
13053 from_imag is NULL, use expansion through memory. If result is not
13054 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13055 the parts directly to inner mode, if even that fails, use expansion
13058 PR middle-end/83623
13059 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13060 check for bswap in mode rather than HImode and use that in expand_unop
13063 Copyright (C) 2018 Free Software Foundation, Inc.
13065 Copying and distribution of this file, with or without modification,
13066 are permitted in any medium without royalty provided the copyright
13067 notice and this notice are preserved.