1 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
5 code generation for cases where splatting a value is not useful.
6 * simplify-rtx.c (simplify_ternary_operation): Simplify
7 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
11 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
13 * config.gcc: Support "goldmont-plus".
14 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
16 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17 PROCESSOR_GOLDMONT_PLUS.
18 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
19 (processor_target_table): Add "goldmont-plus".
20 (PTA_GOLDMONT_PLUS): Define.
21 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
22 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
23 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
24 (fold_builtin_cpu): Add "goldmont-plus".
25 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
26 (ix86_option_override_internal): Add "goldmont-plus".
27 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
28 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
29 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
30 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
32 2018-05-17 Richard Biener <rguenther@suse.de>
34 PR tree-optimization/85757
35 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
36 remove defs that only feed that PHI from further processing.
38 2018-05-16 Jim Wilson <jimw@sifive.com>
40 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
42 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
44 2018-05-16 Mark Wielaard <mark@klomp.org>
46 * dwarf2out.c (count_index_strings): New function.
47 (output_indirect_strings): Call count_index_strings and generate
48 header for dwarf_version >= 5.
50 2018-05-16 Mark Wielaard <mark@klomp.org>
52 * dwarf2out.c (dwarf_FORM): New function.
53 (set_indirect_string): Use dwarf_FORM.
54 (reset_indirect_string): Likewise.
55 (size_of_die): Likewise.
56 (value_format): Likewise.
57 (output_die): Likewise.
58 (add_skeleton_AT_string): Likewise.
59 (output_macinfo_op): Likewise.
60 (index_string): Likewise.
61 (output_index_string_offset): Likewise.
62 (output_index_string): Likewise.
63 (count_index_strings): Likewise.
65 2018-05-16 Carl Love <cel@us.ibm.com>
67 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
68 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
70 2018-05-16 Martin Jambor <mjambor@suse.cz>
72 * ipa-prop.c (ipa_free_all_edge_args): Remove.
73 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
75 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
77 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
78 (fnma<mode>4): Likewise.
79 (fms<mode>4): Likewise.
80 (fnms<mode>4): Likewise.
81 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
82 (aarch64_fnma<mode>4): Likewise.
83 (aarch64_fms<mode>4): Likewise.
84 (aarch64_fnms<mode>4): Likewise.
85 (aarch64_fnmadd<mode>4): Likewise.
87 2018-05-16 Jason Merrill <jason@redhat.com>
89 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
91 2018-05-16 Richard Biener <rguenther@suse.de>
93 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
94 (dump_stmt_cost): Declare.
95 (add_stmt_cost): Dump cost we add.
96 (add_stmt_costs): New function.
97 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
99 (vect_analyze_stmt): Adjust prototype.
100 (vectorizable_condition): Likewise.
101 (vectorizable_live_operation): Likewise.
102 (vectorizable_reduction): Likewise.
103 (vectorizable_induction): Likewise.
104 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
105 cost vector to pass to vectorizable_ and record afterwards.
106 (vect_model_reduction_cost): Take cost vector argument and adjust.
107 (vect_model_induction_cost): Likewise.
108 (vectorizable_reduction): Likewise.
109 (vectorizable_induction): Likewise.
110 (vectorizable_live_operation): Likewise.
111 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
112 SLP_TREE_NUMBER_OF_VEC_STMTS.
113 (vect_analyze_slp_cost_1): Remove.
114 (vect_analyze_slp_cost): Likewise.
115 (vect_slp_analyze_node_operations): Take visited args and
116 a target cost vector. Avoid processing already visited stmt sets.
117 (vect_slp_analyze_operations): Use a local cost vector to gather
118 costs and register those of non-discarded instances.
119 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
120 (vect_schedule_slp_instance): Remove copying of
121 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
123 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
124 adding cost. Record cost entry location.
125 (vect_prologue_cost_for_slp_op): Function to compute cost of
126 a constant or invariant generated for SLP vect in the prologue,
127 split out from vect_analyze_slp_cost_1.
128 (vect_model_simple_cost): Make static. Adjust for SLP costing.
129 (vect_model_promotion_demotion_cost): Likewise.
130 (vect_model_store_cost): Likewise, make static.
131 (vect_model_load_cost): Likewise.
132 (vectorizable_bswap): Add cost vector arg and adjust.
133 (vectorizable_call): Likewise.
134 (vectorizable_simd_clone_call): Likewise.
135 (vectorizable_conversion): Likewise.
136 (vectorizable_assignment): Likewise.
137 (vectorizable_shift): Likewise.
138 (vectorizable_operation): Likewise.
139 (vectorizable_store): Likewise.
140 (vectorizable_load): Likewise.
141 (vectorizable_condition): Likewise.
142 (vectorizable_comparison): Likewise.
143 (can_vectorize_live_stmts): Likewise.
144 (vect_analyze_stmt): Likewise.
145 (vect_transform_stmt): Adjust calls to vectorizable_*.
146 * tree-vectorizer.c: Include gimple-pretty-print.h.
147 (dump_stmt_cost): New function.
149 2018-05-16 Richard Biener <rguenther@suse.de>
151 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
152 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
153 * tree-ssa-dse.c: Include tree-ssa-loop.h.
154 (check_name): New callback.
155 (dse_classify_store): Track cycles via a visited bitmap of PHI
156 defs and simplify handling of in-loop and across loop dead stores
157 and properly fail for loop-variant refs. Handle byte-tracking with
158 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
161 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
163 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
164 (vect_get_mask_type_for_stmt): Likewise.
165 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
167 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
168 to determine the statement's vector type and the vector type that
169 should be used for calculating nunits. Deal with cases in which
170 the type has to be deferred.
171 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
172 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
173 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
174 (vect_determine_vf_for_stmt): New functions, split out from...
175 (vect_determine_vectorization_factor): ...here.
176 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
177 (vect_get_mask_type_for_stmt): New functions, split out from
178 vect_determine_vectorization_factor.
180 2018-05-16 Richard Biener <rguenther@suse.de>
182 * tree-cfg.c (verify_gimple_assign_ternary): Properly
183 verify the [VEC_]COND_EXPR embedded comparison.
185 2018-05-15 Martin Sebor <msebor@redhat.com>
187 PR tree-optimization/85753
188 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
189 RECORD_TYPE in addition to ARRAY_TYPE.
191 2018-05-15 Martin Sebor <msebor@redhat.com>
194 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
196 2018-05-15 Richard Biener <rguenther@suse.de>
198 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
199 add by_clobber_p one. Change algorithm to collect all defs
200 representing uses we need to walk and try reducing them to
201 a single one before failing.
202 (dse_dom_walker::dse_optimize_stmt): Adjust.
204 2018-05-13 Mark Wielaard <mark@klomp.org>
206 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
207 (size_of_loc_descr): Likewise.
208 (output_loc_operands): Likewise.
209 (output_loc_operands_raw): Likewise.
210 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
211 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
212 (hash_loc_operands): Likewise.
213 (compare_loc_operands): Likewise.
215 2018-05-14 Mark Wielaard <mark@klomp.org>
217 * dwarf2out.c (count_index_addrs): New function.
218 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
220 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
222 PR tree-optimization/83648
223 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
224 return value as malloc candidate.
226 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
229 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
230 as true in call to suggest_attribute.
232 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
234 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
237 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
239 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
241 (aarch64_crypto_aesd_fused): Likewise.
243 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
245 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
246 (movsi_aarch64): Likewise.
247 (load_pairsi): Likewise.
248 (load_pairdi): Likewise.
249 (store_pairsi): Likewise.
250 (store_pairdi): Likewise.
251 (load_pairsf): Likewise.
252 (load_pairdf): Likewise.
253 (store_pairsf): Likewise.
254 (store_pairdf): Likewise.
255 (zero_extend): Likewise.
256 (trunc): Swap alternatives.
257 (fcvt_target): Add '?' to prefer w over r.
259 2018-05-14 Jakub Jelinek <jakub@redhat.com>
262 * config/i386/i386.md: Disallow non-commutative arithmetics in
263 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
264 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
265 in the peephole2 before it.
267 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
269 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
270 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
271 (ix86_handle_option): Handle -mcldemote.
272 * config.gcc: New header.
273 * config/i386/cldemoteintrin.h: New file.
274 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
275 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
277 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
278 OPTION_MASK_ISA_CLDEMOTE.
279 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
280 (ix86_valid_target_attribute_inner_p): Ditto.
281 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
282 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
283 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
284 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
285 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
287 * config/i386/i386.opt: Add -mcldemote.
288 * config/i386/x86intrin.h: New header.
289 * doc/invoke.texi: Add -mcldemote.
291 2018-05-14 Richard Biener <rguenther@suse.de>
293 * doc/match-and-simplify.texi: Adjust :s documentation.
295 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
297 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
298 intended memcpy size.
299 (REORDER_45): Likewise.
301 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
304 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
305 * vec.c (qsort_chk): Use gcc_qsort.
306 * Makefile.in (OBJS-libcommon): Add sort.o.
307 (build/sort.o): New target. Use it...
308 (BUILD_RTL): ... here, and...
309 (build/gencfn-macros): ... here, and...
310 (build/genmatch): ... here.
312 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
313 Chung-Ju Wu <jasonwucj@gmail.com>
315 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
316 * config/nds32/nds32-graywolf.md: New file.
317 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
318 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
320 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
321 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
322 * config/nds32/nds32.md (pipeline_model): Add graywolf.
323 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
324 * config/nds32/pipelines.md: Include n15 settings.
326 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
327 Chung-Ju Wu <jasonwucj@gmail.com>
329 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
330 * config/nds32/nds32-n13.md: New file.
331 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
332 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
334 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
335 * config/nds32/nds32.md (pipeline_model): Add n13.
336 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
337 * config/nds32/pipelines.md: Include n13 settings.
339 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
340 Chung-Ju Wu <jasonwucj@gmail.com>
342 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
343 * config/nds32/nds32-n10.md: New file.
344 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
345 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
347 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
348 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
349 * config/nds32/nds32.md (pipeline_model): Add n10.
350 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
351 * config/nds32/pipelines.md: Include n10 settings.
353 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
354 Kito Cheng <kito.cheng@gmail.com>
355 Chung-Ju Wu <jasonwucj@gmail.com>
357 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
358 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
359 Add enum values for DSP extension instructions.
360 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
362 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
363 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
365 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
366 * config/nds32/nds32-dspext.md: New file for DSP implementation.
367 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
368 * config/nds32/nds32-intrinsic.md: Likewise.
369 * config/nds32/nds32_intrinsic.h: Likewise.
370 * config/nds32/nds32-md-auxiliary.c: Likewise.
371 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
372 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
373 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
374 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
375 * config/nds32/nds32-protos.h: New declarations for DSP extension.
376 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
377 TYPE_DMAC in switch statement.
378 * config/nds32/nds32.c: New checking and implementation for DSP
379 extension instructions.
380 * config/nds32/nds32.h: Likewise.
381 * config/nds32/nds32.md: Likewise.
382 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
383 * config/nds32/predicates.md: Implement new predicates for DSP
386 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
388 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
389 Reformat alternatives and attributes so it is easier to identify
390 which constraints/attributes go with which instruction.
391 (mov<mode>_hardfloat32, FMOVE64): Likewise.
392 (mov<mode>_softfloat32, FMOVE64): Likewise.
393 (mov<mode>_hardfloat64, FMOVE64): Likewise.
394 (mov<mode>_softfloat64, FMOVE64): Likewise.
396 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
398 * doc/extend.texi (PowerPC Built-in Functions): Rename this
400 (Basic PowerPC Built-in Functions): The new name of the
401 subsection previously known as "PowerPC Built-in Functions".
402 (Basic PowerPC Built-in Functions Available on all Configurations):
404 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
405 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
406 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
407 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
409 2018-05-11 Martin Jambor <mjambor@suse.cz>
412 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
415 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
418 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
420 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
422 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
423 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
424 (ix86_handle_option): Handle -mwaitpkg.
425 * config.gcc: New header.
426 * config/i386/cpuid.h (bit_WAITPKG): New bit.
427 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
428 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
430 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
431 OPTION_MASK_ISA_WAITPKG.
432 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
433 (ix86_option_override_internal): Add PTA_WAITPKG.
434 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
435 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
437 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
438 __builtin_ia32_umwait and __builtin_ia32_tpause.
439 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
440 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
441 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
442 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
443 UNSPECV_TPAUSE): New.
444 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
445 * config/i386/i386.opt: Add -mwaitpkg.
446 * config/i386/waitpkgintrin.h: New file.
447 * config/i386/x86intrin.h: New header.
448 * doc/invoke.texi: Add -mwaitpkg.
450 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
453 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
455 (cortex-m0): Use armv6s-m isa.
456 (cortex-m0plus): Likewise.
457 (cortex-m1): Likewise.
458 (cortex-m0.small-multiply): Likewise.
459 (cortex-m0plus.small-multiply): Likewise.
460 (cortex-m1.small-multiply): Likewise.
462 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
463 Jakub Jelinek <jakub@redhat.com>
465 PR tree-optimization/85692
466 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
467 source permute as well.
469 2018-05-11 Martin Liska <mliska@suse.cz>
472 * doc/extend.texi: Document LLVM style format for no_sanitize
475 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
477 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
478 mode_supports_vsx_dform_quad to mode_supports_dq_form.
479 (mode_supports_vsx_dform_quad): Likewise.
480 (mode_supports_vmx_dform): Move these functions to be next to the
481 other mode_supports functions.
482 (mode_supports_dq_form): Likewise.
483 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
484 mode_supports_dq_form.
485 (reg_offset_addressing_ok_p): Likewise.
486 (offsettable_ok_by_alignment): Likewise.
487 (rs6000_legitimate_offset_address_p): Likewise.
488 (legitimate_lo_sum_address_p): Likewise.
489 (rs6000_legitimize_address): Likewise.
490 (rs6000_legitimize_reload_address): Likewise.
491 (rs6000_secondary_reload_inner): Likewise.
492 (rs6000_preferred_reload_class): Likewise.
493 (rs6000_output_move_128bit): Likewise.
495 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
497 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
498 Generate SImode target register for null target.
499 <case IX86_BUILTIN_XGETBV>: Ditto.
500 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
501 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
503 2018-05-10 Carl Love <cel@us.ibm.com>
505 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
506 dcbtt and dcbtstt if operands[2] is 0.
508 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
511 * config/i386/sse.md (usadv64qi): New expander.
513 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
515 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
516 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
517 -maltivec=be support.
518 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
519 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
520 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
521 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
522 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
523 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
524 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
525 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
526 altivec_vsumsws): Adjust.
527 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
528 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
529 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
531 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
532 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
533 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
534 (altivec_lve<VI_char>x): Delete expand.
535 (*altivec_lve<VI_char>x_internal): Rename to...
536 (altivec_lve<VI_char>x): ... this.
537 (altivec_lvxl_<mode>): Delete expand.
538 (*altivec_lvxl_<mode>_internal): Rename to ...
539 (altivec_lvxl_<mode>): ... this.
540 (altivec_stvxl_<mode>): Delete expand.
541 (*altivec_stvxl_<mode>_internal): Rename to ...
542 (altivec_stvxl_<mode>): ... this.
543 (altivec_stve<VI_char>x): Delete expand.
544 (*altivec_stve<VI_char>x_internal): Rename to ...
545 (altivec_stve<VI_char>x): ... this.
546 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
547 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
548 reduc_plus_scal_<mode>): Adjust.
549 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
551 (rs6000_cpu_cpp_builtins): Adjust.
552 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
553 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
554 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
555 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
556 -maltivec=be support.
557 (rs6000_split_vec_extract_var): Adjust.
558 (rs6000_split_v4si_init): Adjust.
559 (swap_selector_for_mode): Delete.
560 (altivec_expand_lvx_be, altivec_expand_stvx_be,
561 altivec_expand_stvex_be): Delete.
562 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
563 -maltivec=be support.
564 (rs6000_gimple_fold_builtin): Ditto.
565 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
567 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
568 (TARGET_DIRECT_MOVE_64BIT): Adjust.
569 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
570 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
571 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
572 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
573 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
574 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
575 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
576 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
577 anonymous split): Adjust.
578 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
579 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
581 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
583 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
584 when --with-gxx-include-dir is also specified.
585 * configure: Regenerate.
587 2018-05-09 Jim Wilson <jimw@sifive.com>
590 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
591 * config/riscv/t-withmultilib: New.
592 * config/riscv/withmultilib.h: New.
593 * doc/install.texi: Document RISC-V --with-multilib-list support.
595 2018-05-09 Richard Biener <rguenther@suse.de>
597 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
599 (vect_bb_vectorization_profitable_p): Adjust. Compute
600 actual scalar cost using the cost vector and the add_stmt_cost
603 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
605 PR rtl-optimization/85645
606 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
607 in the REG_CFA_REGISTER note for LR, don't leave it empty.
609 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
611 PR rtl-optimization/85645
612 * shrink-wrap.c (spread_components): Return a boolean saying if
613 anything was changed.
614 (try_shrink_wrapping_separate): Iterate spread_components until
615 nothing changes anymore.
617 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
619 PR rtl-optimization/85645
620 * regrename.c (build_def_use): Also kill the chains that include the
621 destination of a REG_CFA_REGISTER note.
623 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
625 PR rtl-optimization/85645
626 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
627 insn that has a REG_CFA_REGISTER note.
629 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
631 * cfgexpand.c (expand_clobber): New function.
632 (expand_gimple_stmt_1): Use it.
633 * tree-vect-stmts.c (vect_clobber_variable): New function,
635 (vectorizable_simd_clone_call): ...here.
636 (vectorizable_store): Emit a clobber either side of an
637 IFN_STORE_LANES sequence.
638 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
640 2018-05-09 Tom de Vries <tom@codesourcery.com>
643 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
644 (define_insn "trap_if_false"): Add exit after trap.
646 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
648 PR rtl-optimization/85638
649 * bb-reorder.c: Include common/common-target.h.
650 (create_forwarder_block): New function extracted from...
651 (fix_up_crossing_landing_pad): ...here. Rename into...
652 (dw2_fix_up_crossing_landing_pad): ...this.
653 (sjlj_fix_up_crossing_landing_pad): New function.
654 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
655 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
656 from both partitions and exit the loop after one iteration.
658 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
661 * doc/extend.texi (PowerPC Built-in Functions): Rename this
663 (Basic PowerPC Built-in Functions): The new name of the
664 subsection previously known as "PowerPC Built-in Functions".
665 (Basic PowerPC Built-in Functions Available on all Configurations):
667 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
669 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
671 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
673 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
676 2018-05-08 Jim Wilson <jimw@sifive.com>
678 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
679 (LD_EMUL_SUFFIX): New.
682 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
684 * doc/extend.texi (PowerPC Built-in Functions): Rename this
686 (Basic PowerPC Built-in Functions): The new name of the
687 subsection previously known as "PowerPC Built-in Functions".
688 (Basic PowerPC Built-in Functions Available on all Configurations):
690 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
692 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
694 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
696 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
699 2018-05-08 Jakub Jelinek <jakub@redhat.com>
702 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
703 after cmpelim optimization.
705 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
707 * config.gcc: Support "goldmont".
708 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
709 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
711 * config/i386/i386.c (m_GOLDMONT): Define.
712 (processor_target_table): Add "goldmont".
713 (PTA_GOLDMONT): Define.
714 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
715 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
716 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
717 (fold_builtin_cpu): Add "goldmont".
718 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
719 (ix86_option_override_internal): Add "goldmont".
720 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
721 (processor_type): Add PROCESSOR_GOLDMONT.
722 * config/i386/i386.md: Add CPU "glm".
723 * config/i386/glm.md: New file.
724 * config/i386/x86-tune.def: Add m_GOLDMONT.
725 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
727 2018-05-08 Jakub Jelinek <jakub@redhat.com>
730 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
732 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
733 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
734 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
737 * config/i386/i386.c (ix86_fold_builtin): Handle
738 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
741 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
742 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
744 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
747 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
748 (check_arch): Likewise.
749 (check_fpu): Return the result rather than printing it.
750 (end arch): Fix operator precedence.
752 (END): Print the result from check_fpu.
754 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
755 Alan Hayward <alan.hayward@arm.com>
756 David Sherwood <david.sherwood@arm.com>
758 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
759 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
760 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
761 (*fcmuo<mode>_and): New patterns.
763 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
765 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
766 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
767 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
768 (cmp_op, sve_imm_con): New code attributes.
769 (SVE_COND_INT_CMP, imm_con): Delete.
770 (cmp_op): Remove above unspecs from int attribute.
771 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
773 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
774 comparison-specific unspecs.
775 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
776 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
777 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
778 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
779 (*vec_fcm<cmp_op><mode>): Rename to...
780 (*fcm<cmp_op><mode>): ...this and adjust likewise.
781 (*vec_fcmuo<mode>): Rename to...
782 (*fcmuo<mode>): ...this and adjust likewise.
783 (*pred_fcm<cmp_op><mode>): New pattern.
784 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
785 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
787 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
789 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
790 (aarch64_emit_sve_predicated_cond): New function.
791 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
792 (aarch64_emit_unspec_cond_or): Replace with...
793 (aarch64_emit_sve_or_conds): ...this new function. Use
794 aarch64_emit_sve_ptrue_op for the individual comparisons and
795 aarch64_emit_binop to OR them together.
796 (aarch64_emit_inverted_unspec_cond): Replace with...
797 (aarch64_emit_sve_inverted_cond): ...this new function. Use
798 aarch64_emit_sve_ptrue_op for the comparison and
799 aarch64_emit_unop to invert the result.
800 (aarch64_expand_sve_vec_cmp_float): Update after the above
801 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
803 2018-05-07 Nathan Sidwell <nathan@acm.org>
805 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
806 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
807 (Backwards Compatibility): Likewise.
809 2018-05-07 Luis Machado <luis.machado@linaro.org>
813 2018-05-07 Luis Machado <luis.machado@linaro.org>
815 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
816 <prefetch_dynamic_strides>: New const bool field.
817 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
818 prefetch_dynamic_strides.
819 (exynosm1_prefetch_tune): Likewise.
820 (thunderxt88_prefetch_tune): Likewise.
821 (thunderx_prefetch_tune): Likewise.
822 (thunderx2t99_prefetch_tune): Likewise.
823 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
824 (aarch64_override_options_internal): Update to set
825 PARAM_PREFETCH_DYNAMIC_STRIDES.
826 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
827 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
828 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
829 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
830 prefetch-dynamic-strides setting.
832 2018-05-07 Luis Machado <luis.machado@linaro.org>
834 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
835 <minimum_stride>: New const int field.
836 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
837 minimum_stride field.
838 (exynosm1_prefetch_tune): Likewise.
839 (thunderxt88_prefetch_tune): Likewise.
840 (thunderx_prefetch_tune): Likewise.
841 (thunderx2t99_prefetch_tune): Likewise.
842 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
843 (aarch64_override_options_internal): Update to set
844 PARAM_PREFETCH_MINIMUM_STRIDE.
845 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
846 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
847 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
848 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
849 stride is constant and is below the minimum stride threshold.
851 2018-05-07 Luis Machado <luis.machado@linaro.org>
853 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
856 2018-05-07 Luis Machado <luis.machado@linaro.org>
858 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
859 <prefetch_dynamic_strides>: New const bool field.
860 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
861 prefetch_dynamic_strides.
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 prefetch_dynamic_strides to false.
867 (aarch64_override_options_internal): Update to set
868 PARAM_PREFETCH_DYNAMIC_STRIDES.
869 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
870 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
871 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
872 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
873 prefetch-dynamic-strides setting.
875 2018-05-07 Luis Machado <luis.machado@linaro.org>
877 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
878 <minimum_stride>: New const int field.
879 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
880 minimum_stride field.
881 (exynosm1_prefetch_tune): Likewise.
882 (thunderxt88_prefetch_tune): Likewise.
883 (thunderx_prefetch_tune): Likewise.
884 (thunderx2t99_prefetch_tune): Likewise.
885 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
886 (aarch64_override_options_internal): Update to set
887 PARAM_PREFETCH_MINIMUM_STRIDE.
888 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
889 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
890 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
891 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
892 stride is constant and is below the minimum stride threshold.
894 2018-05-06 Jakub Jelinek <jakub@redhat.com>
897 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
898 the type is addressable. Don't force op into register if it has
901 2018-05-05 Roland McGrath <mcgrathr@google.com>
904 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
905 any section for which we don't know a specific type it should have,
906 regardless of name. Previously this was done only for the exact
907 names ".init_array", ".fini_array", and ".preinit_array".
908 (default_elf_asm_named_section): Add comment about
909 relationship with default_section_type_flags and SECTION_NOTYPE.
910 (get_section): Don't consider it a type conflict if one side has
911 SECTION_NOTYPE and the other doesn't, as long as neither has the
912 SECTION_BSS et al used in the default_section_type_flags logic.
914 2018-05-05 Tom de Vries <tom@codesourcery.com>
917 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
918 (workaround_barsyncs): New function.
919 (nvptx_reorg): Use workaround_barsyncs.
920 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
921 (define_expand "nvptx_membar_cta"): New define_expand.
922 (define_insn "*nvptx_membar_cta"): New insn.
924 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
926 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
927 To improve optimization opportunities.
928 * builtin-types.def: The new needed builtin types for the above.
930 2018-05-04 Richard Biener <rguenther@suse.de>
932 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
933 * gimple-ssa-store-merging.c
934 (imm_store_chain_info::output_merged_store): Remove redundant create,
935 release split_store vector contents on failure.
936 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
937 scalar stmt vector on cache hit.
939 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
941 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
943 * config.gcc (powerpc-xilinx-eabi*): Remove.
944 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
946 (fusion_addis_mem_combo_load): Ditto.
947 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
949 (rs6000_cpu_cpp_builtins): Ditto.
950 * config/rs6000/rs6000-linux.c
951 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
952 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
953 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
955 (rs6000_setup_reg_addr_masks): Ditto.
956 (rs6000_init_hard_regno_mode_ok): Ditto.
957 (rs6000_option_override_internal): Ditto.
958 (legitimate_lo_sum_address_p): Ditto.
959 (rs6000_legitimize_address): Ditto.
960 (rs6000_legitimize_reload_address): Ditto.
961 (rs6000_legitimate_address_p): Ditto.
962 (abi_v4_pass_in_fpr): Ditto.
963 (setup_incoming_varargs): Ditto.
964 (rs6000_gimplify_va_arg): Ditto.
965 (rs6000_split_multireg_move): Ditto.
966 (rs6000_savres_strategy): Ditto.
967 (rs6000_emit_prologue_components): Ditto.
968 (rs6000_emit_epilogue_components): Ditto.
969 (rs6000_emit_prologue): Ditto.
970 (rs6000_emit_epilogue): Ditto.
971 (rs6000_elf_file_end): Ditto.
972 (rs6000_function_value): Ditto.
973 (rs6000_libcall_value): Ditto.
974 * config/rs6000/rs6000.h: Ditto.
975 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
976 (TARGET_MINMAX): ... this. New.
977 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
978 * config/rs6000/rs6000.md: Remove Xilinx FP support.
979 (*movsi_internal1_single): Delete.
980 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
981 mfpu=, mxilinx-fpu): Delete.
982 * config/rs6000/singlefp.h: Delete.
983 * config/rs6000/sysv4.h: Remove Xilinx FP support.
984 * config/rs6000/t-rs6000: Ditto.
985 * config/rs6000/t-xilinx: Delete.
986 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
987 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
988 (VStype_simple): Delete.
989 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
990 * config/rs6000/xfpu.h: Delete.
991 * config/rs6000/xfpu.md: Delete.
992 * config/rs6000/xilinx.h: Delete.
993 * config/rs6000/xilinx.opt: Delete.
994 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
995 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
997 2018-05-04 Tom de Vries <tom@codesourcery.com>
1000 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1003 2018-05-04 Richard Biener <rguenther@suse.de>
1006 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1007 (expand_complex_div_wide): Likewise.
1008 (expand_complex_operations_1): Likewise.
1009 (expand_complex_libcall): Preserve EH info of the original stmt.
1010 (tree_lower_complex): Handle removed blocks.
1011 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1012 on complex multiplication and division libcall builtins.
1014 2018-05-04 Richard Biener <rguenther@suse.de>
1017 * fold-const.c (negate_expr_p): Restrict negation of operand
1018 zero of a division to when we know that can happen without
1020 (fold_negate_expr_1): Likewise.
1022 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1025 * real.h (real_nextafter): Declare.
1026 * real.c (real_nextafter): New function.
1027 * fold-const-call.c (fold_const_nextafter): New function.
1028 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1029 CASE_CFN_NEXTTOWARD.
1030 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1031 even when arg1_mode is different from arg0_mode.
1033 2018-05-03 Nathan Sidwell <nathan@acm.org>
1035 * doc/extend.texi (Deprecated Features): Remove
1037 (Backwards Compatibility): Likewise.
1038 * doc/invoke.texi (C++ Language Options): Likewise.
1039 (C++ Dialect Options): Likewise.
1041 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1044 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1045 _mm512_mask_mullox_epi64): New intrinsics.
1047 2018-05-03 Tom de Vries <tom@codesourcery.com>
1050 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1051 dump files): Add offload-tree.
1053 2018-05-03 Richard Biener <rguenther@suse.de>
1055 PR tree-optimization/85615
1056 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1057 to loops not nested in BBs loop father to avoid creating multi-entry
1060 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1062 PR tree-optimization/70291
1063 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1064 arguments. Change return type to tree. Emit libcall as a new
1065 statement rather than replacing existing one when inplace_p is true.
1066 (expand_complex_multiplication_components): New function.
1067 (expand_complex_multiplication): Expand floating-point complex
1068 multiplication using the above.
1069 (expand_complex_division): Rename inner_type parameter to type.
1070 Update expand_complex_libcall call-site.
1071 (expand_complex_operations_1): Update expand_complex_multiplication
1072 and expand_complex_division call-sites.
1074 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1077 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1078 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1079 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1080 the highest significant bit of the shift count mask is clear. In
1081 check whether and[sq]i3 is needed verify that all significant bits
1082 of the shift count other than the highest are set.
1084 2018-05-02 Tom de Vries <tom@codesourcery.com>
1087 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1088 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1089 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1090 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1091 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1092 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1093 __builtin_goacc_parlevel_size.
1095 2018-05-02 Richard Biener <rguenther@suse.de>
1097 PR tree-optimization/85597
1098 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1099 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1102 2018-05-02 Tom de Vries <tom@codesourcery.com>
1105 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1106 dump files): Add ltrans-tree.
1108 2018-05-02 Tom de Vries <tom@codesourcery.com>
1111 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1112 dump files): Add wpa-ipa.
1114 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1116 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1117 powerpc*-*-linux*paired* target.
1118 * config/rs6000/750cl.h: Delete.
1119 * config/rs6000/paired.h: Delete.
1120 * config/rs6000/paired.md: Delete.
1121 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1123 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1124 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1125 comment. Remove paired float support.
1126 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1127 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1129 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1130 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1132 * config/rs6000/rs6000.c: Remove paired float support.
1133 (paired_expand_vector_init, paired_expand_vector_move,
1134 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1135 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1136 paired_expand_builtin, paired_expand_predicate_builtin,
1137 paired_init_builtins): Delete.
1138 * config/rs6000/rs6000.h: Remove paired float support.
1139 * config/rs6000/rs6000.md: Remove paired float support.
1140 (move_from_CR_ov_bit): Delete.
1141 * config/rs6000/rs6000.opt (mpaired): Delete.
1142 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1143 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1145 2018-05-02 Richard Biener <rguenther@suse.de>
1148 * gimplify.c (gimplify_save_expr): When in SSA form allow
1149 SAVE_EXPRs to compute to SSA vars.
1151 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1154 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1155 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1156 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1157 clobber operands[2], instead use a new pseudo. Formatting fixes.
1159 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1161 PR tree-optimization/85586
1162 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1163 exit early for statements in the same group if the accesses are
1166 2018-05-02 Tom de Vries <tom@codesourcery.com>
1169 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1172 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1174 PR tree-optimization/85143
1175 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1177 2018-05-01 Tom de Vries <tom@codesourcery.com>
1180 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1181 not found" error message.
1183 2018-05-01 Tom de Vries <tom@codesourcery.com>
1186 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1187 * vec.c (test_ordered_remove_if): New function.
1188 (vec_c_tests): Call test_ordered_remove_if.
1189 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1190 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1191 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1192 VEC_ORDERED_REMOVE_IF.
1194 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1196 PR tree-optimization/82665
1197 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1198 pointer subtraction where arguments come from a memchr call.
1200 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1202 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1203 --push-state --as-needed and --pop-state instead of --as-needed and
1204 --no-as-needed if ld supports it.
1205 * configure: Regenerated.
1208 * doc/install.texi2html: Replace _002d with - and _002a with * in
1209 generated html files using sed.
1211 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1214 * gcc-rich-location.c (blank_line_before_p): New function.
1215 (use_new_line): New function.
1216 (gcc_rich_location::add_fixit_insert_formatted): New function.
1217 * gcc-rich-location.h
1218 (gcc_rich_location::add_fixit_insert_formatted): New function.
1220 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1222 * selftest.c (assert_streq): Rename "expected" and "actual" to
1223 "val1" and "val2". Extend NULL-handling to cover both inputs
1224 symmetrically, while still requiring both to be non-NULL for a pass.
1225 * selftest.h (assert_streq): Rename "expected" and "actual" to
1227 (ASSERT_EQ): Likewise.
1228 (ASSERT_EQ_AT): Likewise.
1229 (ASSERT_KNOWN_EQ): Likewise.
1230 (ASSERT_KNOWN_EQ_AT): Likewise.
1231 (ASSERT_NE): Likewise.
1232 (ASSERT_MAYBE_NE): Likewise.
1233 (ASSERT_MAYBE_NE_AT): Likewise.
1234 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1235 the assertion to pass.
1236 (ASSERT_STREQ_AT): Likewise.
1238 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1240 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1241 interaction with -pie.
1243 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1245 * selftest.h: Fix alphabetization of per-source-file selftest
1248 2018-04-30 Jason Merrill <jason@redhat.com>
1250 PR c++/61982 - dead stores to destroyed objects.
1251 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1254 2018-04-30 Jason Merrill <jason@redhat.com>
1256 * tree.c (build_clobber): New.
1257 * tree.h: Declare it.
1258 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1260 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1262 * diagnostic-show-locus.c (layout::layout): Update for
1263 location_get_source_line returning a char_span.
1264 (struct char_span): Move to input.h.
1265 (struct correction): Update for fields in char_span becoming
1267 (struct source_line): Update for location_get_source_line
1268 returning a char_span.
1269 (layout::print_line): Likewise.
1270 * edit-context.c (edited_file::print_content): Likewise.
1271 (edited_file::print_diff_hunk): Likewise.
1272 (edited_file::print_run_of_changed_lines): Likewise.
1273 (edited_file::get_num_lines): Likewise.
1274 (edited_line::edited_line): Likewise.
1275 * final.c (asm_show_source): Likewise.
1276 * input.c (location_get_source_line): Convert return type
1277 from const char * to char_span, losing the final "line_len"
1279 (dump_location_info): Update for the above.
1280 (get_substring_ranges_for_loc): Likewise. Use a char_span
1281 when handling the literal within the line.
1282 (test_reading_source_line): Update for location_get_source_line
1283 returning a char_span.
1284 * input.h (class char_span): Move here from
1285 diagnostic-show-locus.c, converting from a struct to a class.
1286 Make data members private.
1287 (char_span::operator bool): New.
1288 (char_span::length): New.
1289 (char_span::get_buffer): New.
1290 (char_span::operator[]): New.
1291 (char_span::subspan): Make const.
1292 (char_span::xstrdup): New.
1293 (location_get_source_line): Convert return type from const char *
1294 to char_span, losing the final "line_size" param.
1296 2018-04-30 Jan Hubicka <jh@suse.cz>
1298 * lto-wrapper.c (ltrans_priorities): New static var.
1299 (cmp_priority): New.
1300 (run_gcc): Read priorities and if doing parallel build order
1301 the Makefile by them.
1303 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1305 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1307 2018-04-30 Richard Biener <rguenther@suse.de>
1309 * tree-cfg.c (verify_address): Remove base argument, add
1310 flag whether to check TREE_ADDRESSABLE and do that.
1311 (verify_expr): Remove.
1312 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1313 (verify_gimple_assign_single): Likewise.
1314 (verify_gimple_switch): Likewise.
1315 (verify_expr_location_1): Dereference tp once. Add (disabled)
1316 piece from verify_expr.
1317 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1319 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1321 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1323 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1325 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1326 (small_data_pattern): Likewise.
1327 (arc_rewrite_small_data): Likewise.
1328 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1329 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1330 (get_symbol_alignment): New function.
1331 (legitimate_small_data_address_p): Likewise.
1332 (legitimate_scaled_address): Update, call
1333 legitimate_small_data_address_p.
1334 (output_sdata): New static variable.
1335 (arc_print_operand): Update how we handle small data operands.
1336 (arc_print_operand_address): Likewise.
1337 (arc_legitimate_address_p): Update, use
1338 legitimate_small_data_address_p.
1339 (arc_rewrite_small_data_p): Remove.
1340 (arc_rewrite_small_data_1): Likewise.
1341 (arc_rewrite_small_data): Likewise.
1342 (small_data_pattern): Likewise.
1343 (compact_sda_memory_operand): Update to use
1344 legitimate_small_data_address_p and get_symbol_alignment.
1345 (prepare_move_operands): Don't rewite sdata pattern.
1346 (prepare_extend_operands): Remove.
1347 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1349 (zero_extendqisi2): Likewise.
1350 (zero_extendhisi2): Likewise.
1351 (extendqihi2): Likewise.
1352 (extendqisi2): Likewise.
1353 (extendhisi2): Likewise.
1357 * config/arc/constraints.md (Usd): Change it to memory constraint.
1359 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1361 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1362 as source of std instructions.
1363 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1364 allow 6-bit constants as source for store instructions.
1365 (movdi_insn): Update instruction pattern to allow 6-bit constants
1366 as source for store instructions.
1368 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1370 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1372 2018-04-30 Nathan Sidwell <nathan@acm.org>
1373 Sandra Loosemore <sandra@codesourcery.com>
1375 * dumpfile.c (dump_open): Allow '-' for stdout.
1376 * doc/invoke.texi (Developer Options): Document dump filename
1377 determination early. Document stdin/stdout selection.
1379 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1381 Microblaze Target: PIC data text relative
1383 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1384 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1386 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1387 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1388 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1389 New addressing mode for data-text relative position indepenedent code.
1390 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1391 'ADDRESS_SYMBOLIC_TXT_REL'.
1392 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1393 (microblaze_legitimate_pic_operand): Exclude function calls from
1394 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1395 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1397 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1398 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1399 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1400 for 'address + offset'.
1401 (microblaze_expand_prologue): Add new function prologue call for
1403 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1404 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1405 table in case of TARGET_PIC_DATA_TEXT_REL.
1406 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1407 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1408 Add new macros 'UNSPEC_TEXT',
1409 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1410 + exclude function calls from 'UNSPEC_PLT' in case of data text
1412 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1413 new target hook for generating address diff vector tables in case of
1415 * doc/tm.texi : Regenerate.
1416 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1417 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1418 of addr diff vector generation.
1419 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1420 target hook definition.
1421 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1422 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1423 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1424 option for data text relative.
1426 2018-04-30 Richard Biener <rguenther@suse.de>
1428 * tree-chrec.h (evolution_function_is_constant_p): Remove
1430 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1432 2018-04-30 Richard Biener <rguenther@suse.de>
1435 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1437 2018-04-30 Richard Biener <rguenther@suse.de>
1439 PR tree-optimization/28364
1440 PR tree-optimization/85275
1441 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1442 copying first exit test.
1444 2018-04-28 Mark Wielaard <mark@klomp.org>
1446 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1448 (dwarf_AT): Handle DW_AT_addr_base.
1449 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1451 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1454 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1455 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1456 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1457 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1459 2018-04-28 Richard Biener <rguenther@suse.de>
1461 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1462 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1463 to reflect use. Only add interesting stmts.
1465 2018-04-27 Martin Jambor <mjambor@suse.cz>
1468 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1469 the jump function allows for passing through aggregate values.
1471 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1473 * input.h (in_system_header_at): Convert from macro to inline
1475 (from_macro_expansion_at): Likewise.
1476 (from_macro_definition_at): Likewise.
1478 2018-04-27 Jeff Law <law@redhat.com>
1480 * config.gcc: Mark tile* targets as deprecated/obsolete.
1482 2018-04-27 Richard Biener <rguenther@suse.de>
1484 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1487 2018-04-27 Richard Biener <rguenther@suse.de>
1489 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1491 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1493 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1494 with Yd constraint. Set "preferred_for_speed" attribute from
1495 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1497 (*movdi_internal): Ditto.
1498 (movti_interunit splitters): Remove
1499 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1500 (movdi_interunit splitters): Ditto.
1501 * config/i386/constraints.md (Ye): Remove.
1502 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1504 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1507 * config/aarch64/constraints.md (Usg): Limit to 31.
1510 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1512 PR tree-optimization/85529
1513 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1514 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1515 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1516 zero extension or masking of the MSB bit.
1517 (optimize_range_tests): Add FIRST_BB argument, pass it through
1518 to optimize_range_tests_var_bound.
1519 (maybe_optimize_range_tests, reassociate_bb): Adjust
1520 optimize_range_tests callers.
1522 2018-04-26 Richard Biener <rguenther@suse.de>
1523 Jakub Jelinek <jakub@redhat.com>
1525 * cgraph.h (symbol_table): Just declare debug method here.
1526 * symtab.c (symbol_table::debug): Define.
1528 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1530 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1532 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1534 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1535 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1536 (*movdi_internal): Substitute Yi and Yj constraint with x
1537 and Ym and Yn constraint with y constraint. Update "isa"
1538 attribute and set "preferred_for_speed" attribute from
1539 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1540 (*movsi_internal): Ditto.
1541 (*movdf_internal): Ditto.
1542 (*movsf_internal): Ditto.
1543 (*zero_extendsidi2): Ditto.
1544 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1545 (sse2_loadld): Ditto.
1546 (*vec_extract<ssevecmodelower>_0): Ditto.
1547 (*vec_extractv4si_0_zext_sse4): Ditto.
1548 (vec_concatv2di): Ditto.
1549 (*vec_dup<mode>): Ditto.
1550 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1551 * config/i386/constraints.md (Yi): Remove.
1556 2018-04-26 Nathan Sidwell <nathan@acm.org>
1558 * dumpfile.c (dump_open): New.
1559 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1560 (dump_finish): Detect stdio/stderr by value not name.
1562 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1564 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1566 2018-04-26 Tom de Vries <tom@codesourcery.com>
1569 * config/nvptx/nvptx.c (verify_neutering_jumps)
1570 (verify_neutering_labels): New function
1571 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1573 2018-04-26 Tom de Vries <tom@codesourcery.com>
1576 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1577 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1580 2018-04-26 Richard Biener <rguenther@suse.de>
1581 Tom de Vries <tom@codesourcery.com>
1584 * lto-streamer-out.c (output_function): Fixup loops if required to match
1585 discovery done in the reader.
1587 2018-04-26 Richard Biener <rguenther@suse.de>
1589 PR tree-optimization/85116
1590 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1591 have a loop exit from the single latch predecessor. Remove
1592 case of header with just condition.
1593 (ch_base::copy_headers): Exclude infinite loops from any
1595 (pass_ch::execute): Record exits.
1597 2018-04-26 Richard Biener <rguenther@suse.de>
1599 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1600 prologue cost vector and pass it to vect_get_load_cost.
1601 (vect_get_peeling_costs_all_drs): Likewise.
1602 (vect_peeling_hash_get_lowest_cost): Likewise.
1603 (vect_enhance_data_refs_alignment): Likewise.
1605 2018-04-26 Richard Biener <rguenther@suse.de>
1608 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1609 checking of integer<->pointer conversions.
1610 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1611 sign-/zero-extending pointer types.
1612 (expand_omp_for_static_chunk): Likewise.
1614 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1615 Jean Lee <xiaoyur347@gmail.com>
1617 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1618 (TARGET_ASAN_SHADOW_OFFSET): Define.
1619 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1620 true for -fsanitize=address.
1622 2018-04-25 Mark Wielaard <mark@klomp.org>
1624 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1627 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1629 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1630 than "alu", remove explicit "memory" and "imm_disp" attributes.
1631 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1634 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1635 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1636 gen_lowpart_no_emit.
1638 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1641 * config/i386/i386.c (ix86_expand_builtin): Change memory
1642 operand to XI, extend p0 to Pmode.
1643 * config/i386/i386.md: Change unspec volatile and operand
1644 1 mode to XI, change operand 0 mode to P.
1646 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1648 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1649 GET_MODE_MASK before any checking.
1650 (nds32_can_use_bset_p): Likewise.
1651 (nds32_can_use_btgl_p): Likewise.
1653 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1655 * config/nds32/nds32-doubleword.md: New define_split pattern for
1656 illegal register number.
1658 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1660 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1662 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1664 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1666 2018-04-25 Richard Biener <rguenther@suse.de>
1668 * lto-streamer.h (LTO_major_version): Bump to 8.
1670 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1672 * BASE-VER: Set to 9.0.0.
1674 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1676 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1677 in __abskf2 and __powikf2.
1679 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1682 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1683 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1684 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1685 Use the above on operand 2. Reindent.
1686 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1688 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1691 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1693 * config/i386/i386.opt (mcet): Removed.
1694 * doc/install.texi: Remove -mcet documentation.
1695 * doc/invoke.texi: Likewise.
1697 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1700 * doc/install.texi: Remove -mcet from bootstrap-cet.
1702 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1705 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1706 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1710 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1711 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1712 containing a CONST_VECTOR.
1714 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1716 * doc/install.texi: Update newlib dependency for nvptx.
1718 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1721 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1722 instead of INTVAL when shifting x left.
1724 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1726 PR tree-optimization/85478
1727 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1728 vect_grouped_store_supported for single element vectors.
1730 2018-04-24 Richard Biener <rguenther@suse.de>
1733 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1734 load cost increase to the case of non-constant step.
1736 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1739 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1740 destination if any_malformed_asm.
1742 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1745 * expr.c (store_field): In the bitfield case, if the value comes from
1746 a function call and is returned in registers by means of a PARALLEL,
1747 do not change the mode of the temporary unless BLKmode and VOIDmode.
1749 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1751 PR rtl-optimization/85423
1752 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1753 dependencies to debug insns when the previous insn is non-debug.
1755 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1757 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1758 enums into a single definition.
1759 (fls): Fix predicates and printing.
1762 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1764 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1765 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1766 and short u6 immediate.
1767 (check_if_valid_sleep_operand): Remove.
1768 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1770 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1772 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1773 flag_always_save_lp condition.
1774 * config/nds32/nds32.opt (malways-save-lp): New option.
1776 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1778 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1779 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1780 * config/nds32/nds32.h
1781 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1782 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1784 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1786 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1787 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1789 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1790 Chung-Ju Wu <jasonwucj@gmail.com>
1792 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1794 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1795 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1797 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1799 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1801 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1803 * config/nds32/nds32-protos.h (nds32_data_alignment,
1804 nds32_local_alignment): Declare.
1805 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1806 nds32_local_alignment): New functions.
1807 (TARGET_CONSTANT_ALIGNMENT): Define.
1808 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1810 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1812 * config/nds32/nds32.c
1813 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1814 (TARGET_MODES_TIEABLE_P): Likewise.
1816 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1818 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1821 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1822 Chung-Ju Wu <jasonwucj@gmail.com>
1824 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1825 for unaligned access.
1826 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1828 * config/nds32/nds32-intrinsic.md: Likewise.
1829 * config/nds32/nds32_intrinsic.h: Likewise.
1830 * config/nds32/nds32.h (nds32_builtins): Likewise.
1831 * config/nds32/nds32.opt (munaligned-access): New option.
1832 * config/nds32/nds32.c (nds32_asm_file_start): Display
1833 flag_unaligned_access status.
1835 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1837 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1838 -mno-relax is present.
1839 * config/riscv/linux.h (LINK_SPEC): Ditto.
1841 2018-04-20 Martin Sebor <msebor@redhat.com>
1844 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1846 (gimple_fold_builtin_stxcpy_chk): Same.
1847 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1849 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1852 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1853 __powikf2 when long double is IEEE 128-bit.
1855 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1857 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1858 step to make sure stack always aligned.
1860 2018-04-20 Carl Love <cel@us.ibm.com>
1863 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1864 size check for arg0.
1866 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1867 Tom de Vries <tom@codesourcery.com>
1870 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1871 Emit insns for calls too.
1872 (nvptx_find_par): Always look for worker-level predecessor insn.
1873 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1875 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1876 (nvptx_process_pars): Propagate frames for calls.
1878 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1881 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1883 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1884 (ix86_handle_option): Don't handle OPT_mibt.
1885 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1887 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1889 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1890 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1891 (ix86_target_macros): Define __CET__ with flag_cf_protection
1892 for -fcf-protection.
1893 * config/i386/i386.c (isa2_opts): Remove -mibt.
1894 * config/i386/i386.h (TARGET_IBT): Removed.
1895 (TARGET_IBT_P): Likewise.
1896 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1897 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1898 * config/i386/i386.opt (mcet): Update help message.
1901 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1902 -mcet as an alias for -mshstk.
1904 2018-04-20 Richard Biener <rguenther@suse.de>
1907 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1908 complexity by forcing a single use of the multiply operand.
1910 2018-04-20 Martin Jambor <mjambor@suse.cz>
1913 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1914 recursion dependency to only apply to non-clones.
1916 2018-04-20 Martin Jambor <mjambor@suse.cz>
1919 * ipa-cp.c (create_specialized_node): Check that clones of
1920 self-recursive edges exist during IPA-CP.
1922 2018-04-19 Toon Moene <toon@moene.org>
1924 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1927 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1929 PR tree-optimization/85467
1930 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1931 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1932 VECTOR_CST element to type.
1934 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1937 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1938 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1939 (builtin_longjmp): Likewise.
1940 (save_stack_nonlocal): New pattern.
1941 (restore_stack_nonlocal): Likewise.
1943 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1946 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1947 Replace ASM_OUTPUT_LABEL with fprintf.
1949 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1952 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1953 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1954 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1955 define __IBT__ and __SHSTK__ for -fcf-protection.
1956 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1958 (ix86_trampoline_init): Likewise.
1959 (x86_output_mi_thunk): Likewise.
1960 (ix86_notrack_prefixed_insn_p): Likewise.
1961 (ix86_option_override_internal): Don't disallow -fcf-protection.
1962 * config/i386/i386.md (rdssp<mode>): Also enable for
1964 (incssp<mode>): Likewise.
1965 (nop_endbr): Likewise.
1966 * config/i386/i386.opt (mcet): Change help message to built-in
1970 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1971 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1972 enable CET built-in functions.
1974 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1976 * common/config/i386/i386-common.c
1977 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1978 OPTION_MASK_ISA_MOVDIRI_UNSET,
1979 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1980 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1981 * config.gcc (movdirintrin.h): New header.
1982 * config/i386/cpuid.h (bit_MOVDIRI,
1983 bit_MOVDIR64B): New bits.
1984 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1986 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1987 (VOID, PVOID, PCVOID)): New function types.
1988 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1989 __builtin_ia32_directstoreu_u64,
1990 __builtin_ia32_movdir64b): New builtins.
1991 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1992 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1994 (ix86_valid_target_attribute_inner_p): Ditto.
1995 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1996 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1997 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1998 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1999 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2000 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2001 (movdiri<mode>, movdir64b_<mode>): New.
2002 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2003 * config/i386/immintrin.h: Include movdirintrin.h.
2004 * config/i386/movdirintrin.h: New file.
2005 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2007 2018-04-19 Richard Biener <rguenther@suse.de>
2010 * cfg.c (clear_bb_flags): When loop state says we have
2011 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2013 2018-04-19 Richard Biener <rguenther@suse.de>
2015 PR tree-optimization/84737
2016 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2017 copying restrict info.
2018 (vect_setup_realignment): Use it.
2019 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2020 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2021 the first DR to all generated stores.
2022 (vectorizable_load): Likewise for loads.
2024 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2026 PR tree-optimization/85446
2027 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2028 the integral and pointer types to have the same precision.
2030 * doc/install.texi: Document --disable-cet being the default and
2033 2018-04-18 Martin Liska <mliska@suse.cz>
2035 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2038 2018-04-18 Martin Liska <mliska@suse.cz>
2041 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2044 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2045 arguments if they are comparable.
2047 2018-04-18 Martin Liska <mliska@suse.cz>
2050 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2053 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2056 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2059 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2060 ENDBR after calling __morestack.
2062 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2065 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2066 by using gcc_base_ver to generate a gcc_driver_version, and use
2067 it when generating GCC_DRIVER_NAME.
2068 * configure: Regenerate.
2070 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2073 * config.gcc: Obsolete powerpc*-*-*spe*.
2075 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2078 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2079 (stabstr_D): Change type of unum from unsigned int to
2080 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2083 2018-04-17 Jim Wilson <jimw@sifive.com>
2086 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2087 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2088 Set arg_pointer_offset after using pretend_args_size.
2090 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2092 PR rtl-optimization/85431
2093 * dse.c (record_store): Ignore zero width stores.
2096 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2097 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2098 __builtin_stack_restore rather than after it.
2099 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2100 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2101 argument instead of virtual_dynamic_stack_rtx.
2103 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2105 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2107 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2108 Add note to error message to explain internal mapping of overloaded
2109 built-in function name to non-overloaded built-in function name.
2110 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2113 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2116 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2117 where the inputs overlap with the output.
2119 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2121 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2122 (=v, v) alternative and explicit "memory" attribute.
2123 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2124 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2126 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2127 "sselog1" type instead of "sselog".
2128 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2129 "sselog". Remove explicit "memory" attribute.
2130 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2131 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2133 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2134 "sselog1" type instead of "sselog". Remove explicit "memory"
2136 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2137 use "sselog1" type instead of "sselog". Remove explicit "memory"
2139 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2140 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2142 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2143 "sselog1" type instead of "sselog". Remove explicit "memory"
2145 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2146 use "sselog1" type instead of "sselog". Remove explicit "memory"
2150 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2153 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2156 2018-04-17 Martin Jambor <mjambor@suse.cz>
2159 * ipa-cp.c (create_specialized_node): Call
2160 expand_all_artificial_thunks if necessary.
2162 2018-04-17 Martin Liska <mliska@suse.cz>
2165 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2166 in message, remote space in between '_G' and '('.
2168 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2171 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2172 avx512f_vmcmp<mode>3<round_saeonly_name>,
2173 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2174 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2175 avx512f_rndscale<mode><round_saeonly_name>,
2176 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2177 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2178 Use %<iptr>2 instead of %2 for -masm=intel.
2179 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2180 avx512f_vcvttss2usi<round_saeonly_name>,
2181 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2183 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2184 avx512f_vcvttsd2usi<round_saeonly_name>,
2185 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2186 Use %q1 instead of %1 for -masm=intel.
2187 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2188 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2189 of %3 for -masm=intel.
2190 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2192 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2194 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2196 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2197 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2199 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2201 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2202 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2203 %g1 and one with %0 and %1.
2204 (avx512er_vmrcp28<mode><round_saeonly_name>,
2205 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2207 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2208 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2209 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2210 of %0 and %{%4%} for -masm=intel.
2211 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2212 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2213 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2214 order of %0 and %{%5%}%{z%} for -masm=intel.
2216 2018-04-17 Jan Hubicka <jh@suse.cz>
2219 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2221 2018-04-17 Martin Liska <mliska@suse.cz>
2224 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2225 for target_clone error message. Make default implementation
2226 clone to be a local declaration.
2227 (separate_attrs): Add new argument and check for an empty
2229 (expand_target_clones): Handle it.
2230 (ipa_target_clone): Make redirection just for target_clones
2233 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2234 Tom de Vries <tom@codesourcery.com>
2237 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2238 tiled basic blocks without omp continue statements.
2240 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2243 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2244 vec_extract expression as having side effects to make sure it gets
2247 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2250 * config/i386/i386.c (get_builtin_code_for_version): Check
2253 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2256 * gcc/config.gcc: Support "skylake".
2257 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2259 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2260 (processor_target_table): Add "skylake".
2261 (ix86_option_override_internal): Add "skylake".
2262 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2263 PROCESSOR_CANNONLAKE.
2264 (get_builtin_code_for_version): Fix priority for
2265 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2266 PROCESSOR_SKYLAKE-AVX512.
2267 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2268 (processor_type): Add PROCESSOR_SKYLAKE.
2270 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2271 Jason Merrill <jason@redhat.com>
2274 * convert.c (convert_to_integer_1): Use direct recursion for
2275 enumeral types and types with a precision less than the number
2276 of bits in their mode.
2278 2018-04-16 Julia Koval <julia.koval@intel.com>
2281 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2282 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2284 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2287 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2288 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2289 and -mno-direct-move.
2291 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2294 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2295 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2296 Ensure negative shifts result in {0}.
2298 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2300 PR rtl-optimization/79916
2301 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2302 regs (if any) to define how to gnerate SD moves when LRA is in
2305 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2307 PR rtl-optimization/85393
2308 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2309 * except.c (expand_dw2_landing_pad_for_region): Make static.
2310 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2311 a label and unconditional jump to old_bb, rather than
2312 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2315 PR rtl-optimization/85376
2316 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2317 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2318 instead of a specific value.
2320 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2321 Bin Cheng <bin.cheng@arm.com>
2323 PR tree-optimization/82965
2324 PR tree-optimization/83991
2325 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2326 by_profile_only parameter.
2327 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2328 information if the loop was predicted to iterate too many times.
2329 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2331 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2334 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2337 2018-04-13 Martin Liska <mliska@suse.cz>
2338 Jakub Jelinek <jakub@redhat.com>
2341 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2342 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2343 * builtins.c (expand_builtin_memory_copy_args): Use
2344 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2345 handle dest_addr == pc_rtx.
2347 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2350 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2351 asked to not generate direct moves.
2352 (fix_trunc<mode>si2_stfiwx): Similar.
2353 (fix_trunc<mode>si2_internal): Similar.
2355 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2358 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2359 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2360 lookup if dest in some wider mode is known to be const0_rtx and
2361 if so, record permanent equivalence for it to be ZERO_EXTEND of
2362 the narrower mode destination.
2364 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2366 * lto-streamer-out.c (output_function): Revert 259346.
2367 * omp-expand.c (expand_oacc_for): Likewise.
2369 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2371 PR rtl-optimization/85354
2372 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2373 * sel-sched.c (sel_global_init): ... here.
2375 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2378 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2379 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2380 mode for PE-COFF targets.
2381 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2382 (i386_pe_asm_lto_end): Likewise.
2383 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2384 (TARGET_ASM_LTO_END): Likewise.
2385 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2386 (i386_pe_asm_lto_start): New function.
2387 (i386_pe_asm_lto_end): Likewise.
2389 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2390 Richard Biener <rguenther@suse.de>
2393 * lto-streamer-out.c (output_function): Fix CFG loop state before
2395 * omp-expand.c (expand_oacc_for): Handle calls to internal
2396 functions like regular functions.
2398 2018-04-12 Richard Biener <rguenther@suse.de>
2401 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2402 for the early LTO debug to properly generate references to it
2403 during DIE emission. Do not re-use that for the skeleton for
2405 (dwarf2out_early_finish): Likewise.
2407 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2410 * config/i386/sse.md
2411 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2412 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2413 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2414 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2415 and output is a reg, avoid creating invalid lowpart subreg, but
2416 instead split into a 512-bit move. Don't split if not AVX512VL,
2417 input is xmm16+ reg and output is a mem.
2418 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2419 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2420 xmm16+ reg and output is a mem.
2422 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2424 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2425 also for flag_dwarf2_cfi_asm.
2427 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2429 PR rtl-optimization/85342
2430 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2431 a bool scalar var inside of the loop instead. Don't try to update
2432 recog_data.operand after failed apply_change_group.
2434 2018-04-12 Tom de Vries <tom@codesourcery.com>
2437 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2438 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2439 array with flexible array member as array without given dimension.
2440 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2441 argument for undefined param to true.
2443 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2446 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2447 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2448 from PowerPC section.
2449 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2450 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2451 help text that is too long.
2452 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2453 help text that is too long.
2454 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2455 help text that is too long.
2457 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2459 * config/alpha/alpha.md (stack_probe_internal): Rename
2460 from "probe_stack". Update all callers.
2462 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2464 PR rtl-optimization/84566
2465 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2466 sched_macro_fuse_insns.
2468 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2471 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2472 (compute_block_dependences): ... from here.
2474 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2476 PR tree-optimization/85331
2477 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2478 from int to HOST_WIDE_INT.
2480 2018-04-11 Martin Jambor <mjambor@suse.cz>
2483 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2484 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2485 not the same as the source val.
2486 (cgraph_edge_brings_value_p): New parameter.
2487 (gather_edges_for_value): Pass destination value to
2488 cgraph_edge_brings_value_p.
2489 (perhaps_add_new_callers): Likewise.
2490 (get_info_about_necessary_edges): Likewise and exclude values brought
2491 only by self-recursive edges.
2492 (create_specialized_node): Redirect only clones of self-calling edges.
2493 (+self_recursive_pass_through_p): New function.
2494 (find_more_scalar_values_for_callers_subset): Use it.
2495 (find_aggregate_values_for_callers_subset): Likewise.
2496 (known_aggs_to_agg_replacement_list): Removed.
2497 (decide_whether_version_node): Re-calculate known constants for all
2498 remaining context clones.
2500 2018-04-11 Richard Biener <rguenther@suse.de>
2503 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2504 from early DWARF output.
2505 (dwarf2out_early_finish): Output line info unconditionally into
2506 early DWARF and add reference to it.
2508 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2511 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2512 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2513 other than V2DFmode using iptr mode attribute.
2514 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2516 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2518 PR rtl-optimization/84659
2519 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2521 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2524 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2526 (output_loc_list): Pass address of a dummy size variable even in the
2527 locview handling loop.
2528 (index_location_lists): Add comment on why skip_loc_list_entry can't
2531 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2534 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2537 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2540 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2541 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2542 and -mstring-compare-inline-limit.
2544 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2547 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2548 for stack clash protection in a register whenever we need it to be in
2551 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2553 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2554 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2556 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2559 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2561 (mlong-double-): Ditto.
2562 * config/rs6000/sysv4.opt (msdata=): Ditto.
2563 (mtls-size=): Ditto.
2565 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2567 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2568 erroneous entries for
2569 "vector int vec_ldl (int, long int *)", and
2570 "vector unsigned int vec_ldl (int, unsigned long int *)".
2571 Add comments and entries for
2572 "vector bool char vec_ldl (int, bool char *)",
2573 "vector bool short vec_ldl (int, bool short *)",
2574 "vector bool int vec_ldl (int, bool int *)",
2575 "vector bool long long vec_ldl (int, bool long long *)",
2576 "vector pixel vec_ldl (int, pixel *)",
2577 "vector long long vec_ldl (int, long long *)",
2578 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2579 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2580 type tree bool_long_long_type_node and correct definition of
2581 bool_V2DI_type_node to make reference to this new type tree.
2582 (rs6000_mangle_type): Replace erroneous reference to
2583 bool_long_type_node with bool_long_long_type_node.
2584 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2585 comments to emphasize sign distinctions for char and int types and
2586 replace RS6000_BTI_bool_long constant with
2587 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2588 use of RS6000_BTI_pixel.
2589 (bool_long_type_node): Remove this macro definition.
2590 (bool_long_long_type_node): New macro definition
2592 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2594 PR rtl-optimization/85300
2595 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2596 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2597 simplify_unary_operation fails.
2599 2018-04-10 Martin Liska <mliska@suse.cz>
2601 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2602 cgraph_edge and ipa_ref.
2604 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2608 * config/i386/sse.md
2609 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2610 computation of the VEC_MERGE selector from mask.
2611 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2612 Fix decoding of the VEC_MERGE selector into mask.
2614 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2616 PR tree-optimization/85286
2617 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2619 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2621 * final.c (final_1): Set insn_last_address as well as
2622 insn_current_address.
2624 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2627 * explow.c (emit_stack_probe): Call validize_mem on memory location
2628 before passing it to gen_probe_stack. Create address operand and
2629 legitimize it for the probe_stack_address case.
2631 2018-04-09 Jan Hubicka <jh@suse.cz>
2634 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2635 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2636 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2638 (free_lang_data): Rebuild type inheritance graph.
2640 2018-04-09 Martin Sebor <msebor@redhat.com>
2642 * invoke.texi (-finline-small-functions): Mention other optimization
2644 (-findirect-inlining, -fpartial-inlining): Same.
2645 (-finline-functions-called-once): Same.
2646 (-freorder-blocks-and-partition): Same.
2648 2018-04-09 Jan Hubicka <jh@suse.cz>
2651 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2652 jumps; choose last target that matches the criteria (i.e.
2653 no partition changes for non-crossing jumps).
2654 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2655 support for redirecting crossing jumps to non-crossing.
2657 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2659 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2660 also for naked functions.
2662 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2664 * config/arc/arc.md (add_shift): New pattern.
2665 (add_shift2): Likewise.
2666 (sub_shift): Likewise.
2667 (sub_shift_cmp0_noout): Likewise.
2668 (compare_si_ashiftsi): Likewise.
2669 (xbfu_cmp0_noout): New combine pattern.
2670 (xbfu_cmp0"): Likewise.
2671 (movsi_set_cc_insn): Place the predicable variant first.
2672 (commutative_binary_cmp0_noout): Remove clobber.
2673 (commutative_binary_cmp0): New pattern.
2674 (noncommutative_binary_cmp0): Likewise.
2675 (noncommutative_binary_cmp0_noout): Likewise.
2676 (noncommutative_binary_comparison_result_used): Removed.
2677 (rsub_cmp0): New pattern.
2678 (rsub_cmp0_noout): Likewise.
2679 (extzvsi): Changed, keep only meaningful variants.
2680 (SQH, SEZ): New iterators.
2681 (SQH_postfix): New mode attribute.
2682 (SEZ_prefix): New code attribute.
2683 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2684 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2685 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2687 (noncommutative_operator): Check the availability of barrel
2690 2018-04-09 Richard Biener <rguenther@suse.de>
2692 PR tree-optimization/85284
2693 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2694 Only use the niter constraining form of simple_iv when the exit
2697 2018-04-09 Tom de Vries <tom@codesourcery.com>
2700 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2701 (define_expand "*memory_barrier"): New define_expand.
2702 (define_insn "memory_barrier"): New insn.
2704 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2706 PR rtl-optimization/80463
2707 PR rtl-optimization/83972
2708 PR rtl-optimization/83480
2710 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2711 correct producer for the insn.
2712 (tidy_control_flow): Fixup seqnos in case of debug insns.
2714 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2716 PR rtl-optimization/83913
2718 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2719 different sched-times when merging exprs.
2721 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2723 PR rtl-optimization/83962
2725 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2726 tidy_fallthru_edge and tidy_control_flow.
2728 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2730 PR rtl-optimization/83530
2732 * sel-sched.c (force_next_insn): New global variable.
2733 (remove_insn_for_debug): When force_next_insn is true, also leave only
2734 next insn in the ready list.
2735 (sel_sched_region): When the region wasn't scheduled, make another pass
2736 over it with force_next_insn set to 1.
2738 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2740 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2742 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2743 for interrupt control.
2744 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2745 functions for interrupt control.
2746 * config/nds32/nds32-intrinsic.md: Likewise.
2747 * config/nds32/nds32_intrinsic.h: Likewise.
2748 * config/nds32/nds32.h (nds32_builtins): Likewise.
2750 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2752 * config/nds32/nds32.c (nds32_init_machine_status,
2753 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2754 strict_aligned_p field.
2755 (nds32_expand_to_rtl_hook): New function.
2756 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2757 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2759 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2760 Chung-Ju Wu <jasonwucj@gmail.com>
2762 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2763 * config/nds32/nds32-n7.md: New file.
2764 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2765 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2767 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2768 * config/nds32/nds32.md (pipeline_model): Add n7.
2769 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2770 * config/nds32/pipelines.md: Include n7 settings.
2772 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2773 Chung-Ju Wu <jasonwucj@gmail.com>
2775 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2776 * config/nds32/nds32-e8.md: New file.
2777 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2778 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2780 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2781 * config/nds32/nds32.md (pipeline_model): Add e8.
2782 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2783 * config/nds32/pipelines.md: Include e8 settings.
2785 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2786 Chung-Ju Wu <jasonwucj@gmail.com>
2788 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2789 * config/nds32/nds32-n8.md: New file.
2790 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2791 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2793 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2794 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2795 * config/nds32/nds32.md (pipeline_model): Add n8.
2796 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2797 * config/nds32/pipelines.md: Include n8 settings.
2799 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2800 Chung-Ju Wu <jasonwucj@gmail.com>
2802 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2803 * config/nds32/nds32-n9-2r1w.md: New file.
2804 * config/nds32/nds32-n9-3r2w.md: New file.
2805 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2806 nds32_register_ports): New or modify for cpu n9.
2807 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2809 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2810 * config/nds32/nds32-utils.c: New file.
2811 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2812 TARGET_MUL_SLOW): Define.
2813 * config/nds32/nds32.md (pipeline_model): New attribute.
2814 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2815 New options that support cpu n9.
2816 * config/nds32/pipelines.md: Include n9 settings.
2817 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2819 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2821 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2822 information if necessary.
2823 (output_cond_branch_compare_zero): Likewise.
2824 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2825 (nds32_target_alignment): Refine for alignment.
2826 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2827 (FUNCTION_BOUNDARY): Modify.
2828 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2830 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2832 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2834 * config/nds32/constants.md (unspec_volatile_element): Add values for
2835 TLB operation and data prefetch.
2836 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2837 functions for TLB operation and data prefetch.
2838 * config/nds32/nds32-intrinsic.md: Likewise.
2839 * config/nds32/nds32_intrinsic.h: Likewise.
2840 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2841 (nds32_print_operand): Likewise.
2842 * config/nds32/nds32.h (nds32_builtins): Likewise.
2844 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2845 Andrew Pinski <pinsika@gcc.gnu.org>
2848 * match.pd: Use constant_boolean_node of correct type instead of
2849 boolean_true_node or boolean_false_node for simplifying
2850 pointer comparisons to zero.
2852 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2854 PR tree-optimization/80021
2855 * tree.c (verify_type_variant): Make error call in verify_variant_match
2856 translatable and remove final full stop.
2858 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2860 * config/nds32/constants.md (unspec_volatile_element): Add
2861 UNSPEC_VOLATILE_EH_RETURN.
2862 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2863 nds32_output_stack_pop): Support dwarf exception handling process.
2864 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2865 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2866 exception handling process.
2867 (nds32_compute_stack_frame): Likewise.
2868 (nds32_return_addr_rtx): Likewise.
2869 (nds32_initial_elimination_offset): Likewise.
2870 (nds32_expand_prologue): Likewise.
2871 (nds32_expand_epilogue): Likewise.
2872 (nds32_dynamic_chain_address): New function.
2873 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2875 (DYNAMIC_CHAIN_ADDRESS): Define.
2876 (EH_RETURN_DATA_REGNO): Define.
2877 (EH_RETURN_STACKADJ_RTX): Define.
2878 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2879 patterns for dwarf exception handling.
2881 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2883 * config/nds32/nds32.h: Clean up obsolete macros.
2885 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2887 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2888 Add enum values for particular instructions.
2889 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2890 particular intrinsic functions.
2891 * config/nds32/nds32-intrinsic.md: Likewise.
2892 * config/nds32/nds32_intrinsic.h: Likewise.
2893 * config/nds32/nds32.h (nds32_builtins): Likewise.
2894 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2895 (btst, ave): New patterns for particular instructions.
2897 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2899 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2900 Add enum values for atomic load/store and memory sync.
2901 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2903 * config/nds32/nds32-intrinsic.md: Likewise.
2904 * config/nds32/nds32_intrinsic.h: Likewise.
2905 * config/nds32/nds32.h (nds32_builtins): Likewise.
2907 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2909 PR tree-optimization/85257
2910 * fold-const.c (native_encode_vector): If not all elts could fit
2911 and off is -1, return 0 rather than offset.
2912 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2913 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2914 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2915 adjust buffer in native_interpret_expr call.
2917 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2919 * config/nds32/constants.md (unspec_volatile_element): Add cache
2920 control enum values.
2921 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2922 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2923 * config/nds32/nds32.c (nds32_cctl_names): New.
2924 (nds32_print_operand): Handle cache control register names.
2925 * config/nds32/nds32.h (nds32_builtins): New enum values.
2926 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2928 * config/nds32/nds32.md (type): Add mmu.
2929 * config/nds32/pipelines.md (simple_insn): Add mmu.
2931 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2933 * config/nds32/nds32.md (type): Remove call.
2934 * config/nds32/pipelines.md (simple_insn): Likewise.
2936 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2938 * config/nds32/constants.md (unspec_volatile_element): Add
2939 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2940 UNSPEC_VOLATILE_FMFCFG.
2941 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2942 description for fmfcfg and fmfcsr.
2943 (bdesc_1arg): Add fmtcsr.
2944 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2945 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2946 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2947 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2948 unspec_fmfcfg): New patterns.
2949 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2950 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2951 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2952 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2953 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2954 __nds32__fmfcfg): Define.
2956 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2958 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2959 intrinsic register names.
2960 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2961 intrinsic register enum values and macros.
2963 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2965 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2966 for load/store addressing form.
2967 (nds32_print_operand_address): Likewise.
2969 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2972 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2973 based on LABEL_REF. Remove useless assertion.
2974 (pic_address_needs_scratch): Fix formatting.
2975 (sparc_legitimize_pic_address): Minor tweaks.
2976 (sparc_delegitimize_address): Adjust assertion accordingly.
2977 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2978 into symbolic_operand.
2979 (movsi_high_pic_label_ref): Likewise.
2980 (movsi_lo_sum_pic_label_ref): Likewise.
2981 (movdi_pic_label_ref): Likewise.
2982 (movdi_high_pic_label_ref): Likewise.
2983 (movdi_lo_sum_pic_label_ref): Likewise.
2985 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2987 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2988 custom LIB_SPEC setup.
2990 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2991 Kito Cheng <kito.cheng@gmail.com>
2993 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2994 * config/riscv/freebsd.h: New.
2996 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2998 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2999 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3002 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3003 Kito Cheng <kito.cheng@gmail.com>
3005 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3006 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3007 * config/nds32/nds32-protos.h (nds32_output_call,
3008 nds32_output_return): Declare.
3009 * config/nds32/nds32.md: Refine all the call and return patterns.
3011 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3014 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3015 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3017 PR rtl-optimization/84872
3018 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3019 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3022 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3024 * expr.c (copy_blkmode_to_reg): Revert 254862.
3025 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3027 2018-04-06 Richard Biener <rguenther@suse.de>
3030 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3031 after seeing a component reference with an adjacent field. Treat
3032 refs to arrays at struct end of external decls similar to
3033 refs to unconstrained commons.
3035 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3038 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3039 look through SAVE_EXPRs with non-side-effects argument. Adjust
3041 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3044 2018-04-06 Richard Biener <rguenther@suse.de>
3047 * alias.c (find_base_term): New wrapper around find_base_term
3048 unwinding CSELIB_VAL_PTR changes.
3049 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3052 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3054 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3056 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3057 constant definitions.
3058 ("nop"): lr 0,0 -> nopr r0
3059 ("nop_lr0", "nop_lr1"): New insn definitions.
3061 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3063 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3064 NDS32_V3PUSH_AVAILABLE_P macro.
3066 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3067 Chung-Ju Wu <jasonwucj@gmail.com>
3069 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3070 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3071 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3072 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3073 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3074 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3075 * config/nds32/constraints.md: New constraints and checking for hard
3076 float configuration.
3077 * config/nds32/iterators.md: New mode iterator and attribute for hard
3078 float configuration.
3079 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3081 * config/nds32/nds32-fpu.md: New file.
3082 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3083 deal with hard float code generation.
3084 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3086 (abi_type, float_reg_number): New enum type.
3087 * config/nds32/nds32-predicates.c: New predicates for hard float.
3088 * config/nds32/nds32-protos.h: Declare functions for hard float.
3089 * config/nds32/nds32.c: Implementation for hard float configuration.
3090 * config/nds32/nds32.h: Definitions for hard float configuration.
3091 * config/nds32/nds32.md: Include hard float machine description and
3092 modify patterns for hard float configuration.
3093 * config/nds32/nds32.opt: New options for hard float configuration.
3094 * config/nds32/predicates.md: New predicates for hard float
3097 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3099 * common/config/nds32/nds32-common.c
3100 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3102 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3105 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3106 CONSTRUCTOR_ELT (ctor, ...)->value.
3108 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3111 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3113 2018-04-05 Tom de Vries <tom@codesourcery.com>
3116 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3119 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3120 Kito Cheng <kito.cheng@gmail.com>
3122 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3123 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3124 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3125 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3127 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3128 Kito Cheng <kito.cheng@gmail.com>
3130 * config/nds32/constraints.md (Ufe): New memory constraint.
3131 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3132 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3133 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3135 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3136 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3138 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3140 * config/nds32/nds32.md: Use optimize_size in the condition for
3141 alu-shift instructions.
3143 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3145 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3147 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3149 * config/nds32/nds32.md (negsi2): Refine pattern.
3151 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3152 Chung-Ju Wu <jasonwucj@gmail.com>
3154 * config/nds32/iterators.md (shift_rotate): New code iterator.
3155 (shift): New code attribute.
3156 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3157 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3158 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3159 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3160 bit-wise operations.
3161 (andsi3, *andsi3): Ditto.
3162 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3163 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3164 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3165 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3166 nds32_ior_operand, nds32_xor_operand): New predicates.
3168 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3170 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3171 (addsi3, subsi3): ... this.
3173 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3175 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3177 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3179 * config/nds32/nds32.md: Adjust indention.
3181 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3183 * config/nds32/nds32.md (feature): New attribute.
3185 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3187 * config/nds32/nds32.md (subtype): New attribute.
3189 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3192 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3193 expansion to perform a bitwise AND of the argument followed by a
3194 boolean negation of the result.
3196 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3198 PR rtl-optimization/84878
3199 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3200 the basic block. Assert the use reference is not artificial and that
3201 it has an associated insn.
3203 2018-04-04 Michael Matz <matz@suse.de>
3205 * builtins.c (compute_objsize): Pass correct operand
3206 to array_at_struct_end_p.
3208 2018-04-04 Richard Biener <rguenther@suse.de>
3211 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3212 from contexts for DINFO_LEVEL_TERSE and below.
3214 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3216 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3217 resiter_operand condition.
3218 * config/nds32/nds32.md (*move<mode>): Ditto.
3220 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3221 Monk Chiang <sh.chiang04@gmail.com>
3223 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3225 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3227 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3229 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3230 Kito Cheng <kito.cheng@gmail.com>
3232 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3233 nds32_cond_code_str, output_cond_branch,
3234 output_cond_branch_compare_zero, nds32_expand_cbranch,
3235 nds32_expand_cstore, nds32_expand_movcc,
3236 nds32_output_cbranchsi4_equality_zero,
3237 nds32_output_cbranchsi4_equality_reg,
3238 nds32_output_cbranchsi4_equality_reg_or_const_int,
3239 nds32_output_cbranchsi4_greater_less_zero: New functions.
3240 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3241 nds32_expand_cstore, nds32_expand_movcc,
3242 nds32_output_cbranchsi4_equality_zero,
3243 nds32_output_cbranchsi4_equality_reg,
3244 nds32_output_cbranchsi4_equality_reg_or_const_int,
3245 nds32_output_cbranchsi4_greater_less_zero): Declare.
3246 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3247 nds32_rimm11s_operand): New predicates.
3248 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3249 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3252 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3254 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3255 * config/nds32/nds32.md: Ditto.
3256 * config/nds32/pipelines.md: Ditto.
3258 2018-04-04 Richard Biener <rguenther@suse.de>
3260 PR tree-optimization/85168
3261 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3262 propagating abnormals.
3264 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3266 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3268 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3269 Kito Cheng <kito.cheng@gmail.com>
3271 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3272 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3273 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3274 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3275 * config/nds32/nds32.md (sibcall_internal): New.
3276 (sibcall_register): Remove.
3277 (sibcall_immediate): Remove.
3278 (sibcall_value_internal): New.
3279 (sibcall_value_register): Remove.
3280 (sibcall_value_immediate): Remove.
3281 * config/nds32/predicates.md (nds32_general_register_operand): New.
3282 (nds32_call_address_operand): New.
3284 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3286 PR rtl-optimization/85167
3287 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3288 bb_defs if *split_p, instead preinitialize it to NULL.
3290 PR tree-optimization/85156
3291 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3292 evaluating the argument multiple times.
3294 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3296 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3298 (_mm_cvtpd_ps): Likewise.
3299 (_mm_cvttpd_epi32): Likewise.
3300 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3301 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3302 vector, pixel, and bool following altivec.h include.
3304 2018-04-03 Martin Sebor <msebor@redhat.com>
3306 * doc/extend.texi (Common Function Attributes): Clarify.
3307 (const attribute): Likewise.
3308 (pure attribute): Likewise.
3310 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3313 * config/i386/i386.c (ix86_expand_vector_set): Use
3314 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3316 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3318 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3319 instructions when changing rounding bits to preserve precision bits
3320 in the x87 control word.
3322 2018-04-03 Martin Liska <mliska@suse.cz>
3324 PR tree-optimization/82491
3325 * rtl.h (strip_offset_and_add): Replace += suboffset with
3326 poly_uint64 () + suboffset.
3328 2018-03-29 Martin Liska <mliska@suse.cz>
3329 Martin Jambor <mjambor@suse.cz>
3332 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3333 param_type is not an integral or pointer type.
3335 2018-04-03 Richard Biener <rguenther@suse.de>
3337 * sese.h (recompute_all_dominators): Remove.
3339 2018-04-02 Martin Sebor <msebor@redhat.com>
3341 * doc/invoke.texi (-Wrestrict): Fix typos.
3343 2018-04-02 Jim Wilson <jimw@sifive.com>
3345 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3346 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3347 (<optab>di3, <optab>si3_extend): Likewise.
3348 (<optab>si3_mask, <optab>si3_mask_1): New.
3349 (<optab>di3_mask, <optab>di3_mask_1): New.
3350 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3351 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3352 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3354 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3356 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3359 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3361 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3362 (nds32_canonicalize_comparison): New function.
3364 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3365 Kito Cheng <kito.cheng@gmail.com>
3366 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3368 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3369 * config/nds32/constants.md (unspec_volatile_element): Add
3370 UNSPEC_VOLATILE_RELAX_GROUP.
3371 * config/nds32/nds32-relax-opt.c: New file.
3372 * config/nds32/nds32-predicates.c
3373 (nds32_symbol_load_store_p): New function.
3374 * config/nds32/nds32-protos.h
3375 (nds32_symbol_load_store_p): Declare function.
3376 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3377 * config/nds32/nds32.c
3378 (nds32_register_pass): New function to register pass.
3379 (nds32_register_passes): New function to register passes.
3380 * config/nds32/nds32.md (relax_group): New pattern.
3381 * config/nds32/nds32.opt (mrelax-hint): New option.
3382 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3384 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3386 * config/nds32/t-nds32: Modify files dependency.
3388 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3390 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3391 (PROFILE_HOOK): Define its implementation.
3393 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3395 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3396 type and 32-bit size.
3398 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3401 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3402 (V_128_256): New mode iterator.
3403 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3404 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3405 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3407 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3408 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3410 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3413 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3414 NaN inputs correctly.
3416 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3419 * config/rs6000/vsx.md (??r): New mode attribute.
3420 (*vsx_mov<mode>_64bit): Use it.
3421 (*vsx_mov<mode>_32bit): Likewise.
3423 2018-03-30 Martin Sebor <msebor@redhat.com>
3425 PR tree-optimization/84818
3426 * builtins.c (check_access): Use warning_n.
3428 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3431 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3433 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3436 2018-03-30 Julia Koval <julia.koval@intel.com>
3439 * x86-tune.def (movx, partial_reg_dependency): Enable for
3442 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3445 * lra-constraints.c (process_alt_operands): Move setting
3446 this_alternative_matches below.
3448 2018-03-29 Martin Liska <mliska@suse.cz>
3451 * doc/invoke.texi: Document how LTO works with debug info.
3452 Describe auto-load support of binutils. Mention 'x86-64'
3453 as valid option value of -march option.
3455 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3457 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3460 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3461 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3462 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3465 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3468 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3469 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3470 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3471 for RS6000_BTM_POWERPC64.
3472 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3473 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3474 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3479 2018-03-28 Carl Love <cel@us.ibm.com>
3482 2017-09-27 Carl Love <cel@us.ibm.com>
3484 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3485 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3486 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3489 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3491 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3492 instead of __vector bool.
3493 (_mm_max_pu8): Likewise.
3494 (_mm_min_pi16): Likewise.
3496 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3499 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3500 (DIVWEUO): Likewise.
3502 (DIVDEUO): Likewise.
3503 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3504 DIVWEUO and DIVDEUO.
3505 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3506 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3507 (div_extend): Likewise.
3508 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3510 (__builtin_divweuo): Likewise.
3511 (__builtin_divdeo): Likewise.
3512 (__builtin_divdeuo): Likewise.
3514 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3517 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3518 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3520 PR tree-optimization/82004
3521 * gimple-match-head.c (optimize_pow_to_exp): New function.
3522 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3523 Don't fold to exp if optimize_pow_to_exp is false.
3525 2018-03-28 Martin Liska <mliska@suse.cz>
3528 * calls.c (initialize_argument_information): Fix trailing space.
3529 * common.opt: Fix typo and provide better explanation for
3530 -fsanitize-coverage option.
3531 * config/i386/i386.opt: Fix typo.
3533 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3534 Martin Liska <mliska@suse.cz>
3537 * gimplify.c (asan_poison_variable): Don't do the check for
3538 gimplify_omp_ctxp here.
3539 (gimplify_decl_expr): Do it here.
3540 (gimplify_target_expr): Likewise.
3542 2018-03-28 Martin Liska <mliska@suse.cz>
3545 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3546 chkp_type_bounds_count if MPX is not enabled.
3548 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3550 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3552 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3555 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3556 function to create the function decl for complex long double
3557 multiply and divide for -mabi=ieeelongdouble.
3558 (init_float128_ieee): Call it.
3560 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3563 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3564 -fcf-protection=branch -mibt.
3565 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3567 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3570 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3572 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3575 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3576 extern array declarations.
3578 2018-03-27 Richard Biener <rguenther@suse.de>
3581 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3582 explicit single_use checks.
3584 2018-03-27 Richard Biener <rguenther@suse.de>
3586 PR tree-optimization/85082
3587 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3590 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3592 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3593 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3594 Turn on fasynchronous-unwind-tables and funwind-tables.
3596 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3599 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3600 (*bmi_blsr_<mode>_ccz): Ditto.
3602 2018-03-26 Tom de Vries <tom@codesourcery.com>
3604 PR tree-optimization/85063
3605 * omp-general.c (offloading_function_p): New function. Factor out
3607 * omp-offload.c (pass_omp_target_link::gate): ... here.
3608 * omp-general.h (offloading_function_p): Declare.
3609 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3610 with attribute omp declare target for offloading functions.
3612 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3614 PR tree-optimization/84005
3615 * tree-data-ref.h (get_base_for_alignment): Declare.
3616 * tree-data-ref.c (get_base_for_alignment_1): New function.
3617 (get_base_for_alignment): Likewise.
3618 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3619 get_base_for_alignment to find a suitable base object, instead
3620 of always using drb->base_address.
3622 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3625 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3626 known size by default.
3628 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3631 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3632 and non BLKmode operands.
3634 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3637 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3638 Clean up attributes.
3640 2018-03-23 Richard Biener <rguenther@suse.de>
3643 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3644 we are going to emit early debug for LTO.
3646 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3649 * function.c (match_asm_constraints_1): Don't optimize if input
3650 doesn't satisfy general_operand predicate for output's mode.
3653 * alias.c (write_dependence_p): Don't require for x_canonicalized
3654 non-VOIDmode if x has VOIDmode.
3657 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3658 just don't try to optimize it rather than assert it never happens.
3660 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3662 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3663 macro expansions for definition of ST_INTERNAL_<mode> and
3664 LD_INTERNAL_<mode> builtins.
3665 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3667 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3669 (altivec_expand_st_builtin): Likewise.
3670 (altivec_expand_builtin): Remove calls to deleted functions.
3671 (rs6000_address_for_altivec): Delete this function.
3672 * config/rs6000/vector.md: Remove expands for
3673 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3675 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3678 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3679 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3680 re-computing once computed.
3681 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3682 (arm_init_machine_status): Initialize
3683 machine->static_chain_stack_bytes.
3685 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3688 * doc/extend.texi: Add four new prototypes for vec_ld.
3689 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3690 definitions for more logical presentation.
3691 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3692 entries for V1TI variants of __builtin_altivec_ld builtin.
3693 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3694 handling of V1TI variant of LVX icode pattern.
3695 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3696 (rs6000_gimple_fold_builtin): Likewise.
3697 (altivec_init_builtins): Add code to define
3698 __builtin_altivec_lvx_v1ti function.
3700 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3703 * function.c (match_asm_constraints_1): Don't do the optimization
3704 if input isn't a REG, SUBREG, MEM or constant.
3706 2018-03-22 Tom de Vries <tom@codesourcery.com>
3708 PR tree-optimization/84956
3709 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3710 bb_has_abnormal_pred.
3712 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3715 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3716 DECL_INITIAL (decl) to decl at the end.
3717 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3720 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3722 * doc/extend.texi (__builtin_tgmath): Document when complex
3723 integer types are treated as _Complex _Float64.
3725 2018-03-21 Tom de Vries <tom@codesourcery.com>
3727 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3729 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3731 PR tree-optimization/84960
3732 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3733 if it is ENTRY block, move them into single succ of ENTRY in that case.
3735 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3737 PR tree-optimization/84811
3738 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3739 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3740 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3741 (known_subrange_p): Update accordingly. Cast each value involved
3742 in the size comparison, rather than casting the result of the
3745 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3747 PR tree-optimization/84982
3748 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3749 by flipping the least significant bit rather than all bits from
3750 bitpos to bitpos + bitsize - 1.
3752 2018-03-21 Nathan Sidwell <nathan@acm.org>
3754 * doc/extend.texi (Deprecated Features): Remove mention of
3755 long-deleted deprecations.
3757 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3760 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3761 * configure: Regenerate.
3763 2018-03-21 Tom de Vries <tom@codesourcery.com>
3765 PR tree-optimization/83126
3766 * tree-parloops.c (num_phis): New function.
3767 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3769 2018-03-21 Nathan Sidwell <nathan@acm.org>
3771 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3772 mention anon-struct/union members and trailing attributes.
3774 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3776 PR tree-optimization/84969
3777 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3778 builtin memset partitions if they set different rhs values.
3780 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3782 PR rtl-optimization/84989
3783 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3784 VEC_DUPLICATE with scalar result mode.
3786 2018-03-21 Martin Liska <mliska@suse.cz>
3789 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3790 not intended return statement.
3792 2018-03-21 Martin Liska <mliska@suse.cz>
3795 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3796 (chkp_find_bound_slots_1): Limit number of iterations.
3798 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3801 * Minor grammar fixes for x86 options.
3803 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3806 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3807 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3809 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3812 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3813 (altivec_lvx_<mode>_1op): Likewise.
3814 (altivec_stvx_<mode>_2op): Likewise.
3815 (altivec_stvx_<mode>_1op): Likewise.
3816 (altivec_lvx_<VM2:mode>): New define_expand.
3817 (altivec_stvx_<VM2:mode>): Likewise.
3818 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3819 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3820 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3821 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3822 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3823 (rs6000_gen_lvx): Likewise.
3824 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3825 (altivec_expand_stv_builtin): Likewise.
3826 (altivec_expand_builtin): Likewise.
3827 * config/rs6000/vector.md: Likewise.
3829 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3832 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3835 2018-03-20 Richard Biener <rguenther@suse.de>
3838 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3839 sign-conversions as zero, fall back to standard scalar_stmt
3842 2018-03-20 Martin Liska <mliska@suse.cz>
3845 * predict.c (rebuild_frequencies): Handle case when we have
3846 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3848 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3851 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3852 flag_section_anchors.
3853 * varasm.c (use_blocks_for_decl_p): Remove hack for
3854 dw2_force_const_mem.
3857 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3859 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3860 be created, use lowpart_subreg of operands[0] rather than operands[0]
3862 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3863 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3864 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3865 and n constraint instead of aarch64_shift_imm_di and Usd.
3866 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3867 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3869 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3872 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3873 to favor GPR over NEON registers.
3874 (<shift>di3_neon): Likewise.
3876 2018-03-20 Tom de Vries <tom@codesourcery.com>
3879 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3880 (nvptx_process_pars): Emit bar.sync asap and alap.
3882 2018-03-20 Tom de Vries <tom@codesourcery.com>
3885 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3886 seen_label if seen_label is already set.
3888 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3891 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3892 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3893 Use 1U instead of 1. Formatting fixes.
3896 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3897 instead of TREE_TYPE (s1) for the return value.
3899 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3901 PR tree-optimization/84946
3902 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3903 bitsize + bitsize in poly_uint64 rather than poly_int64.
3906 * dwarf2asm.c: Include fold-const.c.
3907 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3908 of decl rather than decl itself.
3910 PR rtl-optimization/84643
3911 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3913 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3916 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3917 calling assemble_variable.
3919 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3922 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3923 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3925 2018-03-19 Jim Wilson <jimw@sifive.com>
3928 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3929 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3930 (riscv_first_stack_step): Likewise.
3931 (riscv_option_override): Use STACK_BOUNDARY instead of
3933 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3935 (BIGGEST_ALIGNMENT): Set to 128.
3936 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3937 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3940 2018-03-19 Richard Biener <rguenther@suse.de>
3942 PR tree-optimization/84933
3943 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3944 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3946 2018-03-19 Richard Biener <rguenther@suse.de>
3948 PR tree-optimization/84859
3949 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3950 (cond_if_else_store_replacement): Perform sinking operation on
3951 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3952 Generalize what a BB with a single eligible store is.
3954 2018-03-19 Richard Biener <rguenther@suse.de>
3956 PR tree-optimization/84929
3957 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3958 chrec_is_positive against non-chrec arg.
3960 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3963 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3965 2018-03-18 Martin Liska <mliska@suse.cz>
3967 PR rtl-optimization/84635
3968 * regrename.c (build_def_use): Use matches_mode only when
3971 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3973 PR tree-optimization/84913
3974 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3975 vectorize chains of COND_EXPRs.
3977 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3979 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3981 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3983 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3985 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3987 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3989 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3990 Kito Cheng <kito.cheng@gmail.com>
3992 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3993 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3994 (nds32_adjust_reg_alloc_order): New function.
3995 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3997 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3999 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4000 nds32_print_operand, nds32_print_operand_address): Use
4001 HOST_WIDE_INT_PRINT_DEC instead.
4003 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4005 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4007 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4010 * config/i386/i386.c (initial_ix86_tune_features,
4011 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4013 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4014 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4015 rather than 1u << ix86_tune. Formatting fix.
4016 (ix86_option_override_internal): Change ix86_arch_mask from
4017 unsigned int to unsigned HOST_WIDE_INT, initialize to
4018 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4019 (ix86_function_specific_restore): Likewise.
4021 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4024 * postreload.c (reload_combine_recognize_pattern): Perform
4025 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4026 truncate_int_for_mode the result for the destination's mode.
4029 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4030 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4032 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4035 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4038 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4039 regno_allocno_class_array and sorted_pseudos.
4040 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4041 insns where regno is used.
4043 2018-03-16 Martin Liska <mliska@suse.cz>
4046 * multiple_target.c (create_dispatcher_calls): Redirect
4047 reference in the symbol table.
4049 2018-03-16 Martin Liska <mliska@suse.cz>
4052 * multiple_target.c (create_dispatcher_calls): Redirect also
4055 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4059 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4060 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4061 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4063 2018-03-16 Julia Koval <julia.koval@intel.com>
4065 * doc/invoke.texi (Skylake Server): Add CLWB.
4066 Cannonlake): Remove CLWB.
4068 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4070 PR tree-optimization/84841
4071 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4073 (FLOAT_ONE_CONST_TYPE): Define.
4074 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4075 (sort_by_operand_rank): Put entries with higher constant_type last
4076 rather than first to match comments.
4078 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4080 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4083 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4086 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4087 check and use address of target always.
4089 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4092 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4093 (indirect_thunk_bnd_needed): Likewise.
4094 (indirect_thunks_used): Likewise.
4095 (indirect_thunks_bnd_used): Likewise.
4096 (indirect_return_needed): New.
4097 (indirect_return_bnd_needed): Likewise.
4098 (output_indirect_thunk_function): Add a bool argument for
4100 (output_indirect_thunk_function): Don't generate alias for
4101 function return thunk.
4102 (ix86_code_end): Call output_indirect_thunk_function to generate
4103 function return thunks.
4104 (ix86_output_function_return): Set indirect_return_bnd_needed
4105 and indirect_return_needed instead of indirect_thunk_bnd_needed
4106 and indirect_thunk_needed.
4108 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4110 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4111 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4112 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4114 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4115 Paul Hua <paul.hua.gm@gmail.com>
4118 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4120 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4122 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4123 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4126 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4129 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4130 instead of GET_MODE_SIZE when comparing Units.
4132 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4135 * varasm.c (hash_section): Return an unchangeble hash value
4136 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4137 Return !aarch64_can_use_per_function_literal_pools_p ().
4139 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4142 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4143 rather than address of cmode as last argument to prepare_cmp_insn.
4145 2018-03-15 Julia Koval <julia.koval@intel.com>
4147 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4148 F_AVX512VNNI, F_AVX512BITALG): New.
4150 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4153 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4154 insn for floating-point loads and stores.
4156 2018-03-14 Carl Love <cel@us.ibm.com>
4158 * config/rs6000/rs6000-c.c: Add macro definitions for
4159 ALTIVEC_BUILTIN_VEC_PERMXOR.
4160 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4161 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4162 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4163 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4165 * config/doc/extend.texi: Add prototypes for vec_permxor.
4167 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4170 * diagnostic-show-locus.c (class layout_point): Convert m_line
4171 from int to linenum_type.
4172 (line_span::comparator): Use linenum "compare" function when
4173 comparing line numbers.
4174 (test_line_span): New function.
4175 (layout_range::contains_point): Convert param "row" from int to
4177 (layout_range::intersects_line_p): Likewise.
4178 (layout::will_show_line_p): Likewise.
4179 (layout::print_source_line): Likewise.
4180 (layout::should_print_annotation_line_p): Likewise.
4181 (layout::print_annotation_line): Likewise.
4182 (layout::print_leading_fixits): Likewise.
4183 (layout::annotation_line_showed_range_p): Likewise.
4184 (struct line_corrections): Likewise for field m_row.
4185 (line_corrections::line_corrections): Likewise for param "row".
4186 (layout::print_trailing_fixits): Likewise.
4187 (layout::get_state_at_point): Likewise.
4188 (layout::get_x_bound_for_row): Likewise.
4189 (layout::print_line): Likewise.
4190 (diagnostic_show_locus): Likewise for locals "last_line" and
4192 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4193 * input.c (selftest::test_linenum_comparisons): New function.
4194 (selftest::input_c_tests): Call it.
4195 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4196 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4197 * selftest.h (ASSERT_GT): New macro.
4198 (ASSERT_GT_AT): New macro.
4199 (ASSERT_LT): New macro.
4200 (ASSERT_LT_AT): New macro.
4202 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4204 PR rtl-optimization/84780
4205 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4207 2018-03-14 Martin Liska <mliska@suse.cz>
4209 * tree.c (record_node_allocation_statistics): Use
4210 get_stats_node_kind.
4211 (get_stats_node_kind): New function extracted from
4212 record_node_allocation_statistics.
4213 (free_node): Use get_stats_node_kind.
4215 2018-03-14 Richard Biener <rguenther@suse.de>
4217 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4218 that the value-set of ANTIC_IN doesn't grow.
4221 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4223 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4224 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4226 2018-03-14 Julia Koval <julia.koval@intel.com>
4228 * config.gcc (icelake-client, icelake-server): New.
4230 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4231 (initial_ix86_arch_features): Ditto.
4232 (PTA_SKYLAKE): Add SGX.
4233 (PTA_ICELAKE): Remove.
4234 (PTA_ICELAKE_CLIENT): New.
4235 (PTA_ICELAKE_SERVER): New.
4236 (ix86_option_override_internal): Split up icelake on icelake client and
4238 (get_builtin_code_for_version): Ditto.
4239 (fold_builtin_cpu): Ditto.
4240 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4241 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4242 * config/i386/i386.h (processor_type): Ditto.
4243 * doc/invoke.texi: Ditto.
4245 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4248 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4249 INTEGER_CST offset, add it together with bitpos / 8 and
4250 sign extend based on POINTER_SIZE.
4254 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4257 * config/i386/constraints.md (Yc): New register constraint.
4258 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4259 Use Yc constraint for alternative 2 of operand 0. Remove
4260 preferred_for_speed attribute.
4262 2018-03-14 Richard Biener <rguenther@suse.de>
4264 PR tree-optimization/84830
4265 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4266 with the old one to avoid oscillations.
4268 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4271 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4273 (assign_by_spills): Return a flag of reload assignment failure.
4274 Do not process the reload assignment failures. Do not spill other
4275 reload pseudos if they has the same reg class. Update n if
4277 (lra_assign): Add a return arg. Set up from the result of
4278 assign_by_spills call.
4279 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4280 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4281 usage_insns if it is not NULL.
4282 (spill_hard_reg_in_range): New function.
4283 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4284 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4285 function prototypes.
4286 (lra_assign): Change prototype.
4287 * lra.c (lra): Add code to deal with fails by splitting hard reg
4290 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4292 * config/riscv/riscv.opt (mrelax): New option.
4293 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4294 "norelax" when riscv_mrelax is disabled.
4295 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4297 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4300 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4301 reassociation for int modes.
4303 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4305 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4306 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4308 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4309 * config/aarch64/aarch64-sve.md
4310 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4311 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4312 (*extend<mode><Vwide>2): Rename to...
4313 (aarch64_sve_extend<mode><Vwide>2): ...this.
4314 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4315 renaming the old pattern to...
4316 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4318 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4319 define_expand, renaming the old pattern to...
4320 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4321 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4322 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4323 account when deciding which SVE instruction the optab should use.
4324 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4326 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4328 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4329 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4330 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4331 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4332 (tlsdesc_small_<mode>): Turn a define_expand and use
4333 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4334 (tlsdesc_small_advsimd_<mode>): ...this.
4335 (tlsdesc_small_sve_<mode>): New pattern.
4337 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4339 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4340 (UNSPEC_UMUL_HIGHPART): New constants.
4341 (MUL_HIGHPART): New int iteraor.
4342 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4343 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4345 (*<su>mul<mode>3_highpart): New define_insn.
4347 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4350 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4353 2018-03-13 Martin Liska <mliska@suse.cz>
4356 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4358 (sem_item_optimizer::~sem_item_optimizer): Release it.
4359 (sem_item_optimizer::merge_classes): Register variable aliases.
4360 (sem_item_optimizer::fixup_pt_set): New function.
4361 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4362 * ipa-icf.h: Declare new variables and functions.
4364 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4367 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4368 integer_pow2p@2 and test integer_pow2p in condition.
4369 (A < 0 ? C : 0): Similarly for @1.
4372 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4373 characters starting at p contain '\0' character, don't look beyond
4377 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4378 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4381 * reg-stack.c (change_stack): Change update_end var from int to
4382 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4383 also call set_block_for_insn on the newly added insns and rescan.
4386 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4387 on the last operand.
4390 * tree.c (stabilize_reference_1): Return save_expr (e) for
4391 STATEMENT_LIST even if it doesn't have side-effects.
4393 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4395 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4397 2018-03-12 Renlin Li <renlin.li@arm.com>
4399 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4400 aarch64_output_scalar_simd_mov_immediate.
4402 2018-03-12 Martin Sebor <msebor@redhat.com>
4404 PR tree-optimization/83456
4405 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4406 for perfectly overlapping calls to memcpy.
4407 (gimple_fold_builtin_memory_chk): Same.
4408 (gimple_fold_builtin_strcpy): Handle no-warning.
4409 (gimple_fold_builtin_stxcpy_chk): Same.
4410 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4412 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4414 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4415 parameter. Use it for SFmode.
4416 (rs6000_function_arg_advance_1): Adjust.
4417 (rs6000_function_arg): Adjust.
4418 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4420 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4422 PR rtl-optimization/84169
4423 PR rtl-optimization/84780
4424 * combine.c (can_combine_p): Check for a 2-insn combination whether
4425 the destination register is used between the two insns, too.
4427 2018-03-12 Richard Biener <rguenther@suse.de>
4429 PR tree-optimization/84803
4430 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4431 for refs DR analysis didn't process.
4433 2018-03-12 Richard Biener <rguenther@suse.de>
4435 PR tree-optimization/84777
4436 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4437 force-vectorize loops ignore whether we are optimizing for size.
4439 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4441 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4442 (TARGET_MD_ASM_ADJUST): Define.
4444 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4445 Kito Cheng <kito.cheng@gmail.com>
4446 Chung-Ju Wu <jasonwucj@gmail.com>
4448 * config/nds32/nds32.c (nds32_compute_stack_frame,
4449 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4450 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4451 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4452 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4453 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4454 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4455 * config/nds32/nds32.md (prologue, epilogue): Use macro
4456 NDS32_V3PUSH_AVAILABLE_P to do checking.
4458 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4461 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4462 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4463 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4464 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4465 addition of most attributes on !orig_type_die or the attribute not
4466 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4468 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4469 Chung-Ju Wu <jasonwucj@gmail.com>
4471 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4473 * config/nds32/nds32.opt (mvh): New option.
4475 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4476 Chung-Ju Wu <jasonwucj@gmail.com>
4478 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4480 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4481 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4484 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4485 Chung-Ju Wu <jasonwucj@gmail.com>
4487 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4489 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4490 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4492 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4493 Kito Cheng <kito.cheng@gmail.com>
4494 Chung-Ju Wu <jasonwucj@gmail.com>
4496 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4497 UNSPEC_FFMISM and UNSPEC_FLMISM.
4498 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4499 for ffb, ffmism and flmism.
4500 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4501 (unspec_ffmism): Ditto.
4502 (unspec_flmism): Ditto.
4503 (nds32_expand_builtin_impl): Check if string extension is available.
4504 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4505 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4507 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4510 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4513 * lra-assigns.c (assign_by_spills): Return a flag of reload
4514 assignment failure. Do not process the reload assignment
4515 failures. Do not spill other reload pseudos if they has the same
4517 (lra_assign): Add a return arg. Set up from the result of
4518 assign_by_spills call.
4519 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4520 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4521 usage_insns if it is not NULL.
4522 (spill_hard_reg_in_range): New function.
4523 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4524 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4525 function prototypes.
4526 (lra_assign): Change prototype.
4527 * lra.c (lra): Add code to deal with fails by splitting hard reg
4530 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4533 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4535 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4538 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4539 (dw_val_node): Add val_symbolic_view.
4540 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4541 (symview_upper_bound): New.
4542 (new_line_info_table): Initialize symviews_since_reset.
4543 (dwarf2out_source_line): Count symviews_since_reset and set
4544 symview_upper_bound.
4545 (dw_val_equal_p): Handle symview.
4546 (add_AT_symview): New.
4547 (print_dw_val): Handle symview.
4548 (attr_checksum, attr_checksum_ordered): Likewise.
4549 (same_dw_val_p, size_of_die): Likewise.
4550 (value_format, output_die): Likewise.
4551 (add_high_low_attributes): Use add_AT_symview for entry_view.
4552 (dwarf2out_finish): Reset symview_upper_bound, clear
4555 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4558 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4559 Add strict argument and use it.
4560 (rs6000_split_multireg_move): Update for new strict argument.
4561 (mem_operand_gpr): Disallow all non-offsettable addresses.
4562 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4564 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4567 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4568 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4569 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4572 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4573 decl, use remap_type if we want to use the type.
4575 2018-03-09 Martin Sebor <msebor@redhat.com>
4577 PR tree-optimization/84526
4578 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4580 (builtin_access::generic_overlap): Be prepared to handle non-array
4583 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4585 PR rtl-optimization/84682
4586 * lra-constraints.c (process_address_1): Check is_address flag
4587 for address constraints.
4588 (process_alt_operands): Likewise.
4589 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4590 preprocess_constraints.
4591 * recog.h (preprocess_constraints): Add oploc parameter.
4593 * recog.c (preprocess_constraints): Test address_operand for
4594 CT_ADDRESS constraints.
4596 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4599 * lra-assigns.c (assign_by_spills): Return a flag of reload
4600 assignment failure. Do not process the reload assignment
4601 failures. Do not spill other reload pseudos if they has the same
4603 (lra_assign): Add a return arg. Set up from the result of
4604 assign_by_spills call.
4605 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4606 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4607 usage_insns if it is not NULL.
4608 (spill_hard_reg_in_range): New function.
4609 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4610 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4611 function prototypes.
4612 (lra_assign): Change prototype.
4613 * lra.c (lra): Add code to deal with fails by splitting hard reg
4616 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4619 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4620 Accept complain bool parameter. Only emit errors if it is true.
4621 (arm_parse_cpu_option_name): Likewise.
4622 (arm_target_thumb_only): Adjust callers of the above.
4623 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4624 prototype to take a default true bool parameter.
4625 (arm_parse_arch_option_name): Likewise.
4627 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4628 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4632 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4633 * configure: Regenerate.
4634 * configure.ac ("linker --version-script option"): New.
4635 ("linker soname option"): New.
4637 2018-03-09 Richard Biener <rguenther@suse.de>
4639 PR tree-optimization/84775
4640 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4641 immediate uses of predicate stmts and mark them modified.
4644 PR tree-optimization/84178
4645 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4647 (version_loop_for_if_conversion): Delay update_ssa call.
4648 (tree_if_conversion): Delay update_ssa until after predicate
4651 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4654 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4655 when the function accesses prior frames.
4657 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4660 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4661 gen_llsym, otherwise call maybe_gen_llsym.
4664 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4665 has ',' character inside of it.
4667 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4670 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4671 as clobbering CC_REGNUM.
4673 2018-03-08 Richard Biener <rguenther@suse.de>
4676 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4677 (follow_copies_to_constant): Do not follow SSA names registered
4680 2018-03-08 Richard Biener <rguenther@suse.de>
4682 PR tree-optimization/84178
4683 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4685 (version_loop_for_if_conversion): Delay update_ssa call.
4686 (tree_if_conversion): Delay update_ssa until after predicate
4689 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4691 PR tree-optimization/84178
4692 * tree-if-conv.c (release_bb_predicate): Remove the
4693 the assertion that the stmts have NULL use_ops.
4694 Discard the statements, asserting that they haven't
4695 yet been added to a BB.
4697 2018-03-08 Richard Biener <rguenther@suse.de>
4699 PR tree-optimization/84746
4700 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4701 (phi_translate): Pass in destination ANTIC_OUT set.
4702 (phi_translate_1): Likewise. For a simplified result lookup
4703 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4704 (phi_translate_set): Adjust.
4705 (do_pre_regular_insertion): Likewise.
4706 (do_pre_partial_partial_insertion): Likewise.
4708 2018-03-08 Martin Liska <mliska@suse.cz>
4710 PR gcov-profile/84735
4711 * doc/gcov.texi: Document usage of profile files.
4712 * gcov-io.h: Document changes in the format.
4714 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4718 * dwarf2out.c (struct dw_line_info_table): Update comments for
4720 (FORCE_RESET_NEXT_VIEW): New.
4721 (FORCE_RESETTING_VIEW_P): New.
4722 (RESETTING_VIEW_P): Check for -1 too.
4723 (ZERO_VIEW_P): Likewise.
4724 (new_line_info_table): Force-reset next view.
4725 (dwarf2out_begin_function): Likewise.
4726 (dwarf2out_source_line): Simplify zero_view_p initialization.
4727 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4728 view directly. Omit view when omitting .loc at line 0.
4730 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4732 PR tree-optimization/84740
4733 * tree-switch-conversion.c (process_switch): Call build_constructors
4734 only if info.phi_count is non-zero.
4736 PR tree-optimization/84739
4737 * tree-tailcall.c (find_tail_calls): Check call arguments against
4738 DECL_ARGUMENTS (current_function_decl) rather than
4739 DECL_ARGUMENTS (func) when checking for tail recursion.
4741 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4743 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4744 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4745 Volker Reichelt's entry and add entries for people that perform
4746 GCC fuzzy testing and report numerous bugs.
4748 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4751 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4752 readonly data in sdata, if that is disabled.
4753 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4754 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4755 -mreadonly-in-sdata option.
4757 2018-03-07 Martin Sebor <msebor@redhat.com>
4759 PR tree-optimization/84468
4760 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4761 basic block when looking for nul assignment.
4763 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4766 * except.h (output_function_exception_table): Adjust prototype.
4767 * except.c (output_function_exception_table): Remove FNNAME parameter
4768 and add SECTION parameter. Ouput one part of the table at a time.
4769 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4770 the first part of the exception table and emit unwind directives.
4771 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4772 (i386_pe_seh_cold_init): Likewise.
4773 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4774 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4775 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4776 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4777 * config/i386/winnt.c: Include except.h.
4778 (struct seh_frame_state): Add reg_offset, after_prologue and
4779 in_cold_section fields.
4780 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4781 (i386_pe_seh_cold_init): New function.
4782 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4783 to seh->in_cold_section.
4784 (seh_emit_push): Record the offset of the push.
4785 (seh_emit_save): Record the offet of the save.
4786 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4787 Test seh->after_prologue to disregard the epilogue.
4788 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4789 (i386_pe_end_cold_function): New function.
4791 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4794 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4795 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4798 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4800 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4801 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4804 * multiple_target.c: Include tree-inline.h and intl.h.
4805 (expand_target_clones): Diagnose and fail if node->definition and
4806 !tree_versionable_function_p (node->decl).
4808 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4810 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4812 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4813 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4814 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4816 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4819 * combine.c (try_combine): Use reg_or_subregno instead of handling
4820 just paradoxical SUBREGs and REGs.
4822 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4824 * config/arc/arc.c (arc_finalize_pic): Remove function.
4825 (arc_must_save_register): We use single base PIC register, remove
4826 checks to save/restore the PIC register.
4827 (arc_expand_prologue): Likewise.
4828 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4830 (arc_verify_short): Likewise.
4831 (arc_attr_type): Likewise.
4832 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4833 (walk_stores): Likewise.
4834 (arc_address_cost): Make it static.
4835 (arc_verify_short): Likewise.
4836 (branch_dest): Likewise.
4837 (arc_attr_type): Likewise.
4838 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4839 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4840 (arc_final_prescan_insn): Remove inserting the nops due to
4841 hardware hazards. It is done in reorg step.
4842 (insn_length_variant_t): Remove.
4843 (insn_length_parameters_t): Likewise.
4844 (arc_insn_length_parameters): Likewise.
4845 (arc_get_insn_variants): Likewise.
4846 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4848 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4851 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4854 PR tree-optimization/84687
4855 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4857 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4859 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4861 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4862 Rename to ppc_speculation_barrier.
4863 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4864 __builtin_ppc_speculation_barrier.
4866 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4869 * combine.c (combine_simplify_rtx): Don't try to simplify if
4870 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4873 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4875 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4876 to 32 bytes when compiling for POWER9.
4878 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4881 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4882 regparm >= 3 with no arg reg available also for calls with
4883 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4886 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4888 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4890 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4893 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4895 2018-03-05 Richard Biener <rguenther@suse.de>
4897 PR tree-optimization/84486
4898 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4899 When inserting a __builtin_assume_aligned call set the LHS
4900 SSA name alignment info accordingly.
4902 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4904 PR tree-optimization/84114
4905 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4906 Avoid reassociation of FLOAT_MODE addition.
4908 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4910 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4911 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4912 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4913 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4914 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4915 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4916 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4918 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4919 __builtin_ia32_wbinvd): New builtins.
4920 (SPECIAL_ARGS2): New.
4921 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4922 (SPECIAL_ARGS2): New.
4923 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4924 (ix86_valid_target_attribute_inner_p): Ditto.
4925 (ix86_init_mmx_sse_builtins): Add special_args2.
4926 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4927 TARGET_WBNOINVD_P): New.
4928 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4929 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4930 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4931 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4932 * config/i386/pconfigintrin.h: New file.
4933 * config/i386/wbnoinvdintrin.h: Ditto.
4934 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4935 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4937 2018-03-05 Richard Biener <rguenther@suse.de>
4939 PR tree-optimization/84670
4940 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4942 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4943 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4944 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4945 doesn't grow if all successors (recursively) were visited at least
4948 2018-03-05 Richard Biener <rguenther@suse.de>
4950 PR tree-optimization/84650
4951 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4952 if executed in the loop pipeline.
4954 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4956 * doc/configfiles.texi (Configuration Files): Move info about
4957 conditionalizing $target-protos.h to...
4958 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4959 differs from $target-protos.h.
4961 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4962 Chung-Ju Wu <jasonwucj@gmail.com>
4964 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4965 * config/nds32/nds32-multiple.md (setmemsi): Define.
4966 * config/nds32/nds32-memory-manipulation.c
4967 (nds32_gen_dup_4_byte_to_word_value): New.
4968 (emit_setmem_word_loop): New.
4969 (emit_setmem_byte_loop): New.
4970 (nds32_expand_setmem_loop): New.
4971 (nds32_expand_setmem_loop_v3m): New.
4972 (nds32_expand_setmem_unroll): New.
4973 (nds32_expand_setmem): New.
4975 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4976 Chung-Ju Wu <jasonwucj@gmail.com>
4978 * config/nds32/nds32-memory-manipulation.c
4979 (nds32_emit_load_store): New.
4980 (nds32_emit_post_inc_load_store): New.
4981 (nds32_emit_mem_move): New.
4982 (nds32_emit_mem_move_block): New.
4983 (nds32_expand_movmemsi_loop_unknown_size): New.
4984 (nds32_expand_movmemsi_loop_known_size): New.
4985 (nds32_expand_movmemsi_loop): New.
4986 (nds32_expand_movmemsi_unroll): New.
4987 (nds32_expand_movmemqi): Rename ...
4988 (nds32_expand_movmemsi): ... to this.
4989 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4990 (movmemsi): ... to this.
4991 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4992 (nds32_expand_movmemsi): ... to this.
4994 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4995 Monk Chiang <sh.chiang04@gmail.com>
4996 Chung-Ju Wu <jasonwucj@gmail.com>
4998 * config/nds32/nds32-protos.h
4999 (nds32_expand_load_multiple): New arguments.
5000 (nds32_expand_store_multiple): Ditto.
5001 (nds32_valid_multiple_load_store): Rename ...
5002 (nds32_valid_multiple_load_store_p): ... to this.
5003 * config/nds32/nds32-memory-manipulation.c
5004 (nds32_expand_load_multiple): Refine implementation.
5005 (nds32_expand_store_multiple): Ditto.
5006 * config/nds32/nds32-multiple.md
5007 (load_multiple): Update nds32_expand_load_multiple interface.
5008 (store_multiple): Update nds32_expand_store_multiple interface.
5009 * config/nds32/nds32-predicates.c
5010 (nds32_valid_multiple_load_store): Rename ...
5011 (nds32_valid_multiple_load_store_p): ... to this and refine
5013 * config/nds32/predicates.md
5014 (nds32_load_multiple_and_update_address_operation): New predicate.
5015 (nds32_store_multiple_and_update_address_operation): New predicate.
5017 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5018 Chung-Ju Wu <jasonwucj@gmail.com>
5020 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5021 (combo): New attribute.
5022 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5024 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5026 * config/nds32/nds32.opt: Change -mcmodel= default value.
5028 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5029 Monk Chiang <sh.chiang04@gmail.com>
5030 Chung-Ju Wu <jasonwucj@gmail.com>
5032 * config/nds32/constants.md (unspec_element): New enum.
5033 * config/nds32/constraints.md (Umw): New constraint.
5034 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5035 * config/nds32/nds32-intrinsic.md: Likewise.
5036 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5037 (nds32_valid_smw_lwm_base_p): New.
5038 (nds32_output_smw_single_word): New.
5039 (nds32_output_lmw_single_word): New.
5040 (nds32_expand_unaligned_load): New.
5041 (nds32_expand_unaligned_store): New.
5042 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5043 (nds32_output_smw_single_word): Declare.
5044 (nds32_output_lmw_single_word): Declare.
5045 (nds32_expand_unaligned_load): Declare.
5046 (nds32_expand_unaligned_store): Declare.
5047 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5048 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5049 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5050 NDS32_BUILTIN_UASTORE_DW.
5051 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5054 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5055 Kito Cheng <kito.cheng@gmail.com>
5056 Chung-Ju Wu <jasonwucj@gmail.com>
5058 * config/nds32/nds32-intrinsic.c
5059 (nds32_expand_builtin_null_ftype_reg): Delete.
5060 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5061 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5062 (nds32_read_argument): New.
5063 (nds32_legitimize_target): Ditto.
5064 (nds32_legitimize_argument): Ditto.
5065 (nds32_check_constant_argument): Ditto.
5066 (nds32_expand_unop_builtin): Ditto.
5067 (nds32_expand_unopimm_builtin): Ditto.
5068 (nds32_expand_binop_builtin): Ditto.
5069 (nds32_builtin_decl_impl): Ditto.
5070 (builtin_description): Ditto.
5071 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5072 (nds32_init_builtins_impl): Ditto.
5073 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5074 (nds32_builtin_decl): New.
5075 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5076 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5078 2018-03-02 Jeff Law <law@redhat.com>
5080 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5081 (redundant_insn, fill_simple_delay_slots): Likewise.
5082 (fill_slots_from_thread): Likewise.
5083 * resource.c (mark_referenced_resources): Likewise.
5084 (mark_set_resources, find_dead_or_set_registers): Likewise.
5086 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5088 * substring-locations.h (format_warning_va): Formatting fix for
5090 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5092 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5093 * substring-locations.c: Include intl.h.
5094 (format_warning_va): Turned into small wrapper around
5095 format_warning_n_va, renamed to ...
5096 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5097 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5099 (format_warning_at_substring_n): New function.
5100 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5101 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5102 format_warning_at_substring with just a shorter name instead of
5103 const function pointer.
5104 (fmtwarn_n): New function.
5105 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5106 appropriate, get rid of all the fmtstr temporaries, move conditionals
5107 with G_() wrapped string literals directly into fmtwarn arguments,
5108 cast dir.len to (int), formatting fixes.
5110 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5112 * doc/invoke.texi: Remove "Cilk Plus" references.
5114 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5115 Richard Biener <rguenther@suse.de>
5118 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5119 for error or warning attributes if CALL_FROM_THUNK_P is set.
5122 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5125 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5126 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5129 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5130 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5132 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5133 instead of -1U in last predictors element's probability member.
5135 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5138 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5139 arguments if they are comparable.
5141 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5143 PR tree-optimization/84634
5144 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5145 masks and masked_loop_p with a single loop_masks, making sure it's
5146 null for bb vectorization.
5148 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5150 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5151 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5152 loop->force_vectorize to check whether there is no alias.
5154 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5157 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5159 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5161 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5162 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5163 instead of a loop around prev_real_insn.
5164 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5168 * config/i386/i386.c (ix86_print_operand): Use conditional
5169 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5172 2018-03-02 Richard Biener <rguenther@suse.de>
5174 PR tree-optimization/84427
5175 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5176 (bitmap_set_subtract_values): Rewrite to handle multiple
5179 (prune_clobbered_mems): Likewise.
5180 (phi_translate): Take edge instead of pred/phiblock.
5181 (phi_translate_1): Likewise.
5182 (phi_translate_set): Likewise. Insert all translated
5183 exprs for a value into the set, keeping possibly multiple
5184 expressions per value.
5185 (compute_antic_aux): Adjust for phi_translate changes.
5186 When intersecting union the expressions and prune those
5187 not in the final value set, keeping possibly multiple
5188 expressions per value. Do not use value-insertion
5189 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5190 all expressions. Add verification that the value-sets
5191 only shrink during iteration.
5192 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5193 (do_pre_regular_insertion): Likewise.
5194 (do_pre_partial_partial_insertion): Likewise.
5196 2018-03-02 Richard Biener <rguenther@suse.de>
5199 * config/darwin.c (saved_debug_info_level): New static global.
5200 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5201 (darwin_asm_lto_end): Restore debug info generation settings.
5203 2018-03-01 Martin Liska <mliska@suse.cz>
5206 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5209 2018-03-01 Richard Biener <rguenther@suse.de>
5212 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5213 type annotation with LTO when debug was disabled at compile-time.
5215 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5217 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5219 (mips_final_postscan_insn): Likewise.
5221 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5223 PR rtl-optimization/84528
5224 * alias.c (init_alias_target): Add commentary.
5225 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5226 a unique base value if the frame pointer is not eliminated
5227 to the stack pointer.
5229 2018-03-01 Tom de Vries <tom@codesourcery.com>
5231 PR rtl-optimization/83327
5232 * lra-int.h (hard_regs_spilled_into): Declare.
5233 * lra.c (hard_regs_spilled_into): Define.
5234 (init_reg_info): Init hard_regs_spilled_into.
5235 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5236 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5237 (process_bb_lives): Handle hard_regs_spilled_into.
5238 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5239 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5241 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5243 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5244 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5245 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5246 * config/rs6000/aix72.h: New file.
5248 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5250 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5251 instead of warning_at with conditional singular and plural messages
5255 * stor-layout.c (update_alignment_for_field): For
5256 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5257 && !DECL_PACKED (field), do the alignment update, just use
5258 only desired_align instead of MAX (type_align, desired_align)
5260 (place_field): Don't do known_align < desired_align handling
5261 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5262 is non-NULL, instead do it after rli->prev_field handling and
5263 only if not within a bitfield word. For DECL_PACKED (field)
5264 use type_align of BITS_PER_UNIT.
5266 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5268 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5269 superfluous parentheses and trailing spaces.
5271 2018-02-28 Richard Biener <rguenther@suse.de>
5273 PR tree-optimization/84584
5274 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5275 SCoPs with fake exit edge.
5277 2018-02-28 Martin Liska <mliska@suse.cz>
5280 * timevar.c (timer::print): Fix format to properly print 100%
5283 2018-02-28 Richard Biener <rguenther@suse.de>
5286 * genmatch.c (capture_info::walk_match): Do not mark
5287 captured expressions without operands as expr_p given
5288 they act more like predicates and should be subject to
5289 "lost tail" side-effect preserving.
5291 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5293 PR rtl-optimization/81611
5294 * auto-inc-dec.c (attempt_change): Move dead note from
5295 mem_insn if it's the next use of regno
5296 (find_address): Take address use of reg holding
5297 non-incremented value. Add parm to limit search to the named
5299 (merge_in_block): Attempt to use a mem insn that is the next
5300 use of the original regno.
5302 2018-02-27 Martin Sebor <msebor@redhat.com>
5305 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5306 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5308 2018-02-27 Martin Sebor <msebor@redhat.com>
5310 PR translation/84207
5311 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5312 n argument to unsigned HOST_WIDE_INT.
5313 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5314 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5315 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5316 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5318 2018-02-27 Richard Biener <rguenther@suse.de>
5320 PR tree-optimization/84512
5321 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5322 Do not use the estimate returned from record_stmt_cost for
5323 the scalar iteration cost but sum properly using add_stmt_cost.
5325 2018-02-27 Richard Biener <rguenther@suse.de>
5327 PR tree-optimization/84466
5328 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5329 Adjust last change to less strictly validate use operands.
5331 2018-02-27 Martin Liska <mliska@suse.cz>
5333 PR gcov-profile/84548
5334 * gcov.c (process_file): Allow partial overlap and consider it
5335 also as group functions.
5336 (output_lines): Properly calculate range of lines for a group.
5338 2018-02-27 Martin Liska <mliska@suse.cz>
5340 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5341 'ggc' suffixes. Change first column width.
5342 (timer::print): Fix formatting of the column.
5344 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5346 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5347 preserve inline entry blocks for the sake of debug inline
5348 entry point markers alone.
5349 (remove_unused_locals): Suggest in comments a better place to
5350 force the preservation of inline entry blocks that are
5351 otherwise unused, but do not preserve them.
5353 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5355 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5357 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5360 * config/i386/constraints.md (Bs): Replace
5361 ix86_indirect_branch_register with
5362 TARGET_INDIRECT_BRANCH_REGISTER.
5364 * config/i386/i386.md (indirect_jump): Likewise.
5365 (tablejump): Likewise.
5366 (*sibcall_memory): Likewise.
5367 (*sibcall_value_memory): Likewise.
5368 Peepholes of indirect call and jump via memory: Likewise.
5369 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5370 (*sibcall_value_GOT_32): Likewise.
5371 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5372 (GOT_memory_operand): Likewise.
5373 (call_insn_operand): Likewise.
5374 (sibcall_insn_operand): Likewise.
5375 (GOT32_symbol_operand): Likewise.
5376 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5378 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5380 PR rtl-optimization/83496
5381 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5382 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5383 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5384 redundant insn, if any.
5385 (relax_delay_slots): Likewise.
5386 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5388 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5390 PR tree-optimization/83965
5391 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5392 that grouped statements are part of a reduction chain. Return
5393 true if the statement is not marked as a reduction itself but
5395 (vect_recog_dot_prod_pattern): Don't check whether the statement
5396 is part of a group here.
5397 (vect_recog_sad_pattern): Likewise.
5398 (vect_recog_widen_sum_pattern): Likewise.
5400 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5403 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5405 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5408 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5410 (ix86_output_indirect_function_return): New prototype.
5411 (ix86_split_simple_return_pop_internal): Likewise.
5412 * config/i386/i386.c (indirect_return_via_cx): New.
5413 (indirect_return_via_cx_bnd): Likewise.
5414 (indirect_thunk_name): Handle return va CX_REG.
5415 (output_indirect_thunk_function): Create alias for
5416 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5417 (ix86_output_indirect_jmp): Remove the bool argument.
5418 (ix86_output_indirect_function_return): New function.
5419 (ix86_split_simple_return_pop_internal): Likewise.
5420 * config/i386/i386.md (*indirect_jump): Don't pass false
5421 to ix86_output_indirect_jmp.
5422 (*tablejump_1): Likewise.
5423 (simple_return_pop_internal): Change it to define_insn_and_split.
5424 Call ix86_split_simple_return_pop_internal to split it for
5426 (simple_return_indirect_internal): Call
5427 ix86_output_indirect_function_return instead of
5428 ix86_output_indirect_jmp.
5430 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5433 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5434 memset and value initialization afterwards.
5436 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5438 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5440 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5443 * common/config/aarch64/aarch64-common.c
5444 (aarch_option_optimization_table[]): Switch
5445 off fomit-frame-pointer
5447 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5448 Chung-Ju Wu <jasonwucj@gmail.com>
5450 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5452 (store_multiple): Ditto.
5454 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5456 * config.gcc: Add --with-cpu support for nds32 target.
5457 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5458 * config/nds32/nds32.opt: Add -mcpu= option.
5460 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5462 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5463 isel=yes): Warn for these deprecated options.
5465 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5467 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5468 ISA_2_5_MASKS_EMBEDDED.
5470 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5472 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5473 p->max as pointers rather than using iterative_hash_expr.
5475 2018-02-23 Carl Love <cel@us.ibm.com>
5477 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5478 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5480 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5481 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5482 P8V_BUILTIN_VEC_VUNSIGNED2.
5484 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5487 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5488 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5490 * lra-constraints.c (curr_insn_transform): Set up
5491 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5493 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5494 * lra-eliminations.c (spill_pseudos): Ditto.
5495 (process_insn_for_elimination): Ditto.
5496 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5497 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5498 LRA_NON_CLOBBERED_ALT.
5500 2018-02-22 Martin Sebor <msebor@redhat.com>
5502 PR tree-optimization/84480
5503 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5504 to maybe_diag_stxncpy_trunc. Call it.
5505 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5506 from gimple_fold_builtin_strcpy. Print inlining stack.
5507 (handle_builtin_stxncpy): Print inlining stack.
5508 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5510 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5513 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5514 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5515 and -fcheck-pointer-bounds are used together.
5516 (indirect_thunk_prefix): New enum.
5517 (indirect_thunk_need_prefix): New function.
5518 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5519 "_nt" instead of "_bnd" for NOTRACK prefix.
5520 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5521 (output_indirect_thunk_function): Likewise.
5523 (ix86_code_end): Update output_indirect_thunk_function calls.
5524 (ix86_output_indirect_branch_via_reg): Replace
5525 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5526 (ix86_output_indirect_branch_via_push): Likewise.
5527 (ix86_output_function_return): Likewise.
5528 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5529 incompatible with -fcf-protection=branch and
5530 -fcheck-pointer-bounds.
5532 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5535 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5536 Change gcc_assert call to output_operand_lossage.
5538 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5540 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5542 2018-02-22 DJ Delorie <dj@redhat.com>
5543 Sebastian Perta <sebastian.perta@renesas.com>
5544 Oleg Endo <olegendo@gcc.gnu.org>
5546 * config/rx/rx.c (rx_rtx_costs): New function.
5547 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5549 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5551 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5553 2018-02-22 Martin Liska <mliska@suse.cz>
5556 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5557 Add "native" as a possible value.
5559 2018-02-22 Martin Liska <mliska@suse.cz>
5562 * config/i386/i386.c (ix86_option_override_internal):
5563 Add "native" as a possible value for -march and -mtune.
5565 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5568 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5569 to all type variants.
5571 PR tree-optimization/84503
5572 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5573 width as info->bitpos + info->bitsize - start.
5574 (merged_store_group::merge_overlapping): Simplify width computation.
5575 (check_no_overlap): New function.
5576 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5577 start + width and last_order of the group, fail if check_no_overlap
5579 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5580 to group if check_no_overlap fails.
5582 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5584 * config/rs6000/altivec.md: Delete contraint arguments to
5585 define_expand, define_split, and define_peephole2, and in
5586 define_insn_and_split if always unused.
5587 * config/rs6000/darwin.md: Ditto.
5588 * config/rs6000/dfp.md: Ditto.
5589 * config/rs6000/rs6000.md: Ditto.
5590 * config/rs6000/sync.md: Ditto.
5591 * config/rs6000/vector.md: Ditto.
5592 * config/rs6000/vsx.md: Ditto.
5594 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5596 * config/rs6000/altivec.md: Write output control strings as braced
5597 blocks instead of double-quoted strings.
5598 * config/rs6000/darwin.md: Ditto.
5599 * config/rs6000/rs6000.md: Ditto.
5600 * config/rs6000/vector.md: Ditto.
5601 * config/rs6000/vsx.md: Ditto.
5603 2018-02-21 Jason Merrill <jason@redhat.com>
5605 PR c++/84314 - ICE with templates and fastcall attribute.
5606 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5608 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5610 * ipa-cp.c (determine_versionability): Fix comment typos.
5612 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5615 * ipa-cp.c (determine_versionability): Do not version functions caling
5618 2018-02-21 Martin Liska <mliska@suse.cz>
5621 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5622 Add "native" as a possible value.
5623 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5624 the macro when native cpu detection is available.
5626 2018-02-21 Martin Liska <mliska@suse.cz>
5629 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5630 Add "native" as a possible value.
5631 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5632 when native cpu detection is available.
5634 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5635 Martin Sebor <msebor@redhat.com>
5637 PR tree-optimization/84478
5638 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5640 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5641 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5642 support which is conservatively correct, for 2 only stay conservative
5643 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5644 argument to the 2 argument get_range_strlen, adjust 6 arg
5645 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5647 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5648 (gimple_fold_builtin_strlen): Pass true as last argument to
5651 2018-02-20 Martin Sebor <msebor@redhat.com>
5654 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5655 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5656 (builtin_memref::builtin_memref): Factor out parts into
5657 set_base_and_offset and call it.
5659 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5662 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5663 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5664 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5665 search at the associated MODE_INT.
5667 2018-02-20 Jeff Law <law@redhat.com>
5670 PR tree-optimization/81592
5672 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5673 for range data rather than using global data.
5674 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5675 range data rather than using global data.
5676 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5677 pass it to children as needed.
5678 (struct directive::fmtresult): Similarly.
5679 (struct directive::set_width): Similarly.
5680 (struct directive::set_precision): Similarly.
5681 (format_integer, format_directive, parse_directive): Similarly.
5682 (format_none): Accept unnamed vr_values parameter.
5683 (format_percent, format_floating, format_character): Similarly.
5684 (format_string, format_plain): Similarly.
5685 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5686 the EVRP range analyzer for range data rather than using global data.
5687 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5688 gimple-ssa-evrp-analyze.h
5689 (class sprintf_dom_walker): Add after_dom_children member function.
5690 Add evrp_range_analyzer member.
5691 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5692 range analyzer as needed.
5693 (sprintf_dom_walker::after_dom_children): New member function.
5694 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5696 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5697 (evrp_range_analyzer::pop_to_marker): Likewise.
5699 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5701 PR tree-optimization/84419
5702 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5703 with the required type if its current type is compatible but
5706 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5709 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5710 after vectorization.
5712 2018-02-20 Martin Liska <mliska@suse.cz>
5715 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5716 possible values if we don't have a hint.
5718 2018-02-20 Martin Liska <mliska@suse.cz>
5722 * final.c (shorten_branches): Build align_tab array with one
5724 * opts.c (finish_options): Add alignment option limit check.
5725 (MAX_CODE_ALIGN): Likewise.
5726 (MAX_CODE_ALIGN_VALUE): Likewise.
5727 * doc/invoke.texi: Document maximum allowed option value for
5728 all -falign-* options.
5730 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5733 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5734 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5735 * var-tracking.c (emit_note_insn_var_location): Remove all references
5736 to NOTE_INSN_CALL_ARG_LOCATION.
5737 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5738 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5739 Use copy_rtx_if_shared.
5740 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5741 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5742 (dwarf2out_var_location): Remove handling of
5743 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5745 * final.c (final_scan_insn): Remove all references to
5746 NOTE_INSN_CALL_ARG_LOCATION.
5747 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5748 before dumping final insns.
5749 * except.c (emit_note_eh_region_end): Remove all references to
5750 NOTE_INSN_CALL_ARG_LOCATION.
5751 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5752 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5753 * config/arc/arc.c (hwloop_optimize): Likewise.
5754 * config/arm/arm.c (create_fix_barrier): Likewise.
5755 * config/s390/s390.c (s390_chunkify_start): Likewise.
5756 * config/sh/sh.c (find_barrier): Likewise.
5757 * config/i386/i386.c (rest_of_insert_endbranch,
5758 ix86_seh_fixup_eh_fallthru): Likewise.
5759 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5760 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5761 * config/frv/frv.c (frv_function_prologue): Likewise.
5762 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5764 (note_outside_basic_block_p): Remove all references to
5765 NOTE_INSN_CALL_ARG_LOCATION.
5766 * gengtype.c (adjust_field_rtx_def): Likewise.
5767 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5769 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5770 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5773 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5776 PR tree-optimization/84452
5777 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5778 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5781 2018-02-19 Martin Liska <mliska@suse.cz>
5784 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5786 2018-02-19 Martin Liska <mliska@suse.cz>
5787 Richard Sandiford <richard.sandiford@linaro.org>
5789 PR tree-optimization/82491
5790 * gimple-fold.c (get_base_constructor): Make earlier bail out
5793 2018-02-19 Carl Love <cel@us.ibm.com>
5795 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5796 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5798 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5799 P8V_BUILTIN_VEC_NEG.
5801 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5803 * config/rl78/rl78.md (movdf): New define expand.
5805 2018-02-19 Martin Liska <mliska@suse.cz>
5808 * doc/invoke.texi: Fix typo.
5809 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5811 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5813 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5814 handle rs6000_single_float and rs6000_double_float specially for
5817 2018-02-16 Jeff Law <law@redhat.com>
5819 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5822 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5825 * ipa-inline.c (inline_small_functions): Fix a typo.
5827 2018-02-16 Nathan Sidwell <nathan@acm.org>
5829 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5831 2018-02-16 Carl Love <cel@us.ibm.com>
5833 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5834 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5835 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5836 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5837 expansion to P8V_BUILTIN_VEC_FLOAT2.
5839 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5841 PR rtl-optimization/70023
5842 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5843 src_regno into account.
5845 2018-02-16 Carl Love <cel@us.ibm.com>
5847 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5848 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5849 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5850 * config/rs6000/rs6000.c: Remove case statements for
5851 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5852 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5853 and P9V_BUILTIN_VEC_VINSERT4B.
5854 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5855 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5856 * config/rs6000/vsx.md:
5857 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5860 2018-02-16 Carl Love <cel@us.ibm.com>
5862 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5864 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5866 * config/rs6000/rs6000-c.c: Add the definitions for
5867 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5868 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5869 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5870 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5871 definition for insert4b and define insn *insert3b_internal.
5872 * doc/extend.texi: Add documentation for vec_extract4b.
5874 2018-02-16 Nathan Sidwell <nathan@acm.org>
5876 * doc/extend.texi (Backwards Compatibility): Mention friend
5877 injection. Note for-scope is deprecated.
5878 * doc/invoke.texi (-ffriend-injection): Deprecate.
5880 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5882 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5883 that moved to I2, also allow destinations that are a paradoxical
5884 subreg (instead of a normal reg).
5886 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5889 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5892 2018-02-16 Richard Biener <rguenther@suse.de>
5894 PR tree-optimization/84037
5895 PR tree-optimization/84016
5897 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5898 Adjust vec_construct for the fact we need additional higher latency
5899 128bit inserts for AVX256 and AVX512 vector builds.
5900 (ix86_add_stmt_cost): Scale vector construction cost for
5903 2018-02-16 Richard Biener <rguenther@suse.de>
5905 PR tree-optimization/84417
5906 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5907 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5908 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5910 2018-02-16 Martin Liska <mliska@suse.cz>
5913 * internal-fn.def (ASAN_CHECK): Set proper flags.
5914 (ASAN_MARK): Likewise.
5916 2018-02-16 Julia Koval <julia.koval@intel.com>
5918 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5919 from PTA_CANNONLAKE.
5921 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5924 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5925 Use ++iter rather than iter++ for std::list iterators.
5926 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5927 defer deleting them until all nodes in the forest are processed. Do
5928 free even leaf nodes. Change to_process into auto_vec.
5931 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5932 * vec.h (vec_default_construct): Use memset instead of placement new
5933 if BROKEN_VALUE_INITIALIZATION is defined.
5934 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5935 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5938 PR rtl-optimization/83723
5939 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5940 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5941 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5943 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5945 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5947 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5949 PR rtl-optimization/81443
5950 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5951 from inner REGs to paradoxical SUBREGs.
5953 2018-02-16 Richard Biener <rguenther@suse.de>
5955 PR tree-optimization/84399
5956 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5957 For operands we can analyze at their definition make sure we can
5958 analyze them at each use as well.
5960 2018-02-16 Richard Biener <rguenther@suse.de>
5962 PR tree-optimization/84190
5963 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5964 volatile accesses if the decl isn't volatile.
5966 2018-02-15 Jason Merrill <jason@redhat.com>
5968 PR c++/84314 - ICE with templates and fastcall attribute.
5969 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5970 TYPE_CANONICAL on an existing type.
5972 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5974 PR tree-optimization/84383
5975 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5976 dstoff nor call operand_equal_p if dstbase is NULL.
5978 PR tree-optimization/84334
5979 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5980 also a CONSTANT_CLASS_P, punt.
5982 2018-02-14 Jim Wilson <jimw@sifive.com>
5984 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5985 first SMALL_OPERAND check. New local min_second_step. Move assert
5986 to where locals are set. Add TARGET_RVC support.
5987 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5989 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5991 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5993 2018-02-14 Martin Sebor <msebor@redhat.com>
5995 PR tree-optimization/83698
5996 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5997 arrays constrain the offset range to their bounds.
5998 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5999 (builtin_access::overlap): Avoid setting the size of overlap if it's
6001 (maybe_diag_overlap): Also consider arrays when deciding what values
6002 of offsets to include in diagnostics.
6004 2018-02-14 Martin Sebor <msebor@redhat.com>
6007 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6008 that correspond to the kind of a declaration.
6010 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6013 * config/pa/pa.md: Load address of PIC label using the linkage table
6014 if the label is nonlocal.
6016 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6018 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6019 warning message if user requests -maltivec=be.
6020 * doc/invoke.texi: Document deprecation of -maltivec=be.
6022 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6025 * config/rs6000/rs6000-c.c: Update definitions for
6026 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6027 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6029 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6032 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6033 add _get_ssp intrinsics. Remove argument from
6034 __builtin_ia32_rdssp[d|q].
6035 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6036 * config/i386/i386-builtin.def: Remove argument from
6037 __builtin_ia32_rdssp[d|q].
6038 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6039 ix86_expand_special_args_builtin for _rdssp[d|q].
6040 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6041 Clear register before usage.
6042 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6043 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6045 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6047 PR tree-optimization/84357
6048 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6049 operand 1 of an ARRAY_REF too.
6051 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6054 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6055 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6057 (set_of_reg): New struct.
6058 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6059 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6060 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6062 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6063 Split into bitclr, bitset, bitinvert patterns if appropriate.
6064 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6065 use rx_fuse_in_memory_bitop.
6066 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6067 to named insn, correct maximum insn length.
6069 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6072 * machmode.def: Define a complex mode for PARTIAL_INT.
6073 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6075 * doc/rtl.texi: Document CSPImode.
6076 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6078 (msp430_hard_regno_nregs_with_padding): Likewise.
6080 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6083 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6085 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6087 PR rtl-optimization/84169
6088 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6089 we generated a parallel as new i3 and we split that to new i2 and i3
6090 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6091 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6092 those to i2, not i1. Partially rewrite this scan code.
6094 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6097 * stor-layout.c (place_field): For variable length fields, adjust
6098 offset_align afterwards not just based on the field's alignment,
6099 but also on the size.
6102 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6103 of exps and logs in the use_exp2 case.
6105 2018-02-13 Jeff Law <law@redhat.com>
6107 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6110 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6113 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6117 * common.opt (gas-loc-support, gas-locview-support): New.
6118 (ginline-points, ginternal-reset-location-views): New.
6119 * doc/invoke.texi: Document them. Use @itemx where intended.
6120 (gvariable-location-views): Adjust.
6121 * target.def (reset_location_view): New.
6122 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6123 (TARGET_RESET_LOCATION_VIEW): New.
6124 * doc/tm.texi: Rebuilt.
6125 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6126 (dwarf2out_default_as_locview_support): New.
6127 (output_asm_line_debug_info): Use option variables.
6128 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6129 (output_loc_list): Likewise.
6130 (add_high_low_attributes): Check option variables.
6131 Don't output entry view attribute in strict mode.
6132 (gen_inlined_subroutine_die): Check option variables.
6133 (dwarf2out_inline_entry): Likewise.
6134 (init_sections_and_labels): Likewise.
6135 (dwarf2out_early_finish): Likewise.
6136 (maybe_reset_location_view): New, from...
6137 (dwarf2out_var_location): ... here. Call it.
6138 * debug.h (dwarf2out_default_as_loc_support): Declare.
6139 (dwarf2out_default_as_locview_support): Declare.
6140 * hooks.c (hook_int_rtx_insn_0): New.
6141 * hooks.h (hook_int_rtx_insn_0): Declare.
6142 * toplev.c (process_options): Take -gas-loc-support and
6143 -gas-locview-support from dwarf2out. Enable
6144 -gvariable-location-views by default only with locview
6145 assembler support. Enable -ginternal-reset-location-views by
6146 default only if the target defines the corresponding hook.
6147 Enable -ginline-points by default if location views are
6148 enabled; force it disabled if statement frontiers are
6150 * tree-inline.c (expand_call_inline): Check option variables.
6151 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6153 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6155 PR tree-optimization/84321
6156 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6157 handling. Also check whether the anti-range contains any values
6158 that satisfy the mask; switch to a VR_RANGE if not.
6160 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6163 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6165 2018-02-13 Martin Jambor <mjambor@suse.cz>
6168 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6169 of call statements, also set location of a load to a temporary.
6171 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6173 * config/rl78/rl78.c (add_vector_labels): New function.
6174 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6175 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6176 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6177 which checks that no arguments are passed.
6178 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6179 * doc/extend.texi: Documentation for the new attribute.
6181 2018-02-13 Andreas Schwab <schwab@suse.de>
6183 * config/riscv/linux.h (CPP_SPEC): Define.
6185 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6188 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6189 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6190 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6191 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6192 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6193 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6194 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6195 temporarily for AES and PCLMUL builtins.
6197 PR tree-optimization/84339
6198 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6199 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6203 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6204 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6205 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6207 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6209 * omp-simd-clone.h: New file.
6210 * omp-simd-clone.c: Include omp-simd-clone.h.
6211 (expand_simd_clones): No longer static.
6212 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6213 cgraph.h and omp-simd-clone.h.
6214 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6215 (vect_recog_widen_shift_pattern): Formatting fix.
6216 (vect_pattern_recog_1): Don't check optab for calls.
6219 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6220 operands[2] into a REG before using gen_lowpart on it.
6222 2018-02-12 Jeff Law <law@redhat.com>
6225 * config/sh/sh.c (find_barrier): Consider a sibling call
6228 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6229 successfully back substituting a reg.
6231 2018-02-12 Richard Biener <rguenther@suse.de>
6233 PR tree-optimization/84037
6234 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6235 parameter, move visited init to caller.
6236 (vect_slp_analyze_operations): Separate cost from validity
6237 check, initialize visited once for all instances.
6238 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6240 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6242 (vect_model_promotion_demotion_cost): Likewise.
6243 (vectorizable_bswap): Guard cost modeling with !slp_node
6244 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6246 (vectorizable_call): Likewise.
6247 (vectorizable_conversion): Likewise.
6248 (vectorizable_assignment): Likewise.
6249 (vectorizable_shift): Likewise.
6250 (vectorizable_operation): Likewise.
6251 (vectorizable_store): Likewise.
6252 (vectorizable_load): Likewise.
6253 (vectorizable_condition): Likewise.
6254 (vectorizable_comparison): Likewise.
6256 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6259 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6260 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6261 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6263 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6266 * params.def (inline-min-speedup): Increase from 8 to 15.
6267 (max-inline-insns-auto): Decrease from 40 to 30.
6268 * ipa-split.c (consider_split): Add some buffer for function to
6269 be considered inlining candidate.
6270 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6273 2018-02-12 Richard Biener <rguenther@suse.de>
6275 PR tree-optimization/84037
6276 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6277 matched stmts if we cannot swap the non-matched ones.
6279 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6281 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6282 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6283 _mm_maskz_scalef_round_ss): New intrinsics.
6284 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6285 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6286 __builtin_ia32_scalefss_round): Remove.
6287 (__builtin_ia32_scalefsd_mask_round,
6288 __builtin_ia32_scalefss_mask_round): New intrinsics.
6289 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6290 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6291 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6292 "<round_constraint>")): Changed to ...
6293 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6294 "<round_scalar_constraint>")): ... this.
6295 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6296 %0, %1, %2<round_op3>}"): Changed to ...
6297 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6298 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6299 %2<round_scalar_mask_op3>}"): ... this.
6300 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6302 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6304 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6305 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6306 (_mm_maskz_sqrt_round_ss): New intrinsics.
6307 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6308 (__builtin_ia32_sqrtsd_mask_round)
6309 (__builtin_ia32_sqrtss_mask_round): New builtins.
6310 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6311 (__builtin_ia32_sqrtss_round): Remove.
6312 (__builtin_ia32_sqrtsd_mask_round)
6313 (__builtin_ia32_sqrtss_mask_round): New builtins.
6314 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6315 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6316 ((match_operand:VF_128 1 "vector_operand"
6317 "xBm,<round_constraint>")): Changed to ...
6318 ((match_operand:VF_128 1 "vector_operand"
6319 "xBm,<round_scalar_constraint>")): ... this.
6320 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6321 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6322 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6323 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6324 %<iptr>1<round_scalar_mask_op3>}): ... this.
6325 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6326 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6328 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6331 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6332 Cast vec_cmpeq result to correct type.
6333 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6334 Cast vec_cmpgt result to correct type.
6336 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6338 * final.c (final_scan_insn_1): Renamed from...
6339 (final_scan_insn): ... this. New wrapper, to recover
6340 seen from the outermost call in recursive ones.
6341 * config/sparc/sparc.c (output_return): Drop seen from call.
6342 (output_sibcall): Likewise.
6343 * config/visium/visium.c (output_branch): Likewise.
6345 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6347 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6350 2018-02-10 Alan Modra <amodra@gmail.com>
6353 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6354 Specify LR as an input.
6356 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6359 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6360 remove_member_access_dummy_vars): New functions.
6361 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6362 lower_omp_1, execute_lower_omp): Use them.
6364 PR rtl-optimization/84308
6365 * shrink-wrap.c (spread_components): Release todo vector.
6367 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6369 PR rtl-optimization/57193
6370 * ira-color.c (struct allocno_color_data): Add member
6371 conflict_allocno_hard_prefs.
6372 (update_conflict_allocno_hard_prefs): New.
6373 (bucket_allocno_compare_func): Add a preference based on
6374 conflict_allocno_hard_prefs.
6375 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6376 (color_allocnos): Remove a dead code. Initiate
6377 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6379 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6382 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6383 constraint from =wa to wa. Avoid a subreg on the output operand,
6384 instead use a pseudo and subreg it in a move.
6385 (p9_xxbrd_<mode>): Changed to ...
6386 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6387 (p9_xxbrd_v2df): New expander.
6388 (p9_xxbrw_<mode>): Changed to ...
6389 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6390 (p9_xxbrw_v4sf): New expander.
6392 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6394 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6396 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6399 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6400 multiply in 32-bit mode.
6401 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6402 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6405 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6407 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6408 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6409 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6410 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6412 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6415 * dwarf2out.c (is_trivial_indirect_ref): New function.
6416 (dwarf2out_late_global_decl): Do not generate a location
6417 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6418 and that are not defined in the current unit.
6420 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6422 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6423 instead of a libcall for UNORDERED.
6425 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6428 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6429 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6431 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6434 * config/s390/s390.c (s390_set_current_function): Invoke
6435 s390_indirect_branch_settings also if fndecl didn't change.
6437 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6439 * config/rs6000/rs6000.md (blockage): Set length to zero.
6441 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6443 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6445 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6448 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6449 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6453 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6454 PARALLEL incoming that failed vt_get_decl_and_offset check.
6457 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6458 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6460 (get_variable_section): For decls in named .bss* sections pass true as
6461 second argument to bss_initializer_p.
6463 2018-02-09 Marek Polacek <polacek@redhat.com>
6464 Jakub Jelinek <jakub@redhat.com>
6467 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6468 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6469 Sync some changes from cxx_fold_indirect_ref.
6471 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6473 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6475 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6476 (BLOCK_INLINE_ENTRY_LABEL): New.
6477 (dwarf2out_var_location): Disregard inline entry markers.
6478 (inline_entry_data): New struct.
6479 (inline_entry_data_hasher): New hashtable type.
6480 (inline_entry_data_hasher::hash): New.
6481 (inline_entry_data_hasher::equal): New.
6482 (inline_entry_data_table): New variable.
6483 (add_high_low_attributes): Add DW_AT_entry_pc and
6484 DW_AT_GNU_entry_view attributes if a pending entry is found
6485 in inline_entry_data_table. Add old entry_pc attribute only
6486 if debug nonbinding markers are disabled.
6487 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6488 markers are enabled.
6489 (block_within_block_p, dwarf2out_inline_entry): New.
6490 (dwarf2out_finish): Check that no entries remained in
6491 inline_entry_data_table.
6492 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6493 (final_scan_insn, notice_source_line): Likewise.
6494 (rest_of_clean_state): Skip inline entry markers.
6495 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6497 * gimple.c (gimple_build_debug_inline_entry): New.
6498 * gimple.h (enum gimple_debug_subcode): Add
6499 GIMPLE_DEBUG_INLINE_ENTRY.
6500 (gimple_build_debug_inline_entry): Declare.
6501 (gimple_debug_inline_entry_p): New.
6502 (gimple_debug_nonbind_marker_p): Adjust.
6503 * insn-notes.def (INLINE_ENTRY): New.
6504 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6505 inline entry marker notes.
6506 (print_insn): Likewise.
6507 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6508 (INSN_DEBUG_MARKER_KIND): Likewise.
6509 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6510 * tree-inline.c (expand_call_inline): Build and insert
6511 debug_inline_entry stmt.
6512 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6513 inline entry blocks early, if nonbind markers are enabled.
6514 (dump_scope_block): Dump fragment info.
6515 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6516 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6517 (gimple_build_debug_inline_entry): New.
6518 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6519 Enable/disable inline entry points too.
6520 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6521 (DEBUG_INSN): Describe inline entry markers.
6523 * common.opt (gvariable-location-views): New.
6524 (gvariable-location-views=incompat5): New.
6525 * config.in: Rebuilt.
6526 * configure: Rebuilt.
6527 * configure.ac: Test assembler for view support.
6528 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6529 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6530 * dwarf2out.c (var_loc_view): New typedef.
6531 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6532 (dwarf2out_locviews_in_attribute): New.
6533 (dwarf2out_locviews_in_loclist): New.
6534 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6535 (enum dw_line_info_opcode): Add LI_adv_address.
6536 (struct dw_line_info_table): Add view.
6537 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6538 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6539 (zero_view_p): New variable.
6540 (ZERO_VIEW_P): New macro.
6541 (output_asm_line_debug_info): New.
6542 (struct var_loc_node): Add view.
6543 (add_AT_view_list, AT_loc_list): New.
6544 (add_var_loc_to_decl): Add view param. Test it against last.
6545 (new_loc_list): Add view params. Record them.
6546 (AT_loc_list_ptr): Handle loc and view lists.
6547 (view_list_to_loc_list_val_node): New.
6548 (print_dw_val): Handle dw_val_class_view_list.
6549 (size_of_die): Likewise.
6550 (value_format): Likewise.
6551 (loc_list_has_views): New.
6552 (gen_llsym): Set vl_symbol too.
6553 (maybe_gen_llsym, skip_loc_list_entry): New.
6554 (dwarf2out_maybe_output_loclist_view_pair): New.
6555 (output_loc_list): Output view list or entries too.
6556 (output_view_list_offset): New.
6557 (output_die): Handle dw_val_class_view_list.
6558 (output_dwarf_version): New.
6559 (output_compilation_unit_header): Use it.
6560 (output_skeleton_debug_sections): Likewise.
6561 (output_rnglists, output_line_info): Likewise.
6562 (output_pubnames, output_aranges): Update version comments.
6563 (output_one_line_info_table): Output view numbers in asm comments.
6564 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6565 Call maybe_gen_llsym.
6566 (loc_list_from_tree_1): Adjust.
6567 (add_AT_location_description): Create view list attribute if
6568 needed, check it's absent otherwise.
6569 (convert_cfa_to_fb_loc_list): Adjust.
6570 (maybe_emit_file): Call output_asm_line_debug_info for test.
6571 (dwarf2out_var_location): Reset views as needed. Precompute
6572 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6573 attribute. Set view.
6574 (new_line_info_table): Reset next view.
6575 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6576 (dwarf2out_source_line): Likewise. Output view resets and labels to
6577 the assembler, or select appropriate line info opcodes.
6578 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6579 (optimize_string_length): Catch it. Adjust.
6580 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6581 dw_val_class_view_list, and remove it if no longer needed.
6582 (hash_loc_list): Hash view numbers.
6583 (loc_list_hasher::equal): Compare them.
6584 (optimize_location_lists): Check whether a view list symbol is
6585 needed, and whether the locview attribute is present, and
6586 whether they match. Remove the locview attribute if no longer
6588 (index_location_lists): Call skip_loc_list_entry for test.
6589 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6590 Use output_dwarf_version.
6591 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6592 (struct dw_val_node): Add val_view_list.
6593 * final.c (SEEN_NEXT_VIEW): New.
6594 (set_next_view_needed): New.
6595 (clear_next_view_needed): New.
6596 (maybe_output_next_view): New.
6597 (final_start_function): Rename to...
6598 (final_start_function_1): ... this. Take pointer to FIRST,
6599 add SEEN parameter. Emit param bindings in the initial view.
6600 (final_start_function): Reintroduce SEEN-less interface.
6601 (final): Rename to...
6602 (final_1): ... this. Take SEEN parameter. Output final pending
6603 next view at the end.
6604 (final): Reintroduce seen-less interface.
6605 (final_scan_insn): Output pending next view before switching
6606 sections or ending a block. Mark the next view as needed when
6607 outputting variable locations. Notify debug backend of section
6608 changes, and of location view changes.
6609 (rest_of_handle_final): Adjust.
6610 * toplev.c (process_options): Autodetect value for debug variable
6611 location views option. Warn on incompat5 without -gdwarf-5.
6612 * doc/invoke.texi (gvariable-location-views): New.
6613 (gvariable-location-views=incompat5): New.
6614 (gno-variable-location-views): New.
6616 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6618 PR tree-optimization/84136
6619 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6620 that the result of find_edge is non-NULL.
6622 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6625 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6626 storing integer register in SImode. Fix cost of 256 and 512
6627 byte aligned SSE register store.
6629 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6631 * config/i386/i386.c (ix86_multiplication_cost): Fix
6632 multiplication cost for TARGET_AVX512DQ.
6634 2018-02-08 Marek Polacek <polacek@redhat.com>
6636 PR tree-optimization/84238
6637 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6640 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6642 PR tree-optimization/84265
6643 * tree-vect-stmts.c (vectorizable_store): Don't treat
6644 VMAT_CONTIGUOUS accesses as grouped.
6645 (vectorizable_load): Likewise.
6647 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6649 PR tree-optimization/81635
6650 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6651 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6652 (test_round_for_mask): New functions.
6653 (wide_int_cc_tests): Call test_round_for_mask.
6654 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6655 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6656 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6657 range returned by get_range_info.
6659 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6662 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6663 * symtab.c: Include builtins.h
6664 (symtab_node::output_to_lto_symbol_table_p): Move here
6665 from lto-streamer-out.c:output_symbol_p.
6666 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6667 (output_symbol_p): Move all logic to symtab.c
6668 (produce_symtab): Update.
6670 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6672 * config/s390/s390-opts.h (enum indirect_branch): Define.
6673 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6674 (s390_indirect_branch_via_thunk)
6675 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6676 (enum s390_indirect_branch_type): Define.
6677 * config/s390/s390.c (struct s390_frame_layout, struct
6678 machine_function): Remove.
6679 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6680 (indirect_branch_table_label_no, indirect_branch_table_name):
6682 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6683 (enum s390_indirect_branch_option): Define.
6684 (s390_return_addr_from_memory): New function.
6685 (s390_handle_string_attribute): New function.
6686 (s390_attribute_table): Add new attribute handler.
6687 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6688 (s390_indirect_branch_via_thunk): New function.
6689 (s390_indirect_branch_via_inline_thunk): New function.
6690 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6691 sibling call optimization for non z10 compiles.
6692 (s390_emit_call): Force indirect branch target to be a single
6693 register. Add r1 clobber for non-z10 compiles.
6694 (s390_emit_epilogue): Emit return jump via return_use expander.
6695 (s390_reorg): Handle JUMP_INSNs as execute targets.
6696 (s390_option_override_internal): Perform validity checks for the
6697 new command line options.
6698 (s390_indirect_branch_attrvalue): New function.
6699 (s390_indirect_branch_settings): New function.
6700 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6701 (s390_output_indirect_thunk_function): New function.
6702 (s390_code_end): Implement target hook.
6703 (s390_case_values_threshold): Implement target hook.
6704 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6706 * config/s390/s390.h (struct s390_frame_layout)
6707 (struct machine_function): Move here from s390.c.
6708 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6709 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6710 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6711 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6712 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6713 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6714 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6715 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6716 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6717 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6718 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6719 (mnemonic attribute): Add values which aren't recognized
6721 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6722 pattern for branch conversion. Fix mnemonic attribute.
6723 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6724 indirect branch via thunk if requested.
6725 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6726 ("*indirect_jump"): Disable for branch conversion using out of
6728 ("indirect_jump_via_thunk<mode>_z10")
6729 ("indirect_jump_via_thunk<mode>")
6730 ("indirect_jump_via_inlinethunk<mode>_z10")
6731 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6732 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6733 ("casesi_jump_via_inlinethunk<mode>_z10")
6734 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6735 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6736 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6737 ("*indirect2_jump"): Disable for branch conversion.
6738 ("casesi_jump"): Turn into expander and expand patterns for branch
6740 ("return_use"): New expander.
6741 ("*return"): Emit return via thunk and rename it to ...
6742 ("*return<mode>"): ... this one.
6743 * config/s390/s390.opt: Add new options and and enum for the
6746 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6748 * lra-constraints.c (match_reload): Unconditionally use
6749 gen_lowpart_SUBREG, rather than selecting between that
6750 and equivalent gen_rtx_SUBREG code.
6752 2018-02-08 Richard Biener <rguenther@suse.de>
6754 PR tree-optimization/84233
6755 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6756 changed flag instead of boguously re-using phi_inserted.
6758 2018-02-08 Martin Jambor <mjambor@suse.cz>
6760 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6761 static local variables.
6763 2018-02-08 Richard Biener <rguenther@suse.de>
6765 PR tree-optimization/84278
6766 * tree-vect-stmts.c (vectorizable_store): When looking for
6767 smaller vector types to perform grouped strided loads/stores
6768 make sure the mode is supported by the target.
6769 (vectorizable_load): Likewise.
6771 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6773 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6774 Increase LDP/STP opportunities by adding adjacent callee-saves.
6776 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6778 PR rtl-optimization/84068
6779 PR rtl-optimization/83459
6780 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6782 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6784 PR tree-optimization/84224
6785 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6786 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6789 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6792 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6793 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6794 count, remove one USE.
6796 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6798 * doc/install.texi (Configuration): Document the
6799 --with-long-double-format={ibm,ieee} PowerPC configuration
6803 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6804 Convert from define_expand to be define_insn_and_split. Rework
6805 float/double/_Float128 conversions to QI/HI/SImode to work with
6806 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6807 conversions to QI/HImode types did a store and then a load to
6808 truncate the value. For conversions to VSX registers, don't split
6809 the insn, instead emit the code directly. Use the code iterator
6810 any_fix to combine signed and unsigned conversions.
6811 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6812 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6813 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6814 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6815 (fix_<mode>di2_hw): Likewise.
6816 (fixuns_<mode>di2_hw): Likewise.
6817 (fix_<mode>si2_hw): Likewise.
6818 (fixuns_<mode>si2_hw): Likewise.
6819 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6820 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6821 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6822 fix<uns>_trunc<SFDF:mode>si2_p8.
6823 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6825 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6826 (fix<uns>_<mode>_mem): Likewise.
6827 (fctiw<u>z_<mode>_mem): Likewise.
6828 (fix<uns>_<mode>_mem): Likewise.
6829 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6830 the register allocator from doing a direct move to the GPRs to do
6831 a store, and instead use the ISA 3.0 store byte/half-word from
6832 vector register instruction. For IEEE 128-bit floating point,
6833 also optimize stores of 32-bit ints.
6834 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6836 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6838 * genextract.c (push_pathstr_operand): New function to support
6840 (walk_rtx): Call push_pathstr_operand.
6841 (print_path): Support [a-zA-Z].
6843 2018-02-07 Richard Biener <rguenther@suse.de>
6845 PR tree-optimization/84037
6846 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6847 (cse_and_gimplify_to_preheader): Declare.
6848 (vect_get_place_in_interleaving_chain): Likewise.
6849 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6851 (_loop_vec_info::~_loop_vec_info): Delete it.
6852 (cse_and_gimplify_to_preheader): New function.
6853 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6854 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6855 (vectorizable_load): Likewise. For grouped stores always base
6856 the IV on the first element.
6857 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6858 condition before gimplifying.
6860 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6862 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6863 *DIV_EXPR and *MOD_EXPR.
6865 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6868 * config/i386/i386.c (ix86_option_override_internal): Mask out
6869 the CF_SET bit when checking -fcf-protection.
6871 2018-02-07 Tom de Vries <tom@codesourcery.com>
6874 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6877 2018-02-07 Richard Biener <rguenther@suse.de>
6879 PR tree-optimization/84204
6880 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6883 PR tree-optimization/84205
6884 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6885 special-case isl_ast_op_zdiv_r.
6887 PR tree-optimization/84223
6888 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6889 Only add conditions from within the region.
6890 (gather_bbs::after_dom_children): Adjust.
6892 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6895 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6896 * config/avr/avr.md: Only post-reload split REG-REG moves if
6897 either register is GENERAL_REG_P.
6899 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6901 PR tree-optimization/84235
6902 * tree-ssa-scopedtables.c
6903 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6904 if the subtraction is performed in floating point type where NaNs are
6905 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6906 build 1. Formatting fix.
6908 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6911 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6912 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6913 and skip it regardless of bb boundaries. Use CALL_P macro,
6914 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6915 unnecessarily, formatting fix.
6917 2018-02-06 Michael Collison <michael.collison@arm.com>
6919 * config/arm/thumb2.md:
6920 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6921 (*thumb_mov_notscc): Ditto.
6923 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6926 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6927 unsigned_fix, not "s".
6929 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6931 * configure.ac (gcc_fn_eh_frame_ro): New function.
6932 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6933 correct .eh_frame permissions.
6934 * configure: Regenerate.
6936 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6938 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6941 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6943 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6944 Display warning message for -mno-speculate-indirect-jumps.
6946 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6948 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6950 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6952 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6954 PR tree-optimization/84225
6955 * tree-eh.c (find_trapping_overflow): Only call
6956 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6958 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6961 * config/i386/i386.c: Reimplement the check of possible options
6962 -mibt/-mshstk conbination. Change error messages.
6963 * doc/invoke.texi: Fix a typo: remove extra '='.
6965 2018-02-06 Marek Polacek <polacek@redhat.com>
6967 PR tree-optimization/84228
6968 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6970 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6973 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6974 emitted arch directives.
6975 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6976 __ARM_FEATURE_COPROC before changing architectures.
6978 2018-02-06 Richard Biener <rguenther@suse.de>
6980 * config/i386/i386.c (print_reg): Fix typo.
6981 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6983 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6985 * configure: Regenerate.
6987 2018-02-05 Martin Sebor <msebor@redhat.com>
6989 PR tree-optimization/83369
6990 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6993 2018-02-05 Martin Liska <mliska@suse.cz>
6995 * doc/invoke.texi: Cherry-pick upstream r323995.
6997 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6999 * ira.c (ira_init_register_move_cost): Adjust comment.
7001 2018-02-05 Martin Liska <mliska@suse.cz>
7003 PR gcov-profile/84137
7004 * doc/gcov.texi: Fix typo in documentation.
7006 2018-02-05 Martin Liska <mliska@suse.cz>
7008 PR gcov-profile/83879
7009 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7010 using dlopen functionality.
7012 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7014 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7015 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7016 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7017 _mm_maskz_range_round_ss): New intrinsics.
7018 (__builtin_ia32_rangesd128_round)
7019 (__builtin_ia32_rangess128_round): Remove.
7020 (__builtin_ia32_rangesd128_mask_round,
7021 __builtin_ia32_rangess128_mask_round): New builtins.
7022 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7023 __builtin_ia32_rangess128_round): Remove.
7024 (__builtin_ia32_rangesd128_mask_round,
7025 __builtin_ia32_rangess128_mask_round): New builtins.
7026 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7027 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7028 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7029 "<round_saeonly_constraint>")): Changed to ...
7030 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7031 "<round_saeonly_scalar_constraint>")): ... this.
7032 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7033 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7034 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7035 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7036 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7038 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7040 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7042 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7043 Remove all values except native, 8540 and 8548.
7045 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7047 * config/i386/i386.c (ix86_output_function_return): Pass
7048 INVALID_REGNUM, instead of -1, as invalid register number to
7049 indirect_thunk_name and output_indirect_thunk.
7051 2018-02-02 Julia Koval <julia.koval@intel.com>
7053 * config.gcc: Add -march=icelake.
7054 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7055 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7056 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7057 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7058 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7059 (processor_target_table): Add icelake.
7060 (ix86_option_override_internal): Handle new PTAs.
7061 (get_builtin_code_for_version): Handle icelake.
7062 (M_INTEL_COREI7_ICELAKE): New.
7063 (fold_builtin_cpu): Handle icelake.
7064 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7065 * doc/invoke.texi: Add -march=icelake.
7067 2018-02-02 Julia Koval <julia.koval@intel.com>
7069 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7070 to wide_int_bitmask.
7071 * wide-int-bitmask.h: New.
7073 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7076 * config/i386/i386.md: Replace Pmode with word_mode in
7077 builtin_setjmp_setup and builtin_longjmp to support x32.
7079 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7083 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7085 (rs6000_supported_cpu_names): New static variable.
7086 (linux_cpu_translation_table): Likewise.
7087 (elf_platform) <cpu>: Define new static variable and use it.
7088 Translate kernel AT_PLATFORM name to canonical name if needed.
7089 Error if platform name is unknown.
7091 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7094 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7096 2018-02-01 Jeff Law <law@redhat.com>
7099 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7100 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7101 the scratch if RELEASE_VIA_POP is false.
7102 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7103 If we have to save a temporary register, decrement SIZE appropriately.
7104 Pass new arguments to release_scratch_register_on_entry.
7105 (ix86_adjust_stack_and_probe): Likewise.
7106 (ix86_emit_probe_stack_range): Pass new arguments to
7107 release_scratch_register_on_entry.
7109 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7111 PR rtl-optimization/84157
7112 * combine.c (change_zero_ext): Use REG_P predicate in
7113 front of HARD_REGISTER_P predicate.
7115 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7117 * config/avr/avr.c (avr_option_override): Move disabling of
7118 -fdelete-null-pointer-checks to...
7119 * common/config/avr/avr-common.c (avr_option_optimization_table):
7122 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7124 PR tree-optimization/81635
7125 * tree-data-ref.c (split_constant_offset_1): For types that
7126 wrap on overflow, try to use range info to prove that wrapping
7129 2018-02-01 Renlin Li <renlin.li@arm.com>
7132 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7134 (aarch64_register_move_cost): Likewise.
7135 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7137 (REG_CLASS_NAMES): Likewise.
7138 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7139 TAILCALL_ADDR_REGS. Remove IP registers.
7140 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7142 2018-02-01 Richard Biener <rguenther@suse.de>
7144 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7145 for specifying RPO order and allow NULL for that.
7146 * domwalk.c (dom_walker::dom_walker): Likewise.
7147 (dom_walker::walk): Handle NULL RPO order.
7148 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7150 (rewrite_update_dom_walker): Likewise.
7151 (mark_def_dom_walker): Likewise.
7153 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7155 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7156 (aarch64_maybe_expand_sve_subreg_move): Declare.
7157 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7158 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7160 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7161 that are semantically a reverse operation.
7162 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7163 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7164 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7166 (aarch64_can_change_mode_class): For big-endian, forbid changes
7167 between two SVE modes if they have different element sizes.
7169 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7171 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7172 the TImode handling for big-endian targets.
7174 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7176 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7177 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7179 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7180 Remove BSWAP handing for big-endian targets and use the form of
7181 LD1RQ appropariate for the mode.
7183 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7185 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7186 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7189 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7192 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7193 check for operands that need to go through aarch64_sve_reload_be.
7195 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7197 PR tree-optimization/81661
7198 PR tree-optimization/84117
7199 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7200 * tree-eh.c: Include gimplify.h.
7201 (find_trapping_overflow, replace_trapping_overflow,
7202 rewrite_to_non_trapping_overflow): New functions.
7203 * tree-vect-loop.c: Include tree-eh.h.
7204 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7205 * tree-data-ref.c: Include tree-eh.h.
7206 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7208 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7210 PR rtl-optimization/84123
7211 * combine.c (change_zero_ext): Check if hard register satisfies
7212 can_change_dest_mode before calling gen_lowpart_SUBREG.
7214 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7217 * ira.c (ira_init_register_move_cost): Remove assert.
7219 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7221 PR rtl-optimization/84071
7222 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7223 * doc/tm.texi: Regenerate.
7225 2018-01-31 Richard Biener <rguenther@suse.de>
7227 PR tree-optimization/84132
7228 * tree-data-ref.c (analyze_miv_subscript): Properly
7229 check whether evolution_function_is_affine_multivariate_p
7230 before calling gcd_of_steps_may_divide_p.
7232 2018-01-31 Julia Koval <julia.koval@intel.com>
7235 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7236 * config/i386/i386.md (rdpid_rex64) New.
7237 (rdpid): Make 32bit only.
7239 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7242 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7243 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7245 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7248 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7250 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7252 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7254 PR rtl-optimization/84071
7255 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7256 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7258 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7260 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7261 (arc_attribute_table): Add 'aux' attribute.
7262 (arc_in_small_data_p): Consider aux like variables.
7263 (arc_is_aux_reg_p): New function.
7264 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7265 (arc_get_aux_arg): New function.
7266 (prepare_move_operands): Handle aux-register access.
7267 (arc_handle_aux_attribute): New function.
7268 * doc/extend.texi (ARC Variable attributes): Add subsection.
7270 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7272 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7273 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7274 (arc_attribute_table): Add 'uncached' attribute.
7275 (arc_print_operand): Print '.di' flag for uncached memory
7277 (arc_in_small_data_p): Do not consider for small data the uncached
7279 (arc_is_uncached_mem_p): New function.
7280 * config/arc/predicates.md (compact_store_memory_operand): Check
7281 for uncached memory accesses.
7282 (nonvol_nonimm_operand): Likewise.
7283 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7285 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7288 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7289 falign-loops=): Add Optimization flag.
7291 2018-01-30 Jeff Law <law@redhat.com>
7294 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7295 INT_REGISTERS_SAVED. Check it prior to calling
7296 get_scratch_register_on_entry.
7297 (ix86_adjust_stack_and_probe): Similarly.
7298 (ix86_emit_probe_stack_range): Similarly.
7299 (ix86_expand_prologue): Corresponding changes.
7301 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7304 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7305 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7307 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7310 * lra-constraints.c (curr_insn_transform): Process AND in the
7313 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7315 PR rtl-optimization/83986
7316 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7317 dependence against last_pending_memory_flush in addition to
7320 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7322 PR tree-optimization/81611
7323 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7326 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7329 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7332 2018-01-30 Richard Biener <rguenther@suse.de>
7333 Jakub Jelinek <jakub@redhat.com>
7335 PR tree-optimization/84111
7336 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7337 inner loops added during recursion, as they don't have up-to-date
7340 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7343 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7344 (can_inline_edge_by_limits_p): ... here.
7345 (can_early_inline_edge_p, check_callers,
7346 update_caller_keys, update_callee_keys, recursive_inlining,
7347 add_new_edges_to_heap, speculation_useful_p,
7348 inline_small_functions,
7349 inline_small_functions, flatten_function,
7350 inline_to_all_callers_1): Update.
7352 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7354 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7357 2018-01-30 Richard Biener <rguenther@suse.de>
7359 PR tree-optimization/83008
7360 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7361 invariant and constant vector uses in stmts when they need
7364 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7367 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7368 * configure: Regenerate.
7370 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7372 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7374 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7375 Use gen_rtx_REG rather than gen_lowpart.
7377 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7379 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7380 rather than 0 when creating partial subregs.
7382 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7384 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7387 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7390 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7391 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7392 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7393 flags. This restores the settings used before the 2017-07-24.
7394 Turning off pre increment/decrement/modify allows IVOPTS to
7395 optimize DF/SF loops where the index is an int.
7397 2018-01-29 Richard Biener <rguenther@suse.de>
7398 Kelvin Nilsen <kelvin@gcc.gnu.org>
7401 * tree-vect-stmts.c (vectorizable_call): Don't call
7402 targetm.vectorize_builtin_md_vectorized_function if callee is
7405 2018-01-22 Carl Love <cel@us.ibm.com>
7407 * doc/extend.tex: Fix typo in second arg in
7408 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7410 2018-01-29 Richard Biener <rguenther@suse.de>
7412 PR tree-optimization/84086
7413 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7414 (flush_ssaname_freelist): When SSA names were released reset
7415 the SCEV hash table.
7417 2018-01-29 Richard Biener <rguenther@suse.de>
7419 PR tree-optimization/84057
7420 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7421 removed paths when removing edges.
7423 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7425 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7426 -mfunction-return=@var{choice}.
7428 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7431 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7433 (layout::print_source_line): Likewise.
7434 (test_get_line_width_without_trailing_whitespace): Add test cases.
7436 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7439 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7442 2018-01-26 Jim Wilson <jimw@sifive.com>
7444 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7446 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7449 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7451 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7452 and CMP + SUB-immediate -> SUBS.
7454 2018-01-26 Martin Sebor <msebor@redhat.com>
7456 PR tree-optimization/83896
7457 * tree-ssa-strlen.c (get_string_len): Rename...
7458 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7459 Avoid assuming length is constant.
7460 (handle_char_store): Use HOST_WIDE_INT for string length.
7462 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7465 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7466 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7468 2018-01-26 Richard Biener <rguenther@suse.de>
7470 PR rtl-optimization/84003
7471 * dse.c (record_store): Only record redundant stores when
7472 the earlier store aliases at least all accesses the later one does.
7474 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7476 PR rtl-optimization/83985
7477 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7478 REG_CFA_RESTORE insns.
7479 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7480 REG_CFA_RESTORE insns here.
7483 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7484 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7486 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7488 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7489 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7490 (arc_init): Likewise.
7491 (arc_override_options): Likewise.
7492 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7494 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7496 * config/arc/arc.h (TARGET_DBNZ): Define.
7497 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7498 properly set the tune attribute.
7499 (dbnz): Use TARGET_DBNZ guard.
7500 * config/arc/arc.opt (mtune): Add core3 option.
7502 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7504 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7505 recognize new pic like addresses.
7506 (arc_delegitimize_address): Clean up.
7508 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7510 * config/arc/arc-arches.def: Option mrf16 valid for all
7512 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7513 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7514 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7515 * config/arc/arc-tables.opt: Regenerate.
7516 * config/arc/arc.c (arc_conditional_register_usage): Handle
7517 reduced register file case.
7518 (arc_file_start): Set must have build attributes.
7519 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7521 * config/arc/arc.opt (mrf16): Add new option.
7522 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7523 * config/arc/genmultilib.awk: Handle new mrf16 option.
7524 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7525 * config/arc/t-multilib: Regenerate.
7526 * doc/invoke.texi (ARC Options): Document mrf16 option.
7528 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7530 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7531 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7532 (arc_attribute_table): Add 'secure_call' attribute.
7533 (arc_print_operand): Print secure call operand.
7534 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7536 (arc_is_secure_call_p): New function. * config/arc/arc.md
7537 (call_i): Add support for sjli instruction.
7538 (call_value_i): Likewise.
7539 * config/arc/constraints.md (Csc): New constraint.
7541 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7542 John Eric Martin <John.Martin@emmicro-us.com>
7544 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7545 * config/arc/arc.c (_arc_jli_section): New struct.
7546 (arc_jli_section): New type.
7547 (rc_jli_sections): New static variable.
7548 (arc_handle_jli_attribute): New function.
7549 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7550 (arc_file_end): New function.
7551 (TARGET_ASM_FILE_END): Define.
7552 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7553 (arc_add_jli_section): New function.
7554 (jli_call_scan): Likewise.
7555 (arc_reorg): Call jli_call_scan.
7556 (arc_output_addsi): Remove 'S' from printing asm operand.
7557 (arc_is_jli_call_p): New function.
7558 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7560 (movhi_insn): Likewise.
7561 (movsi_insn): Likewise.
7562 (movsi_set_cc_insn): Likewise.
7563 (loadqi_update): Likewise.
7564 (load_zeroextendqisi_update): Likewise.
7565 (load_signextendqisi_update): Likewise.
7566 (loadhi_update): Likewise.
7567 (load_zeroextendhisi_update): Likewise.
7568 (load_signextendhisi_update): Likewise.
7569 (loadsi_update): Likewise.
7570 (loadsf_update): Likewise.
7571 (movsicc_insn): Likewise.
7572 (bset_insn): Likewise.
7573 (bxor_insn): Likewise.
7574 (bclr_insn): Likewise.
7575 (bmsk_insn): Likewise.
7576 (bicsi3_insn): Likewise.
7577 (cmpsi_cc_c_insn): Likewise.
7578 (movsi_ne): Likewise.
7579 (movsi_cond_exec): Likewise.
7580 (clrsbsi2): Likewise.
7591 (call_i): Remove 'S' asm letter, add jli instruction.
7592 (call_value_i): Likewise.
7593 * config/arc/arc.op (mjli-always): New option.
7594 * config/arc/constraints.md (Cji): New constraint.
7595 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7597 (subsf3_fpx): Likewise.
7598 (mulsf3_fpx): Likewise.
7599 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7601 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7603 * doc/invoke.texi (ARC): Document mjli-always option.
7605 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7607 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7608 avoid addition with 0 and use incw and decw where possible.
7610 2018-01-26 Richard Biener <rguenther@suse.de>
7612 PR tree-optimization/81082
7613 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7614 association if it requires casting to unsigned.
7615 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7616 from fold_plusminus_mult_expr to catch important cases late when
7617 range info is available.
7619 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7621 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7622 * configure.ac (hidden_linkonce): New test.
7623 * configure: Regenerate.
7624 * config.in: Regenerate.
7626 2018-01-26 Julia Koval <julia.koval@intel.com>
7628 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7629 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7630 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7631 _mm_mask_bitshuffle_epi64_mask): Fix type.
7632 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7633 USI_FTYPE_V4DI_V4DI_USI): Remove.
7634 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7635 __builtin_ia32_vpshufbitqmb256_mask,
7636 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7637 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7638 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7640 2018-01-26 Alan Modra <amodra@gmail.com>
7643 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7644 UNSPEC_VBPERMQ. Sort other unspecs.
7646 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7648 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7650 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7653 * predict.c (drop_profile): Do not push/pop cfun; update also
7655 (handle_missing_profiles): Fix logic looking for zero profiles.
7657 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7660 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7661 on functions with #pragma omp declare simd or functions with simd
7663 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7664 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7665 Remove trailing \n from warning_at calls.
7667 2018-01-25 Tom de Vries <tom@codesourcery.com>
7670 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7671 for neutered workers.
7673 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7676 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7677 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7679 2018-01-24 Jeff Law <law@redhat.com>
7682 * i386.c (get_probe_interval): Move to earlier point.
7683 (ix86_compute_frame_layout): If -fstack-clash-protection and
7684 the frame is larger than the probe interval, then use pushes
7685 to save registers rather than reg->mem moves.
7686 (ix86_expand_prologue): Remove conditional for int_registers_saved
7689 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7692 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7693 min/max for never referenced object.
7695 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7698 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7700 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7701 attributes from DECL_ATTRIBUTES (decl) without affecting
7702 DECL_ATTRIBUTES (current_function_decl).
7703 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7704 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7706 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7708 PR tree-optimization/83979
7709 * fold-const.c (fold_comparison): Use constant_boolean_node
7710 instead of boolean_{true,false}_node.
7712 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7714 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7717 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7719 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7720 Simplify the clause that sets the length attribute.
7721 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7722 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7723 clause that sets the length attribute.
7724 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7726 2018-01-24 Tom de Vries <tom@codesourcery.com>
7729 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7730 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7731 Add strict parameter.
7732 (prevent_branch_around_nothing): Insert dummy insn between branch to
7733 label and label with no ptx insn inbetween.
7734 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7736 2018-01-24 Tom de Vries <tom@codesourcery.com>
7739 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7740 for neutered threads in warp.
7741 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7743 2018-01-24 Richard Biener <rguenther@suse.de>
7745 PR tree-optimization/83176
7746 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7749 2018-01-24 Richard Biener <rguenther@suse.de>
7751 PR tree-optimization/82819
7752 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7753 code generating pluses that are no-ops in the target precision.
7755 2018-01-24 Richard Biener <rguenther@suse.de>
7758 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7760 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7762 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7763 to merge probabilities.
7764 * predict.c (probably_never_executed): Also mark as cold functions
7765 with global 0 profile and guessed local profile.
7766 * profile-count.c (profile_probability::combine_with_count): New
7768 * profile-count.h (profile_probability::operator*,
7769 profile_probability::operator*=, profile_probability::operator/,
7770 profile_probability::operator/=): Reduce precision to adjusted
7771 and set value to guessed on contradictory divisions.
7772 (profile_probability::combine_with_freq): Remove.
7773 (profile_probability::combine_wiht_count): Declare.
7774 (profile_count::force_nonzero):: Set to adjusted.
7775 (profile_count::probability_in):: Set quality to adjusted.
7776 * tree-ssa-tail-merge.c (replace_block_by): Use
7779 2018-01-23 Andrew Waterman <andrew@sifive.com>
7780 Jim Wilson <jimw@sifive.com>
7782 * config/riscv/riscv.c (riscv_stack_boundary): New.
7783 (riscv_option_override): Set riscv_stack_boundary. Handle
7784 riscv_preferred_stack_boundary_arg.
7785 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7786 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7787 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7788 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7789 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7790 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7792 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7795 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7796 of struct ix86_frame.
7797 (ix86_expand_epilogue): Likewise. Add a local variable for
7798 the reg_save_offset field in struct ix86_frame.
7800 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7802 PR tree-optimization/82604
7803 * tree-loop-distribution.c (enum partition_kind): New enum item
7804 PKIND_PARTIAL_MEMSET.
7805 (partition_builtin_p): Support above new enum item.
7806 (generate_code_for_partition): Ditto.
7807 (compute_access_range): Differentiate cases that equality can be
7808 proven at all loops, the innermost loops or no loops.
7809 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7810 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7811 (finalize_partitions, distribute_loop): Don't fuse partition of
7812 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7813 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7816 2018-01-23 Martin Liska <mliska@suse.cz>
7818 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7819 order to ignore the predictor.
7820 (PRED_POLYMORPHIC_CALL): Likewise.
7821 (PRED_RECURSIVE_CALL): Likewise.
7823 2018-01-23 Martin Liska <mliska@suse.cz>
7825 * tree-profile.c (tree_profiling): Print function header to
7826 aware reader which function we are working on.
7827 * value-prof.c (gimple_find_values_to_profile): Do not print
7828 not interesting value histograms.
7830 2018-01-23 Martin Liska <mliska@suse.cz>
7832 * profile-count.h (enum profile_quality): Add
7833 profile_uninitialized as the first value. Do not number values
7834 as they are zero based.
7835 (profile_count::verify): Update sanity check.
7836 (profile_probability::verify): Likewise.
7838 2018-01-23 Nathan Sidwell <nathan@acm.org>
7840 * doc/invoke.texi (ffor-scope): Deprecate.
7842 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7844 PR tree-optimization/83510
7845 * domwalk.c (set_all_edges_as_executable): New function.
7846 (dom_walker::dom_walker): Convert bool param
7847 "skip_unreachable_blocks" to enum reachability. Move setup of
7848 edge flags to set_all_edges_as_executable and only do it when
7849 reachability is REACHABLE_BLOCKS.
7850 * domwalk.h (enum dom_walker::reachability): New enum.
7851 (dom_walker::dom_walker): Convert bool param
7852 "skip_unreachable_blocks" to enum reachability.
7853 (set_all_edges_as_executable): New decl.
7854 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7855 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7857 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7858 but converting true to REACHABLE_BLOCKS.
7859 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7861 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7862 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7863 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7865 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7866 if check_all_array_refs will be called.
7868 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7870 * tree.c (selftest::test_location_wrappers): Add more test
7873 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7875 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7876 (selftest::test_bit_in_range): Likewise.
7878 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7881 * doc/sourcebuild.texi (vect_float): Say that the selector
7882 only describes the situation when -funsafe-math-optimizations is on.
7883 (vect_float_strict): Document.
7885 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7887 PR tree-optimization/83965
7888 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7889 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7890 instead of checking only for a reduction.
7891 (vect_recog_widen_sum_pattern): Likewise.
7893 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7895 * predict.c (probably_never_executed): Only use precise profile info.
7896 (compute_function_frequency): Skip after inlining hack since we now
7897 have quality checking.
7899 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7901 * profile-count.h (profile_probability::very_unlikely,
7902 profile_probability::unlikely, profile_probability::even): Set
7903 precision to guessed.
7905 2018-01-23 Richard Biener <rguenther@suse.de>
7907 PR tree-optimization/83963
7908 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7909 Properly terminate dominator walk when crossing the exit edge not
7910 when visiting its source block.
7912 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7915 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7916 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7918 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7920 PR tree-optimization/83957
7921 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7922 semicolon after for body surrounded by braces.
7924 PR tree-optimization/83081
7925 * profile-count.h (profile_probability::split): New method.
7926 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7927 Use profile_probability::split.
7928 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7929 when splitting a single conditional jump into 2.
7931 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7933 PR tree-optimization/69452
7934 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7937 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7939 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7940 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7941 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7943 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7945 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7946 * config/rl78/rl78.md (movdi): New define_expand.
7947 * config/rl78/rl78.c (rl78_split_movdi): New function.
7949 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7952 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7954 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7955 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7956 128-bit to produce an UNSPEC move to get the double word with the
7957 signbit and then a shift directly to do signbit.
7958 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7959 implementation with a new version that just does either a direct
7960 move or a regular move. Move memory interface to separate insns.
7961 Move insns so they are next to the expander.
7962 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7963 with signbit move. Split big and little endian case.
7964 (signbit<mode>2_dm_mem_le): Likewise.
7965 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7966 (signbit<mode>2_dm2): Likewise.
7968 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7970 * config/rl78/rl78.md (anddi3): New define_expand.
7972 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7974 * config/rl78/rl78.md (umindi3): New define_expand.
7976 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7978 * config/rl78/rl78.md (smindi3): New define_expand.
7980 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7982 * config/rl78/rl78.md (smaxdi3): New define_expand.
7984 2018-01-22 Carl Love <cel@us.ibm.com>
7986 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7987 LVX_V1TI): Add macro expansion.
7988 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7989 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7990 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7991 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7992 Change check to determine if the instruction is a byte reversing
7993 entry. Fix typo in comment.
7994 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7995 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7996 Add def_builtin calls for new builtins.
7997 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7998 Add define_insn expansion.
8000 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8002 * config/rl78/rl78.md (umaxdi3): New define_expand.
8004 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8006 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8007 for non-QImode registers.
8009 2018-01-22 Richard Biener <rguenther@suse.de>
8011 PR tree-optimization/83963
8012 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8013 including the loop exit block.
8014 (scop_detection::merge_sese): Likewise.
8015 (scop_detection::add_scop): Do it here instead.
8017 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8019 * doc/sourcebuild.texi (arm_softfloat): Document.
8021 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8024 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8025 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8026 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8028 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8029 David Edelsohn <dje.gcc@gmail.com>
8032 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8033 Change "crset eq" to "crset 2".
8034 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8035 (*call_indirect_aix<mode>_nospec): Likewise.
8036 (*call_value_indirect_aix<mode>_nospec): Likewise.
8037 (*call_indirect_elfv2<mode>_nospec): Likewise.
8038 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8039 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8040 change assembly output from . to $.
8041 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8042 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8043 (*tablejump<mode>_internal1_nospec): Likewise.
8045 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8048 * config/sh/sh_optimize_sett_clrt.cc:
8049 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8051 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8053 PR tree-optimization/83940
8054 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8055 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8056 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8057 use it to pass back the definition type.
8058 (vect_check_store_rhs): Likewise rhs_dt_out.
8059 (vect_build_gather_load_calls): Add a mask_dt argument and use
8060 it instead of a call to vect_is_simple_use.
8061 (vectorizable_store): Update calls to vect_check_load_store_mask
8062 and vect_check_store_rhs. Use the dt returned by the latter instead
8063 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8064 instead of calls to vect_is_simple_use. Pass the scalar rather
8065 than the vector operand to vect_is_simple_use when handling
8066 second and subsequent copies of an rhs value.
8067 (vectorizable_load): Update calls to vect_check_load_store_mask
8068 and vect_build_gather_load_calls. Use the cached mask_dt and
8069 gs_info.offset_dt instead of calls to vect_is_simple_use.
8071 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8074 * tree-emutls.c: Include gimplify.h.
8075 (lower_emutls_2): New function.
8076 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8077 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8078 it before further processing.
8081 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8082 UINTVAL (trueop1) instead of INTVAL (op1).
8084 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8088 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8089 INCOMING_FRAME_SP_OFFSET if not defined.
8090 (scan_trace): Add ENTRY argument. If true and
8091 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8092 emit a note to adjust the CFA offset.
8093 (create_cfi_notes): Adjust scan_trace callers.
8094 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8095 INCOMING_FRAME_SP_OFFSET in the CIE.
8096 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8097 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8099 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8100 * doc/tm.texi: Regenerated.
8102 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8104 PR rtl-optimization/83147
8105 * lra-constraints.c (remove_inheritance_pseudos): Use
8106 lra_substitute_pseudo_within_insn.
8108 2018-01-19 Tom de Vries <tom@codesourcery.com>
8109 Cesar Philippidis <cesar@codesourcery.com>
8112 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8114 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8117 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8118 spaces for function labels.
8120 2018-01-19 Martin Liska <mliska@suse.cz>
8122 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8123 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8124 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8125 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8126 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8127 (PRED_CONST_RETURN): Change from 69 to 65.
8128 (PRED_NULL_RETURN): Change from 91 to 71.
8129 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8130 (PRED_LOOP_GUARD): Change from 66 to 73.
8132 2018-01-19 Martin Liska <mliska@suse.cz>
8134 * predict.c (predict_insn_def): Add new assert.
8135 (struct branch_predictor): Change type to signed integer.
8136 (test_prediction_value_range): Amend test to cover
8138 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8139 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8140 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8141 (PRED_LOOP_IV_COMPARE): Likewise.
8142 * predict.h (PROB_UNINITIALIZED): Define new constant.
8144 2018-01-19 Martin Liska <mliska@suse.cz>
8146 * predict.c (dump_prediction): Add new format for
8147 analyze_brprob.py script which is enabled with -details
8149 * profile-count.h (precise_p): New function.
8151 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8153 PR tree-optimization/83922
8154 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8155 there are no statements that need masking.
8156 (vect_active_double_reduction_p): New function.
8157 (vect_analyze_loop_operations): Use it when handling phis that
8158 are not in the loop header.
8160 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8162 PR tree-optimization/83914
8163 * tree-vect-loop.c (vectorizable_induction): Don't convert
8164 init_expr or apply the peeling adjustment for inductions
8165 that are nested within the vectorized loop.
8167 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8169 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8172 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8176 * function.h (gimplify_parameters): Add gimple_seq * argument.
8177 * function.c: Include gimple.h and options.h.
8178 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8179 for the added local temporaries if needed.
8180 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8181 if there are any parameter cleanups, wrap whole body into a
8182 try/finally with the cleanups.
8184 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8187 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8188 Use GET_MODE_CLASS for scalar floating point.
8190 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8194 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8195 Fix call of call_cgraph_insertion_hooks.
8197 2018-01-18 Martin Sebor <msebor@redhat.com>
8199 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8201 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8204 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8207 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8210 * common.opt: (-ffile-prefix-map): New option.
8211 * opts.c (common_handle_option): Defer it.
8212 * opts-global.c (handle_common_deferred_options): Handle it.
8213 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8214 * file-prefix-map.h: New file.
8215 (remap_debug_filename, add_debug_prefix_map): ...here.
8216 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8217 * final.c (debug_prefix_map, add_debug_prefix_map
8218 remap_debug_filename): Move to...
8219 * file-prefix-map.c: New file.
8220 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8221 generalize, get rid of alloca(), use strrchr() instead of strchr().
8222 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8223 Implement in terms of add_prefix_map().
8224 (remap_macro_filename, remap_debug_filename): Implement in term of
8226 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8227 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8228 * dbxout.c: Include file-prefix-map.h.
8229 * varasm.c: Likewise.
8230 * vmsdbgout.c: Likewise.
8231 * xcoffout.c: Likewise.
8232 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8233 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8234 * doc/invoke.texi (-ffile-prefix-map): Document.
8235 (-fdebug-prefix-map): Update description.
8237 2018-01-18 Martin Liska <mliska@suse.cz>
8239 * config/i386/i386.c (indirect_thunk_name): Document that also
8241 (output_indirect_thunk): Document why both instructions
8242 (pause and lfence) are generated.
8244 2018-01-18 Richard Biener <rguenther@suse.de>
8246 PR tree-optimization/83887
8247 * graphite-scop-detection.c
8248 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8249 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8250 (scop_detection::merge_sese): Re-implement with a flood-fill
8251 algorithm that properly finds a SESE region if it exists.
8253 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8256 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8257 pointer_diff optimizations use view_convert instead of convert.
8259 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8261 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8262 Generate different code for -mno-speculate-indirect-jumps.
8263 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8264 (*call_indirect_aix<mode>): Disable for
8265 -mno-speculate-indirect-jumps.
8266 (*call_indirect_aix<mode>_nospec): New define_insn.
8267 (*call_value_indirect_aix<mode>): Disable for
8268 -mno-speculate-indirect-jumps.
8269 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8270 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8271 -mno-speculate-indirect-jumps.
8272 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8274 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8276 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8277 long double type, set the flags for noting the default long double
8278 type, even if we don't pass or return a long double type.
8280 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8283 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8286 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8288 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8289 support for merge[hl].
8290 (fold_mergehl_helper): New helper function.
8291 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8292 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8293 (altivec_vmrglw_direct): Add xxmrglw insn.
8295 2018-01-17 Andrew Waterman <andrew@sifive.com>
8297 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8298 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8300 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8303 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8304 call the lto_location_cache before reading the
8305 DECL_SOURCE_LOCATION of the types.
8307 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8308 Richard Sandiford <richard.sandiford@linaro.org>
8310 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8311 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8312 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8313 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8314 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8316 * config/aarch64/constraints.md (aarch64_movti_operand):
8318 * config/aarch64/predicates.md (Uti): Add new constraint.
8320 2018-01-17 Carl Love <cel@us.ibm.com>
8322 * config/rs6000/vsx.md (define_expand xl_len_r,
8323 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8324 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8326 (define_expand, define_insn): Move the shift left from the
8327 define_insn to the define_expand for lxvl and stxvl instructions.
8328 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8329 and XL_LEN_R definitions to PURE.
8331 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8333 * config/i386/i386.c (indirect_thunk_name): Declare regno
8334 as unsigned int. Compare regno with INVALID_REGNUM.
8335 (output_indirect_thunk): Ditto.
8336 (output_indirect_thunk_function): Ditto.
8337 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8338 in the call to output_indirect_thunk_function.
8340 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8343 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8344 rather than the size of inner_type to determine the stack slot size
8345 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8347 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8350 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8353 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8355 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8356 endian Linux systems to optionally enable multilibs for selecting
8357 the long double type if the user configured an explicit type.
8358 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8359 have no long double multilibs if not defined.
8360 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8361 warn if the user used -mabi={ieee,ibm}longdouble and we built
8362 multilibs for long double.
8363 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8364 appropriate multilib option.
8365 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8367 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8368 for building long double multilibs.
8369 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8371 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8373 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8376 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8378 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8381 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8384 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8387 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8389 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8390 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8392 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8394 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8395 different rtl trees depending on TARGET_64BIT.
8396 (rs6000_gen_lvx): Likewise.
8398 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8400 * config/visium/visium.md (nop): Tweak comment.
8401 (hazard_nop): Likewise.
8403 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8405 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8406 -mspeculate-indirect-jumps.
8407 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8408 for -mno-speculate-indirect-jumps.
8409 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8410 (*call_value_indirect_elfv2<mode>): Disable for
8411 -mno-speculate-indirect-jumps.
8412 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8413 (indirect_jump): Emit different RTL for
8414 -mno-speculate-indirect-jumps.
8415 (*indirect_jump<mode>): Disable for
8416 -mno-speculate-indirect-jumps.
8417 (*indirect_jump<mode>_nospec): New define_insn.
8418 (tablejump): Emit different RTL for
8419 -mno-speculate-indirect-jumps.
8420 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8421 (tablejumpsi_nospec): New define_expand.
8422 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8423 (tablejumpdi_nospec): New define_expand.
8424 (*tablejump<mode>_internal1): Disable for
8425 -mno-speculate-indirect-jumps.
8426 (*tablejump<mode>_internal1_nospec): New define_insn.
8427 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8430 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8432 * caller-save.c (insert_save): Drop unnecessary parameter. All
8435 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8436 Richard Biener <rguenth@suse.de>
8439 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8440 return early, inline manually is_gimple_sizepos. Make sure if we
8441 call gimplify_expr we don't end up with a gimple constant.
8442 * tree.c (variably_modified_type_p): Don't return true for
8443 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8444 * gimplify.h (is_gimple_sizepos): Remove.
8446 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8448 PR tree-optimization/83857
8449 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8450 vectorizable_live_operation for pure SLP statements.
8451 (vectorizable_live_operation): Handle PHIs.
8453 2018-01-16 Richard Biener <rguenther@suse.de>
8455 PR tree-optimization/83867
8456 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8457 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8459 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8462 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8463 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8464 If off is not INTEGER_CST, issue a may not be aligned warning
8465 rather than isn't aligned. Use isn%'t rather than isn't.
8466 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8468 <case MULT_EXPR>: Improve the case when bottom and one of the
8469 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8470 operand, in that case check if the other operand is multiple of
8471 bottom divided by the INTEGER_CST operand.
8473 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8476 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8477 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8478 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8479 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8480 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8481 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8482 (pa_function_arg_size): New function.
8484 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8486 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8487 in a separate statement.
8489 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8491 PR tree-optimization/83847
8492 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8493 group gathers and scatters.
8495 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8497 PR rtl-optimization/86620
8498 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8500 PR rtl-optimization/83213
8501 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8502 to last if both are JUMP_INSNs.
8504 PR tree-optimization/83843
8505 * gimple-ssa-store-merging.c
8506 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8507 store_immediate_info for bswap/nop orig_stores.
8509 2018-01-15 Andrew Waterman <andrew@sifive.com>
8511 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8513 <UDIV>: Increase cost if !TARGET_DIV.
8515 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8517 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8518 (define_attr "cr_logical_3op"): New.
8519 (cceq_ior_compare): Adjust.
8520 (cceq_ior_compare_complement): Adjust.
8521 (*cceq_rev_compare): Adjust.
8522 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8523 (is_cracked_insn): Adjust.
8524 (insn_must_be_first_in_group): Adjust.
8525 * config/rs6000/40x.md: Adjust.
8526 * config/rs6000/440.md: Adjust.
8527 * config/rs6000/476.md: Adjust.
8528 * config/rs6000/601.md: Adjust.
8529 * config/rs6000/603.md: Adjust.
8530 * config/rs6000/6xx.md: Adjust.
8531 * config/rs6000/7450.md: Adjust.
8532 * config/rs6000/7xx.md: Adjust.
8533 * config/rs6000/8540.md: Adjust.
8534 * config/rs6000/cell.md: Adjust.
8535 * config/rs6000/e300c2c3.md: Adjust.
8536 * config/rs6000/e500mc.md: Adjust.
8537 * config/rs6000/e500mc64.md: Adjust.
8538 * config/rs6000/e5500.md: Adjust.
8539 * config/rs6000/e6500.md: Adjust.
8540 * config/rs6000/mpc.md: Adjust.
8541 * config/rs6000/power4.md: Adjust.
8542 * config/rs6000/power5.md: Adjust.
8543 * config/rs6000/power6.md: Adjust.
8544 * config/rs6000/power7.md: Adjust.
8545 * config/rs6000/power8.md: Adjust.
8546 * config/rs6000/power9.md: Adjust.
8547 * config/rs6000/rs64.md: Adjust.
8548 * config/rs6000/titan.md: Adjust.
8550 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8552 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8553 ix86_indirect_branch_register logic.
8555 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8557 * config/i386/constraints.md (Bs): Update
8558 ix86_indirect_branch_register check. Don't check
8559 ix86_indirect_branch_register with GOT_memory_operand.
8561 * config/i386/predicates.md (GOT_memory_operand): Don't check
8562 ix86_indirect_branch_register here.
8563 (GOT32_symbol_operand): Likewise.
8565 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8567 * config/i386/predicates.md (constant_call_address_operand):
8568 Rewrite ix86_indirect_branch_register logic.
8569 (sibcall_insn_operand): Likewise.
8571 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8573 * config/i386/constraints.md (Bs): Replace
8574 ix86_indirect_branch_thunk_register with
8575 ix86_indirect_branch_register.
8577 * config/i386/i386.md (indirect_jump): Likewise.
8578 (tablejump): Likewise.
8579 (*sibcall_memory): Likewise.
8580 (*sibcall_value_memory): Likewise.
8581 Peepholes of indirect call and jump via memory: Likewise.
8582 * config/i386/i386.opt: Likewise.
8583 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8584 (GOT_memory_operand): Likewise.
8585 (call_insn_operand): Likewise.
8586 (sibcall_insn_operand): Likewise.
8587 (GOT32_symbol_operand): Likewise.
8589 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8592 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8593 type rather than type addr's type points to.
8594 (expand_omp_atomic_mutex): Likewise.
8595 (expand_omp_atomic): Likewise.
8597 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8600 * config/i386/i386.c (output_indirect_thunk_function): Use
8601 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8602 for __x86_return_thunk.
8604 2018-01-15 Richard Biener <rguenther@suse.de>
8607 * expmed.c (extract_bit_field_1): Fix typo.
8609 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8612 * config/arm/iterators.md (VF): New mode iterator.
8613 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8614 Remove integer-related logic from pattern.
8615 (neon_vabd<mode>_3): Likewise.
8617 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8620 * common.opt (fstrict-overflow): No longer an alias.
8621 (fwrapv-pointer): New option.
8622 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8623 also for pointer types based on flag_wrapv_pointer.
8624 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8625 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8626 opts->x_flag_wrapv got set.
8627 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8628 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8629 POINTER_TYPE_OVERFLOW_UNDEFINED.
8630 * match.pd: Likewise in address comparison pattern.
8631 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8633 2018-01-15 Richard Biener <rguenther@suse.de>
8636 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8637 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8638 Reset type names to their identifier if their TYPE_DECL doesn't
8639 have linkage (and thus is used for ODR and devirt).
8640 (save_debug_info_for_decl): Remove.
8641 (save_debug_info_for_type): Likewise.
8642 (add_tree_to_fld_list): Adjust.
8643 * tree-pretty-print.c (dump_generic_node): Make dumping of
8644 type names more robust.
8646 2018-01-15 Richard Biener <rguenther@suse.de>
8648 * BASE-VER: Bump to 8.0.1.
8650 2018-01-14 Martin Sebor <msebor@redhat.com>
8653 * builtins.c (check_access): Avoid warning when the no-warning bit
8656 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8658 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8659 * ira-color (allocno_hard_regs_compare): Likewise.
8661 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8664 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8665 Use .pushsection/.popsection.
8667 2018-01-14 Martin Sebor <msebor@redhat.com>
8670 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8672 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8674 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8675 entry from extra_headers.
8676 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8677 extra_headers, make the list bitwise identical to the i?86-*-* one.
8679 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8681 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8682 -mcmodel=large with -mindirect-branch=thunk,
8683 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8684 -mfunction-return=thunk-extern.
8685 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8686 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8687 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8689 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8691 * config/i386/i386.c (print_reg): Print the name of the full
8692 integer register without '%'.
8693 (ix86_print_operand): Handle 'V'.
8694 * doc/extend.texi: Document 'V' modifier.
8696 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8698 * config/i386/constraints.md (Bs): Disallow memory operand for
8699 -mindirect-branch-register.
8701 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8702 (GOT_memory_operand): Likewise.
8703 (call_insn_operand): Likewise.
8704 (sibcall_insn_operand): Likewise.
8705 (GOT32_symbol_operand): Likewise.
8706 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8707 for -mindirect-branch-register.
8708 (tablejump): Likewise.
8709 (*sibcall_memory): Likewise.
8710 (*sibcall_value_memory): Likewise.
8711 Disallow peepholes of indirect call and jump via memory for
8712 -mindirect-branch-register.
8713 (*call_pop): Replace m with Bw.
8714 (*call_value_pop): Likewise.
8715 (*sibcall_pop_memory): Replace m with Bs.
8716 * config/i386/i386.opt (mindirect-branch-register): New option.
8717 * doc/invoke.texi: Document -mindirect-branch-register option.
8719 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8721 * config/i386/i386-protos.h (ix86_output_function_return): New.
8722 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8723 set function_return_type.
8724 (indirect_thunk_name): Add ret_p to indicate thunk for function
8726 (output_indirect_thunk_function): Pass false to
8727 indirect_thunk_name.
8728 (ix86_output_indirect_branch_via_reg): Likewise.
8729 (ix86_output_indirect_branch_via_push): Likewise.
8730 (output_indirect_thunk_function): Create alias for function
8731 return thunk if regno < 0.
8732 (ix86_output_function_return): New function.
8733 (ix86_handle_fndecl_attribute): Handle function_return.
8734 (ix86_attribute_table): Add function_return.
8735 * config/i386/i386.h (machine_function): Add
8736 function_return_type.
8737 * config/i386/i386.md (simple_return_internal): Use
8738 ix86_output_function_return.
8739 (simple_return_internal_long): Likewise.
8740 * config/i386/i386.opt (mfunction-return=): New option.
8741 (indirect_branch): Mention -mfunction-return=.
8742 * doc/extend.texi: Document function_return function attribute.
8743 * doc/invoke.texi: Document -mfunction-return= option.
8745 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8747 * config/i386/i386-opts.h (indirect_branch): New.
8748 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8749 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8750 with local indirect jump when converting indirect call and jump.
8751 (ix86_set_indirect_branch_type): New.
8752 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8753 (indirectlabelno): New.
8754 (indirect_thunk_needed): Likewise.
8755 (indirect_thunk_bnd_needed): Likewise.
8756 (indirect_thunks_used): Likewise.
8757 (indirect_thunks_bnd_used): Likewise.
8758 (INDIRECT_LABEL): Likewise.
8759 (indirect_thunk_name): Likewise.
8760 (output_indirect_thunk): Likewise.
8761 (output_indirect_thunk_function): Likewise.
8762 (ix86_output_indirect_branch_via_reg): Likewise.
8763 (ix86_output_indirect_branch_via_push): Likewise.
8764 (ix86_output_indirect_branch): Likewise.
8765 (ix86_output_indirect_jmp): Likewise.
8766 (ix86_code_end): Call output_indirect_thunk_function if needed.
8767 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8769 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8770 (ix86_attribute_table): Add indirect_branch.
8771 * config/i386/i386.h (machine_function): Add indirect_branch_type
8772 and has_local_indirect_jump.
8773 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8775 (tablejump): Likewise.
8776 (*indirect_jump): Use ix86_output_indirect_jmp.
8777 (*tablejump_1): Likewise.
8778 (simple_return_indirect_internal): Likewise.
8779 * config/i386/i386.opt (mindirect-branch=): New option.
8780 (indirect_branch): New.
8783 (thunk-inline): Likewise.
8784 (thunk-extern): Likewise.
8785 * doc/extend.texi: Document indirect_branch function attribute.
8786 * doc/invoke.texi: Document -mindirect-branch= option.
8788 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8791 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8793 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8795 * ipa-inline.c (want_inline_small_function_p): Return false if
8796 inlining has already failed with CIF_FINAL_ERROR.
8797 (update_caller_keys): Call want_inline_small_function_p before
8799 (update_callee_keys): Likewise.
8801 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8803 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8805 (rs6000_quadword_masked_address_p): Likewise.
8806 (quad_aligned_load_p): Likewise.
8807 (quad_aligned_store_p): Likewise.
8808 (const_load_sequence_p): Add comment to describe the outer-most loop.
8809 (mimic_memory_attributes_and_flags): New function.
8810 (rs6000_gen_stvx): Likewise.
8811 (replace_swapped_aligned_store): Likewise.
8812 (rs6000_gen_lvx): Likewise.
8813 (replace_swapped_aligned_load): Likewise.
8814 (replace_swapped_load_constant): Capitalize argument name in
8815 comment describing this function.
8816 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8817 and stores that access quad-word aligned addresses and replace
8818 with stvx or lvx instructions when appropriate.
8819 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8820 New function prototype.
8821 (rs6000_quadword_masked_address_p): Likewise.
8822 (rs6000_gen_lvx): Likewise.
8823 (rs6000_gen_stvx): Likewise.
8824 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8825 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8826 when memory address is aligned.
8827 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8828 this split to select lvx instruction when memory address is aligned.
8829 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8830 instruction when memory address is aligned.
8831 (*vsx_le_perm_load_v16qi): Likewise.
8832 (four unnamed splitters): Modify to select the stvx instruction
8833 when memory is aligned.
8835 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8837 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8838 which appears in the queue multiple times.
8840 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8841 Alan Hayward <alan.hayward@arm.com>
8842 David Sherwood <david.sherwood@arm.com>
8844 * tree-vectorizer.h (vec_lower_bound): New structure.
8845 (_loop_vec_info): Add check_nonzero and lower_bounds.
8846 (LOOP_VINFO_CHECK_NONZERO): New macro.
8847 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8848 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8849 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8850 fields. Make seg_len the distance travelled, not including the
8852 (dr_direction_indicator): Declare.
8853 (dr_zero_step_indicator): Likewise.
8854 (dr_known_forward_stride_p): Likewise.
8855 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8857 (runtime_alias_check_p): Allow runtime alias checks with
8859 (operator ==): Compare access_size and align.
8860 (prune_runtime_alias_test_list): Rework for new distinction between
8861 the access_size and seg_len.
8862 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8864 (get_segment_min_max): New function.
8865 (create_intersect_range_checks): Use it.
8866 (dr_step_indicator): New function.
8867 (dr_direction_indicator): Likewise.
8868 (dr_zero_step_indicator): Likewise.
8869 (dr_known_forward_stride_p): Likewise.
8870 * tree-loop-distribution.c (data_ref_segment_size): Return
8871 DR_STEP * (niters - 1).
8872 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8874 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8875 (vect_preserves_scalar_order_p): New function, split out from...
8876 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8877 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8878 (vect_vfa_access_size): New function.
8879 (vect_vfa_align): Likewise.
8880 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8881 (dump_lower_bound): New function.
8882 (vect_check_lower_bound): Likewise.
8883 (vect_small_gap_p): Likewise.
8884 (vectorizable_with_step_bound_p): Likewise.
8885 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8886 depencies if the vectorization factor is 1. Convert the checks
8887 for nonzero steps into checks on the bounds of DR_STEP. Try using
8888 a bunds check for variable steps if the minimum required step is
8889 relatively small. Update calls to the dr_with_seg_len
8890 constructor and to vect_compile_time_alias.
8891 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8893 (vect_loop_versioning): Call it.
8894 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8896 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8898 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8899 Alan Hayward <alan.hayward@arm.com>
8900 David Sherwood <david.sherwood@arm.com>
8902 * doc/sourcebuild.texi (vect_scatter_store): Document.
8903 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8905 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8907 * genopinit.c (main): Add supports_vec_scatter_store and
8908 supports_vec_scatter_store_cached to target_optabs.
8909 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8910 IFN_MASK_SCATTER_STORE.
8911 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8913 * internal-fn.h (internal_store_fn_p): Declare.
8914 (internal_fn_stored_value_index): Likewise.
8915 * internal-fn.c (scatter_store_direct): New macro.
8916 (expand_scatter_store_optab_fn): New function.
8917 (direct_scatter_store_optab_supported_p): New macro.
8918 (internal_store_fn_p): New function.
8919 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8920 IFN_MASK_SCATTER_STORE.
8921 (internal_fn_mask_index): Likewise.
8922 (internal_fn_stored_value_index): New function.
8923 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8925 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8926 * optabs-query.c (supports_vec_scatter_store_p): New function.
8927 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8928 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8929 true for scatter stores.
8930 (vect_gather_scatter_fn_p): Handle scatter stores too.
8931 (vect_check_gather_scatter): Consider using scatter stores if
8932 supports_vec_scatter_store_p.
8933 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8935 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8936 internal_fn_stored_value_index.
8937 (check_load_store_masking): Handle scatter stores too.
8938 (vect_get_store_rhs): Make public.
8939 (vectorizable_call): Use internal_store_fn_p.
8940 (vectorizable_store): Handle scatter store internal functions.
8941 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8942 when deciding whether the end of the group has been reached.
8943 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8944 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8945 (mask_scatter_store<mode>): New insns.
8947 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8948 Alan Hayward <alan.hayward@arm.com>
8949 David Sherwood <david.sherwood@arm.com>
8951 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8952 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8953 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8955 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8956 Use vect_truncate_gather_scatter_offset if we can't treat the
8957 operation as a normal gather load or scatter store.
8958 (get_group_load_store_type): Take the gather_scatter_info
8959 as argument. Try using a gather load or scatter store for
8960 single-element groups.
8961 (get_load_store_type): Update calls to get_group_load_store_type
8962 and vect_use_strided_gather_scatters_p.
8964 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8965 Alan Hayward <alan.hayward@arm.com>
8966 David Sherwood <david.sherwood@arm.com>
8968 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8969 optional tree argument.
8970 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8972 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8973 but continue to use the current value as a fallback.
8974 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8975 to compare the updates.
8976 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8977 (get_load_store_type): Use it when handling a strided access.
8978 (vect_get_strided_load_store_ops): New function.
8979 (vect_get_data_ptr_increment): Likewise.
8980 (vectorizable_load): Handle strided gather loads. Always pass
8981 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8984 Alan Hayward <alan.hayward@arm.com>
8985 David Sherwood <david.sherwood@arm.com>
8987 * doc/md.texi (gather_load@var{m}): Document.
8988 (mask_gather_load@var{m}): Likewise.
8989 * genopinit.c (main): Add supports_vec_gather_load and
8990 supports_vec_gather_load_cached to target_optabs.
8991 * optabs-tree.c (init_tree_optimization_optabs): Use
8992 ggc_cleared_alloc to allocate target_optabs.
8993 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8994 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8996 * internal-fn.h (internal_load_fn_p): Declare.
8997 (internal_gather_scatter_fn_p): Likewise.
8998 (internal_fn_mask_index): Likewise.
8999 (internal_gather_scatter_fn_supported_p): Likewise.
9000 * internal-fn.c (gather_load_direct): New macro.
9001 (expand_gather_load_optab_fn): New function.
9002 (direct_gather_load_optab_supported_p): New macro.
9003 (direct_internal_fn_optab): New function.
9004 (internal_load_fn_p): Likewise.
9005 (internal_gather_scatter_fn_p): Likewise.
9006 (internal_fn_mask_index): Likewise.
9007 (internal_gather_scatter_fn_supported_p): Likewise.
9008 * optabs-query.c (supports_at_least_one_mode_p): New function.
9009 (supports_vec_gather_load_p): Likewise.
9010 * optabs-query.h (supports_vec_gather_load_p): Declare.
9011 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9012 and memory_type field.
9013 (NUM_PATTERNS): Bump to 15.
9014 * tree-vect-data-refs.c: Include internal-fn.h.
9015 (vect_gather_scatter_fn_p): New function.
9016 (vect_describe_gather_scatter_call): Likewise.
9017 (vect_check_gather_scatter): Try using internal functions for
9018 gather loads. Recognize existing calls to a gather load function.
9019 (vect_analyze_data_refs): Consider using gather loads if
9020 supports_vec_gather_load_p.
9021 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9022 (vect_get_gather_scatter_offset_type): Likewise.
9023 (vect_convert_mask_for_vectype): Likewise.
9024 (vect_add_conversion_to_patterm): Likewise.
9025 (vect_try_gather_scatter_pattern): Likewise.
9026 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9027 (vect_vect_recog_func_ptrs): Add it.
9028 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9029 internal_fn_mask_index and internal_gather_scatter_fn_p.
9030 (check_load_store_masking): Take the gather_scatter_info as an
9031 argument and handle gather loads.
9032 (vect_get_gather_scatter_ops): New function.
9033 (vectorizable_call): Check internal_load_fn_p.
9034 (vectorizable_load): Likewise. Handle gather load internal
9036 (vectorizable_store): Update call to check_load_store_masking.
9037 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9038 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9039 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9040 (aarch64_gather_scale_operand_d): New predicates.
9041 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9042 (mask_gather_load<mode>): New insns.
9044 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9045 Alan Hayward <alan.hayward@arm.com>
9046 David Sherwood <david.sherwood@arm.com>
9048 * optabs.def (fold_left_plus_optab): New optab.
9049 * doc/md.texi (fold_left_plus_@var{m}): Document.
9050 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9051 * internal-fn.c (fold_left_direct): Define.
9052 (expand_fold_left_optab_fn): Likewise.
9053 (direct_fold_left_optab_supported_p): Likewise.
9054 * fold-const-call.c (fold_const_fold_left): New function.
9055 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9056 * tree-parloops.c (valid_reduction_p): New function.
9057 (gather_scalar_reductions): Use it.
9058 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9059 (vect_finish_replace_stmt): Declare.
9060 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9061 (needs_fold_left_reduction_p): New function, split out from...
9062 (vect_is_simple_reduction): ...here. Accept reductions that
9063 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9064 (vect_force_simple_reduction): Also store the reduction type in
9065 the assignment's STMT_VINFO_REDUC_TYPE.
9066 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9067 (merge_with_identity): New function.
9068 (vect_expand_fold_left): Likewise.
9069 (vectorize_fold_left_reduction): Likewise.
9070 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9071 scalar phi in place for it. Check for target support and reject
9072 cases that would reassociate the operation. Defer the transform
9073 phase to vectorize_fold_left_reduction.
9074 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9075 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9076 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9078 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9080 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9081 call to ifc_temp_var.
9083 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9084 Alan Hayward <alan.hayward@arm.com>
9085 David Sherwood <david.sherwood@arm.com>
9087 * target.def (legitimize_address_displacement): Take the original
9088 offset as a poly_int.
9089 * targhooks.h (default_legitimize_address_displacement): Update
9091 * targhooks.c (default_legitimize_address_displacement): Likewise.
9092 * doc/tm.texi: Regenerate.
9093 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9094 as an argument, moving assert of ad->disp == ad->disp_term to...
9095 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9096 Try calling targetm.legitimize_address_displacement before expanding
9097 the address rather than afterwards, and adjust for the new interface.
9098 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9099 Match the new hook interface. Handle SVE addresses.
9100 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9103 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9105 * Makefile.in (OBJS): Add early-remat.o.
9106 * target.def (select_early_remat_modes): New hook.
9107 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9108 * doc/tm.texi: Regenerate.
9109 * targhooks.h (default_select_early_remat_modes): Declare.
9110 * targhooks.c (default_select_early_remat_modes): New function.
9111 * timevar.def (TV_EARLY_REMAT): New timevar.
9112 * passes.def (pass_early_remat): New pass.
9113 * tree-pass.h (make_pass_early_remat): Declare.
9114 * early-remat.c: New file.
9115 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9117 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9120 Alan Hayward <alan.hayward@arm.com>
9121 David Sherwood <david.sherwood@arm.com>
9123 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9124 vfm1 with a bound_epilog parameter.
9125 (vect_do_peeling): Update calls accordingly, and move the prologue
9126 call earlier in the function. Treat the base bound_epilog as 0 for
9127 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9128 this base when peeling for gaps.
9129 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9130 with fully-masked loops.
9131 (vect_estimate_min_profitable_iters): Handle the single peeled
9132 iteration in that case.
9134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9135 Alan Hayward <alan.hayward@arm.com>
9136 David Sherwood <david.sherwood@arm.com>
9138 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9139 single-element interleaving even if the size is not a power of 2.
9140 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9141 accesses for single-element interleaving if the group size is
9144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9145 Alan Hayward <alan.hayward@arm.com>
9146 David Sherwood <david.sherwood@arm.com>
9148 * doc/md.texi (fold_extract_last_@var{m}): Document.
9149 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9150 * optabs.def (fold_extract_last_optab): New optab.
9151 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9152 * internal-fn.c (fold_extract_direct): New macro.
9153 (expand_fold_extract_optab_fn): Likewise.
9154 (direct_fold_extract_optab_supported_p): Likewise.
9155 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9156 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9157 EXTRACT_LAST_REDUCTION.
9158 (get_initial_def_for_reduction): Do not create an initial vector
9159 for EXTRACT_LAST_REDUCTION reductions.
9160 (vectorizable_reduction): Leave the scalar phi in place for
9161 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9162 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9163 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9164 transform phase to vectorizable_condition.
9165 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9167 (vect_finish_stmt_generation): ...here.
9168 (vect_finish_replace_stmt): New function.
9169 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9170 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9172 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9174 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9175 Alan Hayward <alan.hayward@arm.com>
9176 David Sherwood <david.sherwood@arm.com>
9178 * doc/md.texi (extract_last_@var{m}): Document.
9179 * optabs.def (extract_last_optab): New optab.
9180 * internal-fn.def (EXTRACT_LAST): New internal function.
9181 * internal-fn.c (cond_unary_direct): New macro.
9182 (expand_cond_unary_optab_fn): Likewise.
9183 (direct_cond_unary_optab_supported_p): Likewise.
9184 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9185 loops using EXTRACT_LAST.
9186 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9187 (extract_last_<mode>): ...this optab.
9188 (vec_extract<mode><Vel>): Update accordingly.
9190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9191 Alan Hayward <alan.hayward@arm.com>
9192 David Sherwood <david.sherwood@arm.com>
9194 * target.def (empty_mask_is_expensive): New hook.
9195 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9196 * doc/tm.texi: Regenerate.
9197 * targhooks.h (default_empty_mask_is_expensive): Declare.
9198 * targhooks.c (default_empty_mask_is_expensive): New function.
9199 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9200 if the target says that empty masks are expensive.
9201 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9203 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9205 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9206 Alan Hayward <alan.hayward@arm.com>
9207 David Sherwood <david.sherwood@arm.com>
9209 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9210 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9211 (vect_use_loop_mask_for_alignment_p): New function.
9212 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9213 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9214 niters_skip argument. Make sure that the first niters_skip elements
9215 of the first iteration are inactive.
9216 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9217 Update call to vect_set_loop_masks_directly.
9218 (get_misalign_in_elems): New function, split out from...
9219 (vect_gen_prolog_loop_niters): ...here.
9220 (vect_update_init_of_dr): Take a code argument that specifies whether
9221 the adjustment should be added or subtracted.
9222 (vect_update_init_of_drs): Likewise.
9223 (vect_prepare_for_masked_peels): New function.
9224 (vect_do_peeling): Skip prologue peeling if we're using a mask
9225 instead. Update call to vect_update_inits_of_drs.
9226 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9228 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9229 alignment. Do not include the number of peeled iterations in
9230 the minimum threshold in that case.
9231 (vectorizable_induction): Adjust the start value down by
9232 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9233 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9234 Take the number of skipped iterations into account when calculating
9236 * tree-vect-stmts.c (vect_gen_while_not): New function.
9238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9239 Alan Hayward <alan.hayward@arm.com>
9240 David Sherwood <david.sherwood@arm.com>
9242 * doc/sourcebuild.texi (vect_fully_masked): Document.
9243 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9245 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9247 (vect_analyze_loop_2): ...here. Don't check the vectorization
9248 factor against the number of loop iterations if the loop is
9251 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9252 Alan Hayward <alan.hayward@arm.com>
9253 David Sherwood <david.sherwood@arm.com>
9255 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9256 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9257 (dump_groups): Update accordingly.
9258 (iv_use::mem_type): New member variable.
9259 (address_p): New function.
9260 (record_use): Add a mem_type argument and initialize the new
9262 (record_group_use): Add a mem_type argument. Use address_p.
9263 Remove obsolete null checks of base_object. Update call to record_use.
9264 (find_interesting_uses_op): Update call to record_group_use.
9265 (find_interesting_uses_cond): Likewise.
9266 (find_interesting_uses_address): Likewise.
9267 (get_mem_type_for_internal_fn): New function.
9268 (find_address_like_use): Likewise.
9269 (find_interesting_uses_stmt): Try find_address_like_use before
9270 calling find_interesting_uses_op.
9271 (addr_offset_valid_p): Use the iv mem_type field as the type
9272 of the addressed memory.
9273 (add_autoinc_candidates): Likewise.
9274 (get_address_cost): Likewise.
9275 (split_small_address_groups_p): Use address_p.
9276 (split_address_groups): Likewise.
9277 (add_iv_candidate_for_use): Likewise.
9278 (autoinc_possible_for_pair): Likewise.
9279 (rewrite_groups): Likewise.
9280 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9281 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9282 (get_alias_ptr_type_for_ptr_address): New function.
9283 (rewrite_use_address): Rewrite address uses in calls that were
9284 identified by find_address_like_use.
9286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9287 Alan Hayward <alan.hayward@arm.com>
9288 David Sherwood <david.sherwood@arm.com>
9290 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9292 * gimple-expr.h (is_gimple_addressable: Likewise.
9293 * gimple-expr.c (is_gimple_address): Likewise.
9294 * internal-fn.c (expand_call_mem_ref): New function.
9295 (expand_mask_load_optab_fn): Use it.
9296 (expand_mask_store_optab_fn): Likewise.
9298 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9299 Alan Hayward <alan.hayward@arm.com>
9300 David Sherwood <david.sherwood@arm.com>
9302 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9303 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9304 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9305 (cond_umax@var{mode}): Document.
9306 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9307 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9308 (cond_umin_optab, cond_umax_optab): New optabs.
9309 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9310 (COND_IOR, COND_XOR): New internal functions.
9311 * internal-fn.h (get_conditional_internal_fn): Declare.
9312 * internal-fn.c (cond_binary_direct): New macro.
9313 (expand_cond_binary_optab_fn): Likewise.
9314 (direct_cond_binary_optab_supported_p): Likewise.
9315 (get_conditional_internal_fn): New function.
9316 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9317 Cope with reduction statements that are vectorized as calls rather
9319 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9320 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9321 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9322 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9323 (UNSPEC_COND_EOR): New unspecs.
9324 (optab): Add mappings for them.
9325 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9326 (sve_int_op, sve_fp_op): New int attributes.
9328 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9329 Alan Hayward <alan.hayward@arm.com>
9330 David Sherwood <david.sherwood@arm.com>
9332 * optabs.def (while_ult_optab): New optab.
9333 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9334 * internal-fn.def (WHILE_ULT): New internal function.
9335 * internal-fn.h (direct_internal_fn_supported_p): New override
9336 that takes two types as argument.
9337 * internal-fn.c (while_direct): New macro.
9338 (expand_while_optab_fn): New function.
9339 (convert_optab_supported_p): Likewise.
9340 (direct_while_optab_supported_p): New macro.
9341 * wide-int.h (wi::udiv_ceil): New function.
9342 * tree-vectorizer.h (rgroup_masks): New structure.
9343 (vec_loop_masks): New typedef.
9344 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9346 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9347 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9348 (vect_max_vf): New function.
9349 (slpeel_make_loop_iterate_ntimes): Delete.
9350 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9351 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9352 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9353 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9354 internal-fn.h, stor-layout.h and optabs-query.h.
9355 (vect_set_loop_mask): New function.
9356 (add_preheader_seq): Likewise.
9357 (add_header_seq): Likewise.
9358 (interleave_supported_p): Likewise.
9359 (vect_maybe_permute_loop_masks): Likewise.
9360 (vect_set_loop_masks_directly): Likewise.
9361 (vect_set_loop_condition_masked): Likewise.
9362 (vect_set_loop_condition_unmasked): New function, split out from
9363 slpeel_make_loop_iterate_ntimes.
9364 (slpeel_make_loop_iterate_ntimes): Rename to..
9365 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9366 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9367 (vect_do_peeling): Update call accordingly.
9368 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9370 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9371 mask_compare_type, can_fully_mask_p and fully_masked_p.
9372 (release_vec_loop_masks): New function.
9373 (_loop_vec_info): Use it to free the loop masks.
9374 (can_produce_all_loop_masks_p): New function.
9375 (vect_get_max_nscalars_per_iter): Likewise.
9376 (vect_verify_full_masking): Likewise.
9377 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9378 retries, and free the mask rgroups before retrying. Check loop-wide
9379 reasons for disallowing fully-masked loops. Make the final decision
9380 about whether use a fully-masked loop or not.
9381 (vect_estimate_min_profitable_iters): Do not assume that peeling
9382 for the number of iterations will be needed for fully-masked loops.
9383 (vectorizable_reduction): Disable fully-masked loops.
9384 (vectorizable_live_operation): Likewise.
9385 (vect_halve_mask_nunits): New function.
9386 (vect_double_mask_nunits): Likewise.
9387 (vect_record_loop_mask): Likewise.
9388 (vect_get_loop_mask): Likewise.
9389 (vect_transform_loop): Handle the case in which the final loop
9390 iteration might handle a partial vector. Call vect_set_loop_condition
9391 instead of slpeel_make_loop_iterate_ntimes.
9392 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9393 (check_load_store_masking): New function.
9394 (prepare_load_store_mask): Likewise.
9395 (vectorizable_store): Handle fully-masked loops.
9396 (vectorizable_load): Likewise.
9397 (supportable_widening_operation): Use vect_halve_mask_nunits for
9399 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9400 (vect_gen_while): New function.
9401 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9402 (aarch64_uqdec<mode>): New insn.
9404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9405 Alan Hayward <alan.hayward@arm.com>
9406 David Sherwood <david.sherwood@arm.com>
9408 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9409 (reduc_xor_scal_optab): New optabs.
9410 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9411 (reduc_xor_scal_@var{m}): Document.
9412 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9413 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9415 * fold-const-call.c (fold_const_call): Handle them.
9416 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9417 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9418 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9419 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9420 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9421 (UNSPEC_XORV): New unspecs.
9422 (optab): Add entries for them.
9423 (BITWISEV): New int iterator.
9424 (bit_reduc_op): New int attributes.
9426 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9427 Alan Hayward <alan.hayward@arm.com>
9428 David Sherwood <david.sherwood@arm.com>
9430 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9431 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9432 * optabs.def (vec_shl_insert_optab): New optab.
9433 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9434 (duplicate_and_interleave): Likewise.
9435 * tree-vect-loop.c: Include internal-fn.h.
9436 (neutral_op_for_slp_reduction): New function, split out from
9437 get_initial_defs_for_reduction.
9438 (get_initial_def_for_reduction): Handle option 2 for variable-length
9439 vectors by loading the neutral value into a vector and then shifting
9440 the initial value into element 0.
9441 (get_initial_defs_for_reduction): Replace the code argument with
9442 the neutral value calculated by neutral_op_for_slp_reduction.
9443 Use gimple_build_vector for constant-length vectors.
9444 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9445 but the first group_size elements have a neutral value.
9446 Use duplicate_and_interleave otherwise.
9447 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9448 Update call to get_initial_defs_for_reduction. Handle SLP
9449 reductions for variable-length vectors by creating one vector
9450 result for each scalar result, with the elements associated
9451 with other scalar results stubbed out with the neutral value.
9452 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9453 Require IFN_VEC_SHL_INSERT for double reductions on
9454 variable-length vectors, or SLP reductions that have
9455 a neutral value. Require can_duplicate_and_interleave_p
9456 support for variable-length unchained SLP reductions if there
9457 is no neutral value, such as for MIN/MAX reductions. Also require
9458 the number of vector elements to be a multiple of the number of
9459 SLP statements when doing variable-length unchained SLP reductions.
9460 Update call to vect_create_epilog_for_reduction.
9461 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9462 and remove initial values.
9463 (duplicate_and_interleave): Make public.
9464 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9465 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9468 Alan Hayward <alan.hayward@arm.com>
9469 David Sherwood <david.sherwood@arm.com>
9471 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9472 (can_duplicate_and_interleave_p): New function.
9473 (vect_get_and_check_slp_defs): Take the vector of statements
9474 rather than just the current one. Remove excess parentheses.
9475 Restriction rejectinon of vect_constant_def and vect_external_def
9476 for variable-length vectors to boolean types, or types for which
9477 can_duplicate_and_interleave_p is false.
9478 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9479 (duplicate_and_interleave): New function.
9480 (vect_get_constant_vectors): Use gimple_build_vector for
9481 constant-length vectors and suitable variable-length constant
9482 vectors. Use duplicate_and_interleave for other variable-length
9483 vectors. Don't defer the update when inserting new statements.
9485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9486 Alan Hayward <alan.hayward@arm.com>
9487 David Sherwood <david.sherwood@arm.com>
9489 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9490 min_profitable_iters doesn't go negative.
9492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9493 Alan Hayward <alan.hayward@arm.com>
9494 David Sherwood <david.sherwood@arm.com>
9496 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9497 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9498 * optabs.def (vec_mask_load_lanes_optab): New optab.
9499 (vec_mask_store_lanes_optab): Likewise.
9500 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9501 (MASK_STORE_LANES): Likewise.
9502 * internal-fn.c (mask_load_lanes_direct): New macro.
9503 (mask_store_lanes_direct): Likewise.
9504 (expand_mask_load_optab_fn): Handle masked operations.
9505 (expand_mask_load_lanes_optab_fn): New macro.
9506 (expand_mask_store_optab_fn): Handle masked operations.
9507 (expand_mask_store_lanes_optab_fn): New macro.
9508 (direct_mask_load_lanes_optab_supported_p): Likewise.
9509 (direct_mask_store_lanes_optab_supported_p): Likewise.
9510 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9512 (vect_load_lanes_supported): Likewise.
9513 * tree-vect-data-refs.c (strip_conversion): New function.
9514 (can_group_stmts_p): Likewise.
9515 (vect_analyze_data_ref_accesses): Use it instead of checking
9516 for a pair of assignments.
9517 (vect_store_lanes_supported): Take a masked_p parameter.
9518 (vect_load_lanes_supported): Likewise.
9519 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9520 vect_store_lanes_supported and vect_load_lanes_supported.
9521 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9522 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9523 parameter. Don't allow gaps for masked accesses.
9524 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9525 and vect_load_lanes_supported.
9526 (get_load_store_type): Take a masked_p parameter and update
9527 call to get_group_load_store_type.
9528 (vectorizable_store): Update call to get_load_store_type.
9529 Handle IFN_MASK_STORE_LANES.
9530 (vectorizable_load): Update call to get_load_store_type.
9531 Handle IFN_MASK_LOAD_LANES.
9533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9534 Alan Hayward <alan.hayward@arm.com>
9535 David Sherwood <david.sherwood@arm.com>
9537 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9539 * config/aarch64/aarch64-protos.h
9540 (aarch64_sve_struct_memory_operand_p): Declare.
9541 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9542 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9543 (VPRED, vpred): Handle SVE structure modes.
9544 * config/aarch64/constraints.md (Utx): New constraint.
9545 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9546 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9547 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9548 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9549 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9550 structure modes. Split into pieces after RA.
9551 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9552 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9554 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9555 SVE structure modes.
9556 (aarch64_classify_address): Likewise.
9557 (sizetochar): Move earlier in file.
9558 (aarch64_print_operand): Handle SVE register lists.
9559 (aarch64_array_mode): New function.
9560 (aarch64_sve_struct_memory_operand_p): Likewise.
9561 (TARGET_ARRAY_MODE): Redefine.
9563 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9564 Alan Hayward <alan.hayward@arm.com>
9565 David Sherwood <david.sherwood@arm.com>
9567 * target.def (array_mode): New target hook.
9568 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9569 * doc/tm.texi: Regenerate.
9570 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9571 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9572 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9574 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9578 Alan Hayward <alan.hayward@arm.com>
9579 David Sherwood <david.sherwood@arm.com>
9581 * fold-const.c (fold_binary_loc): Check the argument types
9582 rather than the result type when testing for a vector operation.
9584 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9586 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9587 * doc/tm.texi: Regenerate.
9589 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9590 Alan Hayward <alan.hayward@arm.com>
9591 David Sherwood <david.sherwood@arm.com>
9593 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9594 (sve): Document new AArch64 extension.
9595 * doc/md.texi (w): Extend the description of the AArch64
9596 constraint to include SVE vectors.
9597 (Upl, Upa): Document new AArch64 predicate constraints.
9598 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9600 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9601 (msve-vector-bits=): New option.
9602 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9603 SVE when these are disabled.
9604 (sve): New extension.
9605 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9606 modes. Adjust their number of units based on aarch64_sve_vg.
9607 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9608 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9609 aarch64_addr_query_type.
9610 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9611 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9612 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9613 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9614 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9615 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9616 (aarch64_simd_imm_zero_p): Delete.
9617 (aarch64_check_zero_based_sve_index_immediate): Declare.
9618 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9619 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9620 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9621 (aarch64_sve_float_mul_immediate_p): Likewise.
9622 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9624 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9625 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9626 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9627 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9628 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9629 (aarch64_regmode_natural_size): Likewise.
9630 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9631 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9633 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9634 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9635 for VG and the SVE predicate registers.
9636 (V_ALIASES): Add a "z"-prefixed alias.
9637 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9638 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9639 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9640 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9641 (REG_CLASS_NAMES): Add entries for them.
9642 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9643 and the predicate registers.
9644 (aarch64_sve_vg): Declare.
9645 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9646 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9647 (REGMODE_NATURAL_SIZE): Define.
9648 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9650 * config/aarch64/aarch64.c: Include cfgrtl.h.
9651 (simd_immediate_info): Add a constructor for series vectors,
9652 and an associated step field.
9653 (aarch64_sve_vg): New variable.
9654 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9655 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9656 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9657 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9658 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9659 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9660 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9661 (aarch64_get_mask_mode): New functions.
9662 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9663 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9664 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9665 predicate modes and predicate registers. Explicitly restrict
9666 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9667 to store a vector mode if it is recognized by
9668 aarch64_classify_vector_mode.
9669 (aarch64_regmode_natural_size): New function.
9670 (aarch64_hard_regno_caller_save_mode): Return the original mode
9672 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9673 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9674 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9675 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9677 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9678 does not overlap dest if the function is frame-related. Handle
9680 (aarch64_split_add_offset): New function.
9681 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9682 them aarch64_add_offset.
9683 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9684 and update call to aarch64_sub_sp.
9685 (aarch64_add_cfa_expression): New function.
9686 (aarch64_expand_prologue): Pass extra temporary registers to the
9687 functions above. Handle the case in which we need to emit new
9688 DW_CFA_expressions for registers that were originally saved
9689 relative to the stack pointer, but now have to be expressed
9690 relative to the frame pointer.
9691 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9693 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9694 IP0 and IP1 values for SVE frames.
9695 (aarch64_expand_vec_series): New function.
9696 (aarch64_expand_sve_widened_duplicate): Likewise.
9697 (aarch64_expand_sve_const_vector): Likewise.
9698 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9699 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9700 into the register, rather than emitting a SET directly.
9701 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9702 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9703 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9704 (offset_9bit_signed_scaled_p): New functions.
9705 (aarch64_replicate_bitmask_imm): New function.
9706 (aarch64_bitmask_imm): Use it.
9707 (aarch64_cannot_force_const_mem): Reject expressions involving
9708 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9709 (aarch64_classify_index): Handle SVE indices, by requiring
9710 a plain register index with a scale that matches the element size.
9711 (aarch64_classify_address): Handle SVE addresses. Assert that
9712 the mode of the address is VOIDmode or an integer mode.
9713 Update call to aarch64_classify_symbol.
9714 (aarch64_classify_symbolic_expression): Update call to
9715 aarch64_classify_symbol.
9716 (aarch64_const_vec_all_in_range_p): New function.
9717 (aarch64_print_vector_float_operand): Likewise.
9718 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9719 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9720 and the FP immediates 1.0 and 0.5.
9721 (aarch64_print_address_internal): Handle SVE addresses.
9722 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9723 (aarch64_regno_regclass): Handle predicate registers.
9724 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9726 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9727 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9728 (aarch64_convert_sve_vector_bits): New function.
9729 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9730 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9732 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9733 Handle SVE vector and predicate modes. Accept VL-based constants
9734 that need only one temporary register, and VL offsets that require
9735 no temporary registers.
9736 (aarch64_conditional_register_usage): Mark the predicate registers
9737 as fixed if SVE isn't available.
9738 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9739 Return true for SVE vector and predicate modes.
9740 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9741 rather than an unsigned int. Handle SVE modes.
9742 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9744 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9746 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9747 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9748 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9749 (aarch64_sve_float_mul_immediate_p): New functions.
9750 (aarch64_sve_valid_immediate): New function.
9751 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9752 Explicitly reject structure modes. Check for INDEX constants.
9753 Handle PTRUE and PFALSE constants.
9754 (aarch64_check_zero_based_sve_index_immediate): New function.
9755 (aarch64_simd_imm_zero_p): Delete.
9756 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9757 vector modes. Accept constants in the range of CNT[BHWD].
9758 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9759 ask for an Advanced SIMD mode.
9760 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9761 (aarch64_simd_vector_alignment): Handle SVE predicates.
9762 (aarch64_vectorize_preferred_vector_alignment): New function.
9763 (aarch64_simd_vector_alignment_reachable): Use it instead of
9765 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9766 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9768 (MAX_VECT_LEN): Delete.
9769 (expand_vec_perm_d): Add a vec_flags field.
9770 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9771 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9772 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9774 (aarch64_evpc_rev): Rename to...
9775 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9776 (aarch64_evpc_rev_global): New function.
9777 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9778 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9780 (aarch64_evpc_sve_tbl): New function.
9781 (aarch64_expand_vec_perm_const_1): Update after rename of
9782 aarch64_evpc_rev. Handle SVE permutes too, trying
9783 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9784 than aarch64_evpc_tbl.
9785 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9786 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9787 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9788 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9789 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9790 (aarch64_expand_sve_vcond): New functions.
9791 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9792 of aarch64_vector_mode_p.
9793 (aarch64_dwarf_poly_indeterminate_value): New function.
9794 (aarch64_compute_pressure_classes): Likewise.
9795 (aarch64_can_change_mode_class): Likewise.
9796 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9797 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9798 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9799 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9800 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9801 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9802 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9803 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9805 (Dn, Dl, Dr): Accept const as well as const_vector.
9806 (Dz): Likewise. Compare against CONST0_RTX.
9807 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9808 of "vector" where appropriate.
9809 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9810 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9811 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9812 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9813 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9814 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9815 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9816 (v_int_equiv): Extend to SVE modes.
9817 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9819 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9820 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9821 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9822 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9823 (SVE_COND_FP_CMP): New int iterators.
9824 (perm_hilo): Handle the new unpack unspecs.
9825 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9827 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9828 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9829 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9830 (aarch64_equality_operator, aarch64_constant_vector_operand)
9831 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9832 (aarch64_sve_nonimmediate_operand): Likewise.
9833 (aarch64_sve_general_operand): Likewise.
9834 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9835 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9836 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9837 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9838 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9839 (aarch64_sve_float_arith_immediate): Likewise.
9840 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9841 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9842 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9843 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9844 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9845 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9846 (aarch64_sve_float_arith_operand): Likewise.
9847 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9848 (aarch64_sve_float_mul_operand): Likewise.
9849 (aarch64_sve_vec_perm_operand): Likewise.
9850 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9851 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9852 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9853 as well as const_vector.
9854 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9855 in file. Use CONST0_RTX and CONSTM1_RTX.
9856 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9857 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9858 Use aarch64_simd_imm_zero.
9859 * config/aarch64/aarch64-sve.md: New file.
9860 * config/aarch64/aarch64.md: Include it.
9861 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9862 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9863 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9864 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9865 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9866 (sve): New attribute.
9867 (enabled): Disable instructions with the sve attribute unless
9869 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9870 aarch64_expand_mov_immediate.
9871 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9872 CNT[BHSD] immediates.
9873 (movti): Split CONST_POLY_INT moves into two halves.
9874 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9875 Split additions that need a temporary here if the destination
9876 is the stack pointer.
9877 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9878 (*add<mode>3_poly_1): New instruction.
9879 (set_clobber_cc): New expander.
9881 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9883 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9884 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9885 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9886 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9887 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9888 Change innermode from fixed_mode_size to machine_mode.
9889 (simplify_subreg): Update call accordingly. Handle a constant-sized
9890 subreg of a variable-length CONST_VECTOR.
9892 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9893 Alan Hayward <alan.hayward@arm.com>
9894 David Sherwood <david.sherwood@arm.com>
9896 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9897 (add_offset_to_base): New function, split out from...
9898 (create_mem_ref): ...here. When handling a scale other than 1,
9899 check first whether the address is valid without the offset.
9900 Add it into the base if so, leaving the index and scale as-is.
9902 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9905 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9906 fold_for_warn before checking if arg2 is INTEGER_CST.
9908 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9910 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9911 (store_multiple_operation): Delete.
9912 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9913 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9914 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9915 guarded by TARGET_STRING.
9916 (rs6000_output_load_multiple): Delete.
9917 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9918 OPTION_MASK_STRING / TARGET_STRING handling.
9919 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9920 (const rs6000_opt_masks) <"string">: Change mask to 0.
9921 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9922 (MASK_STRING): Delete.
9923 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9925 (load_multiple): Delete.
9932 (store_multiple): Delete.
9939 (movmemsi_8reg): Delete.
9940 (corresponding unnamed define_insn): Delete.
9941 (movmemsi_6reg): Delete.
9942 (corresponding unnamed define_insn): Delete.
9943 (movmemsi_4reg): Delete.
9944 (corresponding unnamed define_insn): Delete.
9945 (movmemsi_2reg): Delete.
9946 (corresponding unnamed define_insn): Delete.
9947 (movmemsi_1reg): Delete.
9948 (corresponding unnamed define_insn): Delete.
9949 * config/rs6000/rs6000.opt (mno-string): New.
9950 (mstring): Replace by deprecation warning stub.
9951 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9953 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9955 * regrename.c (regrename_do_replace): If replacing the same
9956 reg multiple times, try to reuse last created gen_raw_REG.
9959 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9960 main to workaround a bug in GDB.
9962 2018-01-12 Tom de Vries <tom@codesourcery.com>
9965 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9967 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9969 PR rtl-optimization/80481
9970 * ira-color.c (get_cap_member): New function.
9971 (allocnos_conflict_by_live_ranges_p): Use it.
9972 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9973 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9975 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9978 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9979 (*saddl_se_1): Ditto.
9981 (*ssubl_se_1): Ditto.
9983 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9985 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9986 rather than wi::to_widest for DR_INITs.
9987 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9988 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9989 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9991 (vect_analyze_group_access_1): Note that here.
9993 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9995 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9996 polynomial type sizes.
9998 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10000 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10001 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10002 (gimple_add_tmp_var): Likewise.
10004 2018-01-12 Martin Liska <mliska@suse.cz>
10006 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10007 (gimple_alloc_sizes): Likewise.
10008 (dump_gimple_statistics): Use PRIu64 in printf format.
10009 * gimple.h: Change uint64_t to int.
10011 2018-01-12 Martin Liska <mliska@suse.cz>
10013 * tree-core.h: Use uint64_t instead of int.
10014 * tree.c (tree_node_counts): Likewise.
10015 (tree_node_sizes): Likewise.
10016 (dump_tree_statistics): Use PRIu64 in printf format.
10018 2018-01-12 Martin Liska <mliska@suse.cz>
10020 * Makefile.in: As qsort_chk is implemented in vec.c, add
10021 vec.o to linkage of gencfn-macros.
10022 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10023 passing the info to record_node_allocation_statistics.
10024 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10026 * ggc-common.c (struct ggc_usage): Add operator== and use
10027 it in operator< and compare function.
10028 * mem-stats.h (struct mem_usage): Likewise.
10029 * vec.c (struct vec_usage): Remove operator< and compare
10030 function. Can be simply inherited.
10032 2018-01-12 Martin Jambor <mjambor@suse.cz>
10035 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10036 * tree-ssa-math-opts.c: Include domwalk.h.
10037 (convert_mult_to_fma_1): New function.
10038 (fma_transformation_info): New type.
10039 (fma_deferring_state): Likewise.
10040 (cancel_fma_deferring): New function.
10041 (result_of_phi): Likewise.
10042 (last_fma_candidate_feeds_initial_phi): Likewise.
10043 (convert_mult_to_fma): Added deferring logic, split actual
10044 transformation to convert_mult_to_fma_1.
10045 (math_opts_dom_walker): New type.
10046 (math_opts_dom_walker::after_dom_children): New method, body moved
10047 here from pass_optimize_widening_mul::execute, added deferring logic
10049 (pass_optimize_widening_mul::execute): Moved most of code to
10050 math_opts_dom_walker::after_dom_children.
10051 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10052 * config/i386/i386.c (ix86_option_override_internal): Added
10053 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10055 2018-01-12 Richard Biener <rguenther@suse.de>
10058 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10059 inline instance vars.
10061 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10064 * config/rx/rx.c (rx_is_restricted_memory_address):
10065 Handle SUBREG case.
10067 2018-01-12 Richard Biener <rguenther@suse.de>
10069 PR tree-optimization/80846
10070 * target.def (split_reduction): New target hook.
10071 * targhooks.c (default_split_reduction): New function.
10072 * targhooks.h (default_split_reduction): Declare.
10073 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10074 target requests first reduce vectors by combining low and high
10076 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10077 (get_vectype_for_scalar_type_and_size): Export.
10078 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10079 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10080 * doc/tm.texi: Regenerate.
10081 * config/i386/i386.c (ix86_split_reduction): Implement
10082 TARGET_VECTORIZE_SPLIT_REDUCTION.
10084 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10087 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10088 in PIC mode except for TARGET_VXWORKS_RTP.
10089 * config/sparc/sparc.c: Include cfgrtl.h.
10090 (TARGET_INIT_PIC_REG): Define.
10091 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10092 (sparc_pic_register_p): New predicate.
10093 (sparc_legitimate_address_p): Use it.
10094 (sparc_legitimize_pic_address): Likewise.
10095 (sparc_delegitimize_address): Likewise.
10096 (sparc_mode_dependent_address_p): Likewise.
10097 (gen_load_pcrel_sym): Remove 4th parameter.
10098 (load_got_register): Adjust call to above. Remove obsolete stuff.
10099 (sparc_expand_prologue): Do not call load_got_register here.
10100 (sparc_flat_expand_prologue): Likewise.
10101 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10102 (sparc_use_pseudo_pic_reg): New function.
10103 (sparc_init_pic_reg): Likewise.
10104 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10105 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10107 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10109 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10110 Add item for branch_cost.
10112 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10114 PR rtl-optimization/83565
10115 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10116 not extend the result to a larger mode for rotate operations.
10117 (num_sign_bit_copies1): Likewise.
10119 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10122 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10124 Use values-Xc.o for -pedantic.
10125 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10127 2018-01-12 Martin Liska <mliska@suse.cz>
10130 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10132 (possible_polymorphic_call_targets): Use it.
10133 (ipa_devirt): Likewise.
10135 2018-01-12 Martin Liska <mliska@suse.cz>
10137 * profile-count.h (enum profile_quality): Use 0 as invalid
10138 enum value of profile_quality.
10140 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10142 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10143 -mext-string options.
10145 2018-01-12 Richard Biener <rguenther@suse.de>
10147 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10148 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10149 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10151 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10153 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10155 * configure.ac (--with-long-double-format): Add support for the
10156 configuration option to change the default long double format on
10158 * config.gcc (powerpc*-linux*-*): Likewise.
10159 * configure: Regenerate.
10160 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10161 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10162 used without modification.
10164 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10166 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10167 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10168 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10169 MISC_BUILTIN_SPEC_BARRIER.
10170 (rs6000_init_builtins): Likewise.
10171 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10173 (speculation_barrier): New define_insn.
10174 * doc/extend.texi: Document __builtin_speculation_barrier.
10176 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10179 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10180 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10181 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10183 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10184 integral modes instead of "ss" and "sd".
10185 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10186 vectors with 32-bit and 64-bit elements.
10187 (vecdupssescalarmodesuffix): New mode attribute.
10188 (vec_dup<mode>): Use it.
10190 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10193 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10194 frame if argument is passed on stack.
10196 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10199 * ree.c (combine_reaching_defs): Optimize also
10200 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10201 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10203 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10205 PR middle-end/83189
10206 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10208 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10210 PR middle-end/83718
10211 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10212 after they are computed.
10214 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10216 PR tree-optimization/83695
10217 * gimple-loop-linterchange.cc
10218 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10219 reset cached scev information after interchange.
10220 (pass_linterchange::execute): Remove call to scev_reset_htab.
10222 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10224 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10225 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10226 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10227 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10228 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10229 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10230 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10231 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10232 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10233 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10234 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10235 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10236 (V_lane_reg): Likewise.
10237 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10239 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10240 (vfmal_lane_low<mode>_intrinsic,
10241 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10242 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10243 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10244 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10245 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10246 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10248 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10250 * config/arm/arm-cpus.in (fp16fml): New feature.
10251 (ALL_SIMD): Add fp16fml.
10252 (armv8.2-a): Add fp16fml as an option.
10253 (armv8.3-a): Likewise.
10254 (armv8.4-a): Add fp16fml as part of fp16.
10255 * config/arm/arm.h (TARGET_FP16FML): Define.
10256 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10258 * config/arm/arm-modes.def (V2HF): Define.
10259 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10260 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10261 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10262 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10263 vfmsl_low, vfmsl_high): New set of builtins.
10264 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10265 (vfml_op): New code attribute.
10266 (VFMLHALVES): New int iterator.
10267 (VFML, VFMLSEL): New mode attributes.
10268 (V_reg): Define mapping for V2HF.
10269 (V_hi, V_lo): New mode attributes.
10270 (VF_constraint): Likewise.
10271 (vfml_half, vfml_half_selector): New int attributes.
10272 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10274 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10275 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10277 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10278 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10279 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10280 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10282 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10283 Document new effective target and option set.
10285 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10287 * config/arm/arm-cpus.in (armv8_4): New feature.
10288 (ARMv8_4a): New fgroup.
10289 (armv8.4-a): New arch.
10290 * config/arm/arm-tables.opt: Regenerate.
10291 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10292 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10293 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10294 Add matching rules for -march=armv8.4-a and extensions.
10295 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10297 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10300 * config/rx/rx.md (BW): New mode attribute.
10301 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10303 2018-01-11 Richard Biener <rguenther@suse.de>
10305 PR tree-optimization/83435
10306 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10307 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10308 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10310 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10311 Alan Hayward <alan.hayward@arm.com>
10312 David Sherwood <david.sherwood@arm.com>
10314 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10316 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10317 (aarch64_print_address_internal): Use it to check for a zero offset.
10319 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10320 Alan Hayward <alan.hayward@arm.com>
10321 David Sherwood <david.sherwood@arm.com>
10323 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10324 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10325 Return a poly_int64 rather than a HOST_WIDE_INT.
10326 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10327 rather than a HOST_WIDE_INT.
10328 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10329 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10330 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10331 final_offset from HOST_WIDE_INT to poly_int64.
10332 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10333 to_constant when getting the number of units in an Advanced SIMD
10335 (aarch64_builtin_vectorized_function): Check for a constant number
10337 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10339 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10340 attribute instead of GET_MODE_NUNITS.
10341 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10342 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10343 GET_MODE_SIZE for fixed-size registers.
10344 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10345 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10346 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10347 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10348 (aarch64_print_operand, aarch64_print_address_internal)
10349 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10350 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10351 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10352 Handle polynomial GET_MODE_SIZE.
10353 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10354 wider than SImode without modification.
10355 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10356 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10357 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10358 passing and returning SVE modes.
10359 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10360 rather than GEN_INT.
10361 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10362 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10363 (aarch64_allocate_and_probe_stack_space): Likewise.
10364 (aarch64_layout_frame): Cope with polynomial offsets.
10365 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10366 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10367 polynomial offsets.
10368 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10369 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10370 poly_int64 rather than a HOST_WIDE_INT.
10371 (aarch64_get_separate_components, aarch64_process_components)
10372 (aarch64_expand_prologue, aarch64_expand_epilogue)
10373 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10374 (aarch64_anchor_offset): New function, split out from...
10375 (aarch64_legitimize_address): ...here.
10376 (aarch64_builtin_vectorization_cost): Handle polynomial
10377 TYPE_VECTOR_SUBPARTS.
10378 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10380 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10381 number of elements from the PARALLEL rather than the mode.
10382 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10383 rather than GET_MODE_BITSIZE.
10384 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10385 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10386 (aarch64_expand_vec_perm_const_1): Handle polynomial
10387 d->perm.length () and d->perm elements.
10388 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10389 Apply to_constant to d->perm elements.
10390 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10391 polynomial CONST_VECTOR_NUNITS.
10392 (aarch64_move_pointer): Take amount as a poly_int64 rather
10394 (aarch64_progress_pointer): Avoid temporary variable.
10395 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10396 the mode attribute instead of GET_MODE.
10398 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10399 Alan Hayward <alan.hayward@arm.com>
10400 David Sherwood <david.sherwood@arm.com>
10402 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10403 x exists before using it.
10404 (aarch64_add_constant_internal): Rename to...
10405 (aarch64_add_offset_1): ...this. Replace regnum with separate
10406 src and dest rtxes. Handle the case in which they're different,
10407 including when the offset is zero. Replace scratchreg with an rtx.
10408 Use 2 additions if there is no spare register into which we can
10409 move a 16-bit constant.
10410 (aarch64_add_constant): Delete.
10411 (aarch64_add_offset): Replace reg with separate src and dest
10412 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10413 Use aarch64_add_offset_1.
10414 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10415 an rtx rather than an int. Take the delta as a poly_int64
10416 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10417 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10418 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10419 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10420 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10421 and aarch64_add_sp.
10422 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10423 aarch64_add_constant.
10425 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10427 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10428 Use scalar_float_mode.
10430 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10432 * config/aarch64/aarch64-simd.md
10433 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10434 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10435 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10436 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10437 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10438 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10439 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10440 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10441 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10442 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10444 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10447 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10448 targ_options->x_arm_arch_string is non NULL.
10450 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10452 * config/aarch64/aarch64.h
10453 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10455 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10458 * expmed.c (emit_store_flag_force): Swap if const op0
10459 and change VOIDmode to mode of op0.
10461 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10463 PR rtl-optimization/83761
10464 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10465 than bytes to mode_for_size.
10467 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10469 PR middle-end/83189
10470 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10471 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10474 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10476 PR middle-end/83575
10477 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10478 when in layout mode.
10479 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10480 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10483 2018-01-10 Michael Collison <michael.collison@arm.com>
10485 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10486 * config/aarch64/aarch64-option-extension.def: Add
10487 AARCH64_OPT_EXTENSION of 'fp16fml'.
10488 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10489 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10490 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10491 * config/aarch64/constraints.md (Ui7): New constraint.
10492 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10493 (VFMLA_SEL_W): Ditto.
10496 (VFMLA16_LOW): New int iterator.
10497 (VFMLA16_HIGH): Ditto.
10498 (UNSPEC_FMLAL): New unspec.
10499 (UNSPEC_FMLSL): Ditto.
10500 (UNSPEC_FMLAL2): Ditto.
10501 (UNSPEC_FMLSL2): Ditto.
10502 (f16mac): New code attribute.
10503 * config/aarch64/aarch64-simd-builtins.def
10504 (aarch64_fmlal_lowv2sf): Ditto.
10505 (aarch64_fmlsl_lowv2sf): Ditto.
10506 (aarch64_fmlalq_lowv4sf): Ditto.
10507 (aarch64_fmlslq_lowv4sf): Ditto.
10508 (aarch64_fmlal_highv2sf): Ditto.
10509 (aarch64_fmlsl_highv2sf): Ditto.
10510 (aarch64_fmlalq_highv4sf): Ditto.
10511 (aarch64_fmlslq_highv4sf): Ditto.
10512 (aarch64_fmlal_lane_lowv2sf): Ditto.
10513 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10514 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10515 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10516 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10517 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10518 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10519 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10520 (aarch64_fmlal_lane_highv2sf): Ditto.
10521 (aarch64_fmlsl_lane_highv2sf): Ditto.
10522 (aarch64_fmlal_laneq_highv2sf): Ditto.
10523 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10524 (aarch64_fmlalq_lane_highv4sf): Ditto.
10525 (aarch64_fmlsl_lane_highv4sf): Ditto.
10526 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10527 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10528 * config/aarch64/aarch64-simd.md:
10529 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10530 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10531 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10532 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10533 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10534 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10535 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10536 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10537 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10538 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10539 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10540 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10541 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10542 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10543 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10544 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10545 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10546 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10547 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10548 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10549 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10550 (vfmlsl_low_u32): Ditto.
10551 (vfmlalq_low_u32): Ditto.
10552 (vfmlslq_low_u32): Ditto.
10553 (vfmlal_high_u32): Ditto.
10554 (vfmlsl_high_u32): Ditto.
10555 (vfmlalq_high_u32): Ditto.
10556 (vfmlslq_high_u32): Ditto.
10557 (vfmlal_lane_low_u32): Ditto.
10558 (vfmlsl_lane_low_u32): Ditto.
10559 (vfmlal_laneq_low_u32): Ditto.
10560 (vfmlsl_laneq_low_u32): Ditto.
10561 (vfmlalq_lane_low_u32): Ditto.
10562 (vfmlslq_lane_low_u32): Ditto.
10563 (vfmlalq_laneq_low_u32): Ditto.
10564 (vfmlslq_laneq_low_u32): Ditto.
10565 (vfmlal_lane_high_u32): Ditto.
10566 (vfmlsl_lane_high_u32): Ditto.
10567 (vfmlal_laneq_high_u32): Ditto.
10568 (vfmlsl_laneq_high_u32): Ditto.
10569 (vfmlalq_lane_high_u32): Ditto.
10570 (vfmlslq_lane_high_u32): Ditto.
10571 (vfmlalq_laneq_high_u32): Ditto.
10572 (vfmlslq_laneq_high_u32): Ditto.
10573 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10574 (AARCH64_FL_FOR_ARCH8_4): New.
10575 (AARCH64_ISA_F16FML): New ISA flag.
10576 (TARGET_F16FML): New feature flag for fp16fml.
10577 (doc/invoke.texi): Document new fp16fml option.
10579 2018-01-10 Michael Collison <michael.collison@arm.com>
10581 * config/aarch64/aarch64-builtins.c:
10582 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10583 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10584 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10585 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10586 (AARCH64_ISA_SHA3): New ISA flag.
10587 (TARGET_SHA3): New feature flag for sha3.
10588 * config/aarch64/iterators.md (sha512_op): New int attribute.
10589 (CRYPTO_SHA512): New int iterator.
10590 (UNSPEC_SHA512H): New unspec.
10591 (UNSPEC_SHA512H2): Ditto.
10592 (UNSPEC_SHA512SU0): Ditto.
10593 (UNSPEC_SHA512SU1): Ditto.
10594 * config/aarch64/aarch64-simd-builtins.def
10595 (aarch64_crypto_sha512hqv2di): New builtin.
10596 (aarch64_crypto_sha512h2qv2di): Ditto.
10597 (aarch64_crypto_sha512su0qv2di): Ditto.
10598 (aarch64_crypto_sha512su1qv2di): Ditto.
10599 (aarch64_eor3qv8hi): Ditto.
10600 (aarch64_rax1qv2di): Ditto.
10601 (aarch64_xarqv2di): Ditto.
10602 (aarch64_bcaxqv8hi): Ditto.
10603 * config/aarch64/aarch64-simd.md:
10604 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10605 (aarch64_crypto_sha512su0qv2di): Ditto.
10606 (aarch64_crypto_sha512su1qv2di): Ditto.
10607 (aarch64_eor3qv8hi): Ditto.
10608 (aarch64_rax1qv2di): Ditto.
10609 (aarch64_xarqv2di): Ditto.
10610 (aarch64_bcaxqv8hi): Ditto.
10611 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10612 (vsha512h2q_u64): Ditto.
10613 (vsha512su0q_u64): Ditto.
10614 (vsha512su1q_u64): Ditto.
10615 (veor3q_u16): Ditto.
10616 (vrax1q_u64): Ditto.
10617 (vxarq_u64): Ditto.
10618 (vbcaxq_u16): Ditto.
10619 * config/arm/types.md (crypto_sha512): New type attribute.
10620 (crypto_sha3): Ditto.
10621 (doc/invoke.texi): Document new sha3 option.
10623 2018-01-10 Michael Collison <michael.collison@arm.com>
10625 * config/aarch64/aarch64-builtins.c:
10626 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10627 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10628 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10629 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10630 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10631 (AARCH64_ISA_SM4): New ISA flag.
10632 (TARGET_SM4): New feature flag for sm4.
10633 * config/aarch64/aarch64-simd-builtins.def
10634 (aarch64_sm3ss1qv4si): Ditto.
10635 (aarch64_sm3tt1aq4si): Ditto.
10636 (aarch64_sm3tt1bq4si): Ditto.
10637 (aarch64_sm3tt2aq4si): Ditto.
10638 (aarch64_sm3tt2bq4si): Ditto.
10639 (aarch64_sm3partw1qv4si): Ditto.
10640 (aarch64_sm3partw2qv4si): Ditto.
10641 (aarch64_sm4eqv4si): Ditto.
10642 (aarch64_sm4ekeyqv4si): Ditto.
10643 * config/aarch64/aarch64-simd.md:
10644 (aarch64_sm3ss1qv4si): Ditto.
10645 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10646 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10647 (aarch64_sm4eqv4si): Ditto.
10648 (aarch64_sm4ekeyqv4si): Ditto.
10649 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10650 (sm3part_op): Ditto.
10651 (CRYPTO_SM3TT): Ditto.
10652 (CRYPTO_SM3PART): Ditto.
10653 (UNSPEC_SM3SS1): New unspec.
10654 (UNSPEC_SM3TT1A): Ditto.
10655 (UNSPEC_SM3TT1B): Ditto.
10656 (UNSPEC_SM3TT2A): Ditto.
10657 (UNSPEC_SM3TT2B): Ditto.
10658 (UNSPEC_SM3PARTW1): Ditto.
10659 (UNSPEC_SM3PARTW2): Ditto.
10660 (UNSPEC_SM4E): Ditto.
10661 (UNSPEC_SM4EKEY): Ditto.
10662 * config/aarch64/constraints.md (Ui2): New constraint.
10663 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10664 * config/arm/types.md (crypto_sm3): New type attribute.
10665 (crypto_sm4): Ditto.
10666 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10667 (vsm3tt1aq_u32): Ditto.
10668 (vsm3tt1bq_u32): Ditto.
10669 (vsm3tt2aq_u32): Ditto.
10670 (vsm3tt2bq_u32): Ditto.
10671 (vsm3partw1q_u32): Ditto.
10672 (vsm3partw2q_u32): Ditto.
10673 (vsm4eq_u32): Ditto.
10674 (vsm4ekeyq_u32): Ditto.
10675 (doc/invoke.texi): Document new sm4 option.
10677 2018-01-10 Michael Collison <michael.collison@arm.com>
10679 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10680 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10681 (AARCH64_FL_FOR_ARCH8_4): New.
10682 (AARCH64_FL_V8_4): New flag.
10683 (doc/invoke.texi): Document new armv8.4-a option.
10685 2018-01-10 Michael Collison <michael.collison@arm.com>
10687 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10688 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10689 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10690 * config/aarch64/aarch64-option-extension.def: Add
10691 AARCH64_OPT_EXTENSION of 'sha2'.
10692 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10693 (crypto): Disable sha2 and aes if crypto disabled.
10694 (crypto): Enable aes and sha2 if enabled.
10695 (simd): Disable sha2 and aes if simd disabled.
10696 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10698 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10699 (TARGET_SHA2): New feature flag for sha2.
10700 (TARGET_AES): New feature flag for aes.
10701 * config/aarch64/aarch64-simd.md:
10702 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10703 conditional on TARGET_AES.
10704 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10705 (aarch64_crypto_sha1hsi): Make pattern conditional
10707 (aarch64_crypto_sha1hv4si): Ditto.
10708 (aarch64_be_crypto_sha1hv4si): Ditto.
10709 (aarch64_crypto_sha1su1v4si): Ditto.
10710 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10711 (aarch64_crypto_sha1su0v4si): Ditto.
10712 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10713 (aarch64_crypto_sha256su0v4si): Ditto.
10714 (aarch64_crypto_sha256su1v4si): Ditto.
10715 (doc/invoke.texi): Document new aes and sha2 options.
10717 2018-01-10 Martin Sebor <msebor@redhat.com>
10719 PR tree-optimization/83781
10720 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10723 2018-01-11 Martin Sebor <msebor@gmail.com>
10724 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10726 PR tree-optimization/83501
10727 PR tree-optimization/81703
10729 * tree-ssa-strlen.c (get_string_cst): Rename...
10730 (get_string_len): ...to this. Handle global constants.
10731 (handle_char_store): Adjust.
10733 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10734 Jim Wilson <jimw@sifive.com>
10736 * config/riscv/riscv-protos.h (riscv_output_return): New.
10737 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10738 (riscv_attribute_table, riscv_output_return),
10739 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10740 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10741 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10743 (riscv_expand_prologue): Add early return for naked function.
10744 (riscv_expand_epilogue): Likewise.
10745 (riscv_function_ok_for_sibcall): Return false for naked function.
10746 (riscv_set_current_function): New.
10747 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10748 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10749 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10750 * doc/extend.texi (RISC-V Function Attributes): New.
10752 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10754 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10755 check for 128-bit long double before checking TCmode.
10756 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10757 128-bit long doubles before checking TFmode or TCmode.
10758 (FLOAT128_IBM_P): Likewise.
10760 2018-01-10 Martin Sebor <msebor@redhat.com>
10762 PR tree-optimization/83671
10763 * builtins.c (c_strlen): Unconditionally return zero for the empty
10765 Use -Warray-bounds for warnings.
10766 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10767 for non-constant array indices with COMPONENT_REF, arrays of
10768 arrays, and pointers to arrays.
10769 (gimple_fold_builtin_strlen): Determine and set length range for
10770 non-constant character arrays.
10772 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10774 PR middle-end/81897
10775 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10778 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10780 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10782 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10785 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10786 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10787 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10788 indexed_or_indirect_operand predicate.
10789 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10790 (*vsx_le_perm_load_v8hi): Likewise.
10791 (*vsx_le_perm_load_v16qi): Likewise.
10792 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10793 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10794 (*vsx_le_perm_store_v8hi): Likewise.
10795 (*vsx_le_perm_store_v16qi): Likewise.
10796 (eight unnamed splitters): Likewise.
10798 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10800 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10801 * config/rs6000/emmintrin.h: Likewise.
10802 * config/rs6000/mmintrin.h: Likewise.
10803 * config/rs6000/xmmintrin.h: Likewise.
10805 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10808 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10810 * tree.c (tree_nop_conversion): Return true for location wrapper
10812 (maybe_wrap_with_location): New function.
10813 (selftest::check_strip_nops): New function.
10814 (selftest::test_location_wrappers): New function.
10815 (selftest::tree_c_tests): Call it.
10816 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10817 (maybe_wrap_with_location): New decl.
10818 (EXPR_LOCATION_WRAPPER_P): New macro.
10819 (location_wrapper_p): New inline function.
10820 (tree_strip_any_location_wrapper): New inline function.
10822 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10825 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10826 stack_realign_offset for the largest alignment of stack slot
10828 (ix86_find_max_used_stack_alignment): New function.
10829 (ix86_finalize_stack_frame_flags): Use it. Set
10830 max_used_stack_alignment if we don't realign stack.
10831 * config/i386/i386.h (machine_function): Add
10832 max_used_stack_alignment.
10834 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10836 * config/arm/arm.opt (-mbranch-cost): New option.
10837 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10840 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10843 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10844 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10846 2018-01-10 Richard Biener <rguenther@suse.de>
10849 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10850 early out so it also covers the case where we have a non-NULL
10853 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10855 PR tree-optimization/83753
10856 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10857 for non-strided grouped accesses if the number of elements is 1.
10859 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10862 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10863 * i386.h (TARGET_USE_GATHER): Define.
10864 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10866 2018-01-10 Martin Liska <mliska@suse.cz>
10869 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10870 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10872 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10873 CLEANUP_NO_PARTITIONING is not set.
10875 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10877 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10878 for vectors, as a partial revert of r254296.
10879 * rtl.h (const_vec_p): Delete.
10880 (const_vec_duplicate_p): Don't test for vector CONSTs.
10881 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10882 * expmed.c (make_tree): Likewise.
10885 * common.md (E, F): Use CONSTANT_P instead of checking for
10887 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10888 checking for CONST_VECTOR.
10890 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10892 PR middle-end/83575
10893 * predict.c (force_edge_cold): Handle in more sane way edges
10894 with no prediction.
10896 2018-01-09 Carl Love <cel@us.ibm.com>
10898 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10900 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10901 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10902 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10903 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10904 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10905 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10906 * config/rs6000/rs6000-protos.h: Add extern defition for
10907 rs6000_generate_float2_double_code.
10908 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10910 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10911 (float2_v2df): Add define_expand.
10913 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10916 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10917 op_mode in the force_to_mode call.
10919 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10921 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10922 instead of checking each element individually.
10923 (aarch64_evpc_uzp): Likewise.
10924 (aarch64_evpc_zip): Likewise.
10925 (aarch64_evpc_ext): Likewise.
10926 (aarch64_evpc_rev): Likewise.
10927 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10928 instead of checking each element individually. Return true without
10930 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10931 whether all selected elements come from the same input, instead of
10932 checking each element individually. Remove calls to gen_rtx_REG,
10933 start_sequence and end_sequence and instead assert that no rtl is
10936 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10938 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10939 order of HIGH and CONST checks.
10941 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10943 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10944 if the destination isn't an SSA_NAME.
10946 2018-01-09 Richard Biener <rguenther@suse.de>
10948 PR tree-optimization/83668
10949 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10951 (canonicalize_loop_form): ... here, renamed from ...
10952 (canonicalize_loop_closed_ssa_form): ... this and amended to
10953 swap successor edges for loop exit blocks to make us use
10954 the RPO order we need for initial schedule generation.
10956 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10958 PR tree-optimization/64811
10959 * match.pd: When optimizing comparisons with Inf, avoid
10960 introducing or losing exceptions from comparisons with NaN.
10962 2018-01-09 Martin Liska <mliska@suse.cz>
10965 * asan.c (shadow_mem_size): Add gcc_assert.
10967 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10969 Don't save registers in main().
10972 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10973 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10974 * config/avr/avr.c (avr_set_current_function): Don't error if
10975 naked, OS_task or OS_main are specified at the same time.
10976 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10978 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10980 * common/config/avr/avr-common.c (avr_option_optimization_table):
10981 Switch on -mmain-is-OS_task for optimizing compilations.
10983 2018-01-09 Richard Biener <rguenther@suse.de>
10985 PR tree-optimization/83572
10986 * graphite.c: Include cfganal.h.
10987 (graphite_transform_loops): Connect infinite loops to exit
10988 and remove fake edges at the end.
10990 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10992 * ipa-inline.c (edge_badness): Revert accidental checkin.
10994 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10997 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10998 symbols; not inline clones.
11000 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11003 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11004 hard registers. Formatting fixes.
11006 PR preprocessor/83722
11007 * gcc.c (try_generate_repro): Pass
11008 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11009 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11012 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11013 Kito Cheng <kito.cheng@gmail.com>
11015 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11016 (riscv_leaf_function_p): Delete.
11017 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11019 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11021 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11023 (do_ifelse): New function.
11024 (do_isel): New function.
11025 (do_sub3): New function.
11026 (do_add3): New function.
11027 (do_load_mask_compare): New function.
11028 (do_overlap_load_compare): New function.
11029 (expand_compare_loop): New function.
11030 (expand_block_compare): Call expand_compare_loop() when appropriate.
11031 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11032 option description.
11033 (-mblock-compare-inline-loop-limit): New option.
11035 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11038 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11039 Reverse order of second and third operands in first alternative.
11040 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11041 of first and second elements in UNSPEC_VPERMR vector.
11042 (altivec_expand_vec_perm_le): Likewise.
11044 2018-01-08 Jeff Law <law@redhat.com>
11046 PR rtl-optimizatin/81308
11047 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11048 (process_switch): If group_case_labels makes a change, then set
11050 (pass_convert_switch::execute): If a switch is converted, then
11051 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11053 PR rtl-optimization/81308
11054 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11057 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11059 PR target/83663 - Revert r255946
11060 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11061 generation for cases where splatting a value is not useful.
11062 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11063 across a vec_duplicate and a paradoxical subreg forming a vector
11064 mode to a vec_concat.
11066 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11068 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11069 -march=armv8.3-a variants.
11070 * config/arm/t-multilib: Likewise.
11071 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11073 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11075 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11077 (cceq_ior_compare_complement): Give it a name so I can use it, and
11078 change boolean_or_operator predicate to boolean_operator so it can
11079 be used to generate a crand.
11080 (eqne): New code iterator.
11081 (bd/bd_neg): New code_attrs.
11082 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11083 a single define_insn.
11084 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11085 decrement (bdnzt/bdnzf/bdzt/bdzf).
11086 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11087 with the new names of the branch decrement patterns, and added the
11088 names of the branch decrement conditional patterns.
11090 2018-01-08 Richard Biener <rguenther@suse.de>
11092 PR tree-optimization/83563
11093 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11096 2018-01-08 Richard Biener <rguenther@suse.de>
11098 PR middle-end/83713
11099 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11101 2018-01-08 Richard Biener <rguenther@suse.de>
11103 PR tree-optimization/83685
11104 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11105 references to abnormals.
11107 2018-01-08 Richard Biener <rguenther@suse.de>
11110 * dwarf2out.c (output_indirect_strings): Handle empty
11111 skeleton_debug_str_hash.
11112 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11114 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11116 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11117 (emit_store_direct): Likewise.
11118 (arc_trampoline_adjust_address): Likewise.
11119 (arc_asm_trampoline_template): New function.
11120 (arc_initialize_trampoline): Use asm_trampoline_template.
11121 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11122 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11123 * config/arc/arc.md (flush_icache): Delete pattern.
11125 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11127 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11128 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11131 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11134 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11135 by not USED_FOR_TARGET.
11136 (make_pass_resolve_sw_modes): Likewise.
11138 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11140 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11143 2018-01-08 Richard Biener <rguenther@suse.de>
11145 PR middle-end/83580
11146 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11148 2018-01-08 Richard Biener <rguenther@suse.de>
11150 PR middle-end/83517
11151 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11153 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11155 PR middle-end/81897
11156 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11157 basic blocks with a small number of successors.
11158 (convert_control_dep_chain_into_preds): Improve handling of
11160 (dump_predicates): Split apart into...
11161 (dump_pred_chain): ...here...
11162 (dump_pred_info): ...and here.
11163 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11164 (can_chain_union_be_invalidated_p): Improve check for invalidation
11166 (uninit_uses_cannot_happen): Avoid unnecessary if
11167 convert_control_dep_chain_into_preds yielded nothing.
11169 2018-01-06 Martin Sebor <msebor@redhat.com>
11171 PR tree-optimization/83640
11172 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11173 subtracting negative offset from size.
11174 (builtin_access::overlap): Adjust offset bounds of the access to fall
11175 within the size of the object if possible.
11177 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11179 PR rtl-optimization/83699
11180 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11181 extract_bit_field_as_subreg to cases in which the extracted
11182 value is also a vector.
11184 * lra-constraints.c (process_alt_operands): Test for the equivalence
11185 substitutions when detecting a possible reload cycle.
11187 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11190 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11191 by default if flag_selective_schedling{,2}. Formatting fixes.
11193 PR rtl-optimization/83682
11194 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11195 if it has non-VECTOR_MODE element mode.
11196 (vec_duplicate_p): Likewise.
11198 PR middle-end/83694
11199 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11200 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11202 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11205 * config/i386/i386-builtin.def
11206 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11207 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11208 Require also OPTION_MASK_ISA_AVX512F in addition to
11209 OPTION_MASK_ISA_GFNI.
11210 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11211 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11212 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11213 to OPTION_MASK_ISA_GFNI.
11214 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11215 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11216 OPTION_MASK_ISA_AVX512BW.
11217 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11218 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11219 addition to OPTION_MASK_ISA_GFNI.
11220 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11221 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11222 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11223 to OPTION_MASK_ISA_GFNI.
11224 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11225 a requirement for all ISAs rather than any of them with a few
11227 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11229 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11230 bitmasks to be enabled with 3 exceptions, instead of requiring any
11231 enabled ISA with lots of exceptions.
11232 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11233 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11234 Change avx512bw in isa attribute to avx512f.
11235 * config/i386/sgxintrin.h: Add license boilerplate.
11236 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11237 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11238 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11239 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11241 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11242 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11243 temporarily sse2 rather than sse if not enabled already.
11246 * config/i386/sse.md (VI248_VLBW): Rename to ...
11247 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11248 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11249 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11250 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11251 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11252 mode iterator instead of VI248_VLBW.
11254 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11256 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11257 (record_modified): Skip clobbers; add debug output.
11258 (param_change_prob): Use sreal frequencies.
11260 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11262 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11263 punt for user-aligned variables.
11265 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11267 * tree-chrec.c (chrec_contains_symbols): Return true for
11270 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11273 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11274 of (x|y) == x for BICS pattern.
11276 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11278 PR tree-optimization/83605
11279 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11280 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11283 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11285 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11286 * config/epiphany/rtems.h: New file.
11288 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11289 Uros Bizjak <ubizjak@gmail.com>
11292 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11293 QIreg_operand instead of register_operand predicate.
11294 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11295 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11296 comments instead of -fmitigate[-_]rop.
11298 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11301 * cgraphunit.c (symbol_table::compile): Switch to text_section
11302 before calling assembly_start debug hook.
11303 * run-rtl-passes.c (run_rtl_passes): Likewise.
11306 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11308 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11309 range_int_cst_p rather than !symbolic_range_p before calling
11310 extract_range_from_multiplicative_op_1.
11312 2018-01-04 Jeff Law <law@redhat.com>
11314 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11315 redundant test in assertion.
11317 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11319 * doc/rtl.texi: Document machine_mode wrapper classes.
11321 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11323 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11324 using tree_to_uhwi.
11326 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11328 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11329 the VEC_PERM_EXPR fold to fail.
11331 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11334 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11335 to switched_sections.
11337 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11340 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11341 test for d.testing.
11343 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11346 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11347 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11349 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11352 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11353 is BLKmode and bitpos not zero or mode change is needed.
11355 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11358 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11361 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11364 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11365 instead of MULT rtx. Update all corresponding splitters.
11366 (*saddl_se): Ditto.
11367 (*ssub<modesuffix>): Ditto.
11368 (*ssubl_se): Ditto.
11369 (*cmp_sadd_di): Update split patterns.
11370 (*cmp_sadd_si): Ditto.
11371 (*cmp_sadd_sidi): Ditto.
11372 (*cmp_ssub_di): Ditto.
11373 (*cmp_ssub_si): Ditto.
11374 (*cmp_ssub_sidi): Ditto.
11375 * config/alpha/predicates.md (const23_operand): New predicate.
11376 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11377 Look for ASHIFT, not MULT inner operand.
11378 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11380 2018-01-04 Martin Liska <mliska@suse.cz>
11382 PR gcov-profile/83669
11383 * gcov.c (output_intermediate_file): Add version to intermediate
11385 * doc/gcov.texi: Document new field 'version' in intermediate
11386 file format. Fix location of '-k' option of gcov command.
11388 2018-01-04 Martin Liska <mliska@suse.cz>
11391 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11393 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11395 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11397 2018-01-03 Martin Sebor <msebor@redhat.com>
11399 PR tree-optimization/83655
11400 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11401 checking calls with invalid arguments.
11403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11405 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11406 (vectorizable_mask_load_store): Delete.
11407 (vectorizable_call): Return false for masked loads and stores.
11408 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11409 instead of gimple_assign_rhs1.
11410 (vectorizable_load): Handle IFN_MASK_LOAD.
11411 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11415 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11417 (vectorizable_mask_load_store): ...here.
11418 (vectorizable_load): ...and here.
11420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11422 * tree-vect-stmts.c (vect_build_all_ones_mask)
11423 (vect_build_zero_merge_argument): New functions, split out from...
11424 (vectorizable_load): ...here.
11426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11428 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11430 (vectorizable_mask_load_store): ...here.
11431 (vectorizable_store): ...and here.
11433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11435 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11437 (vectorizable_mask_load_store): ...here.
11439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11441 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11442 (vect_model_store_cost): Take a vec_load_store_type instead of a
11444 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11445 (vect_model_store_cost): Take a vec_load_store_type instead of a
11447 (vectorizable_mask_load_store): Update accordingly.
11448 (vectorizable_store): Likewise.
11449 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11453 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11454 IFN_MASK_LOAD calls here rather than...
11455 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11458 Alan Hayward <alan.hayward@arm.com>
11459 David Sherwood <david.sherwood@arm.com>
11461 * expmed.c (extract_bit_field_1): For vector extracts,
11462 fall back to extract_bit_field_as_subreg if vec_extract
11465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11466 Alan Hayward <alan.hayward@arm.com>
11467 David Sherwood <david.sherwood@arm.com>
11469 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11470 they are variable or constant sized.
11471 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11472 slots for constant-sized data.
11474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11475 Alan Hayward <alan.hayward@arm.com>
11476 David Sherwood <david.sherwood@arm.com>
11478 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11479 handling COND_EXPRs with boolean comparisons, try to find a better
11480 basis for the mask type than the boolean itself.
11482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11484 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11485 is calculated and how it can be overridden.
11486 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11487 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11489 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11493 Alan Hayward <alan.hayward@arm.com>
11494 David Sherwood <david.sherwood@arm.com>
11496 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11497 Remove the mode argument.
11498 (aarch64_simd_valid_immediate): Remove the mode and inverse
11500 * config/aarch64/iterators.md (bitsize): New iterator.
11501 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11502 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11503 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11504 aarch64_simd_valid_immediate.
11505 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11506 (aarch64_reg_or_bic_imm): Likewise.
11507 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11508 with an insn_type enum and msl with a modifier_type enum.
11509 Replace element_width with a scalar_mode. Change the shift
11510 to unsigned int. Add constructors for scalar_float_mode and
11511 scalar_int_mode elements.
11512 (aarch64_vect_float_const_representable_p): Delete.
11513 (aarch64_can_const_movi_rtx_p)
11514 (aarch64_simd_scalar_immediate_valid_for_move)
11515 (aarch64_simd_make_constant): Update call to
11516 aarch64_simd_valid_immediate.
11517 (aarch64_advsimd_valid_immediate_hs): New function.
11518 (aarch64_advsimd_valid_immediate): Likewise.
11519 (aarch64_simd_valid_immediate): Remove mode and inverse
11520 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11521 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11522 and aarch64_float_const_representable_p on the result.
11523 (aarch64_output_simd_mov_immediate): Remove mode argument.
11524 Update call to aarch64_simd_valid_immediate and use of
11525 simd_immediate_info.
11526 (aarch64_output_scalar_simd_mov_immediate): Update call
11529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11530 Alan Hayward <alan.hayward@arm.com>
11531 David Sherwood <david.sherwood@arm.com>
11533 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11534 (mode_nunits): Likewise CONST_MODE_NUNITS.
11535 * machmode.def (ADJUST_NUNITS): Document.
11536 * genmodes.c (mode_data::need_nunits_adj): New field.
11537 (blank_mode): Update accordingly.
11538 (adj_nunits): New variable.
11539 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11541 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11542 listed in adj_nunits.
11543 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11544 listed in adj_nunits. Don't emit case statements for such modes.
11545 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11546 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11547 nothing if adj_nunits is nonnull.
11548 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11549 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11550 (emit_mode_fbit): Update use of print_maybe_const_decl.
11551 (emit_move_size): Likewise. Treat the array as non-const
11553 (emit_mode_adjustments): Handle adj_nunits.
11555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11557 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11558 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11559 (VECTOR_MODES): Use it.
11560 (make_vector_modes): Take the prefix as an argument.
11562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11563 Alan Hayward <alan.hayward@arm.com>
11564 David Sherwood <david.sherwood@arm.com>
11566 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11567 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11568 for MODE_VECTOR_BOOL.
11569 * machmode.def (VECTOR_BOOL_MODE): Document.
11570 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11571 (make_vector_bool_mode): New function.
11572 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11574 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11575 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11577 * stor-layout.c (int_mode_for_mode): Likewise.
11578 * tree.c (build_vector_type_for_mode): Likewise.
11579 * varasm.c (output_constant_pool_2): Likewise.
11580 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11581 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11582 for MODE_VECTOR_BOOL.
11583 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11584 of mode class checks.
11585 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11586 instead of a list of mode class checks.
11587 (expand_vector_scalar_condition): Likewise.
11588 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11591 Alan Hayward <alan.hayward@arm.com>
11592 David Sherwood <david.sherwood@arm.com>
11594 * machmode.h (mode_size): Change from unsigned short to
11596 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11597 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11598 or if measurement_type is not polynomial.
11599 (fixed_size_mode::includes_p): Check for constant-sized modes.
11600 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11601 return a poly_uint16 rather than an unsigned short.
11602 (emit_mode_size): Change the type of mode_size from unsigned short
11603 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11604 (emit_mode_adjustments): Cope with polynomial vector sizes.
11605 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11607 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11609 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11610 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11611 * caller-save.c (setup_save_areas): Likewise.
11612 (replace_reg_with_saved_mem): Likewise.
11613 * calls.c (emit_library_call_value_1): Likewise.
11614 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11615 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11616 (gen_lowpart_for_combine): Likewise.
11617 * convert.c (convert_to_integer_1): Likewise.
11618 * cse.c (equiv_constant, cse_insn): Likewise.
11619 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11620 (cselib_subst_to_values): Likewise.
11621 * dce.c (word_dce_process_block): Likewise.
11622 * df-problems.c (df_word_lr_mark_ref): Likewise.
11623 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11624 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11625 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11626 (rtl_for_decl_location): Likewise.
11627 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11628 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11629 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11630 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11631 (expand_expr_real_1): Likewise.
11632 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11633 (pad_below): Likewise.
11634 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11635 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11636 * ira.c (get_subreg_tracking_sizes): Likewise.
11637 * ira-build.c (ira_create_allocno_objects): Likewise.
11638 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11639 (ira_sort_regnos_for_alter_reg): Likewise.
11640 * ira-costs.c (record_operand_costs): Likewise.
11641 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11642 (resolve_simple_move): Likewise.
11643 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11644 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11645 (lra_constraints): Likewise.
11646 (CONST_POOL_OK_P): Reject variable-sized modes.
11647 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11648 (add_pseudo_to_slot, lra_spill): Likewise.
11649 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11650 * optabs-query.c (get_best_extraction_insn): Likewise.
11651 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11652 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11653 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11654 * recog.c (offsettable_address_addr_space_p): Likewise.
11655 * regcprop.c (maybe_mode_change): Likewise.
11656 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11657 * regrename.c (build_def_use): Likewise.
11658 * regstat.c (dump_reg_info): Likewise.
11659 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11660 (find_reloads, find_reloads_subreg_address): Likewise.
11661 * reload1.c (eliminate_regs_1): Likewise.
11662 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11663 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11664 (simplify_binary_operation_1, simplify_subreg): Likewise.
11665 * targhooks.c (default_function_arg_padding): Likewise.
11666 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11667 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11668 (verify_gimple_assign_ternary): Likewise.
11669 * tree-inline.c (estimate_move_cost): Likewise.
11670 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11671 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11672 (get_address_cost_ainc): Likewise.
11673 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11674 (vect_supportable_dr_alignment): Likewise.
11675 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11676 (vectorizable_reduction): Likewise.
11677 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11678 (vectorizable_operation, vectorizable_load): Likewise.
11679 * tree.c (build_same_sized_truth_vector_type): Likewise.
11680 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11681 * var-tracking.c (emit_note_insn_var_location): Likewise.
11682 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11683 (ADDR_VEC_ALIGN): Likewise.
11685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11686 Alan Hayward <alan.hayward@arm.com>
11687 David Sherwood <david.sherwood@arm.com>
11689 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11691 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11692 or if measurement_type is polynomial.
11693 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11694 * combine.c (make_extraction): Likewise.
11695 * dse.c (find_shift_sequence): Likewise.
11696 * dwarf2out.c (mem_loc_descriptor): Likewise.
11697 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11698 (extract_bit_field, extract_low_bits): Likewise.
11699 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11700 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11701 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11702 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11703 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11704 * reload.c (find_reloads): Likewise.
11705 * reload1.c (alter_reg): Likewise.
11706 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11707 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11708 * tree-if-conv.c (predicate_mem_writes): Likewise.
11709 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11710 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11711 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11712 * valtrack.c (dead_debug_insert_temp): Likewise.
11713 * varasm.c (mergeable_constant_section): Likewise.
11714 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11717 Alan Hayward <alan.hayward@arm.com>
11718 David Sherwood <david.sherwood@arm.com>
11720 * expr.c (expand_assignment): Cope with polynomial mode sizes
11721 when assigning to a CONCAT.
11723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11724 Alan Hayward <alan.hayward@arm.com>
11725 David Sherwood <david.sherwood@arm.com>
11727 * machmode.h (mode_precision): Change from unsigned short to
11729 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11731 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11732 or if measurement_type is not polynomial.
11733 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11734 in which the mode is already known to be a scalar_int_mode.
11735 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11736 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11738 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11739 for GET_MODE_PRECISION.
11740 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11741 for GET_MODE_PRECISION.
11742 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11744 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11745 (expand_field_assignment, make_extraction): Likewise.
11746 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11747 (get_last_value): Likewise.
11748 * convert.c (convert_to_integer_1): Likewise.
11749 * cse.c (cse_insn): Likewise.
11750 * expr.c (expand_expr_real_1): Likewise.
11751 * lra-constraints.c (simplify_operand_subreg): Likewise.
11752 * optabs-query.c (can_atomic_load_p): Likewise.
11753 * optabs.c (expand_atomic_load): Likewise.
11754 (expand_atomic_store): Likewise.
11755 * ree.c (combine_reaching_defs): Likewise.
11756 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11757 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11758 * tree.h (type_has_mode_precision_p): Likewise.
11759 * ubsan.c (instrument_si_overflow): Likewise.
11761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11762 Alan Hayward <alan.hayward@arm.com>
11763 David Sherwood <david.sherwood@arm.com>
11765 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11766 polynomial numbers of units.
11767 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11768 (valid_vector_subparts_p): New function.
11769 (build_vector_type): Remove temporary shim and take the number
11770 of units as a poly_uint64 rather than an int.
11771 (build_opaque_vector_type): Take the number of units as a
11772 poly_uint64 rather than an int.
11773 * tree.c (build_vector_from_ctor): Handle polynomial
11774 TYPE_VECTOR_SUBPARTS.
11775 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11776 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11777 (build_vector_from_val): If the number of units is variable,
11778 use build_vec_duplicate_cst for constant operands and
11779 VEC_DUPLICATE_EXPR otherwise.
11780 (make_vector_type): Remove temporary is_constant ().
11781 (build_vector_type, build_opaque_vector_type): Take the number of
11782 units as a poly_uint64 rather than an int.
11783 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11785 * cfgexpand.c (expand_debug_expr): Likewise.
11786 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11787 (store_constructor, expand_expr_real_1): Likewise.
11788 (const_scalar_mask_from_tree): Likewise.
11789 * fold-const-call.c (fold_const_reduction): Likewise.
11790 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11791 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11792 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11793 (fold_relational_const): Likewise.
11794 (native_interpret_vector): Likewise. Change the size from an
11795 int to an unsigned int.
11796 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11797 TYPE_VECTOR_SUBPARTS.
11798 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11799 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11800 duplicating a non-constant operand into a variable-length vector.
11801 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11802 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11803 * ipa-icf.c (sem_variable::equals): Likewise.
11804 * match.pd: Likewise.
11805 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11806 * print-tree.c (print_node): Likewise.
11807 * stor-layout.c (layout_type): Likewise.
11808 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11809 * tree-cfg.c (verify_gimple_comparison): Likewise.
11810 (verify_gimple_assign_binary): Likewise.
11811 (verify_gimple_assign_ternary): Likewise.
11812 (verify_gimple_assign_single): Likewise.
11813 * tree-pretty-print.c (dump_generic_node): Likewise.
11814 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11815 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11816 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11817 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11818 (vect_shift_permute_load_chain): Likewise.
11819 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11820 (expand_vector_condition, optimize_vector_constructor): Likewise.
11821 (lower_vec_perm, get_compute_type): Likewise.
11822 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11823 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11824 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11825 (vect_recog_mask_conversion_pattern): Likewise.
11826 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11827 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11828 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11829 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11830 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11831 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11832 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11833 (supportable_widening_operation): Likewise.
11834 (supportable_narrowing_operation): Likewise.
11835 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11837 * varasm.c (output_constant): Likewise.
11839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11840 Alan Hayward <alan.hayward@arm.com>
11841 David Sherwood <david.sherwood@arm.com>
11843 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11844 so that both the length == 3 and length != 3 cases set up their
11845 own permute vectors. Add comments explaining why we know the
11846 number of elements is constant.
11847 (vect_permute_load_chain): Likewise.
11849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11850 Alan Hayward <alan.hayward@arm.com>
11851 David Sherwood <david.sherwood@arm.com>
11853 * machmode.h (mode_nunits): Change from unsigned char to
11855 (ONLY_FIXED_SIZE_MODES): New macro.
11856 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11857 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11858 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11860 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11861 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11862 or if measurement_type is not polynomial.
11863 * genmodes.c (ZERO_COEFFS): New macro.
11864 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11866 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11867 Use ZERO_COEFFS when emitting initializers.
11868 * data-streamer.h (bp_pack_poly_value): New function.
11869 (bp_unpack_poly_value): Likewise.
11870 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11871 for GET_MODE_NUNITS.
11872 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11873 for GET_MODE_NUNITS.
11874 * tree.c (make_vector_type): Remove temporary shim and make
11875 the real function take the number of units as a poly_uint64
11876 rather than an int.
11877 (build_vector_type_for_mode): Handle polynomial nunits.
11878 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11879 * emit-rtl.c (const_vec_series_p_1): Likewise.
11880 (gen_rtx_CONST_VECTOR): Likewise.
11881 * fold-const.c (test_vec_duplicate_folding): Likewise.
11882 * genrecog.c (validate_pattern): Likewise.
11883 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11884 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11885 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11886 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11887 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11888 * rtlanal.c (subreg_get_info): Likewise.
11889 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11890 (vect_grouped_load_supported): Likewise.
11891 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11892 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11893 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11894 (simplify_const_unary_operation, simplify_binary_operation_1)
11895 (simplify_const_binary_operation, simplify_ternary_operation)
11896 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11897 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11898 instead of CONST_VECTOR_NUNITS.
11899 * varasm.c (output_constant_pool_2): Likewise.
11900 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11901 explicit-encoded elements in the XVEC for variable-length vectors.
11903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11905 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11908 Alan Hayward <alan.hayward@arm.com>
11909 David Sherwood <david.sherwood@arm.com>
11911 * coretypes.h (fixed_size_mode): Declare.
11912 (fixed_size_mode_pod): New typedef.
11913 * builtins.h (target_builtins::x_apply_args_mode)
11914 (target_builtins::x_apply_result_mode): Change type to
11915 fixed_size_mode_pod.
11916 * builtins.c (apply_args_size, apply_result_size, result_vector)
11917 (expand_builtin_apply_args_1, expand_builtin_apply)
11918 (expand_builtin_return): Update accordingly.
11920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11922 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11923 * cselib.c (cselib_hash_rtx): Likewise.
11924 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11925 CONST_VECTOR encoding.
11927 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11928 Jeff Law <law@redhat.com>
11931 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11932 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11933 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11934 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11937 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11938 explicitly probe *sp in a noreturn function if there were any callee
11939 register saves or frame pointer is needed.
11941 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11944 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11945 BLKmode for ternary, binary or unary expressions.
11948 * var-tracking.c (delete_vta_debug_insn): New inline function.
11949 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11950 insns from get_insns () to NULL instead of each bb separately.
11951 Use delete_vta_debug_insn. No longer static.
11952 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11953 delete_vta_debug_insns callers.
11954 * rtl.h (delete_vta_debug_insns): Declare.
11955 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11956 instead of variable_tracking_main.
11958 2018-01-03 Martin Sebor <msebor@redhat.com>
11960 PR tree-optimization/83603
11961 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11962 arguments past the endof the argument list in functions declared
11963 without a prototype.
11964 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11965 Avoid checking when arguments are null.
11967 2018-01-03 Martin Sebor <msebor@redhat.com>
11970 * doc/extend.texi (attribute const): Fix a typo.
11971 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11972 issuing -Wsuggest-attribute for void functions.
11974 2018-01-03 Martin Sebor <msebor@redhat.com>
11976 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11977 offset_int::from instead of wide_int::to_shwi.
11978 (maybe_diag_overlap): Remove assertion.
11979 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11980 * gimple-ssa-sprintf.c (format_directive): Same.
11981 (parse_directive): Same.
11982 (sprintf_dom_walker::compute_format_length): Same.
11983 (try_substitute_return_value): Same.
11985 2018-01-03 Jeff Law <law@redhat.com>
11987 PR middle-end/83654
11988 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11989 non-constant residual for zero at runtime and avoid probing in
11990 that case. Reorganize code for trailing problem to mirror handling
11993 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11995 PR tree-optimization/83501
11996 * tree-ssa-strlen.c (get_string_cst): New.
11997 (handle_char_store): Call get_string_cst.
11999 2018-01-03 Martin Liska <mliska@suse.cz>
12001 PR tree-optimization/83593
12002 * tree-ssa-strlen.c: Include tree-cfg.h.
12003 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12004 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12005 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12007 (strlen_dom_walker::before_dom_children): Call
12008 gimple_purge_dead_eh_edges. Dump tranformation with details
12010 (strlen_dom_walker::before_dom_children): Update call by adding
12011 new argument cleanup_eh.
12012 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12014 2018-01-03 Martin Liska <mliska@suse.cz>
12017 * cif-code.def (VARIADIC_THUNK): New enum value.
12018 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12021 2018-01-03 Jan Beulich <jbeulich@suse.com>
12023 * sse.md (mov<mode>_internal): Tighten condition for when to use
12024 vmovdqu<ssescalarsize> for TI and OI modes.
12026 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12028 Update copyright years.
12030 2018-01-03 Martin Liska <mliska@suse.cz>
12033 * ipa-visibility.c (function_and_variable_visibility): Skip
12034 functions with noipa attribure.
12036 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12038 * gcc.c (process_command): Update copyright notice dates.
12039 * gcov-dump.c (print_version): Ditto.
12040 * gcov.c (print_version): Ditto.
12041 * gcov-tool.c (print_version): Ditto.
12042 * gengtype.c (create_file): Ditto.
12043 * doc/cpp.texi: Bump @copying's copyright year.
12044 * doc/cppinternals.texi: Ditto.
12045 * doc/gcc.texi: Ditto.
12046 * doc/gccint.texi: Ditto.
12047 * doc/gcov.texi: Ditto.
12048 * doc/install.texi: Ditto.
12049 * doc/invoke.texi: Ditto.
12051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12053 * vector-builder.h (vector_builder::m_full_nelts): Change from
12054 unsigned int to poly_uint64.
12055 (vector_builder::full_nelts): Update prototype accordingly.
12056 (vector_builder::new_vector): Likewise.
12057 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12058 (vector_builder::operator ==): Likewise.
12059 (vector_builder::finalize): Likewise.
12060 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12061 Take the number of elements as a poly_uint64 rather than an
12063 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12064 from unsigned int to poly_uint64.
12065 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12066 (vec_perm_indices::new_vector): Likewise.
12067 (vec_perm_indices::length): Likewise.
12068 (vec_perm_indices::nelts_per_input): Likewise.
12069 (vec_perm_indices::input_nelts): Likewise.
12070 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12071 number of elements per input as a poly_uint64 rather than an
12072 unsigned int. Use the original encoding for variable-length
12073 vectors, rather than clamping each individual element.
12074 For the second and subsequent elements in each pattern,
12075 clamp the step and base before clamping their sum.
12076 (vec_perm_indices::series_p): Handle polynomial element counts.
12077 (vec_perm_indices::all_in_range_p): Likewise.
12078 (vec_perm_indices_to_tree): Likewise.
12079 (vec_perm_indices_to_rtx): Likewise.
12080 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12081 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12082 (tree_vector_builder::new_binary_operation): Handle polynomial
12083 element counts. Return false if we need to know the number
12084 of elements at compile time.
12085 * fold-const.c (fold_vec_perm): Punt if the number of elements
12086 isn't known at compile time.
12088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12090 * vec-perm-indices.h (vec_perm_builder): Change element type
12091 from HOST_WIDE_INT to poly_int64.
12092 (vec_perm_indices::element_type): Update accordingly.
12093 (vec_perm_indices::clamp): Handle polynomial element_types.
12094 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12095 (vec_perm_indices::all_in_range_p): Likewise.
12096 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12098 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12099 polynomial vec_perm_indices element types.
12100 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12101 * fold-const.c (fold_vec_perm): Likewise.
12102 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12103 * tree-vect-generic.c (lower_vec_perm): Likewise.
12104 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12105 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12106 element type to HOST_WIDE_INT.
12108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12109 Alan Hayward <alan.hayward@arm.com>
12110 David Sherwood <david.sherwood@arm.com>
12112 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12113 rather than an int. Use plus_constant.
12114 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12115 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12118 Alan Hayward <alan.hayward@arm.com>
12119 David Sherwood <david.sherwood@arm.com>
12121 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12122 a HOST_WIDE_INT to a poly_int64.
12124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12125 Alan Hayward <alan.hayward@arm.com>
12126 David Sherwood <david.sherwood@arm.com>
12128 * calls.c (load_register_parameters): Cope with polynomial
12129 mode sizes. Require a constant size for BLKmode parameters
12130 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12131 forces a parameter to be padded at the lsb end in order to
12132 fill a complete number of words, require the parameter size
12133 to be ordered wrt UNITS_PER_WORD.
12135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12136 Alan Hayward <alan.hayward@arm.com>
12137 David Sherwood <david.sherwood@arm.com>
12139 * reload1.c (spill_stack_slot_width): Change element type
12140 from unsigned int to poly_uint64_pod.
12141 (alter_reg): Treat mode sizes as polynomial.
12143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12144 Alan Hayward <alan.hayward@arm.com>
12145 David Sherwood <david.sherwood@arm.com>
12147 * reload.c (complex_word_subreg_p): New function.
12148 (reload_inner_reg_of_subreg, push_reload): Use it.
12150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12151 Alan Hayward <alan.hayward@arm.com>
12152 David Sherwood <david.sherwood@arm.com>
12154 * lra-constraints.c (process_alt_operands): Reject matched
12155 operands whose sizes aren't ordered.
12156 (match_reload): Refer to this check here.
12158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12159 Alan Hayward <alan.hayward@arm.com>
12160 David Sherwood <david.sherwood@arm.com>
12162 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12163 that the mode size is in the set {1, 2, 4, 8, 16}.
12165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12166 Alan Hayward <alan.hayward@arm.com>
12167 David Sherwood <david.sherwood@arm.com>
12169 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12170 Use plus_constant instead of gen_rtx_PLUS.
12172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12173 Alan Hayward <alan.hayward@arm.com>
12174 David Sherwood <david.sherwood@arm.com>
12176 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12177 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12178 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12179 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12180 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12181 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12182 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12183 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12184 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12185 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12187 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12188 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12189 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12190 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12191 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12192 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12193 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12194 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12195 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12196 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12198 * expr.c (emit_move_resolve_push): Treat the input and result
12199 of PUSH_ROUNDING as a poly_int64.
12200 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12201 (emit_push_insn): Likewise.
12202 * lra-eliminations.c (mark_not_eliminable): Likewise.
12203 * recog.c (push_operand): Likewise.
12204 * reload1.c (elimination_effects): Likewise.
12205 * rtlanal.c (nonzero_bits1): Likewise.
12206 * calls.c (store_one_arg): Likewise. Require the padding to be
12207 known at compile time.
12209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12210 Alan Hayward <alan.hayward@arm.com>
12211 David Sherwood <david.sherwood@arm.com>
12213 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12214 Use plus_constant instead of gen_rtx_PLUS.
12216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12217 Alan Hayward <alan.hayward@arm.com>
12218 David Sherwood <david.sherwood@arm.com>
12220 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12221 rather than an int.
12223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12224 Alan Hayward <alan.hayward@arm.com>
12225 David Sherwood <david.sherwood@arm.com>
12227 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12228 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12229 via stack temporaries. Treat the mode size as polynomial too.
12231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12232 Alan Hayward <alan.hayward@arm.com>
12233 David Sherwood <david.sherwood@arm.com>
12235 * expr.c (expand_expr_real_2): When handling conversions involving
12236 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12237 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12238 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12247 Alan Hayward <alan.hayward@arm.com>
12248 David Sherwood <david.sherwood@arm.com>
12250 * combine.c (can_change_dest_mode): Handle polynomial
12251 REGMODE_NATURAL_SIZE.
12252 * expmed.c (store_bit_field_1): Likewise.
12253 * expr.c (store_constructor): Likewise.
12254 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12255 and polynomial REGMODE_NATURAL_SIZE.
12256 (gen_lowpart_common): Likewise.
12257 * reginfo.c (record_subregs_of_mode): Likewise.
12258 * rtlanal.c (read_modify_subreg_p): Likewise.
12260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12261 Alan Hayward <alan.hayward@arm.com>
12262 David Sherwood <david.sherwood@arm.com>
12264 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12265 numbers of elements.
12267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12268 Alan Hayward <alan.hayward@arm.com>
12269 David Sherwood <david.sherwood@arm.com>
12271 * match.pd: Cope with polynomial numbers of vector elements.
12273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12274 Alan Hayward <alan.hayward@arm.com>
12275 David Sherwood <david.sherwood@arm.com>
12277 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12278 in a POINTER_PLUS_EXPR.
12280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12281 Alan Hayward <alan.hayward@arm.com>
12282 David Sherwood <david.sherwood@arm.com>
12284 * omp-simd-clone.c (simd_clone_subparts): New function.
12285 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12286 (ipa_simd_modify_function_body): Likewise.
12288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12289 Alan Hayward <alan.hayward@arm.com>
12290 David Sherwood <david.sherwood@arm.com>
12292 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12293 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12294 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12295 (expand_vector_condition, vector_element): Likewise.
12296 (subparts_gt): New function.
12297 (get_compute_type): Use subparts_gt.
12298 (count_type_subparts): Delete.
12299 (expand_vector_operations_1): Use subparts_gt instead of
12300 count_type_subparts.
12302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12303 Alan Hayward <alan.hayward@arm.com>
12304 David Sherwood <david.sherwood@arm.com>
12306 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12307 (vect_compile_time_alias): ...this new function. Do the calculation
12308 on poly_ints rather than trees.
12309 (vect_prune_runtime_alias_test_list): Update call accordingly.
12311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12312 Alan Hayward <alan.hayward@arm.com>
12313 David Sherwood <david.sherwood@arm.com>
12315 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12317 (vect_schedule_slp_instance): Likewise.
12319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12320 Alan Hayward <alan.hayward@arm.com>
12321 David Sherwood <david.sherwood@arm.com>
12323 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12324 constant and extern definitions for variable-length vectors.
12325 (vect_get_constant_vectors): Note that the number of units
12326 is known to be constant.
12328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12329 Alan Hayward <alan.hayward@arm.com>
12330 David Sherwood <david.sherwood@arm.com>
12332 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12333 of units as polynomial. Choose between WIDE and NARROW based
12336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12337 Alan Hayward <alan.hayward@arm.com>
12338 David Sherwood <david.sherwood@arm.com>
12340 * tree-vect-stmts.c (simd_clone_subparts): New function.
12341 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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-stmts.c (vectorizable_call): Treat the number of
12348 vectors as polynomial. Use build_index_vector for
12349 IFN_GOMP_SIMD_LANE.
12351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12352 Alan Hayward <alan.hayward@arm.com>
12353 David Sherwood <david.sherwood@arm.com>
12355 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12356 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12357 for variable-length vectors.
12358 (vectorizable_mask_load_store): Treat the number of units as
12359 polynomial, asserting that it is constant if the condition has
12360 already been enforced.
12361 (vectorizable_store, vectorizable_load): Likewise.
12363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12364 Alan Hayward <alan.hayward@arm.com>
12365 David Sherwood <david.sherwood@arm.com>
12367 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12368 of units as polynomial. Punt if we can't tell at compile time
12369 which vector contains the final result.
12371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12372 Alan Hayward <alan.hayward@arm.com>
12373 David Sherwood <david.sherwood@arm.com>
12375 * tree-vect-loop.c (vectorizable_induction): Treat the number
12376 of units as polynomial. Punt on SLP inductions. Use an integer
12377 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12378 cast of such a series for variable-length floating-point
12381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12382 Alan Hayward <alan.hayward@arm.com>
12383 David Sherwood <david.sherwood@arm.com>
12385 * tree.h (build_index_vector): Declare.
12386 * tree.c (build_index_vector): New function.
12387 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12388 of units as polynomial, forcibly converting it to a constant if
12389 vectorizable_reduction has already enforced the condition.
12390 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12391 to create a {1,2,3,...} vector.
12392 (vectorizable_reduction): Treat the number of units as polynomial.
12393 Choose vectype_in based on the largest scalar element size rather
12394 than the smallest number of units. Enforce the restrictions
12397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12398 Alan Hayward <alan.hayward@arm.com>
12399 David Sherwood <david.sherwood@arm.com>
12401 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12402 number of units as polynomial.
12404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12405 Alan Hayward <alan.hayward@arm.com>
12406 David Sherwood <david.sherwood@arm.com>
12408 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12409 * target.def (autovectorize_vector_sizes): Return the vector sizes
12410 by pointer, using vector_sizes rather than a bitmask.
12411 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12412 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12413 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12415 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12416 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12417 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12418 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12419 * omp-general.c (omp_max_vf): Likewise.
12420 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12421 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12422 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12423 * tree-vect-slp.c (vect_slp_bb): Likewise.
12424 * doc/tm.texi: Regenerate.
12425 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12427 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12428 the vector size as a poly_uint64 rather than an unsigned int.
12429 (current_vector_size): Change from an unsigned int to a poly_uint64.
12430 (get_vectype_for_scalar_type): Update accordingly.
12431 * tree.h (build_truth_vector_type): Take the size and number of
12432 units as a poly_uint64 rather than an unsigned int.
12433 (build_vector_type): Add a temporary overload that takes
12434 the number of units as a poly_uint64 rather than an unsigned int.
12435 * tree.c (make_vector_type): Likewise.
12436 (build_truth_vector_type): Take the number of units as a poly_uint64
12437 rather than an unsigned int.
12439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12440 Alan Hayward <alan.hayward@arm.com>
12441 David Sherwood <david.sherwood@arm.com>
12443 * target.def (get_mask_mode): Take the number of units and length
12444 as poly_uint64s rather than unsigned ints.
12445 * targhooks.h (default_get_mask_mode): Update accordingly.
12446 * targhooks.c (default_get_mask_mode): Likewise.
12447 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12448 * doc/tm.texi: Regenerate.
12450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12451 Alan Hayward <alan.hayward@arm.com>
12452 David Sherwood <david.sherwood@arm.com>
12454 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12455 * omp-general.c (omp_max_vf): Likewise.
12456 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12457 (expand_omp_simd): Handle polynomial safelen.
12458 * omp-low.c (omplow_simd_context): Add a default constructor.
12459 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12460 (lower_rec_simd_input_clauses): Update accordingly.
12461 (lower_rec_input_clauses): Likewise.
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 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12468 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12469 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12470 (vect_analyze_slp_cost): Likewise.
12471 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12472 (vect_model_load_cost): Likewise.
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 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12479 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12480 from an unsigned int * to a poly_uint64_pod *.
12481 (calculate_unrolling_factor): New function.
12482 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12485 Alan Hayward <alan.hayward@arm.com>
12486 David Sherwood <david.sherwood@arm.com>
12488 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12489 from an unsigned int to a poly_uint64.
12490 (_loop_vec_info::slp_unrolling_factor): Likewise.
12491 (_loop_vec_info::vectorization_factor): Change from an int
12493 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12494 (vect_get_num_vectors): New function.
12495 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12496 (vect_get_num_copies): Use vect_get_num_vectors.
12497 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12498 to an unsigned int *.
12499 (vect_analyze_data_refs): Change min_vf from an int * to a
12501 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12502 than an unsigned HOST_WIDE_INT.
12503 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12504 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12505 to an unsigned int *.
12506 (vect_analyze_data_ref_dependences): Likewise.
12507 (vect_compute_data_ref_alignment): Handle polynomial vf.
12508 (vect_enhance_data_refs_alignment): Likewise.
12509 (vect_prune_runtime_alias_test_list): Likewise.
12510 (vect_shift_permute_load_chain): Likewise.
12511 (vect_supportable_dr_alignment): Likewise.
12512 (dependence_distance_ge_vf): Take the vectorization factor as a
12513 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12514 (vect_analyze_data_refs): Change min_vf from an int * to a
12516 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12517 vfm1 as a poly_uint64 rather than an int. Make the same change
12518 for the returned bound_scalar.
12519 (vect_gen_vector_loop_niters): Handle polynomial vf.
12520 (vect_do_peeling): Likewise. Update call to
12521 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12522 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12524 * tree-vect-loop.c (vect_determine_vectorization_factor)
12525 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12526 (vect_get_known_peeling_cost): Likewise.
12527 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12528 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12529 (vect_transform_loop): Likewise. Use the lowest possible VF when
12530 updating the upper bounds of the loop.
12531 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12532 rather than an int.
12533 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12534 polynomial unroll factors.
12535 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12536 (vect_make_slp_decision): Likewise.
12537 (vect_supported_load_permutation_p): Likewise, and polynomial
12539 (vect_analyze_slp_cost): Handle polynomial vf.
12540 (vect_slp_analyze_node_operations): Likewise.
12541 (vect_slp_analyze_bb_1): Likewise.
12542 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12543 than an unsigned HOST_WIDE_INT.
12544 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12545 (vectorizable_load): Handle polynomial vf.
12546 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12548 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12551 Alan Hayward <alan.hayward@arm.com>
12552 David Sherwood <david.sherwood@arm.com>
12554 * match.pd: Handle bit operations involving three constants
12555 and try to fold one pair.
12557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12559 * tree-vect-loop-manip.c: Include gimple-fold.h.
12560 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12561 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12562 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12563 Add a path that uses a step of VF instead of 1, but disable it
12565 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12566 and niters_no_overflow parameters. Update calls to
12567 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12568 Create a new SSA name if the latter choses to use a ste other
12569 than zero, and return it via niters_vector_mult_vf_var.
12570 * tree-vect-loop.c (vect_transform_loop): Update calls to
12571 vect_do_peeling, vect_gen_vector_loop_niters and
12572 slpeel_make_loop_iterate_ntimes.
12573 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12574 (vect_gen_vector_loop_niters): Update declarations after above changes.
12576 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12578 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12579 128-bit round to integer instructions.
12580 (ceil<mode>2): Likewise.
12581 (btrunc<mode>2): Likewise.
12582 (round<mode>2): Likewise.
12584 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12586 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12587 unaligned VSX load/store on P8/P9.
12588 (expand_block_clear): Allow the use of unaligned VSX
12589 load/store on P8/P9.
12591 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12593 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12595 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12596 swap associated with both a load and a store.
12598 2018-01-02 Andrew Waterman <andrew@sifive.com>
12600 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12601 * config/riscv/riscv.md (clear_cache): Use it.
12603 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12605 * web.c: Remove out-of-date comment.
12607 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12609 * expr.c (fixup_args_size_notes): Check that any existing
12610 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12611 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12612 (emit_single_push_insn): ...here.
12614 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12616 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12617 (const_vector_encoded_nelts): New function.
12618 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12619 (const_vector_int_elt, const_vector_elt): Declare.
12620 * emit-rtl.c (const_vector_int_elt_1): New function.
12621 (const_vector_elt): Likewise.
12622 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12623 of CONST_VECTOR_ELT.
12625 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12627 * expr.c: Include rtx-vector-builder.h.
12628 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12629 directly on the tree encoding.
12630 (const_vector_from_tree): Likewise.
12631 * optabs.c: Include rtx-vector-builder.h.
12632 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12633 sequence of "u" values.
12634 * vec-perm-indices.c: Include rtx-vector-builder.h.
12635 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12636 directly on the vec_perm_indices encoding.
12638 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12640 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12641 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12642 * rtx-vector-builder.h: New file.
12643 * rtx-vector-builder.c: Likewise.
12644 * rtl.h (rtx_def::u2): Add a const_vector field.
12645 (CONST_VECTOR_NPATTERNS): New macro.
12646 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12647 (CONST_VECTOR_DUPLICATE_P): Likewise.
12648 (CONST_VECTOR_STEPPED_P): Likewise.
12649 (CONST_VECTOR_ENCODED_ELT): Likewise.
12650 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12651 (unwrap_const_vec_duplicate): Likewise.
12652 (const_vec_series_p): Check for a non-duplicated vector encoding.
12653 Say that the function only returns true for integer vectors.
12654 * emit-rtl.c: Include rtx-vector-builder.h.
12655 (gen_const_vec_duplicate_1): Delete.
12656 (gen_const_vector): Call gen_const_vec_duplicate instead of
12657 gen_const_vec_duplicate_1.
12658 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12659 (gen_const_vec_duplicate): Use rtx_vector_builder.
12660 (gen_const_vec_series): Likewise.
12661 (gen_rtx_CONST_VECTOR): Likewise.
12662 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12663 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12664 Build a new vector rather than modifying a CONST_VECTOR in-place.
12665 (handle_special_swappables): Update call accordingly.
12666 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12667 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12668 Build a new vector rather than modifying a CONST_VECTOR in-place.
12669 (handle_special_swappables): Update call accordingly.
12671 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12673 * simplify-rtx.c (simplify_const_binary_operation): Use
12674 CONST_VECTOR_ELT instead of XVECEXP.
12676 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12678 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12679 the selector elements to be different from the data elements
12680 if the selector is a VECTOR_CST.
12681 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12682 ssizetype for the selector.
12684 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12686 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12687 before testing each element individually.
12688 * tree-vect-generic.c (lower_vec_perm): Likewise.
12690 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12692 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12693 * selftest-run-tests.c (selftest::run_tests): Call it.
12694 * vector-builder.h (vector_builder::operator ==): New function.
12695 (vector_builder::operator !=): Likewise.
12696 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12697 (vec_perm_indices::all_from_input_p): New function.
12698 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12699 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12700 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12701 instead of reading the VECTOR_CST directly. Detect whether both
12702 vector inputs are the same before constructing the vec_perm_indices,
12703 and update the number of inputs argument accordingly. Use the
12704 utility functions added above. Only construct sel2 if we need to.
12706 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12708 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12709 the broadcast of the low byte.
12710 (expand_mult_highpart): Use an explicit encoding for the permutes.
12711 * optabs-query.c (can_mult_highpart_p): Likewise.
12712 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12713 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12714 (vectorizable_bswap): Likewise.
12715 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12716 explicit encoding for the power-of-2 permutes.
12717 (vect_permute_store_chain): Likewise.
12718 (vect_grouped_load_supported): Likewise.
12719 (vect_permute_load_chain): Likewise.
12721 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12723 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12724 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12725 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12726 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12727 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12728 (vect_gen_perm_mask_any): Likewise.
12730 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12732 * int-vector-builder.h: New file.
12733 * vec-perm-indices.h: Include int-vector-builder.h.
12734 (vec_perm_indices): Redefine as an int_vector_builder.
12735 (auto_vec_perm_indices): Delete.
12736 (vec_perm_builder): Redefine as a stand-alone class.
12737 (vec_perm_indices::vec_perm_indices): New function.
12738 (vec_perm_indices::clamp): Likewise.
12739 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12740 (vec_perm_indices::new_vector): New function.
12741 (vec_perm_indices::new_expanded_vector): Update for new
12742 vec_perm_indices class.
12743 (vec_perm_indices::rotate_inputs): New function.
12744 (vec_perm_indices::all_in_range_p): Operate directly on the
12745 encoded form, without computing elided elements.
12746 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12747 encoding. Update for new vec_perm_indices class.
12748 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12749 the given vec_perm_builder.
12750 (expand_vec_perm_var): Update vec_perm_builder constructor.
12751 (expand_mult_highpart): Use vec_perm_builder instead of
12752 auto_vec_perm_indices.
12753 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12754 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12755 or double series encoding as appropriate.
12756 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12757 vec_perm_indices instead of auto_vec_perm_indices.
12758 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12759 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12760 (vect_permute_store_chain): Likewise.
12761 (vect_grouped_load_supported): Likewise.
12762 (vect_permute_load_chain): Likewise.
12763 (vect_shift_permute_load_chain): Likewise.
12764 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12765 (vect_transform_slp_perm_load): Likewise.
12766 (vect_schedule_slp_instance): Likewise.
12767 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12768 (vectorizable_mask_load_store): Likewise.
12769 (vectorizable_bswap): Likewise.
12770 (vectorizable_store): Likewise.
12771 (vectorizable_load): Likewise.
12772 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12773 vec_perm_indices instead of auto_vec_perm_indices. Use
12774 tree_to_vec_perm_builder to read the vector from a tree.
12775 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12776 vec_perm_builder instead of a vec_perm_indices.
12777 (have_whole_vector_shift): Use vec_perm_builder and
12778 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12779 truncation to calc_vec_perm_mask_for_shift.
12780 (vect_create_epilog_for_reduction): Likewise.
12781 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12782 from auto_vec_perm_indices to vec_perm_indices.
12783 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12784 instead of changing individual elements.
12785 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12786 the vector in d.perm.
12787 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12788 from auto_vec_perm_indices to vec_perm_indices.
12789 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12790 instead of changing individual elements.
12791 (arm_vectorize_vec_perm_const): Use new_vector to install
12792 the vector in d.perm.
12793 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12794 Update vec_perm_builder constructor.
12795 (rs6000_expand_interleave): Likewise.
12796 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12797 (rs6000_expand_interleave): Likewise.
12799 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12801 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12802 to qimode could truncate the indices.
12803 * optabs.c (expand_vec_perm_var): Likewise.
12805 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12807 * Makefile.in (OBJS): Add vec-perm-indices.o.
12808 * vec-perm-indices.h: New file.
12809 * vec-perm-indices.c: Likewise.
12810 * target.h (vec_perm_indices): Replace with a forward class
12812 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12813 * optabs.h: Include vec-perm-indices.h.
12814 (expand_vec_perm): Delete.
12815 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12816 (expand_vec_perm_const): Declare.
12817 * target.def (vec_perm_const_ok): Replace with...
12818 (vec_perm_const): ...this new hook.
12819 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12820 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12821 * doc/tm.texi: Regenerate.
12822 * optabs.def (vec_perm_const): Delete.
12823 * doc/md.texi (vec_perm_const): Likewise.
12824 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12825 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12826 expand_vec_perm for constant permutation vectors. Assert that
12827 the mode of variable permutation vectors is the integer equivalent
12828 of the mode that is being permuted.
12829 * optabs-query.h (selector_fits_mode_p): Declare.
12830 * optabs-query.c: Include vec-perm-indices.h.
12831 (selector_fits_mode_p): New function.
12832 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12833 is defined, instead of checking whether the vec_perm_const_optab
12834 exists. Use targetm.vectorize.vec_perm_const instead of
12835 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12836 fit in the vector mode before using a variable permute.
12837 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12838 vec_perm_indices instead of an rtx.
12839 (expand_vec_perm): Replace with...
12840 (expand_vec_perm_const): ...this new function. Take the selector
12841 as a vec_perm_indices rather than an rtx. Also take the mode of
12842 the selector. Update call to shift_amt_for_vec_perm_mask.
12843 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12844 Use vec_perm_indices::new_expanded_vector to expand the original
12845 selector into bytes. Check whether the indices fit in the vector
12846 mode before using a variable permute.
12847 (expand_vec_perm_var): Make global.
12848 (expand_mult_highpart): Use expand_vec_perm_const.
12849 * fold-const.c: Includes vec-perm-indices.h.
12850 * tree-ssa-forwprop.c: Likewise.
12851 * tree-vect-data-refs.c: Likewise.
12852 * tree-vect-generic.c: Likewise.
12853 * tree-vect-loop.c: Likewise.
12854 * tree-vect-slp.c: Likewise.
12855 * tree-vect-stmts.c: Likewise.
12856 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12858 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12859 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12860 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12861 (aarch64_vectorize_vec_perm_const): ...this new function.
12862 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12863 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12864 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12865 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12866 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12867 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12868 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12870 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12871 check for NEON modes.
12872 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12873 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12874 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12875 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12877 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12878 the old VEC_PERM_CONST conditions.
12879 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12880 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12881 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12882 (ia64_vectorize_vec_perm_const_ok): Merge into...
12883 (ia64_vectorize_vec_perm_const): ...this new function.
12884 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12885 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12886 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12887 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12888 * config/mips/mips.c (mips_expand_vec_perm_const)
12889 (mips_vectorize_vec_perm_const_ok): Merge into...
12890 (mips_vectorize_vec_perm_const): ...this new function.
12891 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12892 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12893 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12894 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12895 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12896 (rs6000_expand_vec_perm_const): Delete.
12897 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12899 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12900 (altivec_expand_vec_perm_const_le): Take each operand individually.
12901 Operate on constant selectors rather than rtxes.
12902 (altivec_expand_vec_perm_const): Likewise. Update call to
12903 altivec_expand_vec_perm_const_le.
12904 (rs6000_expand_vec_perm_const): Delete.
12905 (rs6000_vectorize_vec_perm_const_ok): Delete.
12906 (rs6000_vectorize_vec_perm_const): New function.
12907 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12908 an element count and rtx array.
12909 (rs6000_expand_extract_even): Update call accordingly.
12910 (rs6000_expand_interleave): Likewise.
12911 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12912 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12913 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12914 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12915 (rs6000_expand_vec_perm_const): Delete.
12916 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12917 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12918 (altivec_expand_vec_perm_const_le): Take each operand individually.
12919 Operate on constant selectors rather than rtxes.
12920 (altivec_expand_vec_perm_const): Likewise. Update call to
12921 altivec_expand_vec_perm_const_le.
12922 (rs6000_expand_vec_perm_const): Delete.
12923 (rs6000_vectorize_vec_perm_const_ok): Delete.
12924 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12925 reference to the SPE evmerge intructions.
12926 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12927 an element count and rtx array.
12928 (rs6000_expand_extract_even): Update call accordingly.
12929 (rs6000_expand_interleave): Likewise.
12930 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12931 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12933 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12935 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12937 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12938 vector mode and that that mode matches the mode of the data
12940 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12941 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12942 directly using expand_vec_perm_1 when forcing selectors into
12944 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12946 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12948 * optabs-query.h (can_vec_perm_p): Delete.
12949 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12950 * optabs-query.c (can_vec_perm_p): Split into...
12951 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12952 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12953 particular selector is valid.
12954 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12955 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12956 (vect_grouped_load_supported): Likewise.
12957 (vect_shift_permute_load_chain): Likewise.
12958 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12959 (vect_transform_slp_perm_load): Likewise.
12960 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12961 (vectorizable_bswap): Likewise.
12962 (vect_gen_perm_mask_checked): Likewise.
12963 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12964 implementations of variable permutation vectors into account
12965 when deciding which selector to use.
12966 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12967 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12968 with a false third argument.
12969 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12970 to test whether the constant selector is valid and can_vec_perm_var_p
12971 to test whether a variable selector is valid.
12973 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12975 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12976 * optabs-query.c (can_vec_perm_p): Likewise.
12977 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12978 instead of vec_perm_indices.
12979 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12980 (vect_gen_perm_mask_checked): Likewise,
12981 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12982 (vect_gen_perm_mask_checked): Likewise,
12984 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12986 * optabs-query.h (qimode_for_vec_perm): Declare.
12987 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12988 (qimode_for_vec_perm): ...this new function.
12989 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12991 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12993 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12994 does not have a conditional at the top.
12996 2018-01-02 Richard Biener <rguenther@suse.de>
12998 * ipa-inline.c (big_speedup_p): Fix expression.
13000 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13003 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13006 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13010 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13011 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13012 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13013 cond_taken_branch_cost 3->4.
13015 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13017 PR tree-optimization/83581
13018 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13019 TODO_cleanup_cfg if any changes have been made.
13021 PR middle-end/83608
13022 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13023 convert_modes if target mode has the right side, but different mode
13026 PR middle-end/83609
13027 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13028 last argument when extracting from CONCAT. If either from_real or
13029 from_imag is NULL, use expansion through memory. If result is not
13030 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13031 the parts directly to inner mode, if even that fails, use expansion
13034 PR middle-end/83623
13035 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13036 check for bswap in mode rather than HImode and use that in expand_unop
13039 Copyright (C) 2018 Free Software Foundation, Inc.
13041 Copying and distribution of this file, with or without modification,
13042 are permitted in any medium without royalty provided the copyright
13043 notice and this notice are preserved.