1 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4 _mm512_setzero): New intrinsics.
6 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
7 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
10 code generation for cases where splatting a value is not useful.
11 * simplify-rtx.c (simplify_ternary_operation): Simplify
12 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
15 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
17 * config.gcc: Support "goldmont-plus".
18 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
20 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21 PROCESSOR_GOLDMONT_PLUS.
22 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
23 (processor_target_table): Add "goldmont-plus".
24 (PTA_GOLDMONT_PLUS): Define.
25 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
26 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
27 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
28 (fold_builtin_cpu): Add "goldmont-plus".
29 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
30 (ix86_option_override_internal): Add "goldmont-plus".
31 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
32 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
33 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
34 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
36 2018-05-17 Richard Biener <rguenther@suse.de>
38 PR tree-optimization/85757
39 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
40 remove defs that only feed that PHI from further processing.
42 2018-05-16 Jim Wilson <jimw@sifive.com>
44 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
46 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
48 2018-05-16 Mark Wielaard <mark@klomp.org>
50 * dwarf2out.c (count_index_strings): New function.
51 (output_indirect_strings): Call count_index_strings and generate
52 header for dwarf_version >= 5.
54 2018-05-16 Mark Wielaard <mark@klomp.org>
56 * dwarf2out.c (dwarf_FORM): New function.
57 (set_indirect_string): Use dwarf_FORM.
58 (reset_indirect_string): Likewise.
59 (size_of_die): Likewise.
60 (value_format): Likewise.
61 (output_die): Likewise.
62 (add_skeleton_AT_string): Likewise.
63 (output_macinfo_op): Likewise.
64 (index_string): Likewise.
65 (output_index_string_offset): Likewise.
66 (output_index_string): Likewise.
67 (count_index_strings): Likewise.
69 2018-05-16 Carl Love <cel@us.ibm.com>
71 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
72 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
74 2018-05-16 Martin Jambor <mjambor@suse.cz>
76 * ipa-prop.c (ipa_free_all_edge_args): Remove.
77 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
79 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
81 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
82 (fnma<mode>4): Likewise.
83 (fms<mode>4): Likewise.
84 (fnms<mode>4): Likewise.
85 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
86 (aarch64_fnma<mode>4): Likewise.
87 (aarch64_fms<mode>4): Likewise.
88 (aarch64_fnms<mode>4): Likewise.
89 (aarch64_fnmadd<mode>4): Likewise.
91 2018-05-16 Jason Merrill <jason@redhat.com>
93 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
95 2018-05-16 Richard Biener <rguenther@suse.de>
97 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
98 (dump_stmt_cost): Declare.
99 (add_stmt_cost): Dump cost we add.
100 (add_stmt_costs): New function.
101 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
103 (vect_analyze_stmt): Adjust prototype.
104 (vectorizable_condition): Likewise.
105 (vectorizable_live_operation): Likewise.
106 (vectorizable_reduction): Likewise.
107 (vectorizable_induction): Likewise.
108 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
109 cost vector to pass to vectorizable_ and record afterwards.
110 (vect_model_reduction_cost): Take cost vector argument and adjust.
111 (vect_model_induction_cost): Likewise.
112 (vectorizable_reduction): Likewise.
113 (vectorizable_induction): Likewise.
114 (vectorizable_live_operation): Likewise.
115 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
116 SLP_TREE_NUMBER_OF_VEC_STMTS.
117 (vect_analyze_slp_cost_1): Remove.
118 (vect_analyze_slp_cost): Likewise.
119 (vect_slp_analyze_node_operations): Take visited args and
120 a target cost vector. Avoid processing already visited stmt sets.
121 (vect_slp_analyze_operations): Use a local cost vector to gather
122 costs and register those of non-discarded instances.
123 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
124 (vect_schedule_slp_instance): Remove copying of
125 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
127 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
128 adding cost. Record cost entry location.
129 (vect_prologue_cost_for_slp_op): Function to compute cost of
130 a constant or invariant generated for SLP vect in the prologue,
131 split out from vect_analyze_slp_cost_1.
132 (vect_model_simple_cost): Make static. Adjust for SLP costing.
133 (vect_model_promotion_demotion_cost): Likewise.
134 (vect_model_store_cost): Likewise, make static.
135 (vect_model_load_cost): Likewise.
136 (vectorizable_bswap): Add cost vector arg and adjust.
137 (vectorizable_call): Likewise.
138 (vectorizable_simd_clone_call): Likewise.
139 (vectorizable_conversion): Likewise.
140 (vectorizable_assignment): Likewise.
141 (vectorizable_shift): Likewise.
142 (vectorizable_operation): Likewise.
143 (vectorizable_store): Likewise.
144 (vectorizable_load): Likewise.
145 (vectorizable_condition): Likewise.
146 (vectorizable_comparison): Likewise.
147 (can_vectorize_live_stmts): Likewise.
148 (vect_analyze_stmt): Likewise.
149 (vect_transform_stmt): Adjust calls to vectorizable_*.
150 * tree-vectorizer.c: Include gimple-pretty-print.h.
151 (dump_stmt_cost): New function.
153 2018-05-16 Richard Biener <rguenther@suse.de>
155 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
156 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
157 * tree-ssa-dse.c: Include tree-ssa-loop.h.
158 (check_name): New callback.
159 (dse_classify_store): Track cycles via a visited bitmap of PHI
160 defs and simplify handling of in-loop and across loop dead stores
161 and properly fail for loop-variant refs. Handle byte-tracking with
162 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
165 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
167 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
168 (vect_get_mask_type_for_stmt): Likewise.
169 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
171 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
172 to determine the statement's vector type and the vector type that
173 should be used for calculating nunits. Deal with cases in which
174 the type has to be deferred.
175 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
176 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
177 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
178 (vect_determine_vf_for_stmt): New functions, split out from...
179 (vect_determine_vectorization_factor): ...here.
180 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
181 (vect_get_mask_type_for_stmt): New functions, split out from
182 vect_determine_vectorization_factor.
184 2018-05-16 Richard Biener <rguenther@suse.de>
186 * tree-cfg.c (verify_gimple_assign_ternary): Properly
187 verify the [VEC_]COND_EXPR embedded comparison.
189 2018-05-15 Martin Sebor <msebor@redhat.com>
191 PR tree-optimization/85753
192 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
193 RECORD_TYPE in addition to ARRAY_TYPE.
195 2018-05-15 Martin Sebor <msebor@redhat.com>
198 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
200 2018-05-15 Richard Biener <rguenther@suse.de>
202 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
203 add by_clobber_p one. Change algorithm to collect all defs
204 representing uses we need to walk and try reducing them to
205 a single one before failing.
206 (dse_dom_walker::dse_optimize_stmt): Adjust.
208 2018-05-13 Mark Wielaard <mark@klomp.org>
210 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
211 (size_of_loc_descr): Likewise.
212 (output_loc_operands): Likewise.
213 (output_loc_operands_raw): Likewise.
214 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
215 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
216 (hash_loc_operands): Likewise.
217 (compare_loc_operands): Likewise.
219 2018-05-14 Mark Wielaard <mark@klomp.org>
221 * dwarf2out.c (count_index_addrs): New function.
222 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
224 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
226 PR tree-optimization/83648
227 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
228 return value as malloc candidate.
230 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
233 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
234 as true in call to suggest_attribute.
236 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
238 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
241 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
243 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
245 (aarch64_crypto_aesd_fused): Likewise.
247 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
249 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
250 (movsi_aarch64): Likewise.
251 (load_pairsi): Likewise.
252 (load_pairdi): Likewise.
253 (store_pairsi): Likewise.
254 (store_pairdi): Likewise.
255 (load_pairsf): Likewise.
256 (load_pairdf): Likewise.
257 (store_pairsf): Likewise.
258 (store_pairdf): Likewise.
259 (zero_extend): Likewise.
260 (trunc): Swap alternatives.
261 (fcvt_target): Add '?' to prefer w over r.
263 2018-05-14 Jakub Jelinek <jakub@redhat.com>
266 * config/i386/i386.md: Disallow non-commutative arithmetics in
267 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
268 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
269 in the peephole2 before it.
271 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
273 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
274 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
275 (ix86_handle_option): Handle -mcldemote.
276 * config.gcc: New header.
277 * config/i386/cldemoteintrin.h: New file.
278 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
279 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
281 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
282 OPTION_MASK_ISA_CLDEMOTE.
283 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
284 (ix86_valid_target_attribute_inner_p): Ditto.
285 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
286 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
287 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
288 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
289 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
291 * config/i386/i386.opt: Add -mcldemote.
292 * config/i386/x86intrin.h: New header.
293 * doc/invoke.texi: Add -mcldemote.
295 2018-05-14 Richard Biener <rguenther@suse.de>
297 * doc/match-and-simplify.texi: Adjust :s documentation.
299 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
301 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
302 intended memcpy size.
303 (REORDER_45): Likewise.
305 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
308 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
309 * vec.c (qsort_chk): Use gcc_qsort.
310 * Makefile.in (OBJS-libcommon): Add sort.o.
311 (build/sort.o): New target. Use it...
312 (BUILD_RTL): ... here, and...
313 (build/gencfn-macros): ... here, and...
314 (build/genmatch): ... here.
316 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
317 Chung-Ju Wu <jasonwucj@gmail.com>
319 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
320 * config/nds32/nds32-graywolf.md: New file.
321 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
322 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
324 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
325 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
326 * config/nds32/nds32.md (pipeline_model): Add graywolf.
327 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
328 * config/nds32/pipelines.md: Include n15 settings.
330 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
331 Chung-Ju Wu <jasonwucj@gmail.com>
333 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
334 * config/nds32/nds32-n13.md: New file.
335 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
336 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
338 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
339 * config/nds32/nds32.md (pipeline_model): Add n13.
340 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
341 * config/nds32/pipelines.md: Include n13 settings.
343 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
344 Chung-Ju Wu <jasonwucj@gmail.com>
346 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
347 * config/nds32/nds32-n10.md: New file.
348 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
349 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
351 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
352 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
353 * config/nds32/nds32.md (pipeline_model): Add n10.
354 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
355 * config/nds32/pipelines.md: Include n10 settings.
357 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
358 Kito Cheng <kito.cheng@gmail.com>
359 Chung-Ju Wu <jasonwucj@gmail.com>
361 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
362 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
363 Add enum values for DSP extension instructions.
364 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
366 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
367 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
369 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
370 * config/nds32/nds32-dspext.md: New file for DSP implementation.
371 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
372 * config/nds32/nds32-intrinsic.md: Likewise.
373 * config/nds32/nds32_intrinsic.h: Likewise.
374 * config/nds32/nds32-md-auxiliary.c: Likewise.
375 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
376 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
377 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
378 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
379 * config/nds32/nds32-protos.h: New declarations for DSP extension.
380 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
381 TYPE_DMAC in switch statement.
382 * config/nds32/nds32.c: New checking and implementation for DSP
383 extension instructions.
384 * config/nds32/nds32.h: Likewise.
385 * config/nds32/nds32.md: Likewise.
386 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
387 * config/nds32/predicates.md: Implement new predicates for DSP
390 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
392 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
393 Reformat alternatives and attributes so it is easier to identify
394 which constraints/attributes go with which instruction.
395 (mov<mode>_hardfloat32, FMOVE64): Likewise.
396 (mov<mode>_softfloat32, FMOVE64): Likewise.
397 (mov<mode>_hardfloat64, FMOVE64): Likewise.
398 (mov<mode>_softfloat64, FMOVE64): Likewise.
400 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
402 * doc/extend.texi (PowerPC Built-in Functions): Rename this
404 (Basic PowerPC Built-in Functions): The new name of the
405 subsection previously known as "PowerPC Built-in Functions".
406 (Basic PowerPC Built-in Functions Available on all Configurations):
408 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
409 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
410 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
411 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
413 2018-05-11 Martin Jambor <mjambor@suse.cz>
416 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
419 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
422 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
424 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
426 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
427 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
428 (ix86_handle_option): Handle -mwaitpkg.
429 * config.gcc: New header.
430 * config/i386/cpuid.h (bit_WAITPKG): New bit.
431 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
432 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
434 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
435 OPTION_MASK_ISA_WAITPKG.
436 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
437 (ix86_option_override_internal): Add PTA_WAITPKG.
438 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
439 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
441 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
442 __builtin_ia32_umwait and __builtin_ia32_tpause.
443 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
444 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
445 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
446 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
447 UNSPECV_TPAUSE): New.
448 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
449 * config/i386/i386.opt: Add -mwaitpkg.
450 * config/i386/waitpkgintrin.h: New file.
451 * config/i386/x86intrin.h: New header.
452 * doc/invoke.texi: Add -mwaitpkg.
454 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
457 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
459 (cortex-m0): Use armv6s-m isa.
460 (cortex-m0plus): Likewise.
461 (cortex-m1): Likewise.
462 (cortex-m0.small-multiply): Likewise.
463 (cortex-m0plus.small-multiply): Likewise.
464 (cortex-m1.small-multiply): Likewise.
466 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
467 Jakub Jelinek <jakub@redhat.com>
469 PR tree-optimization/85692
470 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
471 source permute as well.
473 2018-05-11 Martin Liska <mliska@suse.cz>
476 * doc/extend.texi: Document LLVM style format for no_sanitize
479 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
481 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
482 mode_supports_vsx_dform_quad to mode_supports_dq_form.
483 (mode_supports_vsx_dform_quad): Likewise.
484 (mode_supports_vmx_dform): Move these functions to be next to the
485 other mode_supports functions.
486 (mode_supports_dq_form): Likewise.
487 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
488 mode_supports_dq_form.
489 (reg_offset_addressing_ok_p): Likewise.
490 (offsettable_ok_by_alignment): Likewise.
491 (rs6000_legitimate_offset_address_p): Likewise.
492 (legitimate_lo_sum_address_p): Likewise.
493 (rs6000_legitimize_address): Likewise.
494 (rs6000_legitimize_reload_address): Likewise.
495 (rs6000_secondary_reload_inner): Likewise.
496 (rs6000_preferred_reload_class): Likewise.
497 (rs6000_output_move_128bit): Likewise.
499 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
501 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
502 Generate SImode target register for null target.
503 <case IX86_BUILTIN_XGETBV>: Ditto.
504 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
505 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
507 2018-05-10 Carl Love <cel@us.ibm.com>
509 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
510 dcbtt and dcbtstt if operands[2] is 0.
512 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
515 * config/i386/sse.md (usadv64qi): New expander.
517 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
519 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
520 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
521 -maltivec=be support.
522 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
523 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
524 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
525 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
526 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
527 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
528 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
529 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
530 altivec_vsumsws): Adjust.
531 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
532 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
533 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
535 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
536 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
537 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
538 (altivec_lve<VI_char>x): Delete expand.
539 (*altivec_lve<VI_char>x_internal): Rename to...
540 (altivec_lve<VI_char>x): ... this.
541 (altivec_lvxl_<mode>): Delete expand.
542 (*altivec_lvxl_<mode>_internal): Rename to ...
543 (altivec_lvxl_<mode>): ... this.
544 (altivec_stvxl_<mode>): Delete expand.
545 (*altivec_stvxl_<mode>_internal): Rename to ...
546 (altivec_stvxl_<mode>): ... this.
547 (altivec_stve<VI_char>x): Delete expand.
548 (*altivec_stve<VI_char>x_internal): Rename to ...
549 (altivec_stve<VI_char>x): ... this.
550 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
551 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
552 reduc_plus_scal_<mode>): Adjust.
553 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
555 (rs6000_cpu_cpp_builtins): Adjust.
556 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
557 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
558 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
559 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
560 -maltivec=be support.
561 (rs6000_split_vec_extract_var): Adjust.
562 (rs6000_split_v4si_init): Adjust.
563 (swap_selector_for_mode): Delete.
564 (altivec_expand_lvx_be, altivec_expand_stvx_be,
565 altivec_expand_stvex_be): Delete.
566 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
567 -maltivec=be support.
568 (rs6000_gimple_fold_builtin): Ditto.
569 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
571 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
572 (TARGET_DIRECT_MOVE_64BIT): Adjust.
573 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
574 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
575 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
576 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
577 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
578 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
579 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
580 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
581 anonymous split): Adjust.
582 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
583 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
585 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
587 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
588 when --with-gxx-include-dir is also specified.
589 * configure: Regenerate.
591 2018-05-09 Jim Wilson <jimw@sifive.com>
594 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
595 * config/riscv/t-withmultilib: New.
596 * config/riscv/withmultilib.h: New.
597 * doc/install.texi: Document RISC-V --with-multilib-list support.
599 2018-05-09 Richard Biener <rguenther@suse.de>
601 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
603 (vect_bb_vectorization_profitable_p): Adjust. Compute
604 actual scalar cost using the cost vector and the add_stmt_cost
607 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
609 PR rtl-optimization/85645
610 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
611 in the REG_CFA_REGISTER note for LR, don't leave it empty.
613 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
615 PR rtl-optimization/85645
616 * shrink-wrap.c (spread_components): Return a boolean saying if
617 anything was changed.
618 (try_shrink_wrapping_separate): Iterate spread_components until
619 nothing changes anymore.
621 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
623 PR rtl-optimization/85645
624 * regrename.c (build_def_use): Also kill the chains that include the
625 destination of a REG_CFA_REGISTER note.
627 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
629 PR rtl-optimization/85645
630 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
631 insn that has a REG_CFA_REGISTER note.
633 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
635 * cfgexpand.c (expand_clobber): New function.
636 (expand_gimple_stmt_1): Use it.
637 * tree-vect-stmts.c (vect_clobber_variable): New function,
639 (vectorizable_simd_clone_call): ...here.
640 (vectorizable_store): Emit a clobber either side of an
641 IFN_STORE_LANES sequence.
642 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
644 2018-05-09 Tom de Vries <tom@codesourcery.com>
647 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
648 (define_insn "trap_if_false"): Add exit after trap.
650 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
652 PR rtl-optimization/85638
653 * bb-reorder.c: Include common/common-target.h.
654 (create_forwarder_block): New function extracted from...
655 (fix_up_crossing_landing_pad): ...here. Rename into...
656 (dw2_fix_up_crossing_landing_pad): ...this.
657 (sjlj_fix_up_crossing_landing_pad): New function.
658 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
659 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
660 from both partitions and exit the loop after one iteration.
662 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
665 * doc/extend.texi (PowerPC Built-in Functions): Rename this
667 (Basic PowerPC Built-in Functions): The new name of the
668 subsection previously known as "PowerPC Built-in Functions".
669 (Basic PowerPC Built-in Functions Available on all Configurations):
671 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
673 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
675 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
677 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
680 2018-05-08 Jim Wilson <jimw@sifive.com>
682 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
683 (LD_EMUL_SUFFIX): New.
686 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
688 * doc/extend.texi (PowerPC Built-in Functions): Rename this
690 (Basic PowerPC Built-in Functions): The new name of the
691 subsection previously known as "PowerPC Built-in Functions".
692 (Basic PowerPC Built-in Functions Available on all Configurations):
694 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
696 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
698 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
700 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
703 2018-05-08 Jakub Jelinek <jakub@redhat.com>
706 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
707 after cmpelim optimization.
709 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
711 * config.gcc: Support "goldmont".
712 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
713 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
715 * config/i386/i386.c (m_GOLDMONT): Define.
716 (processor_target_table): Add "goldmont".
717 (PTA_GOLDMONT): Define.
718 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
719 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
720 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
721 (fold_builtin_cpu): Add "goldmont".
722 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
723 (ix86_option_override_internal): Add "goldmont".
724 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
725 (processor_type): Add PROCESSOR_GOLDMONT.
726 * config/i386/i386.md: Add CPU "glm".
727 * config/i386/glm.md: New file.
728 * config/i386/x86-tune.def: Add m_GOLDMONT.
729 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
731 2018-05-08 Jakub Jelinek <jakub@redhat.com>
734 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
736 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
737 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
738 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
741 * config/i386/i386.c (ix86_fold_builtin): Handle
742 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
745 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
746 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
748 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
751 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
752 (check_arch): Likewise.
753 (check_fpu): Return the result rather than printing it.
754 (end arch): Fix operator precedence.
756 (END): Print the result from check_fpu.
758 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
759 Alan Hayward <alan.hayward@arm.com>
760 David Sherwood <david.sherwood@arm.com>
762 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
763 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
764 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
765 (*fcmuo<mode>_and): New patterns.
767 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
769 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
770 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
771 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
772 (cmp_op, sve_imm_con): New code attributes.
773 (SVE_COND_INT_CMP, imm_con): Delete.
774 (cmp_op): Remove above unspecs from int attribute.
775 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
777 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
778 comparison-specific unspecs.
779 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
780 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
781 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
782 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
783 (*vec_fcm<cmp_op><mode>): Rename to...
784 (*fcm<cmp_op><mode>): ...this and adjust likewise.
785 (*vec_fcmuo<mode>): Rename to...
786 (*fcmuo<mode>): ...this and adjust likewise.
787 (*pred_fcm<cmp_op><mode>): New pattern.
788 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
789 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
791 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
793 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
794 (aarch64_emit_sve_predicated_cond): New function.
795 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
796 (aarch64_emit_unspec_cond_or): Replace with...
797 (aarch64_emit_sve_or_conds): ...this new function. Use
798 aarch64_emit_sve_ptrue_op for the individual comparisons and
799 aarch64_emit_binop to OR them together.
800 (aarch64_emit_inverted_unspec_cond): Replace with...
801 (aarch64_emit_sve_inverted_cond): ...this new function. Use
802 aarch64_emit_sve_ptrue_op for the comparison and
803 aarch64_emit_unop to invert the result.
804 (aarch64_expand_sve_vec_cmp_float): Update after the above
805 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
807 2018-05-07 Nathan Sidwell <nathan@acm.org>
809 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
810 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
811 (Backwards Compatibility): Likewise.
813 2018-05-07 Luis Machado <luis.machado@linaro.org>
817 2018-05-07 Luis Machado <luis.machado@linaro.org>
819 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
820 <prefetch_dynamic_strides>: New const bool field.
821 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
822 prefetch_dynamic_strides.
823 (exynosm1_prefetch_tune): Likewise.
824 (thunderxt88_prefetch_tune): Likewise.
825 (thunderx_prefetch_tune): Likewise.
826 (thunderx2t99_prefetch_tune): Likewise.
827 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
828 (aarch64_override_options_internal): Update to set
829 PARAM_PREFETCH_DYNAMIC_STRIDES.
830 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
831 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
832 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
833 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
834 prefetch-dynamic-strides setting.
836 2018-05-07 Luis Machado <luis.machado@linaro.org>
838 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
839 <minimum_stride>: New const int field.
840 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
841 minimum_stride field.
842 (exynosm1_prefetch_tune): Likewise.
843 (thunderxt88_prefetch_tune): Likewise.
844 (thunderx_prefetch_tune): Likewise.
845 (thunderx2t99_prefetch_tune): Likewise.
846 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
847 (aarch64_override_options_internal): Update to set
848 PARAM_PREFETCH_MINIMUM_STRIDE.
849 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
850 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
851 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
852 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
853 stride is constant and is below the minimum stride threshold.
855 2018-05-07 Luis Machado <luis.machado@linaro.org>
857 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
860 2018-05-07 Luis Machado <luis.machado@linaro.org>
862 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
863 <prefetch_dynamic_strides>: New const bool field.
864 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
865 prefetch_dynamic_strides.
866 (exynosm1_prefetch_tune): Likewise.
867 (thunderxt88_prefetch_tune): Likewise.
868 (thunderx_prefetch_tune): Likewise.
869 (thunderx2t99_prefetch_tune): Likewise.
870 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
871 (aarch64_override_options_internal): Update to set
872 PARAM_PREFETCH_DYNAMIC_STRIDES.
873 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
874 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
875 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
876 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
877 prefetch-dynamic-strides setting.
879 2018-05-07 Luis Machado <luis.machado@linaro.org>
881 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
882 <minimum_stride>: New const int field.
883 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
884 minimum_stride field.
885 (exynosm1_prefetch_tune): Likewise.
886 (thunderxt88_prefetch_tune): Likewise.
887 (thunderx_prefetch_tune): Likewise.
888 (thunderx2t99_prefetch_tune): Likewise.
889 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
890 (aarch64_override_options_internal): Update to set
891 PARAM_PREFETCH_MINIMUM_STRIDE.
892 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
893 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
894 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
895 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
896 stride is constant and is below the minimum stride threshold.
898 2018-05-06 Jakub Jelinek <jakub@redhat.com>
901 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
902 the type is addressable. Don't force op into register if it has
905 2018-05-05 Roland McGrath <mcgrathr@google.com>
908 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
909 any section for which we don't know a specific type it should have,
910 regardless of name. Previously this was done only for the exact
911 names ".init_array", ".fini_array", and ".preinit_array".
912 (default_elf_asm_named_section): Add comment about
913 relationship with default_section_type_flags and SECTION_NOTYPE.
914 (get_section): Don't consider it a type conflict if one side has
915 SECTION_NOTYPE and the other doesn't, as long as neither has the
916 SECTION_BSS et al used in the default_section_type_flags logic.
918 2018-05-05 Tom de Vries <tom@codesourcery.com>
921 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
922 (workaround_barsyncs): New function.
923 (nvptx_reorg): Use workaround_barsyncs.
924 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
925 (define_expand "nvptx_membar_cta"): New define_expand.
926 (define_insn "*nvptx_membar_cta"): New insn.
928 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
930 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
931 To improve optimization opportunities.
932 * builtin-types.def: The new needed builtin types for the above.
934 2018-05-04 Richard Biener <rguenther@suse.de>
936 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
937 * gimple-ssa-store-merging.c
938 (imm_store_chain_info::output_merged_store): Remove redundant create,
939 release split_store vector contents on failure.
940 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
941 scalar stmt vector on cache hit.
943 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
945 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
947 * config.gcc (powerpc-xilinx-eabi*): Remove.
948 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
950 (fusion_addis_mem_combo_load): Ditto.
951 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
953 (rs6000_cpu_cpp_builtins): Ditto.
954 * config/rs6000/rs6000-linux.c
955 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
956 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
957 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
959 (rs6000_setup_reg_addr_masks): Ditto.
960 (rs6000_init_hard_regno_mode_ok): Ditto.
961 (rs6000_option_override_internal): Ditto.
962 (legitimate_lo_sum_address_p): Ditto.
963 (rs6000_legitimize_address): Ditto.
964 (rs6000_legitimize_reload_address): Ditto.
965 (rs6000_legitimate_address_p): Ditto.
966 (abi_v4_pass_in_fpr): Ditto.
967 (setup_incoming_varargs): Ditto.
968 (rs6000_gimplify_va_arg): Ditto.
969 (rs6000_split_multireg_move): Ditto.
970 (rs6000_savres_strategy): Ditto.
971 (rs6000_emit_prologue_components): Ditto.
972 (rs6000_emit_epilogue_components): Ditto.
973 (rs6000_emit_prologue): Ditto.
974 (rs6000_emit_epilogue): Ditto.
975 (rs6000_elf_file_end): Ditto.
976 (rs6000_function_value): Ditto.
977 (rs6000_libcall_value): Ditto.
978 * config/rs6000/rs6000.h: Ditto.
979 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
980 (TARGET_MINMAX): ... this. New.
981 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
982 * config/rs6000/rs6000.md: Remove Xilinx FP support.
983 (*movsi_internal1_single): Delete.
984 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
985 mfpu=, mxilinx-fpu): Delete.
986 * config/rs6000/singlefp.h: Delete.
987 * config/rs6000/sysv4.h: Remove Xilinx FP support.
988 * config/rs6000/t-rs6000: Ditto.
989 * config/rs6000/t-xilinx: Delete.
990 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
991 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
992 (VStype_simple): Delete.
993 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
994 * config/rs6000/xfpu.h: Delete.
995 * config/rs6000/xfpu.md: Delete.
996 * config/rs6000/xilinx.h: Delete.
997 * config/rs6000/xilinx.opt: Delete.
998 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
999 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1001 2018-05-04 Tom de Vries <tom@codesourcery.com>
1004 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1007 2018-05-04 Richard Biener <rguenther@suse.de>
1010 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1011 (expand_complex_div_wide): Likewise.
1012 (expand_complex_operations_1): Likewise.
1013 (expand_complex_libcall): Preserve EH info of the original stmt.
1014 (tree_lower_complex): Handle removed blocks.
1015 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1016 on complex multiplication and division libcall builtins.
1018 2018-05-04 Richard Biener <rguenther@suse.de>
1021 * fold-const.c (negate_expr_p): Restrict negation of operand
1022 zero of a division to when we know that can happen without
1024 (fold_negate_expr_1): Likewise.
1026 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1029 * real.h (real_nextafter): Declare.
1030 * real.c (real_nextafter): New function.
1031 * fold-const-call.c (fold_const_nextafter): New function.
1032 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1033 CASE_CFN_NEXTTOWARD.
1034 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1035 even when arg1_mode is different from arg0_mode.
1037 2018-05-03 Nathan Sidwell <nathan@acm.org>
1039 * doc/extend.texi (Deprecated Features): Remove
1041 (Backwards Compatibility): Likewise.
1042 * doc/invoke.texi (C++ Language Options): Likewise.
1043 (C++ Dialect Options): Likewise.
1045 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1048 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1049 _mm512_mask_mullox_epi64): New intrinsics.
1051 2018-05-03 Tom de Vries <tom@codesourcery.com>
1054 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1055 dump files): Add offload-tree.
1057 2018-05-03 Richard Biener <rguenther@suse.de>
1059 PR tree-optimization/85615
1060 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1061 to loops not nested in BBs loop father to avoid creating multi-entry
1064 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1066 PR tree-optimization/70291
1067 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1068 arguments. Change return type to tree. Emit libcall as a new
1069 statement rather than replacing existing one when inplace_p is true.
1070 (expand_complex_multiplication_components): New function.
1071 (expand_complex_multiplication): Expand floating-point complex
1072 multiplication using the above.
1073 (expand_complex_division): Rename inner_type parameter to type.
1074 Update expand_complex_libcall call-site.
1075 (expand_complex_operations_1): Update expand_complex_multiplication
1076 and expand_complex_division call-sites.
1078 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1081 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1082 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1083 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1084 the highest significant bit of the shift count mask is clear. In
1085 check whether and[sq]i3 is needed verify that all significant bits
1086 of the shift count other than the highest are set.
1088 2018-05-02 Tom de Vries <tom@codesourcery.com>
1091 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1092 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1093 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1094 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1095 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1096 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1097 __builtin_goacc_parlevel_size.
1099 2018-05-02 Richard Biener <rguenther@suse.de>
1101 PR tree-optimization/85597
1102 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1103 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1106 2018-05-02 Tom de Vries <tom@codesourcery.com>
1109 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1110 dump files): Add ltrans-tree.
1112 2018-05-02 Tom de Vries <tom@codesourcery.com>
1115 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1116 dump files): Add wpa-ipa.
1118 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1120 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1121 powerpc*-*-linux*paired* target.
1122 * config/rs6000/750cl.h: Delete.
1123 * config/rs6000/paired.h: Delete.
1124 * config/rs6000/paired.md: Delete.
1125 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1127 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1128 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1129 comment. Remove paired float support.
1130 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1131 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1133 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1134 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1136 * config/rs6000/rs6000.c: Remove paired float support.
1137 (paired_expand_vector_init, paired_expand_vector_move,
1138 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1139 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1140 paired_expand_builtin, paired_expand_predicate_builtin,
1141 paired_init_builtins): Delete.
1142 * config/rs6000/rs6000.h: Remove paired float support.
1143 * config/rs6000/rs6000.md: Remove paired float support.
1144 (move_from_CR_ov_bit): Delete.
1145 * config/rs6000/rs6000.opt (mpaired): Delete.
1146 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1147 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1149 2018-05-02 Richard Biener <rguenther@suse.de>
1152 * gimplify.c (gimplify_save_expr): When in SSA form allow
1153 SAVE_EXPRs to compute to SSA vars.
1155 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1158 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1159 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1160 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1161 clobber operands[2], instead use a new pseudo. Formatting fixes.
1163 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1165 PR tree-optimization/85586
1166 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1167 exit early for statements in the same group if the accesses are
1170 2018-05-02 Tom de Vries <tom@codesourcery.com>
1173 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1176 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1178 PR tree-optimization/85143
1179 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1181 2018-05-01 Tom de Vries <tom@codesourcery.com>
1184 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1185 not found" error message.
1187 2018-05-01 Tom de Vries <tom@codesourcery.com>
1190 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1191 * vec.c (test_ordered_remove_if): New function.
1192 (vec_c_tests): Call test_ordered_remove_if.
1193 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1194 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1195 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1196 VEC_ORDERED_REMOVE_IF.
1198 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1200 PR tree-optimization/82665
1201 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1202 pointer subtraction where arguments come from a memchr call.
1204 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1206 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1207 --push-state --as-needed and --pop-state instead of --as-needed and
1208 --no-as-needed if ld supports it.
1209 * configure: Regenerated.
1212 * doc/install.texi2html: Replace _002d with - and _002a with * in
1213 generated html files using sed.
1215 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1218 * gcc-rich-location.c (blank_line_before_p): New function.
1219 (use_new_line): New function.
1220 (gcc_rich_location::add_fixit_insert_formatted): New function.
1221 * gcc-rich-location.h
1222 (gcc_rich_location::add_fixit_insert_formatted): New function.
1224 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1226 * selftest.c (assert_streq): Rename "expected" and "actual" to
1227 "val1" and "val2". Extend NULL-handling to cover both inputs
1228 symmetrically, while still requiring both to be non-NULL for a pass.
1229 * selftest.h (assert_streq): Rename "expected" and "actual" to
1231 (ASSERT_EQ): Likewise.
1232 (ASSERT_EQ_AT): Likewise.
1233 (ASSERT_KNOWN_EQ): Likewise.
1234 (ASSERT_KNOWN_EQ_AT): Likewise.
1235 (ASSERT_NE): Likewise.
1236 (ASSERT_MAYBE_NE): Likewise.
1237 (ASSERT_MAYBE_NE_AT): Likewise.
1238 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1239 the assertion to pass.
1240 (ASSERT_STREQ_AT): Likewise.
1242 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1244 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1245 interaction with -pie.
1247 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1249 * selftest.h: Fix alphabetization of per-source-file selftest
1252 2018-04-30 Jason Merrill <jason@redhat.com>
1254 PR c++/61982 - dead stores to destroyed objects.
1255 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1258 2018-04-30 Jason Merrill <jason@redhat.com>
1260 * tree.c (build_clobber): New.
1261 * tree.h: Declare it.
1262 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1264 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1266 * diagnostic-show-locus.c (layout::layout): Update for
1267 location_get_source_line returning a char_span.
1268 (struct char_span): Move to input.h.
1269 (struct correction): Update for fields in char_span becoming
1271 (struct source_line): Update for location_get_source_line
1272 returning a char_span.
1273 (layout::print_line): Likewise.
1274 * edit-context.c (edited_file::print_content): Likewise.
1275 (edited_file::print_diff_hunk): Likewise.
1276 (edited_file::print_run_of_changed_lines): Likewise.
1277 (edited_file::get_num_lines): Likewise.
1278 (edited_line::edited_line): Likewise.
1279 * final.c (asm_show_source): Likewise.
1280 * input.c (location_get_source_line): Convert return type
1281 from const char * to char_span, losing the final "line_len"
1283 (dump_location_info): Update for the above.
1284 (get_substring_ranges_for_loc): Likewise. Use a char_span
1285 when handling the literal within the line.
1286 (test_reading_source_line): Update for location_get_source_line
1287 returning a char_span.
1288 * input.h (class char_span): Move here from
1289 diagnostic-show-locus.c, converting from a struct to a class.
1290 Make data members private.
1291 (char_span::operator bool): New.
1292 (char_span::length): New.
1293 (char_span::get_buffer): New.
1294 (char_span::operator[]): New.
1295 (char_span::subspan): Make const.
1296 (char_span::xstrdup): New.
1297 (location_get_source_line): Convert return type from const char *
1298 to char_span, losing the final "line_size" param.
1300 2018-04-30 Jan Hubicka <jh@suse.cz>
1302 * lto-wrapper.c (ltrans_priorities): New static var.
1303 (cmp_priority): New.
1304 (run_gcc): Read priorities and if doing parallel build order
1305 the Makefile by them.
1307 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1309 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1311 2018-04-30 Richard Biener <rguenther@suse.de>
1313 * tree-cfg.c (verify_address): Remove base argument, add
1314 flag whether to check TREE_ADDRESSABLE and do that.
1315 (verify_expr): Remove.
1316 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1317 (verify_gimple_assign_single): Likewise.
1318 (verify_gimple_switch): Likewise.
1319 (verify_expr_location_1): Dereference tp once. Add (disabled)
1320 piece from verify_expr.
1321 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1323 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1325 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1327 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1329 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1330 (small_data_pattern): Likewise.
1331 (arc_rewrite_small_data): Likewise.
1332 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1333 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1334 (get_symbol_alignment): New function.
1335 (legitimate_small_data_address_p): Likewise.
1336 (legitimate_scaled_address): Update, call
1337 legitimate_small_data_address_p.
1338 (output_sdata): New static variable.
1339 (arc_print_operand): Update how we handle small data operands.
1340 (arc_print_operand_address): Likewise.
1341 (arc_legitimate_address_p): Update, use
1342 legitimate_small_data_address_p.
1343 (arc_rewrite_small_data_p): Remove.
1344 (arc_rewrite_small_data_1): Likewise.
1345 (arc_rewrite_small_data): Likewise.
1346 (small_data_pattern): Likewise.
1347 (compact_sda_memory_operand): Update to use
1348 legitimate_small_data_address_p and get_symbol_alignment.
1349 (prepare_move_operands): Don't rewite sdata pattern.
1350 (prepare_extend_operands): Remove.
1351 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1353 (zero_extendqisi2): Likewise.
1354 (zero_extendhisi2): Likewise.
1355 (extendqihi2): Likewise.
1356 (extendqisi2): Likewise.
1357 (extendhisi2): Likewise.
1361 * config/arc/constraints.md (Usd): Change it to memory constraint.
1363 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1365 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1366 as source of std instructions.
1367 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1368 allow 6-bit constants as source for store instructions.
1369 (movdi_insn): Update instruction pattern to allow 6-bit constants
1370 as source for store instructions.
1372 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1374 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1376 2018-04-30 Nathan Sidwell <nathan@acm.org>
1377 Sandra Loosemore <sandra@codesourcery.com>
1379 * dumpfile.c (dump_open): Allow '-' for stdout.
1380 * doc/invoke.texi (Developer Options): Document dump filename
1381 determination early. Document stdin/stdout selection.
1383 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1385 Microblaze Target: PIC data text relative
1387 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1388 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1390 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1391 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1392 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1393 New addressing mode for data-text relative position indepenedent code.
1394 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1395 'ADDRESS_SYMBOLIC_TXT_REL'.
1396 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1397 (microblaze_legitimate_pic_operand): Exclude function calls from
1398 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1399 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1401 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1402 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1403 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1404 for 'address + offset'.
1405 (microblaze_expand_prologue): Add new function prologue call for
1407 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1408 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1409 table in case of TARGET_PIC_DATA_TEXT_REL.
1410 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1411 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1412 Add new macros 'UNSPEC_TEXT',
1413 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1414 + exclude function calls from 'UNSPEC_PLT' in case of data text
1416 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1417 new target hook for generating address diff vector tables in case of
1419 * doc/tm.texi : Regenerate.
1420 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1421 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1422 of addr diff vector generation.
1423 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1424 target hook definition.
1425 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1426 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1427 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1428 option for data text relative.
1430 2018-04-30 Richard Biener <rguenther@suse.de>
1432 * tree-chrec.h (evolution_function_is_constant_p): Remove
1434 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1436 2018-04-30 Richard Biener <rguenther@suse.de>
1439 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1441 2018-04-30 Richard Biener <rguenther@suse.de>
1443 PR tree-optimization/28364
1444 PR tree-optimization/85275
1445 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1446 copying first exit test.
1448 2018-04-28 Mark Wielaard <mark@klomp.org>
1450 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1452 (dwarf_AT): Handle DW_AT_addr_base.
1453 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1455 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1458 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1459 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1460 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1461 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1463 2018-04-28 Richard Biener <rguenther@suse.de>
1465 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1466 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1467 to reflect use. Only add interesting stmts.
1469 2018-04-27 Martin Jambor <mjambor@suse.cz>
1472 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1473 the jump function allows for passing through aggregate values.
1475 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1477 * input.h (in_system_header_at): Convert from macro to inline
1479 (from_macro_expansion_at): Likewise.
1480 (from_macro_definition_at): Likewise.
1482 2018-04-27 Jeff Law <law@redhat.com>
1484 * config.gcc: Mark tile* targets as deprecated/obsolete.
1486 2018-04-27 Richard Biener <rguenther@suse.de>
1488 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1491 2018-04-27 Richard Biener <rguenther@suse.de>
1493 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1495 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1497 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1498 with Yd constraint. Set "preferred_for_speed" attribute from
1499 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1501 (*movdi_internal): Ditto.
1502 (movti_interunit splitters): Remove
1503 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1504 (movdi_interunit splitters): Ditto.
1505 * config/i386/constraints.md (Ye): Remove.
1506 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1508 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1511 * config/aarch64/constraints.md (Usg): Limit to 31.
1514 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1516 PR tree-optimization/85529
1517 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1518 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1519 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1520 zero extension or masking of the MSB bit.
1521 (optimize_range_tests): Add FIRST_BB argument, pass it through
1522 to optimize_range_tests_var_bound.
1523 (maybe_optimize_range_tests, reassociate_bb): Adjust
1524 optimize_range_tests callers.
1526 2018-04-26 Richard Biener <rguenther@suse.de>
1527 Jakub Jelinek <jakub@redhat.com>
1529 * cgraph.h (symbol_table): Just declare debug method here.
1530 * symtab.c (symbol_table::debug): Define.
1532 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1534 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1536 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1538 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1539 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1540 (*movdi_internal): Substitute Yi and Yj constraint with x
1541 and Ym and Yn constraint with y constraint. Update "isa"
1542 attribute and set "preferred_for_speed" attribute from
1543 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1544 (*movsi_internal): Ditto.
1545 (*movdf_internal): Ditto.
1546 (*movsf_internal): Ditto.
1547 (*zero_extendsidi2): Ditto.
1548 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1549 (sse2_loadld): Ditto.
1550 (*vec_extract<ssevecmodelower>_0): Ditto.
1551 (*vec_extractv4si_0_zext_sse4): Ditto.
1552 (vec_concatv2di): Ditto.
1553 (*vec_dup<mode>): Ditto.
1554 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1555 * config/i386/constraints.md (Yi): Remove.
1560 2018-04-26 Nathan Sidwell <nathan@acm.org>
1562 * dumpfile.c (dump_open): New.
1563 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1564 (dump_finish): Detect stdio/stderr by value not name.
1566 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1568 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1570 2018-04-26 Tom de Vries <tom@codesourcery.com>
1573 * config/nvptx/nvptx.c (verify_neutering_jumps)
1574 (verify_neutering_labels): New function
1575 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1577 2018-04-26 Tom de Vries <tom@codesourcery.com>
1580 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1581 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1584 2018-04-26 Richard Biener <rguenther@suse.de>
1585 Tom de Vries <tom@codesourcery.com>
1588 * lto-streamer-out.c (output_function): Fixup loops if required to match
1589 discovery done in the reader.
1591 2018-04-26 Richard Biener <rguenther@suse.de>
1593 PR tree-optimization/85116
1594 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1595 have a loop exit from the single latch predecessor. Remove
1596 case of header with just condition.
1597 (ch_base::copy_headers): Exclude infinite loops from any
1599 (pass_ch::execute): Record exits.
1601 2018-04-26 Richard Biener <rguenther@suse.de>
1603 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1604 prologue cost vector and pass it to vect_get_load_cost.
1605 (vect_get_peeling_costs_all_drs): Likewise.
1606 (vect_peeling_hash_get_lowest_cost): Likewise.
1607 (vect_enhance_data_refs_alignment): Likewise.
1609 2018-04-26 Richard Biener <rguenther@suse.de>
1612 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1613 checking of integer<->pointer conversions.
1614 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1615 sign-/zero-extending pointer types.
1616 (expand_omp_for_static_chunk): Likewise.
1618 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1619 Jean Lee <xiaoyur347@gmail.com>
1621 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1622 (TARGET_ASAN_SHADOW_OFFSET): Define.
1623 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1624 true for -fsanitize=address.
1626 2018-04-25 Mark Wielaard <mark@klomp.org>
1628 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1631 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1633 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1634 than "alu", remove explicit "memory" and "imm_disp" attributes.
1635 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1638 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1639 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1640 gen_lowpart_no_emit.
1642 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1645 * config/i386/i386.c (ix86_expand_builtin): Change memory
1646 operand to XI, extend p0 to Pmode.
1647 * config/i386/i386.md: Change unspec volatile and operand
1648 1 mode to XI, change operand 0 mode to P.
1650 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1652 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1653 GET_MODE_MASK before any checking.
1654 (nds32_can_use_bset_p): Likewise.
1655 (nds32_can_use_btgl_p): Likewise.
1657 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1659 * config/nds32/nds32-doubleword.md: New define_split pattern for
1660 illegal register number.
1662 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1664 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1666 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1668 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1670 2018-04-25 Richard Biener <rguenther@suse.de>
1672 * lto-streamer.h (LTO_major_version): Bump to 8.
1674 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1676 * BASE-VER: Set to 9.0.0.
1678 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1680 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1681 in __abskf2 and __powikf2.
1683 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1686 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1687 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1688 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1689 Use the above on operand 2. Reindent.
1690 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1692 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1695 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1697 * config/i386/i386.opt (mcet): Removed.
1698 * doc/install.texi: Remove -mcet documentation.
1699 * doc/invoke.texi: Likewise.
1701 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1704 * doc/install.texi: Remove -mcet from bootstrap-cet.
1706 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1709 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1710 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1714 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1715 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1716 containing a CONST_VECTOR.
1718 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1720 * doc/install.texi: Update newlib dependency for nvptx.
1722 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1725 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1726 instead of INTVAL when shifting x left.
1728 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1730 PR tree-optimization/85478
1731 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1732 vect_grouped_store_supported for single element vectors.
1734 2018-04-24 Richard Biener <rguenther@suse.de>
1737 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1738 load cost increase to the case of non-constant step.
1740 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1743 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1744 destination if any_malformed_asm.
1746 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1749 * expr.c (store_field): In the bitfield case, if the value comes from
1750 a function call and is returned in registers by means of a PARALLEL,
1751 do not change the mode of the temporary unless BLKmode and VOIDmode.
1753 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1755 PR rtl-optimization/85423
1756 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1757 dependencies to debug insns when the previous insn is non-debug.
1759 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1761 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1762 enums into a single definition.
1763 (fls): Fix predicates and printing.
1766 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1768 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1769 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1770 and short u6 immediate.
1771 (check_if_valid_sleep_operand): Remove.
1772 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1774 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1776 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1777 flag_always_save_lp condition.
1778 * config/nds32/nds32.opt (malways-save-lp): New option.
1780 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1782 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1783 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1784 * config/nds32/nds32.h
1785 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1786 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1788 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1790 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1791 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1793 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1794 Chung-Ju Wu <jasonwucj@gmail.com>
1796 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1798 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1799 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1801 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1803 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1805 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1807 * config/nds32/nds32-protos.h (nds32_data_alignment,
1808 nds32_local_alignment): Declare.
1809 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1810 nds32_local_alignment): New functions.
1811 (TARGET_CONSTANT_ALIGNMENT): Define.
1812 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1814 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1816 * config/nds32/nds32.c
1817 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1818 (TARGET_MODES_TIEABLE_P): Likewise.
1820 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1822 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1825 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1826 Chung-Ju Wu <jasonwucj@gmail.com>
1828 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1829 for unaligned access.
1830 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1832 * config/nds32/nds32-intrinsic.md: Likewise.
1833 * config/nds32/nds32_intrinsic.h: Likewise.
1834 * config/nds32/nds32.h (nds32_builtins): Likewise.
1835 * config/nds32/nds32.opt (munaligned-access): New option.
1836 * config/nds32/nds32.c (nds32_asm_file_start): Display
1837 flag_unaligned_access status.
1839 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1841 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1842 -mno-relax is present.
1843 * config/riscv/linux.h (LINK_SPEC): Ditto.
1845 2018-04-20 Martin Sebor <msebor@redhat.com>
1848 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1850 (gimple_fold_builtin_stxcpy_chk): Same.
1851 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1853 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1856 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1857 __powikf2 when long double is IEEE 128-bit.
1859 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1861 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1862 step to make sure stack always aligned.
1864 2018-04-20 Carl Love <cel@us.ibm.com>
1867 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1868 size check for arg0.
1870 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1871 Tom de Vries <tom@codesourcery.com>
1874 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1875 Emit insns for calls too.
1876 (nvptx_find_par): Always look for worker-level predecessor insn.
1877 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1879 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1880 (nvptx_process_pars): Propagate frames for calls.
1882 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1885 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1887 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1888 (ix86_handle_option): Don't handle OPT_mibt.
1889 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1891 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1893 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1894 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1895 (ix86_target_macros): Define __CET__ with flag_cf_protection
1896 for -fcf-protection.
1897 * config/i386/i386.c (isa2_opts): Remove -mibt.
1898 * config/i386/i386.h (TARGET_IBT): Removed.
1899 (TARGET_IBT_P): Likewise.
1900 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1901 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1902 * config/i386/i386.opt (mcet): Update help message.
1905 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1906 -mcet as an alias for -mshstk.
1908 2018-04-20 Richard Biener <rguenther@suse.de>
1911 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1912 complexity by forcing a single use of the multiply operand.
1914 2018-04-20 Martin Jambor <mjambor@suse.cz>
1917 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1918 recursion dependency to only apply to non-clones.
1920 2018-04-20 Martin Jambor <mjambor@suse.cz>
1923 * ipa-cp.c (create_specialized_node): Check that clones of
1924 self-recursive edges exist during IPA-CP.
1926 2018-04-19 Toon Moene <toon@moene.org>
1928 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1931 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1933 PR tree-optimization/85467
1934 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1935 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1936 VECTOR_CST element to type.
1938 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1941 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1942 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1943 (builtin_longjmp): Likewise.
1944 (save_stack_nonlocal): New pattern.
1945 (restore_stack_nonlocal): Likewise.
1947 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1950 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1951 Replace ASM_OUTPUT_LABEL with fprintf.
1953 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1956 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1957 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1958 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1959 define __IBT__ and __SHSTK__ for -fcf-protection.
1960 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1962 (ix86_trampoline_init): Likewise.
1963 (x86_output_mi_thunk): Likewise.
1964 (ix86_notrack_prefixed_insn_p): Likewise.
1965 (ix86_option_override_internal): Don't disallow -fcf-protection.
1966 * config/i386/i386.md (rdssp<mode>): Also enable for
1968 (incssp<mode>): Likewise.
1969 (nop_endbr): Likewise.
1970 * config/i386/i386.opt (mcet): Change help message to built-in
1974 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1975 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1976 enable CET built-in functions.
1978 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1980 * common/config/i386/i386-common.c
1981 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1982 OPTION_MASK_ISA_MOVDIRI_UNSET,
1983 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1984 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1985 * config.gcc (movdirintrin.h): New header.
1986 * config/i386/cpuid.h (bit_MOVDIRI,
1987 bit_MOVDIR64B): New bits.
1988 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1990 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1991 (VOID, PVOID, PCVOID)): New function types.
1992 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1993 __builtin_ia32_directstoreu_u64,
1994 __builtin_ia32_movdir64b): New builtins.
1995 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1996 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1998 (ix86_valid_target_attribute_inner_p): Ditto.
1999 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2000 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2001 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2002 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2003 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2004 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2005 (movdiri<mode>, movdir64b_<mode>): New.
2006 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2007 * config/i386/immintrin.h: Include movdirintrin.h.
2008 * config/i386/movdirintrin.h: New file.
2009 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2011 2018-04-19 Richard Biener <rguenther@suse.de>
2014 * cfg.c (clear_bb_flags): When loop state says we have
2015 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2017 2018-04-19 Richard Biener <rguenther@suse.de>
2019 PR tree-optimization/84737
2020 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2021 copying restrict info.
2022 (vect_setup_realignment): Use it.
2023 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2024 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2025 the first DR to all generated stores.
2026 (vectorizable_load): Likewise for loads.
2028 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2030 PR tree-optimization/85446
2031 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2032 the integral and pointer types to have the same precision.
2034 * doc/install.texi: Document --disable-cet being the default and
2037 2018-04-18 Martin Liska <mliska@suse.cz>
2039 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2042 2018-04-18 Martin Liska <mliska@suse.cz>
2045 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2048 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2049 arguments if they are comparable.
2051 2018-04-18 Martin Liska <mliska@suse.cz>
2054 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2057 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2060 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2063 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2064 ENDBR after calling __morestack.
2066 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2069 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2070 by using gcc_base_ver to generate a gcc_driver_version, and use
2071 it when generating GCC_DRIVER_NAME.
2072 * configure: Regenerate.
2074 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2077 * config.gcc: Obsolete powerpc*-*-*spe*.
2079 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2082 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2083 (stabstr_D): Change type of unum from unsigned int to
2084 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2087 2018-04-17 Jim Wilson <jimw@sifive.com>
2090 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2091 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2092 Set arg_pointer_offset after using pretend_args_size.
2094 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2096 PR rtl-optimization/85431
2097 * dse.c (record_store): Ignore zero width stores.
2100 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2101 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2102 __builtin_stack_restore rather than after it.
2103 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2104 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2105 argument instead of virtual_dynamic_stack_rtx.
2107 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2109 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2111 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2112 Add note to error message to explain internal mapping of overloaded
2113 built-in function name to non-overloaded built-in function name.
2114 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2117 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2120 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2121 where the inputs overlap with the output.
2123 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2125 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2126 (=v, v) alternative and explicit "memory" attribute.
2127 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2128 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2130 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2131 "sselog1" type instead of "sselog".
2132 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2133 "sselog". Remove explicit "memory" attribute.
2134 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2135 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2137 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2138 "sselog1" type instead of "sselog". Remove explicit "memory"
2140 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2141 use "sselog1" type instead of "sselog". Remove explicit "memory"
2143 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2144 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2146 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2147 "sselog1" type instead of "sselog". Remove explicit "memory"
2149 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2150 use "sselog1" type instead of "sselog". Remove explicit "memory"
2154 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2157 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2160 2018-04-17 Martin Jambor <mjambor@suse.cz>
2163 * ipa-cp.c (create_specialized_node): Call
2164 expand_all_artificial_thunks if necessary.
2166 2018-04-17 Martin Liska <mliska@suse.cz>
2169 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2170 in message, remote space in between '_G' and '('.
2172 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2175 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2176 avx512f_vmcmp<mode>3<round_saeonly_name>,
2177 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2178 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2179 avx512f_rndscale<mode><round_saeonly_name>,
2180 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2181 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2182 Use %<iptr>2 instead of %2 for -masm=intel.
2183 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2184 avx512f_vcvttss2usi<round_saeonly_name>,
2185 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2187 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2188 avx512f_vcvttsd2usi<round_saeonly_name>,
2189 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2190 Use %q1 instead of %1 for -masm=intel.
2191 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2192 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2193 of %3 for -masm=intel.
2194 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2196 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2198 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2200 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2201 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2203 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2205 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2206 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2207 %g1 and one with %0 and %1.
2208 (avx512er_vmrcp28<mode><round_saeonly_name>,
2209 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2211 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2212 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2213 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2214 of %0 and %{%4%} for -masm=intel.
2215 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2216 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2217 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2218 order of %0 and %{%5%}%{z%} for -masm=intel.
2220 2018-04-17 Jan Hubicka <jh@suse.cz>
2223 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2225 2018-04-17 Martin Liska <mliska@suse.cz>
2228 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2229 for target_clone error message. Make default implementation
2230 clone to be a local declaration.
2231 (separate_attrs): Add new argument and check for an empty
2233 (expand_target_clones): Handle it.
2234 (ipa_target_clone): Make redirection just for target_clones
2237 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2238 Tom de Vries <tom@codesourcery.com>
2241 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2242 tiled basic blocks without omp continue statements.
2244 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2247 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2248 vec_extract expression as having side effects to make sure it gets
2251 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2254 * config/i386/i386.c (get_builtin_code_for_version): Check
2257 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2260 * gcc/config.gcc: Support "skylake".
2261 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2263 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2264 (processor_target_table): Add "skylake".
2265 (ix86_option_override_internal): Add "skylake".
2266 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2267 PROCESSOR_CANNONLAKE.
2268 (get_builtin_code_for_version): Fix priority for
2269 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2270 PROCESSOR_SKYLAKE-AVX512.
2271 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2272 (processor_type): Add PROCESSOR_SKYLAKE.
2274 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2275 Jason Merrill <jason@redhat.com>
2278 * convert.c (convert_to_integer_1): Use direct recursion for
2279 enumeral types and types with a precision less than the number
2280 of bits in their mode.
2282 2018-04-16 Julia Koval <julia.koval@intel.com>
2285 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2286 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2288 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2291 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2292 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2293 and -mno-direct-move.
2295 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2298 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2299 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2300 Ensure negative shifts result in {0}.
2302 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2304 PR rtl-optimization/79916
2305 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2306 regs (if any) to define how to gnerate SD moves when LRA is in
2309 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2311 PR rtl-optimization/85393
2312 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2313 * except.c (expand_dw2_landing_pad_for_region): Make static.
2314 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2315 a label and unconditional jump to old_bb, rather than
2316 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2319 PR rtl-optimization/85376
2320 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2321 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2322 instead of a specific value.
2324 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2325 Bin Cheng <bin.cheng@arm.com>
2327 PR tree-optimization/82965
2328 PR tree-optimization/83991
2329 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2330 by_profile_only parameter.
2331 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2332 information if the loop was predicted to iterate too many times.
2333 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2335 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2338 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2341 2018-04-13 Martin Liska <mliska@suse.cz>
2342 Jakub Jelinek <jakub@redhat.com>
2345 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2346 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2347 * builtins.c (expand_builtin_memory_copy_args): Use
2348 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2349 handle dest_addr == pc_rtx.
2351 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2354 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2355 asked to not generate direct moves.
2356 (fix_trunc<mode>si2_stfiwx): Similar.
2357 (fix_trunc<mode>si2_internal): Similar.
2359 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2362 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2363 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2364 lookup if dest in some wider mode is known to be const0_rtx and
2365 if so, record permanent equivalence for it to be ZERO_EXTEND of
2366 the narrower mode destination.
2368 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2370 * lto-streamer-out.c (output_function): Revert 259346.
2371 * omp-expand.c (expand_oacc_for): Likewise.
2373 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2375 PR rtl-optimization/85354
2376 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2377 * sel-sched.c (sel_global_init): ... here.
2379 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2382 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2383 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2384 mode for PE-COFF targets.
2385 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2386 (i386_pe_asm_lto_end): Likewise.
2387 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2388 (TARGET_ASM_LTO_END): Likewise.
2389 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2390 (i386_pe_asm_lto_start): New function.
2391 (i386_pe_asm_lto_end): Likewise.
2393 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2394 Richard Biener <rguenther@suse.de>
2397 * lto-streamer-out.c (output_function): Fix CFG loop state before
2399 * omp-expand.c (expand_oacc_for): Handle calls to internal
2400 functions like regular functions.
2402 2018-04-12 Richard Biener <rguenther@suse.de>
2405 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2406 for the early LTO debug to properly generate references to it
2407 during DIE emission. Do not re-use that for the skeleton for
2409 (dwarf2out_early_finish): Likewise.
2411 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2414 * config/i386/sse.md
2415 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2416 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2417 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2418 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2419 and output is a reg, avoid creating invalid lowpart subreg, but
2420 instead split into a 512-bit move. Don't split if not AVX512VL,
2421 input is xmm16+ reg and output is a mem.
2422 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2423 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2424 xmm16+ reg and output is a mem.
2426 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2428 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2429 also for flag_dwarf2_cfi_asm.
2431 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2433 PR rtl-optimization/85342
2434 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2435 a bool scalar var inside of the loop instead. Don't try to update
2436 recog_data.operand after failed apply_change_group.
2438 2018-04-12 Tom de Vries <tom@codesourcery.com>
2441 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2442 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2443 array with flexible array member as array without given dimension.
2444 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2445 argument for undefined param to true.
2447 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2450 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2451 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2452 from PowerPC section.
2453 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2454 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2455 help text that is too long.
2456 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2457 help text that is too long.
2458 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2459 help text that is too long.
2461 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2463 * config/alpha/alpha.md (stack_probe_internal): Rename
2464 from "probe_stack". Update all callers.
2466 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2468 PR rtl-optimization/84566
2469 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2470 sched_macro_fuse_insns.
2472 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2475 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2476 (compute_block_dependences): ... from here.
2478 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2480 PR tree-optimization/85331
2481 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2482 from int to HOST_WIDE_INT.
2484 2018-04-11 Martin Jambor <mjambor@suse.cz>
2487 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2488 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2489 not the same as the source val.
2490 (cgraph_edge_brings_value_p): New parameter.
2491 (gather_edges_for_value): Pass destination value to
2492 cgraph_edge_brings_value_p.
2493 (perhaps_add_new_callers): Likewise.
2494 (get_info_about_necessary_edges): Likewise and exclude values brought
2495 only by self-recursive edges.
2496 (create_specialized_node): Redirect only clones of self-calling edges.
2497 (+self_recursive_pass_through_p): New function.
2498 (find_more_scalar_values_for_callers_subset): Use it.
2499 (find_aggregate_values_for_callers_subset): Likewise.
2500 (known_aggs_to_agg_replacement_list): Removed.
2501 (decide_whether_version_node): Re-calculate known constants for all
2502 remaining context clones.
2504 2018-04-11 Richard Biener <rguenther@suse.de>
2507 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2508 from early DWARF output.
2509 (dwarf2out_early_finish): Output line info unconditionally into
2510 early DWARF and add reference to it.
2512 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2515 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2516 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2517 other than V2DFmode using iptr mode attribute.
2518 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2520 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2522 PR rtl-optimization/84659
2523 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2525 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2528 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2530 (output_loc_list): Pass address of a dummy size variable even in the
2531 locview handling loop.
2532 (index_location_lists): Add comment on why skip_loc_list_entry can't
2535 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2538 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2541 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2544 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2545 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2546 and -mstring-compare-inline-limit.
2548 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2551 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2552 for stack clash protection in a register whenever we need it to be in
2555 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2557 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2558 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2560 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2563 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2565 (mlong-double-): Ditto.
2566 * config/rs6000/sysv4.opt (msdata=): Ditto.
2567 (mtls-size=): Ditto.
2569 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2571 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2572 erroneous entries for
2573 "vector int vec_ldl (int, long int *)", and
2574 "vector unsigned int vec_ldl (int, unsigned long int *)".
2575 Add comments and entries for
2576 "vector bool char vec_ldl (int, bool char *)",
2577 "vector bool short vec_ldl (int, bool short *)",
2578 "vector bool int vec_ldl (int, bool int *)",
2579 "vector bool long long vec_ldl (int, bool long long *)",
2580 "vector pixel vec_ldl (int, pixel *)",
2581 "vector long long vec_ldl (int, long long *)",
2582 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2583 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2584 type tree bool_long_long_type_node and correct definition of
2585 bool_V2DI_type_node to make reference to this new type tree.
2586 (rs6000_mangle_type): Replace erroneous reference to
2587 bool_long_type_node with bool_long_long_type_node.
2588 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2589 comments to emphasize sign distinctions for char and int types and
2590 replace RS6000_BTI_bool_long constant with
2591 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2592 use of RS6000_BTI_pixel.
2593 (bool_long_type_node): Remove this macro definition.
2594 (bool_long_long_type_node): New macro definition
2596 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2598 PR rtl-optimization/85300
2599 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2600 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2601 simplify_unary_operation fails.
2603 2018-04-10 Martin Liska <mliska@suse.cz>
2605 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2606 cgraph_edge and ipa_ref.
2608 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2612 * config/i386/sse.md
2613 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2614 computation of the VEC_MERGE selector from mask.
2615 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2616 Fix decoding of the VEC_MERGE selector into mask.
2618 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2620 PR tree-optimization/85286
2621 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2623 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2625 * final.c (final_1): Set insn_last_address as well as
2626 insn_current_address.
2628 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2631 * explow.c (emit_stack_probe): Call validize_mem on memory location
2632 before passing it to gen_probe_stack. Create address operand and
2633 legitimize it for the probe_stack_address case.
2635 2018-04-09 Jan Hubicka <jh@suse.cz>
2638 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2639 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2640 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2642 (free_lang_data): Rebuild type inheritance graph.
2644 2018-04-09 Martin Sebor <msebor@redhat.com>
2646 * invoke.texi (-finline-small-functions): Mention other optimization
2648 (-findirect-inlining, -fpartial-inlining): Same.
2649 (-finline-functions-called-once): Same.
2650 (-freorder-blocks-and-partition): Same.
2652 2018-04-09 Jan Hubicka <jh@suse.cz>
2655 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2656 jumps; choose last target that matches the criteria (i.e.
2657 no partition changes for non-crossing jumps).
2658 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2659 support for redirecting crossing jumps to non-crossing.
2661 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2663 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2664 also for naked functions.
2666 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2668 * config/arc/arc.md (add_shift): New pattern.
2669 (add_shift2): Likewise.
2670 (sub_shift): Likewise.
2671 (sub_shift_cmp0_noout): Likewise.
2672 (compare_si_ashiftsi): Likewise.
2673 (xbfu_cmp0_noout): New combine pattern.
2674 (xbfu_cmp0"): Likewise.
2675 (movsi_set_cc_insn): Place the predicable variant first.
2676 (commutative_binary_cmp0_noout): Remove clobber.
2677 (commutative_binary_cmp0): New pattern.
2678 (noncommutative_binary_cmp0): Likewise.
2679 (noncommutative_binary_cmp0_noout): Likewise.
2680 (noncommutative_binary_comparison_result_used): Removed.
2681 (rsub_cmp0): New pattern.
2682 (rsub_cmp0_noout): Likewise.
2683 (extzvsi): Changed, keep only meaningful variants.
2684 (SQH, SEZ): New iterators.
2685 (SQH_postfix): New mode attribute.
2686 (SEZ_prefix): New code attribute.
2687 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2688 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2689 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2691 (noncommutative_operator): Check the availability of barrel
2694 2018-04-09 Richard Biener <rguenther@suse.de>
2696 PR tree-optimization/85284
2697 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2698 Only use the niter constraining form of simple_iv when the exit
2701 2018-04-09 Tom de Vries <tom@codesourcery.com>
2704 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2705 (define_expand "*memory_barrier"): New define_expand.
2706 (define_insn "memory_barrier"): New insn.
2708 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2710 PR rtl-optimization/80463
2711 PR rtl-optimization/83972
2712 PR rtl-optimization/83480
2714 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2715 correct producer for the insn.
2716 (tidy_control_flow): Fixup seqnos in case of debug insns.
2718 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2720 PR rtl-optimization/83913
2722 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2723 different sched-times when merging exprs.
2725 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2727 PR rtl-optimization/83962
2729 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2730 tidy_fallthru_edge and tidy_control_flow.
2732 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2734 PR rtl-optimization/83530
2736 * sel-sched.c (force_next_insn): New global variable.
2737 (remove_insn_for_debug): When force_next_insn is true, also leave only
2738 next insn in the ready list.
2739 (sel_sched_region): When the region wasn't scheduled, make another pass
2740 over it with force_next_insn set to 1.
2742 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2744 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2746 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2747 for interrupt control.
2748 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2749 functions for interrupt control.
2750 * config/nds32/nds32-intrinsic.md: Likewise.
2751 * config/nds32/nds32_intrinsic.h: Likewise.
2752 * config/nds32/nds32.h (nds32_builtins): Likewise.
2754 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2756 * config/nds32/nds32.c (nds32_init_machine_status,
2757 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2758 strict_aligned_p field.
2759 (nds32_expand_to_rtl_hook): New function.
2760 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2761 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2763 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2764 Chung-Ju Wu <jasonwucj@gmail.com>
2766 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2767 * config/nds32/nds32-n7.md: New file.
2768 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2769 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2771 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2772 * config/nds32/nds32.md (pipeline_model): Add n7.
2773 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2774 * config/nds32/pipelines.md: Include n7 settings.
2776 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2777 Chung-Ju Wu <jasonwucj@gmail.com>
2779 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2780 * config/nds32/nds32-e8.md: New file.
2781 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2782 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2784 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2785 * config/nds32/nds32.md (pipeline_model): Add e8.
2786 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2787 * config/nds32/pipelines.md: Include e8 settings.
2789 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2790 Chung-Ju Wu <jasonwucj@gmail.com>
2792 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2793 * config/nds32/nds32-n8.md: New file.
2794 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2795 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2797 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2798 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2799 * config/nds32/nds32.md (pipeline_model): Add n8.
2800 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2801 * config/nds32/pipelines.md: Include n8 settings.
2803 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2804 Chung-Ju Wu <jasonwucj@gmail.com>
2806 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2807 * config/nds32/nds32-n9-2r1w.md: New file.
2808 * config/nds32/nds32-n9-3r2w.md: New file.
2809 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2810 nds32_register_ports): New or modify for cpu n9.
2811 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2813 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2814 * config/nds32/nds32-utils.c: New file.
2815 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2816 TARGET_MUL_SLOW): Define.
2817 * config/nds32/nds32.md (pipeline_model): New attribute.
2818 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2819 New options that support cpu n9.
2820 * config/nds32/pipelines.md: Include n9 settings.
2821 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2823 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2825 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2826 information if necessary.
2827 (output_cond_branch_compare_zero): Likewise.
2828 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2829 (nds32_target_alignment): Refine for alignment.
2830 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2831 (FUNCTION_BOUNDARY): Modify.
2832 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2834 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2836 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2838 * config/nds32/constants.md (unspec_volatile_element): Add values for
2839 TLB operation and data prefetch.
2840 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2841 functions for TLB operation and data prefetch.
2842 * config/nds32/nds32-intrinsic.md: Likewise.
2843 * config/nds32/nds32_intrinsic.h: Likewise.
2844 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2845 (nds32_print_operand): Likewise.
2846 * config/nds32/nds32.h (nds32_builtins): Likewise.
2848 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2849 Andrew Pinski <pinsika@gcc.gnu.org>
2852 * match.pd: Use constant_boolean_node of correct type instead of
2853 boolean_true_node or boolean_false_node for simplifying
2854 pointer comparisons to zero.
2856 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2858 PR tree-optimization/80021
2859 * tree.c (verify_type_variant): Make error call in verify_variant_match
2860 translatable and remove final full stop.
2862 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2864 * config/nds32/constants.md (unspec_volatile_element): Add
2865 UNSPEC_VOLATILE_EH_RETURN.
2866 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2867 nds32_output_stack_pop): Support dwarf exception handling process.
2868 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2869 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2870 exception handling process.
2871 (nds32_compute_stack_frame): Likewise.
2872 (nds32_return_addr_rtx): Likewise.
2873 (nds32_initial_elimination_offset): Likewise.
2874 (nds32_expand_prologue): Likewise.
2875 (nds32_expand_epilogue): Likewise.
2876 (nds32_dynamic_chain_address): New function.
2877 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2879 (DYNAMIC_CHAIN_ADDRESS): Define.
2880 (EH_RETURN_DATA_REGNO): Define.
2881 (EH_RETURN_STACKADJ_RTX): Define.
2882 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2883 patterns for dwarf exception handling.
2885 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2887 * config/nds32/nds32.h: Clean up obsolete macros.
2889 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2891 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2892 Add enum values for particular instructions.
2893 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2894 particular intrinsic functions.
2895 * config/nds32/nds32-intrinsic.md: Likewise.
2896 * config/nds32/nds32_intrinsic.h: Likewise.
2897 * config/nds32/nds32.h (nds32_builtins): Likewise.
2898 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2899 (btst, ave): New patterns for particular instructions.
2901 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2903 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2904 Add enum values for atomic load/store and memory sync.
2905 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2907 * config/nds32/nds32-intrinsic.md: Likewise.
2908 * config/nds32/nds32_intrinsic.h: Likewise.
2909 * config/nds32/nds32.h (nds32_builtins): Likewise.
2911 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2913 PR tree-optimization/85257
2914 * fold-const.c (native_encode_vector): If not all elts could fit
2915 and off is -1, return 0 rather than offset.
2916 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2917 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2918 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2919 adjust buffer in native_interpret_expr call.
2921 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2923 * config/nds32/constants.md (unspec_volatile_element): Add cache
2924 control enum values.
2925 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2926 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2927 * config/nds32/nds32.c (nds32_cctl_names): New.
2928 (nds32_print_operand): Handle cache control register names.
2929 * config/nds32/nds32.h (nds32_builtins): New enum values.
2930 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2932 * config/nds32/nds32.md (type): Add mmu.
2933 * config/nds32/pipelines.md (simple_insn): Add mmu.
2935 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2937 * config/nds32/nds32.md (type): Remove call.
2938 * config/nds32/pipelines.md (simple_insn): Likewise.
2940 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2942 * config/nds32/constants.md (unspec_volatile_element): Add
2943 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2944 UNSPEC_VOLATILE_FMFCFG.
2945 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2946 description for fmfcfg and fmfcsr.
2947 (bdesc_1arg): Add fmtcsr.
2948 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2949 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2950 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2951 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2952 unspec_fmfcfg): New patterns.
2953 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2954 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2955 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2956 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2957 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2958 __nds32__fmfcfg): Define.
2960 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2962 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2963 intrinsic register names.
2964 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2965 intrinsic register enum values and macros.
2967 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2969 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2970 for load/store addressing form.
2971 (nds32_print_operand_address): Likewise.
2973 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2976 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2977 based on LABEL_REF. Remove useless assertion.
2978 (pic_address_needs_scratch): Fix formatting.
2979 (sparc_legitimize_pic_address): Minor tweaks.
2980 (sparc_delegitimize_address): Adjust assertion accordingly.
2981 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2982 into symbolic_operand.
2983 (movsi_high_pic_label_ref): Likewise.
2984 (movsi_lo_sum_pic_label_ref): Likewise.
2985 (movdi_pic_label_ref): Likewise.
2986 (movdi_high_pic_label_ref): Likewise.
2987 (movdi_lo_sum_pic_label_ref): Likewise.
2989 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2991 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2992 custom LIB_SPEC setup.
2994 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2995 Kito Cheng <kito.cheng@gmail.com>
2997 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2998 * config/riscv/freebsd.h: New.
3000 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3002 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3003 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3006 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3007 Kito Cheng <kito.cheng@gmail.com>
3009 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3010 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3011 * config/nds32/nds32-protos.h (nds32_output_call,
3012 nds32_output_return): Declare.
3013 * config/nds32/nds32.md: Refine all the call and return patterns.
3015 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3018 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3019 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3021 PR rtl-optimization/84872
3022 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3023 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3026 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3028 * expr.c (copy_blkmode_to_reg): Revert 254862.
3029 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3031 2018-04-06 Richard Biener <rguenther@suse.de>
3034 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3035 after seeing a component reference with an adjacent field. Treat
3036 refs to arrays at struct end of external decls similar to
3037 refs to unconstrained commons.
3039 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3042 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3043 look through SAVE_EXPRs with non-side-effects argument. Adjust
3045 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3048 2018-04-06 Richard Biener <rguenther@suse.de>
3051 * alias.c (find_base_term): New wrapper around find_base_term
3052 unwinding CSELIB_VAL_PTR changes.
3053 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3056 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3058 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3060 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3061 constant definitions.
3062 ("nop"): lr 0,0 -> nopr r0
3063 ("nop_lr0", "nop_lr1"): New insn definitions.
3065 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3067 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3068 NDS32_V3PUSH_AVAILABLE_P macro.
3070 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3071 Chung-Ju Wu <jasonwucj@gmail.com>
3073 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3074 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3075 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3076 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3077 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3078 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3079 * config/nds32/constraints.md: New constraints and checking for hard
3080 float configuration.
3081 * config/nds32/iterators.md: New mode iterator and attribute for hard
3082 float configuration.
3083 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3085 * config/nds32/nds32-fpu.md: New file.
3086 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3087 deal with hard float code generation.
3088 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3090 (abi_type, float_reg_number): New enum type.
3091 * config/nds32/nds32-predicates.c: New predicates for hard float.
3092 * config/nds32/nds32-protos.h: Declare functions for hard float.
3093 * config/nds32/nds32.c: Implementation for hard float configuration.
3094 * config/nds32/nds32.h: Definitions for hard float configuration.
3095 * config/nds32/nds32.md: Include hard float machine description and
3096 modify patterns for hard float configuration.
3097 * config/nds32/nds32.opt: New options for hard float configuration.
3098 * config/nds32/predicates.md: New predicates for hard float
3101 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3103 * common/config/nds32/nds32-common.c
3104 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3106 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3109 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3110 CONSTRUCTOR_ELT (ctor, ...)->value.
3112 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3115 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3117 2018-04-05 Tom de Vries <tom@codesourcery.com>
3120 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3123 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3124 Kito Cheng <kito.cheng@gmail.com>
3126 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3127 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3128 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3129 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3131 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3132 Kito Cheng <kito.cheng@gmail.com>
3134 * config/nds32/constraints.md (Ufe): New memory constraint.
3135 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3136 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3137 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3139 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3140 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3142 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3144 * config/nds32/nds32.md: Use optimize_size in the condition for
3145 alu-shift instructions.
3147 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3149 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3151 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3153 * config/nds32/nds32.md (negsi2): Refine pattern.
3155 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3156 Chung-Ju Wu <jasonwucj@gmail.com>
3158 * config/nds32/iterators.md (shift_rotate): New code iterator.
3159 (shift): New code attribute.
3160 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3161 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3162 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3163 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3164 bit-wise operations.
3165 (andsi3, *andsi3): Ditto.
3166 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3167 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3168 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3169 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3170 nds32_ior_operand, nds32_xor_operand): New predicates.
3172 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3174 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3175 (addsi3, subsi3): ... this.
3177 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3179 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3181 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3183 * config/nds32/nds32.md: Adjust indention.
3185 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3187 * config/nds32/nds32.md (feature): New attribute.
3189 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3191 * config/nds32/nds32.md (subtype): New attribute.
3193 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3196 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3197 expansion to perform a bitwise AND of the argument followed by a
3198 boolean negation of the result.
3200 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3202 PR rtl-optimization/84878
3203 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3204 the basic block. Assert the use reference is not artificial and that
3205 it has an associated insn.
3207 2018-04-04 Michael Matz <matz@suse.de>
3209 * builtins.c (compute_objsize): Pass correct operand
3210 to array_at_struct_end_p.
3212 2018-04-04 Richard Biener <rguenther@suse.de>
3215 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3216 from contexts for DINFO_LEVEL_TERSE and below.
3218 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3220 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3221 resiter_operand condition.
3222 * config/nds32/nds32.md (*move<mode>): Ditto.
3224 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3225 Monk Chiang <sh.chiang04@gmail.com>
3227 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3229 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3231 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3233 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3234 Kito Cheng <kito.cheng@gmail.com>
3236 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3237 nds32_cond_code_str, output_cond_branch,
3238 output_cond_branch_compare_zero, nds32_expand_cbranch,
3239 nds32_expand_cstore, nds32_expand_movcc,
3240 nds32_output_cbranchsi4_equality_zero,
3241 nds32_output_cbranchsi4_equality_reg,
3242 nds32_output_cbranchsi4_equality_reg_or_const_int,
3243 nds32_output_cbranchsi4_greater_less_zero: New functions.
3244 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3245 nds32_expand_cstore, nds32_expand_movcc,
3246 nds32_output_cbranchsi4_equality_zero,
3247 nds32_output_cbranchsi4_equality_reg,
3248 nds32_output_cbranchsi4_equality_reg_or_const_int,
3249 nds32_output_cbranchsi4_greater_less_zero): Declare.
3250 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3251 nds32_rimm11s_operand): New predicates.
3252 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3253 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3256 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3258 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3259 * config/nds32/nds32.md: Ditto.
3260 * config/nds32/pipelines.md: Ditto.
3262 2018-04-04 Richard Biener <rguenther@suse.de>
3264 PR tree-optimization/85168
3265 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3266 propagating abnormals.
3268 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3270 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3272 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3273 Kito Cheng <kito.cheng@gmail.com>
3275 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3276 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3277 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3278 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3279 * config/nds32/nds32.md (sibcall_internal): New.
3280 (sibcall_register): Remove.
3281 (sibcall_immediate): Remove.
3282 (sibcall_value_internal): New.
3283 (sibcall_value_register): Remove.
3284 (sibcall_value_immediate): Remove.
3285 * config/nds32/predicates.md (nds32_general_register_operand): New.
3286 (nds32_call_address_operand): New.
3288 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3290 PR rtl-optimization/85167
3291 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3292 bb_defs if *split_p, instead preinitialize it to NULL.
3294 PR tree-optimization/85156
3295 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3296 evaluating the argument multiple times.
3298 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3300 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3302 (_mm_cvtpd_ps): Likewise.
3303 (_mm_cvttpd_epi32): Likewise.
3304 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3305 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3306 vector, pixel, and bool following altivec.h include.
3308 2018-04-03 Martin Sebor <msebor@redhat.com>
3310 * doc/extend.texi (Common Function Attributes): Clarify.
3311 (const attribute): Likewise.
3312 (pure attribute): Likewise.
3314 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3317 * config/i386/i386.c (ix86_expand_vector_set): Use
3318 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3320 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3322 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3323 instructions when changing rounding bits to preserve precision bits
3324 in the x87 control word.
3326 2018-04-03 Martin Liska <mliska@suse.cz>
3328 PR tree-optimization/82491
3329 * rtl.h (strip_offset_and_add): Replace += suboffset with
3330 poly_uint64 () + suboffset.
3332 2018-03-29 Martin Liska <mliska@suse.cz>
3333 Martin Jambor <mjambor@suse.cz>
3336 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3337 param_type is not an integral or pointer type.
3339 2018-04-03 Richard Biener <rguenther@suse.de>
3341 * sese.h (recompute_all_dominators): Remove.
3343 2018-04-02 Martin Sebor <msebor@redhat.com>
3345 * doc/invoke.texi (-Wrestrict): Fix typos.
3347 2018-04-02 Jim Wilson <jimw@sifive.com>
3349 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3350 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3351 (<optab>di3, <optab>si3_extend): Likewise.
3352 (<optab>si3_mask, <optab>si3_mask_1): New.
3353 (<optab>di3_mask, <optab>di3_mask_1): New.
3354 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3355 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3356 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3358 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3360 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3363 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3365 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3366 (nds32_canonicalize_comparison): New function.
3368 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3369 Kito Cheng <kito.cheng@gmail.com>
3370 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3372 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3373 * config/nds32/constants.md (unspec_volatile_element): Add
3374 UNSPEC_VOLATILE_RELAX_GROUP.
3375 * config/nds32/nds32-relax-opt.c: New file.
3376 * config/nds32/nds32-predicates.c
3377 (nds32_symbol_load_store_p): New function.
3378 * config/nds32/nds32-protos.h
3379 (nds32_symbol_load_store_p): Declare function.
3380 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3381 * config/nds32/nds32.c
3382 (nds32_register_pass): New function to register pass.
3383 (nds32_register_passes): New function to register passes.
3384 * config/nds32/nds32.md (relax_group): New pattern.
3385 * config/nds32/nds32.opt (mrelax-hint): New option.
3386 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3388 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3390 * config/nds32/t-nds32: Modify files dependency.
3392 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3394 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3395 (PROFILE_HOOK): Define its implementation.
3397 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3399 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3400 type and 32-bit size.
3402 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3405 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3406 (V_128_256): New mode iterator.
3407 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3408 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3409 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3411 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3412 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3414 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3417 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3418 NaN inputs correctly.
3420 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3423 * config/rs6000/vsx.md (??r): New mode attribute.
3424 (*vsx_mov<mode>_64bit): Use it.
3425 (*vsx_mov<mode>_32bit): Likewise.
3427 2018-03-30 Martin Sebor <msebor@redhat.com>
3429 PR tree-optimization/84818
3430 * builtins.c (check_access): Use warning_n.
3432 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3435 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3437 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3440 2018-03-30 Julia Koval <julia.koval@intel.com>
3443 * x86-tune.def (movx, partial_reg_dependency): Enable for
3446 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3449 * lra-constraints.c (process_alt_operands): Move setting
3450 this_alternative_matches below.
3452 2018-03-29 Martin Liska <mliska@suse.cz>
3455 * doc/invoke.texi: Document how LTO works with debug info.
3456 Describe auto-load support of binutils. Mention 'x86-64'
3457 as valid option value of -march option.
3459 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3461 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3464 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3465 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3466 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3469 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3472 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3473 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3474 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3475 for RS6000_BTM_POWERPC64.
3476 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3477 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3478 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3483 2018-03-28 Carl Love <cel@us.ibm.com>
3486 2017-09-27 Carl Love <cel@us.ibm.com>
3488 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3489 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3490 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3493 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3495 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3496 instead of __vector bool.
3497 (_mm_max_pu8): Likewise.
3498 (_mm_min_pi16): Likewise.
3500 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3503 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3504 (DIVWEUO): Likewise.
3506 (DIVDEUO): Likewise.
3507 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3508 DIVWEUO and DIVDEUO.
3509 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3510 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3511 (div_extend): Likewise.
3512 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3514 (__builtin_divweuo): Likewise.
3515 (__builtin_divdeo): Likewise.
3516 (__builtin_divdeuo): Likewise.
3518 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3521 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3522 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3524 PR tree-optimization/82004
3525 * gimple-match-head.c (optimize_pow_to_exp): New function.
3526 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3527 Don't fold to exp if optimize_pow_to_exp is false.
3529 2018-03-28 Martin Liska <mliska@suse.cz>
3532 * calls.c (initialize_argument_information): Fix trailing space.
3533 * common.opt: Fix typo and provide better explanation for
3534 -fsanitize-coverage option.
3535 * config/i386/i386.opt: Fix typo.
3537 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3538 Martin Liska <mliska@suse.cz>
3541 * gimplify.c (asan_poison_variable): Don't do the check for
3542 gimplify_omp_ctxp here.
3543 (gimplify_decl_expr): Do it here.
3544 (gimplify_target_expr): Likewise.
3546 2018-03-28 Martin Liska <mliska@suse.cz>
3549 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3550 chkp_type_bounds_count if MPX is not enabled.
3552 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3554 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3556 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3559 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3560 function to create the function decl for complex long double
3561 multiply and divide for -mabi=ieeelongdouble.
3562 (init_float128_ieee): Call it.
3564 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3567 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3568 -fcf-protection=branch -mibt.
3569 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3571 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3574 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3576 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3579 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3580 extern array declarations.
3582 2018-03-27 Richard Biener <rguenther@suse.de>
3585 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3586 explicit single_use checks.
3588 2018-03-27 Richard Biener <rguenther@suse.de>
3590 PR tree-optimization/85082
3591 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3594 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3596 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3597 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3598 Turn on fasynchronous-unwind-tables and funwind-tables.
3600 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3603 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3604 (*bmi_blsr_<mode>_ccz): Ditto.
3606 2018-03-26 Tom de Vries <tom@codesourcery.com>
3608 PR tree-optimization/85063
3609 * omp-general.c (offloading_function_p): New function. Factor out
3611 * omp-offload.c (pass_omp_target_link::gate): ... here.
3612 * omp-general.h (offloading_function_p): Declare.
3613 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3614 with attribute omp declare target for offloading functions.
3616 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3618 PR tree-optimization/84005
3619 * tree-data-ref.h (get_base_for_alignment): Declare.
3620 * tree-data-ref.c (get_base_for_alignment_1): New function.
3621 (get_base_for_alignment): Likewise.
3622 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3623 get_base_for_alignment to find a suitable base object, instead
3624 of always using drb->base_address.
3626 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3629 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3630 known size by default.
3632 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3635 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3636 and non BLKmode operands.
3638 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3641 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3642 Clean up attributes.
3644 2018-03-23 Richard Biener <rguenther@suse.de>
3647 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3648 we are going to emit early debug for LTO.
3650 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3653 * function.c (match_asm_constraints_1): Don't optimize if input
3654 doesn't satisfy general_operand predicate for output's mode.
3657 * alias.c (write_dependence_p): Don't require for x_canonicalized
3658 non-VOIDmode if x has VOIDmode.
3661 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3662 just don't try to optimize it rather than assert it never happens.
3664 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3666 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3667 macro expansions for definition of ST_INTERNAL_<mode> and
3668 LD_INTERNAL_<mode> builtins.
3669 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3671 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3673 (altivec_expand_st_builtin): Likewise.
3674 (altivec_expand_builtin): Remove calls to deleted functions.
3675 (rs6000_address_for_altivec): Delete this function.
3676 * config/rs6000/vector.md: Remove expands for
3677 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3679 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3682 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3683 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3684 re-computing once computed.
3685 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3686 (arm_init_machine_status): Initialize
3687 machine->static_chain_stack_bytes.
3689 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3692 * doc/extend.texi: Add four new prototypes for vec_ld.
3693 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3694 definitions for more logical presentation.
3695 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3696 entries for V1TI variants of __builtin_altivec_ld builtin.
3697 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3698 handling of V1TI variant of LVX icode pattern.
3699 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3700 (rs6000_gimple_fold_builtin): Likewise.
3701 (altivec_init_builtins): Add code to define
3702 __builtin_altivec_lvx_v1ti function.
3704 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3707 * function.c (match_asm_constraints_1): Don't do the optimization
3708 if input isn't a REG, SUBREG, MEM or constant.
3710 2018-03-22 Tom de Vries <tom@codesourcery.com>
3712 PR tree-optimization/84956
3713 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3714 bb_has_abnormal_pred.
3716 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3719 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3720 DECL_INITIAL (decl) to decl at the end.
3721 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3724 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3726 * doc/extend.texi (__builtin_tgmath): Document when complex
3727 integer types are treated as _Complex _Float64.
3729 2018-03-21 Tom de Vries <tom@codesourcery.com>
3731 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3733 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3735 PR tree-optimization/84960
3736 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3737 if it is ENTRY block, move them into single succ of ENTRY in that case.
3739 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3741 PR tree-optimization/84811
3742 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3743 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3744 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3745 (known_subrange_p): Update accordingly. Cast each value involved
3746 in the size comparison, rather than casting the result of the
3749 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3751 PR tree-optimization/84982
3752 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3753 by flipping the least significant bit rather than all bits from
3754 bitpos to bitpos + bitsize - 1.
3756 2018-03-21 Nathan Sidwell <nathan@acm.org>
3758 * doc/extend.texi (Deprecated Features): Remove mention of
3759 long-deleted deprecations.
3761 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3764 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3765 * configure: Regenerate.
3767 2018-03-21 Tom de Vries <tom@codesourcery.com>
3769 PR tree-optimization/83126
3770 * tree-parloops.c (num_phis): New function.
3771 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3773 2018-03-21 Nathan Sidwell <nathan@acm.org>
3775 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3776 mention anon-struct/union members and trailing attributes.
3778 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3780 PR tree-optimization/84969
3781 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3782 builtin memset partitions if they set different rhs values.
3784 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3786 PR rtl-optimization/84989
3787 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3788 VEC_DUPLICATE with scalar result mode.
3790 2018-03-21 Martin Liska <mliska@suse.cz>
3793 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3794 not intended return statement.
3796 2018-03-21 Martin Liska <mliska@suse.cz>
3799 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3800 (chkp_find_bound_slots_1): Limit number of iterations.
3802 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3805 * Minor grammar fixes for x86 options.
3807 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3810 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3811 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3813 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3816 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3817 (altivec_lvx_<mode>_1op): Likewise.
3818 (altivec_stvx_<mode>_2op): Likewise.
3819 (altivec_stvx_<mode>_1op): Likewise.
3820 (altivec_lvx_<VM2:mode>): New define_expand.
3821 (altivec_stvx_<VM2:mode>): Likewise.
3822 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3823 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3824 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3825 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3826 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3827 (rs6000_gen_lvx): Likewise.
3828 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3829 (altivec_expand_stv_builtin): Likewise.
3830 (altivec_expand_builtin): Likewise.
3831 * config/rs6000/vector.md: Likewise.
3833 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3836 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3839 2018-03-20 Richard Biener <rguenther@suse.de>
3842 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3843 sign-conversions as zero, fall back to standard scalar_stmt
3846 2018-03-20 Martin Liska <mliska@suse.cz>
3849 * predict.c (rebuild_frequencies): Handle case when we have
3850 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3852 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3855 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3856 flag_section_anchors.
3857 * varasm.c (use_blocks_for_decl_p): Remove hack for
3858 dw2_force_const_mem.
3861 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3863 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3864 be created, use lowpart_subreg of operands[0] rather than operands[0]
3866 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3867 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3868 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3869 and n constraint instead of aarch64_shift_imm_di and Usd.
3870 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3871 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3873 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3876 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3877 to favor GPR over NEON registers.
3878 (<shift>di3_neon): Likewise.
3880 2018-03-20 Tom de Vries <tom@codesourcery.com>
3883 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3884 (nvptx_process_pars): Emit bar.sync asap and alap.
3886 2018-03-20 Tom de Vries <tom@codesourcery.com>
3889 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3890 seen_label if seen_label is already set.
3892 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3895 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3896 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3897 Use 1U instead of 1. Formatting fixes.
3900 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3901 instead of TREE_TYPE (s1) for the return value.
3903 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3905 PR tree-optimization/84946
3906 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3907 bitsize + bitsize in poly_uint64 rather than poly_int64.
3910 * dwarf2asm.c: Include fold-const.c.
3911 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3912 of decl rather than decl itself.
3914 PR rtl-optimization/84643
3915 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3917 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3920 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3921 calling assemble_variable.
3923 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3926 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3927 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3929 2018-03-19 Jim Wilson <jimw@sifive.com>
3932 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3933 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3934 (riscv_first_stack_step): Likewise.
3935 (riscv_option_override): Use STACK_BOUNDARY instead of
3937 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3939 (BIGGEST_ALIGNMENT): Set to 128.
3940 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3941 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3944 2018-03-19 Richard Biener <rguenther@suse.de>
3946 PR tree-optimization/84933
3947 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3948 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3950 2018-03-19 Richard Biener <rguenther@suse.de>
3952 PR tree-optimization/84859
3953 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3954 (cond_if_else_store_replacement): Perform sinking operation on
3955 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3956 Generalize what a BB with a single eligible store is.
3958 2018-03-19 Richard Biener <rguenther@suse.de>
3960 PR tree-optimization/84929
3961 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3962 chrec_is_positive against non-chrec arg.
3964 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3967 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3969 2018-03-18 Martin Liska <mliska@suse.cz>
3971 PR rtl-optimization/84635
3972 * regrename.c (build_def_use): Use matches_mode only when
3975 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3977 PR tree-optimization/84913
3978 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3979 vectorize chains of COND_EXPRs.
3981 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3983 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3985 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3987 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3989 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3991 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3993 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3994 Kito Cheng <kito.cheng@gmail.com>
3996 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3997 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3998 (nds32_adjust_reg_alloc_order): New function.
3999 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4001 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4003 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4004 nds32_print_operand, nds32_print_operand_address): Use
4005 HOST_WIDE_INT_PRINT_DEC instead.
4007 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4009 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4011 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4014 * config/i386/i386.c (initial_ix86_tune_features,
4015 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4017 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4018 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4019 rather than 1u << ix86_tune. Formatting fix.
4020 (ix86_option_override_internal): Change ix86_arch_mask from
4021 unsigned int to unsigned HOST_WIDE_INT, initialize to
4022 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4023 (ix86_function_specific_restore): Likewise.
4025 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4028 * postreload.c (reload_combine_recognize_pattern): Perform
4029 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4030 truncate_int_for_mode the result for the destination's mode.
4033 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4034 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4036 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4039 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4042 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4043 regno_allocno_class_array and sorted_pseudos.
4044 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4045 insns where regno is used.
4047 2018-03-16 Martin Liska <mliska@suse.cz>
4050 * multiple_target.c (create_dispatcher_calls): Redirect
4051 reference in the symbol table.
4053 2018-03-16 Martin Liska <mliska@suse.cz>
4056 * multiple_target.c (create_dispatcher_calls): Redirect also
4059 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4063 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4064 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4065 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4067 2018-03-16 Julia Koval <julia.koval@intel.com>
4069 * doc/invoke.texi (Skylake Server): Add CLWB.
4070 Cannonlake): Remove CLWB.
4072 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4074 PR tree-optimization/84841
4075 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4077 (FLOAT_ONE_CONST_TYPE): Define.
4078 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4079 (sort_by_operand_rank): Put entries with higher constant_type last
4080 rather than first to match comments.
4082 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4084 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4087 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4090 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4091 check and use address of target always.
4093 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4096 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4097 (indirect_thunk_bnd_needed): Likewise.
4098 (indirect_thunks_used): Likewise.
4099 (indirect_thunks_bnd_used): Likewise.
4100 (indirect_return_needed): New.
4101 (indirect_return_bnd_needed): Likewise.
4102 (output_indirect_thunk_function): Add a bool argument for
4104 (output_indirect_thunk_function): Don't generate alias for
4105 function return thunk.
4106 (ix86_code_end): Call output_indirect_thunk_function to generate
4107 function return thunks.
4108 (ix86_output_function_return): Set indirect_return_bnd_needed
4109 and indirect_return_needed instead of indirect_thunk_bnd_needed
4110 and indirect_thunk_needed.
4112 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4114 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4115 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4116 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4118 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4119 Paul Hua <paul.hua.gm@gmail.com>
4122 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4124 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4126 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4127 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4130 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4133 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4134 instead of GET_MODE_SIZE when comparing Units.
4136 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4139 * varasm.c (hash_section): Return an unchangeble hash value
4140 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4141 Return !aarch64_can_use_per_function_literal_pools_p ().
4143 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4146 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4147 rather than address of cmode as last argument to prepare_cmp_insn.
4149 2018-03-15 Julia Koval <julia.koval@intel.com>
4151 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4152 F_AVX512VNNI, F_AVX512BITALG): New.
4154 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4157 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4158 insn for floating-point loads and stores.
4160 2018-03-14 Carl Love <cel@us.ibm.com>
4162 * config/rs6000/rs6000-c.c: Add macro definitions for
4163 ALTIVEC_BUILTIN_VEC_PERMXOR.
4164 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4165 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4166 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4167 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4169 * config/doc/extend.texi: Add prototypes for vec_permxor.
4171 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4174 * diagnostic-show-locus.c (class layout_point): Convert m_line
4175 from int to linenum_type.
4176 (line_span::comparator): Use linenum "compare" function when
4177 comparing line numbers.
4178 (test_line_span): New function.
4179 (layout_range::contains_point): Convert param "row" from int to
4181 (layout_range::intersects_line_p): Likewise.
4182 (layout::will_show_line_p): Likewise.
4183 (layout::print_source_line): Likewise.
4184 (layout::should_print_annotation_line_p): Likewise.
4185 (layout::print_annotation_line): Likewise.
4186 (layout::print_leading_fixits): Likewise.
4187 (layout::annotation_line_showed_range_p): Likewise.
4188 (struct line_corrections): Likewise for field m_row.
4189 (line_corrections::line_corrections): Likewise for param "row".
4190 (layout::print_trailing_fixits): Likewise.
4191 (layout::get_state_at_point): Likewise.
4192 (layout::get_x_bound_for_row): Likewise.
4193 (layout::print_line): Likewise.
4194 (diagnostic_show_locus): Likewise for locals "last_line" and
4196 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4197 * input.c (selftest::test_linenum_comparisons): New function.
4198 (selftest::input_c_tests): Call it.
4199 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4200 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4201 * selftest.h (ASSERT_GT): New macro.
4202 (ASSERT_GT_AT): New macro.
4203 (ASSERT_LT): New macro.
4204 (ASSERT_LT_AT): New macro.
4206 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4208 PR rtl-optimization/84780
4209 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4211 2018-03-14 Martin Liska <mliska@suse.cz>
4213 * tree.c (record_node_allocation_statistics): Use
4214 get_stats_node_kind.
4215 (get_stats_node_kind): New function extracted from
4216 record_node_allocation_statistics.
4217 (free_node): Use get_stats_node_kind.
4219 2018-03-14 Richard Biener <rguenther@suse.de>
4221 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4222 that the value-set of ANTIC_IN doesn't grow.
4225 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4227 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4228 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4230 2018-03-14 Julia Koval <julia.koval@intel.com>
4232 * config.gcc (icelake-client, icelake-server): New.
4234 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4235 (initial_ix86_arch_features): Ditto.
4236 (PTA_SKYLAKE): Add SGX.
4237 (PTA_ICELAKE): Remove.
4238 (PTA_ICELAKE_CLIENT): New.
4239 (PTA_ICELAKE_SERVER): New.
4240 (ix86_option_override_internal): Split up icelake on icelake client and
4242 (get_builtin_code_for_version): Ditto.
4243 (fold_builtin_cpu): Ditto.
4244 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4245 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4246 * config/i386/i386.h (processor_type): Ditto.
4247 * doc/invoke.texi: Ditto.
4249 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4252 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4253 INTEGER_CST offset, add it together with bitpos / 8 and
4254 sign extend based on POINTER_SIZE.
4258 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4261 * config/i386/constraints.md (Yc): New register constraint.
4262 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4263 Use Yc constraint for alternative 2 of operand 0. Remove
4264 preferred_for_speed attribute.
4266 2018-03-14 Richard Biener <rguenther@suse.de>
4268 PR tree-optimization/84830
4269 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4270 with the old one to avoid oscillations.
4272 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4275 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4277 (assign_by_spills): Return a flag of reload assignment failure.
4278 Do not process the reload assignment failures. Do not spill other
4279 reload pseudos if they has the same reg class. Update n if
4281 (lra_assign): Add a return arg. Set up from the result of
4282 assign_by_spills call.
4283 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4284 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4285 usage_insns if it is not NULL.
4286 (spill_hard_reg_in_range): New function.
4287 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4288 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4289 function prototypes.
4290 (lra_assign): Change prototype.
4291 * lra.c (lra): Add code to deal with fails by splitting hard reg
4294 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4296 * config/riscv/riscv.opt (mrelax): New option.
4297 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4298 "norelax" when riscv_mrelax is disabled.
4299 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4301 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4304 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4305 reassociation for int modes.
4307 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4309 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4310 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4312 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4313 * config/aarch64/aarch64-sve.md
4314 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4315 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4316 (*extend<mode><Vwide>2): Rename to...
4317 (aarch64_sve_extend<mode><Vwide>2): ...this.
4318 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4319 renaming the old pattern to...
4320 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4322 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4323 define_expand, renaming the old pattern to...
4324 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4325 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4326 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4327 account when deciding which SVE instruction the optab should use.
4328 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4330 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4332 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4333 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4334 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4335 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4336 (tlsdesc_small_<mode>): Turn a define_expand and use
4337 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4338 (tlsdesc_small_advsimd_<mode>): ...this.
4339 (tlsdesc_small_sve_<mode>): New pattern.
4341 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4343 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4344 (UNSPEC_UMUL_HIGHPART): New constants.
4345 (MUL_HIGHPART): New int iteraor.
4346 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4347 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4349 (*<su>mul<mode>3_highpart): New define_insn.
4351 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4354 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4357 2018-03-13 Martin Liska <mliska@suse.cz>
4360 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4362 (sem_item_optimizer::~sem_item_optimizer): Release it.
4363 (sem_item_optimizer::merge_classes): Register variable aliases.
4364 (sem_item_optimizer::fixup_pt_set): New function.
4365 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4366 * ipa-icf.h: Declare new variables and functions.
4368 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4371 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4372 integer_pow2p@2 and test integer_pow2p in condition.
4373 (A < 0 ? C : 0): Similarly for @1.
4376 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4377 characters starting at p contain '\0' character, don't look beyond
4381 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4382 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4385 * reg-stack.c (change_stack): Change update_end var from int to
4386 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4387 also call set_block_for_insn on the newly added insns and rescan.
4390 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4391 on the last operand.
4394 * tree.c (stabilize_reference_1): Return save_expr (e) for
4395 STATEMENT_LIST even if it doesn't have side-effects.
4397 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4399 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4401 2018-03-12 Renlin Li <renlin.li@arm.com>
4403 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4404 aarch64_output_scalar_simd_mov_immediate.
4406 2018-03-12 Martin Sebor <msebor@redhat.com>
4408 PR tree-optimization/83456
4409 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4410 for perfectly overlapping calls to memcpy.
4411 (gimple_fold_builtin_memory_chk): Same.
4412 (gimple_fold_builtin_strcpy): Handle no-warning.
4413 (gimple_fold_builtin_stxcpy_chk): Same.
4414 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4416 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4418 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4419 parameter. Use it for SFmode.
4420 (rs6000_function_arg_advance_1): Adjust.
4421 (rs6000_function_arg): Adjust.
4422 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4424 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4426 PR rtl-optimization/84169
4427 PR rtl-optimization/84780
4428 * combine.c (can_combine_p): Check for a 2-insn combination whether
4429 the destination register is used between the two insns, too.
4431 2018-03-12 Richard Biener <rguenther@suse.de>
4433 PR tree-optimization/84803
4434 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4435 for refs DR analysis didn't process.
4437 2018-03-12 Richard Biener <rguenther@suse.de>
4439 PR tree-optimization/84777
4440 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4441 force-vectorize loops ignore whether we are optimizing for size.
4443 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4445 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4446 (TARGET_MD_ASM_ADJUST): Define.
4448 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4449 Kito Cheng <kito.cheng@gmail.com>
4450 Chung-Ju Wu <jasonwucj@gmail.com>
4452 * config/nds32/nds32.c (nds32_compute_stack_frame,
4453 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4454 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4455 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4456 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4457 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4458 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4459 * config/nds32/nds32.md (prologue, epilogue): Use macro
4460 NDS32_V3PUSH_AVAILABLE_P to do checking.
4462 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4465 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4466 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4467 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4468 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4469 addition of most attributes on !orig_type_die or the attribute not
4470 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4472 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4473 Chung-Ju Wu <jasonwucj@gmail.com>
4475 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4477 * config/nds32/nds32.opt (mvh): New option.
4479 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4480 Chung-Ju Wu <jasonwucj@gmail.com>
4482 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4484 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4485 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4488 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4489 Chung-Ju Wu <jasonwucj@gmail.com>
4491 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4493 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4494 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4496 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4497 Kito Cheng <kito.cheng@gmail.com>
4498 Chung-Ju Wu <jasonwucj@gmail.com>
4500 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4501 UNSPEC_FFMISM and UNSPEC_FLMISM.
4502 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4503 for ffb, ffmism and flmism.
4504 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4505 (unspec_ffmism): Ditto.
4506 (unspec_flmism): Ditto.
4507 (nds32_expand_builtin_impl): Check if string extension is available.
4508 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4509 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4511 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4514 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4517 * lra-assigns.c (assign_by_spills): Return a flag of reload
4518 assignment failure. Do not process the reload assignment
4519 failures. Do not spill other reload pseudos if they has the same
4521 (lra_assign): Add a return arg. Set up from the result of
4522 assign_by_spills call.
4523 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4524 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4525 usage_insns if it is not NULL.
4526 (spill_hard_reg_in_range): New function.
4527 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4528 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4529 function prototypes.
4530 (lra_assign): Change prototype.
4531 * lra.c (lra): Add code to deal with fails by splitting hard reg
4534 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4537 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4539 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4542 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4543 (dw_val_node): Add val_symbolic_view.
4544 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4545 (symview_upper_bound): New.
4546 (new_line_info_table): Initialize symviews_since_reset.
4547 (dwarf2out_source_line): Count symviews_since_reset and set
4548 symview_upper_bound.
4549 (dw_val_equal_p): Handle symview.
4550 (add_AT_symview): New.
4551 (print_dw_val): Handle symview.
4552 (attr_checksum, attr_checksum_ordered): Likewise.
4553 (same_dw_val_p, size_of_die): Likewise.
4554 (value_format, output_die): Likewise.
4555 (add_high_low_attributes): Use add_AT_symview for entry_view.
4556 (dwarf2out_finish): Reset symview_upper_bound, clear
4559 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4562 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4563 Add strict argument and use it.
4564 (rs6000_split_multireg_move): Update for new strict argument.
4565 (mem_operand_gpr): Disallow all non-offsettable addresses.
4566 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4568 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4571 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4572 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4573 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4576 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4577 decl, use remap_type if we want to use the type.
4579 2018-03-09 Martin Sebor <msebor@redhat.com>
4581 PR tree-optimization/84526
4582 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4584 (builtin_access::generic_overlap): Be prepared to handle non-array
4587 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4589 PR rtl-optimization/84682
4590 * lra-constraints.c (process_address_1): Check is_address flag
4591 for address constraints.
4592 (process_alt_operands): Likewise.
4593 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4594 preprocess_constraints.
4595 * recog.h (preprocess_constraints): Add oploc parameter.
4597 * recog.c (preprocess_constraints): Test address_operand for
4598 CT_ADDRESS constraints.
4600 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4603 * lra-assigns.c (assign_by_spills): Return a flag of reload
4604 assignment failure. Do not process the reload assignment
4605 failures. Do not spill other reload pseudos if they has the same
4607 (lra_assign): Add a return arg. Set up from the result of
4608 assign_by_spills call.
4609 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4610 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4611 usage_insns if it is not NULL.
4612 (spill_hard_reg_in_range): New function.
4613 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4614 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4615 function prototypes.
4616 (lra_assign): Change prototype.
4617 * lra.c (lra): Add code to deal with fails by splitting hard reg
4620 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4623 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4624 Accept complain bool parameter. Only emit errors if it is true.
4625 (arm_parse_cpu_option_name): Likewise.
4626 (arm_target_thumb_only): Adjust callers of the above.
4627 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4628 prototype to take a default true bool parameter.
4629 (arm_parse_arch_option_name): Likewise.
4631 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4632 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4636 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4637 * configure: Regenerate.
4638 * configure.ac ("linker --version-script option"): New.
4639 ("linker soname option"): New.
4641 2018-03-09 Richard Biener <rguenther@suse.de>
4643 PR tree-optimization/84775
4644 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4645 immediate uses of predicate stmts and mark them modified.
4648 PR tree-optimization/84178
4649 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4651 (version_loop_for_if_conversion): Delay update_ssa call.
4652 (tree_if_conversion): Delay update_ssa until after predicate
4655 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4658 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4659 when the function accesses prior frames.
4661 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4664 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4665 gen_llsym, otherwise call maybe_gen_llsym.
4668 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4669 has ',' character inside of it.
4671 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4674 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4675 as clobbering CC_REGNUM.
4677 2018-03-08 Richard Biener <rguenther@suse.de>
4680 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4681 (follow_copies_to_constant): Do not follow SSA names registered
4684 2018-03-08 Richard Biener <rguenther@suse.de>
4686 PR tree-optimization/84178
4687 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4689 (version_loop_for_if_conversion): Delay update_ssa call.
4690 (tree_if_conversion): Delay update_ssa until after predicate
4693 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4695 PR tree-optimization/84178
4696 * tree-if-conv.c (release_bb_predicate): Remove the
4697 the assertion that the stmts have NULL use_ops.
4698 Discard the statements, asserting that they haven't
4699 yet been added to a BB.
4701 2018-03-08 Richard Biener <rguenther@suse.de>
4703 PR tree-optimization/84746
4704 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4705 (phi_translate): Pass in destination ANTIC_OUT set.
4706 (phi_translate_1): Likewise. For a simplified result lookup
4707 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4708 (phi_translate_set): Adjust.
4709 (do_pre_regular_insertion): Likewise.
4710 (do_pre_partial_partial_insertion): Likewise.
4712 2018-03-08 Martin Liska <mliska@suse.cz>
4714 PR gcov-profile/84735
4715 * doc/gcov.texi: Document usage of profile files.
4716 * gcov-io.h: Document changes in the format.
4718 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4722 * dwarf2out.c (struct dw_line_info_table): Update comments for
4724 (FORCE_RESET_NEXT_VIEW): New.
4725 (FORCE_RESETTING_VIEW_P): New.
4726 (RESETTING_VIEW_P): Check for -1 too.
4727 (ZERO_VIEW_P): Likewise.
4728 (new_line_info_table): Force-reset next view.
4729 (dwarf2out_begin_function): Likewise.
4730 (dwarf2out_source_line): Simplify zero_view_p initialization.
4731 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4732 view directly. Omit view when omitting .loc at line 0.
4734 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4736 PR tree-optimization/84740
4737 * tree-switch-conversion.c (process_switch): Call build_constructors
4738 only if info.phi_count is non-zero.
4740 PR tree-optimization/84739
4741 * tree-tailcall.c (find_tail_calls): Check call arguments against
4742 DECL_ARGUMENTS (current_function_decl) rather than
4743 DECL_ARGUMENTS (func) when checking for tail recursion.
4745 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4747 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4748 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4749 Volker Reichelt's entry and add entries for people that perform
4750 GCC fuzzy testing and report numerous bugs.
4752 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4755 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4756 readonly data in sdata, if that is disabled.
4757 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4758 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4759 -mreadonly-in-sdata option.
4761 2018-03-07 Martin Sebor <msebor@redhat.com>
4763 PR tree-optimization/84468
4764 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4765 basic block when looking for nul assignment.
4767 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4770 * except.h (output_function_exception_table): Adjust prototype.
4771 * except.c (output_function_exception_table): Remove FNNAME parameter
4772 and add SECTION parameter. Ouput one part of the table at a time.
4773 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4774 the first part of the exception table and emit unwind directives.
4775 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4776 (i386_pe_seh_cold_init): Likewise.
4777 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4778 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4779 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4780 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4781 * config/i386/winnt.c: Include except.h.
4782 (struct seh_frame_state): Add reg_offset, after_prologue and
4783 in_cold_section fields.
4784 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4785 (i386_pe_seh_cold_init): New function.
4786 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4787 to seh->in_cold_section.
4788 (seh_emit_push): Record the offset of the push.
4789 (seh_emit_save): Record the offet of the save.
4790 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4791 Test seh->after_prologue to disregard the epilogue.
4792 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4793 (i386_pe_end_cold_function): New function.
4795 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4798 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4799 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4802 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4804 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4805 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4808 * multiple_target.c: Include tree-inline.h and intl.h.
4809 (expand_target_clones): Diagnose and fail if node->definition and
4810 !tree_versionable_function_p (node->decl).
4812 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4814 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4816 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4817 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4818 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4820 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4823 * combine.c (try_combine): Use reg_or_subregno instead of handling
4824 just paradoxical SUBREGs and REGs.
4826 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4828 * config/arc/arc.c (arc_finalize_pic): Remove function.
4829 (arc_must_save_register): We use single base PIC register, remove
4830 checks to save/restore the PIC register.
4831 (arc_expand_prologue): Likewise.
4832 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4834 (arc_verify_short): Likewise.
4835 (arc_attr_type): Likewise.
4836 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4837 (walk_stores): Likewise.
4838 (arc_address_cost): Make it static.
4839 (arc_verify_short): Likewise.
4840 (branch_dest): Likewise.
4841 (arc_attr_type): Likewise.
4842 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4843 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4844 (arc_final_prescan_insn): Remove inserting the nops due to
4845 hardware hazards. It is done in reorg step.
4846 (insn_length_variant_t): Remove.
4847 (insn_length_parameters_t): Likewise.
4848 (arc_insn_length_parameters): Likewise.
4849 (arc_get_insn_variants): Likewise.
4850 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4852 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4855 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4858 PR tree-optimization/84687
4859 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4861 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4863 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4865 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4866 Rename to ppc_speculation_barrier.
4867 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4868 __builtin_ppc_speculation_barrier.
4870 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4873 * combine.c (combine_simplify_rtx): Don't try to simplify if
4874 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4877 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4879 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4880 to 32 bytes when compiling for POWER9.
4882 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4885 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4886 regparm >= 3 with no arg reg available also for calls with
4887 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4890 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4892 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4894 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4897 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4899 2018-03-05 Richard Biener <rguenther@suse.de>
4901 PR tree-optimization/84486
4902 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4903 When inserting a __builtin_assume_aligned call set the LHS
4904 SSA name alignment info accordingly.
4906 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4908 PR tree-optimization/84114
4909 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4910 Avoid reassociation of FLOAT_MODE addition.
4912 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4914 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4915 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4916 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4917 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4918 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4919 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4920 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4922 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4923 __builtin_ia32_wbinvd): New builtins.
4924 (SPECIAL_ARGS2): New.
4925 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4926 (SPECIAL_ARGS2): New.
4927 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4928 (ix86_valid_target_attribute_inner_p): Ditto.
4929 (ix86_init_mmx_sse_builtins): Add special_args2.
4930 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4931 TARGET_WBNOINVD_P): New.
4932 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4933 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4934 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4935 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4936 * config/i386/pconfigintrin.h: New file.
4937 * config/i386/wbnoinvdintrin.h: Ditto.
4938 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4939 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4941 2018-03-05 Richard Biener <rguenther@suse.de>
4943 PR tree-optimization/84670
4944 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4946 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4947 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4948 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4949 doesn't grow if all successors (recursively) were visited at least
4952 2018-03-05 Richard Biener <rguenther@suse.de>
4954 PR tree-optimization/84650
4955 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4956 if executed in the loop pipeline.
4958 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4960 * doc/configfiles.texi (Configuration Files): Move info about
4961 conditionalizing $target-protos.h to...
4962 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4963 differs from $target-protos.h.
4965 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4966 Chung-Ju Wu <jasonwucj@gmail.com>
4968 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4969 * config/nds32/nds32-multiple.md (setmemsi): Define.
4970 * config/nds32/nds32-memory-manipulation.c
4971 (nds32_gen_dup_4_byte_to_word_value): New.
4972 (emit_setmem_word_loop): New.
4973 (emit_setmem_byte_loop): New.
4974 (nds32_expand_setmem_loop): New.
4975 (nds32_expand_setmem_loop_v3m): New.
4976 (nds32_expand_setmem_unroll): New.
4977 (nds32_expand_setmem): New.
4979 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4980 Chung-Ju Wu <jasonwucj@gmail.com>
4982 * config/nds32/nds32-memory-manipulation.c
4983 (nds32_emit_load_store): New.
4984 (nds32_emit_post_inc_load_store): New.
4985 (nds32_emit_mem_move): New.
4986 (nds32_emit_mem_move_block): New.
4987 (nds32_expand_movmemsi_loop_unknown_size): New.
4988 (nds32_expand_movmemsi_loop_known_size): New.
4989 (nds32_expand_movmemsi_loop): New.
4990 (nds32_expand_movmemsi_unroll): New.
4991 (nds32_expand_movmemqi): Rename ...
4992 (nds32_expand_movmemsi): ... to this.
4993 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4994 (movmemsi): ... to this.
4995 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4996 (nds32_expand_movmemsi): ... to this.
4998 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4999 Monk Chiang <sh.chiang04@gmail.com>
5000 Chung-Ju Wu <jasonwucj@gmail.com>
5002 * config/nds32/nds32-protos.h
5003 (nds32_expand_load_multiple): New arguments.
5004 (nds32_expand_store_multiple): Ditto.
5005 (nds32_valid_multiple_load_store): Rename ...
5006 (nds32_valid_multiple_load_store_p): ... to this.
5007 * config/nds32/nds32-memory-manipulation.c
5008 (nds32_expand_load_multiple): Refine implementation.
5009 (nds32_expand_store_multiple): Ditto.
5010 * config/nds32/nds32-multiple.md
5011 (load_multiple): Update nds32_expand_load_multiple interface.
5012 (store_multiple): Update nds32_expand_store_multiple interface.
5013 * config/nds32/nds32-predicates.c
5014 (nds32_valid_multiple_load_store): Rename ...
5015 (nds32_valid_multiple_load_store_p): ... to this and refine
5017 * config/nds32/predicates.md
5018 (nds32_load_multiple_and_update_address_operation): New predicate.
5019 (nds32_store_multiple_and_update_address_operation): New predicate.
5021 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5022 Chung-Ju Wu <jasonwucj@gmail.com>
5024 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5025 (combo): New attribute.
5026 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5028 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5030 * config/nds32/nds32.opt: Change -mcmodel= default value.
5032 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5033 Monk Chiang <sh.chiang04@gmail.com>
5034 Chung-Ju Wu <jasonwucj@gmail.com>
5036 * config/nds32/constants.md (unspec_element): New enum.
5037 * config/nds32/constraints.md (Umw): New constraint.
5038 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5039 * config/nds32/nds32-intrinsic.md: Likewise.
5040 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5041 (nds32_valid_smw_lwm_base_p): New.
5042 (nds32_output_smw_single_word): New.
5043 (nds32_output_lmw_single_word): New.
5044 (nds32_expand_unaligned_load): New.
5045 (nds32_expand_unaligned_store): New.
5046 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5047 (nds32_output_smw_single_word): Declare.
5048 (nds32_output_lmw_single_word): Declare.
5049 (nds32_expand_unaligned_load): Declare.
5050 (nds32_expand_unaligned_store): Declare.
5051 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5052 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5053 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5054 NDS32_BUILTIN_UASTORE_DW.
5055 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5058 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5059 Kito Cheng <kito.cheng@gmail.com>
5060 Chung-Ju Wu <jasonwucj@gmail.com>
5062 * config/nds32/nds32-intrinsic.c
5063 (nds32_expand_builtin_null_ftype_reg): Delete.
5064 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5065 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5066 (nds32_read_argument): New.
5067 (nds32_legitimize_target): Ditto.
5068 (nds32_legitimize_argument): Ditto.
5069 (nds32_check_constant_argument): Ditto.
5070 (nds32_expand_unop_builtin): Ditto.
5071 (nds32_expand_unopimm_builtin): Ditto.
5072 (nds32_expand_binop_builtin): Ditto.
5073 (nds32_builtin_decl_impl): Ditto.
5074 (builtin_description): Ditto.
5075 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5076 (nds32_init_builtins_impl): Ditto.
5077 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5078 (nds32_builtin_decl): New.
5079 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5080 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5082 2018-03-02 Jeff Law <law@redhat.com>
5084 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5085 (redundant_insn, fill_simple_delay_slots): Likewise.
5086 (fill_slots_from_thread): Likewise.
5087 * resource.c (mark_referenced_resources): Likewise.
5088 (mark_set_resources, find_dead_or_set_registers): Likewise.
5090 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5092 * substring-locations.h (format_warning_va): Formatting fix for
5094 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5096 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5097 * substring-locations.c: Include intl.h.
5098 (format_warning_va): Turned into small wrapper around
5099 format_warning_n_va, renamed to ...
5100 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5101 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5103 (format_warning_at_substring_n): New function.
5104 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5105 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5106 format_warning_at_substring with just a shorter name instead of
5107 const function pointer.
5108 (fmtwarn_n): New function.
5109 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5110 appropriate, get rid of all the fmtstr temporaries, move conditionals
5111 with G_() wrapped string literals directly into fmtwarn arguments,
5112 cast dir.len to (int), formatting fixes.
5114 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5116 * doc/invoke.texi: Remove "Cilk Plus" references.
5118 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5119 Richard Biener <rguenther@suse.de>
5122 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5123 for error or warning attributes if CALL_FROM_THUNK_P is set.
5126 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5129 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5130 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5133 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5134 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5136 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5137 instead of -1U in last predictors element's probability member.
5139 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5142 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5143 arguments if they are comparable.
5145 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5147 PR tree-optimization/84634
5148 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5149 masks and masked_loop_p with a single loop_masks, making sure it's
5150 null for bb vectorization.
5152 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5154 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5155 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5156 loop->force_vectorize to check whether there is no alias.
5158 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5161 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5163 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5165 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5166 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5167 instead of a loop around prev_real_insn.
5168 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5172 * config/i386/i386.c (ix86_print_operand): Use conditional
5173 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5176 2018-03-02 Richard Biener <rguenther@suse.de>
5178 PR tree-optimization/84427
5179 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5180 (bitmap_set_subtract_values): Rewrite to handle multiple
5183 (prune_clobbered_mems): Likewise.
5184 (phi_translate): Take edge instead of pred/phiblock.
5185 (phi_translate_1): Likewise.
5186 (phi_translate_set): Likewise. Insert all translated
5187 exprs for a value into the set, keeping possibly multiple
5188 expressions per value.
5189 (compute_antic_aux): Adjust for phi_translate changes.
5190 When intersecting union the expressions and prune those
5191 not in the final value set, keeping possibly multiple
5192 expressions per value. Do not use value-insertion
5193 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5194 all expressions. Add verification that the value-sets
5195 only shrink during iteration.
5196 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5197 (do_pre_regular_insertion): Likewise.
5198 (do_pre_partial_partial_insertion): Likewise.
5200 2018-03-02 Richard Biener <rguenther@suse.de>
5203 * config/darwin.c (saved_debug_info_level): New static global.
5204 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5205 (darwin_asm_lto_end): Restore debug info generation settings.
5207 2018-03-01 Martin Liska <mliska@suse.cz>
5210 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5213 2018-03-01 Richard Biener <rguenther@suse.de>
5216 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5217 type annotation with LTO when debug was disabled at compile-time.
5219 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5221 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5223 (mips_final_postscan_insn): Likewise.
5225 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5227 PR rtl-optimization/84528
5228 * alias.c (init_alias_target): Add commentary.
5229 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5230 a unique base value if the frame pointer is not eliminated
5231 to the stack pointer.
5233 2018-03-01 Tom de Vries <tom@codesourcery.com>
5235 PR rtl-optimization/83327
5236 * lra-int.h (hard_regs_spilled_into): Declare.
5237 * lra.c (hard_regs_spilled_into): Define.
5238 (init_reg_info): Init hard_regs_spilled_into.
5239 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5240 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5241 (process_bb_lives): Handle hard_regs_spilled_into.
5242 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5243 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5245 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5247 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5248 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5249 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5250 * config/rs6000/aix72.h: New file.
5252 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5254 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5255 instead of warning_at with conditional singular and plural messages
5259 * stor-layout.c (update_alignment_for_field): For
5260 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5261 && !DECL_PACKED (field), do the alignment update, just use
5262 only desired_align instead of MAX (type_align, desired_align)
5264 (place_field): Don't do known_align < desired_align handling
5265 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5266 is non-NULL, instead do it after rli->prev_field handling and
5267 only if not within a bitfield word. For DECL_PACKED (field)
5268 use type_align of BITS_PER_UNIT.
5270 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5272 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5273 superfluous parentheses and trailing spaces.
5275 2018-02-28 Richard Biener <rguenther@suse.de>
5277 PR tree-optimization/84584
5278 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5279 SCoPs with fake exit edge.
5281 2018-02-28 Martin Liska <mliska@suse.cz>
5284 * timevar.c (timer::print): Fix format to properly print 100%
5287 2018-02-28 Richard Biener <rguenther@suse.de>
5290 * genmatch.c (capture_info::walk_match): Do not mark
5291 captured expressions without operands as expr_p given
5292 they act more like predicates and should be subject to
5293 "lost tail" side-effect preserving.
5295 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5297 PR rtl-optimization/81611
5298 * auto-inc-dec.c (attempt_change): Move dead note from
5299 mem_insn if it's the next use of regno
5300 (find_address): Take address use of reg holding
5301 non-incremented value. Add parm to limit search to the named
5303 (merge_in_block): Attempt to use a mem insn that is the next
5304 use of the original regno.
5306 2018-02-27 Martin Sebor <msebor@redhat.com>
5309 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5310 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5312 2018-02-27 Martin Sebor <msebor@redhat.com>
5314 PR translation/84207
5315 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5316 n argument to unsigned HOST_WIDE_INT.
5317 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5318 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5319 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5320 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5322 2018-02-27 Richard Biener <rguenther@suse.de>
5324 PR tree-optimization/84512
5325 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5326 Do not use the estimate returned from record_stmt_cost for
5327 the scalar iteration cost but sum properly using add_stmt_cost.
5329 2018-02-27 Richard Biener <rguenther@suse.de>
5331 PR tree-optimization/84466
5332 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5333 Adjust last change to less strictly validate use operands.
5335 2018-02-27 Martin Liska <mliska@suse.cz>
5337 PR gcov-profile/84548
5338 * gcov.c (process_file): Allow partial overlap and consider it
5339 also as group functions.
5340 (output_lines): Properly calculate range of lines for a group.
5342 2018-02-27 Martin Liska <mliska@suse.cz>
5344 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5345 'ggc' suffixes. Change first column width.
5346 (timer::print): Fix formatting of the column.
5348 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5350 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5351 preserve inline entry blocks for the sake of debug inline
5352 entry point markers alone.
5353 (remove_unused_locals): Suggest in comments a better place to
5354 force the preservation of inline entry blocks that are
5355 otherwise unused, but do not preserve them.
5357 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5359 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5361 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5364 * config/i386/constraints.md (Bs): Replace
5365 ix86_indirect_branch_register with
5366 TARGET_INDIRECT_BRANCH_REGISTER.
5368 * config/i386/i386.md (indirect_jump): Likewise.
5369 (tablejump): Likewise.
5370 (*sibcall_memory): Likewise.
5371 (*sibcall_value_memory): Likewise.
5372 Peepholes of indirect call and jump via memory: Likewise.
5373 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5374 (*sibcall_value_GOT_32): Likewise.
5375 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5376 (GOT_memory_operand): Likewise.
5377 (call_insn_operand): Likewise.
5378 (sibcall_insn_operand): Likewise.
5379 (GOT32_symbol_operand): Likewise.
5380 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5382 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5384 PR rtl-optimization/83496
5385 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5386 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5387 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5388 redundant insn, if any.
5389 (relax_delay_slots): Likewise.
5390 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5392 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5394 PR tree-optimization/83965
5395 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5396 that grouped statements are part of a reduction chain. Return
5397 true if the statement is not marked as a reduction itself but
5399 (vect_recog_dot_prod_pattern): Don't check whether the statement
5400 is part of a group here.
5401 (vect_recog_sad_pattern): Likewise.
5402 (vect_recog_widen_sum_pattern): Likewise.
5404 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5407 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5409 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5412 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5414 (ix86_output_indirect_function_return): New prototype.
5415 (ix86_split_simple_return_pop_internal): Likewise.
5416 * config/i386/i386.c (indirect_return_via_cx): New.
5417 (indirect_return_via_cx_bnd): Likewise.
5418 (indirect_thunk_name): Handle return va CX_REG.
5419 (output_indirect_thunk_function): Create alias for
5420 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5421 (ix86_output_indirect_jmp): Remove the bool argument.
5422 (ix86_output_indirect_function_return): New function.
5423 (ix86_split_simple_return_pop_internal): Likewise.
5424 * config/i386/i386.md (*indirect_jump): Don't pass false
5425 to ix86_output_indirect_jmp.
5426 (*tablejump_1): Likewise.
5427 (simple_return_pop_internal): Change it to define_insn_and_split.
5428 Call ix86_split_simple_return_pop_internal to split it for
5430 (simple_return_indirect_internal): Call
5431 ix86_output_indirect_function_return instead of
5432 ix86_output_indirect_jmp.
5434 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5437 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5438 memset and value initialization afterwards.
5440 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5442 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5444 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5447 * common/config/aarch64/aarch64-common.c
5448 (aarch_option_optimization_table[]): Switch
5449 off fomit-frame-pointer
5451 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5452 Chung-Ju Wu <jasonwucj@gmail.com>
5454 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5456 (store_multiple): Ditto.
5458 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5460 * config.gcc: Add --with-cpu support for nds32 target.
5461 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5462 * config/nds32/nds32.opt: Add -mcpu= option.
5464 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5466 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5467 isel=yes): Warn for these deprecated options.
5469 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5471 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5472 ISA_2_5_MASKS_EMBEDDED.
5474 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5476 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5477 p->max as pointers rather than using iterative_hash_expr.
5479 2018-02-23 Carl Love <cel@us.ibm.com>
5481 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5482 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5484 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5485 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5486 P8V_BUILTIN_VEC_VUNSIGNED2.
5488 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5491 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5492 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5494 * lra-constraints.c (curr_insn_transform): Set up
5495 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5497 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5498 * lra-eliminations.c (spill_pseudos): Ditto.
5499 (process_insn_for_elimination): Ditto.
5500 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5501 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5502 LRA_NON_CLOBBERED_ALT.
5504 2018-02-22 Martin Sebor <msebor@redhat.com>
5506 PR tree-optimization/84480
5507 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5508 to maybe_diag_stxncpy_trunc. Call it.
5509 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5510 from gimple_fold_builtin_strcpy. Print inlining stack.
5511 (handle_builtin_stxncpy): Print inlining stack.
5512 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5514 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5517 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5518 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5519 and -fcheck-pointer-bounds are used together.
5520 (indirect_thunk_prefix): New enum.
5521 (indirect_thunk_need_prefix): New function.
5522 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5523 "_nt" instead of "_bnd" for NOTRACK prefix.
5524 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5525 (output_indirect_thunk_function): Likewise.
5527 (ix86_code_end): Update output_indirect_thunk_function calls.
5528 (ix86_output_indirect_branch_via_reg): Replace
5529 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5530 (ix86_output_indirect_branch_via_push): Likewise.
5531 (ix86_output_function_return): Likewise.
5532 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5533 incompatible with -fcf-protection=branch and
5534 -fcheck-pointer-bounds.
5536 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5539 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5540 Change gcc_assert call to output_operand_lossage.
5542 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5544 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5546 2018-02-22 DJ Delorie <dj@redhat.com>
5547 Sebastian Perta <sebastian.perta@renesas.com>
5548 Oleg Endo <olegendo@gcc.gnu.org>
5550 * config/rx/rx.c (rx_rtx_costs): New function.
5551 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5553 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5555 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5557 2018-02-22 Martin Liska <mliska@suse.cz>
5560 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5561 Add "native" as a possible value.
5563 2018-02-22 Martin Liska <mliska@suse.cz>
5566 * config/i386/i386.c (ix86_option_override_internal):
5567 Add "native" as a possible value for -march and -mtune.
5569 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5572 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5573 to all type variants.
5575 PR tree-optimization/84503
5576 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5577 width as info->bitpos + info->bitsize - start.
5578 (merged_store_group::merge_overlapping): Simplify width computation.
5579 (check_no_overlap): New function.
5580 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5581 start + width and last_order of the group, fail if check_no_overlap
5583 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5584 to group if check_no_overlap fails.
5586 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5588 * config/rs6000/altivec.md: Delete contraint arguments to
5589 define_expand, define_split, and define_peephole2, and in
5590 define_insn_and_split if always unused.
5591 * config/rs6000/darwin.md: Ditto.
5592 * config/rs6000/dfp.md: Ditto.
5593 * config/rs6000/rs6000.md: Ditto.
5594 * config/rs6000/sync.md: Ditto.
5595 * config/rs6000/vector.md: Ditto.
5596 * config/rs6000/vsx.md: Ditto.
5598 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5600 * config/rs6000/altivec.md: Write output control strings as braced
5601 blocks instead of double-quoted strings.
5602 * config/rs6000/darwin.md: Ditto.
5603 * config/rs6000/rs6000.md: Ditto.
5604 * config/rs6000/vector.md: Ditto.
5605 * config/rs6000/vsx.md: Ditto.
5607 2018-02-21 Jason Merrill <jason@redhat.com>
5609 PR c++/84314 - ICE with templates and fastcall attribute.
5610 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5612 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5614 * ipa-cp.c (determine_versionability): Fix comment typos.
5616 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5619 * ipa-cp.c (determine_versionability): Do not version functions caling
5622 2018-02-21 Martin Liska <mliska@suse.cz>
5625 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5626 Add "native" as a possible value.
5627 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5628 the macro when native cpu detection is available.
5630 2018-02-21 Martin Liska <mliska@suse.cz>
5633 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5634 Add "native" as a possible value.
5635 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5636 when native cpu detection is available.
5638 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5639 Martin Sebor <msebor@redhat.com>
5641 PR tree-optimization/84478
5642 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5644 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5645 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5646 support which is conservatively correct, for 2 only stay conservative
5647 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5648 argument to the 2 argument get_range_strlen, adjust 6 arg
5649 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5651 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5652 (gimple_fold_builtin_strlen): Pass true as last argument to
5655 2018-02-20 Martin Sebor <msebor@redhat.com>
5658 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5659 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5660 (builtin_memref::builtin_memref): Factor out parts into
5661 set_base_and_offset and call it.
5663 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5666 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5667 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5668 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5669 search at the associated MODE_INT.
5671 2018-02-20 Jeff Law <law@redhat.com>
5674 PR tree-optimization/81592
5676 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5677 for range data rather than using global data.
5678 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5679 range data rather than using global data.
5680 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5681 pass it to children as needed.
5682 (struct directive::fmtresult): Similarly.
5683 (struct directive::set_width): Similarly.
5684 (struct directive::set_precision): Similarly.
5685 (format_integer, format_directive, parse_directive): Similarly.
5686 (format_none): Accept unnamed vr_values parameter.
5687 (format_percent, format_floating, format_character): Similarly.
5688 (format_string, format_plain): Similarly.
5689 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5690 the EVRP range analyzer for range data rather than using global data.
5691 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5692 gimple-ssa-evrp-analyze.h
5693 (class sprintf_dom_walker): Add after_dom_children member function.
5694 Add evrp_range_analyzer member.
5695 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5696 range analyzer as needed.
5697 (sprintf_dom_walker::after_dom_children): New member function.
5698 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5700 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5701 (evrp_range_analyzer::pop_to_marker): Likewise.
5703 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5705 PR tree-optimization/84419
5706 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5707 with the required type if its current type is compatible but
5710 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5713 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5714 after vectorization.
5716 2018-02-20 Martin Liska <mliska@suse.cz>
5719 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5720 possible values if we don't have a hint.
5722 2018-02-20 Martin Liska <mliska@suse.cz>
5726 * final.c (shorten_branches): Build align_tab array with one
5728 * opts.c (finish_options): Add alignment option limit check.
5729 (MAX_CODE_ALIGN): Likewise.
5730 (MAX_CODE_ALIGN_VALUE): Likewise.
5731 * doc/invoke.texi: Document maximum allowed option value for
5732 all -falign-* options.
5734 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5737 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5738 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5739 * var-tracking.c (emit_note_insn_var_location): Remove all references
5740 to NOTE_INSN_CALL_ARG_LOCATION.
5741 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5742 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5743 Use copy_rtx_if_shared.
5744 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5745 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5746 (dwarf2out_var_location): Remove handling of
5747 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5749 * final.c (final_scan_insn): Remove all references to
5750 NOTE_INSN_CALL_ARG_LOCATION.
5751 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5752 before dumping final insns.
5753 * except.c (emit_note_eh_region_end): Remove all references to
5754 NOTE_INSN_CALL_ARG_LOCATION.
5755 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5756 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5757 * config/arc/arc.c (hwloop_optimize): Likewise.
5758 * config/arm/arm.c (create_fix_barrier): Likewise.
5759 * config/s390/s390.c (s390_chunkify_start): Likewise.
5760 * config/sh/sh.c (find_barrier): Likewise.
5761 * config/i386/i386.c (rest_of_insert_endbranch,
5762 ix86_seh_fixup_eh_fallthru): Likewise.
5763 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5764 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5765 * config/frv/frv.c (frv_function_prologue): Likewise.
5766 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5768 (note_outside_basic_block_p): Remove all references to
5769 NOTE_INSN_CALL_ARG_LOCATION.
5770 * gengtype.c (adjust_field_rtx_def): Likewise.
5771 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5773 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5774 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5777 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5780 PR tree-optimization/84452
5781 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5782 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5785 2018-02-19 Martin Liska <mliska@suse.cz>
5788 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5790 2018-02-19 Martin Liska <mliska@suse.cz>
5791 Richard Sandiford <richard.sandiford@linaro.org>
5793 PR tree-optimization/82491
5794 * gimple-fold.c (get_base_constructor): Make earlier bail out
5797 2018-02-19 Carl Love <cel@us.ibm.com>
5799 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5800 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5802 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5803 P8V_BUILTIN_VEC_NEG.
5805 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5807 * config/rl78/rl78.md (movdf): New define expand.
5809 2018-02-19 Martin Liska <mliska@suse.cz>
5812 * doc/invoke.texi: Fix typo.
5813 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5815 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5817 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5818 handle rs6000_single_float and rs6000_double_float specially for
5821 2018-02-16 Jeff Law <law@redhat.com>
5823 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5826 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5829 * ipa-inline.c (inline_small_functions): Fix a typo.
5831 2018-02-16 Nathan Sidwell <nathan@acm.org>
5833 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5835 2018-02-16 Carl Love <cel@us.ibm.com>
5837 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5838 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5839 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5840 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5841 expansion to P8V_BUILTIN_VEC_FLOAT2.
5843 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5845 PR rtl-optimization/70023
5846 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5847 src_regno into account.
5849 2018-02-16 Carl Love <cel@us.ibm.com>
5851 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5852 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5853 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5854 * config/rs6000/rs6000.c: Remove case statements for
5855 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5856 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5857 and P9V_BUILTIN_VEC_VINSERT4B.
5858 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5859 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5860 * config/rs6000/vsx.md:
5861 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5864 2018-02-16 Carl Love <cel@us.ibm.com>
5866 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5868 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5870 * config/rs6000/rs6000-c.c: Add the definitions for
5871 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5872 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5873 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5874 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5875 definition for insert4b and define insn *insert3b_internal.
5876 * doc/extend.texi: Add documentation for vec_extract4b.
5878 2018-02-16 Nathan Sidwell <nathan@acm.org>
5880 * doc/extend.texi (Backwards Compatibility): Mention friend
5881 injection. Note for-scope is deprecated.
5882 * doc/invoke.texi (-ffriend-injection): Deprecate.
5884 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5886 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5887 that moved to I2, also allow destinations that are a paradoxical
5888 subreg (instead of a normal reg).
5890 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5893 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5896 2018-02-16 Richard Biener <rguenther@suse.de>
5898 PR tree-optimization/84037
5899 PR tree-optimization/84016
5901 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5902 Adjust vec_construct for the fact we need additional higher latency
5903 128bit inserts for AVX256 and AVX512 vector builds.
5904 (ix86_add_stmt_cost): Scale vector construction cost for
5907 2018-02-16 Richard Biener <rguenther@suse.de>
5909 PR tree-optimization/84417
5910 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5911 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5912 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5914 2018-02-16 Martin Liska <mliska@suse.cz>
5917 * internal-fn.def (ASAN_CHECK): Set proper flags.
5918 (ASAN_MARK): Likewise.
5920 2018-02-16 Julia Koval <julia.koval@intel.com>
5922 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5923 from PTA_CANNONLAKE.
5925 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5928 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5929 Use ++iter rather than iter++ for std::list iterators.
5930 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5931 defer deleting them until all nodes in the forest are processed. Do
5932 free even leaf nodes. Change to_process into auto_vec.
5935 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5936 * vec.h (vec_default_construct): Use memset instead of placement new
5937 if BROKEN_VALUE_INITIALIZATION is defined.
5938 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5939 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5942 PR rtl-optimization/83723
5943 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5944 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5945 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5947 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5949 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5951 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5953 PR rtl-optimization/81443
5954 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5955 from inner REGs to paradoxical SUBREGs.
5957 2018-02-16 Richard Biener <rguenther@suse.de>
5959 PR tree-optimization/84399
5960 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5961 For operands we can analyze at their definition make sure we can
5962 analyze them at each use as well.
5964 2018-02-16 Richard Biener <rguenther@suse.de>
5966 PR tree-optimization/84190
5967 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5968 volatile accesses if the decl isn't volatile.
5970 2018-02-15 Jason Merrill <jason@redhat.com>
5972 PR c++/84314 - ICE with templates and fastcall attribute.
5973 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5974 TYPE_CANONICAL on an existing type.
5976 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5978 PR tree-optimization/84383
5979 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5980 dstoff nor call operand_equal_p if dstbase is NULL.
5982 PR tree-optimization/84334
5983 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5984 also a CONSTANT_CLASS_P, punt.
5986 2018-02-14 Jim Wilson <jimw@sifive.com>
5988 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5989 first SMALL_OPERAND check. New local min_second_step. Move assert
5990 to where locals are set. Add TARGET_RVC support.
5991 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5993 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5995 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5997 2018-02-14 Martin Sebor <msebor@redhat.com>
5999 PR tree-optimization/83698
6000 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6001 arrays constrain the offset range to their bounds.
6002 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6003 (builtin_access::overlap): Avoid setting the size of overlap if it's
6005 (maybe_diag_overlap): Also consider arrays when deciding what values
6006 of offsets to include in diagnostics.
6008 2018-02-14 Martin Sebor <msebor@redhat.com>
6011 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6012 that correspond to the kind of a declaration.
6014 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6017 * config/pa/pa.md: Load address of PIC label using the linkage table
6018 if the label is nonlocal.
6020 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6022 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6023 warning message if user requests -maltivec=be.
6024 * doc/invoke.texi: Document deprecation of -maltivec=be.
6026 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6029 * config/rs6000/rs6000-c.c: Update definitions for
6030 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6031 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6033 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6036 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6037 add _get_ssp intrinsics. Remove argument from
6038 __builtin_ia32_rdssp[d|q].
6039 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6040 * config/i386/i386-builtin.def: Remove argument from
6041 __builtin_ia32_rdssp[d|q].
6042 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6043 ix86_expand_special_args_builtin for _rdssp[d|q].
6044 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6045 Clear register before usage.
6046 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6047 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6049 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6051 PR tree-optimization/84357
6052 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6053 operand 1 of an ARRAY_REF too.
6055 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6058 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6059 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6061 (set_of_reg): New struct.
6062 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6063 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6064 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6066 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6067 Split into bitclr, bitset, bitinvert patterns if appropriate.
6068 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6069 use rx_fuse_in_memory_bitop.
6070 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6071 to named insn, correct maximum insn length.
6073 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6076 * machmode.def: Define a complex mode for PARTIAL_INT.
6077 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6079 * doc/rtl.texi: Document CSPImode.
6080 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6082 (msp430_hard_regno_nregs_with_padding): Likewise.
6084 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6087 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6089 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6091 PR rtl-optimization/84169
6092 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6093 we generated a parallel as new i3 and we split that to new i2 and i3
6094 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6095 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6096 those to i2, not i1. Partially rewrite this scan code.
6098 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6101 * stor-layout.c (place_field): For variable length fields, adjust
6102 offset_align afterwards not just based on the field's alignment,
6103 but also on the size.
6106 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6107 of exps and logs in the use_exp2 case.
6109 2018-02-13 Jeff Law <law@redhat.com>
6111 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6114 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6117 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6121 * common.opt (gas-loc-support, gas-locview-support): New.
6122 (ginline-points, ginternal-reset-location-views): New.
6123 * doc/invoke.texi: Document them. Use @itemx where intended.
6124 (gvariable-location-views): Adjust.
6125 * target.def (reset_location_view): New.
6126 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6127 (TARGET_RESET_LOCATION_VIEW): New.
6128 * doc/tm.texi: Rebuilt.
6129 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6130 (dwarf2out_default_as_locview_support): New.
6131 (output_asm_line_debug_info): Use option variables.
6132 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6133 (output_loc_list): Likewise.
6134 (add_high_low_attributes): Check option variables.
6135 Don't output entry view attribute in strict mode.
6136 (gen_inlined_subroutine_die): Check option variables.
6137 (dwarf2out_inline_entry): Likewise.
6138 (init_sections_and_labels): Likewise.
6139 (dwarf2out_early_finish): Likewise.
6140 (maybe_reset_location_view): New, from...
6141 (dwarf2out_var_location): ... here. Call it.
6142 * debug.h (dwarf2out_default_as_loc_support): Declare.
6143 (dwarf2out_default_as_locview_support): Declare.
6144 * hooks.c (hook_int_rtx_insn_0): New.
6145 * hooks.h (hook_int_rtx_insn_0): Declare.
6146 * toplev.c (process_options): Take -gas-loc-support and
6147 -gas-locview-support from dwarf2out. Enable
6148 -gvariable-location-views by default only with locview
6149 assembler support. Enable -ginternal-reset-location-views by
6150 default only if the target defines the corresponding hook.
6151 Enable -ginline-points by default if location views are
6152 enabled; force it disabled if statement frontiers are
6154 * tree-inline.c (expand_call_inline): Check option variables.
6155 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6157 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6159 PR tree-optimization/84321
6160 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6161 handling. Also check whether the anti-range contains any values
6162 that satisfy the mask; switch to a VR_RANGE if not.
6164 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6167 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6169 2018-02-13 Martin Jambor <mjambor@suse.cz>
6172 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6173 of call statements, also set location of a load to a temporary.
6175 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6177 * config/rl78/rl78.c (add_vector_labels): New function.
6178 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6179 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6180 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6181 which checks that no arguments are passed.
6182 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6183 * doc/extend.texi: Documentation for the new attribute.
6185 2018-02-13 Andreas Schwab <schwab@suse.de>
6187 * config/riscv/linux.h (CPP_SPEC): Define.
6189 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6192 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6193 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6194 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6195 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6196 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6197 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6198 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6199 temporarily for AES and PCLMUL builtins.
6201 PR tree-optimization/84339
6202 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6203 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6207 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6208 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6209 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6211 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6213 * omp-simd-clone.h: New file.
6214 * omp-simd-clone.c: Include omp-simd-clone.h.
6215 (expand_simd_clones): No longer static.
6216 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6217 cgraph.h and omp-simd-clone.h.
6218 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6219 (vect_recog_widen_shift_pattern): Formatting fix.
6220 (vect_pattern_recog_1): Don't check optab for calls.
6223 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6224 operands[2] into a REG before using gen_lowpart on it.
6226 2018-02-12 Jeff Law <law@redhat.com>
6229 * config/sh/sh.c (find_barrier): Consider a sibling call
6232 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6233 successfully back substituting a reg.
6235 2018-02-12 Richard Biener <rguenther@suse.de>
6237 PR tree-optimization/84037
6238 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6239 parameter, move visited init to caller.
6240 (vect_slp_analyze_operations): Separate cost from validity
6241 check, initialize visited once for all instances.
6242 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6244 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6246 (vect_model_promotion_demotion_cost): Likewise.
6247 (vectorizable_bswap): Guard cost modeling with !slp_node
6248 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6250 (vectorizable_call): Likewise.
6251 (vectorizable_conversion): Likewise.
6252 (vectorizable_assignment): Likewise.
6253 (vectorizable_shift): Likewise.
6254 (vectorizable_operation): Likewise.
6255 (vectorizable_store): Likewise.
6256 (vectorizable_load): Likewise.
6257 (vectorizable_condition): Likewise.
6258 (vectorizable_comparison): Likewise.
6260 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6263 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6264 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6265 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6267 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6270 * params.def (inline-min-speedup): Increase from 8 to 15.
6271 (max-inline-insns-auto): Decrease from 40 to 30.
6272 * ipa-split.c (consider_split): Add some buffer for function to
6273 be considered inlining candidate.
6274 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6277 2018-02-12 Richard Biener <rguenther@suse.de>
6279 PR tree-optimization/84037
6280 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6281 matched stmts if we cannot swap the non-matched ones.
6283 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6285 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6286 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6287 _mm_maskz_scalef_round_ss): New intrinsics.
6288 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6289 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6290 __builtin_ia32_scalefss_round): Remove.
6291 (__builtin_ia32_scalefsd_mask_round,
6292 __builtin_ia32_scalefss_mask_round): New intrinsics.
6293 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6294 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6295 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6296 "<round_constraint>")): Changed to ...
6297 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6298 "<round_scalar_constraint>")): ... this.
6299 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6300 %0, %1, %2<round_op3>}"): Changed to ...
6301 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6302 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6303 %2<round_scalar_mask_op3>}"): ... this.
6304 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6306 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6308 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6309 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6310 (_mm_maskz_sqrt_round_ss): New intrinsics.
6311 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6312 (__builtin_ia32_sqrtsd_mask_round)
6313 (__builtin_ia32_sqrtss_mask_round): New builtins.
6314 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6315 (__builtin_ia32_sqrtss_round): Remove.
6316 (__builtin_ia32_sqrtsd_mask_round)
6317 (__builtin_ia32_sqrtss_mask_round): New builtins.
6318 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6319 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6320 ((match_operand:VF_128 1 "vector_operand"
6321 "xBm,<round_constraint>")): Changed to ...
6322 ((match_operand:VF_128 1 "vector_operand"
6323 "xBm,<round_scalar_constraint>")): ... this.
6324 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6325 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6326 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6327 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6328 %<iptr>1<round_scalar_mask_op3>}): ... this.
6329 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6330 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6332 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6335 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6336 Cast vec_cmpeq result to correct type.
6337 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6338 Cast vec_cmpgt result to correct type.
6340 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6342 * final.c (final_scan_insn_1): Renamed from...
6343 (final_scan_insn): ... this. New wrapper, to recover
6344 seen from the outermost call in recursive ones.
6345 * config/sparc/sparc.c (output_return): Drop seen from call.
6346 (output_sibcall): Likewise.
6347 * config/visium/visium.c (output_branch): Likewise.
6349 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6351 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6354 2018-02-10 Alan Modra <amodra@gmail.com>
6357 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6358 Specify LR as an input.
6360 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6363 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6364 remove_member_access_dummy_vars): New functions.
6365 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6366 lower_omp_1, execute_lower_omp): Use them.
6368 PR rtl-optimization/84308
6369 * shrink-wrap.c (spread_components): Release todo vector.
6371 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6373 PR rtl-optimization/57193
6374 * ira-color.c (struct allocno_color_data): Add member
6375 conflict_allocno_hard_prefs.
6376 (update_conflict_allocno_hard_prefs): New.
6377 (bucket_allocno_compare_func): Add a preference based on
6378 conflict_allocno_hard_prefs.
6379 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6380 (color_allocnos): Remove a dead code. Initiate
6381 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6383 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6386 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6387 constraint from =wa to wa. Avoid a subreg on the output operand,
6388 instead use a pseudo and subreg it in a move.
6389 (p9_xxbrd_<mode>): Changed to ...
6390 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6391 (p9_xxbrd_v2df): New expander.
6392 (p9_xxbrw_<mode>): Changed to ...
6393 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6394 (p9_xxbrw_v4sf): New expander.
6396 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6398 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6400 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6403 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6404 multiply in 32-bit mode.
6405 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6406 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6409 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6411 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6412 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6413 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6414 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6416 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6419 * dwarf2out.c (is_trivial_indirect_ref): New function.
6420 (dwarf2out_late_global_decl): Do not generate a location
6421 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6422 and that are not defined in the current unit.
6424 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6426 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6427 instead of a libcall for UNORDERED.
6429 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6432 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6433 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6435 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6438 * config/s390/s390.c (s390_set_current_function): Invoke
6439 s390_indirect_branch_settings also if fndecl didn't change.
6441 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6443 * config/rs6000/rs6000.md (blockage): Set length to zero.
6445 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6447 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6449 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6452 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6453 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6457 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6458 PARALLEL incoming that failed vt_get_decl_and_offset check.
6461 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6462 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6464 (get_variable_section): For decls in named .bss* sections pass true as
6465 second argument to bss_initializer_p.
6467 2018-02-09 Marek Polacek <polacek@redhat.com>
6468 Jakub Jelinek <jakub@redhat.com>
6471 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6472 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6473 Sync some changes from cxx_fold_indirect_ref.
6475 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6477 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6479 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6480 (BLOCK_INLINE_ENTRY_LABEL): New.
6481 (dwarf2out_var_location): Disregard inline entry markers.
6482 (inline_entry_data): New struct.
6483 (inline_entry_data_hasher): New hashtable type.
6484 (inline_entry_data_hasher::hash): New.
6485 (inline_entry_data_hasher::equal): New.
6486 (inline_entry_data_table): New variable.
6487 (add_high_low_attributes): Add DW_AT_entry_pc and
6488 DW_AT_GNU_entry_view attributes if a pending entry is found
6489 in inline_entry_data_table. Add old entry_pc attribute only
6490 if debug nonbinding markers are disabled.
6491 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6492 markers are enabled.
6493 (block_within_block_p, dwarf2out_inline_entry): New.
6494 (dwarf2out_finish): Check that no entries remained in
6495 inline_entry_data_table.
6496 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6497 (final_scan_insn, notice_source_line): Likewise.
6498 (rest_of_clean_state): Skip inline entry markers.
6499 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6501 * gimple.c (gimple_build_debug_inline_entry): New.
6502 * gimple.h (enum gimple_debug_subcode): Add
6503 GIMPLE_DEBUG_INLINE_ENTRY.
6504 (gimple_build_debug_inline_entry): Declare.
6505 (gimple_debug_inline_entry_p): New.
6506 (gimple_debug_nonbind_marker_p): Adjust.
6507 * insn-notes.def (INLINE_ENTRY): New.
6508 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6509 inline entry marker notes.
6510 (print_insn): Likewise.
6511 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6512 (INSN_DEBUG_MARKER_KIND): Likewise.
6513 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6514 * tree-inline.c (expand_call_inline): Build and insert
6515 debug_inline_entry stmt.
6516 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6517 inline entry blocks early, if nonbind markers are enabled.
6518 (dump_scope_block): Dump fragment info.
6519 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6520 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6521 (gimple_build_debug_inline_entry): New.
6522 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6523 Enable/disable inline entry points too.
6524 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6525 (DEBUG_INSN): Describe inline entry markers.
6527 * common.opt (gvariable-location-views): New.
6528 (gvariable-location-views=incompat5): New.
6529 * config.in: Rebuilt.
6530 * configure: Rebuilt.
6531 * configure.ac: Test assembler for view support.
6532 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6533 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6534 * dwarf2out.c (var_loc_view): New typedef.
6535 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6536 (dwarf2out_locviews_in_attribute): New.
6537 (dwarf2out_locviews_in_loclist): New.
6538 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6539 (enum dw_line_info_opcode): Add LI_adv_address.
6540 (struct dw_line_info_table): Add view.
6541 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6542 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6543 (zero_view_p): New variable.
6544 (ZERO_VIEW_P): New macro.
6545 (output_asm_line_debug_info): New.
6546 (struct var_loc_node): Add view.
6547 (add_AT_view_list, AT_loc_list): New.
6548 (add_var_loc_to_decl): Add view param. Test it against last.
6549 (new_loc_list): Add view params. Record them.
6550 (AT_loc_list_ptr): Handle loc and view lists.
6551 (view_list_to_loc_list_val_node): New.
6552 (print_dw_val): Handle dw_val_class_view_list.
6553 (size_of_die): Likewise.
6554 (value_format): Likewise.
6555 (loc_list_has_views): New.
6556 (gen_llsym): Set vl_symbol too.
6557 (maybe_gen_llsym, skip_loc_list_entry): New.
6558 (dwarf2out_maybe_output_loclist_view_pair): New.
6559 (output_loc_list): Output view list or entries too.
6560 (output_view_list_offset): New.
6561 (output_die): Handle dw_val_class_view_list.
6562 (output_dwarf_version): New.
6563 (output_compilation_unit_header): Use it.
6564 (output_skeleton_debug_sections): Likewise.
6565 (output_rnglists, output_line_info): Likewise.
6566 (output_pubnames, output_aranges): Update version comments.
6567 (output_one_line_info_table): Output view numbers in asm comments.
6568 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6569 Call maybe_gen_llsym.
6570 (loc_list_from_tree_1): Adjust.
6571 (add_AT_location_description): Create view list attribute if
6572 needed, check it's absent otherwise.
6573 (convert_cfa_to_fb_loc_list): Adjust.
6574 (maybe_emit_file): Call output_asm_line_debug_info for test.
6575 (dwarf2out_var_location): Reset views as needed. Precompute
6576 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6577 attribute. Set view.
6578 (new_line_info_table): Reset next view.
6579 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6580 (dwarf2out_source_line): Likewise. Output view resets and labels to
6581 the assembler, or select appropriate line info opcodes.
6582 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6583 (optimize_string_length): Catch it. Adjust.
6584 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6585 dw_val_class_view_list, and remove it if no longer needed.
6586 (hash_loc_list): Hash view numbers.
6587 (loc_list_hasher::equal): Compare them.
6588 (optimize_location_lists): Check whether a view list symbol is
6589 needed, and whether the locview attribute is present, and
6590 whether they match. Remove the locview attribute if no longer
6592 (index_location_lists): Call skip_loc_list_entry for test.
6593 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6594 Use output_dwarf_version.
6595 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6596 (struct dw_val_node): Add val_view_list.
6597 * final.c (SEEN_NEXT_VIEW): New.
6598 (set_next_view_needed): New.
6599 (clear_next_view_needed): New.
6600 (maybe_output_next_view): New.
6601 (final_start_function): Rename to...
6602 (final_start_function_1): ... this. Take pointer to FIRST,
6603 add SEEN parameter. Emit param bindings in the initial view.
6604 (final_start_function): Reintroduce SEEN-less interface.
6605 (final): Rename to...
6606 (final_1): ... this. Take SEEN parameter. Output final pending
6607 next view at the end.
6608 (final): Reintroduce seen-less interface.
6609 (final_scan_insn): Output pending next view before switching
6610 sections or ending a block. Mark the next view as needed when
6611 outputting variable locations. Notify debug backend of section
6612 changes, and of location view changes.
6613 (rest_of_handle_final): Adjust.
6614 * toplev.c (process_options): Autodetect value for debug variable
6615 location views option. Warn on incompat5 without -gdwarf-5.
6616 * doc/invoke.texi (gvariable-location-views): New.
6617 (gvariable-location-views=incompat5): New.
6618 (gno-variable-location-views): New.
6620 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6622 PR tree-optimization/84136
6623 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6624 that the result of find_edge is non-NULL.
6626 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6629 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6630 storing integer register in SImode. Fix cost of 256 and 512
6631 byte aligned SSE register store.
6633 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6635 * config/i386/i386.c (ix86_multiplication_cost): Fix
6636 multiplication cost for TARGET_AVX512DQ.
6638 2018-02-08 Marek Polacek <polacek@redhat.com>
6640 PR tree-optimization/84238
6641 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6644 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6646 PR tree-optimization/84265
6647 * tree-vect-stmts.c (vectorizable_store): Don't treat
6648 VMAT_CONTIGUOUS accesses as grouped.
6649 (vectorizable_load): Likewise.
6651 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6653 PR tree-optimization/81635
6654 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6655 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6656 (test_round_for_mask): New functions.
6657 (wide_int_cc_tests): Call test_round_for_mask.
6658 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6659 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6660 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6661 range returned by get_range_info.
6663 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6666 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6667 * symtab.c: Include builtins.h
6668 (symtab_node::output_to_lto_symbol_table_p): Move here
6669 from lto-streamer-out.c:output_symbol_p.
6670 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6671 (output_symbol_p): Move all logic to symtab.c
6672 (produce_symtab): Update.
6674 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6676 * config/s390/s390-opts.h (enum indirect_branch): Define.
6677 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6678 (s390_indirect_branch_via_thunk)
6679 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6680 (enum s390_indirect_branch_type): Define.
6681 * config/s390/s390.c (struct s390_frame_layout, struct
6682 machine_function): Remove.
6683 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6684 (indirect_branch_table_label_no, indirect_branch_table_name):
6686 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6687 (enum s390_indirect_branch_option): Define.
6688 (s390_return_addr_from_memory): New function.
6689 (s390_handle_string_attribute): New function.
6690 (s390_attribute_table): Add new attribute handler.
6691 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6692 (s390_indirect_branch_via_thunk): New function.
6693 (s390_indirect_branch_via_inline_thunk): New function.
6694 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6695 sibling call optimization for non z10 compiles.
6696 (s390_emit_call): Force indirect branch target to be a single
6697 register. Add r1 clobber for non-z10 compiles.
6698 (s390_emit_epilogue): Emit return jump via return_use expander.
6699 (s390_reorg): Handle JUMP_INSNs as execute targets.
6700 (s390_option_override_internal): Perform validity checks for the
6701 new command line options.
6702 (s390_indirect_branch_attrvalue): New function.
6703 (s390_indirect_branch_settings): New function.
6704 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6705 (s390_output_indirect_thunk_function): New function.
6706 (s390_code_end): Implement target hook.
6707 (s390_case_values_threshold): Implement target hook.
6708 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6710 * config/s390/s390.h (struct s390_frame_layout)
6711 (struct machine_function): Move here from s390.c.
6712 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6713 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6714 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6715 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6716 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6717 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6718 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6719 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6720 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6721 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6722 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6723 (mnemonic attribute): Add values which aren't recognized
6725 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6726 pattern for branch conversion. Fix mnemonic attribute.
6727 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6728 indirect branch via thunk if requested.
6729 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6730 ("*indirect_jump"): Disable for branch conversion using out of
6732 ("indirect_jump_via_thunk<mode>_z10")
6733 ("indirect_jump_via_thunk<mode>")
6734 ("indirect_jump_via_inlinethunk<mode>_z10")
6735 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6736 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6737 ("casesi_jump_via_inlinethunk<mode>_z10")
6738 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6739 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6740 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6741 ("*indirect2_jump"): Disable for branch conversion.
6742 ("casesi_jump"): Turn into expander and expand patterns for branch
6744 ("return_use"): New expander.
6745 ("*return"): Emit return via thunk and rename it to ...
6746 ("*return<mode>"): ... this one.
6747 * config/s390/s390.opt: Add new options and and enum for the
6750 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6752 * lra-constraints.c (match_reload): Unconditionally use
6753 gen_lowpart_SUBREG, rather than selecting between that
6754 and equivalent gen_rtx_SUBREG code.
6756 2018-02-08 Richard Biener <rguenther@suse.de>
6758 PR tree-optimization/84233
6759 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6760 changed flag instead of boguously re-using phi_inserted.
6762 2018-02-08 Martin Jambor <mjambor@suse.cz>
6764 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6765 static local variables.
6767 2018-02-08 Richard Biener <rguenther@suse.de>
6769 PR tree-optimization/84278
6770 * tree-vect-stmts.c (vectorizable_store): When looking for
6771 smaller vector types to perform grouped strided loads/stores
6772 make sure the mode is supported by the target.
6773 (vectorizable_load): Likewise.
6775 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6777 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6778 Increase LDP/STP opportunities by adding adjacent callee-saves.
6780 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6782 PR rtl-optimization/84068
6783 PR rtl-optimization/83459
6784 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6786 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6788 PR tree-optimization/84224
6789 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6790 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6793 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6796 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6797 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6798 count, remove one USE.
6800 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6802 * doc/install.texi (Configuration): Document the
6803 --with-long-double-format={ibm,ieee} PowerPC configuration
6807 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6808 Convert from define_expand to be define_insn_and_split. Rework
6809 float/double/_Float128 conversions to QI/HI/SImode to work with
6810 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6811 conversions to QI/HImode types did a store and then a load to
6812 truncate the value. For conversions to VSX registers, don't split
6813 the insn, instead emit the code directly. Use the code iterator
6814 any_fix to combine signed and unsigned conversions.
6815 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6816 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6817 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6818 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6819 (fix_<mode>di2_hw): Likewise.
6820 (fixuns_<mode>di2_hw): Likewise.
6821 (fix_<mode>si2_hw): Likewise.
6822 (fixuns_<mode>si2_hw): Likewise.
6823 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6824 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6825 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6826 fix<uns>_trunc<SFDF:mode>si2_p8.
6827 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6829 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6830 (fix<uns>_<mode>_mem): Likewise.
6831 (fctiw<u>z_<mode>_mem): Likewise.
6832 (fix<uns>_<mode>_mem): Likewise.
6833 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6834 the register allocator from doing a direct move to the GPRs to do
6835 a store, and instead use the ISA 3.0 store byte/half-word from
6836 vector register instruction. For IEEE 128-bit floating point,
6837 also optimize stores of 32-bit ints.
6838 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6840 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6842 * genextract.c (push_pathstr_operand): New function to support
6844 (walk_rtx): Call push_pathstr_operand.
6845 (print_path): Support [a-zA-Z].
6847 2018-02-07 Richard Biener <rguenther@suse.de>
6849 PR tree-optimization/84037
6850 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6851 (cse_and_gimplify_to_preheader): Declare.
6852 (vect_get_place_in_interleaving_chain): Likewise.
6853 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6855 (_loop_vec_info::~_loop_vec_info): Delete it.
6856 (cse_and_gimplify_to_preheader): New function.
6857 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6858 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6859 (vectorizable_load): Likewise. For grouped stores always base
6860 the IV on the first element.
6861 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6862 condition before gimplifying.
6864 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6866 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6867 *DIV_EXPR and *MOD_EXPR.
6869 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6872 * config/i386/i386.c (ix86_option_override_internal): Mask out
6873 the CF_SET bit when checking -fcf-protection.
6875 2018-02-07 Tom de Vries <tom@codesourcery.com>
6878 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6881 2018-02-07 Richard Biener <rguenther@suse.de>
6883 PR tree-optimization/84204
6884 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6887 PR tree-optimization/84205
6888 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6889 special-case isl_ast_op_zdiv_r.
6891 PR tree-optimization/84223
6892 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6893 Only add conditions from within the region.
6894 (gather_bbs::after_dom_children): Adjust.
6896 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6899 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6900 * config/avr/avr.md: Only post-reload split REG-REG moves if
6901 either register is GENERAL_REG_P.
6903 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6905 PR tree-optimization/84235
6906 * tree-ssa-scopedtables.c
6907 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6908 if the subtraction is performed in floating point type where NaNs are
6909 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6910 build 1. Formatting fix.
6912 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6915 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6916 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6917 and skip it regardless of bb boundaries. Use CALL_P macro,
6918 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6919 unnecessarily, formatting fix.
6921 2018-02-06 Michael Collison <michael.collison@arm.com>
6923 * config/arm/thumb2.md:
6924 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6925 (*thumb_mov_notscc): Ditto.
6927 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6930 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6931 unsigned_fix, not "s".
6933 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6935 * configure.ac (gcc_fn_eh_frame_ro): New function.
6936 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6937 correct .eh_frame permissions.
6938 * configure: Regenerate.
6940 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6942 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6945 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6947 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6948 Display warning message for -mno-speculate-indirect-jumps.
6950 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6952 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6954 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6956 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6958 PR tree-optimization/84225
6959 * tree-eh.c (find_trapping_overflow): Only call
6960 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6962 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6965 * config/i386/i386.c: Reimplement the check of possible options
6966 -mibt/-mshstk conbination. Change error messages.
6967 * doc/invoke.texi: Fix a typo: remove extra '='.
6969 2018-02-06 Marek Polacek <polacek@redhat.com>
6971 PR tree-optimization/84228
6972 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6974 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6977 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6978 emitted arch directives.
6979 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6980 __ARM_FEATURE_COPROC before changing architectures.
6982 2018-02-06 Richard Biener <rguenther@suse.de>
6984 * config/i386/i386.c (print_reg): Fix typo.
6985 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6987 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6989 * configure: Regenerate.
6991 2018-02-05 Martin Sebor <msebor@redhat.com>
6993 PR tree-optimization/83369
6994 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6997 2018-02-05 Martin Liska <mliska@suse.cz>
6999 * doc/invoke.texi: Cherry-pick upstream r323995.
7001 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7003 * ira.c (ira_init_register_move_cost): Adjust comment.
7005 2018-02-05 Martin Liska <mliska@suse.cz>
7007 PR gcov-profile/84137
7008 * doc/gcov.texi: Fix typo in documentation.
7010 2018-02-05 Martin Liska <mliska@suse.cz>
7012 PR gcov-profile/83879
7013 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7014 using dlopen functionality.
7016 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7018 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7019 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7020 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7021 _mm_maskz_range_round_ss): New intrinsics.
7022 (__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/i386-builtin.def (__builtin_ia32_rangesd128_round,
7027 __builtin_ia32_rangess128_round): Remove.
7028 (__builtin_ia32_rangesd128_mask_round,
7029 __builtin_ia32_rangess128_mask_round): New builtins.
7030 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7031 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7032 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7033 "<round_saeonly_constraint>")): Changed to ...
7034 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7035 "<round_saeonly_scalar_constraint>")): ... this.
7036 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7037 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7038 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7039 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7040 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7042 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7044 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7046 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7047 Remove all values except native, 8540 and 8548.
7049 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7051 * config/i386/i386.c (ix86_output_function_return): Pass
7052 INVALID_REGNUM, instead of -1, as invalid register number to
7053 indirect_thunk_name and output_indirect_thunk.
7055 2018-02-02 Julia Koval <julia.koval@intel.com>
7057 * config.gcc: Add -march=icelake.
7058 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7059 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7060 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7061 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7062 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7063 (processor_target_table): Add icelake.
7064 (ix86_option_override_internal): Handle new PTAs.
7065 (get_builtin_code_for_version): Handle icelake.
7066 (M_INTEL_COREI7_ICELAKE): New.
7067 (fold_builtin_cpu): Handle icelake.
7068 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7069 * doc/invoke.texi: Add -march=icelake.
7071 2018-02-02 Julia Koval <julia.koval@intel.com>
7073 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7074 to wide_int_bitmask.
7075 * wide-int-bitmask.h: New.
7077 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7080 * config/i386/i386.md: Replace Pmode with word_mode in
7081 builtin_setjmp_setup and builtin_longjmp to support x32.
7083 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7087 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7089 (rs6000_supported_cpu_names): New static variable.
7090 (linux_cpu_translation_table): Likewise.
7091 (elf_platform) <cpu>: Define new static variable and use it.
7092 Translate kernel AT_PLATFORM name to canonical name if needed.
7093 Error if platform name is unknown.
7095 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7098 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7100 2018-02-01 Jeff Law <law@redhat.com>
7103 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7104 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7105 the scratch if RELEASE_VIA_POP is false.
7106 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7107 If we have to save a temporary register, decrement SIZE appropriately.
7108 Pass new arguments to release_scratch_register_on_entry.
7109 (ix86_adjust_stack_and_probe): Likewise.
7110 (ix86_emit_probe_stack_range): Pass new arguments to
7111 release_scratch_register_on_entry.
7113 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7115 PR rtl-optimization/84157
7116 * combine.c (change_zero_ext): Use REG_P predicate in
7117 front of HARD_REGISTER_P predicate.
7119 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7121 * config/avr/avr.c (avr_option_override): Move disabling of
7122 -fdelete-null-pointer-checks to...
7123 * common/config/avr/avr-common.c (avr_option_optimization_table):
7126 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7128 PR tree-optimization/81635
7129 * tree-data-ref.c (split_constant_offset_1): For types that
7130 wrap on overflow, try to use range info to prove that wrapping
7133 2018-02-01 Renlin Li <renlin.li@arm.com>
7136 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7138 (aarch64_register_move_cost): Likewise.
7139 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7141 (REG_CLASS_NAMES): Likewise.
7142 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7143 TAILCALL_ADDR_REGS. Remove IP registers.
7144 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7146 2018-02-01 Richard Biener <rguenther@suse.de>
7148 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7149 for specifying RPO order and allow NULL for that.
7150 * domwalk.c (dom_walker::dom_walker): Likewise.
7151 (dom_walker::walk): Handle NULL RPO order.
7152 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7154 (rewrite_update_dom_walker): Likewise.
7155 (mark_def_dom_walker): Likewise.
7157 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7159 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7160 (aarch64_maybe_expand_sve_subreg_move): Declare.
7161 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7162 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7164 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7165 that are semantically a reverse operation.
7166 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7167 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7168 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7170 (aarch64_can_change_mode_class): For big-endian, forbid changes
7171 between two SVE modes if they have different element sizes.
7173 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7175 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7176 the TImode handling for big-endian targets.
7178 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7180 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7181 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7183 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7184 Remove BSWAP handing for big-endian targets and use the form of
7185 LD1RQ appropariate for the mode.
7187 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7189 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7190 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7193 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7196 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7197 check for operands that need to go through aarch64_sve_reload_be.
7199 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7201 PR tree-optimization/81661
7202 PR tree-optimization/84117
7203 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7204 * tree-eh.c: Include gimplify.h.
7205 (find_trapping_overflow, replace_trapping_overflow,
7206 rewrite_to_non_trapping_overflow): New functions.
7207 * tree-vect-loop.c: Include tree-eh.h.
7208 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7209 * tree-data-ref.c: Include tree-eh.h.
7210 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7212 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7214 PR rtl-optimization/84123
7215 * combine.c (change_zero_ext): Check if hard register satisfies
7216 can_change_dest_mode before calling gen_lowpart_SUBREG.
7218 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7221 * ira.c (ira_init_register_move_cost): Remove assert.
7223 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7225 PR rtl-optimization/84071
7226 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7227 * doc/tm.texi: Regenerate.
7229 2018-01-31 Richard Biener <rguenther@suse.de>
7231 PR tree-optimization/84132
7232 * tree-data-ref.c (analyze_miv_subscript): Properly
7233 check whether evolution_function_is_affine_multivariate_p
7234 before calling gcd_of_steps_may_divide_p.
7236 2018-01-31 Julia Koval <julia.koval@intel.com>
7239 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7240 * config/i386/i386.md (rdpid_rex64) New.
7241 (rdpid): Make 32bit only.
7243 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7246 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7247 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7249 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7252 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7254 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7256 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7258 PR rtl-optimization/84071
7259 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7260 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7262 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7264 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7265 (arc_attribute_table): Add 'aux' attribute.
7266 (arc_in_small_data_p): Consider aux like variables.
7267 (arc_is_aux_reg_p): New function.
7268 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7269 (arc_get_aux_arg): New function.
7270 (prepare_move_operands): Handle aux-register access.
7271 (arc_handle_aux_attribute): New function.
7272 * doc/extend.texi (ARC Variable attributes): Add subsection.
7274 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7276 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7277 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7278 (arc_attribute_table): Add 'uncached' attribute.
7279 (arc_print_operand): Print '.di' flag for uncached memory
7281 (arc_in_small_data_p): Do not consider for small data the uncached
7283 (arc_is_uncached_mem_p): New function.
7284 * config/arc/predicates.md (compact_store_memory_operand): Check
7285 for uncached memory accesses.
7286 (nonvol_nonimm_operand): Likewise.
7287 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7289 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7292 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7293 falign-loops=): Add Optimization flag.
7295 2018-01-30 Jeff Law <law@redhat.com>
7298 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7299 INT_REGISTERS_SAVED. Check it prior to calling
7300 get_scratch_register_on_entry.
7301 (ix86_adjust_stack_and_probe): Similarly.
7302 (ix86_emit_probe_stack_range): Similarly.
7303 (ix86_expand_prologue): Corresponding changes.
7305 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7308 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7309 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7311 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7314 * lra-constraints.c (curr_insn_transform): Process AND in the
7317 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7319 PR rtl-optimization/83986
7320 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7321 dependence against last_pending_memory_flush in addition to
7324 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7326 PR tree-optimization/81611
7327 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7330 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7333 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7336 2018-01-30 Richard Biener <rguenther@suse.de>
7337 Jakub Jelinek <jakub@redhat.com>
7339 PR tree-optimization/84111
7340 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7341 inner loops added during recursion, as they don't have up-to-date
7344 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7347 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7348 (can_inline_edge_by_limits_p): ... here.
7349 (can_early_inline_edge_p, check_callers,
7350 update_caller_keys, update_callee_keys, recursive_inlining,
7351 add_new_edges_to_heap, speculation_useful_p,
7352 inline_small_functions,
7353 inline_small_functions, flatten_function,
7354 inline_to_all_callers_1): Update.
7356 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7358 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7361 2018-01-30 Richard Biener <rguenther@suse.de>
7363 PR tree-optimization/83008
7364 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7365 invariant and constant vector uses in stmts when they need
7368 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7371 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7372 * configure: Regenerate.
7374 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7376 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7378 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7379 Use gen_rtx_REG rather than gen_lowpart.
7381 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7383 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7384 rather than 0 when creating partial subregs.
7386 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7388 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7391 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7394 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7395 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7396 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7397 flags. This restores the settings used before the 2017-07-24.
7398 Turning off pre increment/decrement/modify allows IVOPTS to
7399 optimize DF/SF loops where the index is an int.
7401 2018-01-29 Richard Biener <rguenther@suse.de>
7402 Kelvin Nilsen <kelvin@gcc.gnu.org>
7405 * tree-vect-stmts.c (vectorizable_call): Don't call
7406 targetm.vectorize_builtin_md_vectorized_function if callee is
7409 2018-01-22 Carl Love <cel@us.ibm.com>
7411 * doc/extend.tex: Fix typo in second arg in
7412 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7414 2018-01-29 Richard Biener <rguenther@suse.de>
7416 PR tree-optimization/84086
7417 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7418 (flush_ssaname_freelist): When SSA names were released reset
7419 the SCEV hash table.
7421 2018-01-29 Richard Biener <rguenther@suse.de>
7423 PR tree-optimization/84057
7424 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7425 removed paths when removing edges.
7427 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7429 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7430 -mfunction-return=@var{choice}.
7432 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7435 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7437 (layout::print_source_line): Likewise.
7438 (test_get_line_width_without_trailing_whitespace): Add test cases.
7440 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7443 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7446 2018-01-26 Jim Wilson <jimw@sifive.com>
7448 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7450 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7453 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7455 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7456 and CMP + SUB-immediate -> SUBS.
7458 2018-01-26 Martin Sebor <msebor@redhat.com>
7460 PR tree-optimization/83896
7461 * tree-ssa-strlen.c (get_string_len): Rename...
7462 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7463 Avoid assuming length is constant.
7464 (handle_char_store): Use HOST_WIDE_INT for string length.
7466 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7469 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7470 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7472 2018-01-26 Richard Biener <rguenther@suse.de>
7474 PR rtl-optimization/84003
7475 * dse.c (record_store): Only record redundant stores when
7476 the earlier store aliases at least all accesses the later one does.
7478 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7480 PR rtl-optimization/83985
7481 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7482 REG_CFA_RESTORE insns.
7483 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7484 REG_CFA_RESTORE insns here.
7487 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7488 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7490 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7492 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7493 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7494 (arc_init): Likewise.
7495 (arc_override_options): Likewise.
7496 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7498 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7500 * config/arc/arc.h (TARGET_DBNZ): Define.
7501 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7502 properly set the tune attribute.
7503 (dbnz): Use TARGET_DBNZ guard.
7504 * config/arc/arc.opt (mtune): Add core3 option.
7506 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7508 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7509 recognize new pic like addresses.
7510 (arc_delegitimize_address): Clean up.
7512 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7514 * config/arc/arc-arches.def: Option mrf16 valid for all
7516 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7517 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7518 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7519 * config/arc/arc-tables.opt: Regenerate.
7520 * config/arc/arc.c (arc_conditional_register_usage): Handle
7521 reduced register file case.
7522 (arc_file_start): Set must have build attributes.
7523 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7525 * config/arc/arc.opt (mrf16): Add new option.
7526 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7527 * config/arc/genmultilib.awk: Handle new mrf16 option.
7528 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7529 * config/arc/t-multilib: Regenerate.
7530 * doc/invoke.texi (ARC Options): Document mrf16 option.
7532 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7534 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7535 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7536 (arc_attribute_table): Add 'secure_call' attribute.
7537 (arc_print_operand): Print secure call operand.
7538 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7540 (arc_is_secure_call_p): New function. * config/arc/arc.md
7541 (call_i): Add support for sjli instruction.
7542 (call_value_i): Likewise.
7543 * config/arc/constraints.md (Csc): New constraint.
7545 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7546 John Eric Martin <John.Martin@emmicro-us.com>
7548 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7549 * config/arc/arc.c (_arc_jli_section): New struct.
7550 (arc_jli_section): New type.
7551 (rc_jli_sections): New static variable.
7552 (arc_handle_jli_attribute): New function.
7553 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7554 (arc_file_end): New function.
7555 (TARGET_ASM_FILE_END): Define.
7556 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7557 (arc_add_jli_section): New function.
7558 (jli_call_scan): Likewise.
7559 (arc_reorg): Call jli_call_scan.
7560 (arc_output_addsi): Remove 'S' from printing asm operand.
7561 (arc_is_jli_call_p): New function.
7562 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7564 (movhi_insn): Likewise.
7565 (movsi_insn): Likewise.
7566 (movsi_set_cc_insn): Likewise.
7567 (loadqi_update): Likewise.
7568 (load_zeroextendqisi_update): Likewise.
7569 (load_signextendqisi_update): Likewise.
7570 (loadhi_update): Likewise.
7571 (load_zeroextendhisi_update): Likewise.
7572 (load_signextendhisi_update): Likewise.
7573 (loadsi_update): Likewise.
7574 (loadsf_update): Likewise.
7575 (movsicc_insn): Likewise.
7576 (bset_insn): Likewise.
7577 (bxor_insn): Likewise.
7578 (bclr_insn): Likewise.
7579 (bmsk_insn): Likewise.
7580 (bicsi3_insn): Likewise.
7581 (cmpsi_cc_c_insn): Likewise.
7582 (movsi_ne): Likewise.
7583 (movsi_cond_exec): Likewise.
7584 (clrsbsi2): Likewise.
7595 (call_i): Remove 'S' asm letter, add jli instruction.
7596 (call_value_i): Likewise.
7597 * config/arc/arc.op (mjli-always): New option.
7598 * config/arc/constraints.md (Cji): New constraint.
7599 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7601 (subsf3_fpx): Likewise.
7602 (mulsf3_fpx): Likewise.
7603 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7605 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7607 * doc/invoke.texi (ARC): Document mjli-always option.
7609 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7611 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7612 avoid addition with 0 and use incw and decw where possible.
7614 2018-01-26 Richard Biener <rguenther@suse.de>
7616 PR tree-optimization/81082
7617 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7618 association if it requires casting to unsigned.
7619 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7620 from fold_plusminus_mult_expr to catch important cases late when
7621 range info is available.
7623 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7625 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7626 * configure.ac (hidden_linkonce): New test.
7627 * configure: Regenerate.
7628 * config.in: Regenerate.
7630 2018-01-26 Julia Koval <julia.koval@intel.com>
7632 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7633 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7634 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7635 _mm_mask_bitshuffle_epi64_mask): Fix type.
7636 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7637 USI_FTYPE_V4DI_V4DI_USI): Remove.
7638 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7639 __builtin_ia32_vpshufbitqmb256_mask,
7640 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7641 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7642 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7644 2018-01-26 Alan Modra <amodra@gmail.com>
7647 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7648 UNSPEC_VBPERMQ. Sort other unspecs.
7650 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7652 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7654 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7657 * predict.c (drop_profile): Do not push/pop cfun; update also
7659 (handle_missing_profiles): Fix logic looking for zero profiles.
7661 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7664 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7665 on functions with #pragma omp declare simd or functions with simd
7667 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7668 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7669 Remove trailing \n from warning_at calls.
7671 2018-01-25 Tom de Vries <tom@codesourcery.com>
7674 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7675 for neutered workers.
7677 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7680 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7681 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7683 2018-01-24 Jeff Law <law@redhat.com>
7686 * i386.c (get_probe_interval): Move to earlier point.
7687 (ix86_compute_frame_layout): If -fstack-clash-protection and
7688 the frame is larger than the probe interval, then use pushes
7689 to save registers rather than reg->mem moves.
7690 (ix86_expand_prologue): Remove conditional for int_registers_saved
7693 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7696 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7697 min/max for never referenced object.
7699 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7702 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7704 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7705 attributes from DECL_ATTRIBUTES (decl) without affecting
7706 DECL_ATTRIBUTES (current_function_decl).
7707 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7708 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7710 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7712 PR tree-optimization/83979
7713 * fold-const.c (fold_comparison): Use constant_boolean_node
7714 instead of boolean_{true,false}_node.
7716 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7718 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7721 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7723 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7724 Simplify the clause that sets the length attribute.
7725 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7726 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7727 clause that sets the length attribute.
7728 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7730 2018-01-24 Tom de Vries <tom@codesourcery.com>
7733 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7734 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7735 Add strict parameter.
7736 (prevent_branch_around_nothing): Insert dummy insn between branch to
7737 label and label with no ptx insn inbetween.
7738 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7740 2018-01-24 Tom de Vries <tom@codesourcery.com>
7743 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7744 for neutered threads in warp.
7745 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7747 2018-01-24 Richard Biener <rguenther@suse.de>
7749 PR tree-optimization/83176
7750 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7753 2018-01-24 Richard Biener <rguenther@suse.de>
7755 PR tree-optimization/82819
7756 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7757 code generating pluses that are no-ops in the target precision.
7759 2018-01-24 Richard Biener <rguenther@suse.de>
7762 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7764 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7766 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7767 to merge probabilities.
7768 * predict.c (probably_never_executed): Also mark as cold functions
7769 with global 0 profile and guessed local profile.
7770 * profile-count.c (profile_probability::combine_with_count): New
7772 * profile-count.h (profile_probability::operator*,
7773 profile_probability::operator*=, profile_probability::operator/,
7774 profile_probability::operator/=): Reduce precision to adjusted
7775 and set value to guessed on contradictory divisions.
7776 (profile_probability::combine_with_freq): Remove.
7777 (profile_probability::combine_wiht_count): Declare.
7778 (profile_count::force_nonzero):: Set to adjusted.
7779 (profile_count::probability_in):: Set quality to adjusted.
7780 * tree-ssa-tail-merge.c (replace_block_by): Use
7783 2018-01-23 Andrew Waterman <andrew@sifive.com>
7784 Jim Wilson <jimw@sifive.com>
7786 * config/riscv/riscv.c (riscv_stack_boundary): New.
7787 (riscv_option_override): Set riscv_stack_boundary. Handle
7788 riscv_preferred_stack_boundary_arg.
7789 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7790 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7791 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7792 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7793 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7794 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7796 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7799 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7800 of struct ix86_frame.
7801 (ix86_expand_epilogue): Likewise. Add a local variable for
7802 the reg_save_offset field in struct ix86_frame.
7804 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7806 PR tree-optimization/82604
7807 * tree-loop-distribution.c (enum partition_kind): New enum item
7808 PKIND_PARTIAL_MEMSET.
7809 (partition_builtin_p): Support above new enum item.
7810 (generate_code_for_partition): Ditto.
7811 (compute_access_range): Differentiate cases that equality can be
7812 proven at all loops, the innermost loops or no loops.
7813 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7814 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7815 (finalize_partitions, distribute_loop): Don't fuse partition of
7816 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7817 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7820 2018-01-23 Martin Liska <mliska@suse.cz>
7822 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7823 order to ignore the predictor.
7824 (PRED_POLYMORPHIC_CALL): Likewise.
7825 (PRED_RECURSIVE_CALL): Likewise.
7827 2018-01-23 Martin Liska <mliska@suse.cz>
7829 * tree-profile.c (tree_profiling): Print function header to
7830 aware reader which function we are working on.
7831 * value-prof.c (gimple_find_values_to_profile): Do not print
7832 not interesting value histograms.
7834 2018-01-23 Martin Liska <mliska@suse.cz>
7836 * profile-count.h (enum profile_quality): Add
7837 profile_uninitialized as the first value. Do not number values
7838 as they are zero based.
7839 (profile_count::verify): Update sanity check.
7840 (profile_probability::verify): Likewise.
7842 2018-01-23 Nathan Sidwell <nathan@acm.org>
7844 * doc/invoke.texi (ffor-scope): Deprecate.
7846 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7848 PR tree-optimization/83510
7849 * domwalk.c (set_all_edges_as_executable): New function.
7850 (dom_walker::dom_walker): Convert bool param
7851 "skip_unreachable_blocks" to enum reachability. Move setup of
7852 edge flags to set_all_edges_as_executable and only do it when
7853 reachability is REACHABLE_BLOCKS.
7854 * domwalk.h (enum dom_walker::reachability): New enum.
7855 (dom_walker::dom_walker): Convert bool param
7856 "skip_unreachable_blocks" to enum reachability.
7857 (set_all_edges_as_executable): New decl.
7858 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7859 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7861 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7862 but converting true to REACHABLE_BLOCKS.
7863 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7865 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7866 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7867 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7869 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7870 if check_all_array_refs will be called.
7872 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7874 * tree.c (selftest::test_location_wrappers): Add more test
7877 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7879 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7880 (selftest::test_bit_in_range): Likewise.
7882 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7885 * doc/sourcebuild.texi (vect_float): Say that the selector
7886 only describes the situation when -funsafe-math-optimizations is on.
7887 (vect_float_strict): Document.
7889 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7891 PR tree-optimization/83965
7892 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7893 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7894 instead of checking only for a reduction.
7895 (vect_recog_widen_sum_pattern): Likewise.
7897 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7899 * predict.c (probably_never_executed): Only use precise profile info.
7900 (compute_function_frequency): Skip after inlining hack since we now
7901 have quality checking.
7903 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7905 * profile-count.h (profile_probability::very_unlikely,
7906 profile_probability::unlikely, profile_probability::even): Set
7907 precision to guessed.
7909 2018-01-23 Richard Biener <rguenther@suse.de>
7911 PR tree-optimization/83963
7912 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7913 Properly terminate dominator walk when crossing the exit edge not
7914 when visiting its source block.
7916 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7919 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7920 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7922 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7924 PR tree-optimization/83957
7925 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7926 semicolon after for body surrounded by braces.
7928 PR tree-optimization/83081
7929 * profile-count.h (profile_probability::split): New method.
7930 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7931 Use profile_probability::split.
7932 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7933 when splitting a single conditional jump into 2.
7935 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7937 PR tree-optimization/69452
7938 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7941 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7943 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7944 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7945 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7947 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7949 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7950 * config/rl78/rl78.md (movdi): New define_expand.
7951 * config/rl78/rl78.c (rl78_split_movdi): New function.
7953 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7956 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7958 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7959 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7960 128-bit to produce an UNSPEC move to get the double word with the
7961 signbit and then a shift directly to do signbit.
7962 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7963 implementation with a new version that just does either a direct
7964 move or a regular move. Move memory interface to separate insns.
7965 Move insns so they are next to the expander.
7966 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7967 with signbit move. Split big and little endian case.
7968 (signbit<mode>2_dm_mem_le): Likewise.
7969 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7970 (signbit<mode>2_dm2): Likewise.
7972 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7974 * config/rl78/rl78.md (anddi3): New define_expand.
7976 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7978 * config/rl78/rl78.md (umindi3): New define_expand.
7980 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7982 * config/rl78/rl78.md (smindi3): New define_expand.
7984 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7986 * config/rl78/rl78.md (smaxdi3): New define_expand.
7988 2018-01-22 Carl Love <cel@us.ibm.com>
7990 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7991 LVX_V1TI): Add macro expansion.
7992 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7993 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7994 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7995 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7996 Change check to determine if the instruction is a byte reversing
7997 entry. Fix typo in comment.
7998 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7999 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8000 Add def_builtin calls for new builtins.
8001 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8002 Add define_insn expansion.
8004 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8006 * config/rl78/rl78.md (umaxdi3): New define_expand.
8008 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8010 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8011 for non-QImode registers.
8013 2018-01-22 Richard Biener <rguenther@suse.de>
8015 PR tree-optimization/83963
8016 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8017 including the loop exit block.
8018 (scop_detection::merge_sese): Likewise.
8019 (scop_detection::add_scop): Do it here instead.
8021 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8023 * doc/sourcebuild.texi (arm_softfloat): Document.
8025 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8028 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8029 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8030 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8032 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8033 David Edelsohn <dje.gcc@gmail.com>
8036 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8037 Change "crset eq" to "crset 2".
8038 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8039 (*call_indirect_aix<mode>_nospec): Likewise.
8040 (*call_value_indirect_aix<mode>_nospec): Likewise.
8041 (*call_indirect_elfv2<mode>_nospec): Likewise.
8042 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8043 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8044 change assembly output from . to $.
8045 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8046 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8047 (*tablejump<mode>_internal1_nospec): Likewise.
8049 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8052 * config/sh/sh_optimize_sett_clrt.cc:
8053 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8055 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8057 PR tree-optimization/83940
8058 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8059 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8060 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8061 use it to pass back the definition type.
8062 (vect_check_store_rhs): Likewise rhs_dt_out.
8063 (vect_build_gather_load_calls): Add a mask_dt argument and use
8064 it instead of a call to vect_is_simple_use.
8065 (vectorizable_store): Update calls to vect_check_load_store_mask
8066 and vect_check_store_rhs. Use the dt returned by the latter instead
8067 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8068 instead of calls to vect_is_simple_use. Pass the scalar rather
8069 than the vector operand to vect_is_simple_use when handling
8070 second and subsequent copies of an rhs value.
8071 (vectorizable_load): Update calls to vect_check_load_store_mask
8072 and vect_build_gather_load_calls. Use the cached mask_dt and
8073 gs_info.offset_dt instead of calls to vect_is_simple_use.
8075 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8078 * tree-emutls.c: Include gimplify.h.
8079 (lower_emutls_2): New function.
8080 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8081 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8082 it before further processing.
8085 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8086 UINTVAL (trueop1) instead of INTVAL (op1).
8088 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8092 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8093 INCOMING_FRAME_SP_OFFSET if not defined.
8094 (scan_trace): Add ENTRY argument. If true and
8095 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8096 emit a note to adjust the CFA offset.
8097 (create_cfi_notes): Adjust scan_trace callers.
8098 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8099 INCOMING_FRAME_SP_OFFSET in the CIE.
8100 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8101 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8103 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8104 * doc/tm.texi: Regenerated.
8106 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8108 PR rtl-optimization/83147
8109 * lra-constraints.c (remove_inheritance_pseudos): Use
8110 lra_substitute_pseudo_within_insn.
8112 2018-01-19 Tom de Vries <tom@codesourcery.com>
8113 Cesar Philippidis <cesar@codesourcery.com>
8116 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8118 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8121 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8122 spaces for function labels.
8124 2018-01-19 Martin Liska <mliska@suse.cz>
8126 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8127 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8128 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8129 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8130 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8131 (PRED_CONST_RETURN): Change from 69 to 65.
8132 (PRED_NULL_RETURN): Change from 91 to 71.
8133 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8134 (PRED_LOOP_GUARD): Change from 66 to 73.
8136 2018-01-19 Martin Liska <mliska@suse.cz>
8138 * predict.c (predict_insn_def): Add new assert.
8139 (struct branch_predictor): Change type to signed integer.
8140 (test_prediction_value_range): Amend test to cover
8142 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8143 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8144 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8145 (PRED_LOOP_IV_COMPARE): Likewise.
8146 * predict.h (PROB_UNINITIALIZED): Define new constant.
8148 2018-01-19 Martin Liska <mliska@suse.cz>
8150 * predict.c (dump_prediction): Add new format for
8151 analyze_brprob.py script which is enabled with -details
8153 * profile-count.h (precise_p): New function.
8155 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8157 PR tree-optimization/83922
8158 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8159 there are no statements that need masking.
8160 (vect_active_double_reduction_p): New function.
8161 (vect_analyze_loop_operations): Use it when handling phis that
8162 are not in the loop header.
8164 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8166 PR tree-optimization/83914
8167 * tree-vect-loop.c (vectorizable_induction): Don't convert
8168 init_expr or apply the peeling adjustment for inductions
8169 that are nested within the vectorized loop.
8171 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8173 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8176 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8180 * function.h (gimplify_parameters): Add gimple_seq * argument.
8181 * function.c: Include gimple.h and options.h.
8182 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8183 for the added local temporaries if needed.
8184 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8185 if there are any parameter cleanups, wrap whole body into a
8186 try/finally with the cleanups.
8188 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8191 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8192 Use GET_MODE_CLASS for scalar floating point.
8194 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8198 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8199 Fix call of call_cgraph_insertion_hooks.
8201 2018-01-18 Martin Sebor <msebor@redhat.com>
8203 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8205 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8208 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8211 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8214 * common.opt: (-ffile-prefix-map): New option.
8215 * opts.c (common_handle_option): Defer it.
8216 * opts-global.c (handle_common_deferred_options): Handle it.
8217 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8218 * file-prefix-map.h: New file.
8219 (remap_debug_filename, add_debug_prefix_map): ...here.
8220 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8221 * final.c (debug_prefix_map, add_debug_prefix_map
8222 remap_debug_filename): Move to...
8223 * file-prefix-map.c: New file.
8224 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8225 generalize, get rid of alloca(), use strrchr() instead of strchr().
8226 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8227 Implement in terms of add_prefix_map().
8228 (remap_macro_filename, remap_debug_filename): Implement in term of
8230 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8231 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8232 * dbxout.c: Include file-prefix-map.h.
8233 * varasm.c: Likewise.
8234 * vmsdbgout.c: Likewise.
8235 * xcoffout.c: Likewise.
8236 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8237 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8238 * doc/invoke.texi (-ffile-prefix-map): Document.
8239 (-fdebug-prefix-map): Update description.
8241 2018-01-18 Martin Liska <mliska@suse.cz>
8243 * config/i386/i386.c (indirect_thunk_name): Document that also
8245 (output_indirect_thunk): Document why both instructions
8246 (pause and lfence) are generated.
8248 2018-01-18 Richard Biener <rguenther@suse.de>
8250 PR tree-optimization/83887
8251 * graphite-scop-detection.c
8252 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8253 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8254 (scop_detection::merge_sese): Re-implement with a flood-fill
8255 algorithm that properly finds a SESE region if it exists.
8257 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8260 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8261 pointer_diff optimizations use view_convert instead of convert.
8263 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8265 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8266 Generate different code for -mno-speculate-indirect-jumps.
8267 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8268 (*call_indirect_aix<mode>): Disable for
8269 -mno-speculate-indirect-jumps.
8270 (*call_indirect_aix<mode>_nospec): New define_insn.
8271 (*call_value_indirect_aix<mode>): Disable for
8272 -mno-speculate-indirect-jumps.
8273 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8274 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8275 -mno-speculate-indirect-jumps.
8276 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8278 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8280 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8281 long double type, set the flags for noting the default long double
8282 type, even if we don't pass or return a long double type.
8284 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8287 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8290 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8292 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8293 support for merge[hl].
8294 (fold_mergehl_helper): New helper function.
8295 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8296 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8297 (altivec_vmrglw_direct): Add xxmrglw insn.
8299 2018-01-17 Andrew Waterman <andrew@sifive.com>
8301 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8302 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8304 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8307 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8308 call the lto_location_cache before reading the
8309 DECL_SOURCE_LOCATION of the types.
8311 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8312 Richard Sandiford <richard.sandiford@linaro.org>
8314 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8315 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8316 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8317 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8318 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8320 * config/aarch64/constraints.md (aarch64_movti_operand):
8322 * config/aarch64/predicates.md (Uti): Add new constraint.
8324 2018-01-17 Carl Love <cel@us.ibm.com>
8326 * config/rs6000/vsx.md (define_expand xl_len_r,
8327 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8328 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8330 (define_expand, define_insn): Move the shift left from the
8331 define_insn to the define_expand for lxvl and stxvl instructions.
8332 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8333 and XL_LEN_R definitions to PURE.
8335 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8337 * config/i386/i386.c (indirect_thunk_name): Declare regno
8338 as unsigned int. Compare regno with INVALID_REGNUM.
8339 (output_indirect_thunk): Ditto.
8340 (output_indirect_thunk_function): Ditto.
8341 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8342 in the call to output_indirect_thunk_function.
8344 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8347 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8348 rather than the size of inner_type to determine the stack slot size
8349 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8351 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8354 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8357 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8359 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8360 endian Linux systems to optionally enable multilibs for selecting
8361 the long double type if the user configured an explicit type.
8362 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8363 have no long double multilibs if not defined.
8364 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8365 warn if the user used -mabi={ieee,ibm}longdouble and we built
8366 multilibs for long double.
8367 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8368 appropriate multilib option.
8369 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8371 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8372 for building long double multilibs.
8373 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8375 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8377 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8380 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8382 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8385 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8388 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8391 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8393 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8394 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8396 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8398 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8399 different rtl trees depending on TARGET_64BIT.
8400 (rs6000_gen_lvx): Likewise.
8402 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8404 * config/visium/visium.md (nop): Tweak comment.
8405 (hazard_nop): Likewise.
8407 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8409 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8410 -mspeculate-indirect-jumps.
8411 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8412 for -mno-speculate-indirect-jumps.
8413 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8414 (*call_value_indirect_elfv2<mode>): Disable for
8415 -mno-speculate-indirect-jumps.
8416 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8417 (indirect_jump): Emit different RTL for
8418 -mno-speculate-indirect-jumps.
8419 (*indirect_jump<mode>): Disable for
8420 -mno-speculate-indirect-jumps.
8421 (*indirect_jump<mode>_nospec): New define_insn.
8422 (tablejump): Emit different RTL for
8423 -mno-speculate-indirect-jumps.
8424 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8425 (tablejumpsi_nospec): New define_expand.
8426 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8427 (tablejumpdi_nospec): New define_expand.
8428 (*tablejump<mode>_internal1): Disable for
8429 -mno-speculate-indirect-jumps.
8430 (*tablejump<mode>_internal1_nospec): New define_insn.
8431 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8434 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8436 * caller-save.c (insert_save): Drop unnecessary parameter. All
8439 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8440 Richard Biener <rguenth@suse.de>
8443 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8444 return early, inline manually is_gimple_sizepos. Make sure if we
8445 call gimplify_expr we don't end up with a gimple constant.
8446 * tree.c (variably_modified_type_p): Don't return true for
8447 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8448 * gimplify.h (is_gimple_sizepos): Remove.
8450 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8452 PR tree-optimization/83857
8453 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8454 vectorizable_live_operation for pure SLP statements.
8455 (vectorizable_live_operation): Handle PHIs.
8457 2018-01-16 Richard Biener <rguenther@suse.de>
8459 PR tree-optimization/83867
8460 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8461 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8463 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8466 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8467 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8468 If off is not INTEGER_CST, issue a may not be aligned warning
8469 rather than isn't aligned. Use isn%'t rather than isn't.
8470 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8472 <case MULT_EXPR>: Improve the case when bottom and one of the
8473 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8474 operand, in that case check if the other operand is multiple of
8475 bottom divided by the INTEGER_CST operand.
8477 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8480 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8481 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8482 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8483 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8484 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8485 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8486 (pa_function_arg_size): New function.
8488 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8490 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8491 in a separate statement.
8493 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8495 PR tree-optimization/83847
8496 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8497 group gathers and scatters.
8499 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8501 PR rtl-optimization/86620
8502 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8504 PR rtl-optimization/83213
8505 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8506 to last if both are JUMP_INSNs.
8508 PR tree-optimization/83843
8509 * gimple-ssa-store-merging.c
8510 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8511 store_immediate_info for bswap/nop orig_stores.
8513 2018-01-15 Andrew Waterman <andrew@sifive.com>
8515 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8517 <UDIV>: Increase cost if !TARGET_DIV.
8519 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8521 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8522 (define_attr "cr_logical_3op"): New.
8523 (cceq_ior_compare): Adjust.
8524 (cceq_ior_compare_complement): Adjust.
8525 (*cceq_rev_compare): Adjust.
8526 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8527 (is_cracked_insn): Adjust.
8528 (insn_must_be_first_in_group): Adjust.
8529 * config/rs6000/40x.md: Adjust.
8530 * config/rs6000/440.md: Adjust.
8531 * config/rs6000/476.md: Adjust.
8532 * config/rs6000/601.md: Adjust.
8533 * config/rs6000/603.md: Adjust.
8534 * config/rs6000/6xx.md: Adjust.
8535 * config/rs6000/7450.md: Adjust.
8536 * config/rs6000/7xx.md: Adjust.
8537 * config/rs6000/8540.md: Adjust.
8538 * config/rs6000/cell.md: Adjust.
8539 * config/rs6000/e300c2c3.md: Adjust.
8540 * config/rs6000/e500mc.md: Adjust.
8541 * config/rs6000/e500mc64.md: Adjust.
8542 * config/rs6000/e5500.md: Adjust.
8543 * config/rs6000/e6500.md: Adjust.
8544 * config/rs6000/mpc.md: Adjust.
8545 * config/rs6000/power4.md: Adjust.
8546 * config/rs6000/power5.md: Adjust.
8547 * config/rs6000/power6.md: Adjust.
8548 * config/rs6000/power7.md: Adjust.
8549 * config/rs6000/power8.md: Adjust.
8550 * config/rs6000/power9.md: Adjust.
8551 * config/rs6000/rs64.md: Adjust.
8552 * config/rs6000/titan.md: Adjust.
8554 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8556 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8557 ix86_indirect_branch_register logic.
8559 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8561 * config/i386/constraints.md (Bs): Update
8562 ix86_indirect_branch_register check. Don't check
8563 ix86_indirect_branch_register with GOT_memory_operand.
8565 * config/i386/predicates.md (GOT_memory_operand): Don't check
8566 ix86_indirect_branch_register here.
8567 (GOT32_symbol_operand): Likewise.
8569 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8571 * config/i386/predicates.md (constant_call_address_operand):
8572 Rewrite ix86_indirect_branch_register logic.
8573 (sibcall_insn_operand): Likewise.
8575 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8577 * config/i386/constraints.md (Bs): Replace
8578 ix86_indirect_branch_thunk_register with
8579 ix86_indirect_branch_register.
8581 * config/i386/i386.md (indirect_jump): Likewise.
8582 (tablejump): Likewise.
8583 (*sibcall_memory): Likewise.
8584 (*sibcall_value_memory): Likewise.
8585 Peepholes of indirect call and jump via memory: Likewise.
8586 * config/i386/i386.opt: Likewise.
8587 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8588 (GOT_memory_operand): Likewise.
8589 (call_insn_operand): Likewise.
8590 (sibcall_insn_operand): Likewise.
8591 (GOT32_symbol_operand): Likewise.
8593 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8596 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8597 type rather than type addr's type points to.
8598 (expand_omp_atomic_mutex): Likewise.
8599 (expand_omp_atomic): Likewise.
8601 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8604 * config/i386/i386.c (output_indirect_thunk_function): Use
8605 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8606 for __x86_return_thunk.
8608 2018-01-15 Richard Biener <rguenther@suse.de>
8611 * expmed.c (extract_bit_field_1): Fix typo.
8613 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8616 * config/arm/iterators.md (VF): New mode iterator.
8617 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8618 Remove integer-related logic from pattern.
8619 (neon_vabd<mode>_3): Likewise.
8621 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8624 * common.opt (fstrict-overflow): No longer an alias.
8625 (fwrapv-pointer): New option.
8626 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8627 also for pointer types based on flag_wrapv_pointer.
8628 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8629 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8630 opts->x_flag_wrapv got set.
8631 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8632 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8633 POINTER_TYPE_OVERFLOW_UNDEFINED.
8634 * match.pd: Likewise in address comparison pattern.
8635 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8637 2018-01-15 Richard Biener <rguenther@suse.de>
8640 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8641 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8642 Reset type names to their identifier if their TYPE_DECL doesn't
8643 have linkage (and thus is used for ODR and devirt).
8644 (save_debug_info_for_decl): Remove.
8645 (save_debug_info_for_type): Likewise.
8646 (add_tree_to_fld_list): Adjust.
8647 * tree-pretty-print.c (dump_generic_node): Make dumping of
8648 type names more robust.
8650 2018-01-15 Richard Biener <rguenther@suse.de>
8652 * BASE-VER: Bump to 8.0.1.
8654 2018-01-14 Martin Sebor <msebor@redhat.com>
8657 * builtins.c (check_access): Avoid warning when the no-warning bit
8660 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8662 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8663 * ira-color (allocno_hard_regs_compare): Likewise.
8665 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8668 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8669 Use .pushsection/.popsection.
8671 2018-01-14 Martin Sebor <msebor@redhat.com>
8674 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8676 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8678 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8679 entry from extra_headers.
8680 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8681 extra_headers, make the list bitwise identical to the i?86-*-* one.
8683 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8685 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8686 -mcmodel=large with -mindirect-branch=thunk,
8687 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8688 -mfunction-return=thunk-extern.
8689 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8690 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8691 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8693 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8695 * config/i386/i386.c (print_reg): Print the name of the full
8696 integer register without '%'.
8697 (ix86_print_operand): Handle 'V'.
8698 * doc/extend.texi: Document 'V' modifier.
8700 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8702 * config/i386/constraints.md (Bs): Disallow memory operand for
8703 -mindirect-branch-register.
8705 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8706 (GOT_memory_operand): Likewise.
8707 (call_insn_operand): Likewise.
8708 (sibcall_insn_operand): Likewise.
8709 (GOT32_symbol_operand): Likewise.
8710 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8711 for -mindirect-branch-register.
8712 (tablejump): Likewise.
8713 (*sibcall_memory): Likewise.
8714 (*sibcall_value_memory): Likewise.
8715 Disallow peepholes of indirect call and jump via memory for
8716 -mindirect-branch-register.
8717 (*call_pop): Replace m with Bw.
8718 (*call_value_pop): Likewise.
8719 (*sibcall_pop_memory): Replace m with Bs.
8720 * config/i386/i386.opt (mindirect-branch-register): New option.
8721 * doc/invoke.texi: Document -mindirect-branch-register option.
8723 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8725 * config/i386/i386-protos.h (ix86_output_function_return): New.
8726 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8727 set function_return_type.
8728 (indirect_thunk_name): Add ret_p to indicate thunk for function
8730 (output_indirect_thunk_function): Pass false to
8731 indirect_thunk_name.
8732 (ix86_output_indirect_branch_via_reg): Likewise.
8733 (ix86_output_indirect_branch_via_push): Likewise.
8734 (output_indirect_thunk_function): Create alias for function
8735 return thunk if regno < 0.
8736 (ix86_output_function_return): New function.
8737 (ix86_handle_fndecl_attribute): Handle function_return.
8738 (ix86_attribute_table): Add function_return.
8739 * config/i386/i386.h (machine_function): Add
8740 function_return_type.
8741 * config/i386/i386.md (simple_return_internal): Use
8742 ix86_output_function_return.
8743 (simple_return_internal_long): Likewise.
8744 * config/i386/i386.opt (mfunction-return=): New option.
8745 (indirect_branch): Mention -mfunction-return=.
8746 * doc/extend.texi: Document function_return function attribute.
8747 * doc/invoke.texi: Document -mfunction-return= option.
8749 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8751 * config/i386/i386-opts.h (indirect_branch): New.
8752 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8753 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8754 with local indirect jump when converting indirect call and jump.
8755 (ix86_set_indirect_branch_type): New.
8756 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8757 (indirectlabelno): New.
8758 (indirect_thunk_needed): Likewise.
8759 (indirect_thunk_bnd_needed): Likewise.
8760 (indirect_thunks_used): Likewise.
8761 (indirect_thunks_bnd_used): Likewise.
8762 (INDIRECT_LABEL): Likewise.
8763 (indirect_thunk_name): Likewise.
8764 (output_indirect_thunk): Likewise.
8765 (output_indirect_thunk_function): Likewise.
8766 (ix86_output_indirect_branch_via_reg): Likewise.
8767 (ix86_output_indirect_branch_via_push): Likewise.
8768 (ix86_output_indirect_branch): Likewise.
8769 (ix86_output_indirect_jmp): Likewise.
8770 (ix86_code_end): Call output_indirect_thunk_function if needed.
8771 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8773 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8774 (ix86_attribute_table): Add indirect_branch.
8775 * config/i386/i386.h (machine_function): Add indirect_branch_type
8776 and has_local_indirect_jump.
8777 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8779 (tablejump): Likewise.
8780 (*indirect_jump): Use ix86_output_indirect_jmp.
8781 (*tablejump_1): Likewise.
8782 (simple_return_indirect_internal): Likewise.
8783 * config/i386/i386.opt (mindirect-branch=): New option.
8784 (indirect_branch): New.
8787 (thunk-inline): Likewise.
8788 (thunk-extern): Likewise.
8789 * doc/extend.texi: Document indirect_branch function attribute.
8790 * doc/invoke.texi: Document -mindirect-branch= option.
8792 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8795 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8797 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8799 * ipa-inline.c (want_inline_small_function_p): Return false if
8800 inlining has already failed with CIF_FINAL_ERROR.
8801 (update_caller_keys): Call want_inline_small_function_p before
8803 (update_callee_keys): Likewise.
8805 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8807 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8809 (rs6000_quadword_masked_address_p): Likewise.
8810 (quad_aligned_load_p): Likewise.
8811 (quad_aligned_store_p): Likewise.
8812 (const_load_sequence_p): Add comment to describe the outer-most loop.
8813 (mimic_memory_attributes_and_flags): New function.
8814 (rs6000_gen_stvx): Likewise.
8815 (replace_swapped_aligned_store): Likewise.
8816 (rs6000_gen_lvx): Likewise.
8817 (replace_swapped_aligned_load): Likewise.
8818 (replace_swapped_load_constant): Capitalize argument name in
8819 comment describing this function.
8820 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8821 and stores that access quad-word aligned addresses and replace
8822 with stvx or lvx instructions when appropriate.
8823 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8824 New function prototype.
8825 (rs6000_quadword_masked_address_p): Likewise.
8826 (rs6000_gen_lvx): Likewise.
8827 (rs6000_gen_stvx): Likewise.
8828 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8829 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8830 when memory address is aligned.
8831 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8832 this split to select lvx instruction when memory address is aligned.
8833 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8834 instruction when memory address is aligned.
8835 (*vsx_le_perm_load_v16qi): Likewise.
8836 (four unnamed splitters): Modify to select the stvx instruction
8837 when memory is aligned.
8839 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8841 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8842 which appears in the queue multiple times.
8844 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8845 Alan Hayward <alan.hayward@arm.com>
8846 David Sherwood <david.sherwood@arm.com>
8848 * tree-vectorizer.h (vec_lower_bound): New structure.
8849 (_loop_vec_info): Add check_nonzero and lower_bounds.
8850 (LOOP_VINFO_CHECK_NONZERO): New macro.
8851 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8852 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8853 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8854 fields. Make seg_len the distance travelled, not including the
8856 (dr_direction_indicator): Declare.
8857 (dr_zero_step_indicator): Likewise.
8858 (dr_known_forward_stride_p): Likewise.
8859 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8861 (runtime_alias_check_p): Allow runtime alias checks with
8863 (operator ==): Compare access_size and align.
8864 (prune_runtime_alias_test_list): Rework for new distinction between
8865 the access_size and seg_len.
8866 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8868 (get_segment_min_max): New function.
8869 (create_intersect_range_checks): Use it.
8870 (dr_step_indicator): New function.
8871 (dr_direction_indicator): Likewise.
8872 (dr_zero_step_indicator): Likewise.
8873 (dr_known_forward_stride_p): Likewise.
8874 * tree-loop-distribution.c (data_ref_segment_size): Return
8875 DR_STEP * (niters - 1).
8876 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8878 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8879 (vect_preserves_scalar_order_p): New function, split out from...
8880 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8881 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8882 (vect_vfa_access_size): New function.
8883 (vect_vfa_align): Likewise.
8884 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8885 (dump_lower_bound): New function.
8886 (vect_check_lower_bound): Likewise.
8887 (vect_small_gap_p): Likewise.
8888 (vectorizable_with_step_bound_p): Likewise.
8889 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8890 depencies if the vectorization factor is 1. Convert the checks
8891 for nonzero steps into checks on the bounds of DR_STEP. Try using
8892 a bunds check for variable steps if the minimum required step is
8893 relatively small. Update calls to the dr_with_seg_len
8894 constructor and to vect_compile_time_alias.
8895 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8897 (vect_loop_versioning): Call it.
8898 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8900 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8902 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8903 Alan Hayward <alan.hayward@arm.com>
8904 David Sherwood <david.sherwood@arm.com>
8906 * doc/sourcebuild.texi (vect_scatter_store): Document.
8907 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8909 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8911 * genopinit.c (main): Add supports_vec_scatter_store and
8912 supports_vec_scatter_store_cached to target_optabs.
8913 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8914 IFN_MASK_SCATTER_STORE.
8915 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8917 * internal-fn.h (internal_store_fn_p): Declare.
8918 (internal_fn_stored_value_index): Likewise.
8919 * internal-fn.c (scatter_store_direct): New macro.
8920 (expand_scatter_store_optab_fn): New function.
8921 (direct_scatter_store_optab_supported_p): New macro.
8922 (internal_store_fn_p): New function.
8923 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8924 IFN_MASK_SCATTER_STORE.
8925 (internal_fn_mask_index): Likewise.
8926 (internal_fn_stored_value_index): New function.
8927 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8929 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8930 * optabs-query.c (supports_vec_scatter_store_p): New function.
8931 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8932 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8933 true for scatter stores.
8934 (vect_gather_scatter_fn_p): Handle scatter stores too.
8935 (vect_check_gather_scatter): Consider using scatter stores if
8936 supports_vec_scatter_store_p.
8937 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8939 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8940 internal_fn_stored_value_index.
8941 (check_load_store_masking): Handle scatter stores too.
8942 (vect_get_store_rhs): Make public.
8943 (vectorizable_call): Use internal_store_fn_p.
8944 (vectorizable_store): Handle scatter store internal functions.
8945 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8946 when deciding whether the end of the group has been reached.
8947 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8948 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8949 (mask_scatter_store<mode>): New insns.
8951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8952 Alan Hayward <alan.hayward@arm.com>
8953 David Sherwood <david.sherwood@arm.com>
8955 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8956 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8957 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8959 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8960 Use vect_truncate_gather_scatter_offset if we can't treat the
8961 operation as a normal gather load or scatter store.
8962 (get_group_load_store_type): Take the gather_scatter_info
8963 as argument. Try using a gather load or scatter store for
8964 single-element groups.
8965 (get_load_store_type): Update calls to get_group_load_store_type
8966 and vect_use_strided_gather_scatters_p.
8968 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8969 Alan Hayward <alan.hayward@arm.com>
8970 David Sherwood <david.sherwood@arm.com>
8972 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8973 optional tree argument.
8974 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8976 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8977 but continue to use the current value as a fallback.
8978 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8979 to compare the updates.
8980 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8981 (get_load_store_type): Use it when handling a strided access.
8982 (vect_get_strided_load_store_ops): New function.
8983 (vect_get_data_ptr_increment): Likewise.
8984 (vectorizable_load): Handle strided gather loads. Always pass
8985 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8987 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8988 Alan Hayward <alan.hayward@arm.com>
8989 David Sherwood <david.sherwood@arm.com>
8991 * doc/md.texi (gather_load@var{m}): Document.
8992 (mask_gather_load@var{m}): Likewise.
8993 * genopinit.c (main): Add supports_vec_gather_load and
8994 supports_vec_gather_load_cached to target_optabs.
8995 * optabs-tree.c (init_tree_optimization_optabs): Use
8996 ggc_cleared_alloc to allocate target_optabs.
8997 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8998 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9000 * internal-fn.h (internal_load_fn_p): Declare.
9001 (internal_gather_scatter_fn_p): Likewise.
9002 (internal_fn_mask_index): Likewise.
9003 (internal_gather_scatter_fn_supported_p): Likewise.
9004 * internal-fn.c (gather_load_direct): New macro.
9005 (expand_gather_load_optab_fn): New function.
9006 (direct_gather_load_optab_supported_p): New macro.
9007 (direct_internal_fn_optab): New function.
9008 (internal_load_fn_p): Likewise.
9009 (internal_gather_scatter_fn_p): Likewise.
9010 (internal_fn_mask_index): Likewise.
9011 (internal_gather_scatter_fn_supported_p): Likewise.
9012 * optabs-query.c (supports_at_least_one_mode_p): New function.
9013 (supports_vec_gather_load_p): Likewise.
9014 * optabs-query.h (supports_vec_gather_load_p): Declare.
9015 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9016 and memory_type field.
9017 (NUM_PATTERNS): Bump to 15.
9018 * tree-vect-data-refs.c: Include internal-fn.h.
9019 (vect_gather_scatter_fn_p): New function.
9020 (vect_describe_gather_scatter_call): Likewise.
9021 (vect_check_gather_scatter): Try using internal functions for
9022 gather loads. Recognize existing calls to a gather load function.
9023 (vect_analyze_data_refs): Consider using gather loads if
9024 supports_vec_gather_load_p.
9025 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9026 (vect_get_gather_scatter_offset_type): Likewise.
9027 (vect_convert_mask_for_vectype): Likewise.
9028 (vect_add_conversion_to_patterm): Likewise.
9029 (vect_try_gather_scatter_pattern): Likewise.
9030 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9031 (vect_vect_recog_func_ptrs): Add it.
9032 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9033 internal_fn_mask_index and internal_gather_scatter_fn_p.
9034 (check_load_store_masking): Take the gather_scatter_info as an
9035 argument and handle gather loads.
9036 (vect_get_gather_scatter_ops): New function.
9037 (vectorizable_call): Check internal_load_fn_p.
9038 (vectorizable_load): Likewise. Handle gather load internal
9040 (vectorizable_store): Update call to check_load_store_masking.
9041 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9042 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9043 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9044 (aarch64_gather_scale_operand_d): New predicates.
9045 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9046 (mask_gather_load<mode>): New insns.
9048 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9049 Alan Hayward <alan.hayward@arm.com>
9050 David Sherwood <david.sherwood@arm.com>
9052 * optabs.def (fold_left_plus_optab): New optab.
9053 * doc/md.texi (fold_left_plus_@var{m}): Document.
9054 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9055 * internal-fn.c (fold_left_direct): Define.
9056 (expand_fold_left_optab_fn): Likewise.
9057 (direct_fold_left_optab_supported_p): Likewise.
9058 * fold-const-call.c (fold_const_fold_left): New function.
9059 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9060 * tree-parloops.c (valid_reduction_p): New function.
9061 (gather_scalar_reductions): Use it.
9062 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9063 (vect_finish_replace_stmt): Declare.
9064 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9065 (needs_fold_left_reduction_p): New function, split out from...
9066 (vect_is_simple_reduction): ...here. Accept reductions that
9067 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9068 (vect_force_simple_reduction): Also store the reduction type in
9069 the assignment's STMT_VINFO_REDUC_TYPE.
9070 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9071 (merge_with_identity): New function.
9072 (vect_expand_fold_left): Likewise.
9073 (vectorize_fold_left_reduction): Likewise.
9074 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9075 scalar phi in place for it. Check for target support and reject
9076 cases that would reassociate the operation. Defer the transform
9077 phase to vectorize_fold_left_reduction.
9078 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9079 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9080 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9082 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9084 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9085 call to ifc_temp_var.
9087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9088 Alan Hayward <alan.hayward@arm.com>
9089 David Sherwood <david.sherwood@arm.com>
9091 * target.def (legitimize_address_displacement): Take the original
9092 offset as a poly_int.
9093 * targhooks.h (default_legitimize_address_displacement): Update
9095 * targhooks.c (default_legitimize_address_displacement): Likewise.
9096 * doc/tm.texi: Regenerate.
9097 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9098 as an argument, moving assert of ad->disp == ad->disp_term to...
9099 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9100 Try calling targetm.legitimize_address_displacement before expanding
9101 the address rather than afterwards, and adjust for the new interface.
9102 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9103 Match the new hook interface. Handle SVE addresses.
9104 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9109 * Makefile.in (OBJS): Add early-remat.o.
9110 * target.def (select_early_remat_modes): New hook.
9111 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9112 * doc/tm.texi: Regenerate.
9113 * targhooks.h (default_select_early_remat_modes): Declare.
9114 * targhooks.c (default_select_early_remat_modes): New function.
9115 * timevar.def (TV_EARLY_REMAT): New timevar.
9116 * passes.def (pass_early_remat): New pass.
9117 * tree-pass.h (make_pass_early_remat): Declare.
9118 * early-remat.c: New file.
9119 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9121 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9123 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9124 Alan Hayward <alan.hayward@arm.com>
9125 David Sherwood <david.sherwood@arm.com>
9127 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9128 vfm1 with a bound_epilog parameter.
9129 (vect_do_peeling): Update calls accordingly, and move the prologue
9130 call earlier in the function. Treat the base bound_epilog as 0 for
9131 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9132 this base when peeling for gaps.
9133 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9134 with fully-masked loops.
9135 (vect_estimate_min_profitable_iters): Handle the single peeled
9136 iteration in that case.
9138 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9139 Alan Hayward <alan.hayward@arm.com>
9140 David Sherwood <david.sherwood@arm.com>
9142 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9143 single-element interleaving even if the size is not a power of 2.
9144 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9145 accesses for single-element interleaving if the group size is
9148 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9149 Alan Hayward <alan.hayward@arm.com>
9150 David Sherwood <david.sherwood@arm.com>
9152 * doc/md.texi (fold_extract_last_@var{m}): Document.
9153 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9154 * optabs.def (fold_extract_last_optab): New optab.
9155 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9156 * internal-fn.c (fold_extract_direct): New macro.
9157 (expand_fold_extract_optab_fn): Likewise.
9158 (direct_fold_extract_optab_supported_p): Likewise.
9159 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9160 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9161 EXTRACT_LAST_REDUCTION.
9162 (get_initial_def_for_reduction): Do not create an initial vector
9163 for EXTRACT_LAST_REDUCTION reductions.
9164 (vectorizable_reduction): Leave the scalar phi in place for
9165 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9166 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9167 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9168 transform phase to vectorizable_condition.
9169 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9171 (vect_finish_stmt_generation): ...here.
9172 (vect_finish_replace_stmt): New function.
9173 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9174 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9176 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9178 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9179 Alan Hayward <alan.hayward@arm.com>
9180 David Sherwood <david.sherwood@arm.com>
9182 * doc/md.texi (extract_last_@var{m}): Document.
9183 * optabs.def (extract_last_optab): New optab.
9184 * internal-fn.def (EXTRACT_LAST): New internal function.
9185 * internal-fn.c (cond_unary_direct): New macro.
9186 (expand_cond_unary_optab_fn): Likewise.
9187 (direct_cond_unary_optab_supported_p): Likewise.
9188 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9189 loops using EXTRACT_LAST.
9190 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9191 (extract_last_<mode>): ...this optab.
9192 (vec_extract<mode><Vel>): Update accordingly.
9194 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9195 Alan Hayward <alan.hayward@arm.com>
9196 David Sherwood <david.sherwood@arm.com>
9198 * target.def (empty_mask_is_expensive): New hook.
9199 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9200 * doc/tm.texi: Regenerate.
9201 * targhooks.h (default_empty_mask_is_expensive): Declare.
9202 * targhooks.c (default_empty_mask_is_expensive): New function.
9203 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9204 if the target says that empty masks are expensive.
9205 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9207 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9210 Alan Hayward <alan.hayward@arm.com>
9211 David Sherwood <david.sherwood@arm.com>
9213 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9214 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9215 (vect_use_loop_mask_for_alignment_p): New function.
9216 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9217 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9218 niters_skip argument. Make sure that the first niters_skip elements
9219 of the first iteration are inactive.
9220 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9221 Update call to vect_set_loop_masks_directly.
9222 (get_misalign_in_elems): New function, split out from...
9223 (vect_gen_prolog_loop_niters): ...here.
9224 (vect_update_init_of_dr): Take a code argument that specifies whether
9225 the adjustment should be added or subtracted.
9226 (vect_update_init_of_drs): Likewise.
9227 (vect_prepare_for_masked_peels): New function.
9228 (vect_do_peeling): Skip prologue peeling if we're using a mask
9229 instead. Update call to vect_update_inits_of_drs.
9230 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9232 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9233 alignment. Do not include the number of peeled iterations in
9234 the minimum threshold in that case.
9235 (vectorizable_induction): Adjust the start value down by
9236 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9237 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9238 Take the number of skipped iterations into account when calculating
9240 * tree-vect-stmts.c (vect_gen_while_not): New function.
9242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9243 Alan Hayward <alan.hayward@arm.com>
9244 David Sherwood <david.sherwood@arm.com>
9246 * doc/sourcebuild.texi (vect_fully_masked): Document.
9247 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9249 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9251 (vect_analyze_loop_2): ...here. Don't check the vectorization
9252 factor against the number of loop iterations if the loop is
9255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9256 Alan Hayward <alan.hayward@arm.com>
9257 David Sherwood <david.sherwood@arm.com>
9259 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9260 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9261 (dump_groups): Update accordingly.
9262 (iv_use::mem_type): New member variable.
9263 (address_p): New function.
9264 (record_use): Add a mem_type argument and initialize the new
9266 (record_group_use): Add a mem_type argument. Use address_p.
9267 Remove obsolete null checks of base_object. Update call to record_use.
9268 (find_interesting_uses_op): Update call to record_group_use.
9269 (find_interesting_uses_cond): Likewise.
9270 (find_interesting_uses_address): Likewise.
9271 (get_mem_type_for_internal_fn): New function.
9272 (find_address_like_use): Likewise.
9273 (find_interesting_uses_stmt): Try find_address_like_use before
9274 calling find_interesting_uses_op.
9275 (addr_offset_valid_p): Use the iv mem_type field as the type
9276 of the addressed memory.
9277 (add_autoinc_candidates): Likewise.
9278 (get_address_cost): Likewise.
9279 (split_small_address_groups_p): Use address_p.
9280 (split_address_groups): Likewise.
9281 (add_iv_candidate_for_use): Likewise.
9282 (autoinc_possible_for_pair): Likewise.
9283 (rewrite_groups): Likewise.
9284 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9285 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9286 (get_alias_ptr_type_for_ptr_address): New function.
9287 (rewrite_use_address): Rewrite address uses in calls that were
9288 identified by find_address_like_use.
9290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9291 Alan Hayward <alan.hayward@arm.com>
9292 David Sherwood <david.sherwood@arm.com>
9294 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9296 * gimple-expr.h (is_gimple_addressable: Likewise.
9297 * gimple-expr.c (is_gimple_address): Likewise.
9298 * internal-fn.c (expand_call_mem_ref): New function.
9299 (expand_mask_load_optab_fn): Use it.
9300 (expand_mask_store_optab_fn): Likewise.
9302 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9303 Alan Hayward <alan.hayward@arm.com>
9304 David Sherwood <david.sherwood@arm.com>
9306 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9307 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9308 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9309 (cond_umax@var{mode}): Document.
9310 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9311 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9312 (cond_umin_optab, cond_umax_optab): New optabs.
9313 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9314 (COND_IOR, COND_XOR): New internal functions.
9315 * internal-fn.h (get_conditional_internal_fn): Declare.
9316 * internal-fn.c (cond_binary_direct): New macro.
9317 (expand_cond_binary_optab_fn): Likewise.
9318 (direct_cond_binary_optab_supported_p): Likewise.
9319 (get_conditional_internal_fn): New function.
9320 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9321 Cope with reduction statements that are vectorized as calls rather
9323 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9324 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9325 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9326 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9327 (UNSPEC_COND_EOR): New unspecs.
9328 (optab): Add mappings for them.
9329 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9330 (sve_int_op, sve_fp_op): New int attributes.
9332 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9333 Alan Hayward <alan.hayward@arm.com>
9334 David Sherwood <david.sherwood@arm.com>
9336 * optabs.def (while_ult_optab): New optab.
9337 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9338 * internal-fn.def (WHILE_ULT): New internal function.
9339 * internal-fn.h (direct_internal_fn_supported_p): New override
9340 that takes two types as argument.
9341 * internal-fn.c (while_direct): New macro.
9342 (expand_while_optab_fn): New function.
9343 (convert_optab_supported_p): Likewise.
9344 (direct_while_optab_supported_p): New macro.
9345 * wide-int.h (wi::udiv_ceil): New function.
9346 * tree-vectorizer.h (rgroup_masks): New structure.
9347 (vec_loop_masks): New typedef.
9348 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9350 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9351 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9352 (vect_max_vf): New function.
9353 (slpeel_make_loop_iterate_ntimes): Delete.
9354 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9355 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9356 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9357 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9358 internal-fn.h, stor-layout.h and optabs-query.h.
9359 (vect_set_loop_mask): New function.
9360 (add_preheader_seq): Likewise.
9361 (add_header_seq): Likewise.
9362 (interleave_supported_p): Likewise.
9363 (vect_maybe_permute_loop_masks): Likewise.
9364 (vect_set_loop_masks_directly): Likewise.
9365 (vect_set_loop_condition_masked): Likewise.
9366 (vect_set_loop_condition_unmasked): New function, split out from
9367 slpeel_make_loop_iterate_ntimes.
9368 (slpeel_make_loop_iterate_ntimes): Rename to..
9369 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9370 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9371 (vect_do_peeling): Update call accordingly.
9372 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9374 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9375 mask_compare_type, can_fully_mask_p and fully_masked_p.
9376 (release_vec_loop_masks): New function.
9377 (_loop_vec_info): Use it to free the loop masks.
9378 (can_produce_all_loop_masks_p): New function.
9379 (vect_get_max_nscalars_per_iter): Likewise.
9380 (vect_verify_full_masking): Likewise.
9381 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9382 retries, and free the mask rgroups before retrying. Check loop-wide
9383 reasons for disallowing fully-masked loops. Make the final decision
9384 about whether use a fully-masked loop or not.
9385 (vect_estimate_min_profitable_iters): Do not assume that peeling
9386 for the number of iterations will be needed for fully-masked loops.
9387 (vectorizable_reduction): Disable fully-masked loops.
9388 (vectorizable_live_operation): Likewise.
9389 (vect_halve_mask_nunits): New function.
9390 (vect_double_mask_nunits): Likewise.
9391 (vect_record_loop_mask): Likewise.
9392 (vect_get_loop_mask): Likewise.
9393 (vect_transform_loop): Handle the case in which the final loop
9394 iteration might handle a partial vector. Call vect_set_loop_condition
9395 instead of slpeel_make_loop_iterate_ntimes.
9396 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9397 (check_load_store_masking): New function.
9398 (prepare_load_store_mask): Likewise.
9399 (vectorizable_store): Handle fully-masked loops.
9400 (vectorizable_load): Likewise.
9401 (supportable_widening_operation): Use vect_halve_mask_nunits for
9403 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9404 (vect_gen_while): New function.
9405 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9406 (aarch64_uqdec<mode>): New insn.
9408 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9409 Alan Hayward <alan.hayward@arm.com>
9410 David Sherwood <david.sherwood@arm.com>
9412 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9413 (reduc_xor_scal_optab): New optabs.
9414 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9415 (reduc_xor_scal_@var{m}): Document.
9416 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9417 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9419 * fold-const-call.c (fold_const_call): Handle them.
9420 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9421 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9422 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9423 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9424 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9425 (UNSPEC_XORV): New unspecs.
9426 (optab): Add entries for them.
9427 (BITWISEV): New int iterator.
9428 (bit_reduc_op): New int attributes.
9430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9431 Alan Hayward <alan.hayward@arm.com>
9432 David Sherwood <david.sherwood@arm.com>
9434 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9435 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9436 * optabs.def (vec_shl_insert_optab): New optab.
9437 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9438 (duplicate_and_interleave): Likewise.
9439 * tree-vect-loop.c: Include internal-fn.h.
9440 (neutral_op_for_slp_reduction): New function, split out from
9441 get_initial_defs_for_reduction.
9442 (get_initial_def_for_reduction): Handle option 2 for variable-length
9443 vectors by loading the neutral value into a vector and then shifting
9444 the initial value into element 0.
9445 (get_initial_defs_for_reduction): Replace the code argument with
9446 the neutral value calculated by neutral_op_for_slp_reduction.
9447 Use gimple_build_vector for constant-length vectors.
9448 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9449 but the first group_size elements have a neutral value.
9450 Use duplicate_and_interleave otherwise.
9451 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9452 Update call to get_initial_defs_for_reduction. Handle SLP
9453 reductions for variable-length vectors by creating one vector
9454 result for each scalar result, with the elements associated
9455 with other scalar results stubbed out with the neutral value.
9456 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9457 Require IFN_VEC_SHL_INSERT for double reductions on
9458 variable-length vectors, or SLP reductions that have
9459 a neutral value. Require can_duplicate_and_interleave_p
9460 support for variable-length unchained SLP reductions if there
9461 is no neutral value, such as for MIN/MAX reductions. Also require
9462 the number of vector elements to be a multiple of the number of
9463 SLP statements when doing variable-length unchained SLP reductions.
9464 Update call to vect_create_epilog_for_reduction.
9465 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9466 and remove initial values.
9467 (duplicate_and_interleave): Make public.
9468 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9469 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9471 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9472 Alan Hayward <alan.hayward@arm.com>
9473 David Sherwood <david.sherwood@arm.com>
9475 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9476 (can_duplicate_and_interleave_p): New function.
9477 (vect_get_and_check_slp_defs): Take the vector of statements
9478 rather than just the current one. Remove excess parentheses.
9479 Restriction rejectinon of vect_constant_def and vect_external_def
9480 for variable-length vectors to boolean types, or types for which
9481 can_duplicate_and_interleave_p is false.
9482 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9483 (duplicate_and_interleave): New function.
9484 (vect_get_constant_vectors): Use gimple_build_vector for
9485 constant-length vectors and suitable variable-length constant
9486 vectors. Use duplicate_and_interleave for other variable-length
9487 vectors. Don't defer the update when inserting new statements.
9489 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9490 Alan Hayward <alan.hayward@arm.com>
9491 David Sherwood <david.sherwood@arm.com>
9493 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9494 min_profitable_iters doesn't go negative.
9496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9497 Alan Hayward <alan.hayward@arm.com>
9498 David Sherwood <david.sherwood@arm.com>
9500 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9501 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9502 * optabs.def (vec_mask_load_lanes_optab): New optab.
9503 (vec_mask_store_lanes_optab): Likewise.
9504 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9505 (MASK_STORE_LANES): Likewise.
9506 * internal-fn.c (mask_load_lanes_direct): New macro.
9507 (mask_store_lanes_direct): Likewise.
9508 (expand_mask_load_optab_fn): Handle masked operations.
9509 (expand_mask_load_lanes_optab_fn): New macro.
9510 (expand_mask_store_optab_fn): Handle masked operations.
9511 (expand_mask_store_lanes_optab_fn): New macro.
9512 (direct_mask_load_lanes_optab_supported_p): Likewise.
9513 (direct_mask_store_lanes_optab_supported_p): Likewise.
9514 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9516 (vect_load_lanes_supported): Likewise.
9517 * tree-vect-data-refs.c (strip_conversion): New function.
9518 (can_group_stmts_p): Likewise.
9519 (vect_analyze_data_ref_accesses): Use it instead of checking
9520 for a pair of assignments.
9521 (vect_store_lanes_supported): Take a masked_p parameter.
9522 (vect_load_lanes_supported): Likewise.
9523 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9524 vect_store_lanes_supported and vect_load_lanes_supported.
9525 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9526 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9527 parameter. Don't allow gaps for masked accesses.
9528 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9529 and vect_load_lanes_supported.
9530 (get_load_store_type): Take a masked_p parameter and update
9531 call to get_group_load_store_type.
9532 (vectorizable_store): Update call to get_load_store_type.
9533 Handle IFN_MASK_STORE_LANES.
9534 (vectorizable_load): Update call to get_load_store_type.
9535 Handle IFN_MASK_LOAD_LANES.
9537 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9538 Alan Hayward <alan.hayward@arm.com>
9539 David Sherwood <david.sherwood@arm.com>
9541 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9543 * config/aarch64/aarch64-protos.h
9544 (aarch64_sve_struct_memory_operand_p): Declare.
9545 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9546 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9547 (VPRED, vpred): Handle SVE structure modes.
9548 * config/aarch64/constraints.md (Utx): New constraint.
9549 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9550 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9551 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9552 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9553 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9554 structure modes. Split into pieces after RA.
9555 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9556 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9558 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9559 SVE structure modes.
9560 (aarch64_classify_address): Likewise.
9561 (sizetochar): Move earlier in file.
9562 (aarch64_print_operand): Handle SVE register lists.
9563 (aarch64_array_mode): New function.
9564 (aarch64_sve_struct_memory_operand_p): Likewise.
9565 (TARGET_ARRAY_MODE): Redefine.
9567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9568 Alan Hayward <alan.hayward@arm.com>
9569 David Sherwood <david.sherwood@arm.com>
9571 * target.def (array_mode): New target hook.
9572 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9573 * doc/tm.texi: Regenerate.
9574 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9575 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9576 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9578 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9581 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9582 Alan Hayward <alan.hayward@arm.com>
9583 David Sherwood <david.sherwood@arm.com>
9585 * fold-const.c (fold_binary_loc): Check the argument types
9586 rather than the result type when testing for a vector operation.
9588 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9590 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9591 * doc/tm.texi: Regenerate.
9593 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9594 Alan Hayward <alan.hayward@arm.com>
9595 David Sherwood <david.sherwood@arm.com>
9597 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9598 (sve): Document new AArch64 extension.
9599 * doc/md.texi (w): Extend the description of the AArch64
9600 constraint to include SVE vectors.
9601 (Upl, Upa): Document new AArch64 predicate constraints.
9602 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9604 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9605 (msve-vector-bits=): New option.
9606 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9607 SVE when these are disabled.
9608 (sve): New extension.
9609 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9610 modes. Adjust their number of units based on aarch64_sve_vg.
9611 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9612 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9613 aarch64_addr_query_type.
9614 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9615 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9616 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9617 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9618 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9619 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9620 (aarch64_simd_imm_zero_p): Delete.
9621 (aarch64_check_zero_based_sve_index_immediate): Declare.
9622 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9623 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9624 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9625 (aarch64_sve_float_mul_immediate_p): Likewise.
9626 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9628 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9629 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9630 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9631 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9632 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9633 (aarch64_regmode_natural_size): Likewise.
9634 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9635 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9637 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9638 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9639 for VG and the SVE predicate registers.
9640 (V_ALIASES): Add a "z"-prefixed alias.
9641 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9642 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9643 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9644 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9645 (REG_CLASS_NAMES): Add entries for them.
9646 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9647 and the predicate registers.
9648 (aarch64_sve_vg): Declare.
9649 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9650 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9651 (REGMODE_NATURAL_SIZE): Define.
9652 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9654 * config/aarch64/aarch64.c: Include cfgrtl.h.
9655 (simd_immediate_info): Add a constructor for series vectors,
9656 and an associated step field.
9657 (aarch64_sve_vg): New variable.
9658 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9659 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9660 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9661 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9662 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9663 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9664 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9665 (aarch64_get_mask_mode): New functions.
9666 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9667 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9668 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9669 predicate modes and predicate registers. Explicitly restrict
9670 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9671 to store a vector mode if it is recognized by
9672 aarch64_classify_vector_mode.
9673 (aarch64_regmode_natural_size): New function.
9674 (aarch64_hard_regno_caller_save_mode): Return the original mode
9676 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9677 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9678 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9679 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9681 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9682 does not overlap dest if the function is frame-related. Handle
9684 (aarch64_split_add_offset): New function.
9685 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9686 them aarch64_add_offset.
9687 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9688 and update call to aarch64_sub_sp.
9689 (aarch64_add_cfa_expression): New function.
9690 (aarch64_expand_prologue): Pass extra temporary registers to the
9691 functions above. Handle the case in which we need to emit new
9692 DW_CFA_expressions for registers that were originally saved
9693 relative to the stack pointer, but now have to be expressed
9694 relative to the frame pointer.
9695 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9697 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9698 IP0 and IP1 values for SVE frames.
9699 (aarch64_expand_vec_series): New function.
9700 (aarch64_expand_sve_widened_duplicate): Likewise.
9701 (aarch64_expand_sve_const_vector): Likewise.
9702 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9703 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9704 into the register, rather than emitting a SET directly.
9705 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9706 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9707 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9708 (offset_9bit_signed_scaled_p): New functions.
9709 (aarch64_replicate_bitmask_imm): New function.
9710 (aarch64_bitmask_imm): Use it.
9711 (aarch64_cannot_force_const_mem): Reject expressions involving
9712 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9713 (aarch64_classify_index): Handle SVE indices, by requiring
9714 a plain register index with a scale that matches the element size.
9715 (aarch64_classify_address): Handle SVE addresses. Assert that
9716 the mode of the address is VOIDmode or an integer mode.
9717 Update call to aarch64_classify_symbol.
9718 (aarch64_classify_symbolic_expression): Update call to
9719 aarch64_classify_symbol.
9720 (aarch64_const_vec_all_in_range_p): New function.
9721 (aarch64_print_vector_float_operand): Likewise.
9722 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9723 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9724 and the FP immediates 1.0 and 0.5.
9725 (aarch64_print_address_internal): Handle SVE addresses.
9726 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9727 (aarch64_regno_regclass): Handle predicate registers.
9728 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9730 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9731 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9732 (aarch64_convert_sve_vector_bits): New function.
9733 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9734 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9736 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9737 Handle SVE vector and predicate modes. Accept VL-based constants
9738 that need only one temporary register, and VL offsets that require
9739 no temporary registers.
9740 (aarch64_conditional_register_usage): Mark the predicate registers
9741 as fixed if SVE isn't available.
9742 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9743 Return true for SVE vector and predicate modes.
9744 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9745 rather than an unsigned int. Handle SVE modes.
9746 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9748 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9750 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9751 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9752 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9753 (aarch64_sve_float_mul_immediate_p): New functions.
9754 (aarch64_sve_valid_immediate): New function.
9755 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9756 Explicitly reject structure modes. Check for INDEX constants.
9757 Handle PTRUE and PFALSE constants.
9758 (aarch64_check_zero_based_sve_index_immediate): New function.
9759 (aarch64_simd_imm_zero_p): Delete.
9760 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9761 vector modes. Accept constants in the range of CNT[BHWD].
9762 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9763 ask for an Advanced SIMD mode.
9764 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9765 (aarch64_simd_vector_alignment): Handle SVE predicates.
9766 (aarch64_vectorize_preferred_vector_alignment): New function.
9767 (aarch64_simd_vector_alignment_reachable): Use it instead of
9769 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9770 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9772 (MAX_VECT_LEN): Delete.
9773 (expand_vec_perm_d): Add a vec_flags field.
9774 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9775 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9776 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9778 (aarch64_evpc_rev): Rename to...
9779 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9780 (aarch64_evpc_rev_global): New function.
9781 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9782 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9784 (aarch64_evpc_sve_tbl): New function.
9785 (aarch64_expand_vec_perm_const_1): Update after rename of
9786 aarch64_evpc_rev. Handle SVE permutes too, trying
9787 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9788 than aarch64_evpc_tbl.
9789 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9790 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9791 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9792 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9793 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9794 (aarch64_expand_sve_vcond): New functions.
9795 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9796 of aarch64_vector_mode_p.
9797 (aarch64_dwarf_poly_indeterminate_value): New function.
9798 (aarch64_compute_pressure_classes): Likewise.
9799 (aarch64_can_change_mode_class): Likewise.
9800 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9801 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9802 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9803 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9804 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9805 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9806 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9807 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9809 (Dn, Dl, Dr): Accept const as well as const_vector.
9810 (Dz): Likewise. Compare against CONST0_RTX.
9811 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9812 of "vector" where appropriate.
9813 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9814 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9815 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9816 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9817 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9818 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9819 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9820 (v_int_equiv): Extend to SVE modes.
9821 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9823 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9824 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9825 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9826 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9827 (SVE_COND_FP_CMP): New int iterators.
9828 (perm_hilo): Handle the new unpack unspecs.
9829 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9831 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9832 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9833 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9834 (aarch64_equality_operator, aarch64_constant_vector_operand)
9835 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9836 (aarch64_sve_nonimmediate_operand): Likewise.
9837 (aarch64_sve_general_operand): Likewise.
9838 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9839 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9840 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9841 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9842 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9843 (aarch64_sve_float_arith_immediate): Likewise.
9844 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9845 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9846 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9847 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9848 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9849 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9850 (aarch64_sve_float_arith_operand): Likewise.
9851 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9852 (aarch64_sve_float_mul_operand): Likewise.
9853 (aarch64_sve_vec_perm_operand): Likewise.
9854 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9855 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9856 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9857 as well as const_vector.
9858 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9859 in file. Use CONST0_RTX and CONSTM1_RTX.
9860 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9861 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9862 Use aarch64_simd_imm_zero.
9863 * config/aarch64/aarch64-sve.md: New file.
9864 * config/aarch64/aarch64.md: Include it.
9865 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9866 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9867 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9868 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9869 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9870 (sve): New attribute.
9871 (enabled): Disable instructions with the sve attribute unless
9873 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9874 aarch64_expand_mov_immediate.
9875 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9876 CNT[BHSD] immediates.
9877 (movti): Split CONST_POLY_INT moves into two halves.
9878 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9879 Split additions that need a temporary here if the destination
9880 is the stack pointer.
9881 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9882 (*add<mode>3_poly_1): New instruction.
9883 (set_clobber_cc): New expander.
9885 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9887 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9888 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9889 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9890 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9891 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9892 Change innermode from fixed_mode_size to machine_mode.
9893 (simplify_subreg): Update call accordingly. Handle a constant-sized
9894 subreg of a variable-length CONST_VECTOR.
9896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9897 Alan Hayward <alan.hayward@arm.com>
9898 David Sherwood <david.sherwood@arm.com>
9900 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9901 (add_offset_to_base): New function, split out from...
9902 (create_mem_ref): ...here. When handling a scale other than 1,
9903 check first whether the address is valid without the offset.
9904 Add it into the base if so, leaving the index and scale as-is.
9906 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9909 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9910 fold_for_warn before checking if arg2 is INTEGER_CST.
9912 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9914 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9915 (store_multiple_operation): Delete.
9916 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9917 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9918 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9919 guarded by TARGET_STRING.
9920 (rs6000_output_load_multiple): Delete.
9921 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9922 OPTION_MASK_STRING / TARGET_STRING handling.
9923 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9924 (const rs6000_opt_masks) <"string">: Change mask to 0.
9925 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9926 (MASK_STRING): Delete.
9927 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9929 (load_multiple): Delete.
9936 (store_multiple): Delete.
9943 (movmemsi_8reg): Delete.
9944 (corresponding unnamed define_insn): Delete.
9945 (movmemsi_6reg): Delete.
9946 (corresponding unnamed define_insn): Delete.
9947 (movmemsi_4reg): Delete.
9948 (corresponding unnamed define_insn): Delete.
9949 (movmemsi_2reg): Delete.
9950 (corresponding unnamed define_insn): Delete.
9951 (movmemsi_1reg): Delete.
9952 (corresponding unnamed define_insn): Delete.
9953 * config/rs6000/rs6000.opt (mno-string): New.
9954 (mstring): Replace by deprecation warning stub.
9955 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9957 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9959 * regrename.c (regrename_do_replace): If replacing the same
9960 reg multiple times, try to reuse last created gen_raw_REG.
9963 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9964 main to workaround a bug in GDB.
9966 2018-01-12 Tom de Vries <tom@codesourcery.com>
9969 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9971 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9973 PR rtl-optimization/80481
9974 * ira-color.c (get_cap_member): New function.
9975 (allocnos_conflict_by_live_ranges_p): Use it.
9976 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9977 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9979 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9982 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9983 (*saddl_se_1): Ditto.
9985 (*ssubl_se_1): Ditto.
9987 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9989 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9990 rather than wi::to_widest for DR_INITs.
9991 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9992 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9993 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9995 (vect_analyze_group_access_1): Note that here.
9997 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9999 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10000 polynomial type sizes.
10002 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10004 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10005 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10006 (gimple_add_tmp_var): Likewise.
10008 2018-01-12 Martin Liska <mliska@suse.cz>
10010 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10011 (gimple_alloc_sizes): Likewise.
10012 (dump_gimple_statistics): Use PRIu64 in printf format.
10013 * gimple.h: Change uint64_t to int.
10015 2018-01-12 Martin Liska <mliska@suse.cz>
10017 * tree-core.h: Use uint64_t instead of int.
10018 * tree.c (tree_node_counts): Likewise.
10019 (tree_node_sizes): Likewise.
10020 (dump_tree_statistics): Use PRIu64 in printf format.
10022 2018-01-12 Martin Liska <mliska@suse.cz>
10024 * Makefile.in: As qsort_chk is implemented in vec.c, add
10025 vec.o to linkage of gencfn-macros.
10026 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10027 passing the info to record_node_allocation_statistics.
10028 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10030 * ggc-common.c (struct ggc_usage): Add operator== and use
10031 it in operator< and compare function.
10032 * mem-stats.h (struct mem_usage): Likewise.
10033 * vec.c (struct vec_usage): Remove operator< and compare
10034 function. Can be simply inherited.
10036 2018-01-12 Martin Jambor <mjambor@suse.cz>
10039 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10040 * tree-ssa-math-opts.c: Include domwalk.h.
10041 (convert_mult_to_fma_1): New function.
10042 (fma_transformation_info): New type.
10043 (fma_deferring_state): Likewise.
10044 (cancel_fma_deferring): New function.
10045 (result_of_phi): Likewise.
10046 (last_fma_candidate_feeds_initial_phi): Likewise.
10047 (convert_mult_to_fma): Added deferring logic, split actual
10048 transformation to convert_mult_to_fma_1.
10049 (math_opts_dom_walker): New type.
10050 (math_opts_dom_walker::after_dom_children): New method, body moved
10051 here from pass_optimize_widening_mul::execute, added deferring logic
10053 (pass_optimize_widening_mul::execute): Moved most of code to
10054 math_opts_dom_walker::after_dom_children.
10055 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10056 * config/i386/i386.c (ix86_option_override_internal): Added
10057 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10059 2018-01-12 Richard Biener <rguenther@suse.de>
10062 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10063 inline instance vars.
10065 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10068 * config/rx/rx.c (rx_is_restricted_memory_address):
10069 Handle SUBREG case.
10071 2018-01-12 Richard Biener <rguenther@suse.de>
10073 PR tree-optimization/80846
10074 * target.def (split_reduction): New target hook.
10075 * targhooks.c (default_split_reduction): New function.
10076 * targhooks.h (default_split_reduction): Declare.
10077 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10078 target requests first reduce vectors by combining low and high
10080 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10081 (get_vectype_for_scalar_type_and_size): Export.
10082 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10083 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10084 * doc/tm.texi: Regenerate.
10085 * config/i386/i386.c (ix86_split_reduction): Implement
10086 TARGET_VECTORIZE_SPLIT_REDUCTION.
10088 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10091 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10092 in PIC mode except for TARGET_VXWORKS_RTP.
10093 * config/sparc/sparc.c: Include cfgrtl.h.
10094 (TARGET_INIT_PIC_REG): Define.
10095 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10096 (sparc_pic_register_p): New predicate.
10097 (sparc_legitimate_address_p): Use it.
10098 (sparc_legitimize_pic_address): Likewise.
10099 (sparc_delegitimize_address): Likewise.
10100 (sparc_mode_dependent_address_p): Likewise.
10101 (gen_load_pcrel_sym): Remove 4th parameter.
10102 (load_got_register): Adjust call to above. Remove obsolete stuff.
10103 (sparc_expand_prologue): Do not call load_got_register here.
10104 (sparc_flat_expand_prologue): Likewise.
10105 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10106 (sparc_use_pseudo_pic_reg): New function.
10107 (sparc_init_pic_reg): Likewise.
10108 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10109 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10111 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10113 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10114 Add item for branch_cost.
10116 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10118 PR rtl-optimization/83565
10119 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10120 not extend the result to a larger mode for rotate operations.
10121 (num_sign_bit_copies1): Likewise.
10123 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10126 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10128 Use values-Xc.o for -pedantic.
10129 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10131 2018-01-12 Martin Liska <mliska@suse.cz>
10134 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10136 (possible_polymorphic_call_targets): Use it.
10137 (ipa_devirt): Likewise.
10139 2018-01-12 Martin Liska <mliska@suse.cz>
10141 * profile-count.h (enum profile_quality): Use 0 as invalid
10142 enum value of profile_quality.
10144 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10146 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10147 -mext-string options.
10149 2018-01-12 Richard Biener <rguenther@suse.de>
10151 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10152 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10153 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10155 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10157 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10159 * configure.ac (--with-long-double-format): Add support for the
10160 configuration option to change the default long double format on
10162 * config.gcc (powerpc*-linux*-*): Likewise.
10163 * configure: Regenerate.
10164 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10165 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10166 used without modification.
10168 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10170 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10171 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10172 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10173 MISC_BUILTIN_SPEC_BARRIER.
10174 (rs6000_init_builtins): Likewise.
10175 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10177 (speculation_barrier): New define_insn.
10178 * doc/extend.texi: Document __builtin_speculation_barrier.
10180 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10183 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10184 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10185 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10187 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10188 integral modes instead of "ss" and "sd".
10189 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10190 vectors with 32-bit and 64-bit elements.
10191 (vecdupssescalarmodesuffix): New mode attribute.
10192 (vec_dup<mode>): Use it.
10194 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10197 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10198 frame if argument is passed on stack.
10200 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10203 * ree.c (combine_reaching_defs): Optimize also
10204 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10205 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10207 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10209 PR middle-end/83189
10210 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10212 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10214 PR middle-end/83718
10215 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10216 after they are computed.
10218 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10220 PR tree-optimization/83695
10221 * gimple-loop-linterchange.cc
10222 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10223 reset cached scev information after interchange.
10224 (pass_linterchange::execute): Remove call to scev_reset_htab.
10226 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10228 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10229 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10230 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10231 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10232 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10233 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10234 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10235 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10236 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10237 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10238 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10239 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10240 (V_lane_reg): Likewise.
10241 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10243 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10244 (vfmal_lane_low<mode>_intrinsic,
10245 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10246 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10247 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10248 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10249 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10250 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10252 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10254 * config/arm/arm-cpus.in (fp16fml): New feature.
10255 (ALL_SIMD): Add fp16fml.
10256 (armv8.2-a): Add fp16fml as an option.
10257 (armv8.3-a): Likewise.
10258 (armv8.4-a): Add fp16fml as part of fp16.
10259 * config/arm/arm.h (TARGET_FP16FML): Define.
10260 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10262 * config/arm/arm-modes.def (V2HF): Define.
10263 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10264 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10265 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10266 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10267 vfmsl_low, vfmsl_high): New set of builtins.
10268 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10269 (vfml_op): New code attribute.
10270 (VFMLHALVES): New int iterator.
10271 (VFML, VFMLSEL): New mode attributes.
10272 (V_reg): Define mapping for V2HF.
10273 (V_hi, V_lo): New mode attributes.
10274 (VF_constraint): Likewise.
10275 (vfml_half, vfml_half_selector): New int attributes.
10276 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10278 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10279 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10281 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10282 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10283 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10284 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10286 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10287 Document new effective target and option set.
10289 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10291 * config/arm/arm-cpus.in (armv8_4): New feature.
10292 (ARMv8_4a): New fgroup.
10293 (armv8.4-a): New arch.
10294 * config/arm/arm-tables.opt: Regenerate.
10295 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10296 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10297 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10298 Add matching rules for -march=armv8.4-a and extensions.
10299 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10301 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10304 * config/rx/rx.md (BW): New mode attribute.
10305 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10307 2018-01-11 Richard Biener <rguenther@suse.de>
10309 PR tree-optimization/83435
10310 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10311 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10312 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10314 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10315 Alan Hayward <alan.hayward@arm.com>
10316 David Sherwood <david.sherwood@arm.com>
10318 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10320 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10321 (aarch64_print_address_internal): Use it to check for a zero offset.
10323 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10324 Alan Hayward <alan.hayward@arm.com>
10325 David Sherwood <david.sherwood@arm.com>
10327 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10328 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10329 Return a poly_int64 rather than a HOST_WIDE_INT.
10330 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10331 rather than a HOST_WIDE_INT.
10332 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10333 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10334 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10335 final_offset from HOST_WIDE_INT to poly_int64.
10336 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10337 to_constant when getting the number of units in an Advanced SIMD
10339 (aarch64_builtin_vectorized_function): Check for a constant number
10341 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10343 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10344 attribute instead of GET_MODE_NUNITS.
10345 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10346 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10347 GET_MODE_SIZE for fixed-size registers.
10348 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10349 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10350 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10351 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10352 (aarch64_print_operand, aarch64_print_address_internal)
10353 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10354 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10355 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10356 Handle polynomial GET_MODE_SIZE.
10357 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10358 wider than SImode without modification.
10359 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10360 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10361 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10362 passing and returning SVE modes.
10363 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10364 rather than GEN_INT.
10365 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10366 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10367 (aarch64_allocate_and_probe_stack_space): Likewise.
10368 (aarch64_layout_frame): Cope with polynomial offsets.
10369 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10370 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10371 polynomial offsets.
10372 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10373 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10374 poly_int64 rather than a HOST_WIDE_INT.
10375 (aarch64_get_separate_components, aarch64_process_components)
10376 (aarch64_expand_prologue, aarch64_expand_epilogue)
10377 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10378 (aarch64_anchor_offset): New function, split out from...
10379 (aarch64_legitimize_address): ...here.
10380 (aarch64_builtin_vectorization_cost): Handle polynomial
10381 TYPE_VECTOR_SUBPARTS.
10382 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10384 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10385 number of elements from the PARALLEL rather than the mode.
10386 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10387 rather than GET_MODE_BITSIZE.
10388 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10389 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10390 (aarch64_expand_vec_perm_const_1): Handle polynomial
10391 d->perm.length () and d->perm elements.
10392 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10393 Apply to_constant to d->perm elements.
10394 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10395 polynomial CONST_VECTOR_NUNITS.
10396 (aarch64_move_pointer): Take amount as a poly_int64 rather
10398 (aarch64_progress_pointer): Avoid temporary variable.
10399 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10400 the mode attribute instead of GET_MODE.
10402 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10403 Alan Hayward <alan.hayward@arm.com>
10404 David Sherwood <david.sherwood@arm.com>
10406 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10407 x exists before using it.
10408 (aarch64_add_constant_internal): Rename to...
10409 (aarch64_add_offset_1): ...this. Replace regnum with separate
10410 src and dest rtxes. Handle the case in which they're different,
10411 including when the offset is zero. Replace scratchreg with an rtx.
10412 Use 2 additions if there is no spare register into which we can
10413 move a 16-bit constant.
10414 (aarch64_add_constant): Delete.
10415 (aarch64_add_offset): Replace reg with separate src and dest
10416 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10417 Use aarch64_add_offset_1.
10418 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10419 an rtx rather than an int. Take the delta as a poly_int64
10420 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10421 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10422 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10423 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10424 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10425 and aarch64_add_sp.
10426 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10427 aarch64_add_constant.
10429 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10431 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10432 Use scalar_float_mode.
10434 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10436 * config/aarch64/aarch64-simd.md
10437 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10438 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10439 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10440 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10441 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10442 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10443 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10444 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10445 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10446 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10448 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10451 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10452 targ_options->x_arm_arch_string is non NULL.
10454 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10456 * config/aarch64/aarch64.h
10457 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10459 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10462 * expmed.c (emit_store_flag_force): Swap if const op0
10463 and change VOIDmode to mode of op0.
10465 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10467 PR rtl-optimization/83761
10468 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10469 than bytes to mode_for_size.
10471 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10473 PR middle-end/83189
10474 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10475 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10478 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10480 PR middle-end/83575
10481 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10482 when in layout mode.
10483 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10484 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10487 2018-01-10 Michael Collison <michael.collison@arm.com>
10489 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10490 * config/aarch64/aarch64-option-extension.def: Add
10491 AARCH64_OPT_EXTENSION of 'fp16fml'.
10492 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10493 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10494 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10495 * config/aarch64/constraints.md (Ui7): New constraint.
10496 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10497 (VFMLA_SEL_W): Ditto.
10500 (VFMLA16_LOW): New int iterator.
10501 (VFMLA16_HIGH): Ditto.
10502 (UNSPEC_FMLAL): New unspec.
10503 (UNSPEC_FMLSL): Ditto.
10504 (UNSPEC_FMLAL2): Ditto.
10505 (UNSPEC_FMLSL2): Ditto.
10506 (f16mac): New code attribute.
10507 * config/aarch64/aarch64-simd-builtins.def
10508 (aarch64_fmlal_lowv2sf): Ditto.
10509 (aarch64_fmlsl_lowv2sf): Ditto.
10510 (aarch64_fmlalq_lowv4sf): Ditto.
10511 (aarch64_fmlslq_lowv4sf): Ditto.
10512 (aarch64_fmlal_highv2sf): Ditto.
10513 (aarch64_fmlsl_highv2sf): Ditto.
10514 (aarch64_fmlalq_highv4sf): Ditto.
10515 (aarch64_fmlslq_highv4sf): Ditto.
10516 (aarch64_fmlal_lane_lowv2sf): Ditto.
10517 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10518 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10519 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10520 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10521 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10522 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10523 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10524 (aarch64_fmlal_lane_highv2sf): Ditto.
10525 (aarch64_fmlsl_lane_highv2sf): Ditto.
10526 (aarch64_fmlal_laneq_highv2sf): Ditto.
10527 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10528 (aarch64_fmlalq_lane_highv4sf): Ditto.
10529 (aarch64_fmlsl_lane_highv4sf): Ditto.
10530 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10531 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10532 * config/aarch64/aarch64-simd.md:
10533 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10534 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10535 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10536 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10537 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10538 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10539 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10540 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10541 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10542 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10543 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10544 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10545 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10546 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10547 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10548 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10549 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10550 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10551 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10552 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10553 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10554 (vfmlsl_low_u32): Ditto.
10555 (vfmlalq_low_u32): Ditto.
10556 (vfmlslq_low_u32): Ditto.
10557 (vfmlal_high_u32): Ditto.
10558 (vfmlsl_high_u32): Ditto.
10559 (vfmlalq_high_u32): Ditto.
10560 (vfmlslq_high_u32): Ditto.
10561 (vfmlal_lane_low_u32): Ditto.
10562 (vfmlsl_lane_low_u32): Ditto.
10563 (vfmlal_laneq_low_u32): Ditto.
10564 (vfmlsl_laneq_low_u32): Ditto.
10565 (vfmlalq_lane_low_u32): Ditto.
10566 (vfmlslq_lane_low_u32): Ditto.
10567 (vfmlalq_laneq_low_u32): Ditto.
10568 (vfmlslq_laneq_low_u32): Ditto.
10569 (vfmlal_lane_high_u32): Ditto.
10570 (vfmlsl_lane_high_u32): Ditto.
10571 (vfmlal_laneq_high_u32): Ditto.
10572 (vfmlsl_laneq_high_u32): Ditto.
10573 (vfmlalq_lane_high_u32): Ditto.
10574 (vfmlslq_lane_high_u32): Ditto.
10575 (vfmlalq_laneq_high_u32): Ditto.
10576 (vfmlslq_laneq_high_u32): Ditto.
10577 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10578 (AARCH64_FL_FOR_ARCH8_4): New.
10579 (AARCH64_ISA_F16FML): New ISA flag.
10580 (TARGET_F16FML): New feature flag for fp16fml.
10581 (doc/invoke.texi): Document new fp16fml option.
10583 2018-01-10 Michael Collison <michael.collison@arm.com>
10585 * config/aarch64/aarch64-builtins.c:
10586 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10587 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10588 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10589 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10590 (AARCH64_ISA_SHA3): New ISA flag.
10591 (TARGET_SHA3): New feature flag for sha3.
10592 * config/aarch64/iterators.md (sha512_op): New int attribute.
10593 (CRYPTO_SHA512): New int iterator.
10594 (UNSPEC_SHA512H): New unspec.
10595 (UNSPEC_SHA512H2): Ditto.
10596 (UNSPEC_SHA512SU0): Ditto.
10597 (UNSPEC_SHA512SU1): Ditto.
10598 * config/aarch64/aarch64-simd-builtins.def
10599 (aarch64_crypto_sha512hqv2di): New builtin.
10600 (aarch64_crypto_sha512h2qv2di): Ditto.
10601 (aarch64_crypto_sha512su0qv2di): Ditto.
10602 (aarch64_crypto_sha512su1qv2di): Ditto.
10603 (aarch64_eor3qv8hi): Ditto.
10604 (aarch64_rax1qv2di): Ditto.
10605 (aarch64_xarqv2di): Ditto.
10606 (aarch64_bcaxqv8hi): Ditto.
10607 * config/aarch64/aarch64-simd.md:
10608 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10609 (aarch64_crypto_sha512su0qv2di): Ditto.
10610 (aarch64_crypto_sha512su1qv2di): Ditto.
10611 (aarch64_eor3qv8hi): Ditto.
10612 (aarch64_rax1qv2di): Ditto.
10613 (aarch64_xarqv2di): Ditto.
10614 (aarch64_bcaxqv8hi): Ditto.
10615 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10616 (vsha512h2q_u64): Ditto.
10617 (vsha512su0q_u64): Ditto.
10618 (vsha512su1q_u64): Ditto.
10619 (veor3q_u16): Ditto.
10620 (vrax1q_u64): Ditto.
10621 (vxarq_u64): Ditto.
10622 (vbcaxq_u16): Ditto.
10623 * config/arm/types.md (crypto_sha512): New type attribute.
10624 (crypto_sha3): Ditto.
10625 (doc/invoke.texi): Document new sha3 option.
10627 2018-01-10 Michael Collison <michael.collison@arm.com>
10629 * config/aarch64/aarch64-builtins.c:
10630 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10631 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10632 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10633 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10634 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10635 (AARCH64_ISA_SM4): New ISA flag.
10636 (TARGET_SM4): New feature flag for sm4.
10637 * config/aarch64/aarch64-simd-builtins.def
10638 (aarch64_sm3ss1qv4si): Ditto.
10639 (aarch64_sm3tt1aq4si): Ditto.
10640 (aarch64_sm3tt1bq4si): Ditto.
10641 (aarch64_sm3tt2aq4si): Ditto.
10642 (aarch64_sm3tt2bq4si): Ditto.
10643 (aarch64_sm3partw1qv4si): Ditto.
10644 (aarch64_sm3partw2qv4si): Ditto.
10645 (aarch64_sm4eqv4si): Ditto.
10646 (aarch64_sm4ekeyqv4si): Ditto.
10647 * config/aarch64/aarch64-simd.md:
10648 (aarch64_sm3ss1qv4si): Ditto.
10649 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10650 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10651 (aarch64_sm4eqv4si): Ditto.
10652 (aarch64_sm4ekeyqv4si): Ditto.
10653 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10654 (sm3part_op): Ditto.
10655 (CRYPTO_SM3TT): Ditto.
10656 (CRYPTO_SM3PART): Ditto.
10657 (UNSPEC_SM3SS1): New unspec.
10658 (UNSPEC_SM3TT1A): Ditto.
10659 (UNSPEC_SM3TT1B): Ditto.
10660 (UNSPEC_SM3TT2A): Ditto.
10661 (UNSPEC_SM3TT2B): Ditto.
10662 (UNSPEC_SM3PARTW1): Ditto.
10663 (UNSPEC_SM3PARTW2): Ditto.
10664 (UNSPEC_SM4E): Ditto.
10665 (UNSPEC_SM4EKEY): Ditto.
10666 * config/aarch64/constraints.md (Ui2): New constraint.
10667 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10668 * config/arm/types.md (crypto_sm3): New type attribute.
10669 (crypto_sm4): Ditto.
10670 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10671 (vsm3tt1aq_u32): Ditto.
10672 (vsm3tt1bq_u32): Ditto.
10673 (vsm3tt2aq_u32): Ditto.
10674 (vsm3tt2bq_u32): Ditto.
10675 (vsm3partw1q_u32): Ditto.
10676 (vsm3partw2q_u32): Ditto.
10677 (vsm4eq_u32): Ditto.
10678 (vsm4ekeyq_u32): Ditto.
10679 (doc/invoke.texi): Document new sm4 option.
10681 2018-01-10 Michael Collison <michael.collison@arm.com>
10683 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10684 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10685 (AARCH64_FL_FOR_ARCH8_4): New.
10686 (AARCH64_FL_V8_4): New flag.
10687 (doc/invoke.texi): Document new armv8.4-a option.
10689 2018-01-10 Michael Collison <michael.collison@arm.com>
10691 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10692 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10693 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10694 * config/aarch64/aarch64-option-extension.def: Add
10695 AARCH64_OPT_EXTENSION of 'sha2'.
10696 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10697 (crypto): Disable sha2 and aes if crypto disabled.
10698 (crypto): Enable aes and sha2 if enabled.
10699 (simd): Disable sha2 and aes if simd disabled.
10700 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10702 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10703 (TARGET_SHA2): New feature flag for sha2.
10704 (TARGET_AES): New feature flag for aes.
10705 * config/aarch64/aarch64-simd.md:
10706 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10707 conditional on TARGET_AES.
10708 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10709 (aarch64_crypto_sha1hsi): Make pattern conditional
10711 (aarch64_crypto_sha1hv4si): Ditto.
10712 (aarch64_be_crypto_sha1hv4si): Ditto.
10713 (aarch64_crypto_sha1su1v4si): Ditto.
10714 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10715 (aarch64_crypto_sha1su0v4si): Ditto.
10716 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10717 (aarch64_crypto_sha256su0v4si): Ditto.
10718 (aarch64_crypto_sha256su1v4si): Ditto.
10719 (doc/invoke.texi): Document new aes and sha2 options.
10721 2018-01-10 Martin Sebor <msebor@redhat.com>
10723 PR tree-optimization/83781
10724 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10727 2018-01-11 Martin Sebor <msebor@gmail.com>
10728 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10730 PR tree-optimization/83501
10731 PR tree-optimization/81703
10733 * tree-ssa-strlen.c (get_string_cst): Rename...
10734 (get_string_len): ...to this. Handle global constants.
10735 (handle_char_store): Adjust.
10737 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10738 Jim Wilson <jimw@sifive.com>
10740 * config/riscv/riscv-protos.h (riscv_output_return): New.
10741 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10742 (riscv_attribute_table, riscv_output_return),
10743 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10744 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10745 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10747 (riscv_expand_prologue): Add early return for naked function.
10748 (riscv_expand_epilogue): Likewise.
10749 (riscv_function_ok_for_sibcall): Return false for naked function.
10750 (riscv_set_current_function): New.
10751 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10752 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10753 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10754 * doc/extend.texi (RISC-V Function Attributes): New.
10756 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10758 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10759 check for 128-bit long double before checking TCmode.
10760 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10761 128-bit long doubles before checking TFmode or TCmode.
10762 (FLOAT128_IBM_P): Likewise.
10764 2018-01-10 Martin Sebor <msebor@redhat.com>
10766 PR tree-optimization/83671
10767 * builtins.c (c_strlen): Unconditionally return zero for the empty
10769 Use -Warray-bounds for warnings.
10770 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10771 for non-constant array indices with COMPONENT_REF, arrays of
10772 arrays, and pointers to arrays.
10773 (gimple_fold_builtin_strlen): Determine and set length range for
10774 non-constant character arrays.
10776 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10778 PR middle-end/81897
10779 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10782 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10784 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10786 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10789 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10790 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10791 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10792 indexed_or_indirect_operand predicate.
10793 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10794 (*vsx_le_perm_load_v8hi): Likewise.
10795 (*vsx_le_perm_load_v16qi): Likewise.
10796 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10797 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10798 (*vsx_le_perm_store_v8hi): Likewise.
10799 (*vsx_le_perm_store_v16qi): Likewise.
10800 (eight unnamed splitters): Likewise.
10802 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10804 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10805 * config/rs6000/emmintrin.h: Likewise.
10806 * config/rs6000/mmintrin.h: Likewise.
10807 * config/rs6000/xmmintrin.h: Likewise.
10809 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10812 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10814 * tree.c (tree_nop_conversion): Return true for location wrapper
10816 (maybe_wrap_with_location): New function.
10817 (selftest::check_strip_nops): New function.
10818 (selftest::test_location_wrappers): New function.
10819 (selftest::tree_c_tests): Call it.
10820 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10821 (maybe_wrap_with_location): New decl.
10822 (EXPR_LOCATION_WRAPPER_P): New macro.
10823 (location_wrapper_p): New inline function.
10824 (tree_strip_any_location_wrapper): New inline function.
10826 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10829 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10830 stack_realign_offset for the largest alignment of stack slot
10832 (ix86_find_max_used_stack_alignment): New function.
10833 (ix86_finalize_stack_frame_flags): Use it. Set
10834 max_used_stack_alignment if we don't realign stack.
10835 * config/i386/i386.h (machine_function): Add
10836 max_used_stack_alignment.
10838 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10840 * config/arm/arm.opt (-mbranch-cost): New option.
10841 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10844 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10847 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10848 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10850 2018-01-10 Richard Biener <rguenther@suse.de>
10853 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10854 early out so it also covers the case where we have a non-NULL
10857 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10859 PR tree-optimization/83753
10860 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10861 for non-strided grouped accesses if the number of elements is 1.
10863 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10866 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10867 * i386.h (TARGET_USE_GATHER): Define.
10868 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10870 2018-01-10 Martin Liska <mliska@suse.cz>
10873 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10874 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10876 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10877 CLEANUP_NO_PARTITIONING is not set.
10879 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10881 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10882 for vectors, as a partial revert of r254296.
10883 * rtl.h (const_vec_p): Delete.
10884 (const_vec_duplicate_p): Don't test for vector CONSTs.
10885 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10886 * expmed.c (make_tree): Likewise.
10889 * common.md (E, F): Use CONSTANT_P instead of checking for
10891 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10892 checking for CONST_VECTOR.
10894 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10896 PR middle-end/83575
10897 * predict.c (force_edge_cold): Handle in more sane way edges
10898 with no prediction.
10900 2018-01-09 Carl Love <cel@us.ibm.com>
10902 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10904 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10905 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10906 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10907 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10908 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10909 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10910 * config/rs6000/rs6000-protos.h: Add extern defition for
10911 rs6000_generate_float2_double_code.
10912 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10914 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10915 (float2_v2df): Add define_expand.
10917 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10920 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10921 op_mode in the force_to_mode call.
10923 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10925 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10926 instead of checking each element individually.
10927 (aarch64_evpc_uzp): Likewise.
10928 (aarch64_evpc_zip): Likewise.
10929 (aarch64_evpc_ext): Likewise.
10930 (aarch64_evpc_rev): Likewise.
10931 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10932 instead of checking each element individually. Return true without
10934 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10935 whether all selected elements come from the same input, instead of
10936 checking each element individually. Remove calls to gen_rtx_REG,
10937 start_sequence and end_sequence and instead assert that no rtl is
10940 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10942 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10943 order of HIGH and CONST checks.
10945 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10947 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10948 if the destination isn't an SSA_NAME.
10950 2018-01-09 Richard Biener <rguenther@suse.de>
10952 PR tree-optimization/83668
10953 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10955 (canonicalize_loop_form): ... here, renamed from ...
10956 (canonicalize_loop_closed_ssa_form): ... this and amended to
10957 swap successor edges for loop exit blocks to make us use
10958 the RPO order we need for initial schedule generation.
10960 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10962 PR tree-optimization/64811
10963 * match.pd: When optimizing comparisons with Inf, avoid
10964 introducing or losing exceptions from comparisons with NaN.
10966 2018-01-09 Martin Liska <mliska@suse.cz>
10969 * asan.c (shadow_mem_size): Add gcc_assert.
10971 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10973 Don't save registers in main().
10976 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10977 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10978 * config/avr/avr.c (avr_set_current_function): Don't error if
10979 naked, OS_task or OS_main are specified at the same time.
10980 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10982 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10984 * common/config/avr/avr-common.c (avr_option_optimization_table):
10985 Switch on -mmain-is-OS_task for optimizing compilations.
10987 2018-01-09 Richard Biener <rguenther@suse.de>
10989 PR tree-optimization/83572
10990 * graphite.c: Include cfganal.h.
10991 (graphite_transform_loops): Connect infinite loops to exit
10992 and remove fake edges at the end.
10994 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10996 * ipa-inline.c (edge_badness): Revert accidental checkin.
10998 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11001 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11002 symbols; not inline clones.
11004 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11007 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11008 hard registers. Formatting fixes.
11010 PR preprocessor/83722
11011 * gcc.c (try_generate_repro): Pass
11012 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11013 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11016 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11017 Kito Cheng <kito.cheng@gmail.com>
11019 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11020 (riscv_leaf_function_p): Delete.
11021 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11023 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11025 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11027 (do_ifelse): New function.
11028 (do_isel): New function.
11029 (do_sub3): New function.
11030 (do_add3): New function.
11031 (do_load_mask_compare): New function.
11032 (do_overlap_load_compare): New function.
11033 (expand_compare_loop): New function.
11034 (expand_block_compare): Call expand_compare_loop() when appropriate.
11035 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11036 option description.
11037 (-mblock-compare-inline-loop-limit): New option.
11039 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11042 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11043 Reverse order of second and third operands in first alternative.
11044 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11045 of first and second elements in UNSPEC_VPERMR vector.
11046 (altivec_expand_vec_perm_le): Likewise.
11048 2018-01-08 Jeff Law <law@redhat.com>
11050 PR rtl-optimizatin/81308
11051 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11052 (process_switch): If group_case_labels makes a change, then set
11054 (pass_convert_switch::execute): If a switch is converted, then
11055 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11057 PR rtl-optimization/81308
11058 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11061 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11063 PR target/83663 - Revert r255946
11064 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11065 generation for cases where splatting a value is not useful.
11066 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11067 across a vec_duplicate and a paradoxical subreg forming a vector
11068 mode to a vec_concat.
11070 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11072 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11073 -march=armv8.3-a variants.
11074 * config/arm/t-multilib: Likewise.
11075 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11077 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11079 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11081 (cceq_ior_compare_complement): Give it a name so I can use it, and
11082 change boolean_or_operator predicate to boolean_operator so it can
11083 be used to generate a crand.
11084 (eqne): New code iterator.
11085 (bd/bd_neg): New code_attrs.
11086 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11087 a single define_insn.
11088 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11089 decrement (bdnzt/bdnzf/bdzt/bdzf).
11090 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11091 with the new names of the branch decrement patterns, and added the
11092 names of the branch decrement conditional patterns.
11094 2018-01-08 Richard Biener <rguenther@suse.de>
11096 PR tree-optimization/83563
11097 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11100 2018-01-08 Richard Biener <rguenther@suse.de>
11102 PR middle-end/83713
11103 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11105 2018-01-08 Richard Biener <rguenther@suse.de>
11107 PR tree-optimization/83685
11108 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11109 references to abnormals.
11111 2018-01-08 Richard Biener <rguenther@suse.de>
11114 * dwarf2out.c (output_indirect_strings): Handle empty
11115 skeleton_debug_str_hash.
11116 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11118 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11120 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11121 (emit_store_direct): Likewise.
11122 (arc_trampoline_adjust_address): Likewise.
11123 (arc_asm_trampoline_template): New function.
11124 (arc_initialize_trampoline): Use asm_trampoline_template.
11125 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11126 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11127 * config/arc/arc.md (flush_icache): Delete pattern.
11129 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11131 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11132 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11135 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11138 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11139 by not USED_FOR_TARGET.
11140 (make_pass_resolve_sw_modes): Likewise.
11142 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11144 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11147 2018-01-08 Richard Biener <rguenther@suse.de>
11149 PR middle-end/83580
11150 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11152 2018-01-08 Richard Biener <rguenther@suse.de>
11154 PR middle-end/83517
11155 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11157 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11159 PR middle-end/81897
11160 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11161 basic blocks with a small number of successors.
11162 (convert_control_dep_chain_into_preds): Improve handling of
11164 (dump_predicates): Split apart into...
11165 (dump_pred_chain): ...here...
11166 (dump_pred_info): ...and here.
11167 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11168 (can_chain_union_be_invalidated_p): Improve check for invalidation
11170 (uninit_uses_cannot_happen): Avoid unnecessary if
11171 convert_control_dep_chain_into_preds yielded nothing.
11173 2018-01-06 Martin Sebor <msebor@redhat.com>
11175 PR tree-optimization/83640
11176 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11177 subtracting negative offset from size.
11178 (builtin_access::overlap): Adjust offset bounds of the access to fall
11179 within the size of the object if possible.
11181 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11183 PR rtl-optimization/83699
11184 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11185 extract_bit_field_as_subreg to cases in which the extracted
11186 value is also a vector.
11188 * lra-constraints.c (process_alt_operands): Test for the equivalence
11189 substitutions when detecting a possible reload cycle.
11191 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11194 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11195 by default if flag_selective_schedling{,2}. Formatting fixes.
11197 PR rtl-optimization/83682
11198 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11199 if it has non-VECTOR_MODE element mode.
11200 (vec_duplicate_p): Likewise.
11202 PR middle-end/83694
11203 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11204 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11206 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11209 * config/i386/i386-builtin.def
11210 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11211 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11212 Require also OPTION_MASK_ISA_AVX512F in addition to
11213 OPTION_MASK_ISA_GFNI.
11214 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11215 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11216 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11217 to OPTION_MASK_ISA_GFNI.
11218 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11219 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11220 OPTION_MASK_ISA_AVX512BW.
11221 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11222 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11223 addition to OPTION_MASK_ISA_GFNI.
11224 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11225 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11226 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11227 to OPTION_MASK_ISA_GFNI.
11228 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11229 a requirement for all ISAs rather than any of them with a few
11231 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11233 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11234 bitmasks to be enabled with 3 exceptions, instead of requiring any
11235 enabled ISA with lots of exceptions.
11236 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11237 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11238 Change avx512bw in isa attribute to avx512f.
11239 * config/i386/sgxintrin.h: Add license boilerplate.
11240 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11241 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11242 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11243 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11245 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11246 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11247 temporarily sse2 rather than sse if not enabled already.
11250 * config/i386/sse.md (VI248_VLBW): Rename to ...
11251 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11252 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11253 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11254 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11255 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11256 mode iterator instead of VI248_VLBW.
11258 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11260 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11261 (record_modified): Skip clobbers; add debug output.
11262 (param_change_prob): Use sreal frequencies.
11264 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11266 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11267 punt for user-aligned variables.
11269 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11271 * tree-chrec.c (chrec_contains_symbols): Return true for
11274 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11277 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11278 of (x|y) == x for BICS pattern.
11280 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11282 PR tree-optimization/83605
11283 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11284 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11287 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11289 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11290 * config/epiphany/rtems.h: New file.
11292 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11293 Uros Bizjak <ubizjak@gmail.com>
11296 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11297 QIreg_operand instead of register_operand predicate.
11298 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11299 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11300 comments instead of -fmitigate[-_]rop.
11302 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11305 * cgraphunit.c (symbol_table::compile): Switch to text_section
11306 before calling assembly_start debug hook.
11307 * run-rtl-passes.c (run_rtl_passes): Likewise.
11310 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11312 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11313 range_int_cst_p rather than !symbolic_range_p before calling
11314 extract_range_from_multiplicative_op_1.
11316 2018-01-04 Jeff Law <law@redhat.com>
11318 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11319 redundant test in assertion.
11321 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11323 * doc/rtl.texi: Document machine_mode wrapper classes.
11325 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11327 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11328 using tree_to_uhwi.
11330 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11332 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11333 the VEC_PERM_EXPR fold to fail.
11335 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11338 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11339 to switched_sections.
11341 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11344 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11345 test for d.testing.
11347 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11350 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11351 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11353 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11356 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11357 is BLKmode and bitpos not zero or mode change is needed.
11359 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11362 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11365 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11368 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11369 instead of MULT rtx. Update all corresponding splitters.
11370 (*saddl_se): Ditto.
11371 (*ssub<modesuffix>): Ditto.
11372 (*ssubl_se): Ditto.
11373 (*cmp_sadd_di): Update split patterns.
11374 (*cmp_sadd_si): Ditto.
11375 (*cmp_sadd_sidi): Ditto.
11376 (*cmp_ssub_di): Ditto.
11377 (*cmp_ssub_si): Ditto.
11378 (*cmp_ssub_sidi): Ditto.
11379 * config/alpha/predicates.md (const23_operand): New predicate.
11380 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11381 Look for ASHIFT, not MULT inner operand.
11382 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11384 2018-01-04 Martin Liska <mliska@suse.cz>
11386 PR gcov-profile/83669
11387 * gcov.c (output_intermediate_file): Add version to intermediate
11389 * doc/gcov.texi: Document new field 'version' in intermediate
11390 file format. Fix location of '-k' option of gcov command.
11392 2018-01-04 Martin Liska <mliska@suse.cz>
11395 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11397 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11399 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11401 2018-01-03 Martin Sebor <msebor@redhat.com>
11403 PR tree-optimization/83655
11404 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11405 checking calls with invalid arguments.
11407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11409 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11410 (vectorizable_mask_load_store): Delete.
11411 (vectorizable_call): Return false for masked loads and stores.
11412 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11413 instead of gimple_assign_rhs1.
11414 (vectorizable_load): Handle IFN_MASK_LOAD.
11415 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11419 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11421 (vectorizable_mask_load_store): ...here.
11422 (vectorizable_load): ...and here.
11424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11426 * tree-vect-stmts.c (vect_build_all_ones_mask)
11427 (vect_build_zero_merge_argument): New functions, split out from...
11428 (vectorizable_load): ...here.
11430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11432 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11434 (vectorizable_mask_load_store): ...here.
11435 (vectorizable_store): ...and here.
11437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11439 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11441 (vectorizable_mask_load_store): ...here.
11443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11445 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11446 (vect_model_store_cost): Take a vec_load_store_type instead of a
11448 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11449 (vect_model_store_cost): Take a vec_load_store_type instead of a
11451 (vectorizable_mask_load_store): Update accordingly.
11452 (vectorizable_store): Likewise.
11453 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11457 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11458 IFN_MASK_LOAD calls here rather than...
11459 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11462 Alan Hayward <alan.hayward@arm.com>
11463 David Sherwood <david.sherwood@arm.com>
11465 * expmed.c (extract_bit_field_1): For vector extracts,
11466 fall back to extract_bit_field_as_subreg if vec_extract
11469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11470 Alan Hayward <alan.hayward@arm.com>
11471 David Sherwood <david.sherwood@arm.com>
11473 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11474 they are variable or constant sized.
11475 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11476 slots for constant-sized data.
11478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11479 Alan Hayward <alan.hayward@arm.com>
11480 David Sherwood <david.sherwood@arm.com>
11482 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11483 handling COND_EXPRs with boolean comparisons, try to find a better
11484 basis for the mask type than the boolean itself.
11486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11488 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11489 is calculated and how it can be overridden.
11490 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11491 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11493 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11497 Alan Hayward <alan.hayward@arm.com>
11498 David Sherwood <david.sherwood@arm.com>
11500 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11501 Remove the mode argument.
11502 (aarch64_simd_valid_immediate): Remove the mode and inverse
11504 * config/aarch64/iterators.md (bitsize): New iterator.
11505 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11506 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11507 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11508 aarch64_simd_valid_immediate.
11509 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11510 (aarch64_reg_or_bic_imm): Likewise.
11511 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11512 with an insn_type enum and msl with a modifier_type enum.
11513 Replace element_width with a scalar_mode. Change the shift
11514 to unsigned int. Add constructors for scalar_float_mode and
11515 scalar_int_mode elements.
11516 (aarch64_vect_float_const_representable_p): Delete.
11517 (aarch64_can_const_movi_rtx_p)
11518 (aarch64_simd_scalar_immediate_valid_for_move)
11519 (aarch64_simd_make_constant): Update call to
11520 aarch64_simd_valid_immediate.
11521 (aarch64_advsimd_valid_immediate_hs): New function.
11522 (aarch64_advsimd_valid_immediate): Likewise.
11523 (aarch64_simd_valid_immediate): Remove mode and inverse
11524 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11525 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11526 and aarch64_float_const_representable_p on the result.
11527 (aarch64_output_simd_mov_immediate): Remove mode argument.
11528 Update call to aarch64_simd_valid_immediate and use of
11529 simd_immediate_info.
11530 (aarch64_output_scalar_simd_mov_immediate): Update call
11533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11534 Alan Hayward <alan.hayward@arm.com>
11535 David Sherwood <david.sherwood@arm.com>
11537 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11538 (mode_nunits): Likewise CONST_MODE_NUNITS.
11539 * machmode.def (ADJUST_NUNITS): Document.
11540 * genmodes.c (mode_data::need_nunits_adj): New field.
11541 (blank_mode): Update accordingly.
11542 (adj_nunits): New variable.
11543 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11545 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11546 listed in adj_nunits.
11547 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11548 listed in adj_nunits. Don't emit case statements for such modes.
11549 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11550 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11551 nothing if adj_nunits is nonnull.
11552 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11553 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11554 (emit_mode_fbit): Update use of print_maybe_const_decl.
11555 (emit_move_size): Likewise. Treat the array as non-const
11557 (emit_mode_adjustments): Handle adj_nunits.
11559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11561 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11562 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11563 (VECTOR_MODES): Use it.
11564 (make_vector_modes): Take the prefix as an argument.
11566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11567 Alan Hayward <alan.hayward@arm.com>
11568 David Sherwood <david.sherwood@arm.com>
11570 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11571 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11572 for MODE_VECTOR_BOOL.
11573 * machmode.def (VECTOR_BOOL_MODE): Document.
11574 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11575 (make_vector_bool_mode): New function.
11576 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11578 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11579 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11581 * stor-layout.c (int_mode_for_mode): Likewise.
11582 * tree.c (build_vector_type_for_mode): Likewise.
11583 * varasm.c (output_constant_pool_2): Likewise.
11584 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11585 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11586 for MODE_VECTOR_BOOL.
11587 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11588 of mode class checks.
11589 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11590 instead of a list of mode class checks.
11591 (expand_vector_scalar_condition): Likewise.
11592 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11595 Alan Hayward <alan.hayward@arm.com>
11596 David Sherwood <david.sherwood@arm.com>
11598 * machmode.h (mode_size): Change from unsigned short to
11600 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11601 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11602 or if measurement_type is not polynomial.
11603 (fixed_size_mode::includes_p): Check for constant-sized modes.
11604 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11605 return a poly_uint16 rather than an unsigned short.
11606 (emit_mode_size): Change the type of mode_size from unsigned short
11607 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11608 (emit_mode_adjustments): Cope with polynomial vector sizes.
11609 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11611 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11613 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11614 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11615 * caller-save.c (setup_save_areas): Likewise.
11616 (replace_reg_with_saved_mem): Likewise.
11617 * calls.c (emit_library_call_value_1): Likewise.
11618 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11619 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11620 (gen_lowpart_for_combine): Likewise.
11621 * convert.c (convert_to_integer_1): Likewise.
11622 * cse.c (equiv_constant, cse_insn): Likewise.
11623 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11624 (cselib_subst_to_values): Likewise.
11625 * dce.c (word_dce_process_block): Likewise.
11626 * df-problems.c (df_word_lr_mark_ref): Likewise.
11627 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11628 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11629 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11630 (rtl_for_decl_location): Likewise.
11631 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11632 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11633 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11634 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11635 (expand_expr_real_1): Likewise.
11636 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11637 (pad_below): Likewise.
11638 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11639 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11640 * ira.c (get_subreg_tracking_sizes): Likewise.
11641 * ira-build.c (ira_create_allocno_objects): Likewise.
11642 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11643 (ira_sort_regnos_for_alter_reg): Likewise.
11644 * ira-costs.c (record_operand_costs): Likewise.
11645 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11646 (resolve_simple_move): Likewise.
11647 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11648 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11649 (lra_constraints): Likewise.
11650 (CONST_POOL_OK_P): Reject variable-sized modes.
11651 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11652 (add_pseudo_to_slot, lra_spill): Likewise.
11653 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11654 * optabs-query.c (get_best_extraction_insn): Likewise.
11655 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11656 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11657 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11658 * recog.c (offsettable_address_addr_space_p): Likewise.
11659 * regcprop.c (maybe_mode_change): Likewise.
11660 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11661 * regrename.c (build_def_use): Likewise.
11662 * regstat.c (dump_reg_info): Likewise.
11663 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11664 (find_reloads, find_reloads_subreg_address): Likewise.
11665 * reload1.c (eliminate_regs_1): Likewise.
11666 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11667 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11668 (simplify_binary_operation_1, simplify_subreg): Likewise.
11669 * targhooks.c (default_function_arg_padding): Likewise.
11670 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11671 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11672 (verify_gimple_assign_ternary): Likewise.
11673 * tree-inline.c (estimate_move_cost): Likewise.
11674 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11675 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11676 (get_address_cost_ainc): Likewise.
11677 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11678 (vect_supportable_dr_alignment): Likewise.
11679 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11680 (vectorizable_reduction): Likewise.
11681 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11682 (vectorizable_operation, vectorizable_load): Likewise.
11683 * tree.c (build_same_sized_truth_vector_type): Likewise.
11684 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11685 * var-tracking.c (emit_note_insn_var_location): Likewise.
11686 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11687 (ADDR_VEC_ALIGN): Likewise.
11689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11690 Alan Hayward <alan.hayward@arm.com>
11691 David Sherwood <david.sherwood@arm.com>
11693 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11695 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11696 or if measurement_type is polynomial.
11697 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11698 * combine.c (make_extraction): Likewise.
11699 * dse.c (find_shift_sequence): Likewise.
11700 * dwarf2out.c (mem_loc_descriptor): Likewise.
11701 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11702 (extract_bit_field, extract_low_bits): Likewise.
11703 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11704 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11705 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11706 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11707 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11708 * reload.c (find_reloads): Likewise.
11709 * reload1.c (alter_reg): Likewise.
11710 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11711 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11712 * tree-if-conv.c (predicate_mem_writes): Likewise.
11713 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11714 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11715 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11716 * valtrack.c (dead_debug_insert_temp): Likewise.
11717 * varasm.c (mergeable_constant_section): Likewise.
11718 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11721 Alan Hayward <alan.hayward@arm.com>
11722 David Sherwood <david.sherwood@arm.com>
11724 * expr.c (expand_assignment): Cope with polynomial mode sizes
11725 when assigning to a CONCAT.
11727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11728 Alan Hayward <alan.hayward@arm.com>
11729 David Sherwood <david.sherwood@arm.com>
11731 * machmode.h (mode_precision): Change from unsigned short to
11733 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11735 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11736 or if measurement_type is not polynomial.
11737 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11738 in which the mode is already known to be a scalar_int_mode.
11739 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11740 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11742 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11743 for GET_MODE_PRECISION.
11744 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11745 for GET_MODE_PRECISION.
11746 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11748 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11749 (expand_field_assignment, make_extraction): Likewise.
11750 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11751 (get_last_value): Likewise.
11752 * convert.c (convert_to_integer_1): Likewise.
11753 * cse.c (cse_insn): Likewise.
11754 * expr.c (expand_expr_real_1): Likewise.
11755 * lra-constraints.c (simplify_operand_subreg): Likewise.
11756 * optabs-query.c (can_atomic_load_p): Likewise.
11757 * optabs.c (expand_atomic_load): Likewise.
11758 (expand_atomic_store): Likewise.
11759 * ree.c (combine_reaching_defs): Likewise.
11760 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11761 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11762 * tree.h (type_has_mode_precision_p): Likewise.
11763 * ubsan.c (instrument_si_overflow): Likewise.
11765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11766 Alan Hayward <alan.hayward@arm.com>
11767 David Sherwood <david.sherwood@arm.com>
11769 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11770 polynomial numbers of units.
11771 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11772 (valid_vector_subparts_p): New function.
11773 (build_vector_type): Remove temporary shim and take the number
11774 of units as a poly_uint64 rather than an int.
11775 (build_opaque_vector_type): Take the number of units as a
11776 poly_uint64 rather than an int.
11777 * tree.c (build_vector_from_ctor): Handle polynomial
11778 TYPE_VECTOR_SUBPARTS.
11779 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11780 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11781 (build_vector_from_val): If the number of units is variable,
11782 use build_vec_duplicate_cst for constant operands and
11783 VEC_DUPLICATE_EXPR otherwise.
11784 (make_vector_type): Remove temporary is_constant ().
11785 (build_vector_type, build_opaque_vector_type): Take the number of
11786 units as a poly_uint64 rather than an int.
11787 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11789 * cfgexpand.c (expand_debug_expr): Likewise.
11790 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11791 (store_constructor, expand_expr_real_1): Likewise.
11792 (const_scalar_mask_from_tree): Likewise.
11793 * fold-const-call.c (fold_const_reduction): Likewise.
11794 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11795 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11796 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11797 (fold_relational_const): Likewise.
11798 (native_interpret_vector): Likewise. Change the size from an
11799 int to an unsigned int.
11800 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11801 TYPE_VECTOR_SUBPARTS.
11802 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11803 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11804 duplicating a non-constant operand into a variable-length vector.
11805 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11806 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11807 * ipa-icf.c (sem_variable::equals): Likewise.
11808 * match.pd: Likewise.
11809 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11810 * print-tree.c (print_node): Likewise.
11811 * stor-layout.c (layout_type): Likewise.
11812 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11813 * tree-cfg.c (verify_gimple_comparison): Likewise.
11814 (verify_gimple_assign_binary): Likewise.
11815 (verify_gimple_assign_ternary): Likewise.
11816 (verify_gimple_assign_single): Likewise.
11817 * tree-pretty-print.c (dump_generic_node): Likewise.
11818 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11819 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11820 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11821 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11822 (vect_shift_permute_load_chain): Likewise.
11823 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11824 (expand_vector_condition, optimize_vector_constructor): Likewise.
11825 (lower_vec_perm, get_compute_type): Likewise.
11826 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11827 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11828 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11829 (vect_recog_mask_conversion_pattern): Likewise.
11830 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11831 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11832 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11833 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11834 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11835 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11836 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11837 (supportable_widening_operation): Likewise.
11838 (supportable_narrowing_operation): Likewise.
11839 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11841 * varasm.c (output_constant): Likewise.
11843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11844 Alan Hayward <alan.hayward@arm.com>
11845 David Sherwood <david.sherwood@arm.com>
11847 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11848 so that both the length == 3 and length != 3 cases set up their
11849 own permute vectors. Add comments explaining why we know the
11850 number of elements is constant.
11851 (vect_permute_load_chain): Likewise.
11853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11854 Alan Hayward <alan.hayward@arm.com>
11855 David Sherwood <david.sherwood@arm.com>
11857 * machmode.h (mode_nunits): Change from unsigned char to
11859 (ONLY_FIXED_SIZE_MODES): New macro.
11860 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11861 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11862 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11864 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11865 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11866 or if measurement_type is not polynomial.
11867 * genmodes.c (ZERO_COEFFS): New macro.
11868 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11870 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11871 Use ZERO_COEFFS when emitting initializers.
11872 * data-streamer.h (bp_pack_poly_value): New function.
11873 (bp_unpack_poly_value): Likewise.
11874 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11875 for GET_MODE_NUNITS.
11876 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11877 for GET_MODE_NUNITS.
11878 * tree.c (make_vector_type): Remove temporary shim and make
11879 the real function take the number of units as a poly_uint64
11880 rather than an int.
11881 (build_vector_type_for_mode): Handle polynomial nunits.
11882 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11883 * emit-rtl.c (const_vec_series_p_1): Likewise.
11884 (gen_rtx_CONST_VECTOR): Likewise.
11885 * fold-const.c (test_vec_duplicate_folding): Likewise.
11886 * genrecog.c (validate_pattern): Likewise.
11887 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11888 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11889 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11890 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11891 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11892 * rtlanal.c (subreg_get_info): Likewise.
11893 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11894 (vect_grouped_load_supported): Likewise.
11895 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11896 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11897 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11898 (simplify_const_unary_operation, simplify_binary_operation_1)
11899 (simplify_const_binary_operation, simplify_ternary_operation)
11900 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11901 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11902 instead of CONST_VECTOR_NUNITS.
11903 * varasm.c (output_constant_pool_2): Likewise.
11904 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11905 explicit-encoded elements in the XVEC for variable-length vectors.
11907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11909 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11912 Alan Hayward <alan.hayward@arm.com>
11913 David Sherwood <david.sherwood@arm.com>
11915 * coretypes.h (fixed_size_mode): Declare.
11916 (fixed_size_mode_pod): New typedef.
11917 * builtins.h (target_builtins::x_apply_args_mode)
11918 (target_builtins::x_apply_result_mode): Change type to
11919 fixed_size_mode_pod.
11920 * builtins.c (apply_args_size, apply_result_size, result_vector)
11921 (expand_builtin_apply_args_1, expand_builtin_apply)
11922 (expand_builtin_return): Update accordingly.
11924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11926 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11927 * cselib.c (cselib_hash_rtx): Likewise.
11928 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11929 CONST_VECTOR encoding.
11931 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11932 Jeff Law <law@redhat.com>
11935 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11936 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11937 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11938 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11941 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11942 explicitly probe *sp in a noreturn function if there were any callee
11943 register saves or frame pointer is needed.
11945 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11948 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11949 BLKmode for ternary, binary or unary expressions.
11952 * var-tracking.c (delete_vta_debug_insn): New inline function.
11953 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11954 insns from get_insns () to NULL instead of each bb separately.
11955 Use delete_vta_debug_insn. No longer static.
11956 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11957 delete_vta_debug_insns callers.
11958 * rtl.h (delete_vta_debug_insns): Declare.
11959 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11960 instead of variable_tracking_main.
11962 2018-01-03 Martin Sebor <msebor@redhat.com>
11964 PR tree-optimization/83603
11965 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11966 arguments past the endof the argument list in functions declared
11967 without a prototype.
11968 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11969 Avoid checking when arguments are null.
11971 2018-01-03 Martin Sebor <msebor@redhat.com>
11974 * doc/extend.texi (attribute const): Fix a typo.
11975 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11976 issuing -Wsuggest-attribute for void functions.
11978 2018-01-03 Martin Sebor <msebor@redhat.com>
11980 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11981 offset_int::from instead of wide_int::to_shwi.
11982 (maybe_diag_overlap): Remove assertion.
11983 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11984 * gimple-ssa-sprintf.c (format_directive): Same.
11985 (parse_directive): Same.
11986 (sprintf_dom_walker::compute_format_length): Same.
11987 (try_substitute_return_value): Same.
11989 2018-01-03 Jeff Law <law@redhat.com>
11991 PR middle-end/83654
11992 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11993 non-constant residual for zero at runtime and avoid probing in
11994 that case. Reorganize code for trailing problem to mirror handling
11997 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11999 PR tree-optimization/83501
12000 * tree-ssa-strlen.c (get_string_cst): New.
12001 (handle_char_store): Call get_string_cst.
12003 2018-01-03 Martin Liska <mliska@suse.cz>
12005 PR tree-optimization/83593
12006 * tree-ssa-strlen.c: Include tree-cfg.h.
12007 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12008 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12009 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12011 (strlen_dom_walker::before_dom_children): Call
12012 gimple_purge_dead_eh_edges. Dump tranformation with details
12014 (strlen_dom_walker::before_dom_children): Update call by adding
12015 new argument cleanup_eh.
12016 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12018 2018-01-03 Martin Liska <mliska@suse.cz>
12021 * cif-code.def (VARIADIC_THUNK): New enum value.
12022 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12025 2018-01-03 Jan Beulich <jbeulich@suse.com>
12027 * sse.md (mov<mode>_internal): Tighten condition for when to use
12028 vmovdqu<ssescalarsize> for TI and OI modes.
12030 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12032 Update copyright years.
12034 2018-01-03 Martin Liska <mliska@suse.cz>
12037 * ipa-visibility.c (function_and_variable_visibility): Skip
12038 functions with noipa attribure.
12040 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12042 * gcc.c (process_command): Update copyright notice dates.
12043 * gcov-dump.c (print_version): Ditto.
12044 * gcov.c (print_version): Ditto.
12045 * gcov-tool.c (print_version): Ditto.
12046 * gengtype.c (create_file): Ditto.
12047 * doc/cpp.texi: Bump @copying's copyright year.
12048 * doc/cppinternals.texi: Ditto.
12049 * doc/gcc.texi: Ditto.
12050 * doc/gccint.texi: Ditto.
12051 * doc/gcov.texi: Ditto.
12052 * doc/install.texi: Ditto.
12053 * doc/invoke.texi: Ditto.
12055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12057 * vector-builder.h (vector_builder::m_full_nelts): Change from
12058 unsigned int to poly_uint64.
12059 (vector_builder::full_nelts): Update prototype accordingly.
12060 (vector_builder::new_vector): Likewise.
12061 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12062 (vector_builder::operator ==): Likewise.
12063 (vector_builder::finalize): Likewise.
12064 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12065 Take the number of elements as a poly_uint64 rather than an
12067 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12068 from unsigned int to poly_uint64.
12069 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12070 (vec_perm_indices::new_vector): Likewise.
12071 (vec_perm_indices::length): Likewise.
12072 (vec_perm_indices::nelts_per_input): Likewise.
12073 (vec_perm_indices::input_nelts): Likewise.
12074 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12075 number of elements per input as a poly_uint64 rather than an
12076 unsigned int. Use the original encoding for variable-length
12077 vectors, rather than clamping each individual element.
12078 For the second and subsequent elements in each pattern,
12079 clamp the step and base before clamping their sum.
12080 (vec_perm_indices::series_p): Handle polynomial element counts.
12081 (vec_perm_indices::all_in_range_p): Likewise.
12082 (vec_perm_indices_to_tree): Likewise.
12083 (vec_perm_indices_to_rtx): Likewise.
12084 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12085 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12086 (tree_vector_builder::new_binary_operation): Handle polynomial
12087 element counts. Return false if we need to know the number
12088 of elements at compile time.
12089 * fold-const.c (fold_vec_perm): Punt if the number of elements
12090 isn't known at compile time.
12092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12094 * vec-perm-indices.h (vec_perm_builder): Change element type
12095 from HOST_WIDE_INT to poly_int64.
12096 (vec_perm_indices::element_type): Update accordingly.
12097 (vec_perm_indices::clamp): Handle polynomial element_types.
12098 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12099 (vec_perm_indices::all_in_range_p): Likewise.
12100 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12102 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12103 polynomial vec_perm_indices element types.
12104 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12105 * fold-const.c (fold_vec_perm): Likewise.
12106 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12107 * tree-vect-generic.c (lower_vec_perm): Likewise.
12108 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12109 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12110 element type to HOST_WIDE_INT.
12112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12113 Alan Hayward <alan.hayward@arm.com>
12114 David Sherwood <david.sherwood@arm.com>
12116 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12117 rather than an int. Use plus_constant.
12118 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12119 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12122 Alan Hayward <alan.hayward@arm.com>
12123 David Sherwood <david.sherwood@arm.com>
12125 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12126 a HOST_WIDE_INT to a poly_int64.
12128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12129 Alan Hayward <alan.hayward@arm.com>
12130 David Sherwood <david.sherwood@arm.com>
12132 * calls.c (load_register_parameters): Cope with polynomial
12133 mode sizes. Require a constant size for BLKmode parameters
12134 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12135 forces a parameter to be padded at the lsb end in order to
12136 fill a complete number of words, require the parameter size
12137 to be ordered wrt UNITS_PER_WORD.
12139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12140 Alan Hayward <alan.hayward@arm.com>
12141 David Sherwood <david.sherwood@arm.com>
12143 * reload1.c (spill_stack_slot_width): Change element type
12144 from unsigned int to poly_uint64_pod.
12145 (alter_reg): Treat mode sizes as polynomial.
12147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12148 Alan Hayward <alan.hayward@arm.com>
12149 David Sherwood <david.sherwood@arm.com>
12151 * reload.c (complex_word_subreg_p): New function.
12152 (reload_inner_reg_of_subreg, push_reload): Use it.
12154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12155 Alan Hayward <alan.hayward@arm.com>
12156 David Sherwood <david.sherwood@arm.com>
12158 * lra-constraints.c (process_alt_operands): Reject matched
12159 operands whose sizes aren't ordered.
12160 (match_reload): Refer to this check here.
12162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12163 Alan Hayward <alan.hayward@arm.com>
12164 David Sherwood <david.sherwood@arm.com>
12166 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12167 that the mode size is in the set {1, 2, 4, 8, 16}.
12169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12170 Alan Hayward <alan.hayward@arm.com>
12171 David Sherwood <david.sherwood@arm.com>
12173 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12174 Use plus_constant instead of gen_rtx_PLUS.
12176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12177 Alan Hayward <alan.hayward@arm.com>
12178 David Sherwood <david.sherwood@arm.com>
12180 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12181 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12182 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12183 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12184 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12185 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12186 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12187 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12188 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12189 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12191 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12192 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12193 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12194 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12195 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12196 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12197 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12198 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12199 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12200 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12202 * expr.c (emit_move_resolve_push): Treat the input and result
12203 of PUSH_ROUNDING as a poly_int64.
12204 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12205 (emit_push_insn): Likewise.
12206 * lra-eliminations.c (mark_not_eliminable): Likewise.
12207 * recog.c (push_operand): Likewise.
12208 * reload1.c (elimination_effects): Likewise.
12209 * rtlanal.c (nonzero_bits1): Likewise.
12210 * calls.c (store_one_arg): Likewise. Require the padding to be
12211 known at compile time.
12213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12214 Alan Hayward <alan.hayward@arm.com>
12215 David Sherwood <david.sherwood@arm.com>
12217 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12218 Use plus_constant instead of gen_rtx_PLUS.
12220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12221 Alan Hayward <alan.hayward@arm.com>
12222 David Sherwood <david.sherwood@arm.com>
12224 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12225 rather than an int.
12227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12228 Alan Hayward <alan.hayward@arm.com>
12229 David Sherwood <david.sherwood@arm.com>
12231 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12232 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12233 via stack temporaries. Treat the mode size as polynomial too.
12235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12236 Alan Hayward <alan.hayward@arm.com>
12237 David Sherwood <david.sherwood@arm.com>
12239 * expr.c (expand_expr_real_2): When handling conversions involving
12240 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12241 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12242 as a poly_uint64 too.
12244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12245 Alan Hayward <alan.hayward@arm.com>
12246 David Sherwood <david.sherwood@arm.com>
12248 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12251 Alan Hayward <alan.hayward@arm.com>
12252 David Sherwood <david.sherwood@arm.com>
12254 * combine.c (can_change_dest_mode): Handle polynomial
12255 REGMODE_NATURAL_SIZE.
12256 * expmed.c (store_bit_field_1): Likewise.
12257 * expr.c (store_constructor): Likewise.
12258 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12259 and polynomial REGMODE_NATURAL_SIZE.
12260 (gen_lowpart_common): Likewise.
12261 * reginfo.c (record_subregs_of_mode): Likewise.
12262 * rtlanal.c (read_modify_subreg_p): Likewise.
12264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12265 Alan Hayward <alan.hayward@arm.com>
12266 David Sherwood <david.sherwood@arm.com>
12268 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12269 numbers of elements.
12271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12272 Alan Hayward <alan.hayward@arm.com>
12273 David Sherwood <david.sherwood@arm.com>
12275 * match.pd: Cope with polynomial numbers of vector elements.
12277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12278 Alan Hayward <alan.hayward@arm.com>
12279 David Sherwood <david.sherwood@arm.com>
12281 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12282 in a POINTER_PLUS_EXPR.
12284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12285 Alan Hayward <alan.hayward@arm.com>
12286 David Sherwood <david.sherwood@arm.com>
12288 * omp-simd-clone.c (simd_clone_subparts): New function.
12289 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12290 (ipa_simd_modify_function_body): Likewise.
12292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12293 Alan Hayward <alan.hayward@arm.com>
12294 David Sherwood <david.sherwood@arm.com>
12296 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12297 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12298 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12299 (expand_vector_condition, vector_element): Likewise.
12300 (subparts_gt): New function.
12301 (get_compute_type): Use subparts_gt.
12302 (count_type_subparts): Delete.
12303 (expand_vector_operations_1): Use subparts_gt instead of
12304 count_type_subparts.
12306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12307 Alan Hayward <alan.hayward@arm.com>
12308 David Sherwood <david.sherwood@arm.com>
12310 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12311 (vect_compile_time_alias): ...this new function. Do the calculation
12312 on poly_ints rather than trees.
12313 (vect_prune_runtime_alias_test_list): Update call accordingly.
12315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12316 Alan Hayward <alan.hayward@arm.com>
12317 David Sherwood <david.sherwood@arm.com>
12319 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12321 (vect_schedule_slp_instance): Likewise.
12323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12324 Alan Hayward <alan.hayward@arm.com>
12325 David Sherwood <david.sherwood@arm.com>
12327 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12328 constant and extern definitions for variable-length vectors.
12329 (vect_get_constant_vectors): Note that the number of units
12330 is known to be constant.
12332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12333 Alan Hayward <alan.hayward@arm.com>
12334 David Sherwood <david.sherwood@arm.com>
12336 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12337 of units as polynomial. Choose between WIDE and NARROW based
12340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12341 Alan Hayward <alan.hayward@arm.com>
12342 David Sherwood <david.sherwood@arm.com>
12344 * tree-vect-stmts.c (simd_clone_subparts): New function.
12345 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12348 Alan Hayward <alan.hayward@arm.com>
12349 David Sherwood <david.sherwood@arm.com>
12351 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12352 vectors as polynomial. Use build_index_vector for
12353 IFN_GOMP_SIMD_LANE.
12355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12356 Alan Hayward <alan.hayward@arm.com>
12357 David Sherwood <david.sherwood@arm.com>
12359 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12360 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12361 for variable-length vectors.
12362 (vectorizable_mask_load_store): Treat the number of units as
12363 polynomial, asserting that it is constant if the condition has
12364 already been enforced.
12365 (vectorizable_store, vectorizable_load): Likewise.
12367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12368 Alan Hayward <alan.hayward@arm.com>
12369 David Sherwood <david.sherwood@arm.com>
12371 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12372 of units as polynomial. Punt if we can't tell at compile time
12373 which vector contains the final result.
12375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12376 Alan Hayward <alan.hayward@arm.com>
12377 David Sherwood <david.sherwood@arm.com>
12379 * tree-vect-loop.c (vectorizable_induction): Treat the number
12380 of units as polynomial. Punt on SLP inductions. Use an integer
12381 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12382 cast of such a series for variable-length floating-point
12385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12386 Alan Hayward <alan.hayward@arm.com>
12387 David Sherwood <david.sherwood@arm.com>
12389 * tree.h (build_index_vector): Declare.
12390 * tree.c (build_index_vector): New function.
12391 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12392 of units as polynomial, forcibly converting it to a constant if
12393 vectorizable_reduction has already enforced the condition.
12394 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12395 to create a {1,2,3,...} vector.
12396 (vectorizable_reduction): Treat the number of units as polynomial.
12397 Choose vectype_in based on the largest scalar element size rather
12398 than the smallest number of units. Enforce the restrictions
12401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12402 Alan Hayward <alan.hayward@arm.com>
12403 David Sherwood <david.sherwood@arm.com>
12405 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12406 number of units as polynomial.
12408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12409 Alan Hayward <alan.hayward@arm.com>
12410 David Sherwood <david.sherwood@arm.com>
12412 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12413 * target.def (autovectorize_vector_sizes): Return the vector sizes
12414 by pointer, using vector_sizes rather than a bitmask.
12415 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12416 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12417 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12419 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12420 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12421 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12422 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12423 * omp-general.c (omp_max_vf): Likewise.
12424 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12425 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12426 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12427 * tree-vect-slp.c (vect_slp_bb): Likewise.
12428 * doc/tm.texi: Regenerate.
12429 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12431 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12432 the vector size as a poly_uint64 rather than an unsigned int.
12433 (current_vector_size): Change from an unsigned int to a poly_uint64.
12434 (get_vectype_for_scalar_type): Update accordingly.
12435 * tree.h (build_truth_vector_type): Take the size and number of
12436 units as a poly_uint64 rather than an unsigned int.
12437 (build_vector_type): Add a temporary overload that takes
12438 the number of units as a poly_uint64 rather than an unsigned int.
12439 * tree.c (make_vector_type): Likewise.
12440 (build_truth_vector_type): Take the number of units as a poly_uint64
12441 rather than an unsigned int.
12443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12444 Alan Hayward <alan.hayward@arm.com>
12445 David Sherwood <david.sherwood@arm.com>
12447 * target.def (get_mask_mode): Take the number of units and length
12448 as poly_uint64s rather than unsigned ints.
12449 * targhooks.h (default_get_mask_mode): Update accordingly.
12450 * targhooks.c (default_get_mask_mode): Likewise.
12451 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12452 * doc/tm.texi: Regenerate.
12454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12455 Alan Hayward <alan.hayward@arm.com>
12456 David Sherwood <david.sherwood@arm.com>
12458 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12459 * omp-general.c (omp_max_vf): Likewise.
12460 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12461 (expand_omp_simd): Handle polynomial safelen.
12462 * omp-low.c (omplow_simd_context): Add a default constructor.
12463 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12464 (lower_rec_simd_input_clauses): Update accordingly.
12465 (lower_rec_input_clauses): Likewise.
12467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12468 Alan Hayward <alan.hayward@arm.com>
12469 David Sherwood <david.sherwood@arm.com>
12471 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12472 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12473 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12474 (vect_analyze_slp_cost): Likewise.
12475 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12476 (vect_model_load_cost): Likewise.
12478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12479 Alan Hayward <alan.hayward@arm.com>
12480 David Sherwood <david.sherwood@arm.com>
12482 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12483 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12484 from an unsigned int * to a poly_uint64_pod *.
12485 (calculate_unrolling_factor): New function.
12486 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12489 Alan Hayward <alan.hayward@arm.com>
12490 David Sherwood <david.sherwood@arm.com>
12492 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12493 from an unsigned int to a poly_uint64.
12494 (_loop_vec_info::slp_unrolling_factor): Likewise.
12495 (_loop_vec_info::vectorization_factor): Change from an int
12497 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12498 (vect_get_num_vectors): New function.
12499 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12500 (vect_get_num_copies): Use vect_get_num_vectors.
12501 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12502 to an unsigned int *.
12503 (vect_analyze_data_refs): Change min_vf from an int * to a
12505 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12506 than an unsigned HOST_WIDE_INT.
12507 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12508 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12509 to an unsigned int *.
12510 (vect_analyze_data_ref_dependences): Likewise.
12511 (vect_compute_data_ref_alignment): Handle polynomial vf.
12512 (vect_enhance_data_refs_alignment): Likewise.
12513 (vect_prune_runtime_alias_test_list): Likewise.
12514 (vect_shift_permute_load_chain): Likewise.
12515 (vect_supportable_dr_alignment): Likewise.
12516 (dependence_distance_ge_vf): Take the vectorization factor as a
12517 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12518 (vect_analyze_data_refs): Change min_vf from an int * to a
12520 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12521 vfm1 as a poly_uint64 rather than an int. Make the same change
12522 for the returned bound_scalar.
12523 (vect_gen_vector_loop_niters): Handle polynomial vf.
12524 (vect_do_peeling): Likewise. Update call to
12525 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12526 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12528 * tree-vect-loop.c (vect_determine_vectorization_factor)
12529 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12530 (vect_get_known_peeling_cost): Likewise.
12531 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12532 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12533 (vect_transform_loop): Likewise. Use the lowest possible VF when
12534 updating the upper bounds of the loop.
12535 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12536 rather than an int.
12537 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12538 polynomial unroll factors.
12539 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12540 (vect_make_slp_decision): Likewise.
12541 (vect_supported_load_permutation_p): Likewise, and polynomial
12543 (vect_analyze_slp_cost): Handle polynomial vf.
12544 (vect_slp_analyze_node_operations): Likewise.
12545 (vect_slp_analyze_bb_1): Likewise.
12546 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12547 than an unsigned HOST_WIDE_INT.
12548 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12549 (vectorizable_load): Handle polynomial vf.
12550 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12552 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12555 Alan Hayward <alan.hayward@arm.com>
12556 David Sherwood <david.sherwood@arm.com>
12558 * match.pd: Handle bit operations involving three constants
12559 and try to fold one pair.
12561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12563 * tree-vect-loop-manip.c: Include gimple-fold.h.
12564 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12565 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12566 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12567 Add a path that uses a step of VF instead of 1, but disable it
12569 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12570 and niters_no_overflow parameters. Update calls to
12571 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12572 Create a new SSA name if the latter choses to use a ste other
12573 than zero, and return it via niters_vector_mult_vf_var.
12574 * tree-vect-loop.c (vect_transform_loop): Update calls to
12575 vect_do_peeling, vect_gen_vector_loop_niters and
12576 slpeel_make_loop_iterate_ntimes.
12577 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12578 (vect_gen_vector_loop_niters): Update declarations after above changes.
12580 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12582 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12583 128-bit round to integer instructions.
12584 (ceil<mode>2): Likewise.
12585 (btrunc<mode>2): Likewise.
12586 (round<mode>2): Likewise.
12588 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12590 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12591 unaligned VSX load/store on P8/P9.
12592 (expand_block_clear): Allow the use of unaligned VSX
12593 load/store on P8/P9.
12595 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12597 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12599 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12600 swap associated with both a load and a store.
12602 2018-01-02 Andrew Waterman <andrew@sifive.com>
12604 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12605 * config/riscv/riscv.md (clear_cache): Use it.
12607 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12609 * web.c: Remove out-of-date comment.
12611 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12613 * expr.c (fixup_args_size_notes): Check that any existing
12614 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12615 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12616 (emit_single_push_insn): ...here.
12618 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12620 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12621 (const_vector_encoded_nelts): New function.
12622 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12623 (const_vector_int_elt, const_vector_elt): Declare.
12624 * emit-rtl.c (const_vector_int_elt_1): New function.
12625 (const_vector_elt): Likewise.
12626 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12627 of CONST_VECTOR_ELT.
12629 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12631 * expr.c: Include rtx-vector-builder.h.
12632 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12633 directly on the tree encoding.
12634 (const_vector_from_tree): Likewise.
12635 * optabs.c: Include rtx-vector-builder.h.
12636 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12637 sequence of "u" values.
12638 * vec-perm-indices.c: Include rtx-vector-builder.h.
12639 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12640 directly on the vec_perm_indices encoding.
12642 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12644 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12645 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12646 * rtx-vector-builder.h: New file.
12647 * rtx-vector-builder.c: Likewise.
12648 * rtl.h (rtx_def::u2): Add a const_vector field.
12649 (CONST_VECTOR_NPATTERNS): New macro.
12650 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12651 (CONST_VECTOR_DUPLICATE_P): Likewise.
12652 (CONST_VECTOR_STEPPED_P): Likewise.
12653 (CONST_VECTOR_ENCODED_ELT): Likewise.
12654 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12655 (unwrap_const_vec_duplicate): Likewise.
12656 (const_vec_series_p): Check for a non-duplicated vector encoding.
12657 Say that the function only returns true for integer vectors.
12658 * emit-rtl.c: Include rtx-vector-builder.h.
12659 (gen_const_vec_duplicate_1): Delete.
12660 (gen_const_vector): Call gen_const_vec_duplicate instead of
12661 gen_const_vec_duplicate_1.
12662 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12663 (gen_const_vec_duplicate): Use rtx_vector_builder.
12664 (gen_const_vec_series): Likewise.
12665 (gen_rtx_CONST_VECTOR): Likewise.
12666 * config/powerpcspe/powerpcspe.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.
12670 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12671 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12672 Build a new vector rather than modifying a CONST_VECTOR in-place.
12673 (handle_special_swappables): Update call accordingly.
12675 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12677 * simplify-rtx.c (simplify_const_binary_operation): Use
12678 CONST_VECTOR_ELT instead of XVECEXP.
12680 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12682 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12683 the selector elements to be different from the data elements
12684 if the selector is a VECTOR_CST.
12685 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12686 ssizetype for the selector.
12688 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12690 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12691 before testing each element individually.
12692 * tree-vect-generic.c (lower_vec_perm): Likewise.
12694 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12696 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12697 * selftest-run-tests.c (selftest::run_tests): Call it.
12698 * vector-builder.h (vector_builder::operator ==): New function.
12699 (vector_builder::operator !=): Likewise.
12700 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12701 (vec_perm_indices::all_from_input_p): New function.
12702 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12703 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12704 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12705 instead of reading the VECTOR_CST directly. Detect whether both
12706 vector inputs are the same before constructing the vec_perm_indices,
12707 and update the number of inputs argument accordingly. Use the
12708 utility functions added above. Only construct sel2 if we need to.
12710 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12712 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12713 the broadcast of the low byte.
12714 (expand_mult_highpart): Use an explicit encoding for the permutes.
12715 * optabs-query.c (can_mult_highpart_p): Likewise.
12716 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12717 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12718 (vectorizable_bswap): Likewise.
12719 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12720 explicit encoding for the power-of-2 permutes.
12721 (vect_permute_store_chain): Likewise.
12722 (vect_grouped_load_supported): Likewise.
12723 (vect_permute_load_chain): Likewise.
12725 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12727 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12728 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12729 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12730 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12731 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12732 (vect_gen_perm_mask_any): Likewise.
12734 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12736 * int-vector-builder.h: New file.
12737 * vec-perm-indices.h: Include int-vector-builder.h.
12738 (vec_perm_indices): Redefine as an int_vector_builder.
12739 (auto_vec_perm_indices): Delete.
12740 (vec_perm_builder): Redefine as a stand-alone class.
12741 (vec_perm_indices::vec_perm_indices): New function.
12742 (vec_perm_indices::clamp): Likewise.
12743 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12744 (vec_perm_indices::new_vector): New function.
12745 (vec_perm_indices::new_expanded_vector): Update for new
12746 vec_perm_indices class.
12747 (vec_perm_indices::rotate_inputs): New function.
12748 (vec_perm_indices::all_in_range_p): Operate directly on the
12749 encoded form, without computing elided elements.
12750 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12751 encoding. Update for new vec_perm_indices class.
12752 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12753 the given vec_perm_builder.
12754 (expand_vec_perm_var): Update vec_perm_builder constructor.
12755 (expand_mult_highpart): Use vec_perm_builder instead of
12756 auto_vec_perm_indices.
12757 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12758 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12759 or double series encoding as appropriate.
12760 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12761 vec_perm_indices instead of auto_vec_perm_indices.
12762 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12763 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12764 (vect_permute_store_chain): Likewise.
12765 (vect_grouped_load_supported): Likewise.
12766 (vect_permute_load_chain): Likewise.
12767 (vect_shift_permute_load_chain): Likewise.
12768 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12769 (vect_transform_slp_perm_load): Likewise.
12770 (vect_schedule_slp_instance): Likewise.
12771 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12772 (vectorizable_mask_load_store): Likewise.
12773 (vectorizable_bswap): Likewise.
12774 (vectorizable_store): Likewise.
12775 (vectorizable_load): Likewise.
12776 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12777 vec_perm_indices instead of auto_vec_perm_indices. Use
12778 tree_to_vec_perm_builder to read the vector from a tree.
12779 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12780 vec_perm_builder instead of a vec_perm_indices.
12781 (have_whole_vector_shift): Use vec_perm_builder and
12782 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12783 truncation to calc_vec_perm_mask_for_shift.
12784 (vect_create_epilog_for_reduction): Likewise.
12785 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12786 from auto_vec_perm_indices to vec_perm_indices.
12787 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12788 instead of changing individual elements.
12789 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12790 the vector in d.perm.
12791 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12792 from auto_vec_perm_indices to vec_perm_indices.
12793 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12794 instead of changing individual elements.
12795 (arm_vectorize_vec_perm_const): Use new_vector to install
12796 the vector in d.perm.
12797 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12798 Update vec_perm_builder constructor.
12799 (rs6000_expand_interleave): Likewise.
12800 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12801 (rs6000_expand_interleave): Likewise.
12803 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12805 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12806 to qimode could truncate the indices.
12807 * optabs.c (expand_vec_perm_var): Likewise.
12809 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12811 * Makefile.in (OBJS): Add vec-perm-indices.o.
12812 * vec-perm-indices.h: New file.
12813 * vec-perm-indices.c: Likewise.
12814 * target.h (vec_perm_indices): Replace with a forward class
12816 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12817 * optabs.h: Include vec-perm-indices.h.
12818 (expand_vec_perm): Delete.
12819 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12820 (expand_vec_perm_const): Declare.
12821 * target.def (vec_perm_const_ok): Replace with...
12822 (vec_perm_const): ...this new hook.
12823 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12824 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12825 * doc/tm.texi: Regenerate.
12826 * optabs.def (vec_perm_const): Delete.
12827 * doc/md.texi (vec_perm_const): Likewise.
12828 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12829 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12830 expand_vec_perm for constant permutation vectors. Assert that
12831 the mode of variable permutation vectors is the integer equivalent
12832 of the mode that is being permuted.
12833 * optabs-query.h (selector_fits_mode_p): Declare.
12834 * optabs-query.c: Include vec-perm-indices.h.
12835 (selector_fits_mode_p): New function.
12836 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12837 is defined, instead of checking whether the vec_perm_const_optab
12838 exists. Use targetm.vectorize.vec_perm_const instead of
12839 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12840 fit in the vector mode before using a variable permute.
12841 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12842 vec_perm_indices instead of an rtx.
12843 (expand_vec_perm): Replace with...
12844 (expand_vec_perm_const): ...this new function. Take the selector
12845 as a vec_perm_indices rather than an rtx. Also take the mode of
12846 the selector. Update call to shift_amt_for_vec_perm_mask.
12847 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12848 Use vec_perm_indices::new_expanded_vector to expand the original
12849 selector into bytes. Check whether the indices fit in the vector
12850 mode before using a variable permute.
12851 (expand_vec_perm_var): Make global.
12852 (expand_mult_highpart): Use expand_vec_perm_const.
12853 * fold-const.c: Includes vec-perm-indices.h.
12854 * tree-ssa-forwprop.c: Likewise.
12855 * tree-vect-data-refs.c: Likewise.
12856 * tree-vect-generic.c: Likewise.
12857 * tree-vect-loop.c: Likewise.
12858 * tree-vect-slp.c: Likewise.
12859 * tree-vect-stmts.c: Likewise.
12860 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12862 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12863 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12864 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12865 (aarch64_vectorize_vec_perm_const): ...this new function.
12866 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12867 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12868 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12869 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12870 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12871 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12872 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12874 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12875 check for NEON modes.
12876 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12877 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12878 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12879 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12881 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12882 the old VEC_PERM_CONST conditions.
12883 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12884 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12885 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12886 (ia64_vectorize_vec_perm_const_ok): Merge into...
12887 (ia64_vectorize_vec_perm_const): ...this new function.
12888 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12889 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12890 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12891 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12892 * config/mips/mips.c (mips_expand_vec_perm_const)
12893 (mips_vectorize_vec_perm_const_ok): Merge into...
12894 (mips_vectorize_vec_perm_const): ...this new function.
12895 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12896 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12897 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12898 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12899 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12900 (rs6000_expand_vec_perm_const): Delete.
12901 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12903 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12904 (altivec_expand_vec_perm_const_le): Take each operand individually.
12905 Operate on constant selectors rather than rtxes.
12906 (altivec_expand_vec_perm_const): Likewise. Update call to
12907 altivec_expand_vec_perm_const_le.
12908 (rs6000_expand_vec_perm_const): Delete.
12909 (rs6000_vectorize_vec_perm_const_ok): Delete.
12910 (rs6000_vectorize_vec_perm_const): New function.
12911 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12912 an element count and rtx array.
12913 (rs6000_expand_extract_even): Update call accordingly.
12914 (rs6000_expand_interleave): Likewise.
12915 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12916 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12917 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12918 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12919 (rs6000_expand_vec_perm_const): Delete.
12920 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12921 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12922 (altivec_expand_vec_perm_const_le): Take each operand individually.
12923 Operate on constant selectors rather than rtxes.
12924 (altivec_expand_vec_perm_const): Likewise. Update call to
12925 altivec_expand_vec_perm_const_le.
12926 (rs6000_expand_vec_perm_const): Delete.
12927 (rs6000_vectorize_vec_perm_const_ok): Delete.
12928 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12929 reference to the SPE evmerge intructions.
12930 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12931 an element count and rtx array.
12932 (rs6000_expand_extract_even): Update call accordingly.
12933 (rs6000_expand_interleave): Likewise.
12934 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12935 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12937 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12939 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12941 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12942 vector mode and that that mode matches the mode of the data
12944 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12945 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12946 directly using expand_vec_perm_1 when forcing selectors into
12948 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12950 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12952 * optabs-query.h (can_vec_perm_p): Delete.
12953 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12954 * optabs-query.c (can_vec_perm_p): Split into...
12955 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12956 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12957 particular selector is valid.
12958 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12959 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12960 (vect_grouped_load_supported): Likewise.
12961 (vect_shift_permute_load_chain): Likewise.
12962 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12963 (vect_transform_slp_perm_load): Likewise.
12964 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12965 (vectorizable_bswap): Likewise.
12966 (vect_gen_perm_mask_checked): Likewise.
12967 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12968 implementations of variable permutation vectors into account
12969 when deciding which selector to use.
12970 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12971 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12972 with a false third argument.
12973 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12974 to test whether the constant selector is valid and can_vec_perm_var_p
12975 to test whether a variable selector is valid.
12977 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12979 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12980 * optabs-query.c (can_vec_perm_p): Likewise.
12981 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12982 instead of vec_perm_indices.
12983 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12984 (vect_gen_perm_mask_checked): Likewise,
12985 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12986 (vect_gen_perm_mask_checked): Likewise,
12988 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12990 * optabs-query.h (qimode_for_vec_perm): Declare.
12991 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12992 (qimode_for_vec_perm): ...this new function.
12993 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12995 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12997 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12998 does not have a conditional at the top.
13000 2018-01-02 Richard Biener <rguenther@suse.de>
13002 * ipa-inline.c (big_speedup_p): Fix expression.
13004 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13007 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13010 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13014 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13015 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13016 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13017 cond_taken_branch_cost 3->4.
13019 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13021 PR tree-optimization/83581
13022 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13023 TODO_cleanup_cfg if any changes have been made.
13025 PR middle-end/83608
13026 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13027 convert_modes if target mode has the right side, but different mode
13030 PR middle-end/83609
13031 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13032 last argument when extracting from CONCAT. If either from_real or
13033 from_imag is NULL, use expansion through memory. If result is not
13034 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13035 the parts directly to inner mode, if even that fails, use expansion
13038 PR middle-end/83623
13039 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13040 check for bswap in mode rather than HImode and use that in expand_unop
13043 Copyright (C) 2018 Free Software Foundation, Inc.
13045 Copying and distribution of this file, with or without modification,
13046 are permitted in any medium without royalty provided the copyright
13047 notice and this notice are preserved.