1 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
5 even if the mask is not all ones.
8 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
10 (ix86_gimple_fold_builtin): Likewise.
13 * config/i386/i386.c: Include tree-vector-builder.h.
14 (ix86_vector_shift_count): New function.
15 (ix86_fold_builtin): Fold shift builtins by scalar count.
16 (ix86_gimple_fold_builtin): Likewise.
18 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
19 _mm512_setzero): New intrinsics.
21 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
25 code generation for cases where splatting a value is not useful.
26 * simplify-rtx.c (simplify_ternary_operation): Simplify
27 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
30 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
32 * config.gcc: Support "goldmont-plus".
33 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
35 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
36 PROCESSOR_GOLDMONT_PLUS.
37 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
38 (processor_target_table): Add "goldmont-plus".
39 (PTA_GOLDMONT_PLUS): Define.
40 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
41 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
42 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
43 (fold_builtin_cpu): Add "goldmont-plus".
44 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
45 (ix86_option_override_internal): Add "goldmont-plus".
46 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
47 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
48 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
49 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
51 2018-05-17 Richard Biener <rguenther@suse.de>
53 PR tree-optimization/85757
54 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
55 remove defs that only feed that PHI from further processing.
57 2018-05-16 Jim Wilson <jimw@sifive.com>
59 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
61 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
63 2018-05-16 Mark Wielaard <mark@klomp.org>
65 * dwarf2out.c (count_index_strings): New function.
66 (output_indirect_strings): Call count_index_strings and generate
67 header for dwarf_version >= 5.
69 2018-05-16 Mark Wielaard <mark@klomp.org>
71 * dwarf2out.c (dwarf_FORM): New function.
72 (set_indirect_string): Use dwarf_FORM.
73 (reset_indirect_string): Likewise.
74 (size_of_die): Likewise.
75 (value_format): Likewise.
76 (output_die): Likewise.
77 (add_skeleton_AT_string): Likewise.
78 (output_macinfo_op): Likewise.
79 (index_string): Likewise.
80 (output_index_string_offset): Likewise.
81 (output_index_string): Likewise.
82 (count_index_strings): Likewise.
84 2018-05-16 Carl Love <cel@us.ibm.com>
86 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
87 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
89 2018-05-16 Martin Jambor <mjambor@suse.cz>
91 * ipa-prop.c (ipa_free_all_edge_args): Remove.
92 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
94 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
96 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
97 (fnma<mode>4): Likewise.
98 (fms<mode>4): Likewise.
99 (fnms<mode>4): Likewise.
100 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
101 (aarch64_fnma<mode>4): Likewise.
102 (aarch64_fms<mode>4): Likewise.
103 (aarch64_fnms<mode>4): Likewise.
104 (aarch64_fnmadd<mode>4): Likewise.
106 2018-05-16 Jason Merrill <jason@redhat.com>
108 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
110 2018-05-16 Richard Biener <rguenther@suse.de>
112 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
113 (dump_stmt_cost): Declare.
114 (add_stmt_cost): Dump cost we add.
115 (add_stmt_costs): New function.
116 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
118 (vect_analyze_stmt): Adjust prototype.
119 (vectorizable_condition): Likewise.
120 (vectorizable_live_operation): Likewise.
121 (vectorizable_reduction): Likewise.
122 (vectorizable_induction): Likewise.
123 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
124 cost vector to pass to vectorizable_ and record afterwards.
125 (vect_model_reduction_cost): Take cost vector argument and adjust.
126 (vect_model_induction_cost): Likewise.
127 (vectorizable_reduction): Likewise.
128 (vectorizable_induction): Likewise.
129 (vectorizable_live_operation): Likewise.
130 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
131 SLP_TREE_NUMBER_OF_VEC_STMTS.
132 (vect_analyze_slp_cost_1): Remove.
133 (vect_analyze_slp_cost): Likewise.
134 (vect_slp_analyze_node_operations): Take visited args and
135 a target cost vector. Avoid processing already visited stmt sets.
136 (vect_slp_analyze_operations): Use a local cost vector to gather
137 costs and register those of non-discarded instances.
138 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
139 (vect_schedule_slp_instance): Remove copying of
140 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
142 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
143 adding cost. Record cost entry location.
144 (vect_prologue_cost_for_slp_op): Function to compute cost of
145 a constant or invariant generated for SLP vect in the prologue,
146 split out from vect_analyze_slp_cost_1.
147 (vect_model_simple_cost): Make static. Adjust for SLP costing.
148 (vect_model_promotion_demotion_cost): Likewise.
149 (vect_model_store_cost): Likewise, make static.
150 (vect_model_load_cost): Likewise.
151 (vectorizable_bswap): Add cost vector arg and adjust.
152 (vectorizable_call): Likewise.
153 (vectorizable_simd_clone_call): Likewise.
154 (vectorizable_conversion): Likewise.
155 (vectorizable_assignment): Likewise.
156 (vectorizable_shift): Likewise.
157 (vectorizable_operation): Likewise.
158 (vectorizable_store): Likewise.
159 (vectorizable_load): Likewise.
160 (vectorizable_condition): Likewise.
161 (vectorizable_comparison): Likewise.
162 (can_vectorize_live_stmts): Likewise.
163 (vect_analyze_stmt): Likewise.
164 (vect_transform_stmt): Adjust calls to vectorizable_*.
165 * tree-vectorizer.c: Include gimple-pretty-print.h.
166 (dump_stmt_cost): New function.
168 2018-05-16 Richard Biener <rguenther@suse.de>
170 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
171 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
172 * tree-ssa-dse.c: Include tree-ssa-loop.h.
173 (check_name): New callback.
174 (dse_classify_store): Track cycles via a visited bitmap of PHI
175 defs and simplify handling of in-loop and across loop dead stores
176 and properly fail for loop-variant refs. Handle byte-tracking with
177 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
180 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
182 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
183 (vect_get_mask_type_for_stmt): Likewise.
184 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
186 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
187 to determine the statement's vector type and the vector type that
188 should be used for calculating nunits. Deal with cases in which
189 the type has to be deferred.
190 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
191 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
192 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
193 (vect_determine_vf_for_stmt): New functions, split out from...
194 (vect_determine_vectorization_factor): ...here.
195 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
196 (vect_get_mask_type_for_stmt): New functions, split out from
197 vect_determine_vectorization_factor.
199 2018-05-16 Richard Biener <rguenther@suse.de>
201 * tree-cfg.c (verify_gimple_assign_ternary): Properly
202 verify the [VEC_]COND_EXPR embedded comparison.
204 2018-05-15 Martin Sebor <msebor@redhat.com>
206 PR tree-optimization/85753
207 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
208 RECORD_TYPE in addition to ARRAY_TYPE.
210 2018-05-15 Martin Sebor <msebor@redhat.com>
213 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
215 2018-05-15 Richard Biener <rguenther@suse.de>
217 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
218 add by_clobber_p one. Change algorithm to collect all defs
219 representing uses we need to walk and try reducing them to
220 a single one before failing.
221 (dse_dom_walker::dse_optimize_stmt): Adjust.
223 2018-05-13 Mark Wielaard <mark@klomp.org>
225 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
226 (size_of_loc_descr): Likewise.
227 (output_loc_operands): Likewise.
228 (output_loc_operands_raw): Likewise.
229 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
230 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
231 (hash_loc_operands): Likewise.
232 (compare_loc_operands): Likewise.
234 2018-05-14 Mark Wielaard <mark@klomp.org>
236 * dwarf2out.c (count_index_addrs): New function.
237 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
239 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
241 PR tree-optimization/83648
242 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
243 return value as malloc candidate.
245 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
248 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
249 as true in call to suggest_attribute.
251 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
253 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
256 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
258 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
260 (aarch64_crypto_aesd_fused): Likewise.
262 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
264 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
265 (movsi_aarch64): Likewise.
266 (load_pairsi): Likewise.
267 (load_pairdi): Likewise.
268 (store_pairsi): Likewise.
269 (store_pairdi): Likewise.
270 (load_pairsf): Likewise.
271 (load_pairdf): Likewise.
272 (store_pairsf): Likewise.
273 (store_pairdf): Likewise.
274 (zero_extend): Likewise.
275 (trunc): Swap alternatives.
276 (fcvt_target): Add '?' to prefer w over r.
278 2018-05-14 Jakub Jelinek <jakub@redhat.com>
281 * config/i386/i386.md: Disallow non-commutative arithmetics in
282 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
283 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
284 in the peephole2 before it.
286 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
288 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
289 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
290 (ix86_handle_option): Handle -mcldemote.
291 * config.gcc: New header.
292 * config/i386/cldemoteintrin.h: New file.
293 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
294 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
296 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
297 OPTION_MASK_ISA_CLDEMOTE.
298 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
299 (ix86_valid_target_attribute_inner_p): Ditto.
300 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
301 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
302 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
303 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
304 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
306 * config/i386/i386.opt: Add -mcldemote.
307 * config/i386/x86intrin.h: New header.
308 * doc/invoke.texi: Add -mcldemote.
310 2018-05-14 Richard Biener <rguenther@suse.de>
312 * doc/match-and-simplify.texi: Adjust :s documentation.
314 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
316 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
317 intended memcpy size.
318 (REORDER_45): Likewise.
320 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
323 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
324 * vec.c (qsort_chk): Use gcc_qsort.
325 * Makefile.in (OBJS-libcommon): Add sort.o.
326 (build/sort.o): New target. Use it...
327 (BUILD_RTL): ... here, and...
328 (build/gencfn-macros): ... here, and...
329 (build/genmatch): ... here.
331 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
332 Chung-Ju Wu <jasonwucj@gmail.com>
334 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
335 * config/nds32/nds32-graywolf.md: New file.
336 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
337 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
339 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
340 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
341 * config/nds32/nds32.md (pipeline_model): Add graywolf.
342 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
343 * config/nds32/pipelines.md: Include n15 settings.
345 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
346 Chung-Ju Wu <jasonwucj@gmail.com>
348 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
349 * config/nds32/nds32-n13.md: New file.
350 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
351 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
353 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
354 * config/nds32/nds32.md (pipeline_model): Add n13.
355 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
356 * config/nds32/pipelines.md: Include n13 settings.
358 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
359 Chung-Ju Wu <jasonwucj@gmail.com>
361 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
362 * config/nds32/nds32-n10.md: New file.
363 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
364 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
366 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
367 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
368 * config/nds32/nds32.md (pipeline_model): Add n10.
369 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
370 * config/nds32/pipelines.md: Include n10 settings.
372 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
373 Kito Cheng <kito.cheng@gmail.com>
374 Chung-Ju Wu <jasonwucj@gmail.com>
376 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
377 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
378 Add enum values for DSP extension instructions.
379 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
381 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
382 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
384 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
385 * config/nds32/nds32-dspext.md: New file for DSP implementation.
386 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
387 * config/nds32/nds32-intrinsic.md: Likewise.
388 * config/nds32/nds32_intrinsic.h: Likewise.
389 * config/nds32/nds32-md-auxiliary.c: Likewise.
390 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
391 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
392 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
393 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
394 * config/nds32/nds32-protos.h: New declarations for DSP extension.
395 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
396 TYPE_DMAC in switch statement.
397 * config/nds32/nds32.c: New checking and implementation for DSP
398 extension instructions.
399 * config/nds32/nds32.h: Likewise.
400 * config/nds32/nds32.md: Likewise.
401 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
402 * config/nds32/predicates.md: Implement new predicates for DSP
405 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
407 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
408 Reformat alternatives and attributes so it is easier to identify
409 which constraints/attributes go with which instruction.
410 (mov<mode>_hardfloat32, FMOVE64): Likewise.
411 (mov<mode>_softfloat32, FMOVE64): Likewise.
412 (mov<mode>_hardfloat64, FMOVE64): Likewise.
413 (mov<mode>_softfloat64, FMOVE64): Likewise.
415 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
417 * doc/extend.texi (PowerPC Built-in Functions): Rename this
419 (Basic PowerPC Built-in Functions): The new name of the
420 subsection previously known as "PowerPC Built-in Functions".
421 (Basic PowerPC Built-in Functions Available on all Configurations):
423 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
424 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
425 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
426 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
428 2018-05-11 Martin Jambor <mjambor@suse.cz>
431 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
434 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
437 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
439 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
441 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
442 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
443 (ix86_handle_option): Handle -mwaitpkg.
444 * config.gcc: New header.
445 * config/i386/cpuid.h (bit_WAITPKG): New bit.
446 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
447 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
449 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
450 OPTION_MASK_ISA_WAITPKG.
451 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
452 (ix86_option_override_internal): Add PTA_WAITPKG.
453 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
454 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
456 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
457 __builtin_ia32_umwait and __builtin_ia32_tpause.
458 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
459 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
460 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
461 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
462 UNSPECV_TPAUSE): New.
463 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
464 * config/i386/i386.opt: Add -mwaitpkg.
465 * config/i386/waitpkgintrin.h: New file.
466 * config/i386/x86intrin.h: New header.
467 * doc/invoke.texi: Add -mwaitpkg.
469 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
472 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
474 (cortex-m0): Use armv6s-m isa.
475 (cortex-m0plus): Likewise.
476 (cortex-m1): Likewise.
477 (cortex-m0.small-multiply): Likewise.
478 (cortex-m0plus.small-multiply): Likewise.
479 (cortex-m1.small-multiply): Likewise.
481 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
482 Jakub Jelinek <jakub@redhat.com>
484 PR tree-optimization/85692
485 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
486 source permute as well.
488 2018-05-11 Martin Liska <mliska@suse.cz>
491 * doc/extend.texi: Document LLVM style format for no_sanitize
494 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
496 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
497 mode_supports_vsx_dform_quad to mode_supports_dq_form.
498 (mode_supports_vsx_dform_quad): Likewise.
499 (mode_supports_vmx_dform): Move these functions to be next to the
500 other mode_supports functions.
501 (mode_supports_dq_form): Likewise.
502 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
503 mode_supports_dq_form.
504 (reg_offset_addressing_ok_p): Likewise.
505 (offsettable_ok_by_alignment): Likewise.
506 (rs6000_legitimate_offset_address_p): Likewise.
507 (legitimate_lo_sum_address_p): Likewise.
508 (rs6000_legitimize_address): Likewise.
509 (rs6000_legitimize_reload_address): Likewise.
510 (rs6000_secondary_reload_inner): Likewise.
511 (rs6000_preferred_reload_class): Likewise.
512 (rs6000_output_move_128bit): Likewise.
514 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
516 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
517 Generate SImode target register for null target.
518 <case IX86_BUILTIN_XGETBV>: Ditto.
519 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
520 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
522 2018-05-10 Carl Love <cel@us.ibm.com>
524 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
525 dcbtt and dcbtstt if operands[2] is 0.
527 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
530 * config/i386/sse.md (usadv64qi): New expander.
532 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
534 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
535 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
536 -maltivec=be support.
537 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
538 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
539 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
540 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
541 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
542 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
543 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
544 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
545 altivec_vsumsws): Adjust.
546 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
547 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
548 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
550 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
551 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
552 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
553 (altivec_lve<VI_char>x): Delete expand.
554 (*altivec_lve<VI_char>x_internal): Rename to...
555 (altivec_lve<VI_char>x): ... this.
556 (altivec_lvxl_<mode>): Delete expand.
557 (*altivec_lvxl_<mode>_internal): Rename to ...
558 (altivec_lvxl_<mode>): ... this.
559 (altivec_stvxl_<mode>): Delete expand.
560 (*altivec_stvxl_<mode>_internal): Rename to ...
561 (altivec_stvxl_<mode>): ... this.
562 (altivec_stve<VI_char>x): Delete expand.
563 (*altivec_stve<VI_char>x_internal): Rename to ...
564 (altivec_stve<VI_char>x): ... this.
565 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
566 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
567 reduc_plus_scal_<mode>): Adjust.
568 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
570 (rs6000_cpu_cpp_builtins): Adjust.
571 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
572 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
573 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
574 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
575 -maltivec=be support.
576 (rs6000_split_vec_extract_var): Adjust.
577 (rs6000_split_v4si_init): Adjust.
578 (swap_selector_for_mode): Delete.
579 (altivec_expand_lvx_be, altivec_expand_stvx_be,
580 altivec_expand_stvex_be): Delete.
581 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
582 -maltivec=be support.
583 (rs6000_gimple_fold_builtin): Ditto.
584 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
586 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
587 (TARGET_DIRECT_MOVE_64BIT): Adjust.
588 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
589 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
590 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
591 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
592 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
593 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
594 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
595 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
596 anonymous split): Adjust.
597 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
598 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
600 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
602 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
603 when --with-gxx-include-dir is also specified.
604 * configure: Regenerate.
606 2018-05-09 Jim Wilson <jimw@sifive.com>
609 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
610 * config/riscv/t-withmultilib: New.
611 * config/riscv/withmultilib.h: New.
612 * doc/install.texi: Document RISC-V --with-multilib-list support.
614 2018-05-09 Richard Biener <rguenther@suse.de>
616 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
618 (vect_bb_vectorization_profitable_p): Adjust. Compute
619 actual scalar cost using the cost vector and the add_stmt_cost
622 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
624 PR rtl-optimization/85645
625 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
626 in the REG_CFA_REGISTER note for LR, don't leave it empty.
628 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
630 PR rtl-optimization/85645
631 * shrink-wrap.c (spread_components): Return a boolean saying if
632 anything was changed.
633 (try_shrink_wrapping_separate): Iterate spread_components until
634 nothing changes anymore.
636 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
638 PR rtl-optimization/85645
639 * regrename.c (build_def_use): Also kill the chains that include the
640 destination of a REG_CFA_REGISTER note.
642 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
644 PR rtl-optimization/85645
645 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
646 insn that has a REG_CFA_REGISTER note.
648 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
650 * cfgexpand.c (expand_clobber): New function.
651 (expand_gimple_stmt_1): Use it.
652 * tree-vect-stmts.c (vect_clobber_variable): New function,
654 (vectorizable_simd_clone_call): ...here.
655 (vectorizable_store): Emit a clobber either side of an
656 IFN_STORE_LANES sequence.
657 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
659 2018-05-09 Tom de Vries <tom@codesourcery.com>
662 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
663 (define_insn "trap_if_false"): Add exit after trap.
665 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
667 PR rtl-optimization/85638
668 * bb-reorder.c: Include common/common-target.h.
669 (create_forwarder_block): New function extracted from...
670 (fix_up_crossing_landing_pad): ...here. Rename into...
671 (dw2_fix_up_crossing_landing_pad): ...this.
672 (sjlj_fix_up_crossing_landing_pad): New function.
673 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
674 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
675 from both partitions and exit the loop after one iteration.
677 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
680 * doc/extend.texi (PowerPC Built-in Functions): Rename this
682 (Basic PowerPC Built-in Functions): The new name of the
683 subsection previously known as "PowerPC Built-in Functions".
684 (Basic PowerPC Built-in Functions Available on all Configurations):
686 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
688 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
690 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
692 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
695 2018-05-08 Jim Wilson <jimw@sifive.com>
697 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
698 (LD_EMUL_SUFFIX): New.
701 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
703 * doc/extend.texi (PowerPC Built-in Functions): Rename this
705 (Basic PowerPC Built-in Functions): The new name of the
706 subsection previously known as "PowerPC Built-in Functions".
707 (Basic PowerPC Built-in Functions Available on all Configurations):
709 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
711 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
713 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
715 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
718 2018-05-08 Jakub Jelinek <jakub@redhat.com>
721 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
722 after cmpelim optimization.
724 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
726 * config.gcc: Support "goldmont".
727 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
728 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
730 * config/i386/i386.c (m_GOLDMONT): Define.
731 (processor_target_table): Add "goldmont".
732 (PTA_GOLDMONT): Define.
733 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
734 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
735 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
736 (fold_builtin_cpu): Add "goldmont".
737 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
738 (ix86_option_override_internal): Add "goldmont".
739 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
740 (processor_type): Add PROCESSOR_GOLDMONT.
741 * config/i386/i386.md: Add CPU "glm".
742 * config/i386/glm.md: New file.
743 * config/i386/x86-tune.def: Add m_GOLDMONT.
744 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
746 2018-05-08 Jakub Jelinek <jakub@redhat.com>
749 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
751 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
752 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
753 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
756 * config/i386/i386.c (ix86_fold_builtin): Handle
757 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
760 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
761 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
763 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
766 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
767 (check_arch): Likewise.
768 (check_fpu): Return the result rather than printing it.
769 (end arch): Fix operator precedence.
771 (END): Print the result from check_fpu.
773 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
774 Alan Hayward <alan.hayward@arm.com>
775 David Sherwood <david.sherwood@arm.com>
777 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
778 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
779 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
780 (*fcmuo<mode>_and): New patterns.
782 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
784 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
785 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
786 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
787 (cmp_op, sve_imm_con): New code attributes.
788 (SVE_COND_INT_CMP, imm_con): Delete.
789 (cmp_op): Remove above unspecs from int attribute.
790 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
792 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
793 comparison-specific unspecs.
794 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
795 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
796 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
797 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
798 (*vec_fcm<cmp_op><mode>): Rename to...
799 (*fcm<cmp_op><mode>): ...this and adjust likewise.
800 (*vec_fcmuo<mode>): Rename to...
801 (*fcmuo<mode>): ...this and adjust likewise.
802 (*pred_fcm<cmp_op><mode>): New pattern.
803 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
804 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
806 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
808 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
809 (aarch64_emit_sve_predicated_cond): New function.
810 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
811 (aarch64_emit_unspec_cond_or): Replace with...
812 (aarch64_emit_sve_or_conds): ...this new function. Use
813 aarch64_emit_sve_ptrue_op for the individual comparisons and
814 aarch64_emit_binop to OR them together.
815 (aarch64_emit_inverted_unspec_cond): Replace with...
816 (aarch64_emit_sve_inverted_cond): ...this new function. Use
817 aarch64_emit_sve_ptrue_op for the comparison and
818 aarch64_emit_unop to invert the result.
819 (aarch64_expand_sve_vec_cmp_float): Update after the above
820 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
822 2018-05-07 Nathan Sidwell <nathan@acm.org>
824 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
825 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
826 (Backwards Compatibility): Likewise.
828 2018-05-07 Luis Machado <luis.machado@linaro.org>
832 2018-05-07 Luis Machado <luis.machado@linaro.org>
834 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
835 <prefetch_dynamic_strides>: New const bool field.
836 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
837 prefetch_dynamic_strides.
838 (exynosm1_prefetch_tune): Likewise.
839 (thunderxt88_prefetch_tune): Likewise.
840 (thunderx_prefetch_tune): Likewise.
841 (thunderx2t99_prefetch_tune): Likewise.
842 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
843 (aarch64_override_options_internal): Update to set
844 PARAM_PREFETCH_DYNAMIC_STRIDES.
845 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
846 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
847 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
848 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
849 prefetch-dynamic-strides setting.
851 2018-05-07 Luis Machado <luis.machado@linaro.org>
853 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
854 <minimum_stride>: New const int field.
855 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
856 minimum_stride field.
857 (exynosm1_prefetch_tune): Likewise.
858 (thunderxt88_prefetch_tune): Likewise.
859 (thunderx_prefetch_tune): Likewise.
860 (thunderx2t99_prefetch_tune): Likewise.
861 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
862 (aarch64_override_options_internal): Update to set
863 PARAM_PREFETCH_MINIMUM_STRIDE.
864 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
865 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
866 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
867 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
868 stride is constant and is below the minimum stride threshold.
870 2018-05-07 Luis Machado <luis.machado@linaro.org>
872 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
875 2018-05-07 Luis Machado <luis.machado@linaro.org>
877 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
878 <prefetch_dynamic_strides>: New const bool field.
879 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
880 prefetch_dynamic_strides.
881 (exynosm1_prefetch_tune): Likewise.
882 (thunderxt88_prefetch_tune): Likewise.
883 (thunderx_prefetch_tune): Likewise.
884 (thunderx2t99_prefetch_tune): Likewise.
885 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
886 (aarch64_override_options_internal): Update to set
887 PARAM_PREFETCH_DYNAMIC_STRIDES.
888 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
889 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
890 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
891 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
892 prefetch-dynamic-strides setting.
894 2018-05-07 Luis Machado <luis.machado@linaro.org>
896 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
897 <minimum_stride>: New const int field.
898 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
899 minimum_stride field.
900 (exynosm1_prefetch_tune): Likewise.
901 (thunderxt88_prefetch_tune): Likewise.
902 (thunderx_prefetch_tune): Likewise.
903 (thunderx2t99_prefetch_tune): Likewise.
904 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
905 (aarch64_override_options_internal): Update to set
906 PARAM_PREFETCH_MINIMUM_STRIDE.
907 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
908 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
909 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
910 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
911 stride is constant and is below the minimum stride threshold.
913 2018-05-06 Jakub Jelinek <jakub@redhat.com>
916 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
917 the type is addressable. Don't force op into register if it has
920 2018-05-05 Roland McGrath <mcgrathr@google.com>
923 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
924 any section for which we don't know a specific type it should have,
925 regardless of name. Previously this was done only for the exact
926 names ".init_array", ".fini_array", and ".preinit_array".
927 (default_elf_asm_named_section): Add comment about
928 relationship with default_section_type_flags and SECTION_NOTYPE.
929 (get_section): Don't consider it a type conflict if one side has
930 SECTION_NOTYPE and the other doesn't, as long as neither has the
931 SECTION_BSS et al used in the default_section_type_flags logic.
933 2018-05-05 Tom de Vries <tom@codesourcery.com>
936 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
937 (workaround_barsyncs): New function.
938 (nvptx_reorg): Use workaround_barsyncs.
939 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
940 (define_expand "nvptx_membar_cta"): New define_expand.
941 (define_insn "*nvptx_membar_cta"): New insn.
943 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
945 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
946 To improve optimization opportunities.
947 * builtin-types.def: The new needed builtin types for the above.
949 2018-05-04 Richard Biener <rguenther@suse.de>
951 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
952 * gimple-ssa-store-merging.c
953 (imm_store_chain_info::output_merged_store): Remove redundant create,
954 release split_store vector contents on failure.
955 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
956 scalar stmt vector on cache hit.
958 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
960 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
962 * config.gcc (powerpc-xilinx-eabi*): Remove.
963 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
965 (fusion_addis_mem_combo_load): Ditto.
966 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
968 (rs6000_cpu_cpp_builtins): Ditto.
969 * config/rs6000/rs6000-linux.c
970 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
971 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
972 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
974 (rs6000_setup_reg_addr_masks): Ditto.
975 (rs6000_init_hard_regno_mode_ok): Ditto.
976 (rs6000_option_override_internal): Ditto.
977 (legitimate_lo_sum_address_p): Ditto.
978 (rs6000_legitimize_address): Ditto.
979 (rs6000_legitimize_reload_address): Ditto.
980 (rs6000_legitimate_address_p): Ditto.
981 (abi_v4_pass_in_fpr): Ditto.
982 (setup_incoming_varargs): Ditto.
983 (rs6000_gimplify_va_arg): Ditto.
984 (rs6000_split_multireg_move): Ditto.
985 (rs6000_savres_strategy): Ditto.
986 (rs6000_emit_prologue_components): Ditto.
987 (rs6000_emit_epilogue_components): Ditto.
988 (rs6000_emit_prologue): Ditto.
989 (rs6000_emit_epilogue): Ditto.
990 (rs6000_elf_file_end): Ditto.
991 (rs6000_function_value): Ditto.
992 (rs6000_libcall_value): Ditto.
993 * config/rs6000/rs6000.h: Ditto.
994 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
995 (TARGET_MINMAX): ... this. New.
996 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
997 * config/rs6000/rs6000.md: Remove Xilinx FP support.
998 (*movsi_internal1_single): Delete.
999 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1000 mfpu=, mxilinx-fpu): Delete.
1001 * config/rs6000/singlefp.h: Delete.
1002 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1003 * config/rs6000/t-rs6000: Ditto.
1004 * config/rs6000/t-xilinx: Delete.
1005 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1006 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1007 (VStype_simple): Delete.
1008 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1009 * config/rs6000/xfpu.h: Delete.
1010 * config/rs6000/xfpu.md: Delete.
1011 * config/rs6000/xilinx.h: Delete.
1012 * config/rs6000/xilinx.opt: Delete.
1013 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1014 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1016 2018-05-04 Tom de Vries <tom@codesourcery.com>
1019 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1022 2018-05-04 Richard Biener <rguenther@suse.de>
1025 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1026 (expand_complex_div_wide): Likewise.
1027 (expand_complex_operations_1): Likewise.
1028 (expand_complex_libcall): Preserve EH info of the original stmt.
1029 (tree_lower_complex): Handle removed blocks.
1030 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1031 on complex multiplication and division libcall builtins.
1033 2018-05-04 Richard Biener <rguenther@suse.de>
1036 * fold-const.c (negate_expr_p): Restrict negation of operand
1037 zero of a division to when we know that can happen without
1039 (fold_negate_expr_1): Likewise.
1041 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1044 * real.h (real_nextafter): Declare.
1045 * real.c (real_nextafter): New function.
1046 * fold-const-call.c (fold_const_nextafter): New function.
1047 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1048 CASE_CFN_NEXTTOWARD.
1049 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1050 even when arg1_mode is different from arg0_mode.
1052 2018-05-03 Nathan Sidwell <nathan@acm.org>
1054 * doc/extend.texi (Deprecated Features): Remove
1056 (Backwards Compatibility): Likewise.
1057 * doc/invoke.texi (C++ Language Options): Likewise.
1058 (C++ Dialect Options): Likewise.
1060 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1063 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1064 _mm512_mask_mullox_epi64): New intrinsics.
1066 2018-05-03 Tom de Vries <tom@codesourcery.com>
1069 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1070 dump files): Add offload-tree.
1072 2018-05-03 Richard Biener <rguenther@suse.de>
1074 PR tree-optimization/85615
1075 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1076 to loops not nested in BBs loop father to avoid creating multi-entry
1079 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1081 PR tree-optimization/70291
1082 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1083 arguments. Change return type to tree. Emit libcall as a new
1084 statement rather than replacing existing one when inplace_p is true.
1085 (expand_complex_multiplication_components): New function.
1086 (expand_complex_multiplication): Expand floating-point complex
1087 multiplication using the above.
1088 (expand_complex_division): Rename inner_type parameter to type.
1089 Update expand_complex_libcall call-site.
1090 (expand_complex_operations_1): Update expand_complex_multiplication
1091 and expand_complex_division call-sites.
1093 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1096 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1097 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1098 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1099 the highest significant bit of the shift count mask is clear. In
1100 check whether and[sq]i3 is needed verify that all significant bits
1101 of the shift count other than the highest are set.
1103 2018-05-02 Tom de Vries <tom@codesourcery.com>
1106 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1107 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1108 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1109 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1110 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1111 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1112 __builtin_goacc_parlevel_size.
1114 2018-05-02 Richard Biener <rguenther@suse.de>
1116 PR tree-optimization/85597
1117 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1118 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1121 2018-05-02 Tom de Vries <tom@codesourcery.com>
1124 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1125 dump files): Add ltrans-tree.
1127 2018-05-02 Tom de Vries <tom@codesourcery.com>
1130 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1131 dump files): Add wpa-ipa.
1133 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1135 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1136 powerpc*-*-linux*paired* target.
1137 * config/rs6000/750cl.h: Delete.
1138 * config/rs6000/paired.h: Delete.
1139 * config/rs6000/paired.md: Delete.
1140 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1142 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1143 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1144 comment. Remove paired float support.
1145 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1146 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1148 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1149 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1151 * config/rs6000/rs6000.c: Remove paired float support.
1152 (paired_expand_vector_init, paired_expand_vector_move,
1153 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1154 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1155 paired_expand_builtin, paired_expand_predicate_builtin,
1156 paired_init_builtins): Delete.
1157 * config/rs6000/rs6000.h: Remove paired float support.
1158 * config/rs6000/rs6000.md: Remove paired float support.
1159 (move_from_CR_ov_bit): Delete.
1160 * config/rs6000/rs6000.opt (mpaired): Delete.
1161 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1162 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1164 2018-05-02 Richard Biener <rguenther@suse.de>
1167 * gimplify.c (gimplify_save_expr): When in SSA form allow
1168 SAVE_EXPRs to compute to SSA vars.
1170 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1173 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1174 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1175 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1176 clobber operands[2], instead use a new pseudo. Formatting fixes.
1178 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1180 PR tree-optimization/85586
1181 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1182 exit early for statements in the same group if the accesses are
1185 2018-05-02 Tom de Vries <tom@codesourcery.com>
1188 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1191 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1193 PR tree-optimization/85143
1194 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1196 2018-05-01 Tom de Vries <tom@codesourcery.com>
1199 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1200 not found" error message.
1202 2018-05-01 Tom de Vries <tom@codesourcery.com>
1205 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1206 * vec.c (test_ordered_remove_if): New function.
1207 (vec_c_tests): Call test_ordered_remove_if.
1208 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1209 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1210 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1211 VEC_ORDERED_REMOVE_IF.
1213 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1215 PR tree-optimization/82665
1216 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1217 pointer subtraction where arguments come from a memchr call.
1219 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1221 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1222 --push-state --as-needed and --pop-state instead of --as-needed and
1223 --no-as-needed if ld supports it.
1224 * configure: Regenerated.
1227 * doc/install.texi2html: Replace _002d with - and _002a with * in
1228 generated html files using sed.
1230 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1233 * gcc-rich-location.c (blank_line_before_p): New function.
1234 (use_new_line): New function.
1235 (gcc_rich_location::add_fixit_insert_formatted): New function.
1236 * gcc-rich-location.h
1237 (gcc_rich_location::add_fixit_insert_formatted): New function.
1239 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1241 * selftest.c (assert_streq): Rename "expected" and "actual" to
1242 "val1" and "val2". Extend NULL-handling to cover both inputs
1243 symmetrically, while still requiring both to be non-NULL for a pass.
1244 * selftest.h (assert_streq): Rename "expected" and "actual" to
1246 (ASSERT_EQ): Likewise.
1247 (ASSERT_EQ_AT): Likewise.
1248 (ASSERT_KNOWN_EQ): Likewise.
1249 (ASSERT_KNOWN_EQ_AT): Likewise.
1250 (ASSERT_NE): Likewise.
1251 (ASSERT_MAYBE_NE): Likewise.
1252 (ASSERT_MAYBE_NE_AT): Likewise.
1253 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1254 the assertion to pass.
1255 (ASSERT_STREQ_AT): Likewise.
1257 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1259 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1260 interaction with -pie.
1262 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1264 * selftest.h: Fix alphabetization of per-source-file selftest
1267 2018-04-30 Jason Merrill <jason@redhat.com>
1269 PR c++/61982 - dead stores to destroyed objects.
1270 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1273 2018-04-30 Jason Merrill <jason@redhat.com>
1275 * tree.c (build_clobber): New.
1276 * tree.h: Declare it.
1277 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1279 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1281 * diagnostic-show-locus.c (layout::layout): Update for
1282 location_get_source_line returning a char_span.
1283 (struct char_span): Move to input.h.
1284 (struct correction): Update for fields in char_span becoming
1286 (struct source_line): Update for location_get_source_line
1287 returning a char_span.
1288 (layout::print_line): Likewise.
1289 * edit-context.c (edited_file::print_content): Likewise.
1290 (edited_file::print_diff_hunk): Likewise.
1291 (edited_file::print_run_of_changed_lines): Likewise.
1292 (edited_file::get_num_lines): Likewise.
1293 (edited_line::edited_line): Likewise.
1294 * final.c (asm_show_source): Likewise.
1295 * input.c (location_get_source_line): Convert return type
1296 from const char * to char_span, losing the final "line_len"
1298 (dump_location_info): Update for the above.
1299 (get_substring_ranges_for_loc): Likewise. Use a char_span
1300 when handling the literal within the line.
1301 (test_reading_source_line): Update for location_get_source_line
1302 returning a char_span.
1303 * input.h (class char_span): Move here from
1304 diagnostic-show-locus.c, converting from a struct to a class.
1305 Make data members private.
1306 (char_span::operator bool): New.
1307 (char_span::length): New.
1308 (char_span::get_buffer): New.
1309 (char_span::operator[]): New.
1310 (char_span::subspan): Make const.
1311 (char_span::xstrdup): New.
1312 (location_get_source_line): Convert return type from const char *
1313 to char_span, losing the final "line_size" param.
1315 2018-04-30 Jan Hubicka <jh@suse.cz>
1317 * lto-wrapper.c (ltrans_priorities): New static var.
1318 (cmp_priority): New.
1319 (run_gcc): Read priorities and if doing parallel build order
1320 the Makefile by them.
1322 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1324 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1326 2018-04-30 Richard Biener <rguenther@suse.de>
1328 * tree-cfg.c (verify_address): Remove base argument, add
1329 flag whether to check TREE_ADDRESSABLE and do that.
1330 (verify_expr): Remove.
1331 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1332 (verify_gimple_assign_single): Likewise.
1333 (verify_gimple_switch): Likewise.
1334 (verify_expr_location_1): Dereference tp once. Add (disabled)
1335 piece from verify_expr.
1336 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1338 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1340 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1342 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1344 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1345 (small_data_pattern): Likewise.
1346 (arc_rewrite_small_data): Likewise.
1347 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1348 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1349 (get_symbol_alignment): New function.
1350 (legitimate_small_data_address_p): Likewise.
1351 (legitimate_scaled_address): Update, call
1352 legitimate_small_data_address_p.
1353 (output_sdata): New static variable.
1354 (arc_print_operand): Update how we handle small data operands.
1355 (arc_print_operand_address): Likewise.
1356 (arc_legitimate_address_p): Update, use
1357 legitimate_small_data_address_p.
1358 (arc_rewrite_small_data_p): Remove.
1359 (arc_rewrite_small_data_1): Likewise.
1360 (arc_rewrite_small_data): Likewise.
1361 (small_data_pattern): Likewise.
1362 (compact_sda_memory_operand): Update to use
1363 legitimate_small_data_address_p and get_symbol_alignment.
1364 (prepare_move_operands): Don't rewite sdata pattern.
1365 (prepare_extend_operands): Remove.
1366 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1368 (zero_extendqisi2): Likewise.
1369 (zero_extendhisi2): Likewise.
1370 (extendqihi2): Likewise.
1371 (extendqisi2): Likewise.
1372 (extendhisi2): Likewise.
1376 * config/arc/constraints.md (Usd): Change it to memory constraint.
1378 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1380 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1381 as source of std instructions.
1382 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1383 allow 6-bit constants as source for store instructions.
1384 (movdi_insn): Update instruction pattern to allow 6-bit constants
1385 as source for store instructions.
1387 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1389 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1391 2018-04-30 Nathan Sidwell <nathan@acm.org>
1392 Sandra Loosemore <sandra@codesourcery.com>
1394 * dumpfile.c (dump_open): Allow '-' for stdout.
1395 * doc/invoke.texi (Developer Options): Document dump filename
1396 determination early. Document stdin/stdout selection.
1398 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1400 Microblaze Target: PIC data text relative
1402 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1403 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1405 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1406 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1407 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1408 New addressing mode for data-text relative position indepenedent code.
1409 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1410 'ADDRESS_SYMBOLIC_TXT_REL'.
1411 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1412 (microblaze_legitimate_pic_operand): Exclude function calls from
1413 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1414 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1416 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1417 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1418 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1419 for 'address + offset'.
1420 (microblaze_expand_prologue): Add new function prologue call for
1422 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1423 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1424 table in case of TARGET_PIC_DATA_TEXT_REL.
1425 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1426 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1427 Add new macros 'UNSPEC_TEXT',
1428 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1429 + exclude function calls from 'UNSPEC_PLT' in case of data text
1431 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1432 new target hook for generating address diff vector tables in case of
1434 * doc/tm.texi : Regenerate.
1435 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1436 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1437 of addr diff vector generation.
1438 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1439 target hook definition.
1440 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1441 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1442 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1443 option for data text relative.
1445 2018-04-30 Richard Biener <rguenther@suse.de>
1447 * tree-chrec.h (evolution_function_is_constant_p): Remove
1449 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1451 2018-04-30 Richard Biener <rguenther@suse.de>
1454 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1456 2018-04-30 Richard Biener <rguenther@suse.de>
1458 PR tree-optimization/28364
1459 PR tree-optimization/85275
1460 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1461 copying first exit test.
1463 2018-04-28 Mark Wielaard <mark@klomp.org>
1465 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1467 (dwarf_AT): Handle DW_AT_addr_base.
1468 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1470 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1473 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1474 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1475 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1476 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1478 2018-04-28 Richard Biener <rguenther@suse.de>
1480 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1481 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1482 to reflect use. Only add interesting stmts.
1484 2018-04-27 Martin Jambor <mjambor@suse.cz>
1487 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1488 the jump function allows for passing through aggregate values.
1490 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1492 * input.h (in_system_header_at): Convert from macro to inline
1494 (from_macro_expansion_at): Likewise.
1495 (from_macro_definition_at): Likewise.
1497 2018-04-27 Jeff Law <law@redhat.com>
1499 * config.gcc: Mark tile* targets as deprecated/obsolete.
1501 2018-04-27 Richard Biener <rguenther@suse.de>
1503 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1506 2018-04-27 Richard Biener <rguenther@suse.de>
1508 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1510 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1512 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1513 with Yd constraint. Set "preferred_for_speed" attribute from
1514 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1516 (*movdi_internal): Ditto.
1517 (movti_interunit splitters): Remove
1518 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1519 (movdi_interunit splitters): Ditto.
1520 * config/i386/constraints.md (Ye): Remove.
1521 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1523 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1526 * config/aarch64/constraints.md (Usg): Limit to 31.
1529 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1531 PR tree-optimization/85529
1532 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1533 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1534 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1535 zero extension or masking of the MSB bit.
1536 (optimize_range_tests): Add FIRST_BB argument, pass it through
1537 to optimize_range_tests_var_bound.
1538 (maybe_optimize_range_tests, reassociate_bb): Adjust
1539 optimize_range_tests callers.
1541 2018-04-26 Richard Biener <rguenther@suse.de>
1542 Jakub Jelinek <jakub@redhat.com>
1544 * cgraph.h (symbol_table): Just declare debug method here.
1545 * symtab.c (symbol_table::debug): Define.
1547 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1549 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1551 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1553 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1554 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1555 (*movdi_internal): Substitute Yi and Yj constraint with x
1556 and Ym and Yn constraint with y constraint. Update "isa"
1557 attribute and set "preferred_for_speed" attribute from
1558 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1559 (*movsi_internal): Ditto.
1560 (*movdf_internal): Ditto.
1561 (*movsf_internal): Ditto.
1562 (*zero_extendsidi2): Ditto.
1563 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1564 (sse2_loadld): Ditto.
1565 (*vec_extract<ssevecmodelower>_0): Ditto.
1566 (*vec_extractv4si_0_zext_sse4): Ditto.
1567 (vec_concatv2di): Ditto.
1568 (*vec_dup<mode>): Ditto.
1569 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1570 * config/i386/constraints.md (Yi): Remove.
1575 2018-04-26 Nathan Sidwell <nathan@acm.org>
1577 * dumpfile.c (dump_open): New.
1578 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1579 (dump_finish): Detect stdio/stderr by value not name.
1581 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1583 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1585 2018-04-26 Tom de Vries <tom@codesourcery.com>
1588 * config/nvptx/nvptx.c (verify_neutering_jumps)
1589 (verify_neutering_labels): New function
1590 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1592 2018-04-26 Tom de Vries <tom@codesourcery.com>
1595 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1596 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1599 2018-04-26 Richard Biener <rguenther@suse.de>
1600 Tom de Vries <tom@codesourcery.com>
1603 * lto-streamer-out.c (output_function): Fixup loops if required to match
1604 discovery done in the reader.
1606 2018-04-26 Richard Biener <rguenther@suse.de>
1608 PR tree-optimization/85116
1609 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1610 have a loop exit from the single latch predecessor. Remove
1611 case of header with just condition.
1612 (ch_base::copy_headers): Exclude infinite loops from any
1614 (pass_ch::execute): Record exits.
1616 2018-04-26 Richard Biener <rguenther@suse.de>
1618 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1619 prologue cost vector and pass it to vect_get_load_cost.
1620 (vect_get_peeling_costs_all_drs): Likewise.
1621 (vect_peeling_hash_get_lowest_cost): Likewise.
1622 (vect_enhance_data_refs_alignment): Likewise.
1624 2018-04-26 Richard Biener <rguenther@suse.de>
1627 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1628 checking of integer<->pointer conversions.
1629 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1630 sign-/zero-extending pointer types.
1631 (expand_omp_for_static_chunk): Likewise.
1633 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1634 Jean Lee <xiaoyur347@gmail.com>
1636 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1637 (TARGET_ASAN_SHADOW_OFFSET): Define.
1638 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1639 true for -fsanitize=address.
1641 2018-04-25 Mark Wielaard <mark@klomp.org>
1643 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1646 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1648 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1649 than "alu", remove explicit "memory" and "imm_disp" attributes.
1650 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1653 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1654 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1655 gen_lowpart_no_emit.
1657 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1660 * config/i386/i386.c (ix86_expand_builtin): Change memory
1661 operand to XI, extend p0 to Pmode.
1662 * config/i386/i386.md: Change unspec volatile and operand
1663 1 mode to XI, change operand 0 mode to P.
1665 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1667 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1668 GET_MODE_MASK before any checking.
1669 (nds32_can_use_bset_p): Likewise.
1670 (nds32_can_use_btgl_p): Likewise.
1672 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1674 * config/nds32/nds32-doubleword.md: New define_split pattern for
1675 illegal register number.
1677 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1679 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1681 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1683 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1685 2018-04-25 Richard Biener <rguenther@suse.de>
1687 * lto-streamer.h (LTO_major_version): Bump to 8.
1689 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1691 * BASE-VER: Set to 9.0.0.
1693 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1695 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1696 in __abskf2 and __powikf2.
1698 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1701 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1702 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1703 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1704 Use the above on operand 2. Reindent.
1705 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1707 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1710 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1712 * config/i386/i386.opt (mcet): Removed.
1713 * doc/install.texi: Remove -mcet documentation.
1714 * doc/invoke.texi: Likewise.
1716 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1719 * doc/install.texi: Remove -mcet from bootstrap-cet.
1721 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1724 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1725 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1729 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1730 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1731 containing a CONST_VECTOR.
1733 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1735 * doc/install.texi: Update newlib dependency for nvptx.
1737 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1740 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1741 instead of INTVAL when shifting x left.
1743 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1745 PR tree-optimization/85478
1746 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1747 vect_grouped_store_supported for single element vectors.
1749 2018-04-24 Richard Biener <rguenther@suse.de>
1752 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1753 load cost increase to the case of non-constant step.
1755 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1758 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1759 destination if any_malformed_asm.
1761 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1764 * expr.c (store_field): In the bitfield case, if the value comes from
1765 a function call and is returned in registers by means of a PARALLEL,
1766 do not change the mode of the temporary unless BLKmode and VOIDmode.
1768 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1770 PR rtl-optimization/85423
1771 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1772 dependencies to debug insns when the previous insn is non-debug.
1774 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1776 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1777 enums into a single definition.
1778 (fls): Fix predicates and printing.
1781 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1783 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1784 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1785 and short u6 immediate.
1786 (check_if_valid_sleep_operand): Remove.
1787 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1789 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1791 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1792 flag_always_save_lp condition.
1793 * config/nds32/nds32.opt (malways-save-lp): New option.
1795 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1797 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1798 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1799 * config/nds32/nds32.h
1800 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1801 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1803 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1805 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1806 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1808 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1809 Chung-Ju Wu <jasonwucj@gmail.com>
1811 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1813 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1814 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1816 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1818 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1820 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1822 * config/nds32/nds32-protos.h (nds32_data_alignment,
1823 nds32_local_alignment): Declare.
1824 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1825 nds32_local_alignment): New functions.
1826 (TARGET_CONSTANT_ALIGNMENT): Define.
1827 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1829 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1831 * config/nds32/nds32.c
1832 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1833 (TARGET_MODES_TIEABLE_P): Likewise.
1835 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1837 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1840 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1841 Chung-Ju Wu <jasonwucj@gmail.com>
1843 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1844 for unaligned access.
1845 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1847 * config/nds32/nds32-intrinsic.md: Likewise.
1848 * config/nds32/nds32_intrinsic.h: Likewise.
1849 * config/nds32/nds32.h (nds32_builtins): Likewise.
1850 * config/nds32/nds32.opt (munaligned-access): New option.
1851 * config/nds32/nds32.c (nds32_asm_file_start): Display
1852 flag_unaligned_access status.
1854 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1856 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1857 -mno-relax is present.
1858 * config/riscv/linux.h (LINK_SPEC): Ditto.
1860 2018-04-20 Martin Sebor <msebor@redhat.com>
1863 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1865 (gimple_fold_builtin_stxcpy_chk): Same.
1866 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1868 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1871 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1872 __powikf2 when long double is IEEE 128-bit.
1874 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1876 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1877 step to make sure stack always aligned.
1879 2018-04-20 Carl Love <cel@us.ibm.com>
1882 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1883 size check for arg0.
1885 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1886 Tom de Vries <tom@codesourcery.com>
1889 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1890 Emit insns for calls too.
1891 (nvptx_find_par): Always look for worker-level predecessor insn.
1892 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1894 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1895 (nvptx_process_pars): Propagate frames for calls.
1897 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1900 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1902 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1903 (ix86_handle_option): Don't handle OPT_mibt.
1904 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1906 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1908 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1909 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1910 (ix86_target_macros): Define __CET__ with flag_cf_protection
1911 for -fcf-protection.
1912 * config/i386/i386.c (isa2_opts): Remove -mibt.
1913 * config/i386/i386.h (TARGET_IBT): Removed.
1914 (TARGET_IBT_P): Likewise.
1915 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1916 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1917 * config/i386/i386.opt (mcet): Update help message.
1920 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1921 -mcet as an alias for -mshstk.
1923 2018-04-20 Richard Biener <rguenther@suse.de>
1926 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1927 complexity by forcing a single use of the multiply operand.
1929 2018-04-20 Martin Jambor <mjambor@suse.cz>
1932 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1933 recursion dependency to only apply to non-clones.
1935 2018-04-20 Martin Jambor <mjambor@suse.cz>
1938 * ipa-cp.c (create_specialized_node): Check that clones of
1939 self-recursive edges exist during IPA-CP.
1941 2018-04-19 Toon Moene <toon@moene.org>
1943 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1946 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1948 PR tree-optimization/85467
1949 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1950 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1951 VECTOR_CST element to type.
1953 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1956 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1957 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1958 (builtin_longjmp): Likewise.
1959 (save_stack_nonlocal): New pattern.
1960 (restore_stack_nonlocal): Likewise.
1962 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1965 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1966 Replace ASM_OUTPUT_LABEL with fprintf.
1968 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1971 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1972 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1973 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1974 define __IBT__ and __SHSTK__ for -fcf-protection.
1975 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1977 (ix86_trampoline_init): Likewise.
1978 (x86_output_mi_thunk): Likewise.
1979 (ix86_notrack_prefixed_insn_p): Likewise.
1980 (ix86_option_override_internal): Don't disallow -fcf-protection.
1981 * config/i386/i386.md (rdssp<mode>): Also enable for
1983 (incssp<mode>): Likewise.
1984 (nop_endbr): Likewise.
1985 * config/i386/i386.opt (mcet): Change help message to built-in
1989 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1990 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1991 enable CET built-in functions.
1993 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1995 * common/config/i386/i386-common.c
1996 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1997 OPTION_MASK_ISA_MOVDIRI_UNSET,
1998 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1999 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2000 * config.gcc (movdirintrin.h): New header.
2001 * config/i386/cpuid.h (bit_MOVDIRI,
2002 bit_MOVDIR64B): New bits.
2003 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2005 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2006 (VOID, PVOID, PCVOID)): New function types.
2007 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2008 __builtin_ia32_directstoreu_u64,
2009 __builtin_ia32_movdir64b): New builtins.
2010 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2011 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2013 (ix86_valid_target_attribute_inner_p): Ditto.
2014 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2015 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2016 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2017 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2018 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2019 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2020 (movdiri<mode>, movdir64b_<mode>): New.
2021 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2022 * config/i386/immintrin.h: Include movdirintrin.h.
2023 * config/i386/movdirintrin.h: New file.
2024 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2026 2018-04-19 Richard Biener <rguenther@suse.de>
2029 * cfg.c (clear_bb_flags): When loop state says we have
2030 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2032 2018-04-19 Richard Biener <rguenther@suse.de>
2034 PR tree-optimization/84737
2035 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2036 copying restrict info.
2037 (vect_setup_realignment): Use it.
2038 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2039 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2040 the first DR to all generated stores.
2041 (vectorizable_load): Likewise for loads.
2043 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2045 PR tree-optimization/85446
2046 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2047 the integral and pointer types to have the same precision.
2049 * doc/install.texi: Document --disable-cet being the default and
2052 2018-04-18 Martin Liska <mliska@suse.cz>
2054 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2057 2018-04-18 Martin Liska <mliska@suse.cz>
2060 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2063 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2064 arguments if they are comparable.
2066 2018-04-18 Martin Liska <mliska@suse.cz>
2069 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2072 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2075 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2078 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2079 ENDBR after calling __morestack.
2081 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2084 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2085 by using gcc_base_ver to generate a gcc_driver_version, and use
2086 it when generating GCC_DRIVER_NAME.
2087 * configure: Regenerate.
2089 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2092 * config.gcc: Obsolete powerpc*-*-*spe*.
2094 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2097 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2098 (stabstr_D): Change type of unum from unsigned int to
2099 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2102 2018-04-17 Jim Wilson <jimw@sifive.com>
2105 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2106 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2107 Set arg_pointer_offset after using pretend_args_size.
2109 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2111 PR rtl-optimization/85431
2112 * dse.c (record_store): Ignore zero width stores.
2115 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2116 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2117 __builtin_stack_restore rather than after it.
2118 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2119 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2120 argument instead of virtual_dynamic_stack_rtx.
2122 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2124 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2126 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2127 Add note to error message to explain internal mapping of overloaded
2128 built-in function name to non-overloaded built-in function name.
2129 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2132 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2135 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2136 where the inputs overlap with the output.
2138 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2140 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2141 (=v, v) alternative and explicit "memory" attribute.
2142 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2143 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2145 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2146 "sselog1" type instead of "sselog".
2147 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2148 "sselog". Remove explicit "memory" attribute.
2149 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2150 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2152 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2153 "sselog1" type instead of "sselog". Remove explicit "memory"
2155 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2156 use "sselog1" type instead of "sselog". Remove explicit "memory"
2158 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2159 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2161 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2162 "sselog1" type instead of "sselog". Remove explicit "memory"
2164 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2165 use "sselog1" type instead of "sselog". Remove explicit "memory"
2169 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2172 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2175 2018-04-17 Martin Jambor <mjambor@suse.cz>
2178 * ipa-cp.c (create_specialized_node): Call
2179 expand_all_artificial_thunks if necessary.
2181 2018-04-17 Martin Liska <mliska@suse.cz>
2184 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2185 in message, remote space in between '_G' and '('.
2187 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2190 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2191 avx512f_vmcmp<mode>3<round_saeonly_name>,
2192 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2193 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2194 avx512f_rndscale<mode><round_saeonly_name>,
2195 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2196 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2197 Use %<iptr>2 instead of %2 for -masm=intel.
2198 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2199 avx512f_vcvttss2usi<round_saeonly_name>,
2200 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2202 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2203 avx512f_vcvttsd2usi<round_saeonly_name>,
2204 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2205 Use %q1 instead of %1 for -masm=intel.
2206 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2207 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2208 of %3 for -masm=intel.
2209 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2211 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2213 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2215 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2216 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2218 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2220 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2221 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2222 %g1 and one with %0 and %1.
2223 (avx512er_vmrcp28<mode><round_saeonly_name>,
2224 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2226 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2227 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2228 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2229 of %0 and %{%4%} for -masm=intel.
2230 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2231 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2232 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2233 order of %0 and %{%5%}%{z%} for -masm=intel.
2235 2018-04-17 Jan Hubicka <jh@suse.cz>
2238 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2240 2018-04-17 Martin Liska <mliska@suse.cz>
2243 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2244 for target_clone error message. Make default implementation
2245 clone to be a local declaration.
2246 (separate_attrs): Add new argument and check for an empty
2248 (expand_target_clones): Handle it.
2249 (ipa_target_clone): Make redirection just for target_clones
2252 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2253 Tom de Vries <tom@codesourcery.com>
2256 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2257 tiled basic blocks without omp continue statements.
2259 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2262 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2263 vec_extract expression as having side effects to make sure it gets
2266 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2269 * config/i386/i386.c (get_builtin_code_for_version): Check
2272 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2275 * gcc/config.gcc: Support "skylake".
2276 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2278 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2279 (processor_target_table): Add "skylake".
2280 (ix86_option_override_internal): Add "skylake".
2281 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2282 PROCESSOR_CANNONLAKE.
2283 (get_builtin_code_for_version): Fix priority for
2284 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2285 PROCESSOR_SKYLAKE-AVX512.
2286 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2287 (processor_type): Add PROCESSOR_SKYLAKE.
2289 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2290 Jason Merrill <jason@redhat.com>
2293 * convert.c (convert_to_integer_1): Use direct recursion for
2294 enumeral types and types with a precision less than the number
2295 of bits in their mode.
2297 2018-04-16 Julia Koval <julia.koval@intel.com>
2300 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2301 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2303 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2306 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2307 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2308 and -mno-direct-move.
2310 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2313 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2314 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2315 Ensure negative shifts result in {0}.
2317 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2319 PR rtl-optimization/79916
2320 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2321 regs (if any) to define how to gnerate SD moves when LRA is in
2324 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2326 PR rtl-optimization/85393
2327 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2328 * except.c (expand_dw2_landing_pad_for_region): Make static.
2329 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2330 a label and unconditional jump to old_bb, rather than
2331 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2334 PR rtl-optimization/85376
2335 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2336 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2337 instead of a specific value.
2339 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2340 Bin Cheng <bin.cheng@arm.com>
2342 PR tree-optimization/82965
2343 PR tree-optimization/83991
2344 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2345 by_profile_only parameter.
2346 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2347 information if the loop was predicted to iterate too many times.
2348 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2350 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2353 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2356 2018-04-13 Martin Liska <mliska@suse.cz>
2357 Jakub Jelinek <jakub@redhat.com>
2360 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2361 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2362 * builtins.c (expand_builtin_memory_copy_args): Use
2363 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2364 handle dest_addr == pc_rtx.
2366 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2369 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2370 asked to not generate direct moves.
2371 (fix_trunc<mode>si2_stfiwx): Similar.
2372 (fix_trunc<mode>si2_internal): Similar.
2374 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2377 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2378 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2379 lookup if dest in some wider mode is known to be const0_rtx and
2380 if so, record permanent equivalence for it to be ZERO_EXTEND of
2381 the narrower mode destination.
2383 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2385 * lto-streamer-out.c (output_function): Revert 259346.
2386 * omp-expand.c (expand_oacc_for): Likewise.
2388 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2390 PR rtl-optimization/85354
2391 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2392 * sel-sched.c (sel_global_init): ... here.
2394 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2397 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2398 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2399 mode for PE-COFF targets.
2400 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2401 (i386_pe_asm_lto_end): Likewise.
2402 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2403 (TARGET_ASM_LTO_END): Likewise.
2404 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2405 (i386_pe_asm_lto_start): New function.
2406 (i386_pe_asm_lto_end): Likewise.
2408 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2409 Richard Biener <rguenther@suse.de>
2412 * lto-streamer-out.c (output_function): Fix CFG loop state before
2414 * omp-expand.c (expand_oacc_for): Handle calls to internal
2415 functions like regular functions.
2417 2018-04-12 Richard Biener <rguenther@suse.de>
2420 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2421 for the early LTO debug to properly generate references to it
2422 during DIE emission. Do not re-use that for the skeleton for
2424 (dwarf2out_early_finish): Likewise.
2426 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2429 * config/i386/sse.md
2430 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2431 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2432 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2433 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2434 and output is a reg, avoid creating invalid lowpart subreg, but
2435 instead split into a 512-bit move. Don't split if not AVX512VL,
2436 input is xmm16+ reg and output is a mem.
2437 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2438 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2439 xmm16+ reg and output is a mem.
2441 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2443 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2444 also for flag_dwarf2_cfi_asm.
2446 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2448 PR rtl-optimization/85342
2449 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2450 a bool scalar var inside of the loop instead. Don't try to update
2451 recog_data.operand after failed apply_change_group.
2453 2018-04-12 Tom de Vries <tom@codesourcery.com>
2456 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2457 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2458 array with flexible array member as array without given dimension.
2459 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2460 argument for undefined param to true.
2462 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2465 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2466 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2467 from PowerPC section.
2468 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2469 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2470 help text that is too long.
2471 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2472 help text that is too long.
2473 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2474 help text that is too long.
2476 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2478 * config/alpha/alpha.md (stack_probe_internal): Rename
2479 from "probe_stack". Update all callers.
2481 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2483 PR rtl-optimization/84566
2484 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2485 sched_macro_fuse_insns.
2487 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2490 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2491 (compute_block_dependences): ... from here.
2493 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2495 PR tree-optimization/85331
2496 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2497 from int to HOST_WIDE_INT.
2499 2018-04-11 Martin Jambor <mjambor@suse.cz>
2502 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2503 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2504 not the same as the source val.
2505 (cgraph_edge_brings_value_p): New parameter.
2506 (gather_edges_for_value): Pass destination value to
2507 cgraph_edge_brings_value_p.
2508 (perhaps_add_new_callers): Likewise.
2509 (get_info_about_necessary_edges): Likewise and exclude values brought
2510 only by self-recursive edges.
2511 (create_specialized_node): Redirect only clones of self-calling edges.
2512 (+self_recursive_pass_through_p): New function.
2513 (find_more_scalar_values_for_callers_subset): Use it.
2514 (find_aggregate_values_for_callers_subset): Likewise.
2515 (known_aggs_to_agg_replacement_list): Removed.
2516 (decide_whether_version_node): Re-calculate known constants for all
2517 remaining context clones.
2519 2018-04-11 Richard Biener <rguenther@suse.de>
2522 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2523 from early DWARF output.
2524 (dwarf2out_early_finish): Output line info unconditionally into
2525 early DWARF and add reference to it.
2527 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2530 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2531 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2532 other than V2DFmode using iptr mode attribute.
2533 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2535 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2537 PR rtl-optimization/84659
2538 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2540 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2543 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2545 (output_loc_list): Pass address of a dummy size variable even in the
2546 locview handling loop.
2547 (index_location_lists): Add comment on why skip_loc_list_entry can't
2550 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2553 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2556 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2559 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2560 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2561 and -mstring-compare-inline-limit.
2563 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2566 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2567 for stack clash protection in a register whenever we need it to be in
2570 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2572 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2573 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2575 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2578 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2580 (mlong-double-): Ditto.
2581 * config/rs6000/sysv4.opt (msdata=): Ditto.
2582 (mtls-size=): Ditto.
2584 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2586 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2587 erroneous entries for
2588 "vector int vec_ldl (int, long int *)", and
2589 "vector unsigned int vec_ldl (int, unsigned long int *)".
2590 Add comments and entries for
2591 "vector bool char vec_ldl (int, bool char *)",
2592 "vector bool short vec_ldl (int, bool short *)",
2593 "vector bool int vec_ldl (int, bool int *)",
2594 "vector bool long long vec_ldl (int, bool long long *)",
2595 "vector pixel vec_ldl (int, pixel *)",
2596 "vector long long vec_ldl (int, long long *)",
2597 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2598 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2599 type tree bool_long_long_type_node and correct definition of
2600 bool_V2DI_type_node to make reference to this new type tree.
2601 (rs6000_mangle_type): Replace erroneous reference to
2602 bool_long_type_node with bool_long_long_type_node.
2603 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2604 comments to emphasize sign distinctions for char and int types and
2605 replace RS6000_BTI_bool_long constant with
2606 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2607 use of RS6000_BTI_pixel.
2608 (bool_long_type_node): Remove this macro definition.
2609 (bool_long_long_type_node): New macro definition
2611 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2613 PR rtl-optimization/85300
2614 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2615 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2616 simplify_unary_operation fails.
2618 2018-04-10 Martin Liska <mliska@suse.cz>
2620 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2621 cgraph_edge and ipa_ref.
2623 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2627 * config/i386/sse.md
2628 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2629 computation of the VEC_MERGE selector from mask.
2630 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2631 Fix decoding of the VEC_MERGE selector into mask.
2633 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2635 PR tree-optimization/85286
2636 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2638 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2640 * final.c (final_1): Set insn_last_address as well as
2641 insn_current_address.
2643 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2646 * explow.c (emit_stack_probe): Call validize_mem on memory location
2647 before passing it to gen_probe_stack. Create address operand and
2648 legitimize it for the probe_stack_address case.
2650 2018-04-09 Jan Hubicka <jh@suse.cz>
2653 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2654 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2655 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2657 (free_lang_data): Rebuild type inheritance graph.
2659 2018-04-09 Martin Sebor <msebor@redhat.com>
2661 * invoke.texi (-finline-small-functions): Mention other optimization
2663 (-findirect-inlining, -fpartial-inlining): Same.
2664 (-finline-functions-called-once): Same.
2665 (-freorder-blocks-and-partition): Same.
2667 2018-04-09 Jan Hubicka <jh@suse.cz>
2670 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2671 jumps; choose last target that matches the criteria (i.e.
2672 no partition changes for non-crossing jumps).
2673 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2674 support for redirecting crossing jumps to non-crossing.
2676 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2678 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2679 also for naked functions.
2681 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2683 * config/arc/arc.md (add_shift): New pattern.
2684 (add_shift2): Likewise.
2685 (sub_shift): Likewise.
2686 (sub_shift_cmp0_noout): Likewise.
2687 (compare_si_ashiftsi): Likewise.
2688 (xbfu_cmp0_noout): New combine pattern.
2689 (xbfu_cmp0"): Likewise.
2690 (movsi_set_cc_insn): Place the predicable variant first.
2691 (commutative_binary_cmp0_noout): Remove clobber.
2692 (commutative_binary_cmp0): New pattern.
2693 (noncommutative_binary_cmp0): Likewise.
2694 (noncommutative_binary_cmp0_noout): Likewise.
2695 (noncommutative_binary_comparison_result_used): Removed.
2696 (rsub_cmp0): New pattern.
2697 (rsub_cmp0_noout): Likewise.
2698 (extzvsi): Changed, keep only meaningful variants.
2699 (SQH, SEZ): New iterators.
2700 (SQH_postfix): New mode attribute.
2701 (SEZ_prefix): New code attribute.
2702 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2703 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2704 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2706 (noncommutative_operator): Check the availability of barrel
2709 2018-04-09 Richard Biener <rguenther@suse.de>
2711 PR tree-optimization/85284
2712 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2713 Only use the niter constraining form of simple_iv when the exit
2716 2018-04-09 Tom de Vries <tom@codesourcery.com>
2719 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2720 (define_expand "*memory_barrier"): New define_expand.
2721 (define_insn "memory_barrier"): New insn.
2723 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2725 PR rtl-optimization/80463
2726 PR rtl-optimization/83972
2727 PR rtl-optimization/83480
2729 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2730 correct producer for the insn.
2731 (tidy_control_flow): Fixup seqnos in case of debug insns.
2733 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2735 PR rtl-optimization/83913
2737 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2738 different sched-times when merging exprs.
2740 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2742 PR rtl-optimization/83962
2744 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2745 tidy_fallthru_edge and tidy_control_flow.
2747 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2749 PR rtl-optimization/83530
2751 * sel-sched.c (force_next_insn): New global variable.
2752 (remove_insn_for_debug): When force_next_insn is true, also leave only
2753 next insn in the ready list.
2754 (sel_sched_region): When the region wasn't scheduled, make another pass
2755 over it with force_next_insn set to 1.
2757 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2759 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2761 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2762 for interrupt control.
2763 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2764 functions for interrupt control.
2765 * config/nds32/nds32-intrinsic.md: Likewise.
2766 * config/nds32/nds32_intrinsic.h: Likewise.
2767 * config/nds32/nds32.h (nds32_builtins): Likewise.
2769 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2771 * config/nds32/nds32.c (nds32_init_machine_status,
2772 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2773 strict_aligned_p field.
2774 (nds32_expand_to_rtl_hook): New function.
2775 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2776 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2778 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2779 Chung-Ju Wu <jasonwucj@gmail.com>
2781 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2782 * config/nds32/nds32-n7.md: New file.
2783 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2784 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2786 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2787 * config/nds32/nds32.md (pipeline_model): Add n7.
2788 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2789 * config/nds32/pipelines.md: Include n7 settings.
2791 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2792 Chung-Ju Wu <jasonwucj@gmail.com>
2794 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2795 * config/nds32/nds32-e8.md: New file.
2796 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2797 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2799 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2800 * config/nds32/nds32.md (pipeline_model): Add e8.
2801 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2802 * config/nds32/pipelines.md: Include e8 settings.
2804 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2805 Chung-Ju Wu <jasonwucj@gmail.com>
2807 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2808 * config/nds32/nds32-n8.md: New file.
2809 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2810 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2812 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2813 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2814 * config/nds32/nds32.md (pipeline_model): Add n8.
2815 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2816 * config/nds32/pipelines.md: Include n8 settings.
2818 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2819 Chung-Ju Wu <jasonwucj@gmail.com>
2821 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2822 * config/nds32/nds32-n9-2r1w.md: New file.
2823 * config/nds32/nds32-n9-3r2w.md: New file.
2824 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2825 nds32_register_ports): New or modify for cpu n9.
2826 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2828 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2829 * config/nds32/nds32-utils.c: New file.
2830 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2831 TARGET_MUL_SLOW): Define.
2832 * config/nds32/nds32.md (pipeline_model): New attribute.
2833 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2834 New options that support cpu n9.
2835 * config/nds32/pipelines.md: Include n9 settings.
2836 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2838 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2840 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2841 information if necessary.
2842 (output_cond_branch_compare_zero): Likewise.
2843 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2844 (nds32_target_alignment): Refine for alignment.
2845 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2846 (FUNCTION_BOUNDARY): Modify.
2847 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2849 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2851 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2853 * config/nds32/constants.md (unspec_volatile_element): Add values for
2854 TLB operation and data prefetch.
2855 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2856 functions for TLB operation and data prefetch.
2857 * config/nds32/nds32-intrinsic.md: Likewise.
2858 * config/nds32/nds32_intrinsic.h: Likewise.
2859 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2860 (nds32_print_operand): Likewise.
2861 * config/nds32/nds32.h (nds32_builtins): Likewise.
2863 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2864 Andrew Pinski <pinsika@gcc.gnu.org>
2867 * match.pd: Use constant_boolean_node of correct type instead of
2868 boolean_true_node or boolean_false_node for simplifying
2869 pointer comparisons to zero.
2871 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2873 PR tree-optimization/80021
2874 * tree.c (verify_type_variant): Make error call in verify_variant_match
2875 translatable and remove final full stop.
2877 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2879 * config/nds32/constants.md (unspec_volatile_element): Add
2880 UNSPEC_VOLATILE_EH_RETURN.
2881 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2882 nds32_output_stack_pop): Support dwarf exception handling process.
2883 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2884 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2885 exception handling process.
2886 (nds32_compute_stack_frame): Likewise.
2887 (nds32_return_addr_rtx): Likewise.
2888 (nds32_initial_elimination_offset): Likewise.
2889 (nds32_expand_prologue): Likewise.
2890 (nds32_expand_epilogue): Likewise.
2891 (nds32_dynamic_chain_address): New function.
2892 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2894 (DYNAMIC_CHAIN_ADDRESS): Define.
2895 (EH_RETURN_DATA_REGNO): Define.
2896 (EH_RETURN_STACKADJ_RTX): Define.
2897 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2898 patterns for dwarf exception handling.
2900 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2902 * config/nds32/nds32.h: Clean up obsolete macros.
2904 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2906 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2907 Add enum values for particular instructions.
2908 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2909 particular intrinsic functions.
2910 * config/nds32/nds32-intrinsic.md: Likewise.
2911 * config/nds32/nds32_intrinsic.h: Likewise.
2912 * config/nds32/nds32.h (nds32_builtins): Likewise.
2913 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2914 (btst, ave): New patterns for particular instructions.
2916 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2918 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2919 Add enum values for atomic load/store and memory sync.
2920 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2922 * config/nds32/nds32-intrinsic.md: Likewise.
2923 * config/nds32/nds32_intrinsic.h: Likewise.
2924 * config/nds32/nds32.h (nds32_builtins): Likewise.
2926 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2928 PR tree-optimization/85257
2929 * fold-const.c (native_encode_vector): If not all elts could fit
2930 and off is -1, return 0 rather than offset.
2931 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2932 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2933 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2934 adjust buffer in native_interpret_expr call.
2936 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2938 * config/nds32/constants.md (unspec_volatile_element): Add cache
2939 control enum values.
2940 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2941 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2942 * config/nds32/nds32.c (nds32_cctl_names): New.
2943 (nds32_print_operand): Handle cache control register names.
2944 * config/nds32/nds32.h (nds32_builtins): New enum values.
2945 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2947 * config/nds32/nds32.md (type): Add mmu.
2948 * config/nds32/pipelines.md (simple_insn): Add mmu.
2950 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2952 * config/nds32/nds32.md (type): Remove call.
2953 * config/nds32/pipelines.md (simple_insn): Likewise.
2955 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2957 * config/nds32/constants.md (unspec_volatile_element): Add
2958 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2959 UNSPEC_VOLATILE_FMFCFG.
2960 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2961 description for fmfcfg and fmfcsr.
2962 (bdesc_1arg): Add fmtcsr.
2963 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2964 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2965 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2966 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2967 unspec_fmfcfg): New patterns.
2968 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2969 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2970 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2971 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2972 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2973 __nds32__fmfcfg): Define.
2975 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2977 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2978 intrinsic register names.
2979 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2980 intrinsic register enum values and macros.
2982 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2984 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2985 for load/store addressing form.
2986 (nds32_print_operand_address): Likewise.
2988 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2991 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2992 based on LABEL_REF. Remove useless assertion.
2993 (pic_address_needs_scratch): Fix formatting.
2994 (sparc_legitimize_pic_address): Minor tweaks.
2995 (sparc_delegitimize_address): Adjust assertion accordingly.
2996 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2997 into symbolic_operand.
2998 (movsi_high_pic_label_ref): Likewise.
2999 (movsi_lo_sum_pic_label_ref): Likewise.
3000 (movdi_pic_label_ref): Likewise.
3001 (movdi_high_pic_label_ref): Likewise.
3002 (movdi_lo_sum_pic_label_ref): Likewise.
3004 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3006 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3007 custom LIB_SPEC setup.
3009 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3010 Kito Cheng <kito.cheng@gmail.com>
3012 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3013 * config/riscv/freebsd.h: New.
3015 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3017 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3018 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3021 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3022 Kito Cheng <kito.cheng@gmail.com>
3024 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3025 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3026 * config/nds32/nds32-protos.h (nds32_output_call,
3027 nds32_output_return): Declare.
3028 * config/nds32/nds32.md: Refine all the call and return patterns.
3030 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3033 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3034 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3036 PR rtl-optimization/84872
3037 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3038 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3041 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3043 * expr.c (copy_blkmode_to_reg): Revert 254862.
3044 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3046 2018-04-06 Richard Biener <rguenther@suse.de>
3049 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3050 after seeing a component reference with an adjacent field. Treat
3051 refs to arrays at struct end of external decls similar to
3052 refs to unconstrained commons.
3054 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3057 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3058 look through SAVE_EXPRs with non-side-effects argument. Adjust
3060 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3063 2018-04-06 Richard Biener <rguenther@suse.de>
3066 * alias.c (find_base_term): New wrapper around find_base_term
3067 unwinding CSELIB_VAL_PTR changes.
3068 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3071 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3073 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3075 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3076 constant definitions.
3077 ("nop"): lr 0,0 -> nopr r0
3078 ("nop_lr0", "nop_lr1"): New insn definitions.
3080 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3082 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3083 NDS32_V3PUSH_AVAILABLE_P macro.
3085 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3086 Chung-Ju Wu <jasonwucj@gmail.com>
3088 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3089 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3090 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3091 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3092 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3093 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3094 * config/nds32/constraints.md: New constraints and checking for hard
3095 float configuration.
3096 * config/nds32/iterators.md: New mode iterator and attribute for hard
3097 float configuration.
3098 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3100 * config/nds32/nds32-fpu.md: New file.
3101 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3102 deal with hard float code generation.
3103 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3105 (abi_type, float_reg_number): New enum type.
3106 * config/nds32/nds32-predicates.c: New predicates for hard float.
3107 * config/nds32/nds32-protos.h: Declare functions for hard float.
3108 * config/nds32/nds32.c: Implementation for hard float configuration.
3109 * config/nds32/nds32.h: Definitions for hard float configuration.
3110 * config/nds32/nds32.md: Include hard float machine description and
3111 modify patterns for hard float configuration.
3112 * config/nds32/nds32.opt: New options for hard float configuration.
3113 * config/nds32/predicates.md: New predicates for hard float
3116 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3118 * common/config/nds32/nds32-common.c
3119 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3121 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3124 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3125 CONSTRUCTOR_ELT (ctor, ...)->value.
3127 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3130 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3132 2018-04-05 Tom de Vries <tom@codesourcery.com>
3135 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3138 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3139 Kito Cheng <kito.cheng@gmail.com>
3141 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3142 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3143 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3144 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3146 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3147 Kito Cheng <kito.cheng@gmail.com>
3149 * config/nds32/constraints.md (Ufe): New memory constraint.
3150 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3151 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3152 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3154 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3155 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3157 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3159 * config/nds32/nds32.md: Use optimize_size in the condition for
3160 alu-shift instructions.
3162 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3164 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3166 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3168 * config/nds32/nds32.md (negsi2): Refine pattern.
3170 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3171 Chung-Ju Wu <jasonwucj@gmail.com>
3173 * config/nds32/iterators.md (shift_rotate): New code iterator.
3174 (shift): New code attribute.
3175 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3176 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3177 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3178 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3179 bit-wise operations.
3180 (andsi3, *andsi3): Ditto.
3181 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3182 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3183 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3184 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3185 nds32_ior_operand, nds32_xor_operand): New predicates.
3187 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3189 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3190 (addsi3, subsi3): ... this.
3192 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3194 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3196 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3198 * config/nds32/nds32.md: Adjust indention.
3200 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3202 * config/nds32/nds32.md (feature): New attribute.
3204 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3206 * config/nds32/nds32.md (subtype): New attribute.
3208 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3211 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3212 expansion to perform a bitwise AND of the argument followed by a
3213 boolean negation of the result.
3215 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3217 PR rtl-optimization/84878
3218 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3219 the basic block. Assert the use reference is not artificial and that
3220 it has an associated insn.
3222 2018-04-04 Michael Matz <matz@suse.de>
3224 * builtins.c (compute_objsize): Pass correct operand
3225 to array_at_struct_end_p.
3227 2018-04-04 Richard Biener <rguenther@suse.de>
3230 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3231 from contexts for DINFO_LEVEL_TERSE and below.
3233 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3235 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3236 resiter_operand condition.
3237 * config/nds32/nds32.md (*move<mode>): Ditto.
3239 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3240 Monk Chiang <sh.chiang04@gmail.com>
3242 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3244 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3246 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3248 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3249 Kito Cheng <kito.cheng@gmail.com>
3251 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3252 nds32_cond_code_str, output_cond_branch,
3253 output_cond_branch_compare_zero, nds32_expand_cbranch,
3254 nds32_expand_cstore, nds32_expand_movcc,
3255 nds32_output_cbranchsi4_equality_zero,
3256 nds32_output_cbranchsi4_equality_reg,
3257 nds32_output_cbranchsi4_equality_reg_or_const_int,
3258 nds32_output_cbranchsi4_greater_less_zero: New functions.
3259 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3260 nds32_expand_cstore, nds32_expand_movcc,
3261 nds32_output_cbranchsi4_equality_zero,
3262 nds32_output_cbranchsi4_equality_reg,
3263 nds32_output_cbranchsi4_equality_reg_or_const_int,
3264 nds32_output_cbranchsi4_greater_less_zero): Declare.
3265 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3266 nds32_rimm11s_operand): New predicates.
3267 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3268 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3271 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3273 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3274 * config/nds32/nds32.md: Ditto.
3275 * config/nds32/pipelines.md: Ditto.
3277 2018-04-04 Richard Biener <rguenther@suse.de>
3279 PR tree-optimization/85168
3280 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3281 propagating abnormals.
3283 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3285 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3287 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3288 Kito Cheng <kito.cheng@gmail.com>
3290 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3291 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3292 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3293 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3294 * config/nds32/nds32.md (sibcall_internal): New.
3295 (sibcall_register): Remove.
3296 (sibcall_immediate): Remove.
3297 (sibcall_value_internal): New.
3298 (sibcall_value_register): Remove.
3299 (sibcall_value_immediate): Remove.
3300 * config/nds32/predicates.md (nds32_general_register_operand): New.
3301 (nds32_call_address_operand): New.
3303 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3305 PR rtl-optimization/85167
3306 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3307 bb_defs if *split_p, instead preinitialize it to NULL.
3309 PR tree-optimization/85156
3310 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3311 evaluating the argument multiple times.
3313 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3315 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3317 (_mm_cvtpd_ps): Likewise.
3318 (_mm_cvttpd_epi32): Likewise.
3319 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3320 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3321 vector, pixel, and bool following altivec.h include.
3323 2018-04-03 Martin Sebor <msebor@redhat.com>
3325 * doc/extend.texi (Common Function Attributes): Clarify.
3326 (const attribute): Likewise.
3327 (pure attribute): Likewise.
3329 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3332 * config/i386/i386.c (ix86_expand_vector_set): Use
3333 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3335 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3337 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3338 instructions when changing rounding bits to preserve precision bits
3339 in the x87 control word.
3341 2018-04-03 Martin Liska <mliska@suse.cz>
3343 PR tree-optimization/82491
3344 * rtl.h (strip_offset_and_add): Replace += suboffset with
3345 poly_uint64 () + suboffset.
3347 2018-03-29 Martin Liska <mliska@suse.cz>
3348 Martin Jambor <mjambor@suse.cz>
3351 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3352 param_type is not an integral or pointer type.
3354 2018-04-03 Richard Biener <rguenther@suse.de>
3356 * sese.h (recompute_all_dominators): Remove.
3358 2018-04-02 Martin Sebor <msebor@redhat.com>
3360 * doc/invoke.texi (-Wrestrict): Fix typos.
3362 2018-04-02 Jim Wilson <jimw@sifive.com>
3364 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3365 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3366 (<optab>di3, <optab>si3_extend): Likewise.
3367 (<optab>si3_mask, <optab>si3_mask_1): New.
3368 (<optab>di3_mask, <optab>di3_mask_1): New.
3369 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3370 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3371 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3373 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3375 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3378 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3380 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3381 (nds32_canonicalize_comparison): New function.
3383 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3384 Kito Cheng <kito.cheng@gmail.com>
3385 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3387 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3388 * config/nds32/constants.md (unspec_volatile_element): Add
3389 UNSPEC_VOLATILE_RELAX_GROUP.
3390 * config/nds32/nds32-relax-opt.c: New file.
3391 * config/nds32/nds32-predicates.c
3392 (nds32_symbol_load_store_p): New function.
3393 * config/nds32/nds32-protos.h
3394 (nds32_symbol_load_store_p): Declare function.
3395 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3396 * config/nds32/nds32.c
3397 (nds32_register_pass): New function to register pass.
3398 (nds32_register_passes): New function to register passes.
3399 * config/nds32/nds32.md (relax_group): New pattern.
3400 * config/nds32/nds32.opt (mrelax-hint): New option.
3401 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3403 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3405 * config/nds32/t-nds32: Modify files dependency.
3407 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3409 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3410 (PROFILE_HOOK): Define its implementation.
3412 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3414 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3415 type and 32-bit size.
3417 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3420 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3421 (V_128_256): New mode iterator.
3422 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3423 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3424 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3426 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3427 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3429 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3432 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3433 NaN inputs correctly.
3435 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3438 * config/rs6000/vsx.md (??r): New mode attribute.
3439 (*vsx_mov<mode>_64bit): Use it.
3440 (*vsx_mov<mode>_32bit): Likewise.
3442 2018-03-30 Martin Sebor <msebor@redhat.com>
3444 PR tree-optimization/84818
3445 * builtins.c (check_access): Use warning_n.
3447 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3450 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3452 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3455 2018-03-30 Julia Koval <julia.koval@intel.com>
3458 * x86-tune.def (movx, partial_reg_dependency): Enable for
3461 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3464 * lra-constraints.c (process_alt_operands): Move setting
3465 this_alternative_matches below.
3467 2018-03-29 Martin Liska <mliska@suse.cz>
3470 * doc/invoke.texi: Document how LTO works with debug info.
3471 Describe auto-load support of binutils. Mention 'x86-64'
3472 as valid option value of -march option.
3474 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3476 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3479 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3480 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3481 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3484 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3487 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3488 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3489 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3490 for RS6000_BTM_POWERPC64.
3491 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3492 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3493 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3498 2018-03-28 Carl Love <cel@us.ibm.com>
3501 2017-09-27 Carl Love <cel@us.ibm.com>
3503 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3504 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3505 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3508 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3510 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3511 instead of __vector bool.
3512 (_mm_max_pu8): Likewise.
3513 (_mm_min_pi16): Likewise.
3515 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3518 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3519 (DIVWEUO): Likewise.
3521 (DIVDEUO): Likewise.
3522 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3523 DIVWEUO and DIVDEUO.
3524 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3525 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3526 (div_extend): Likewise.
3527 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3529 (__builtin_divweuo): Likewise.
3530 (__builtin_divdeo): Likewise.
3531 (__builtin_divdeuo): Likewise.
3533 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3536 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3537 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3539 PR tree-optimization/82004
3540 * gimple-match-head.c (optimize_pow_to_exp): New function.
3541 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3542 Don't fold to exp if optimize_pow_to_exp is false.
3544 2018-03-28 Martin Liska <mliska@suse.cz>
3547 * calls.c (initialize_argument_information): Fix trailing space.
3548 * common.opt: Fix typo and provide better explanation for
3549 -fsanitize-coverage option.
3550 * config/i386/i386.opt: Fix typo.
3552 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3553 Martin Liska <mliska@suse.cz>
3556 * gimplify.c (asan_poison_variable): Don't do the check for
3557 gimplify_omp_ctxp here.
3558 (gimplify_decl_expr): Do it here.
3559 (gimplify_target_expr): Likewise.
3561 2018-03-28 Martin Liska <mliska@suse.cz>
3564 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3565 chkp_type_bounds_count if MPX is not enabled.
3567 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3569 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3571 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3574 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3575 function to create the function decl for complex long double
3576 multiply and divide for -mabi=ieeelongdouble.
3577 (init_float128_ieee): Call it.
3579 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3582 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3583 -fcf-protection=branch -mibt.
3584 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3586 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3589 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3591 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3594 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3595 extern array declarations.
3597 2018-03-27 Richard Biener <rguenther@suse.de>
3600 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3601 explicit single_use checks.
3603 2018-03-27 Richard Biener <rguenther@suse.de>
3605 PR tree-optimization/85082
3606 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3609 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3611 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3612 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3613 Turn on fasynchronous-unwind-tables and funwind-tables.
3615 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3618 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3619 (*bmi_blsr_<mode>_ccz): Ditto.
3621 2018-03-26 Tom de Vries <tom@codesourcery.com>
3623 PR tree-optimization/85063
3624 * omp-general.c (offloading_function_p): New function. Factor out
3626 * omp-offload.c (pass_omp_target_link::gate): ... here.
3627 * omp-general.h (offloading_function_p): Declare.
3628 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3629 with attribute omp declare target for offloading functions.
3631 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3633 PR tree-optimization/84005
3634 * tree-data-ref.h (get_base_for_alignment): Declare.
3635 * tree-data-ref.c (get_base_for_alignment_1): New function.
3636 (get_base_for_alignment): Likewise.
3637 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3638 get_base_for_alignment to find a suitable base object, instead
3639 of always using drb->base_address.
3641 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3644 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3645 known size by default.
3647 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3650 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3651 and non BLKmode operands.
3653 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3656 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3657 Clean up attributes.
3659 2018-03-23 Richard Biener <rguenther@suse.de>
3662 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3663 we are going to emit early debug for LTO.
3665 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3668 * function.c (match_asm_constraints_1): Don't optimize if input
3669 doesn't satisfy general_operand predicate for output's mode.
3672 * alias.c (write_dependence_p): Don't require for x_canonicalized
3673 non-VOIDmode if x has VOIDmode.
3676 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3677 just don't try to optimize it rather than assert it never happens.
3679 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3681 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3682 macro expansions for definition of ST_INTERNAL_<mode> and
3683 LD_INTERNAL_<mode> builtins.
3684 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3686 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3688 (altivec_expand_st_builtin): Likewise.
3689 (altivec_expand_builtin): Remove calls to deleted functions.
3690 (rs6000_address_for_altivec): Delete this function.
3691 * config/rs6000/vector.md: Remove expands for
3692 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3694 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3697 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3698 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3699 re-computing once computed.
3700 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3701 (arm_init_machine_status): Initialize
3702 machine->static_chain_stack_bytes.
3704 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3707 * doc/extend.texi: Add four new prototypes for vec_ld.
3708 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3709 definitions for more logical presentation.
3710 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3711 entries for V1TI variants of __builtin_altivec_ld builtin.
3712 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3713 handling of V1TI variant of LVX icode pattern.
3714 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3715 (rs6000_gimple_fold_builtin): Likewise.
3716 (altivec_init_builtins): Add code to define
3717 __builtin_altivec_lvx_v1ti function.
3719 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3722 * function.c (match_asm_constraints_1): Don't do the optimization
3723 if input isn't a REG, SUBREG, MEM or constant.
3725 2018-03-22 Tom de Vries <tom@codesourcery.com>
3727 PR tree-optimization/84956
3728 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3729 bb_has_abnormal_pred.
3731 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3734 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3735 DECL_INITIAL (decl) to decl at the end.
3736 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3739 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3741 * doc/extend.texi (__builtin_tgmath): Document when complex
3742 integer types are treated as _Complex _Float64.
3744 2018-03-21 Tom de Vries <tom@codesourcery.com>
3746 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3748 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3750 PR tree-optimization/84960
3751 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3752 if it is ENTRY block, move them into single succ of ENTRY in that case.
3754 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3756 PR tree-optimization/84811
3757 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3758 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3759 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3760 (known_subrange_p): Update accordingly. Cast each value involved
3761 in the size comparison, rather than casting the result of the
3764 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3766 PR tree-optimization/84982
3767 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3768 by flipping the least significant bit rather than all bits from
3769 bitpos to bitpos + bitsize - 1.
3771 2018-03-21 Nathan Sidwell <nathan@acm.org>
3773 * doc/extend.texi (Deprecated Features): Remove mention of
3774 long-deleted deprecations.
3776 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3779 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3780 * configure: Regenerate.
3782 2018-03-21 Tom de Vries <tom@codesourcery.com>
3784 PR tree-optimization/83126
3785 * tree-parloops.c (num_phis): New function.
3786 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3788 2018-03-21 Nathan Sidwell <nathan@acm.org>
3790 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3791 mention anon-struct/union members and trailing attributes.
3793 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3795 PR tree-optimization/84969
3796 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3797 builtin memset partitions if they set different rhs values.
3799 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3801 PR rtl-optimization/84989
3802 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3803 VEC_DUPLICATE with scalar result mode.
3805 2018-03-21 Martin Liska <mliska@suse.cz>
3808 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3809 not intended return statement.
3811 2018-03-21 Martin Liska <mliska@suse.cz>
3814 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3815 (chkp_find_bound_slots_1): Limit number of iterations.
3817 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3820 * Minor grammar fixes for x86 options.
3822 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3825 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3826 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3828 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3831 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3832 (altivec_lvx_<mode>_1op): Likewise.
3833 (altivec_stvx_<mode>_2op): Likewise.
3834 (altivec_stvx_<mode>_1op): Likewise.
3835 (altivec_lvx_<VM2:mode>): New define_expand.
3836 (altivec_stvx_<VM2:mode>): Likewise.
3837 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3838 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3839 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3840 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3841 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3842 (rs6000_gen_lvx): Likewise.
3843 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3844 (altivec_expand_stv_builtin): Likewise.
3845 (altivec_expand_builtin): Likewise.
3846 * config/rs6000/vector.md: Likewise.
3848 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3851 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3854 2018-03-20 Richard Biener <rguenther@suse.de>
3857 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3858 sign-conversions as zero, fall back to standard scalar_stmt
3861 2018-03-20 Martin Liska <mliska@suse.cz>
3864 * predict.c (rebuild_frequencies): Handle case when we have
3865 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3867 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3870 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3871 flag_section_anchors.
3872 * varasm.c (use_blocks_for_decl_p): Remove hack for
3873 dw2_force_const_mem.
3876 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3878 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3879 be created, use lowpart_subreg of operands[0] rather than operands[0]
3881 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3882 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3883 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3884 and n constraint instead of aarch64_shift_imm_di and Usd.
3885 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3886 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3888 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3891 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3892 to favor GPR over NEON registers.
3893 (<shift>di3_neon): Likewise.
3895 2018-03-20 Tom de Vries <tom@codesourcery.com>
3898 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3899 (nvptx_process_pars): Emit bar.sync asap and alap.
3901 2018-03-20 Tom de Vries <tom@codesourcery.com>
3904 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3905 seen_label if seen_label is already set.
3907 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3910 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3911 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3912 Use 1U instead of 1. Formatting fixes.
3915 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3916 instead of TREE_TYPE (s1) for the return value.
3918 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3920 PR tree-optimization/84946
3921 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3922 bitsize + bitsize in poly_uint64 rather than poly_int64.
3925 * dwarf2asm.c: Include fold-const.c.
3926 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3927 of decl rather than decl itself.
3929 PR rtl-optimization/84643
3930 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3932 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3935 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3936 calling assemble_variable.
3938 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3941 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3942 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3944 2018-03-19 Jim Wilson <jimw@sifive.com>
3947 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3948 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3949 (riscv_first_stack_step): Likewise.
3950 (riscv_option_override): Use STACK_BOUNDARY instead of
3952 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3954 (BIGGEST_ALIGNMENT): Set to 128.
3955 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3956 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3959 2018-03-19 Richard Biener <rguenther@suse.de>
3961 PR tree-optimization/84933
3962 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3963 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3965 2018-03-19 Richard Biener <rguenther@suse.de>
3967 PR tree-optimization/84859
3968 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3969 (cond_if_else_store_replacement): Perform sinking operation on
3970 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3971 Generalize what a BB with a single eligible store is.
3973 2018-03-19 Richard Biener <rguenther@suse.de>
3975 PR tree-optimization/84929
3976 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3977 chrec_is_positive against non-chrec arg.
3979 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3982 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3984 2018-03-18 Martin Liska <mliska@suse.cz>
3986 PR rtl-optimization/84635
3987 * regrename.c (build_def_use): Use matches_mode only when
3990 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3992 PR tree-optimization/84913
3993 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3994 vectorize chains of COND_EXPRs.
3996 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3998 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4000 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4002 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4004 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4006 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4008 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4009 Kito Cheng <kito.cheng@gmail.com>
4011 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4012 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4013 (nds32_adjust_reg_alloc_order): New function.
4014 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4016 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4018 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4019 nds32_print_operand, nds32_print_operand_address): Use
4020 HOST_WIDE_INT_PRINT_DEC instead.
4022 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4024 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4026 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4029 * config/i386/i386.c (initial_ix86_tune_features,
4030 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4032 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4033 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4034 rather than 1u << ix86_tune. Formatting fix.
4035 (ix86_option_override_internal): Change ix86_arch_mask from
4036 unsigned int to unsigned HOST_WIDE_INT, initialize to
4037 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4038 (ix86_function_specific_restore): Likewise.
4040 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4043 * postreload.c (reload_combine_recognize_pattern): Perform
4044 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4045 truncate_int_for_mode the result for the destination's mode.
4048 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4049 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4051 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4054 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4057 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4058 regno_allocno_class_array and sorted_pseudos.
4059 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4060 insns where regno is used.
4062 2018-03-16 Martin Liska <mliska@suse.cz>
4065 * multiple_target.c (create_dispatcher_calls): Redirect
4066 reference in the symbol table.
4068 2018-03-16 Martin Liska <mliska@suse.cz>
4071 * multiple_target.c (create_dispatcher_calls): Redirect also
4074 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4078 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4079 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4080 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4082 2018-03-16 Julia Koval <julia.koval@intel.com>
4084 * doc/invoke.texi (Skylake Server): Add CLWB.
4085 Cannonlake): Remove CLWB.
4087 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4089 PR tree-optimization/84841
4090 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4092 (FLOAT_ONE_CONST_TYPE): Define.
4093 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4094 (sort_by_operand_rank): Put entries with higher constant_type last
4095 rather than first to match comments.
4097 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4099 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4102 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4105 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4106 check and use address of target always.
4108 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4111 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4112 (indirect_thunk_bnd_needed): Likewise.
4113 (indirect_thunks_used): Likewise.
4114 (indirect_thunks_bnd_used): Likewise.
4115 (indirect_return_needed): New.
4116 (indirect_return_bnd_needed): Likewise.
4117 (output_indirect_thunk_function): Add a bool argument for
4119 (output_indirect_thunk_function): Don't generate alias for
4120 function return thunk.
4121 (ix86_code_end): Call output_indirect_thunk_function to generate
4122 function return thunks.
4123 (ix86_output_function_return): Set indirect_return_bnd_needed
4124 and indirect_return_needed instead of indirect_thunk_bnd_needed
4125 and indirect_thunk_needed.
4127 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4129 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4130 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4131 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4133 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4134 Paul Hua <paul.hua.gm@gmail.com>
4137 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4139 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4141 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4142 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4145 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4148 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4149 instead of GET_MODE_SIZE when comparing Units.
4151 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4154 * varasm.c (hash_section): Return an unchangeble hash value
4155 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4156 Return !aarch64_can_use_per_function_literal_pools_p ().
4158 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4161 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4162 rather than address of cmode as last argument to prepare_cmp_insn.
4164 2018-03-15 Julia Koval <julia.koval@intel.com>
4166 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4167 F_AVX512VNNI, F_AVX512BITALG): New.
4169 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4172 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4173 insn for floating-point loads and stores.
4175 2018-03-14 Carl Love <cel@us.ibm.com>
4177 * config/rs6000/rs6000-c.c: Add macro definitions for
4178 ALTIVEC_BUILTIN_VEC_PERMXOR.
4179 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4180 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4181 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4182 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4184 * config/doc/extend.texi: Add prototypes for vec_permxor.
4186 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4189 * diagnostic-show-locus.c (class layout_point): Convert m_line
4190 from int to linenum_type.
4191 (line_span::comparator): Use linenum "compare" function when
4192 comparing line numbers.
4193 (test_line_span): New function.
4194 (layout_range::contains_point): Convert param "row" from int to
4196 (layout_range::intersects_line_p): Likewise.
4197 (layout::will_show_line_p): Likewise.
4198 (layout::print_source_line): Likewise.
4199 (layout::should_print_annotation_line_p): Likewise.
4200 (layout::print_annotation_line): Likewise.
4201 (layout::print_leading_fixits): Likewise.
4202 (layout::annotation_line_showed_range_p): Likewise.
4203 (struct line_corrections): Likewise for field m_row.
4204 (line_corrections::line_corrections): Likewise for param "row".
4205 (layout::print_trailing_fixits): Likewise.
4206 (layout::get_state_at_point): Likewise.
4207 (layout::get_x_bound_for_row): Likewise.
4208 (layout::print_line): Likewise.
4209 (diagnostic_show_locus): Likewise for locals "last_line" and
4211 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4212 * input.c (selftest::test_linenum_comparisons): New function.
4213 (selftest::input_c_tests): Call it.
4214 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4215 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4216 * selftest.h (ASSERT_GT): New macro.
4217 (ASSERT_GT_AT): New macro.
4218 (ASSERT_LT): New macro.
4219 (ASSERT_LT_AT): New macro.
4221 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4223 PR rtl-optimization/84780
4224 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4226 2018-03-14 Martin Liska <mliska@suse.cz>
4228 * tree.c (record_node_allocation_statistics): Use
4229 get_stats_node_kind.
4230 (get_stats_node_kind): New function extracted from
4231 record_node_allocation_statistics.
4232 (free_node): Use get_stats_node_kind.
4234 2018-03-14 Richard Biener <rguenther@suse.de>
4236 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4237 that the value-set of ANTIC_IN doesn't grow.
4240 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4242 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4243 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4245 2018-03-14 Julia Koval <julia.koval@intel.com>
4247 * config.gcc (icelake-client, icelake-server): New.
4249 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4250 (initial_ix86_arch_features): Ditto.
4251 (PTA_SKYLAKE): Add SGX.
4252 (PTA_ICELAKE): Remove.
4253 (PTA_ICELAKE_CLIENT): New.
4254 (PTA_ICELAKE_SERVER): New.
4255 (ix86_option_override_internal): Split up icelake on icelake client and
4257 (get_builtin_code_for_version): Ditto.
4258 (fold_builtin_cpu): Ditto.
4259 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4260 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4261 * config/i386/i386.h (processor_type): Ditto.
4262 * doc/invoke.texi: Ditto.
4264 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4267 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4268 INTEGER_CST offset, add it together with bitpos / 8 and
4269 sign extend based on POINTER_SIZE.
4273 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4276 * config/i386/constraints.md (Yc): New register constraint.
4277 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4278 Use Yc constraint for alternative 2 of operand 0. Remove
4279 preferred_for_speed attribute.
4281 2018-03-14 Richard Biener <rguenther@suse.de>
4283 PR tree-optimization/84830
4284 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4285 with the old one to avoid oscillations.
4287 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4290 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4292 (assign_by_spills): Return a flag of reload assignment failure.
4293 Do not process the reload assignment failures. Do not spill other
4294 reload pseudos if they has the same reg class. Update n if
4296 (lra_assign): Add a return arg. Set up from the result of
4297 assign_by_spills call.
4298 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4299 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4300 usage_insns if it is not NULL.
4301 (spill_hard_reg_in_range): New function.
4302 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4303 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4304 function prototypes.
4305 (lra_assign): Change prototype.
4306 * lra.c (lra): Add code to deal with fails by splitting hard reg
4309 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4311 * config/riscv/riscv.opt (mrelax): New option.
4312 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4313 "norelax" when riscv_mrelax is disabled.
4314 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4316 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4319 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4320 reassociation for int modes.
4322 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4324 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4325 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4327 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4328 * config/aarch64/aarch64-sve.md
4329 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4330 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4331 (*extend<mode><Vwide>2): Rename to...
4332 (aarch64_sve_extend<mode><Vwide>2): ...this.
4333 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4334 renaming the old pattern to...
4335 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4337 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4338 define_expand, renaming the old pattern to...
4339 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4340 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4341 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4342 account when deciding which SVE instruction the optab should use.
4343 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4345 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4347 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4348 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4349 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4350 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4351 (tlsdesc_small_<mode>): Turn a define_expand and use
4352 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4353 (tlsdesc_small_advsimd_<mode>): ...this.
4354 (tlsdesc_small_sve_<mode>): New pattern.
4356 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4358 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4359 (UNSPEC_UMUL_HIGHPART): New constants.
4360 (MUL_HIGHPART): New int iteraor.
4361 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4362 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4364 (*<su>mul<mode>3_highpart): New define_insn.
4366 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4369 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4372 2018-03-13 Martin Liska <mliska@suse.cz>
4375 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4377 (sem_item_optimizer::~sem_item_optimizer): Release it.
4378 (sem_item_optimizer::merge_classes): Register variable aliases.
4379 (sem_item_optimizer::fixup_pt_set): New function.
4380 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4381 * ipa-icf.h: Declare new variables and functions.
4383 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4386 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4387 integer_pow2p@2 and test integer_pow2p in condition.
4388 (A < 0 ? C : 0): Similarly for @1.
4391 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4392 characters starting at p contain '\0' character, don't look beyond
4396 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4397 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4400 * reg-stack.c (change_stack): Change update_end var from int to
4401 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4402 also call set_block_for_insn on the newly added insns and rescan.
4405 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4406 on the last operand.
4409 * tree.c (stabilize_reference_1): Return save_expr (e) for
4410 STATEMENT_LIST even if it doesn't have side-effects.
4412 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4414 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4416 2018-03-12 Renlin Li <renlin.li@arm.com>
4418 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4419 aarch64_output_scalar_simd_mov_immediate.
4421 2018-03-12 Martin Sebor <msebor@redhat.com>
4423 PR tree-optimization/83456
4424 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4425 for perfectly overlapping calls to memcpy.
4426 (gimple_fold_builtin_memory_chk): Same.
4427 (gimple_fold_builtin_strcpy): Handle no-warning.
4428 (gimple_fold_builtin_stxcpy_chk): Same.
4429 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4431 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4433 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4434 parameter. Use it for SFmode.
4435 (rs6000_function_arg_advance_1): Adjust.
4436 (rs6000_function_arg): Adjust.
4437 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4439 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4441 PR rtl-optimization/84169
4442 PR rtl-optimization/84780
4443 * combine.c (can_combine_p): Check for a 2-insn combination whether
4444 the destination register is used between the two insns, too.
4446 2018-03-12 Richard Biener <rguenther@suse.de>
4448 PR tree-optimization/84803
4449 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4450 for refs DR analysis didn't process.
4452 2018-03-12 Richard Biener <rguenther@suse.de>
4454 PR tree-optimization/84777
4455 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4456 force-vectorize loops ignore whether we are optimizing for size.
4458 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4460 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4461 (TARGET_MD_ASM_ADJUST): Define.
4463 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4464 Kito Cheng <kito.cheng@gmail.com>
4465 Chung-Ju Wu <jasonwucj@gmail.com>
4467 * config/nds32/nds32.c (nds32_compute_stack_frame,
4468 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4469 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4470 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4471 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4472 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4473 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4474 * config/nds32/nds32.md (prologue, epilogue): Use macro
4475 NDS32_V3PUSH_AVAILABLE_P to do checking.
4477 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4480 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4481 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4482 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4483 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4484 addition of most attributes on !orig_type_die or the attribute not
4485 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4487 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4488 Chung-Ju Wu <jasonwucj@gmail.com>
4490 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4492 * config/nds32/nds32.opt (mvh): New option.
4494 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4495 Chung-Ju Wu <jasonwucj@gmail.com>
4497 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4499 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4500 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4503 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4504 Chung-Ju Wu <jasonwucj@gmail.com>
4506 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4508 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4509 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4511 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4512 Kito Cheng <kito.cheng@gmail.com>
4513 Chung-Ju Wu <jasonwucj@gmail.com>
4515 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4516 UNSPEC_FFMISM and UNSPEC_FLMISM.
4517 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4518 for ffb, ffmism and flmism.
4519 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4520 (unspec_ffmism): Ditto.
4521 (unspec_flmism): Ditto.
4522 (nds32_expand_builtin_impl): Check if string extension is available.
4523 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4524 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4526 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4529 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4532 * lra-assigns.c (assign_by_spills): Return a flag of reload
4533 assignment failure. Do not process the reload assignment
4534 failures. Do not spill other reload pseudos if they has the same
4536 (lra_assign): Add a return arg. Set up from the result of
4537 assign_by_spills call.
4538 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4539 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4540 usage_insns if it is not NULL.
4541 (spill_hard_reg_in_range): New function.
4542 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4543 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4544 function prototypes.
4545 (lra_assign): Change prototype.
4546 * lra.c (lra): Add code to deal with fails by splitting hard reg
4549 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4552 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4554 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4557 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4558 (dw_val_node): Add val_symbolic_view.
4559 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4560 (symview_upper_bound): New.
4561 (new_line_info_table): Initialize symviews_since_reset.
4562 (dwarf2out_source_line): Count symviews_since_reset and set
4563 symview_upper_bound.
4564 (dw_val_equal_p): Handle symview.
4565 (add_AT_symview): New.
4566 (print_dw_val): Handle symview.
4567 (attr_checksum, attr_checksum_ordered): Likewise.
4568 (same_dw_val_p, size_of_die): Likewise.
4569 (value_format, output_die): Likewise.
4570 (add_high_low_attributes): Use add_AT_symview for entry_view.
4571 (dwarf2out_finish): Reset symview_upper_bound, clear
4574 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4577 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4578 Add strict argument and use it.
4579 (rs6000_split_multireg_move): Update for new strict argument.
4580 (mem_operand_gpr): Disallow all non-offsettable addresses.
4581 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4583 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4586 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4587 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4588 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4591 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4592 decl, use remap_type if we want to use the type.
4594 2018-03-09 Martin Sebor <msebor@redhat.com>
4596 PR tree-optimization/84526
4597 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4599 (builtin_access::generic_overlap): Be prepared to handle non-array
4602 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4604 PR rtl-optimization/84682
4605 * lra-constraints.c (process_address_1): Check is_address flag
4606 for address constraints.
4607 (process_alt_operands): Likewise.
4608 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4609 preprocess_constraints.
4610 * recog.h (preprocess_constraints): Add oploc parameter.
4612 * recog.c (preprocess_constraints): Test address_operand for
4613 CT_ADDRESS constraints.
4615 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4618 * lra-assigns.c (assign_by_spills): Return a flag of reload
4619 assignment failure. Do not process the reload assignment
4620 failures. Do not spill other reload pseudos if they has the same
4622 (lra_assign): Add a return arg. Set up from the result of
4623 assign_by_spills call.
4624 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4625 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4626 usage_insns if it is not NULL.
4627 (spill_hard_reg_in_range): New function.
4628 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4629 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4630 function prototypes.
4631 (lra_assign): Change prototype.
4632 * lra.c (lra): Add code to deal with fails by splitting hard reg
4635 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4638 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4639 Accept complain bool parameter. Only emit errors if it is true.
4640 (arm_parse_cpu_option_name): Likewise.
4641 (arm_target_thumb_only): Adjust callers of the above.
4642 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4643 prototype to take a default true bool parameter.
4644 (arm_parse_arch_option_name): Likewise.
4646 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4647 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4651 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4652 * configure: Regenerate.
4653 * configure.ac ("linker --version-script option"): New.
4654 ("linker soname option"): New.
4656 2018-03-09 Richard Biener <rguenther@suse.de>
4658 PR tree-optimization/84775
4659 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4660 immediate uses of predicate stmts and mark them modified.
4663 PR tree-optimization/84178
4664 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4666 (version_loop_for_if_conversion): Delay update_ssa call.
4667 (tree_if_conversion): Delay update_ssa until after predicate
4670 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4673 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4674 when the function accesses prior frames.
4676 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4679 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4680 gen_llsym, otherwise call maybe_gen_llsym.
4683 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4684 has ',' character inside of it.
4686 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4689 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4690 as clobbering CC_REGNUM.
4692 2018-03-08 Richard Biener <rguenther@suse.de>
4695 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4696 (follow_copies_to_constant): Do not follow SSA names registered
4699 2018-03-08 Richard Biener <rguenther@suse.de>
4701 PR tree-optimization/84178
4702 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4704 (version_loop_for_if_conversion): Delay update_ssa call.
4705 (tree_if_conversion): Delay update_ssa until after predicate
4708 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4710 PR tree-optimization/84178
4711 * tree-if-conv.c (release_bb_predicate): Remove the
4712 the assertion that the stmts have NULL use_ops.
4713 Discard the statements, asserting that they haven't
4714 yet been added to a BB.
4716 2018-03-08 Richard Biener <rguenther@suse.de>
4718 PR tree-optimization/84746
4719 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4720 (phi_translate): Pass in destination ANTIC_OUT set.
4721 (phi_translate_1): Likewise. For a simplified result lookup
4722 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4723 (phi_translate_set): Adjust.
4724 (do_pre_regular_insertion): Likewise.
4725 (do_pre_partial_partial_insertion): Likewise.
4727 2018-03-08 Martin Liska <mliska@suse.cz>
4729 PR gcov-profile/84735
4730 * doc/gcov.texi: Document usage of profile files.
4731 * gcov-io.h: Document changes in the format.
4733 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4737 * dwarf2out.c (struct dw_line_info_table): Update comments for
4739 (FORCE_RESET_NEXT_VIEW): New.
4740 (FORCE_RESETTING_VIEW_P): New.
4741 (RESETTING_VIEW_P): Check for -1 too.
4742 (ZERO_VIEW_P): Likewise.
4743 (new_line_info_table): Force-reset next view.
4744 (dwarf2out_begin_function): Likewise.
4745 (dwarf2out_source_line): Simplify zero_view_p initialization.
4746 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4747 view directly. Omit view when omitting .loc at line 0.
4749 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4751 PR tree-optimization/84740
4752 * tree-switch-conversion.c (process_switch): Call build_constructors
4753 only if info.phi_count is non-zero.
4755 PR tree-optimization/84739
4756 * tree-tailcall.c (find_tail_calls): Check call arguments against
4757 DECL_ARGUMENTS (current_function_decl) rather than
4758 DECL_ARGUMENTS (func) when checking for tail recursion.
4760 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4762 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4763 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4764 Volker Reichelt's entry and add entries for people that perform
4765 GCC fuzzy testing and report numerous bugs.
4767 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4770 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4771 readonly data in sdata, if that is disabled.
4772 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4773 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4774 -mreadonly-in-sdata option.
4776 2018-03-07 Martin Sebor <msebor@redhat.com>
4778 PR tree-optimization/84468
4779 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4780 basic block when looking for nul assignment.
4782 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4785 * except.h (output_function_exception_table): Adjust prototype.
4786 * except.c (output_function_exception_table): Remove FNNAME parameter
4787 and add SECTION parameter. Ouput one part of the table at a time.
4788 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4789 the first part of the exception table and emit unwind directives.
4790 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4791 (i386_pe_seh_cold_init): Likewise.
4792 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4793 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4794 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4795 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4796 * config/i386/winnt.c: Include except.h.
4797 (struct seh_frame_state): Add reg_offset, after_prologue and
4798 in_cold_section fields.
4799 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4800 (i386_pe_seh_cold_init): New function.
4801 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4802 to seh->in_cold_section.
4803 (seh_emit_push): Record the offset of the push.
4804 (seh_emit_save): Record the offet of the save.
4805 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4806 Test seh->after_prologue to disregard the epilogue.
4807 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4808 (i386_pe_end_cold_function): New function.
4810 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4813 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4814 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4817 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4819 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4820 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4823 * multiple_target.c: Include tree-inline.h and intl.h.
4824 (expand_target_clones): Diagnose and fail if node->definition and
4825 !tree_versionable_function_p (node->decl).
4827 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4829 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4831 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4832 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4833 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4835 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4838 * combine.c (try_combine): Use reg_or_subregno instead of handling
4839 just paradoxical SUBREGs and REGs.
4841 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4843 * config/arc/arc.c (arc_finalize_pic): Remove function.
4844 (arc_must_save_register): We use single base PIC register, remove
4845 checks to save/restore the PIC register.
4846 (arc_expand_prologue): Likewise.
4847 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4849 (arc_verify_short): Likewise.
4850 (arc_attr_type): Likewise.
4851 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4852 (walk_stores): Likewise.
4853 (arc_address_cost): Make it static.
4854 (arc_verify_short): Likewise.
4855 (branch_dest): Likewise.
4856 (arc_attr_type): Likewise.
4857 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4858 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4859 (arc_final_prescan_insn): Remove inserting the nops due to
4860 hardware hazards. It is done in reorg step.
4861 (insn_length_variant_t): Remove.
4862 (insn_length_parameters_t): Likewise.
4863 (arc_insn_length_parameters): Likewise.
4864 (arc_get_insn_variants): Likewise.
4865 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4867 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4870 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4873 PR tree-optimization/84687
4874 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4876 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4878 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4880 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4881 Rename to ppc_speculation_barrier.
4882 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4883 __builtin_ppc_speculation_barrier.
4885 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4888 * combine.c (combine_simplify_rtx): Don't try to simplify if
4889 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4892 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4894 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4895 to 32 bytes when compiling for POWER9.
4897 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4900 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4901 regparm >= 3 with no arg reg available also for calls with
4902 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4905 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4907 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4909 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4912 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4914 2018-03-05 Richard Biener <rguenther@suse.de>
4916 PR tree-optimization/84486
4917 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4918 When inserting a __builtin_assume_aligned call set the LHS
4919 SSA name alignment info accordingly.
4921 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4923 PR tree-optimization/84114
4924 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4925 Avoid reassociation of FLOAT_MODE addition.
4927 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4929 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4930 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4931 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4932 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4933 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4934 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4935 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4937 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4938 __builtin_ia32_wbinvd): New builtins.
4939 (SPECIAL_ARGS2): New.
4940 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4941 (SPECIAL_ARGS2): New.
4942 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4943 (ix86_valid_target_attribute_inner_p): Ditto.
4944 (ix86_init_mmx_sse_builtins): Add special_args2.
4945 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4946 TARGET_WBNOINVD_P): New.
4947 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4948 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4949 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4950 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4951 * config/i386/pconfigintrin.h: New file.
4952 * config/i386/wbnoinvdintrin.h: Ditto.
4953 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4954 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4956 2018-03-05 Richard Biener <rguenther@suse.de>
4958 PR tree-optimization/84670
4959 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4961 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4962 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4963 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4964 doesn't grow if all successors (recursively) were visited at least
4967 2018-03-05 Richard Biener <rguenther@suse.de>
4969 PR tree-optimization/84650
4970 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4971 if executed in the loop pipeline.
4973 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4975 * doc/configfiles.texi (Configuration Files): Move info about
4976 conditionalizing $target-protos.h to...
4977 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4978 differs from $target-protos.h.
4980 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4981 Chung-Ju Wu <jasonwucj@gmail.com>
4983 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4984 * config/nds32/nds32-multiple.md (setmemsi): Define.
4985 * config/nds32/nds32-memory-manipulation.c
4986 (nds32_gen_dup_4_byte_to_word_value): New.
4987 (emit_setmem_word_loop): New.
4988 (emit_setmem_byte_loop): New.
4989 (nds32_expand_setmem_loop): New.
4990 (nds32_expand_setmem_loop_v3m): New.
4991 (nds32_expand_setmem_unroll): New.
4992 (nds32_expand_setmem): New.
4994 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4995 Chung-Ju Wu <jasonwucj@gmail.com>
4997 * config/nds32/nds32-memory-manipulation.c
4998 (nds32_emit_load_store): New.
4999 (nds32_emit_post_inc_load_store): New.
5000 (nds32_emit_mem_move): New.
5001 (nds32_emit_mem_move_block): New.
5002 (nds32_expand_movmemsi_loop_unknown_size): New.
5003 (nds32_expand_movmemsi_loop_known_size): New.
5004 (nds32_expand_movmemsi_loop): New.
5005 (nds32_expand_movmemsi_unroll): New.
5006 (nds32_expand_movmemqi): Rename ...
5007 (nds32_expand_movmemsi): ... to this.
5008 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5009 (movmemsi): ... to this.
5010 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5011 (nds32_expand_movmemsi): ... to this.
5013 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5014 Monk Chiang <sh.chiang04@gmail.com>
5015 Chung-Ju Wu <jasonwucj@gmail.com>
5017 * config/nds32/nds32-protos.h
5018 (nds32_expand_load_multiple): New arguments.
5019 (nds32_expand_store_multiple): Ditto.
5020 (nds32_valid_multiple_load_store): Rename ...
5021 (nds32_valid_multiple_load_store_p): ... to this.
5022 * config/nds32/nds32-memory-manipulation.c
5023 (nds32_expand_load_multiple): Refine implementation.
5024 (nds32_expand_store_multiple): Ditto.
5025 * config/nds32/nds32-multiple.md
5026 (load_multiple): Update nds32_expand_load_multiple interface.
5027 (store_multiple): Update nds32_expand_store_multiple interface.
5028 * config/nds32/nds32-predicates.c
5029 (nds32_valid_multiple_load_store): Rename ...
5030 (nds32_valid_multiple_load_store_p): ... to this and refine
5032 * config/nds32/predicates.md
5033 (nds32_load_multiple_and_update_address_operation): New predicate.
5034 (nds32_store_multiple_and_update_address_operation): New predicate.
5036 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5037 Chung-Ju Wu <jasonwucj@gmail.com>
5039 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5040 (combo): New attribute.
5041 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5043 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5045 * config/nds32/nds32.opt: Change -mcmodel= default value.
5047 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5048 Monk Chiang <sh.chiang04@gmail.com>
5049 Chung-Ju Wu <jasonwucj@gmail.com>
5051 * config/nds32/constants.md (unspec_element): New enum.
5052 * config/nds32/constraints.md (Umw): New constraint.
5053 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5054 * config/nds32/nds32-intrinsic.md: Likewise.
5055 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5056 (nds32_valid_smw_lwm_base_p): New.
5057 (nds32_output_smw_single_word): New.
5058 (nds32_output_lmw_single_word): New.
5059 (nds32_expand_unaligned_load): New.
5060 (nds32_expand_unaligned_store): New.
5061 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5062 (nds32_output_smw_single_word): Declare.
5063 (nds32_output_lmw_single_word): Declare.
5064 (nds32_expand_unaligned_load): Declare.
5065 (nds32_expand_unaligned_store): Declare.
5066 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5067 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5068 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5069 NDS32_BUILTIN_UASTORE_DW.
5070 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5073 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5074 Kito Cheng <kito.cheng@gmail.com>
5075 Chung-Ju Wu <jasonwucj@gmail.com>
5077 * config/nds32/nds32-intrinsic.c
5078 (nds32_expand_builtin_null_ftype_reg): Delete.
5079 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5080 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5081 (nds32_read_argument): New.
5082 (nds32_legitimize_target): Ditto.
5083 (nds32_legitimize_argument): Ditto.
5084 (nds32_check_constant_argument): Ditto.
5085 (nds32_expand_unop_builtin): Ditto.
5086 (nds32_expand_unopimm_builtin): Ditto.
5087 (nds32_expand_binop_builtin): Ditto.
5088 (nds32_builtin_decl_impl): Ditto.
5089 (builtin_description): Ditto.
5090 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5091 (nds32_init_builtins_impl): Ditto.
5092 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5093 (nds32_builtin_decl): New.
5094 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5095 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5097 2018-03-02 Jeff Law <law@redhat.com>
5099 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5100 (redundant_insn, fill_simple_delay_slots): Likewise.
5101 (fill_slots_from_thread): Likewise.
5102 * resource.c (mark_referenced_resources): Likewise.
5103 (mark_set_resources, find_dead_or_set_registers): Likewise.
5105 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5107 * substring-locations.h (format_warning_va): Formatting fix for
5109 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5111 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5112 * substring-locations.c: Include intl.h.
5113 (format_warning_va): Turned into small wrapper around
5114 format_warning_n_va, renamed to ...
5115 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5116 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5118 (format_warning_at_substring_n): New function.
5119 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5120 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5121 format_warning_at_substring with just a shorter name instead of
5122 const function pointer.
5123 (fmtwarn_n): New function.
5124 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5125 appropriate, get rid of all the fmtstr temporaries, move conditionals
5126 with G_() wrapped string literals directly into fmtwarn arguments,
5127 cast dir.len to (int), formatting fixes.
5129 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5131 * doc/invoke.texi: Remove "Cilk Plus" references.
5133 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5134 Richard Biener <rguenther@suse.de>
5137 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5138 for error or warning attributes if CALL_FROM_THUNK_P is set.
5141 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5144 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5145 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5148 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5149 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5151 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5152 instead of -1U in last predictors element's probability member.
5154 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5157 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5158 arguments if they are comparable.
5160 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5162 PR tree-optimization/84634
5163 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5164 masks and masked_loop_p with a single loop_masks, making sure it's
5165 null for bb vectorization.
5167 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5169 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5170 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5171 loop->force_vectorize to check whether there is no alias.
5173 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5176 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5178 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5180 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5181 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5182 instead of a loop around prev_real_insn.
5183 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5187 * config/i386/i386.c (ix86_print_operand): Use conditional
5188 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5191 2018-03-02 Richard Biener <rguenther@suse.de>
5193 PR tree-optimization/84427
5194 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5195 (bitmap_set_subtract_values): Rewrite to handle multiple
5198 (prune_clobbered_mems): Likewise.
5199 (phi_translate): Take edge instead of pred/phiblock.
5200 (phi_translate_1): Likewise.
5201 (phi_translate_set): Likewise. Insert all translated
5202 exprs for a value into the set, keeping possibly multiple
5203 expressions per value.
5204 (compute_antic_aux): Adjust for phi_translate changes.
5205 When intersecting union the expressions and prune those
5206 not in the final value set, keeping possibly multiple
5207 expressions per value. Do not use value-insertion
5208 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5209 all expressions. Add verification that the value-sets
5210 only shrink during iteration.
5211 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5212 (do_pre_regular_insertion): Likewise.
5213 (do_pre_partial_partial_insertion): Likewise.
5215 2018-03-02 Richard Biener <rguenther@suse.de>
5218 * config/darwin.c (saved_debug_info_level): New static global.
5219 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5220 (darwin_asm_lto_end): Restore debug info generation settings.
5222 2018-03-01 Martin Liska <mliska@suse.cz>
5225 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5228 2018-03-01 Richard Biener <rguenther@suse.de>
5231 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5232 type annotation with LTO when debug was disabled at compile-time.
5234 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5236 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5238 (mips_final_postscan_insn): Likewise.
5240 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5242 PR rtl-optimization/84528
5243 * alias.c (init_alias_target): Add commentary.
5244 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5245 a unique base value if the frame pointer is not eliminated
5246 to the stack pointer.
5248 2018-03-01 Tom de Vries <tom@codesourcery.com>
5250 PR rtl-optimization/83327
5251 * lra-int.h (hard_regs_spilled_into): Declare.
5252 * lra.c (hard_regs_spilled_into): Define.
5253 (init_reg_info): Init hard_regs_spilled_into.
5254 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5255 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5256 (process_bb_lives): Handle hard_regs_spilled_into.
5257 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5258 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5260 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5262 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5263 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5264 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5265 * config/rs6000/aix72.h: New file.
5267 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5269 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5270 instead of warning_at with conditional singular and plural messages
5274 * stor-layout.c (update_alignment_for_field): For
5275 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5276 && !DECL_PACKED (field), do the alignment update, just use
5277 only desired_align instead of MAX (type_align, desired_align)
5279 (place_field): Don't do known_align < desired_align handling
5280 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5281 is non-NULL, instead do it after rli->prev_field handling and
5282 only if not within a bitfield word. For DECL_PACKED (field)
5283 use type_align of BITS_PER_UNIT.
5285 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5287 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5288 superfluous parentheses and trailing spaces.
5290 2018-02-28 Richard Biener <rguenther@suse.de>
5292 PR tree-optimization/84584
5293 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5294 SCoPs with fake exit edge.
5296 2018-02-28 Martin Liska <mliska@suse.cz>
5299 * timevar.c (timer::print): Fix format to properly print 100%
5302 2018-02-28 Richard Biener <rguenther@suse.de>
5305 * genmatch.c (capture_info::walk_match): Do not mark
5306 captured expressions without operands as expr_p given
5307 they act more like predicates and should be subject to
5308 "lost tail" side-effect preserving.
5310 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5312 PR rtl-optimization/81611
5313 * auto-inc-dec.c (attempt_change): Move dead note from
5314 mem_insn if it's the next use of regno
5315 (find_address): Take address use of reg holding
5316 non-incremented value. Add parm to limit search to the named
5318 (merge_in_block): Attempt to use a mem insn that is the next
5319 use of the original regno.
5321 2018-02-27 Martin Sebor <msebor@redhat.com>
5324 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5325 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5327 2018-02-27 Martin Sebor <msebor@redhat.com>
5329 PR translation/84207
5330 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5331 n argument to unsigned HOST_WIDE_INT.
5332 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5333 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5334 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5335 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5337 2018-02-27 Richard Biener <rguenther@suse.de>
5339 PR tree-optimization/84512
5340 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5341 Do not use the estimate returned from record_stmt_cost for
5342 the scalar iteration cost but sum properly using add_stmt_cost.
5344 2018-02-27 Richard Biener <rguenther@suse.de>
5346 PR tree-optimization/84466
5347 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5348 Adjust last change to less strictly validate use operands.
5350 2018-02-27 Martin Liska <mliska@suse.cz>
5352 PR gcov-profile/84548
5353 * gcov.c (process_file): Allow partial overlap and consider it
5354 also as group functions.
5355 (output_lines): Properly calculate range of lines for a group.
5357 2018-02-27 Martin Liska <mliska@suse.cz>
5359 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5360 'ggc' suffixes. Change first column width.
5361 (timer::print): Fix formatting of the column.
5363 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5365 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5366 preserve inline entry blocks for the sake of debug inline
5367 entry point markers alone.
5368 (remove_unused_locals): Suggest in comments a better place to
5369 force the preservation of inline entry blocks that are
5370 otherwise unused, but do not preserve them.
5372 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5374 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5376 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5379 * config/i386/constraints.md (Bs): Replace
5380 ix86_indirect_branch_register with
5381 TARGET_INDIRECT_BRANCH_REGISTER.
5383 * config/i386/i386.md (indirect_jump): Likewise.
5384 (tablejump): Likewise.
5385 (*sibcall_memory): Likewise.
5386 (*sibcall_value_memory): Likewise.
5387 Peepholes of indirect call and jump via memory: Likewise.
5388 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5389 (*sibcall_value_GOT_32): Likewise.
5390 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5391 (GOT_memory_operand): Likewise.
5392 (call_insn_operand): Likewise.
5393 (sibcall_insn_operand): Likewise.
5394 (GOT32_symbol_operand): Likewise.
5395 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5397 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5399 PR rtl-optimization/83496
5400 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5401 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5402 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5403 redundant insn, if any.
5404 (relax_delay_slots): Likewise.
5405 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5407 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5409 PR tree-optimization/83965
5410 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5411 that grouped statements are part of a reduction chain. Return
5412 true if the statement is not marked as a reduction itself but
5414 (vect_recog_dot_prod_pattern): Don't check whether the statement
5415 is part of a group here.
5416 (vect_recog_sad_pattern): Likewise.
5417 (vect_recog_widen_sum_pattern): Likewise.
5419 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5422 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5424 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5427 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5429 (ix86_output_indirect_function_return): New prototype.
5430 (ix86_split_simple_return_pop_internal): Likewise.
5431 * config/i386/i386.c (indirect_return_via_cx): New.
5432 (indirect_return_via_cx_bnd): Likewise.
5433 (indirect_thunk_name): Handle return va CX_REG.
5434 (output_indirect_thunk_function): Create alias for
5435 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5436 (ix86_output_indirect_jmp): Remove the bool argument.
5437 (ix86_output_indirect_function_return): New function.
5438 (ix86_split_simple_return_pop_internal): Likewise.
5439 * config/i386/i386.md (*indirect_jump): Don't pass false
5440 to ix86_output_indirect_jmp.
5441 (*tablejump_1): Likewise.
5442 (simple_return_pop_internal): Change it to define_insn_and_split.
5443 Call ix86_split_simple_return_pop_internal to split it for
5445 (simple_return_indirect_internal): Call
5446 ix86_output_indirect_function_return instead of
5447 ix86_output_indirect_jmp.
5449 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5452 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5453 memset and value initialization afterwards.
5455 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5457 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5459 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5462 * common/config/aarch64/aarch64-common.c
5463 (aarch_option_optimization_table[]): Switch
5464 off fomit-frame-pointer
5466 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5467 Chung-Ju Wu <jasonwucj@gmail.com>
5469 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5471 (store_multiple): Ditto.
5473 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5475 * config.gcc: Add --with-cpu support for nds32 target.
5476 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5477 * config/nds32/nds32.opt: Add -mcpu= option.
5479 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5481 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5482 isel=yes): Warn for these deprecated options.
5484 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5486 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5487 ISA_2_5_MASKS_EMBEDDED.
5489 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5491 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5492 p->max as pointers rather than using iterative_hash_expr.
5494 2018-02-23 Carl Love <cel@us.ibm.com>
5496 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5497 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5499 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5500 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5501 P8V_BUILTIN_VEC_VUNSIGNED2.
5503 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5506 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5507 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5509 * lra-constraints.c (curr_insn_transform): Set up
5510 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5512 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5513 * lra-eliminations.c (spill_pseudos): Ditto.
5514 (process_insn_for_elimination): Ditto.
5515 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5516 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5517 LRA_NON_CLOBBERED_ALT.
5519 2018-02-22 Martin Sebor <msebor@redhat.com>
5521 PR tree-optimization/84480
5522 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5523 to maybe_diag_stxncpy_trunc. Call it.
5524 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5525 from gimple_fold_builtin_strcpy. Print inlining stack.
5526 (handle_builtin_stxncpy): Print inlining stack.
5527 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5529 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5532 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5533 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5534 and -fcheck-pointer-bounds are used together.
5535 (indirect_thunk_prefix): New enum.
5536 (indirect_thunk_need_prefix): New function.
5537 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5538 "_nt" instead of "_bnd" for NOTRACK prefix.
5539 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5540 (output_indirect_thunk_function): Likewise.
5542 (ix86_code_end): Update output_indirect_thunk_function calls.
5543 (ix86_output_indirect_branch_via_reg): Replace
5544 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5545 (ix86_output_indirect_branch_via_push): Likewise.
5546 (ix86_output_function_return): Likewise.
5547 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5548 incompatible with -fcf-protection=branch and
5549 -fcheck-pointer-bounds.
5551 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5554 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5555 Change gcc_assert call to output_operand_lossage.
5557 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5559 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5561 2018-02-22 DJ Delorie <dj@redhat.com>
5562 Sebastian Perta <sebastian.perta@renesas.com>
5563 Oleg Endo <olegendo@gcc.gnu.org>
5565 * config/rx/rx.c (rx_rtx_costs): New function.
5566 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5568 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5570 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5572 2018-02-22 Martin Liska <mliska@suse.cz>
5575 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5576 Add "native" as a possible value.
5578 2018-02-22 Martin Liska <mliska@suse.cz>
5581 * config/i386/i386.c (ix86_option_override_internal):
5582 Add "native" as a possible value for -march and -mtune.
5584 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5587 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5588 to all type variants.
5590 PR tree-optimization/84503
5591 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5592 width as info->bitpos + info->bitsize - start.
5593 (merged_store_group::merge_overlapping): Simplify width computation.
5594 (check_no_overlap): New function.
5595 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5596 start + width and last_order of the group, fail if check_no_overlap
5598 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5599 to group if check_no_overlap fails.
5601 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5603 * config/rs6000/altivec.md: Delete contraint arguments to
5604 define_expand, define_split, and define_peephole2, and in
5605 define_insn_and_split if always unused.
5606 * config/rs6000/darwin.md: Ditto.
5607 * config/rs6000/dfp.md: Ditto.
5608 * config/rs6000/rs6000.md: Ditto.
5609 * config/rs6000/sync.md: Ditto.
5610 * config/rs6000/vector.md: Ditto.
5611 * config/rs6000/vsx.md: Ditto.
5613 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5615 * config/rs6000/altivec.md: Write output control strings as braced
5616 blocks instead of double-quoted strings.
5617 * config/rs6000/darwin.md: Ditto.
5618 * config/rs6000/rs6000.md: Ditto.
5619 * config/rs6000/vector.md: Ditto.
5620 * config/rs6000/vsx.md: Ditto.
5622 2018-02-21 Jason Merrill <jason@redhat.com>
5624 PR c++/84314 - ICE with templates and fastcall attribute.
5625 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5627 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5629 * ipa-cp.c (determine_versionability): Fix comment typos.
5631 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5634 * ipa-cp.c (determine_versionability): Do not version functions caling
5637 2018-02-21 Martin Liska <mliska@suse.cz>
5640 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5641 Add "native" as a possible value.
5642 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5643 the macro when native cpu detection is available.
5645 2018-02-21 Martin Liska <mliska@suse.cz>
5648 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5649 Add "native" as a possible value.
5650 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5651 when native cpu detection is available.
5653 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5654 Martin Sebor <msebor@redhat.com>
5656 PR tree-optimization/84478
5657 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5659 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5660 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5661 support which is conservatively correct, for 2 only stay conservative
5662 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5663 argument to the 2 argument get_range_strlen, adjust 6 arg
5664 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5666 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5667 (gimple_fold_builtin_strlen): Pass true as last argument to
5670 2018-02-20 Martin Sebor <msebor@redhat.com>
5673 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5674 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5675 (builtin_memref::builtin_memref): Factor out parts into
5676 set_base_and_offset and call it.
5678 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5681 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5682 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5683 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5684 search at the associated MODE_INT.
5686 2018-02-20 Jeff Law <law@redhat.com>
5689 PR tree-optimization/81592
5691 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5692 for range data rather than using global data.
5693 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5694 range data rather than using global data.
5695 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5696 pass it to children as needed.
5697 (struct directive::fmtresult): Similarly.
5698 (struct directive::set_width): Similarly.
5699 (struct directive::set_precision): Similarly.
5700 (format_integer, format_directive, parse_directive): Similarly.
5701 (format_none): Accept unnamed vr_values parameter.
5702 (format_percent, format_floating, format_character): Similarly.
5703 (format_string, format_plain): Similarly.
5704 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5705 the EVRP range analyzer for range data rather than using global data.
5706 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5707 gimple-ssa-evrp-analyze.h
5708 (class sprintf_dom_walker): Add after_dom_children member function.
5709 Add evrp_range_analyzer member.
5710 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5711 range analyzer as needed.
5712 (sprintf_dom_walker::after_dom_children): New member function.
5713 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5715 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5716 (evrp_range_analyzer::pop_to_marker): Likewise.
5718 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5720 PR tree-optimization/84419
5721 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5722 with the required type if its current type is compatible but
5725 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5728 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5729 after vectorization.
5731 2018-02-20 Martin Liska <mliska@suse.cz>
5734 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5735 possible values if we don't have a hint.
5737 2018-02-20 Martin Liska <mliska@suse.cz>
5741 * final.c (shorten_branches): Build align_tab array with one
5743 * opts.c (finish_options): Add alignment option limit check.
5744 (MAX_CODE_ALIGN): Likewise.
5745 (MAX_CODE_ALIGN_VALUE): Likewise.
5746 * doc/invoke.texi: Document maximum allowed option value for
5747 all -falign-* options.
5749 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5752 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5753 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5754 * var-tracking.c (emit_note_insn_var_location): Remove all references
5755 to NOTE_INSN_CALL_ARG_LOCATION.
5756 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5757 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5758 Use copy_rtx_if_shared.
5759 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5760 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5761 (dwarf2out_var_location): Remove handling of
5762 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5764 * final.c (final_scan_insn): Remove all references to
5765 NOTE_INSN_CALL_ARG_LOCATION.
5766 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5767 before dumping final insns.
5768 * except.c (emit_note_eh_region_end): Remove all references to
5769 NOTE_INSN_CALL_ARG_LOCATION.
5770 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5771 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5772 * config/arc/arc.c (hwloop_optimize): Likewise.
5773 * config/arm/arm.c (create_fix_barrier): Likewise.
5774 * config/s390/s390.c (s390_chunkify_start): Likewise.
5775 * config/sh/sh.c (find_barrier): Likewise.
5776 * config/i386/i386.c (rest_of_insert_endbranch,
5777 ix86_seh_fixup_eh_fallthru): Likewise.
5778 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5779 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5780 * config/frv/frv.c (frv_function_prologue): Likewise.
5781 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5783 (note_outside_basic_block_p): Remove all references to
5784 NOTE_INSN_CALL_ARG_LOCATION.
5785 * gengtype.c (adjust_field_rtx_def): Likewise.
5786 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5788 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5789 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5792 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5795 PR tree-optimization/84452
5796 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5797 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5800 2018-02-19 Martin Liska <mliska@suse.cz>
5803 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5805 2018-02-19 Martin Liska <mliska@suse.cz>
5806 Richard Sandiford <richard.sandiford@linaro.org>
5808 PR tree-optimization/82491
5809 * gimple-fold.c (get_base_constructor): Make earlier bail out
5812 2018-02-19 Carl Love <cel@us.ibm.com>
5814 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5815 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5817 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5818 P8V_BUILTIN_VEC_NEG.
5820 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5822 * config/rl78/rl78.md (movdf): New define expand.
5824 2018-02-19 Martin Liska <mliska@suse.cz>
5827 * doc/invoke.texi: Fix typo.
5828 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5830 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5832 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5833 handle rs6000_single_float and rs6000_double_float specially for
5836 2018-02-16 Jeff Law <law@redhat.com>
5838 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5841 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5844 * ipa-inline.c (inline_small_functions): Fix a typo.
5846 2018-02-16 Nathan Sidwell <nathan@acm.org>
5848 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5850 2018-02-16 Carl Love <cel@us.ibm.com>
5852 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5853 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5854 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5855 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5856 expansion to P8V_BUILTIN_VEC_FLOAT2.
5858 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5860 PR rtl-optimization/70023
5861 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5862 src_regno into account.
5864 2018-02-16 Carl Love <cel@us.ibm.com>
5866 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5867 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5868 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5869 * config/rs6000/rs6000.c: Remove case statements for
5870 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5871 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5872 and P9V_BUILTIN_VEC_VINSERT4B.
5873 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5874 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5875 * config/rs6000/vsx.md:
5876 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5879 2018-02-16 Carl Love <cel@us.ibm.com>
5881 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5883 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5885 * config/rs6000/rs6000-c.c: Add the definitions for
5886 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5887 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5888 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5889 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5890 definition for insert4b and define insn *insert3b_internal.
5891 * doc/extend.texi: Add documentation for vec_extract4b.
5893 2018-02-16 Nathan Sidwell <nathan@acm.org>
5895 * doc/extend.texi (Backwards Compatibility): Mention friend
5896 injection. Note for-scope is deprecated.
5897 * doc/invoke.texi (-ffriend-injection): Deprecate.
5899 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5901 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5902 that moved to I2, also allow destinations that are a paradoxical
5903 subreg (instead of a normal reg).
5905 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5908 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5911 2018-02-16 Richard Biener <rguenther@suse.de>
5913 PR tree-optimization/84037
5914 PR tree-optimization/84016
5916 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5917 Adjust vec_construct for the fact we need additional higher latency
5918 128bit inserts for AVX256 and AVX512 vector builds.
5919 (ix86_add_stmt_cost): Scale vector construction cost for
5922 2018-02-16 Richard Biener <rguenther@suse.de>
5924 PR tree-optimization/84417
5925 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5926 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5927 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5929 2018-02-16 Martin Liska <mliska@suse.cz>
5932 * internal-fn.def (ASAN_CHECK): Set proper flags.
5933 (ASAN_MARK): Likewise.
5935 2018-02-16 Julia Koval <julia.koval@intel.com>
5937 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5938 from PTA_CANNONLAKE.
5940 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5943 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5944 Use ++iter rather than iter++ for std::list iterators.
5945 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5946 defer deleting them until all nodes in the forest are processed. Do
5947 free even leaf nodes. Change to_process into auto_vec.
5950 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5951 * vec.h (vec_default_construct): Use memset instead of placement new
5952 if BROKEN_VALUE_INITIALIZATION is defined.
5953 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5954 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5957 PR rtl-optimization/83723
5958 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5959 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5960 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5962 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5964 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5966 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5968 PR rtl-optimization/81443
5969 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5970 from inner REGs to paradoxical SUBREGs.
5972 2018-02-16 Richard Biener <rguenther@suse.de>
5974 PR tree-optimization/84399
5975 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5976 For operands we can analyze at their definition make sure we can
5977 analyze them at each use as well.
5979 2018-02-16 Richard Biener <rguenther@suse.de>
5981 PR tree-optimization/84190
5982 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5983 volatile accesses if the decl isn't volatile.
5985 2018-02-15 Jason Merrill <jason@redhat.com>
5987 PR c++/84314 - ICE with templates and fastcall attribute.
5988 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5989 TYPE_CANONICAL on an existing type.
5991 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5993 PR tree-optimization/84383
5994 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5995 dstoff nor call operand_equal_p if dstbase is NULL.
5997 PR tree-optimization/84334
5998 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5999 also a CONSTANT_CLASS_P, punt.
6001 2018-02-14 Jim Wilson <jimw@sifive.com>
6003 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6004 first SMALL_OPERAND check. New local min_second_step. Move assert
6005 to where locals are set. Add TARGET_RVC support.
6006 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6008 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6010 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6012 2018-02-14 Martin Sebor <msebor@redhat.com>
6014 PR tree-optimization/83698
6015 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6016 arrays constrain the offset range to their bounds.
6017 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6018 (builtin_access::overlap): Avoid setting the size of overlap if it's
6020 (maybe_diag_overlap): Also consider arrays when deciding what values
6021 of offsets to include in diagnostics.
6023 2018-02-14 Martin Sebor <msebor@redhat.com>
6026 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6027 that correspond to the kind of a declaration.
6029 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6032 * config/pa/pa.md: Load address of PIC label using the linkage table
6033 if the label is nonlocal.
6035 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6037 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6038 warning message if user requests -maltivec=be.
6039 * doc/invoke.texi: Document deprecation of -maltivec=be.
6041 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6044 * config/rs6000/rs6000-c.c: Update definitions for
6045 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6046 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6048 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6051 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6052 add _get_ssp intrinsics. Remove argument from
6053 __builtin_ia32_rdssp[d|q].
6054 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6055 * config/i386/i386-builtin.def: Remove argument from
6056 __builtin_ia32_rdssp[d|q].
6057 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6058 ix86_expand_special_args_builtin for _rdssp[d|q].
6059 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6060 Clear register before usage.
6061 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6062 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6064 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6066 PR tree-optimization/84357
6067 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6068 operand 1 of an ARRAY_REF too.
6070 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6073 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6074 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6076 (set_of_reg): New struct.
6077 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6078 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6079 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6081 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6082 Split into bitclr, bitset, bitinvert patterns if appropriate.
6083 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6084 use rx_fuse_in_memory_bitop.
6085 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6086 to named insn, correct maximum insn length.
6088 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6091 * machmode.def: Define a complex mode for PARTIAL_INT.
6092 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6094 * doc/rtl.texi: Document CSPImode.
6095 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6097 (msp430_hard_regno_nregs_with_padding): Likewise.
6099 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6102 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6104 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6106 PR rtl-optimization/84169
6107 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6108 we generated a parallel as new i3 and we split that to new i2 and i3
6109 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6110 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6111 those to i2, not i1. Partially rewrite this scan code.
6113 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6116 * stor-layout.c (place_field): For variable length fields, adjust
6117 offset_align afterwards not just based on the field's alignment,
6118 but also on the size.
6121 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6122 of exps and logs in the use_exp2 case.
6124 2018-02-13 Jeff Law <law@redhat.com>
6126 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6129 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6132 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6136 * common.opt (gas-loc-support, gas-locview-support): New.
6137 (ginline-points, ginternal-reset-location-views): New.
6138 * doc/invoke.texi: Document them. Use @itemx where intended.
6139 (gvariable-location-views): Adjust.
6140 * target.def (reset_location_view): New.
6141 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6142 (TARGET_RESET_LOCATION_VIEW): New.
6143 * doc/tm.texi: Rebuilt.
6144 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6145 (dwarf2out_default_as_locview_support): New.
6146 (output_asm_line_debug_info): Use option variables.
6147 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6148 (output_loc_list): Likewise.
6149 (add_high_low_attributes): Check option variables.
6150 Don't output entry view attribute in strict mode.
6151 (gen_inlined_subroutine_die): Check option variables.
6152 (dwarf2out_inline_entry): Likewise.
6153 (init_sections_and_labels): Likewise.
6154 (dwarf2out_early_finish): Likewise.
6155 (maybe_reset_location_view): New, from...
6156 (dwarf2out_var_location): ... here. Call it.
6157 * debug.h (dwarf2out_default_as_loc_support): Declare.
6158 (dwarf2out_default_as_locview_support): Declare.
6159 * hooks.c (hook_int_rtx_insn_0): New.
6160 * hooks.h (hook_int_rtx_insn_0): Declare.
6161 * toplev.c (process_options): Take -gas-loc-support and
6162 -gas-locview-support from dwarf2out. Enable
6163 -gvariable-location-views by default only with locview
6164 assembler support. Enable -ginternal-reset-location-views by
6165 default only if the target defines the corresponding hook.
6166 Enable -ginline-points by default if location views are
6167 enabled; force it disabled if statement frontiers are
6169 * tree-inline.c (expand_call_inline): Check option variables.
6170 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6172 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6174 PR tree-optimization/84321
6175 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6176 handling. Also check whether the anti-range contains any values
6177 that satisfy the mask; switch to a VR_RANGE if not.
6179 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6182 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6184 2018-02-13 Martin Jambor <mjambor@suse.cz>
6187 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6188 of call statements, also set location of a load to a temporary.
6190 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6192 * config/rl78/rl78.c (add_vector_labels): New function.
6193 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6194 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6195 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6196 which checks that no arguments are passed.
6197 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6198 * doc/extend.texi: Documentation for the new attribute.
6200 2018-02-13 Andreas Schwab <schwab@suse.de>
6202 * config/riscv/linux.h (CPP_SPEC): Define.
6204 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6207 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6208 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6209 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6210 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6211 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6212 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6213 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6214 temporarily for AES and PCLMUL builtins.
6216 PR tree-optimization/84339
6217 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6218 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6222 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6223 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6224 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6226 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6228 * omp-simd-clone.h: New file.
6229 * omp-simd-clone.c: Include omp-simd-clone.h.
6230 (expand_simd_clones): No longer static.
6231 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6232 cgraph.h and omp-simd-clone.h.
6233 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6234 (vect_recog_widen_shift_pattern): Formatting fix.
6235 (vect_pattern_recog_1): Don't check optab for calls.
6238 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6239 operands[2] into a REG before using gen_lowpart on it.
6241 2018-02-12 Jeff Law <law@redhat.com>
6244 * config/sh/sh.c (find_barrier): Consider a sibling call
6247 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6248 successfully back substituting a reg.
6250 2018-02-12 Richard Biener <rguenther@suse.de>
6252 PR tree-optimization/84037
6253 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6254 parameter, move visited init to caller.
6255 (vect_slp_analyze_operations): Separate cost from validity
6256 check, initialize visited once for all instances.
6257 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6259 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6261 (vect_model_promotion_demotion_cost): Likewise.
6262 (vectorizable_bswap): Guard cost modeling with !slp_node
6263 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6265 (vectorizable_call): Likewise.
6266 (vectorizable_conversion): Likewise.
6267 (vectorizable_assignment): Likewise.
6268 (vectorizable_shift): Likewise.
6269 (vectorizable_operation): Likewise.
6270 (vectorizable_store): Likewise.
6271 (vectorizable_load): Likewise.
6272 (vectorizable_condition): Likewise.
6273 (vectorizable_comparison): Likewise.
6275 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6278 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6279 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6280 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6282 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6285 * params.def (inline-min-speedup): Increase from 8 to 15.
6286 (max-inline-insns-auto): Decrease from 40 to 30.
6287 * ipa-split.c (consider_split): Add some buffer for function to
6288 be considered inlining candidate.
6289 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6292 2018-02-12 Richard Biener <rguenther@suse.de>
6294 PR tree-optimization/84037
6295 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6296 matched stmts if we cannot swap the non-matched ones.
6298 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6300 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6301 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6302 _mm_maskz_scalef_round_ss): New intrinsics.
6303 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6304 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6305 __builtin_ia32_scalefss_round): Remove.
6306 (__builtin_ia32_scalefsd_mask_round,
6307 __builtin_ia32_scalefss_mask_round): New intrinsics.
6308 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6309 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6310 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6311 "<round_constraint>")): Changed to ...
6312 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6313 "<round_scalar_constraint>")): ... this.
6314 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6315 %0, %1, %2<round_op3>}"): Changed to ...
6316 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6317 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6318 %2<round_scalar_mask_op3>}"): ... this.
6319 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6321 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6323 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6324 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6325 (_mm_maskz_sqrt_round_ss): New intrinsics.
6326 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6327 (__builtin_ia32_sqrtsd_mask_round)
6328 (__builtin_ia32_sqrtss_mask_round): New builtins.
6329 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6330 (__builtin_ia32_sqrtss_round): Remove.
6331 (__builtin_ia32_sqrtsd_mask_round)
6332 (__builtin_ia32_sqrtss_mask_round): New builtins.
6333 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6334 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6335 ((match_operand:VF_128 1 "vector_operand"
6336 "xBm,<round_constraint>")): Changed to ...
6337 ((match_operand:VF_128 1 "vector_operand"
6338 "xBm,<round_scalar_constraint>")): ... this.
6339 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6340 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6341 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6342 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6343 %<iptr>1<round_scalar_mask_op3>}): ... this.
6344 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6345 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6347 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6350 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6351 Cast vec_cmpeq result to correct type.
6352 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6353 Cast vec_cmpgt result to correct type.
6355 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6357 * final.c (final_scan_insn_1): Renamed from...
6358 (final_scan_insn): ... this. New wrapper, to recover
6359 seen from the outermost call in recursive ones.
6360 * config/sparc/sparc.c (output_return): Drop seen from call.
6361 (output_sibcall): Likewise.
6362 * config/visium/visium.c (output_branch): Likewise.
6364 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6366 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6369 2018-02-10 Alan Modra <amodra@gmail.com>
6372 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6373 Specify LR as an input.
6375 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6378 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6379 remove_member_access_dummy_vars): New functions.
6380 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6381 lower_omp_1, execute_lower_omp): Use them.
6383 PR rtl-optimization/84308
6384 * shrink-wrap.c (spread_components): Release todo vector.
6386 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6388 PR rtl-optimization/57193
6389 * ira-color.c (struct allocno_color_data): Add member
6390 conflict_allocno_hard_prefs.
6391 (update_conflict_allocno_hard_prefs): New.
6392 (bucket_allocno_compare_func): Add a preference based on
6393 conflict_allocno_hard_prefs.
6394 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6395 (color_allocnos): Remove a dead code. Initiate
6396 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6398 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6401 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6402 constraint from =wa to wa. Avoid a subreg on the output operand,
6403 instead use a pseudo and subreg it in a move.
6404 (p9_xxbrd_<mode>): Changed to ...
6405 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6406 (p9_xxbrd_v2df): New expander.
6407 (p9_xxbrw_<mode>): Changed to ...
6408 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6409 (p9_xxbrw_v4sf): New expander.
6411 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6413 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6415 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6418 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6419 multiply in 32-bit mode.
6420 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6421 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6424 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6426 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6427 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6428 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6429 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6431 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6434 * dwarf2out.c (is_trivial_indirect_ref): New function.
6435 (dwarf2out_late_global_decl): Do not generate a location
6436 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6437 and that are not defined in the current unit.
6439 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6441 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6442 instead of a libcall for UNORDERED.
6444 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6447 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6448 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6450 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6453 * config/s390/s390.c (s390_set_current_function): Invoke
6454 s390_indirect_branch_settings also if fndecl didn't change.
6456 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6458 * config/rs6000/rs6000.md (blockage): Set length to zero.
6460 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6462 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6464 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6467 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6468 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6472 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6473 PARALLEL incoming that failed vt_get_decl_and_offset check.
6476 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6477 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6479 (get_variable_section): For decls in named .bss* sections pass true as
6480 second argument to bss_initializer_p.
6482 2018-02-09 Marek Polacek <polacek@redhat.com>
6483 Jakub Jelinek <jakub@redhat.com>
6486 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6487 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6488 Sync some changes from cxx_fold_indirect_ref.
6490 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6492 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6494 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6495 (BLOCK_INLINE_ENTRY_LABEL): New.
6496 (dwarf2out_var_location): Disregard inline entry markers.
6497 (inline_entry_data): New struct.
6498 (inline_entry_data_hasher): New hashtable type.
6499 (inline_entry_data_hasher::hash): New.
6500 (inline_entry_data_hasher::equal): New.
6501 (inline_entry_data_table): New variable.
6502 (add_high_low_attributes): Add DW_AT_entry_pc and
6503 DW_AT_GNU_entry_view attributes if a pending entry is found
6504 in inline_entry_data_table. Add old entry_pc attribute only
6505 if debug nonbinding markers are disabled.
6506 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6507 markers are enabled.
6508 (block_within_block_p, dwarf2out_inline_entry): New.
6509 (dwarf2out_finish): Check that no entries remained in
6510 inline_entry_data_table.
6511 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6512 (final_scan_insn, notice_source_line): Likewise.
6513 (rest_of_clean_state): Skip inline entry markers.
6514 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6516 * gimple.c (gimple_build_debug_inline_entry): New.
6517 * gimple.h (enum gimple_debug_subcode): Add
6518 GIMPLE_DEBUG_INLINE_ENTRY.
6519 (gimple_build_debug_inline_entry): Declare.
6520 (gimple_debug_inline_entry_p): New.
6521 (gimple_debug_nonbind_marker_p): Adjust.
6522 * insn-notes.def (INLINE_ENTRY): New.
6523 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6524 inline entry marker notes.
6525 (print_insn): Likewise.
6526 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6527 (INSN_DEBUG_MARKER_KIND): Likewise.
6528 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6529 * tree-inline.c (expand_call_inline): Build and insert
6530 debug_inline_entry stmt.
6531 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6532 inline entry blocks early, if nonbind markers are enabled.
6533 (dump_scope_block): Dump fragment info.
6534 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6535 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6536 (gimple_build_debug_inline_entry): New.
6537 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6538 Enable/disable inline entry points too.
6539 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6540 (DEBUG_INSN): Describe inline entry markers.
6542 * common.opt (gvariable-location-views): New.
6543 (gvariable-location-views=incompat5): New.
6544 * config.in: Rebuilt.
6545 * configure: Rebuilt.
6546 * configure.ac: Test assembler for view support.
6547 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6548 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6549 * dwarf2out.c (var_loc_view): New typedef.
6550 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6551 (dwarf2out_locviews_in_attribute): New.
6552 (dwarf2out_locviews_in_loclist): New.
6553 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6554 (enum dw_line_info_opcode): Add LI_adv_address.
6555 (struct dw_line_info_table): Add view.
6556 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6557 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6558 (zero_view_p): New variable.
6559 (ZERO_VIEW_P): New macro.
6560 (output_asm_line_debug_info): New.
6561 (struct var_loc_node): Add view.
6562 (add_AT_view_list, AT_loc_list): New.
6563 (add_var_loc_to_decl): Add view param. Test it against last.
6564 (new_loc_list): Add view params. Record them.
6565 (AT_loc_list_ptr): Handle loc and view lists.
6566 (view_list_to_loc_list_val_node): New.
6567 (print_dw_val): Handle dw_val_class_view_list.
6568 (size_of_die): Likewise.
6569 (value_format): Likewise.
6570 (loc_list_has_views): New.
6571 (gen_llsym): Set vl_symbol too.
6572 (maybe_gen_llsym, skip_loc_list_entry): New.
6573 (dwarf2out_maybe_output_loclist_view_pair): New.
6574 (output_loc_list): Output view list or entries too.
6575 (output_view_list_offset): New.
6576 (output_die): Handle dw_val_class_view_list.
6577 (output_dwarf_version): New.
6578 (output_compilation_unit_header): Use it.
6579 (output_skeleton_debug_sections): Likewise.
6580 (output_rnglists, output_line_info): Likewise.
6581 (output_pubnames, output_aranges): Update version comments.
6582 (output_one_line_info_table): Output view numbers in asm comments.
6583 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6584 Call maybe_gen_llsym.
6585 (loc_list_from_tree_1): Adjust.
6586 (add_AT_location_description): Create view list attribute if
6587 needed, check it's absent otherwise.
6588 (convert_cfa_to_fb_loc_list): Adjust.
6589 (maybe_emit_file): Call output_asm_line_debug_info for test.
6590 (dwarf2out_var_location): Reset views as needed. Precompute
6591 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6592 attribute. Set view.
6593 (new_line_info_table): Reset next view.
6594 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6595 (dwarf2out_source_line): Likewise. Output view resets and labels to
6596 the assembler, or select appropriate line info opcodes.
6597 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6598 (optimize_string_length): Catch it. Adjust.
6599 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6600 dw_val_class_view_list, and remove it if no longer needed.
6601 (hash_loc_list): Hash view numbers.
6602 (loc_list_hasher::equal): Compare them.
6603 (optimize_location_lists): Check whether a view list symbol is
6604 needed, and whether the locview attribute is present, and
6605 whether they match. Remove the locview attribute if no longer
6607 (index_location_lists): Call skip_loc_list_entry for test.
6608 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6609 Use output_dwarf_version.
6610 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6611 (struct dw_val_node): Add val_view_list.
6612 * final.c (SEEN_NEXT_VIEW): New.
6613 (set_next_view_needed): New.
6614 (clear_next_view_needed): New.
6615 (maybe_output_next_view): New.
6616 (final_start_function): Rename to...
6617 (final_start_function_1): ... this. Take pointer to FIRST,
6618 add SEEN parameter. Emit param bindings in the initial view.
6619 (final_start_function): Reintroduce SEEN-less interface.
6620 (final): Rename to...
6621 (final_1): ... this. Take SEEN parameter. Output final pending
6622 next view at the end.
6623 (final): Reintroduce seen-less interface.
6624 (final_scan_insn): Output pending next view before switching
6625 sections or ending a block. Mark the next view as needed when
6626 outputting variable locations. Notify debug backend of section
6627 changes, and of location view changes.
6628 (rest_of_handle_final): Adjust.
6629 * toplev.c (process_options): Autodetect value for debug variable
6630 location views option. Warn on incompat5 without -gdwarf-5.
6631 * doc/invoke.texi (gvariable-location-views): New.
6632 (gvariable-location-views=incompat5): New.
6633 (gno-variable-location-views): New.
6635 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6637 PR tree-optimization/84136
6638 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6639 that the result of find_edge is non-NULL.
6641 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6644 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6645 storing integer register in SImode. Fix cost of 256 and 512
6646 byte aligned SSE register store.
6648 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6650 * config/i386/i386.c (ix86_multiplication_cost): Fix
6651 multiplication cost for TARGET_AVX512DQ.
6653 2018-02-08 Marek Polacek <polacek@redhat.com>
6655 PR tree-optimization/84238
6656 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6659 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6661 PR tree-optimization/84265
6662 * tree-vect-stmts.c (vectorizable_store): Don't treat
6663 VMAT_CONTIGUOUS accesses as grouped.
6664 (vectorizable_load): Likewise.
6666 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6668 PR tree-optimization/81635
6669 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6670 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6671 (test_round_for_mask): New functions.
6672 (wide_int_cc_tests): Call test_round_for_mask.
6673 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6674 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6675 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6676 range returned by get_range_info.
6678 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6681 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6682 * symtab.c: Include builtins.h
6683 (symtab_node::output_to_lto_symbol_table_p): Move here
6684 from lto-streamer-out.c:output_symbol_p.
6685 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6686 (output_symbol_p): Move all logic to symtab.c
6687 (produce_symtab): Update.
6689 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6691 * config/s390/s390-opts.h (enum indirect_branch): Define.
6692 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6693 (s390_indirect_branch_via_thunk)
6694 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6695 (enum s390_indirect_branch_type): Define.
6696 * config/s390/s390.c (struct s390_frame_layout, struct
6697 machine_function): Remove.
6698 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6699 (indirect_branch_table_label_no, indirect_branch_table_name):
6701 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6702 (enum s390_indirect_branch_option): Define.
6703 (s390_return_addr_from_memory): New function.
6704 (s390_handle_string_attribute): New function.
6705 (s390_attribute_table): Add new attribute handler.
6706 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6707 (s390_indirect_branch_via_thunk): New function.
6708 (s390_indirect_branch_via_inline_thunk): New function.
6709 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6710 sibling call optimization for non z10 compiles.
6711 (s390_emit_call): Force indirect branch target to be a single
6712 register. Add r1 clobber for non-z10 compiles.
6713 (s390_emit_epilogue): Emit return jump via return_use expander.
6714 (s390_reorg): Handle JUMP_INSNs as execute targets.
6715 (s390_option_override_internal): Perform validity checks for the
6716 new command line options.
6717 (s390_indirect_branch_attrvalue): New function.
6718 (s390_indirect_branch_settings): New function.
6719 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6720 (s390_output_indirect_thunk_function): New function.
6721 (s390_code_end): Implement target hook.
6722 (s390_case_values_threshold): Implement target hook.
6723 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6725 * config/s390/s390.h (struct s390_frame_layout)
6726 (struct machine_function): Move here from s390.c.
6727 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6728 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6729 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6730 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6731 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6732 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6733 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6734 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6735 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6736 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6737 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6738 (mnemonic attribute): Add values which aren't recognized
6740 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6741 pattern for branch conversion. Fix mnemonic attribute.
6742 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6743 indirect branch via thunk if requested.
6744 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6745 ("*indirect_jump"): Disable for branch conversion using out of
6747 ("indirect_jump_via_thunk<mode>_z10")
6748 ("indirect_jump_via_thunk<mode>")
6749 ("indirect_jump_via_inlinethunk<mode>_z10")
6750 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6751 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6752 ("casesi_jump_via_inlinethunk<mode>_z10")
6753 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6754 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6755 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6756 ("*indirect2_jump"): Disable for branch conversion.
6757 ("casesi_jump"): Turn into expander and expand patterns for branch
6759 ("return_use"): New expander.
6760 ("*return"): Emit return via thunk and rename it to ...
6761 ("*return<mode>"): ... this one.
6762 * config/s390/s390.opt: Add new options and and enum for the
6765 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6767 * lra-constraints.c (match_reload): Unconditionally use
6768 gen_lowpart_SUBREG, rather than selecting between that
6769 and equivalent gen_rtx_SUBREG code.
6771 2018-02-08 Richard Biener <rguenther@suse.de>
6773 PR tree-optimization/84233
6774 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6775 changed flag instead of boguously re-using phi_inserted.
6777 2018-02-08 Martin Jambor <mjambor@suse.cz>
6779 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6780 static local variables.
6782 2018-02-08 Richard Biener <rguenther@suse.de>
6784 PR tree-optimization/84278
6785 * tree-vect-stmts.c (vectorizable_store): When looking for
6786 smaller vector types to perform grouped strided loads/stores
6787 make sure the mode is supported by the target.
6788 (vectorizable_load): Likewise.
6790 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6792 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6793 Increase LDP/STP opportunities by adding adjacent callee-saves.
6795 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6797 PR rtl-optimization/84068
6798 PR rtl-optimization/83459
6799 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6801 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6803 PR tree-optimization/84224
6804 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6805 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6808 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6811 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6812 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6813 count, remove one USE.
6815 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6817 * doc/install.texi (Configuration): Document the
6818 --with-long-double-format={ibm,ieee} PowerPC configuration
6822 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6823 Convert from define_expand to be define_insn_and_split. Rework
6824 float/double/_Float128 conversions to QI/HI/SImode to work with
6825 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6826 conversions to QI/HImode types did a store and then a load to
6827 truncate the value. For conversions to VSX registers, don't split
6828 the insn, instead emit the code directly. Use the code iterator
6829 any_fix to combine signed and unsigned conversions.
6830 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6831 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6832 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6833 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6834 (fix_<mode>di2_hw): Likewise.
6835 (fixuns_<mode>di2_hw): Likewise.
6836 (fix_<mode>si2_hw): Likewise.
6837 (fixuns_<mode>si2_hw): Likewise.
6838 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6839 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6840 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6841 fix<uns>_trunc<SFDF:mode>si2_p8.
6842 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6844 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6845 (fix<uns>_<mode>_mem): Likewise.
6846 (fctiw<u>z_<mode>_mem): Likewise.
6847 (fix<uns>_<mode>_mem): Likewise.
6848 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6849 the register allocator from doing a direct move to the GPRs to do
6850 a store, and instead use the ISA 3.0 store byte/half-word from
6851 vector register instruction. For IEEE 128-bit floating point,
6852 also optimize stores of 32-bit ints.
6853 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6855 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6857 * genextract.c (push_pathstr_operand): New function to support
6859 (walk_rtx): Call push_pathstr_operand.
6860 (print_path): Support [a-zA-Z].
6862 2018-02-07 Richard Biener <rguenther@suse.de>
6864 PR tree-optimization/84037
6865 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6866 (cse_and_gimplify_to_preheader): Declare.
6867 (vect_get_place_in_interleaving_chain): Likewise.
6868 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6870 (_loop_vec_info::~_loop_vec_info): Delete it.
6871 (cse_and_gimplify_to_preheader): New function.
6872 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6873 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6874 (vectorizable_load): Likewise. For grouped stores always base
6875 the IV on the first element.
6876 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6877 condition before gimplifying.
6879 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6881 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6882 *DIV_EXPR and *MOD_EXPR.
6884 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6887 * config/i386/i386.c (ix86_option_override_internal): Mask out
6888 the CF_SET bit when checking -fcf-protection.
6890 2018-02-07 Tom de Vries <tom@codesourcery.com>
6893 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6896 2018-02-07 Richard Biener <rguenther@suse.de>
6898 PR tree-optimization/84204
6899 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6902 PR tree-optimization/84205
6903 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6904 special-case isl_ast_op_zdiv_r.
6906 PR tree-optimization/84223
6907 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6908 Only add conditions from within the region.
6909 (gather_bbs::after_dom_children): Adjust.
6911 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6914 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6915 * config/avr/avr.md: Only post-reload split REG-REG moves if
6916 either register is GENERAL_REG_P.
6918 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6920 PR tree-optimization/84235
6921 * tree-ssa-scopedtables.c
6922 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6923 if the subtraction is performed in floating point type where NaNs are
6924 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6925 build 1. Formatting fix.
6927 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6930 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6931 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6932 and skip it regardless of bb boundaries. Use CALL_P macro,
6933 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6934 unnecessarily, formatting fix.
6936 2018-02-06 Michael Collison <michael.collison@arm.com>
6938 * config/arm/thumb2.md:
6939 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6940 (*thumb_mov_notscc): Ditto.
6942 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6945 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6946 unsigned_fix, not "s".
6948 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6950 * configure.ac (gcc_fn_eh_frame_ro): New function.
6951 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6952 correct .eh_frame permissions.
6953 * configure: Regenerate.
6955 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6957 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6960 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6962 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6963 Display warning message for -mno-speculate-indirect-jumps.
6965 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6967 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6969 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6971 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6973 PR tree-optimization/84225
6974 * tree-eh.c (find_trapping_overflow): Only call
6975 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6977 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6980 * config/i386/i386.c: Reimplement the check of possible options
6981 -mibt/-mshstk conbination. Change error messages.
6982 * doc/invoke.texi: Fix a typo: remove extra '='.
6984 2018-02-06 Marek Polacek <polacek@redhat.com>
6986 PR tree-optimization/84228
6987 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6989 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6992 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6993 emitted arch directives.
6994 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6995 __ARM_FEATURE_COPROC before changing architectures.
6997 2018-02-06 Richard Biener <rguenther@suse.de>
6999 * config/i386/i386.c (print_reg): Fix typo.
7000 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7002 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7004 * configure: Regenerate.
7006 2018-02-05 Martin Sebor <msebor@redhat.com>
7008 PR tree-optimization/83369
7009 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7012 2018-02-05 Martin Liska <mliska@suse.cz>
7014 * doc/invoke.texi: Cherry-pick upstream r323995.
7016 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7018 * ira.c (ira_init_register_move_cost): Adjust comment.
7020 2018-02-05 Martin Liska <mliska@suse.cz>
7022 PR gcov-profile/84137
7023 * doc/gcov.texi: Fix typo in documentation.
7025 2018-02-05 Martin Liska <mliska@suse.cz>
7027 PR gcov-profile/83879
7028 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7029 using dlopen functionality.
7031 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7033 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7034 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7035 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7036 _mm_maskz_range_round_ss): New intrinsics.
7037 (__builtin_ia32_rangesd128_round)
7038 (__builtin_ia32_rangess128_round): Remove.
7039 (__builtin_ia32_rangesd128_mask_round,
7040 __builtin_ia32_rangess128_mask_round): New builtins.
7041 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7042 __builtin_ia32_rangess128_round): Remove.
7043 (__builtin_ia32_rangesd128_mask_round,
7044 __builtin_ia32_rangess128_mask_round): New builtins.
7045 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7046 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7047 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7048 "<round_saeonly_constraint>")): Changed to ...
7049 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7050 "<round_saeonly_scalar_constraint>")): ... this.
7051 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7052 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7053 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7054 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7055 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7057 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7059 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7061 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7062 Remove all values except native, 8540 and 8548.
7064 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7066 * config/i386/i386.c (ix86_output_function_return): Pass
7067 INVALID_REGNUM, instead of -1, as invalid register number to
7068 indirect_thunk_name and output_indirect_thunk.
7070 2018-02-02 Julia Koval <julia.koval@intel.com>
7072 * config.gcc: Add -march=icelake.
7073 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7074 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7075 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7076 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7077 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7078 (processor_target_table): Add icelake.
7079 (ix86_option_override_internal): Handle new PTAs.
7080 (get_builtin_code_for_version): Handle icelake.
7081 (M_INTEL_COREI7_ICELAKE): New.
7082 (fold_builtin_cpu): Handle icelake.
7083 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7084 * doc/invoke.texi: Add -march=icelake.
7086 2018-02-02 Julia Koval <julia.koval@intel.com>
7088 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7089 to wide_int_bitmask.
7090 * wide-int-bitmask.h: New.
7092 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7095 * config/i386/i386.md: Replace Pmode with word_mode in
7096 builtin_setjmp_setup and builtin_longjmp to support x32.
7098 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7102 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7104 (rs6000_supported_cpu_names): New static variable.
7105 (linux_cpu_translation_table): Likewise.
7106 (elf_platform) <cpu>: Define new static variable and use it.
7107 Translate kernel AT_PLATFORM name to canonical name if needed.
7108 Error if platform name is unknown.
7110 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7113 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7115 2018-02-01 Jeff Law <law@redhat.com>
7118 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7119 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7120 the scratch if RELEASE_VIA_POP is false.
7121 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7122 If we have to save a temporary register, decrement SIZE appropriately.
7123 Pass new arguments to release_scratch_register_on_entry.
7124 (ix86_adjust_stack_and_probe): Likewise.
7125 (ix86_emit_probe_stack_range): Pass new arguments to
7126 release_scratch_register_on_entry.
7128 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7130 PR rtl-optimization/84157
7131 * combine.c (change_zero_ext): Use REG_P predicate in
7132 front of HARD_REGISTER_P predicate.
7134 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7136 * config/avr/avr.c (avr_option_override): Move disabling of
7137 -fdelete-null-pointer-checks to...
7138 * common/config/avr/avr-common.c (avr_option_optimization_table):
7141 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7143 PR tree-optimization/81635
7144 * tree-data-ref.c (split_constant_offset_1): For types that
7145 wrap on overflow, try to use range info to prove that wrapping
7148 2018-02-01 Renlin Li <renlin.li@arm.com>
7151 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7153 (aarch64_register_move_cost): Likewise.
7154 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7156 (REG_CLASS_NAMES): Likewise.
7157 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7158 TAILCALL_ADDR_REGS. Remove IP registers.
7159 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7161 2018-02-01 Richard Biener <rguenther@suse.de>
7163 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7164 for specifying RPO order and allow NULL for that.
7165 * domwalk.c (dom_walker::dom_walker): Likewise.
7166 (dom_walker::walk): Handle NULL RPO order.
7167 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7169 (rewrite_update_dom_walker): Likewise.
7170 (mark_def_dom_walker): Likewise.
7172 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7174 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7175 (aarch64_maybe_expand_sve_subreg_move): Declare.
7176 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7177 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7179 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7180 that are semantically a reverse operation.
7181 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7182 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7183 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7185 (aarch64_can_change_mode_class): For big-endian, forbid changes
7186 between two SVE modes if they have different element sizes.
7188 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7190 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7191 the TImode handling for big-endian targets.
7193 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7195 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7196 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7198 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7199 Remove BSWAP handing for big-endian targets and use the form of
7200 LD1RQ appropariate for the mode.
7202 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7204 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7205 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7208 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7211 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7212 check for operands that need to go through aarch64_sve_reload_be.
7214 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7216 PR tree-optimization/81661
7217 PR tree-optimization/84117
7218 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7219 * tree-eh.c: Include gimplify.h.
7220 (find_trapping_overflow, replace_trapping_overflow,
7221 rewrite_to_non_trapping_overflow): New functions.
7222 * tree-vect-loop.c: Include tree-eh.h.
7223 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7224 * tree-data-ref.c: Include tree-eh.h.
7225 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7227 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7229 PR rtl-optimization/84123
7230 * combine.c (change_zero_ext): Check if hard register satisfies
7231 can_change_dest_mode before calling gen_lowpart_SUBREG.
7233 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7236 * ira.c (ira_init_register_move_cost): Remove assert.
7238 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7240 PR rtl-optimization/84071
7241 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7242 * doc/tm.texi: Regenerate.
7244 2018-01-31 Richard Biener <rguenther@suse.de>
7246 PR tree-optimization/84132
7247 * tree-data-ref.c (analyze_miv_subscript): Properly
7248 check whether evolution_function_is_affine_multivariate_p
7249 before calling gcd_of_steps_may_divide_p.
7251 2018-01-31 Julia Koval <julia.koval@intel.com>
7254 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7255 * config/i386/i386.md (rdpid_rex64) New.
7256 (rdpid): Make 32bit only.
7258 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7261 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7262 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7264 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7267 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7269 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7271 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7273 PR rtl-optimization/84071
7274 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7275 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7277 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7279 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7280 (arc_attribute_table): Add 'aux' attribute.
7281 (arc_in_small_data_p): Consider aux like variables.
7282 (arc_is_aux_reg_p): New function.
7283 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7284 (arc_get_aux_arg): New function.
7285 (prepare_move_operands): Handle aux-register access.
7286 (arc_handle_aux_attribute): New function.
7287 * doc/extend.texi (ARC Variable attributes): Add subsection.
7289 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7291 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7292 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7293 (arc_attribute_table): Add 'uncached' attribute.
7294 (arc_print_operand): Print '.di' flag for uncached memory
7296 (arc_in_small_data_p): Do not consider for small data the uncached
7298 (arc_is_uncached_mem_p): New function.
7299 * config/arc/predicates.md (compact_store_memory_operand): Check
7300 for uncached memory accesses.
7301 (nonvol_nonimm_operand): Likewise.
7302 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7304 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7307 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7308 falign-loops=): Add Optimization flag.
7310 2018-01-30 Jeff Law <law@redhat.com>
7313 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7314 INT_REGISTERS_SAVED. Check it prior to calling
7315 get_scratch_register_on_entry.
7316 (ix86_adjust_stack_and_probe): Similarly.
7317 (ix86_emit_probe_stack_range): Similarly.
7318 (ix86_expand_prologue): Corresponding changes.
7320 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7323 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7324 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7326 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7329 * lra-constraints.c (curr_insn_transform): Process AND in the
7332 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7334 PR rtl-optimization/83986
7335 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7336 dependence against last_pending_memory_flush in addition to
7339 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7341 PR tree-optimization/81611
7342 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7345 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7348 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7351 2018-01-30 Richard Biener <rguenther@suse.de>
7352 Jakub Jelinek <jakub@redhat.com>
7354 PR tree-optimization/84111
7355 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7356 inner loops added during recursion, as they don't have up-to-date
7359 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7362 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7363 (can_inline_edge_by_limits_p): ... here.
7364 (can_early_inline_edge_p, check_callers,
7365 update_caller_keys, update_callee_keys, recursive_inlining,
7366 add_new_edges_to_heap, speculation_useful_p,
7367 inline_small_functions,
7368 inline_small_functions, flatten_function,
7369 inline_to_all_callers_1): Update.
7371 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7373 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7376 2018-01-30 Richard Biener <rguenther@suse.de>
7378 PR tree-optimization/83008
7379 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7380 invariant and constant vector uses in stmts when they need
7383 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7386 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7387 * configure: Regenerate.
7389 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7391 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7393 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7394 Use gen_rtx_REG rather than gen_lowpart.
7396 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7398 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7399 rather than 0 when creating partial subregs.
7401 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7403 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7406 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7409 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7410 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7411 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7412 flags. This restores the settings used before the 2017-07-24.
7413 Turning off pre increment/decrement/modify allows IVOPTS to
7414 optimize DF/SF loops where the index is an int.
7416 2018-01-29 Richard Biener <rguenther@suse.de>
7417 Kelvin Nilsen <kelvin@gcc.gnu.org>
7420 * tree-vect-stmts.c (vectorizable_call): Don't call
7421 targetm.vectorize_builtin_md_vectorized_function if callee is
7424 2018-01-22 Carl Love <cel@us.ibm.com>
7426 * doc/extend.tex: Fix typo in second arg in
7427 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7429 2018-01-29 Richard Biener <rguenther@suse.de>
7431 PR tree-optimization/84086
7432 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7433 (flush_ssaname_freelist): When SSA names were released reset
7434 the SCEV hash table.
7436 2018-01-29 Richard Biener <rguenther@suse.de>
7438 PR tree-optimization/84057
7439 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7440 removed paths when removing edges.
7442 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7444 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7445 -mfunction-return=@var{choice}.
7447 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7450 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7452 (layout::print_source_line): Likewise.
7453 (test_get_line_width_without_trailing_whitespace): Add test cases.
7455 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7458 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7461 2018-01-26 Jim Wilson <jimw@sifive.com>
7463 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7465 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7468 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7470 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7471 and CMP + SUB-immediate -> SUBS.
7473 2018-01-26 Martin Sebor <msebor@redhat.com>
7475 PR tree-optimization/83896
7476 * tree-ssa-strlen.c (get_string_len): Rename...
7477 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7478 Avoid assuming length is constant.
7479 (handle_char_store): Use HOST_WIDE_INT for string length.
7481 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7484 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7485 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7487 2018-01-26 Richard Biener <rguenther@suse.de>
7489 PR rtl-optimization/84003
7490 * dse.c (record_store): Only record redundant stores when
7491 the earlier store aliases at least all accesses the later one does.
7493 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7495 PR rtl-optimization/83985
7496 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7497 REG_CFA_RESTORE insns.
7498 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7499 REG_CFA_RESTORE insns here.
7502 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7503 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7505 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7507 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7508 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7509 (arc_init): Likewise.
7510 (arc_override_options): Likewise.
7511 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7513 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7515 * config/arc/arc.h (TARGET_DBNZ): Define.
7516 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7517 properly set the tune attribute.
7518 (dbnz): Use TARGET_DBNZ guard.
7519 * config/arc/arc.opt (mtune): Add core3 option.
7521 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7523 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7524 recognize new pic like addresses.
7525 (arc_delegitimize_address): Clean up.
7527 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7529 * config/arc/arc-arches.def: Option mrf16 valid for all
7531 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7532 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7533 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7534 * config/arc/arc-tables.opt: Regenerate.
7535 * config/arc/arc.c (arc_conditional_register_usage): Handle
7536 reduced register file case.
7537 (arc_file_start): Set must have build attributes.
7538 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7540 * config/arc/arc.opt (mrf16): Add new option.
7541 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7542 * config/arc/genmultilib.awk: Handle new mrf16 option.
7543 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7544 * config/arc/t-multilib: Regenerate.
7545 * doc/invoke.texi (ARC Options): Document mrf16 option.
7547 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7549 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7550 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7551 (arc_attribute_table): Add 'secure_call' attribute.
7552 (arc_print_operand): Print secure call operand.
7553 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7555 (arc_is_secure_call_p): New function. * config/arc/arc.md
7556 (call_i): Add support for sjli instruction.
7557 (call_value_i): Likewise.
7558 * config/arc/constraints.md (Csc): New constraint.
7560 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7561 John Eric Martin <John.Martin@emmicro-us.com>
7563 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7564 * config/arc/arc.c (_arc_jli_section): New struct.
7565 (arc_jli_section): New type.
7566 (rc_jli_sections): New static variable.
7567 (arc_handle_jli_attribute): New function.
7568 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7569 (arc_file_end): New function.
7570 (TARGET_ASM_FILE_END): Define.
7571 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7572 (arc_add_jli_section): New function.
7573 (jli_call_scan): Likewise.
7574 (arc_reorg): Call jli_call_scan.
7575 (arc_output_addsi): Remove 'S' from printing asm operand.
7576 (arc_is_jli_call_p): New function.
7577 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7579 (movhi_insn): Likewise.
7580 (movsi_insn): Likewise.
7581 (movsi_set_cc_insn): Likewise.
7582 (loadqi_update): Likewise.
7583 (load_zeroextendqisi_update): Likewise.
7584 (load_signextendqisi_update): Likewise.
7585 (loadhi_update): Likewise.
7586 (load_zeroextendhisi_update): Likewise.
7587 (load_signextendhisi_update): Likewise.
7588 (loadsi_update): Likewise.
7589 (loadsf_update): Likewise.
7590 (movsicc_insn): Likewise.
7591 (bset_insn): Likewise.
7592 (bxor_insn): Likewise.
7593 (bclr_insn): Likewise.
7594 (bmsk_insn): Likewise.
7595 (bicsi3_insn): Likewise.
7596 (cmpsi_cc_c_insn): Likewise.
7597 (movsi_ne): Likewise.
7598 (movsi_cond_exec): Likewise.
7599 (clrsbsi2): Likewise.
7610 (call_i): Remove 'S' asm letter, add jli instruction.
7611 (call_value_i): Likewise.
7612 * config/arc/arc.op (mjli-always): New option.
7613 * config/arc/constraints.md (Cji): New constraint.
7614 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7616 (subsf3_fpx): Likewise.
7617 (mulsf3_fpx): Likewise.
7618 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7620 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7622 * doc/invoke.texi (ARC): Document mjli-always option.
7624 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7626 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7627 avoid addition with 0 and use incw and decw where possible.
7629 2018-01-26 Richard Biener <rguenther@suse.de>
7631 PR tree-optimization/81082
7632 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7633 association if it requires casting to unsigned.
7634 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7635 from fold_plusminus_mult_expr to catch important cases late when
7636 range info is available.
7638 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7640 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7641 * configure.ac (hidden_linkonce): New test.
7642 * configure: Regenerate.
7643 * config.in: Regenerate.
7645 2018-01-26 Julia Koval <julia.koval@intel.com>
7647 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7648 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7649 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7650 _mm_mask_bitshuffle_epi64_mask): Fix type.
7651 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7652 USI_FTYPE_V4DI_V4DI_USI): Remove.
7653 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7654 __builtin_ia32_vpshufbitqmb256_mask,
7655 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7656 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7657 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7659 2018-01-26 Alan Modra <amodra@gmail.com>
7662 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7663 UNSPEC_VBPERMQ. Sort other unspecs.
7665 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7667 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7669 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7672 * predict.c (drop_profile): Do not push/pop cfun; update also
7674 (handle_missing_profiles): Fix logic looking for zero profiles.
7676 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7679 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7680 on functions with #pragma omp declare simd or functions with simd
7682 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7683 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7684 Remove trailing \n from warning_at calls.
7686 2018-01-25 Tom de Vries <tom@codesourcery.com>
7689 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7690 for neutered workers.
7692 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7695 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7696 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7698 2018-01-24 Jeff Law <law@redhat.com>
7701 * i386.c (get_probe_interval): Move to earlier point.
7702 (ix86_compute_frame_layout): If -fstack-clash-protection and
7703 the frame is larger than the probe interval, then use pushes
7704 to save registers rather than reg->mem moves.
7705 (ix86_expand_prologue): Remove conditional for int_registers_saved
7708 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7711 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7712 min/max for never referenced object.
7714 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7717 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7719 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7720 attributes from DECL_ATTRIBUTES (decl) without affecting
7721 DECL_ATTRIBUTES (current_function_decl).
7722 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7723 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7725 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7727 PR tree-optimization/83979
7728 * fold-const.c (fold_comparison): Use constant_boolean_node
7729 instead of boolean_{true,false}_node.
7731 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7733 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7736 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7738 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7739 Simplify the clause that sets the length attribute.
7740 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7741 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7742 clause that sets the length attribute.
7743 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7745 2018-01-24 Tom de Vries <tom@codesourcery.com>
7748 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7749 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7750 Add strict parameter.
7751 (prevent_branch_around_nothing): Insert dummy insn between branch to
7752 label and label with no ptx insn inbetween.
7753 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7755 2018-01-24 Tom de Vries <tom@codesourcery.com>
7758 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7759 for neutered threads in warp.
7760 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7762 2018-01-24 Richard Biener <rguenther@suse.de>
7764 PR tree-optimization/83176
7765 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7768 2018-01-24 Richard Biener <rguenther@suse.de>
7770 PR tree-optimization/82819
7771 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7772 code generating pluses that are no-ops in the target precision.
7774 2018-01-24 Richard Biener <rguenther@suse.de>
7777 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7779 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7781 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7782 to merge probabilities.
7783 * predict.c (probably_never_executed): Also mark as cold functions
7784 with global 0 profile and guessed local profile.
7785 * profile-count.c (profile_probability::combine_with_count): New
7787 * profile-count.h (profile_probability::operator*,
7788 profile_probability::operator*=, profile_probability::operator/,
7789 profile_probability::operator/=): Reduce precision to adjusted
7790 and set value to guessed on contradictory divisions.
7791 (profile_probability::combine_with_freq): Remove.
7792 (profile_probability::combine_wiht_count): Declare.
7793 (profile_count::force_nonzero):: Set to adjusted.
7794 (profile_count::probability_in):: Set quality to adjusted.
7795 * tree-ssa-tail-merge.c (replace_block_by): Use
7798 2018-01-23 Andrew Waterman <andrew@sifive.com>
7799 Jim Wilson <jimw@sifive.com>
7801 * config/riscv/riscv.c (riscv_stack_boundary): New.
7802 (riscv_option_override): Set riscv_stack_boundary. Handle
7803 riscv_preferred_stack_boundary_arg.
7804 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7805 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7806 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7807 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7808 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7809 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7811 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7814 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7815 of struct ix86_frame.
7816 (ix86_expand_epilogue): Likewise. Add a local variable for
7817 the reg_save_offset field in struct ix86_frame.
7819 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7821 PR tree-optimization/82604
7822 * tree-loop-distribution.c (enum partition_kind): New enum item
7823 PKIND_PARTIAL_MEMSET.
7824 (partition_builtin_p): Support above new enum item.
7825 (generate_code_for_partition): Ditto.
7826 (compute_access_range): Differentiate cases that equality can be
7827 proven at all loops, the innermost loops or no loops.
7828 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7829 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7830 (finalize_partitions, distribute_loop): Don't fuse partition of
7831 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7832 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7835 2018-01-23 Martin Liska <mliska@suse.cz>
7837 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7838 order to ignore the predictor.
7839 (PRED_POLYMORPHIC_CALL): Likewise.
7840 (PRED_RECURSIVE_CALL): Likewise.
7842 2018-01-23 Martin Liska <mliska@suse.cz>
7844 * tree-profile.c (tree_profiling): Print function header to
7845 aware reader which function we are working on.
7846 * value-prof.c (gimple_find_values_to_profile): Do not print
7847 not interesting value histograms.
7849 2018-01-23 Martin Liska <mliska@suse.cz>
7851 * profile-count.h (enum profile_quality): Add
7852 profile_uninitialized as the first value. Do not number values
7853 as they are zero based.
7854 (profile_count::verify): Update sanity check.
7855 (profile_probability::verify): Likewise.
7857 2018-01-23 Nathan Sidwell <nathan@acm.org>
7859 * doc/invoke.texi (ffor-scope): Deprecate.
7861 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7863 PR tree-optimization/83510
7864 * domwalk.c (set_all_edges_as_executable): New function.
7865 (dom_walker::dom_walker): Convert bool param
7866 "skip_unreachable_blocks" to enum reachability. Move setup of
7867 edge flags to set_all_edges_as_executable and only do it when
7868 reachability is REACHABLE_BLOCKS.
7869 * domwalk.h (enum dom_walker::reachability): New enum.
7870 (dom_walker::dom_walker): Convert bool param
7871 "skip_unreachable_blocks" to enum reachability.
7872 (set_all_edges_as_executable): New decl.
7873 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7874 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7876 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7877 but converting true to REACHABLE_BLOCKS.
7878 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7880 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7881 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7882 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7884 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7885 if check_all_array_refs will be called.
7887 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7889 * tree.c (selftest::test_location_wrappers): Add more test
7892 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7894 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7895 (selftest::test_bit_in_range): Likewise.
7897 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7900 * doc/sourcebuild.texi (vect_float): Say that the selector
7901 only describes the situation when -funsafe-math-optimizations is on.
7902 (vect_float_strict): Document.
7904 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7906 PR tree-optimization/83965
7907 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7908 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7909 instead of checking only for a reduction.
7910 (vect_recog_widen_sum_pattern): Likewise.
7912 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7914 * predict.c (probably_never_executed): Only use precise profile info.
7915 (compute_function_frequency): Skip after inlining hack since we now
7916 have quality checking.
7918 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7920 * profile-count.h (profile_probability::very_unlikely,
7921 profile_probability::unlikely, profile_probability::even): Set
7922 precision to guessed.
7924 2018-01-23 Richard Biener <rguenther@suse.de>
7926 PR tree-optimization/83963
7927 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7928 Properly terminate dominator walk when crossing the exit edge not
7929 when visiting its source block.
7931 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7934 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7935 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7937 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7939 PR tree-optimization/83957
7940 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7941 semicolon after for body surrounded by braces.
7943 PR tree-optimization/83081
7944 * profile-count.h (profile_probability::split): New method.
7945 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7946 Use profile_probability::split.
7947 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7948 when splitting a single conditional jump into 2.
7950 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7952 PR tree-optimization/69452
7953 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7956 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7958 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7959 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7960 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7962 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7964 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7965 * config/rl78/rl78.md (movdi): New define_expand.
7966 * config/rl78/rl78.c (rl78_split_movdi): New function.
7968 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7971 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7973 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7974 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7975 128-bit to produce an UNSPEC move to get the double word with the
7976 signbit and then a shift directly to do signbit.
7977 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7978 implementation with a new version that just does either a direct
7979 move or a regular move. Move memory interface to separate insns.
7980 Move insns so they are next to the expander.
7981 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7982 with signbit move. Split big and little endian case.
7983 (signbit<mode>2_dm_mem_le): Likewise.
7984 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7985 (signbit<mode>2_dm2): Likewise.
7987 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7989 * config/rl78/rl78.md (anddi3): New define_expand.
7991 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7993 * config/rl78/rl78.md (umindi3): New define_expand.
7995 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7997 * config/rl78/rl78.md (smindi3): New define_expand.
7999 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8001 * config/rl78/rl78.md (smaxdi3): New define_expand.
8003 2018-01-22 Carl Love <cel@us.ibm.com>
8005 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8006 LVX_V1TI): Add macro expansion.
8007 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8008 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8009 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8010 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8011 Change check to determine if the instruction is a byte reversing
8012 entry. Fix typo in comment.
8013 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8014 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8015 Add def_builtin calls for new builtins.
8016 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8017 Add define_insn expansion.
8019 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8021 * config/rl78/rl78.md (umaxdi3): New define_expand.
8023 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8025 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8026 for non-QImode registers.
8028 2018-01-22 Richard Biener <rguenther@suse.de>
8030 PR tree-optimization/83963
8031 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8032 including the loop exit block.
8033 (scop_detection::merge_sese): Likewise.
8034 (scop_detection::add_scop): Do it here instead.
8036 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8038 * doc/sourcebuild.texi (arm_softfloat): Document.
8040 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8043 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8044 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8045 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8047 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8048 David Edelsohn <dje.gcc@gmail.com>
8051 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8052 Change "crset eq" to "crset 2".
8053 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8054 (*call_indirect_aix<mode>_nospec): Likewise.
8055 (*call_value_indirect_aix<mode>_nospec): Likewise.
8056 (*call_indirect_elfv2<mode>_nospec): Likewise.
8057 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8058 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8059 change assembly output from . to $.
8060 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8061 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8062 (*tablejump<mode>_internal1_nospec): Likewise.
8064 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8067 * config/sh/sh_optimize_sett_clrt.cc:
8068 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8070 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8072 PR tree-optimization/83940
8073 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8074 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8075 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8076 use it to pass back the definition type.
8077 (vect_check_store_rhs): Likewise rhs_dt_out.
8078 (vect_build_gather_load_calls): Add a mask_dt argument and use
8079 it instead of a call to vect_is_simple_use.
8080 (vectorizable_store): Update calls to vect_check_load_store_mask
8081 and vect_check_store_rhs. Use the dt returned by the latter instead
8082 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8083 instead of calls to vect_is_simple_use. Pass the scalar rather
8084 than the vector operand to vect_is_simple_use when handling
8085 second and subsequent copies of an rhs value.
8086 (vectorizable_load): Update calls to vect_check_load_store_mask
8087 and vect_build_gather_load_calls. Use the cached mask_dt and
8088 gs_info.offset_dt instead of calls to vect_is_simple_use.
8090 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8093 * tree-emutls.c: Include gimplify.h.
8094 (lower_emutls_2): New function.
8095 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8096 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8097 it before further processing.
8100 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8101 UINTVAL (trueop1) instead of INTVAL (op1).
8103 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8107 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8108 INCOMING_FRAME_SP_OFFSET if not defined.
8109 (scan_trace): Add ENTRY argument. If true and
8110 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8111 emit a note to adjust the CFA offset.
8112 (create_cfi_notes): Adjust scan_trace callers.
8113 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8114 INCOMING_FRAME_SP_OFFSET in the CIE.
8115 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8116 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8118 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8119 * doc/tm.texi: Regenerated.
8121 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8123 PR rtl-optimization/83147
8124 * lra-constraints.c (remove_inheritance_pseudos): Use
8125 lra_substitute_pseudo_within_insn.
8127 2018-01-19 Tom de Vries <tom@codesourcery.com>
8128 Cesar Philippidis <cesar@codesourcery.com>
8131 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8133 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8136 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8137 spaces for function labels.
8139 2018-01-19 Martin Liska <mliska@suse.cz>
8141 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8142 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8143 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8144 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8145 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8146 (PRED_CONST_RETURN): Change from 69 to 65.
8147 (PRED_NULL_RETURN): Change from 91 to 71.
8148 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8149 (PRED_LOOP_GUARD): Change from 66 to 73.
8151 2018-01-19 Martin Liska <mliska@suse.cz>
8153 * predict.c (predict_insn_def): Add new assert.
8154 (struct branch_predictor): Change type to signed integer.
8155 (test_prediction_value_range): Amend test to cover
8157 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8158 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8159 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8160 (PRED_LOOP_IV_COMPARE): Likewise.
8161 * predict.h (PROB_UNINITIALIZED): Define new constant.
8163 2018-01-19 Martin Liska <mliska@suse.cz>
8165 * predict.c (dump_prediction): Add new format for
8166 analyze_brprob.py script which is enabled with -details
8168 * profile-count.h (precise_p): New function.
8170 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8172 PR tree-optimization/83922
8173 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8174 there are no statements that need masking.
8175 (vect_active_double_reduction_p): New function.
8176 (vect_analyze_loop_operations): Use it when handling phis that
8177 are not in the loop header.
8179 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8181 PR tree-optimization/83914
8182 * tree-vect-loop.c (vectorizable_induction): Don't convert
8183 init_expr or apply the peeling adjustment for inductions
8184 that are nested within the vectorized loop.
8186 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8188 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8191 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8195 * function.h (gimplify_parameters): Add gimple_seq * argument.
8196 * function.c: Include gimple.h and options.h.
8197 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8198 for the added local temporaries if needed.
8199 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8200 if there are any parameter cleanups, wrap whole body into a
8201 try/finally with the cleanups.
8203 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8206 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8207 Use GET_MODE_CLASS for scalar floating point.
8209 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8213 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8214 Fix call of call_cgraph_insertion_hooks.
8216 2018-01-18 Martin Sebor <msebor@redhat.com>
8218 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8220 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8223 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8226 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8229 * common.opt: (-ffile-prefix-map): New option.
8230 * opts.c (common_handle_option): Defer it.
8231 * opts-global.c (handle_common_deferred_options): Handle it.
8232 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8233 * file-prefix-map.h: New file.
8234 (remap_debug_filename, add_debug_prefix_map): ...here.
8235 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8236 * final.c (debug_prefix_map, add_debug_prefix_map
8237 remap_debug_filename): Move to...
8238 * file-prefix-map.c: New file.
8239 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8240 generalize, get rid of alloca(), use strrchr() instead of strchr().
8241 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8242 Implement in terms of add_prefix_map().
8243 (remap_macro_filename, remap_debug_filename): Implement in term of
8245 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8246 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8247 * dbxout.c: Include file-prefix-map.h.
8248 * varasm.c: Likewise.
8249 * vmsdbgout.c: Likewise.
8250 * xcoffout.c: Likewise.
8251 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8252 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8253 * doc/invoke.texi (-ffile-prefix-map): Document.
8254 (-fdebug-prefix-map): Update description.
8256 2018-01-18 Martin Liska <mliska@suse.cz>
8258 * config/i386/i386.c (indirect_thunk_name): Document that also
8260 (output_indirect_thunk): Document why both instructions
8261 (pause and lfence) are generated.
8263 2018-01-18 Richard Biener <rguenther@suse.de>
8265 PR tree-optimization/83887
8266 * graphite-scop-detection.c
8267 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8268 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8269 (scop_detection::merge_sese): Re-implement with a flood-fill
8270 algorithm that properly finds a SESE region if it exists.
8272 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8275 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8276 pointer_diff optimizations use view_convert instead of convert.
8278 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8280 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8281 Generate different code for -mno-speculate-indirect-jumps.
8282 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8283 (*call_indirect_aix<mode>): Disable for
8284 -mno-speculate-indirect-jumps.
8285 (*call_indirect_aix<mode>_nospec): New define_insn.
8286 (*call_value_indirect_aix<mode>): Disable for
8287 -mno-speculate-indirect-jumps.
8288 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8289 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8290 -mno-speculate-indirect-jumps.
8291 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8293 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8295 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8296 long double type, set the flags for noting the default long double
8297 type, even if we don't pass or return a long double type.
8299 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8302 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8305 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8307 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8308 support for merge[hl].
8309 (fold_mergehl_helper): New helper function.
8310 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8311 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8312 (altivec_vmrglw_direct): Add xxmrglw insn.
8314 2018-01-17 Andrew Waterman <andrew@sifive.com>
8316 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8317 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8319 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8322 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8323 call the lto_location_cache before reading the
8324 DECL_SOURCE_LOCATION of the types.
8326 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8327 Richard Sandiford <richard.sandiford@linaro.org>
8329 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8330 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8331 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8332 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8333 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8335 * config/aarch64/constraints.md (aarch64_movti_operand):
8337 * config/aarch64/predicates.md (Uti): Add new constraint.
8339 2018-01-17 Carl Love <cel@us.ibm.com>
8341 * config/rs6000/vsx.md (define_expand xl_len_r,
8342 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8343 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8345 (define_expand, define_insn): Move the shift left from the
8346 define_insn to the define_expand for lxvl and stxvl instructions.
8347 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8348 and XL_LEN_R definitions to PURE.
8350 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8352 * config/i386/i386.c (indirect_thunk_name): Declare regno
8353 as unsigned int. Compare regno with INVALID_REGNUM.
8354 (output_indirect_thunk): Ditto.
8355 (output_indirect_thunk_function): Ditto.
8356 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8357 in the call to output_indirect_thunk_function.
8359 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8362 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8363 rather than the size of inner_type to determine the stack slot size
8364 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8366 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8369 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8372 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8374 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8375 endian Linux systems to optionally enable multilibs for selecting
8376 the long double type if the user configured an explicit type.
8377 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8378 have no long double multilibs if not defined.
8379 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8380 warn if the user used -mabi={ieee,ibm}longdouble and we built
8381 multilibs for long double.
8382 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8383 appropriate multilib option.
8384 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8386 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8387 for building long double multilibs.
8388 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8390 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8392 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8395 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8397 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8400 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8403 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8406 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8408 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8409 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8411 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8413 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8414 different rtl trees depending on TARGET_64BIT.
8415 (rs6000_gen_lvx): Likewise.
8417 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8419 * config/visium/visium.md (nop): Tweak comment.
8420 (hazard_nop): Likewise.
8422 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8424 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8425 -mspeculate-indirect-jumps.
8426 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8427 for -mno-speculate-indirect-jumps.
8428 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8429 (*call_value_indirect_elfv2<mode>): Disable for
8430 -mno-speculate-indirect-jumps.
8431 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8432 (indirect_jump): Emit different RTL for
8433 -mno-speculate-indirect-jumps.
8434 (*indirect_jump<mode>): Disable for
8435 -mno-speculate-indirect-jumps.
8436 (*indirect_jump<mode>_nospec): New define_insn.
8437 (tablejump): Emit different RTL for
8438 -mno-speculate-indirect-jumps.
8439 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8440 (tablejumpsi_nospec): New define_expand.
8441 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8442 (tablejumpdi_nospec): New define_expand.
8443 (*tablejump<mode>_internal1): Disable for
8444 -mno-speculate-indirect-jumps.
8445 (*tablejump<mode>_internal1_nospec): New define_insn.
8446 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8449 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8451 * caller-save.c (insert_save): Drop unnecessary parameter. All
8454 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8455 Richard Biener <rguenth@suse.de>
8458 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8459 return early, inline manually is_gimple_sizepos. Make sure if we
8460 call gimplify_expr we don't end up with a gimple constant.
8461 * tree.c (variably_modified_type_p): Don't return true for
8462 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8463 * gimplify.h (is_gimple_sizepos): Remove.
8465 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8467 PR tree-optimization/83857
8468 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8469 vectorizable_live_operation for pure SLP statements.
8470 (vectorizable_live_operation): Handle PHIs.
8472 2018-01-16 Richard Biener <rguenther@suse.de>
8474 PR tree-optimization/83867
8475 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8476 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8478 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8481 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8482 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8483 If off is not INTEGER_CST, issue a may not be aligned warning
8484 rather than isn't aligned. Use isn%'t rather than isn't.
8485 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8487 <case MULT_EXPR>: Improve the case when bottom and one of the
8488 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8489 operand, in that case check if the other operand is multiple of
8490 bottom divided by the INTEGER_CST operand.
8492 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8495 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8496 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8497 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8498 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8499 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8500 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8501 (pa_function_arg_size): New function.
8503 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8505 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8506 in a separate statement.
8508 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8510 PR tree-optimization/83847
8511 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8512 group gathers and scatters.
8514 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8516 PR rtl-optimization/86620
8517 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8519 PR rtl-optimization/83213
8520 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8521 to last if both are JUMP_INSNs.
8523 PR tree-optimization/83843
8524 * gimple-ssa-store-merging.c
8525 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8526 store_immediate_info for bswap/nop orig_stores.
8528 2018-01-15 Andrew Waterman <andrew@sifive.com>
8530 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8532 <UDIV>: Increase cost if !TARGET_DIV.
8534 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8536 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8537 (define_attr "cr_logical_3op"): New.
8538 (cceq_ior_compare): Adjust.
8539 (cceq_ior_compare_complement): Adjust.
8540 (*cceq_rev_compare): Adjust.
8541 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8542 (is_cracked_insn): Adjust.
8543 (insn_must_be_first_in_group): Adjust.
8544 * config/rs6000/40x.md: Adjust.
8545 * config/rs6000/440.md: Adjust.
8546 * config/rs6000/476.md: Adjust.
8547 * config/rs6000/601.md: Adjust.
8548 * config/rs6000/603.md: Adjust.
8549 * config/rs6000/6xx.md: Adjust.
8550 * config/rs6000/7450.md: Adjust.
8551 * config/rs6000/7xx.md: Adjust.
8552 * config/rs6000/8540.md: Adjust.
8553 * config/rs6000/cell.md: Adjust.
8554 * config/rs6000/e300c2c3.md: Adjust.
8555 * config/rs6000/e500mc.md: Adjust.
8556 * config/rs6000/e500mc64.md: Adjust.
8557 * config/rs6000/e5500.md: Adjust.
8558 * config/rs6000/e6500.md: Adjust.
8559 * config/rs6000/mpc.md: Adjust.
8560 * config/rs6000/power4.md: Adjust.
8561 * config/rs6000/power5.md: Adjust.
8562 * config/rs6000/power6.md: Adjust.
8563 * config/rs6000/power7.md: Adjust.
8564 * config/rs6000/power8.md: Adjust.
8565 * config/rs6000/power9.md: Adjust.
8566 * config/rs6000/rs64.md: Adjust.
8567 * config/rs6000/titan.md: Adjust.
8569 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8571 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8572 ix86_indirect_branch_register logic.
8574 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8576 * config/i386/constraints.md (Bs): Update
8577 ix86_indirect_branch_register check. Don't check
8578 ix86_indirect_branch_register with GOT_memory_operand.
8580 * config/i386/predicates.md (GOT_memory_operand): Don't check
8581 ix86_indirect_branch_register here.
8582 (GOT32_symbol_operand): Likewise.
8584 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8586 * config/i386/predicates.md (constant_call_address_operand):
8587 Rewrite ix86_indirect_branch_register logic.
8588 (sibcall_insn_operand): Likewise.
8590 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8592 * config/i386/constraints.md (Bs): Replace
8593 ix86_indirect_branch_thunk_register with
8594 ix86_indirect_branch_register.
8596 * config/i386/i386.md (indirect_jump): Likewise.
8597 (tablejump): Likewise.
8598 (*sibcall_memory): Likewise.
8599 (*sibcall_value_memory): Likewise.
8600 Peepholes of indirect call and jump via memory: Likewise.
8601 * config/i386/i386.opt: Likewise.
8602 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8603 (GOT_memory_operand): Likewise.
8604 (call_insn_operand): Likewise.
8605 (sibcall_insn_operand): Likewise.
8606 (GOT32_symbol_operand): Likewise.
8608 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8611 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8612 type rather than type addr's type points to.
8613 (expand_omp_atomic_mutex): Likewise.
8614 (expand_omp_atomic): Likewise.
8616 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8619 * config/i386/i386.c (output_indirect_thunk_function): Use
8620 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8621 for __x86_return_thunk.
8623 2018-01-15 Richard Biener <rguenther@suse.de>
8626 * expmed.c (extract_bit_field_1): Fix typo.
8628 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8631 * config/arm/iterators.md (VF): New mode iterator.
8632 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8633 Remove integer-related logic from pattern.
8634 (neon_vabd<mode>_3): Likewise.
8636 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8639 * common.opt (fstrict-overflow): No longer an alias.
8640 (fwrapv-pointer): New option.
8641 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8642 also for pointer types based on flag_wrapv_pointer.
8643 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8644 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8645 opts->x_flag_wrapv got set.
8646 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8647 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8648 POINTER_TYPE_OVERFLOW_UNDEFINED.
8649 * match.pd: Likewise in address comparison pattern.
8650 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8652 2018-01-15 Richard Biener <rguenther@suse.de>
8655 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8656 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8657 Reset type names to their identifier if their TYPE_DECL doesn't
8658 have linkage (and thus is used for ODR and devirt).
8659 (save_debug_info_for_decl): Remove.
8660 (save_debug_info_for_type): Likewise.
8661 (add_tree_to_fld_list): Adjust.
8662 * tree-pretty-print.c (dump_generic_node): Make dumping of
8663 type names more robust.
8665 2018-01-15 Richard Biener <rguenther@suse.de>
8667 * BASE-VER: Bump to 8.0.1.
8669 2018-01-14 Martin Sebor <msebor@redhat.com>
8672 * builtins.c (check_access): Avoid warning when the no-warning bit
8675 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8677 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8678 * ira-color (allocno_hard_regs_compare): Likewise.
8680 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8683 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8684 Use .pushsection/.popsection.
8686 2018-01-14 Martin Sebor <msebor@redhat.com>
8689 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8691 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8693 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8694 entry from extra_headers.
8695 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8696 extra_headers, make the list bitwise identical to the i?86-*-* one.
8698 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8700 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8701 -mcmodel=large with -mindirect-branch=thunk,
8702 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8703 -mfunction-return=thunk-extern.
8704 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8705 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8706 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8708 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8710 * config/i386/i386.c (print_reg): Print the name of the full
8711 integer register without '%'.
8712 (ix86_print_operand): Handle 'V'.
8713 * doc/extend.texi: Document 'V' modifier.
8715 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8717 * config/i386/constraints.md (Bs): Disallow memory operand for
8718 -mindirect-branch-register.
8720 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8721 (GOT_memory_operand): Likewise.
8722 (call_insn_operand): Likewise.
8723 (sibcall_insn_operand): Likewise.
8724 (GOT32_symbol_operand): Likewise.
8725 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8726 for -mindirect-branch-register.
8727 (tablejump): Likewise.
8728 (*sibcall_memory): Likewise.
8729 (*sibcall_value_memory): Likewise.
8730 Disallow peepholes of indirect call and jump via memory for
8731 -mindirect-branch-register.
8732 (*call_pop): Replace m with Bw.
8733 (*call_value_pop): Likewise.
8734 (*sibcall_pop_memory): Replace m with Bs.
8735 * config/i386/i386.opt (mindirect-branch-register): New option.
8736 * doc/invoke.texi: Document -mindirect-branch-register option.
8738 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8740 * config/i386/i386-protos.h (ix86_output_function_return): New.
8741 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8742 set function_return_type.
8743 (indirect_thunk_name): Add ret_p to indicate thunk for function
8745 (output_indirect_thunk_function): Pass false to
8746 indirect_thunk_name.
8747 (ix86_output_indirect_branch_via_reg): Likewise.
8748 (ix86_output_indirect_branch_via_push): Likewise.
8749 (output_indirect_thunk_function): Create alias for function
8750 return thunk if regno < 0.
8751 (ix86_output_function_return): New function.
8752 (ix86_handle_fndecl_attribute): Handle function_return.
8753 (ix86_attribute_table): Add function_return.
8754 * config/i386/i386.h (machine_function): Add
8755 function_return_type.
8756 * config/i386/i386.md (simple_return_internal): Use
8757 ix86_output_function_return.
8758 (simple_return_internal_long): Likewise.
8759 * config/i386/i386.opt (mfunction-return=): New option.
8760 (indirect_branch): Mention -mfunction-return=.
8761 * doc/extend.texi: Document function_return function attribute.
8762 * doc/invoke.texi: Document -mfunction-return= option.
8764 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8766 * config/i386/i386-opts.h (indirect_branch): New.
8767 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8768 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8769 with local indirect jump when converting indirect call and jump.
8770 (ix86_set_indirect_branch_type): New.
8771 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8772 (indirectlabelno): New.
8773 (indirect_thunk_needed): Likewise.
8774 (indirect_thunk_bnd_needed): Likewise.
8775 (indirect_thunks_used): Likewise.
8776 (indirect_thunks_bnd_used): Likewise.
8777 (INDIRECT_LABEL): Likewise.
8778 (indirect_thunk_name): Likewise.
8779 (output_indirect_thunk): Likewise.
8780 (output_indirect_thunk_function): Likewise.
8781 (ix86_output_indirect_branch_via_reg): Likewise.
8782 (ix86_output_indirect_branch_via_push): Likewise.
8783 (ix86_output_indirect_branch): Likewise.
8784 (ix86_output_indirect_jmp): Likewise.
8785 (ix86_code_end): Call output_indirect_thunk_function if needed.
8786 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8788 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8789 (ix86_attribute_table): Add indirect_branch.
8790 * config/i386/i386.h (machine_function): Add indirect_branch_type
8791 and has_local_indirect_jump.
8792 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8794 (tablejump): Likewise.
8795 (*indirect_jump): Use ix86_output_indirect_jmp.
8796 (*tablejump_1): Likewise.
8797 (simple_return_indirect_internal): Likewise.
8798 * config/i386/i386.opt (mindirect-branch=): New option.
8799 (indirect_branch): New.
8802 (thunk-inline): Likewise.
8803 (thunk-extern): Likewise.
8804 * doc/extend.texi: Document indirect_branch function attribute.
8805 * doc/invoke.texi: Document -mindirect-branch= option.
8807 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8810 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8812 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8814 * ipa-inline.c (want_inline_small_function_p): Return false if
8815 inlining has already failed with CIF_FINAL_ERROR.
8816 (update_caller_keys): Call want_inline_small_function_p before
8818 (update_callee_keys): Likewise.
8820 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8822 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8824 (rs6000_quadword_masked_address_p): Likewise.
8825 (quad_aligned_load_p): Likewise.
8826 (quad_aligned_store_p): Likewise.
8827 (const_load_sequence_p): Add comment to describe the outer-most loop.
8828 (mimic_memory_attributes_and_flags): New function.
8829 (rs6000_gen_stvx): Likewise.
8830 (replace_swapped_aligned_store): Likewise.
8831 (rs6000_gen_lvx): Likewise.
8832 (replace_swapped_aligned_load): Likewise.
8833 (replace_swapped_load_constant): Capitalize argument name in
8834 comment describing this function.
8835 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8836 and stores that access quad-word aligned addresses and replace
8837 with stvx or lvx instructions when appropriate.
8838 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8839 New function prototype.
8840 (rs6000_quadword_masked_address_p): Likewise.
8841 (rs6000_gen_lvx): Likewise.
8842 (rs6000_gen_stvx): Likewise.
8843 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8844 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8845 when memory address is aligned.
8846 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8847 this split to select lvx instruction when memory address is aligned.
8848 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8849 instruction when memory address is aligned.
8850 (*vsx_le_perm_load_v16qi): Likewise.
8851 (four unnamed splitters): Modify to select the stvx instruction
8852 when memory is aligned.
8854 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8856 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8857 which appears in the queue multiple times.
8859 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8860 Alan Hayward <alan.hayward@arm.com>
8861 David Sherwood <david.sherwood@arm.com>
8863 * tree-vectorizer.h (vec_lower_bound): New structure.
8864 (_loop_vec_info): Add check_nonzero and lower_bounds.
8865 (LOOP_VINFO_CHECK_NONZERO): New macro.
8866 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8867 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8868 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8869 fields. Make seg_len the distance travelled, not including the
8871 (dr_direction_indicator): Declare.
8872 (dr_zero_step_indicator): Likewise.
8873 (dr_known_forward_stride_p): Likewise.
8874 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8876 (runtime_alias_check_p): Allow runtime alias checks with
8878 (operator ==): Compare access_size and align.
8879 (prune_runtime_alias_test_list): Rework for new distinction between
8880 the access_size and seg_len.
8881 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8883 (get_segment_min_max): New function.
8884 (create_intersect_range_checks): Use it.
8885 (dr_step_indicator): New function.
8886 (dr_direction_indicator): Likewise.
8887 (dr_zero_step_indicator): Likewise.
8888 (dr_known_forward_stride_p): Likewise.
8889 * tree-loop-distribution.c (data_ref_segment_size): Return
8890 DR_STEP * (niters - 1).
8891 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8893 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8894 (vect_preserves_scalar_order_p): New function, split out from...
8895 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8896 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8897 (vect_vfa_access_size): New function.
8898 (vect_vfa_align): Likewise.
8899 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8900 (dump_lower_bound): New function.
8901 (vect_check_lower_bound): Likewise.
8902 (vect_small_gap_p): Likewise.
8903 (vectorizable_with_step_bound_p): Likewise.
8904 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8905 depencies if the vectorization factor is 1. Convert the checks
8906 for nonzero steps into checks on the bounds of DR_STEP. Try using
8907 a bunds check for variable steps if the minimum required step is
8908 relatively small. Update calls to the dr_with_seg_len
8909 constructor and to vect_compile_time_alias.
8910 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8912 (vect_loop_versioning): Call it.
8913 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8915 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8918 Alan Hayward <alan.hayward@arm.com>
8919 David Sherwood <david.sherwood@arm.com>
8921 * doc/sourcebuild.texi (vect_scatter_store): Document.
8922 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8924 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8926 * genopinit.c (main): Add supports_vec_scatter_store and
8927 supports_vec_scatter_store_cached to target_optabs.
8928 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8929 IFN_MASK_SCATTER_STORE.
8930 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8932 * internal-fn.h (internal_store_fn_p): Declare.
8933 (internal_fn_stored_value_index): Likewise.
8934 * internal-fn.c (scatter_store_direct): New macro.
8935 (expand_scatter_store_optab_fn): New function.
8936 (direct_scatter_store_optab_supported_p): New macro.
8937 (internal_store_fn_p): New function.
8938 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8939 IFN_MASK_SCATTER_STORE.
8940 (internal_fn_mask_index): Likewise.
8941 (internal_fn_stored_value_index): New function.
8942 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8944 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8945 * optabs-query.c (supports_vec_scatter_store_p): New function.
8946 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8947 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8948 true for scatter stores.
8949 (vect_gather_scatter_fn_p): Handle scatter stores too.
8950 (vect_check_gather_scatter): Consider using scatter stores if
8951 supports_vec_scatter_store_p.
8952 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8954 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8955 internal_fn_stored_value_index.
8956 (check_load_store_masking): Handle scatter stores too.
8957 (vect_get_store_rhs): Make public.
8958 (vectorizable_call): Use internal_store_fn_p.
8959 (vectorizable_store): Handle scatter store internal functions.
8960 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8961 when deciding whether the end of the group has been reached.
8962 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8963 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8964 (mask_scatter_store<mode>): New insns.
8966 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8967 Alan Hayward <alan.hayward@arm.com>
8968 David Sherwood <david.sherwood@arm.com>
8970 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8971 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8972 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8974 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8975 Use vect_truncate_gather_scatter_offset if we can't treat the
8976 operation as a normal gather load or scatter store.
8977 (get_group_load_store_type): Take the gather_scatter_info
8978 as argument. Try using a gather load or scatter store for
8979 single-element groups.
8980 (get_load_store_type): Update calls to get_group_load_store_type
8981 and vect_use_strided_gather_scatters_p.
8983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8984 Alan Hayward <alan.hayward@arm.com>
8985 David Sherwood <david.sherwood@arm.com>
8987 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8988 optional tree argument.
8989 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8991 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8992 but continue to use the current value as a fallback.
8993 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8994 to compare the updates.
8995 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8996 (get_load_store_type): Use it when handling a strided access.
8997 (vect_get_strided_load_store_ops): New function.
8998 (vect_get_data_ptr_increment): Likewise.
8999 (vectorizable_load): Handle strided gather loads. Always pass
9000 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9002 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9003 Alan Hayward <alan.hayward@arm.com>
9004 David Sherwood <david.sherwood@arm.com>
9006 * doc/md.texi (gather_load@var{m}): Document.
9007 (mask_gather_load@var{m}): Likewise.
9008 * genopinit.c (main): Add supports_vec_gather_load and
9009 supports_vec_gather_load_cached to target_optabs.
9010 * optabs-tree.c (init_tree_optimization_optabs): Use
9011 ggc_cleared_alloc to allocate target_optabs.
9012 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9013 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9015 * internal-fn.h (internal_load_fn_p): Declare.
9016 (internal_gather_scatter_fn_p): Likewise.
9017 (internal_fn_mask_index): Likewise.
9018 (internal_gather_scatter_fn_supported_p): Likewise.
9019 * internal-fn.c (gather_load_direct): New macro.
9020 (expand_gather_load_optab_fn): New function.
9021 (direct_gather_load_optab_supported_p): New macro.
9022 (direct_internal_fn_optab): New function.
9023 (internal_load_fn_p): Likewise.
9024 (internal_gather_scatter_fn_p): Likewise.
9025 (internal_fn_mask_index): Likewise.
9026 (internal_gather_scatter_fn_supported_p): Likewise.
9027 * optabs-query.c (supports_at_least_one_mode_p): New function.
9028 (supports_vec_gather_load_p): Likewise.
9029 * optabs-query.h (supports_vec_gather_load_p): Declare.
9030 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9031 and memory_type field.
9032 (NUM_PATTERNS): Bump to 15.
9033 * tree-vect-data-refs.c: Include internal-fn.h.
9034 (vect_gather_scatter_fn_p): New function.
9035 (vect_describe_gather_scatter_call): Likewise.
9036 (vect_check_gather_scatter): Try using internal functions for
9037 gather loads. Recognize existing calls to a gather load function.
9038 (vect_analyze_data_refs): Consider using gather loads if
9039 supports_vec_gather_load_p.
9040 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9041 (vect_get_gather_scatter_offset_type): Likewise.
9042 (vect_convert_mask_for_vectype): Likewise.
9043 (vect_add_conversion_to_patterm): Likewise.
9044 (vect_try_gather_scatter_pattern): Likewise.
9045 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9046 (vect_vect_recog_func_ptrs): Add it.
9047 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9048 internal_fn_mask_index and internal_gather_scatter_fn_p.
9049 (check_load_store_masking): Take the gather_scatter_info as an
9050 argument and handle gather loads.
9051 (vect_get_gather_scatter_ops): New function.
9052 (vectorizable_call): Check internal_load_fn_p.
9053 (vectorizable_load): Likewise. Handle gather load internal
9055 (vectorizable_store): Update call to check_load_store_masking.
9056 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9057 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9058 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9059 (aarch64_gather_scale_operand_d): New predicates.
9060 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9061 (mask_gather_load<mode>): New insns.
9063 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9064 Alan Hayward <alan.hayward@arm.com>
9065 David Sherwood <david.sherwood@arm.com>
9067 * optabs.def (fold_left_plus_optab): New optab.
9068 * doc/md.texi (fold_left_plus_@var{m}): Document.
9069 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9070 * internal-fn.c (fold_left_direct): Define.
9071 (expand_fold_left_optab_fn): Likewise.
9072 (direct_fold_left_optab_supported_p): Likewise.
9073 * fold-const-call.c (fold_const_fold_left): New function.
9074 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9075 * tree-parloops.c (valid_reduction_p): New function.
9076 (gather_scalar_reductions): Use it.
9077 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9078 (vect_finish_replace_stmt): Declare.
9079 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9080 (needs_fold_left_reduction_p): New function, split out from...
9081 (vect_is_simple_reduction): ...here. Accept reductions that
9082 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9083 (vect_force_simple_reduction): Also store the reduction type in
9084 the assignment's STMT_VINFO_REDUC_TYPE.
9085 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9086 (merge_with_identity): New function.
9087 (vect_expand_fold_left): Likewise.
9088 (vectorize_fold_left_reduction): Likewise.
9089 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9090 scalar phi in place for it. Check for target support and reject
9091 cases that would reassociate the operation. Defer the transform
9092 phase to vectorize_fold_left_reduction.
9093 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9094 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9095 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9097 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9099 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9100 call to ifc_temp_var.
9102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9103 Alan Hayward <alan.hayward@arm.com>
9104 David Sherwood <david.sherwood@arm.com>
9106 * target.def (legitimize_address_displacement): Take the original
9107 offset as a poly_int.
9108 * targhooks.h (default_legitimize_address_displacement): Update
9110 * targhooks.c (default_legitimize_address_displacement): Likewise.
9111 * doc/tm.texi: Regenerate.
9112 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9113 as an argument, moving assert of ad->disp == ad->disp_term to...
9114 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9115 Try calling targetm.legitimize_address_displacement before expanding
9116 the address rather than afterwards, and adjust for the new interface.
9117 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9118 Match the new hook interface. Handle SVE addresses.
9119 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9122 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9124 * Makefile.in (OBJS): Add early-remat.o.
9125 * target.def (select_early_remat_modes): New hook.
9126 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9127 * doc/tm.texi: Regenerate.
9128 * targhooks.h (default_select_early_remat_modes): Declare.
9129 * targhooks.c (default_select_early_remat_modes): New function.
9130 * timevar.def (TV_EARLY_REMAT): New timevar.
9131 * passes.def (pass_early_remat): New pass.
9132 * tree-pass.h (make_pass_early_remat): Declare.
9133 * early-remat.c: New file.
9134 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9136 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
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-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9143 vfm1 with a bound_epilog parameter.
9144 (vect_do_peeling): Update calls accordingly, and move the prologue
9145 call earlier in the function. Treat the base bound_epilog as 0 for
9146 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9147 this base when peeling for gaps.
9148 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9149 with fully-masked loops.
9150 (vect_estimate_min_profitable_iters): Handle the single peeled
9151 iteration in that case.
9153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9154 Alan Hayward <alan.hayward@arm.com>
9155 David Sherwood <david.sherwood@arm.com>
9157 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9158 single-element interleaving even if the size is not a power of 2.
9159 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9160 accesses for single-element interleaving if the group size is
9163 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9164 Alan Hayward <alan.hayward@arm.com>
9165 David Sherwood <david.sherwood@arm.com>
9167 * doc/md.texi (fold_extract_last_@var{m}): Document.
9168 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9169 * optabs.def (fold_extract_last_optab): New optab.
9170 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9171 * internal-fn.c (fold_extract_direct): New macro.
9172 (expand_fold_extract_optab_fn): Likewise.
9173 (direct_fold_extract_optab_supported_p): Likewise.
9174 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9175 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9176 EXTRACT_LAST_REDUCTION.
9177 (get_initial_def_for_reduction): Do not create an initial vector
9178 for EXTRACT_LAST_REDUCTION reductions.
9179 (vectorizable_reduction): Leave the scalar phi in place for
9180 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9181 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9182 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9183 transform phase to vectorizable_condition.
9184 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9186 (vect_finish_stmt_generation): ...here.
9187 (vect_finish_replace_stmt): New function.
9188 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9189 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9191 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9193 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9194 Alan Hayward <alan.hayward@arm.com>
9195 David Sherwood <david.sherwood@arm.com>
9197 * doc/md.texi (extract_last_@var{m}): Document.
9198 * optabs.def (extract_last_optab): New optab.
9199 * internal-fn.def (EXTRACT_LAST): New internal function.
9200 * internal-fn.c (cond_unary_direct): New macro.
9201 (expand_cond_unary_optab_fn): Likewise.
9202 (direct_cond_unary_optab_supported_p): Likewise.
9203 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9204 loops using EXTRACT_LAST.
9205 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9206 (extract_last_<mode>): ...this optab.
9207 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
9214 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9215 * doc/tm.texi: Regenerate.
9216 * targhooks.h (default_empty_mask_is_expensive): Declare.
9217 * targhooks.c (default_empty_mask_is_expensive): New function.
9218 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9219 if the target says that empty masks are expensive.
9220 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9222 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9224 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9225 Alan Hayward <alan.hayward@arm.com>
9226 David Sherwood <david.sherwood@arm.com>
9228 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9229 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9230 (vect_use_loop_mask_for_alignment_p): New function.
9231 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9232 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9233 niters_skip argument. Make sure that the first niters_skip elements
9234 of the first iteration are inactive.
9235 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9236 Update call to vect_set_loop_masks_directly.
9237 (get_misalign_in_elems): New function, split out from...
9238 (vect_gen_prolog_loop_niters): ...here.
9239 (vect_update_init_of_dr): Take a code argument that specifies whether
9240 the adjustment should be added or subtracted.
9241 (vect_update_init_of_drs): Likewise.
9242 (vect_prepare_for_masked_peels): New function.
9243 (vect_do_peeling): Skip prologue peeling if we're using a mask
9244 instead. Update call to vect_update_inits_of_drs.
9245 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9247 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9248 alignment. Do not include the number of peeled iterations in
9249 the minimum threshold in that case.
9250 (vectorizable_induction): Adjust the start value down by
9251 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9252 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9253 Take the number of skipped iterations into account when calculating
9255 * tree-vect-stmts.c (vect_gen_while_not): New function.
9257 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9258 Alan Hayward <alan.hayward@arm.com>
9259 David Sherwood <david.sherwood@arm.com>
9261 * doc/sourcebuild.texi (vect_fully_masked): Document.
9262 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9264 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9266 (vect_analyze_loop_2): ...here. Don't check the vectorization
9267 factor against the number of loop iterations if the loop is
9270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9271 Alan Hayward <alan.hayward@arm.com>
9272 David Sherwood <david.sherwood@arm.com>
9274 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9275 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9276 (dump_groups): Update accordingly.
9277 (iv_use::mem_type): New member variable.
9278 (address_p): New function.
9279 (record_use): Add a mem_type argument and initialize the new
9281 (record_group_use): Add a mem_type argument. Use address_p.
9282 Remove obsolete null checks of base_object. Update call to record_use.
9283 (find_interesting_uses_op): Update call to record_group_use.
9284 (find_interesting_uses_cond): Likewise.
9285 (find_interesting_uses_address): Likewise.
9286 (get_mem_type_for_internal_fn): New function.
9287 (find_address_like_use): Likewise.
9288 (find_interesting_uses_stmt): Try find_address_like_use before
9289 calling find_interesting_uses_op.
9290 (addr_offset_valid_p): Use the iv mem_type field as the type
9291 of the addressed memory.
9292 (add_autoinc_candidates): Likewise.
9293 (get_address_cost): Likewise.
9294 (split_small_address_groups_p): Use address_p.
9295 (split_address_groups): Likewise.
9296 (add_iv_candidate_for_use): Likewise.
9297 (autoinc_possible_for_pair): Likewise.
9298 (rewrite_groups): Likewise.
9299 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9300 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9301 (get_alias_ptr_type_for_ptr_address): New function.
9302 (rewrite_use_address): Rewrite address uses in calls that were
9303 identified by find_address_like_use.
9305 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9306 Alan Hayward <alan.hayward@arm.com>
9307 David Sherwood <david.sherwood@arm.com>
9309 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9311 * gimple-expr.h (is_gimple_addressable: Likewise.
9312 * gimple-expr.c (is_gimple_address): Likewise.
9313 * internal-fn.c (expand_call_mem_ref): New function.
9314 (expand_mask_load_optab_fn): Use it.
9315 (expand_mask_store_optab_fn): Likewise.
9317 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9318 Alan Hayward <alan.hayward@arm.com>
9319 David Sherwood <david.sherwood@arm.com>
9321 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9322 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9323 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9324 (cond_umax@var{mode}): Document.
9325 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9326 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9327 (cond_umin_optab, cond_umax_optab): New optabs.
9328 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9329 (COND_IOR, COND_XOR): New internal functions.
9330 * internal-fn.h (get_conditional_internal_fn): Declare.
9331 * internal-fn.c (cond_binary_direct): New macro.
9332 (expand_cond_binary_optab_fn): Likewise.
9333 (direct_cond_binary_optab_supported_p): Likewise.
9334 (get_conditional_internal_fn): New function.
9335 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9336 Cope with reduction statements that are vectorized as calls rather
9338 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9339 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9340 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9341 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9342 (UNSPEC_COND_EOR): New unspecs.
9343 (optab): Add mappings for them.
9344 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9345 (sve_int_op, sve_fp_op): New int attributes.
9347 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9348 Alan Hayward <alan.hayward@arm.com>
9349 David Sherwood <david.sherwood@arm.com>
9351 * optabs.def (while_ult_optab): New optab.
9352 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9353 * internal-fn.def (WHILE_ULT): New internal function.
9354 * internal-fn.h (direct_internal_fn_supported_p): New override
9355 that takes two types as argument.
9356 * internal-fn.c (while_direct): New macro.
9357 (expand_while_optab_fn): New function.
9358 (convert_optab_supported_p): Likewise.
9359 (direct_while_optab_supported_p): New macro.
9360 * wide-int.h (wi::udiv_ceil): New function.
9361 * tree-vectorizer.h (rgroup_masks): New structure.
9362 (vec_loop_masks): New typedef.
9363 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9365 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9366 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9367 (vect_max_vf): New function.
9368 (slpeel_make_loop_iterate_ntimes): Delete.
9369 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9370 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9371 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9372 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9373 internal-fn.h, stor-layout.h and optabs-query.h.
9374 (vect_set_loop_mask): New function.
9375 (add_preheader_seq): Likewise.
9376 (add_header_seq): Likewise.
9377 (interleave_supported_p): Likewise.
9378 (vect_maybe_permute_loop_masks): Likewise.
9379 (vect_set_loop_masks_directly): Likewise.
9380 (vect_set_loop_condition_masked): Likewise.
9381 (vect_set_loop_condition_unmasked): New function, split out from
9382 slpeel_make_loop_iterate_ntimes.
9383 (slpeel_make_loop_iterate_ntimes): Rename to..
9384 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9385 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9386 (vect_do_peeling): Update call accordingly.
9387 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9389 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9390 mask_compare_type, can_fully_mask_p and fully_masked_p.
9391 (release_vec_loop_masks): New function.
9392 (_loop_vec_info): Use it to free the loop masks.
9393 (can_produce_all_loop_masks_p): New function.
9394 (vect_get_max_nscalars_per_iter): Likewise.
9395 (vect_verify_full_masking): Likewise.
9396 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9397 retries, and free the mask rgroups before retrying. Check loop-wide
9398 reasons for disallowing fully-masked loops. Make the final decision
9399 about whether use a fully-masked loop or not.
9400 (vect_estimate_min_profitable_iters): Do not assume that peeling
9401 for the number of iterations will be needed for fully-masked loops.
9402 (vectorizable_reduction): Disable fully-masked loops.
9403 (vectorizable_live_operation): Likewise.
9404 (vect_halve_mask_nunits): New function.
9405 (vect_double_mask_nunits): Likewise.
9406 (vect_record_loop_mask): Likewise.
9407 (vect_get_loop_mask): Likewise.
9408 (vect_transform_loop): Handle the case in which the final loop
9409 iteration might handle a partial vector. Call vect_set_loop_condition
9410 instead of slpeel_make_loop_iterate_ntimes.
9411 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9412 (check_load_store_masking): New function.
9413 (prepare_load_store_mask): Likewise.
9414 (vectorizable_store): Handle fully-masked loops.
9415 (vectorizable_load): Likewise.
9416 (supportable_widening_operation): Use vect_halve_mask_nunits for
9418 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9419 (vect_gen_while): New function.
9420 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9421 (aarch64_uqdec<mode>): New insn.
9423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9424 Alan Hayward <alan.hayward@arm.com>
9425 David Sherwood <david.sherwood@arm.com>
9427 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9428 (reduc_xor_scal_optab): New optabs.
9429 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9430 (reduc_xor_scal_@var{m}): Document.
9431 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9432 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9434 * fold-const-call.c (fold_const_call): Handle them.
9435 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9436 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9437 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9438 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9439 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9440 (UNSPEC_XORV): New unspecs.
9441 (optab): Add entries for them.
9442 (BITWISEV): New int iterator.
9443 (bit_reduc_op): New int attributes.
9445 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9446 Alan Hayward <alan.hayward@arm.com>
9447 David Sherwood <david.sherwood@arm.com>
9449 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9450 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9451 * optabs.def (vec_shl_insert_optab): New optab.
9452 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9453 (duplicate_and_interleave): Likewise.
9454 * tree-vect-loop.c: Include internal-fn.h.
9455 (neutral_op_for_slp_reduction): New function, split out from
9456 get_initial_defs_for_reduction.
9457 (get_initial_def_for_reduction): Handle option 2 for variable-length
9458 vectors by loading the neutral value into a vector and then shifting
9459 the initial value into element 0.
9460 (get_initial_defs_for_reduction): Replace the code argument with
9461 the neutral value calculated by neutral_op_for_slp_reduction.
9462 Use gimple_build_vector for constant-length vectors.
9463 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9464 but the first group_size elements have a neutral value.
9465 Use duplicate_and_interleave otherwise.
9466 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9467 Update call to get_initial_defs_for_reduction. Handle SLP
9468 reductions for variable-length vectors by creating one vector
9469 result for each scalar result, with the elements associated
9470 with other scalar results stubbed out with the neutral value.
9471 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9472 Require IFN_VEC_SHL_INSERT for double reductions on
9473 variable-length vectors, or SLP reductions that have
9474 a neutral value. Require can_duplicate_and_interleave_p
9475 support for variable-length unchained SLP reductions if there
9476 is no neutral value, such as for MIN/MAX reductions. Also require
9477 the number of vector elements to be a multiple of the number of
9478 SLP statements when doing variable-length unchained SLP reductions.
9479 Update call to vect_create_epilog_for_reduction.
9480 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9481 and remove initial values.
9482 (duplicate_and_interleave): Make public.
9483 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9484 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9486 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9487 Alan Hayward <alan.hayward@arm.com>
9488 David Sherwood <david.sherwood@arm.com>
9490 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9491 (can_duplicate_and_interleave_p): New function.
9492 (vect_get_and_check_slp_defs): Take the vector of statements
9493 rather than just the current one. Remove excess parentheses.
9494 Restriction rejectinon of vect_constant_def and vect_external_def
9495 for variable-length vectors to boolean types, or types for which
9496 can_duplicate_and_interleave_p is false.
9497 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9498 (duplicate_and_interleave): New function.
9499 (vect_get_constant_vectors): Use gimple_build_vector for
9500 constant-length vectors and suitable variable-length constant
9501 vectors. Use duplicate_and_interleave for other variable-length
9502 vectors. Don't defer the update when inserting new statements.
9504 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9505 Alan Hayward <alan.hayward@arm.com>
9506 David Sherwood <david.sherwood@arm.com>
9508 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9509 min_profitable_iters doesn't go negative.
9511 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9512 Alan Hayward <alan.hayward@arm.com>
9513 David Sherwood <david.sherwood@arm.com>
9515 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9516 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9517 * optabs.def (vec_mask_load_lanes_optab): New optab.
9518 (vec_mask_store_lanes_optab): Likewise.
9519 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9520 (MASK_STORE_LANES): Likewise.
9521 * internal-fn.c (mask_load_lanes_direct): New macro.
9522 (mask_store_lanes_direct): Likewise.
9523 (expand_mask_load_optab_fn): Handle masked operations.
9524 (expand_mask_load_lanes_optab_fn): New macro.
9525 (expand_mask_store_optab_fn): Handle masked operations.
9526 (expand_mask_store_lanes_optab_fn): New macro.
9527 (direct_mask_load_lanes_optab_supported_p): Likewise.
9528 (direct_mask_store_lanes_optab_supported_p): Likewise.
9529 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9531 (vect_load_lanes_supported): Likewise.
9532 * tree-vect-data-refs.c (strip_conversion): New function.
9533 (can_group_stmts_p): Likewise.
9534 (vect_analyze_data_ref_accesses): Use it instead of checking
9535 for a pair of assignments.
9536 (vect_store_lanes_supported): Take a masked_p parameter.
9537 (vect_load_lanes_supported): Likewise.
9538 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9539 vect_store_lanes_supported and vect_load_lanes_supported.
9540 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9541 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9542 parameter. Don't allow gaps for masked accesses.
9543 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9544 and vect_load_lanes_supported.
9545 (get_load_store_type): Take a masked_p parameter and update
9546 call to get_group_load_store_type.
9547 (vectorizable_store): Update call to get_load_store_type.
9548 Handle IFN_MASK_STORE_LANES.
9549 (vectorizable_load): Update call to get_load_store_type.
9550 Handle IFN_MASK_LOAD_LANES.
9552 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9553 Alan Hayward <alan.hayward@arm.com>
9554 David Sherwood <david.sherwood@arm.com>
9556 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9558 * config/aarch64/aarch64-protos.h
9559 (aarch64_sve_struct_memory_operand_p): Declare.
9560 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9561 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9562 (VPRED, vpred): Handle SVE structure modes.
9563 * config/aarch64/constraints.md (Utx): New constraint.
9564 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9565 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9566 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9567 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9568 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9569 structure modes. Split into pieces after RA.
9570 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9571 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9573 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9574 SVE structure modes.
9575 (aarch64_classify_address): Likewise.
9576 (sizetochar): Move earlier in file.
9577 (aarch64_print_operand): Handle SVE register lists.
9578 (aarch64_array_mode): New function.
9579 (aarch64_sve_struct_memory_operand_p): Likewise.
9580 (TARGET_ARRAY_MODE): Redefine.
9582 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9583 Alan Hayward <alan.hayward@arm.com>
9584 David Sherwood <david.sherwood@arm.com>
9586 * target.def (array_mode): New target hook.
9587 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9588 * doc/tm.texi: Regenerate.
9589 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9590 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9591 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9593 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9596 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9597 Alan Hayward <alan.hayward@arm.com>
9598 David Sherwood <david.sherwood@arm.com>
9600 * fold-const.c (fold_binary_loc): Check the argument types
9601 rather than the result type when testing for a vector operation.
9603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9605 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9606 * doc/tm.texi: Regenerate.
9608 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9609 Alan Hayward <alan.hayward@arm.com>
9610 David Sherwood <david.sherwood@arm.com>
9612 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9613 (sve): Document new AArch64 extension.
9614 * doc/md.texi (w): Extend the description of the AArch64
9615 constraint to include SVE vectors.
9616 (Upl, Upa): Document new AArch64 predicate constraints.
9617 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9619 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9620 (msve-vector-bits=): New option.
9621 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9622 SVE when these are disabled.
9623 (sve): New extension.
9624 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9625 modes. Adjust their number of units based on aarch64_sve_vg.
9626 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9627 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9628 aarch64_addr_query_type.
9629 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9630 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9631 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9632 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9633 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9634 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9635 (aarch64_simd_imm_zero_p): Delete.
9636 (aarch64_check_zero_based_sve_index_immediate): Declare.
9637 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9638 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9639 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9640 (aarch64_sve_float_mul_immediate_p): Likewise.
9641 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9643 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9644 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9645 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9646 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9647 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9648 (aarch64_regmode_natural_size): Likewise.
9649 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9650 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9652 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9653 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9654 for VG and the SVE predicate registers.
9655 (V_ALIASES): Add a "z"-prefixed alias.
9656 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9657 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9658 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9659 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9660 (REG_CLASS_NAMES): Add entries for them.
9661 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9662 and the predicate registers.
9663 (aarch64_sve_vg): Declare.
9664 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9665 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9666 (REGMODE_NATURAL_SIZE): Define.
9667 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9669 * config/aarch64/aarch64.c: Include cfgrtl.h.
9670 (simd_immediate_info): Add a constructor for series vectors,
9671 and an associated step field.
9672 (aarch64_sve_vg): New variable.
9673 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9674 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9675 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9676 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9677 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9678 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9679 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9680 (aarch64_get_mask_mode): New functions.
9681 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9682 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9683 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9684 predicate modes and predicate registers. Explicitly restrict
9685 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9686 to store a vector mode if it is recognized by
9687 aarch64_classify_vector_mode.
9688 (aarch64_regmode_natural_size): New function.
9689 (aarch64_hard_regno_caller_save_mode): Return the original mode
9691 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9692 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9693 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9694 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9696 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9697 does not overlap dest if the function is frame-related. Handle
9699 (aarch64_split_add_offset): New function.
9700 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9701 them aarch64_add_offset.
9702 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9703 and update call to aarch64_sub_sp.
9704 (aarch64_add_cfa_expression): New function.
9705 (aarch64_expand_prologue): Pass extra temporary registers to the
9706 functions above. Handle the case in which we need to emit new
9707 DW_CFA_expressions for registers that were originally saved
9708 relative to the stack pointer, but now have to be expressed
9709 relative to the frame pointer.
9710 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9712 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9713 IP0 and IP1 values for SVE frames.
9714 (aarch64_expand_vec_series): New function.
9715 (aarch64_expand_sve_widened_duplicate): Likewise.
9716 (aarch64_expand_sve_const_vector): Likewise.
9717 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9718 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9719 into the register, rather than emitting a SET directly.
9720 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9721 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9722 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9723 (offset_9bit_signed_scaled_p): New functions.
9724 (aarch64_replicate_bitmask_imm): New function.
9725 (aarch64_bitmask_imm): Use it.
9726 (aarch64_cannot_force_const_mem): Reject expressions involving
9727 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9728 (aarch64_classify_index): Handle SVE indices, by requiring
9729 a plain register index with a scale that matches the element size.
9730 (aarch64_classify_address): Handle SVE addresses. Assert that
9731 the mode of the address is VOIDmode or an integer mode.
9732 Update call to aarch64_classify_symbol.
9733 (aarch64_classify_symbolic_expression): Update call to
9734 aarch64_classify_symbol.
9735 (aarch64_const_vec_all_in_range_p): New function.
9736 (aarch64_print_vector_float_operand): Likewise.
9737 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9738 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9739 and the FP immediates 1.0 and 0.5.
9740 (aarch64_print_address_internal): Handle SVE addresses.
9741 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9742 (aarch64_regno_regclass): Handle predicate registers.
9743 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9745 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9746 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9747 (aarch64_convert_sve_vector_bits): New function.
9748 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9749 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9751 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9752 Handle SVE vector and predicate modes. Accept VL-based constants
9753 that need only one temporary register, and VL offsets that require
9754 no temporary registers.
9755 (aarch64_conditional_register_usage): Mark the predicate registers
9756 as fixed if SVE isn't available.
9757 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9758 Return true for SVE vector and predicate modes.
9759 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9760 rather than an unsigned int. Handle SVE modes.
9761 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9763 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9765 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9766 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9767 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9768 (aarch64_sve_float_mul_immediate_p): New functions.
9769 (aarch64_sve_valid_immediate): New function.
9770 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9771 Explicitly reject structure modes. Check for INDEX constants.
9772 Handle PTRUE and PFALSE constants.
9773 (aarch64_check_zero_based_sve_index_immediate): New function.
9774 (aarch64_simd_imm_zero_p): Delete.
9775 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9776 vector modes. Accept constants in the range of CNT[BHWD].
9777 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9778 ask for an Advanced SIMD mode.
9779 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9780 (aarch64_simd_vector_alignment): Handle SVE predicates.
9781 (aarch64_vectorize_preferred_vector_alignment): New function.
9782 (aarch64_simd_vector_alignment_reachable): Use it instead of
9784 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9785 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9787 (MAX_VECT_LEN): Delete.
9788 (expand_vec_perm_d): Add a vec_flags field.
9789 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9790 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9791 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9793 (aarch64_evpc_rev): Rename to...
9794 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9795 (aarch64_evpc_rev_global): New function.
9796 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9797 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9799 (aarch64_evpc_sve_tbl): New function.
9800 (aarch64_expand_vec_perm_const_1): Update after rename of
9801 aarch64_evpc_rev. Handle SVE permutes too, trying
9802 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9803 than aarch64_evpc_tbl.
9804 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9805 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9806 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9807 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9808 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9809 (aarch64_expand_sve_vcond): New functions.
9810 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9811 of aarch64_vector_mode_p.
9812 (aarch64_dwarf_poly_indeterminate_value): New function.
9813 (aarch64_compute_pressure_classes): Likewise.
9814 (aarch64_can_change_mode_class): Likewise.
9815 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9816 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9817 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9818 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9819 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9820 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9821 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9822 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9824 (Dn, Dl, Dr): Accept const as well as const_vector.
9825 (Dz): Likewise. Compare against CONST0_RTX.
9826 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9827 of "vector" where appropriate.
9828 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9829 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9830 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9831 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9832 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9833 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9834 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9835 (v_int_equiv): Extend to SVE modes.
9836 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9838 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9839 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9840 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9841 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9842 (SVE_COND_FP_CMP): New int iterators.
9843 (perm_hilo): Handle the new unpack unspecs.
9844 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9846 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9847 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9848 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9849 (aarch64_equality_operator, aarch64_constant_vector_operand)
9850 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9851 (aarch64_sve_nonimmediate_operand): Likewise.
9852 (aarch64_sve_general_operand): Likewise.
9853 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9854 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9855 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9856 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9857 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9858 (aarch64_sve_float_arith_immediate): Likewise.
9859 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9860 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9861 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9862 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9863 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9864 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9865 (aarch64_sve_float_arith_operand): Likewise.
9866 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9867 (aarch64_sve_float_mul_operand): Likewise.
9868 (aarch64_sve_vec_perm_operand): Likewise.
9869 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9870 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9871 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9872 as well as const_vector.
9873 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9874 in file. Use CONST0_RTX and CONSTM1_RTX.
9875 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9876 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9877 Use aarch64_simd_imm_zero.
9878 * config/aarch64/aarch64-sve.md: New file.
9879 * config/aarch64/aarch64.md: Include it.
9880 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9881 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9882 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9883 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9884 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9885 (sve): New attribute.
9886 (enabled): Disable instructions with the sve attribute unless
9888 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9889 aarch64_expand_mov_immediate.
9890 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9891 CNT[BHSD] immediates.
9892 (movti): Split CONST_POLY_INT moves into two halves.
9893 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9894 Split additions that need a temporary here if the destination
9895 is the stack pointer.
9896 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9897 (*add<mode>3_poly_1): New instruction.
9898 (set_clobber_cc): New expander.
9900 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9902 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9903 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9904 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9905 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9906 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9907 Change innermode from fixed_mode_size to machine_mode.
9908 (simplify_subreg): Update call accordingly. Handle a constant-sized
9909 subreg of a variable-length CONST_VECTOR.
9911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9912 Alan Hayward <alan.hayward@arm.com>
9913 David Sherwood <david.sherwood@arm.com>
9915 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9916 (add_offset_to_base): New function, split out from...
9917 (create_mem_ref): ...here. When handling a scale other than 1,
9918 check first whether the address is valid without the offset.
9919 Add it into the base if so, leaving the index and scale as-is.
9921 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9924 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9925 fold_for_warn before checking if arg2 is INTEGER_CST.
9927 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9929 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9930 (store_multiple_operation): Delete.
9931 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9932 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9933 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9934 guarded by TARGET_STRING.
9935 (rs6000_output_load_multiple): Delete.
9936 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9937 OPTION_MASK_STRING / TARGET_STRING handling.
9938 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9939 (const rs6000_opt_masks) <"string">: Change mask to 0.
9940 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9941 (MASK_STRING): Delete.
9942 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9944 (load_multiple): Delete.
9951 (store_multiple): Delete.
9958 (movmemsi_8reg): Delete.
9959 (corresponding unnamed define_insn): Delete.
9960 (movmemsi_6reg): Delete.
9961 (corresponding unnamed define_insn): Delete.
9962 (movmemsi_4reg): Delete.
9963 (corresponding unnamed define_insn): Delete.
9964 (movmemsi_2reg): Delete.
9965 (corresponding unnamed define_insn): Delete.
9966 (movmemsi_1reg): Delete.
9967 (corresponding unnamed define_insn): Delete.
9968 * config/rs6000/rs6000.opt (mno-string): New.
9969 (mstring): Replace by deprecation warning stub.
9970 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9972 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9974 * regrename.c (regrename_do_replace): If replacing the same
9975 reg multiple times, try to reuse last created gen_raw_REG.
9978 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9979 main to workaround a bug in GDB.
9981 2018-01-12 Tom de Vries <tom@codesourcery.com>
9984 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9986 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9988 PR rtl-optimization/80481
9989 * ira-color.c (get_cap_member): New function.
9990 (allocnos_conflict_by_live_ranges_p): Use it.
9991 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9992 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9994 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9997 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9998 (*saddl_se_1): Ditto.
10000 (*ssubl_se_1): Ditto.
10002 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10004 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10005 rather than wi::to_widest for DR_INITs.
10006 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10007 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10008 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10010 (vect_analyze_group_access_1): Note that here.
10012 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10014 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10015 polynomial type sizes.
10017 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10019 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10020 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10021 (gimple_add_tmp_var): Likewise.
10023 2018-01-12 Martin Liska <mliska@suse.cz>
10025 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10026 (gimple_alloc_sizes): Likewise.
10027 (dump_gimple_statistics): Use PRIu64 in printf format.
10028 * gimple.h: Change uint64_t to int.
10030 2018-01-12 Martin Liska <mliska@suse.cz>
10032 * tree-core.h: Use uint64_t instead of int.
10033 * tree.c (tree_node_counts): Likewise.
10034 (tree_node_sizes): Likewise.
10035 (dump_tree_statistics): Use PRIu64 in printf format.
10037 2018-01-12 Martin Liska <mliska@suse.cz>
10039 * Makefile.in: As qsort_chk is implemented in vec.c, add
10040 vec.o to linkage of gencfn-macros.
10041 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10042 passing the info to record_node_allocation_statistics.
10043 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10045 * ggc-common.c (struct ggc_usage): Add operator== and use
10046 it in operator< and compare function.
10047 * mem-stats.h (struct mem_usage): Likewise.
10048 * vec.c (struct vec_usage): Remove operator< and compare
10049 function. Can be simply inherited.
10051 2018-01-12 Martin Jambor <mjambor@suse.cz>
10054 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10055 * tree-ssa-math-opts.c: Include domwalk.h.
10056 (convert_mult_to_fma_1): New function.
10057 (fma_transformation_info): New type.
10058 (fma_deferring_state): Likewise.
10059 (cancel_fma_deferring): New function.
10060 (result_of_phi): Likewise.
10061 (last_fma_candidate_feeds_initial_phi): Likewise.
10062 (convert_mult_to_fma): Added deferring logic, split actual
10063 transformation to convert_mult_to_fma_1.
10064 (math_opts_dom_walker): New type.
10065 (math_opts_dom_walker::after_dom_children): New method, body moved
10066 here from pass_optimize_widening_mul::execute, added deferring logic
10068 (pass_optimize_widening_mul::execute): Moved most of code to
10069 math_opts_dom_walker::after_dom_children.
10070 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10071 * config/i386/i386.c (ix86_option_override_internal): Added
10072 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10074 2018-01-12 Richard Biener <rguenther@suse.de>
10077 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10078 inline instance vars.
10080 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10083 * config/rx/rx.c (rx_is_restricted_memory_address):
10084 Handle SUBREG case.
10086 2018-01-12 Richard Biener <rguenther@suse.de>
10088 PR tree-optimization/80846
10089 * target.def (split_reduction): New target hook.
10090 * targhooks.c (default_split_reduction): New function.
10091 * targhooks.h (default_split_reduction): Declare.
10092 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10093 target requests first reduce vectors by combining low and high
10095 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10096 (get_vectype_for_scalar_type_and_size): Export.
10097 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10098 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10099 * doc/tm.texi: Regenerate.
10100 * config/i386/i386.c (ix86_split_reduction): Implement
10101 TARGET_VECTORIZE_SPLIT_REDUCTION.
10103 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10106 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10107 in PIC mode except for TARGET_VXWORKS_RTP.
10108 * config/sparc/sparc.c: Include cfgrtl.h.
10109 (TARGET_INIT_PIC_REG): Define.
10110 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10111 (sparc_pic_register_p): New predicate.
10112 (sparc_legitimate_address_p): Use it.
10113 (sparc_legitimize_pic_address): Likewise.
10114 (sparc_delegitimize_address): Likewise.
10115 (sparc_mode_dependent_address_p): Likewise.
10116 (gen_load_pcrel_sym): Remove 4th parameter.
10117 (load_got_register): Adjust call to above. Remove obsolete stuff.
10118 (sparc_expand_prologue): Do not call load_got_register here.
10119 (sparc_flat_expand_prologue): Likewise.
10120 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10121 (sparc_use_pseudo_pic_reg): New function.
10122 (sparc_init_pic_reg): Likewise.
10123 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10124 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10126 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10128 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10129 Add item for branch_cost.
10131 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10133 PR rtl-optimization/83565
10134 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10135 not extend the result to a larger mode for rotate operations.
10136 (num_sign_bit_copies1): Likewise.
10138 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10141 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10143 Use values-Xc.o for -pedantic.
10144 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10146 2018-01-12 Martin Liska <mliska@suse.cz>
10149 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10151 (possible_polymorphic_call_targets): Use it.
10152 (ipa_devirt): Likewise.
10154 2018-01-12 Martin Liska <mliska@suse.cz>
10156 * profile-count.h (enum profile_quality): Use 0 as invalid
10157 enum value of profile_quality.
10159 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10161 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10162 -mext-string options.
10164 2018-01-12 Richard Biener <rguenther@suse.de>
10166 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10167 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10168 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10170 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10172 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10174 * configure.ac (--with-long-double-format): Add support for the
10175 configuration option to change the default long double format on
10177 * config.gcc (powerpc*-linux*-*): Likewise.
10178 * configure: Regenerate.
10179 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10180 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10181 used without modification.
10183 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10185 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10186 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10187 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10188 MISC_BUILTIN_SPEC_BARRIER.
10189 (rs6000_init_builtins): Likewise.
10190 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10192 (speculation_barrier): New define_insn.
10193 * doc/extend.texi: Document __builtin_speculation_barrier.
10195 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10198 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10199 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10200 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10202 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10203 integral modes instead of "ss" and "sd".
10204 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10205 vectors with 32-bit and 64-bit elements.
10206 (vecdupssescalarmodesuffix): New mode attribute.
10207 (vec_dup<mode>): Use it.
10209 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10212 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10213 frame if argument is passed on stack.
10215 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10218 * ree.c (combine_reaching_defs): Optimize also
10219 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10220 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10222 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10224 PR middle-end/83189
10225 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10227 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10229 PR middle-end/83718
10230 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10231 after they are computed.
10233 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10235 PR tree-optimization/83695
10236 * gimple-loop-linterchange.cc
10237 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10238 reset cached scev information after interchange.
10239 (pass_linterchange::execute): Remove call to scev_reset_htab.
10241 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10243 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10244 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10245 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10246 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10247 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10248 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10249 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10250 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10251 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10252 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10253 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10254 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10255 (V_lane_reg): Likewise.
10256 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10258 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10259 (vfmal_lane_low<mode>_intrinsic,
10260 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10261 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10262 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10263 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10264 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10265 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10267 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10269 * config/arm/arm-cpus.in (fp16fml): New feature.
10270 (ALL_SIMD): Add fp16fml.
10271 (armv8.2-a): Add fp16fml as an option.
10272 (armv8.3-a): Likewise.
10273 (armv8.4-a): Add fp16fml as part of fp16.
10274 * config/arm/arm.h (TARGET_FP16FML): Define.
10275 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10277 * config/arm/arm-modes.def (V2HF): Define.
10278 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10279 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10280 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10281 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10282 vfmsl_low, vfmsl_high): New set of builtins.
10283 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10284 (vfml_op): New code attribute.
10285 (VFMLHALVES): New int iterator.
10286 (VFML, VFMLSEL): New mode attributes.
10287 (V_reg): Define mapping for V2HF.
10288 (V_hi, V_lo): New mode attributes.
10289 (VF_constraint): Likewise.
10290 (vfml_half, vfml_half_selector): New int attributes.
10291 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10293 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10294 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10296 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10297 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10298 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10299 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10301 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10302 Document new effective target and option set.
10304 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10306 * config/arm/arm-cpus.in (armv8_4): New feature.
10307 (ARMv8_4a): New fgroup.
10308 (armv8.4-a): New arch.
10309 * config/arm/arm-tables.opt: Regenerate.
10310 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10311 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10312 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10313 Add matching rules for -march=armv8.4-a and extensions.
10314 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10316 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10319 * config/rx/rx.md (BW): New mode attribute.
10320 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10322 2018-01-11 Richard Biener <rguenther@suse.de>
10324 PR tree-optimization/83435
10325 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10326 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10327 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10329 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10330 Alan Hayward <alan.hayward@arm.com>
10331 David Sherwood <david.sherwood@arm.com>
10333 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10335 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10336 (aarch64_print_address_internal): Use it to check for a zero offset.
10338 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10339 Alan Hayward <alan.hayward@arm.com>
10340 David Sherwood <david.sherwood@arm.com>
10342 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10343 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10344 Return a poly_int64 rather than a HOST_WIDE_INT.
10345 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10346 rather than a HOST_WIDE_INT.
10347 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10348 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10349 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10350 final_offset from HOST_WIDE_INT to poly_int64.
10351 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10352 to_constant when getting the number of units in an Advanced SIMD
10354 (aarch64_builtin_vectorized_function): Check for a constant number
10356 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10358 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10359 attribute instead of GET_MODE_NUNITS.
10360 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10361 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10362 GET_MODE_SIZE for fixed-size registers.
10363 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10364 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10365 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10366 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10367 (aarch64_print_operand, aarch64_print_address_internal)
10368 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10369 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10370 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10371 Handle polynomial GET_MODE_SIZE.
10372 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10373 wider than SImode without modification.
10374 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10375 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10376 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10377 passing and returning SVE modes.
10378 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10379 rather than GEN_INT.
10380 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10381 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10382 (aarch64_allocate_and_probe_stack_space): Likewise.
10383 (aarch64_layout_frame): Cope with polynomial offsets.
10384 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10385 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10386 polynomial offsets.
10387 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10388 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10389 poly_int64 rather than a HOST_WIDE_INT.
10390 (aarch64_get_separate_components, aarch64_process_components)
10391 (aarch64_expand_prologue, aarch64_expand_epilogue)
10392 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10393 (aarch64_anchor_offset): New function, split out from...
10394 (aarch64_legitimize_address): ...here.
10395 (aarch64_builtin_vectorization_cost): Handle polynomial
10396 TYPE_VECTOR_SUBPARTS.
10397 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10399 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10400 number of elements from the PARALLEL rather than the mode.
10401 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10402 rather than GET_MODE_BITSIZE.
10403 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10404 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10405 (aarch64_expand_vec_perm_const_1): Handle polynomial
10406 d->perm.length () and d->perm elements.
10407 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10408 Apply to_constant to d->perm elements.
10409 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10410 polynomial CONST_VECTOR_NUNITS.
10411 (aarch64_move_pointer): Take amount as a poly_int64 rather
10413 (aarch64_progress_pointer): Avoid temporary variable.
10414 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10415 the mode attribute instead of GET_MODE.
10417 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10418 Alan Hayward <alan.hayward@arm.com>
10419 David Sherwood <david.sherwood@arm.com>
10421 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10422 x exists before using it.
10423 (aarch64_add_constant_internal): Rename to...
10424 (aarch64_add_offset_1): ...this. Replace regnum with separate
10425 src and dest rtxes. Handle the case in which they're different,
10426 including when the offset is zero. Replace scratchreg with an rtx.
10427 Use 2 additions if there is no spare register into which we can
10428 move a 16-bit constant.
10429 (aarch64_add_constant): Delete.
10430 (aarch64_add_offset): Replace reg with separate src and dest
10431 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10432 Use aarch64_add_offset_1.
10433 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10434 an rtx rather than an int. Take the delta as a poly_int64
10435 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10436 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10437 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10438 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10439 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10440 and aarch64_add_sp.
10441 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10442 aarch64_add_constant.
10444 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10446 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10447 Use scalar_float_mode.
10449 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10451 * config/aarch64/aarch64-simd.md
10452 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10453 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10454 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10455 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10456 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10457 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10458 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10459 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10460 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10461 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10463 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10466 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10467 targ_options->x_arm_arch_string is non NULL.
10469 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10471 * config/aarch64/aarch64.h
10472 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10474 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10477 * expmed.c (emit_store_flag_force): Swap if const op0
10478 and change VOIDmode to mode of op0.
10480 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10482 PR rtl-optimization/83761
10483 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10484 than bytes to mode_for_size.
10486 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10488 PR middle-end/83189
10489 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10490 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10493 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10495 PR middle-end/83575
10496 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10497 when in layout mode.
10498 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10499 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10502 2018-01-10 Michael Collison <michael.collison@arm.com>
10504 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10505 * config/aarch64/aarch64-option-extension.def: Add
10506 AARCH64_OPT_EXTENSION of 'fp16fml'.
10507 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10508 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10509 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10510 * config/aarch64/constraints.md (Ui7): New constraint.
10511 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10512 (VFMLA_SEL_W): Ditto.
10515 (VFMLA16_LOW): New int iterator.
10516 (VFMLA16_HIGH): Ditto.
10517 (UNSPEC_FMLAL): New unspec.
10518 (UNSPEC_FMLSL): Ditto.
10519 (UNSPEC_FMLAL2): Ditto.
10520 (UNSPEC_FMLSL2): Ditto.
10521 (f16mac): New code attribute.
10522 * config/aarch64/aarch64-simd-builtins.def
10523 (aarch64_fmlal_lowv2sf): Ditto.
10524 (aarch64_fmlsl_lowv2sf): Ditto.
10525 (aarch64_fmlalq_lowv4sf): Ditto.
10526 (aarch64_fmlslq_lowv4sf): Ditto.
10527 (aarch64_fmlal_highv2sf): Ditto.
10528 (aarch64_fmlsl_highv2sf): Ditto.
10529 (aarch64_fmlalq_highv4sf): Ditto.
10530 (aarch64_fmlslq_highv4sf): Ditto.
10531 (aarch64_fmlal_lane_lowv2sf): Ditto.
10532 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10533 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10534 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10535 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10536 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10537 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10538 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10539 (aarch64_fmlal_lane_highv2sf): Ditto.
10540 (aarch64_fmlsl_lane_highv2sf): Ditto.
10541 (aarch64_fmlal_laneq_highv2sf): Ditto.
10542 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10543 (aarch64_fmlalq_lane_highv4sf): Ditto.
10544 (aarch64_fmlsl_lane_highv4sf): Ditto.
10545 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10546 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10547 * config/aarch64/aarch64-simd.md:
10548 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10549 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10550 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10551 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10552 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10553 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10554 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10555 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10556 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10557 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10558 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10559 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10560 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10561 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10562 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10563 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10564 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10565 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10566 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10567 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10568 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10569 (vfmlsl_low_u32): Ditto.
10570 (vfmlalq_low_u32): Ditto.
10571 (vfmlslq_low_u32): Ditto.
10572 (vfmlal_high_u32): Ditto.
10573 (vfmlsl_high_u32): Ditto.
10574 (vfmlalq_high_u32): Ditto.
10575 (vfmlslq_high_u32): Ditto.
10576 (vfmlal_lane_low_u32): Ditto.
10577 (vfmlsl_lane_low_u32): Ditto.
10578 (vfmlal_laneq_low_u32): Ditto.
10579 (vfmlsl_laneq_low_u32): Ditto.
10580 (vfmlalq_lane_low_u32): Ditto.
10581 (vfmlslq_lane_low_u32): Ditto.
10582 (vfmlalq_laneq_low_u32): Ditto.
10583 (vfmlslq_laneq_low_u32): Ditto.
10584 (vfmlal_lane_high_u32): Ditto.
10585 (vfmlsl_lane_high_u32): Ditto.
10586 (vfmlal_laneq_high_u32): Ditto.
10587 (vfmlsl_laneq_high_u32): Ditto.
10588 (vfmlalq_lane_high_u32): Ditto.
10589 (vfmlslq_lane_high_u32): Ditto.
10590 (vfmlalq_laneq_high_u32): Ditto.
10591 (vfmlslq_laneq_high_u32): Ditto.
10592 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10593 (AARCH64_FL_FOR_ARCH8_4): New.
10594 (AARCH64_ISA_F16FML): New ISA flag.
10595 (TARGET_F16FML): New feature flag for fp16fml.
10596 (doc/invoke.texi): Document new fp16fml option.
10598 2018-01-10 Michael Collison <michael.collison@arm.com>
10600 * config/aarch64/aarch64-builtins.c:
10601 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10602 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10603 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10604 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10605 (AARCH64_ISA_SHA3): New ISA flag.
10606 (TARGET_SHA3): New feature flag for sha3.
10607 * config/aarch64/iterators.md (sha512_op): New int attribute.
10608 (CRYPTO_SHA512): New int iterator.
10609 (UNSPEC_SHA512H): New unspec.
10610 (UNSPEC_SHA512H2): Ditto.
10611 (UNSPEC_SHA512SU0): Ditto.
10612 (UNSPEC_SHA512SU1): Ditto.
10613 * config/aarch64/aarch64-simd-builtins.def
10614 (aarch64_crypto_sha512hqv2di): New builtin.
10615 (aarch64_crypto_sha512h2qv2di): Ditto.
10616 (aarch64_crypto_sha512su0qv2di): Ditto.
10617 (aarch64_crypto_sha512su1qv2di): Ditto.
10618 (aarch64_eor3qv8hi): Ditto.
10619 (aarch64_rax1qv2di): Ditto.
10620 (aarch64_xarqv2di): Ditto.
10621 (aarch64_bcaxqv8hi): Ditto.
10622 * config/aarch64/aarch64-simd.md:
10623 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10624 (aarch64_crypto_sha512su0qv2di): Ditto.
10625 (aarch64_crypto_sha512su1qv2di): Ditto.
10626 (aarch64_eor3qv8hi): Ditto.
10627 (aarch64_rax1qv2di): Ditto.
10628 (aarch64_xarqv2di): Ditto.
10629 (aarch64_bcaxqv8hi): Ditto.
10630 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10631 (vsha512h2q_u64): Ditto.
10632 (vsha512su0q_u64): Ditto.
10633 (vsha512su1q_u64): Ditto.
10634 (veor3q_u16): Ditto.
10635 (vrax1q_u64): Ditto.
10636 (vxarq_u64): Ditto.
10637 (vbcaxq_u16): Ditto.
10638 * config/arm/types.md (crypto_sha512): New type attribute.
10639 (crypto_sha3): Ditto.
10640 (doc/invoke.texi): Document new sha3 option.
10642 2018-01-10 Michael Collison <michael.collison@arm.com>
10644 * config/aarch64/aarch64-builtins.c:
10645 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10646 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10647 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10648 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10649 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10650 (AARCH64_ISA_SM4): New ISA flag.
10651 (TARGET_SM4): New feature flag for sm4.
10652 * config/aarch64/aarch64-simd-builtins.def
10653 (aarch64_sm3ss1qv4si): Ditto.
10654 (aarch64_sm3tt1aq4si): Ditto.
10655 (aarch64_sm3tt1bq4si): Ditto.
10656 (aarch64_sm3tt2aq4si): Ditto.
10657 (aarch64_sm3tt2bq4si): Ditto.
10658 (aarch64_sm3partw1qv4si): Ditto.
10659 (aarch64_sm3partw2qv4si): Ditto.
10660 (aarch64_sm4eqv4si): Ditto.
10661 (aarch64_sm4ekeyqv4si): Ditto.
10662 * config/aarch64/aarch64-simd.md:
10663 (aarch64_sm3ss1qv4si): Ditto.
10664 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10665 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10666 (aarch64_sm4eqv4si): Ditto.
10667 (aarch64_sm4ekeyqv4si): Ditto.
10668 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10669 (sm3part_op): Ditto.
10670 (CRYPTO_SM3TT): Ditto.
10671 (CRYPTO_SM3PART): Ditto.
10672 (UNSPEC_SM3SS1): New unspec.
10673 (UNSPEC_SM3TT1A): Ditto.
10674 (UNSPEC_SM3TT1B): Ditto.
10675 (UNSPEC_SM3TT2A): Ditto.
10676 (UNSPEC_SM3TT2B): Ditto.
10677 (UNSPEC_SM3PARTW1): Ditto.
10678 (UNSPEC_SM3PARTW2): Ditto.
10679 (UNSPEC_SM4E): Ditto.
10680 (UNSPEC_SM4EKEY): Ditto.
10681 * config/aarch64/constraints.md (Ui2): New constraint.
10682 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10683 * config/arm/types.md (crypto_sm3): New type attribute.
10684 (crypto_sm4): Ditto.
10685 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10686 (vsm3tt1aq_u32): Ditto.
10687 (vsm3tt1bq_u32): Ditto.
10688 (vsm3tt2aq_u32): Ditto.
10689 (vsm3tt2bq_u32): Ditto.
10690 (vsm3partw1q_u32): Ditto.
10691 (vsm3partw2q_u32): Ditto.
10692 (vsm4eq_u32): Ditto.
10693 (vsm4ekeyq_u32): Ditto.
10694 (doc/invoke.texi): Document new sm4 option.
10696 2018-01-10 Michael Collison <michael.collison@arm.com>
10698 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10699 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10700 (AARCH64_FL_FOR_ARCH8_4): New.
10701 (AARCH64_FL_V8_4): New flag.
10702 (doc/invoke.texi): Document new armv8.4-a option.
10704 2018-01-10 Michael Collison <michael.collison@arm.com>
10706 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10707 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10708 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10709 * config/aarch64/aarch64-option-extension.def: Add
10710 AARCH64_OPT_EXTENSION of 'sha2'.
10711 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10712 (crypto): Disable sha2 and aes if crypto disabled.
10713 (crypto): Enable aes and sha2 if enabled.
10714 (simd): Disable sha2 and aes if simd disabled.
10715 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10717 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10718 (TARGET_SHA2): New feature flag for sha2.
10719 (TARGET_AES): New feature flag for aes.
10720 * config/aarch64/aarch64-simd.md:
10721 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10722 conditional on TARGET_AES.
10723 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10724 (aarch64_crypto_sha1hsi): Make pattern conditional
10726 (aarch64_crypto_sha1hv4si): Ditto.
10727 (aarch64_be_crypto_sha1hv4si): Ditto.
10728 (aarch64_crypto_sha1su1v4si): Ditto.
10729 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10730 (aarch64_crypto_sha1su0v4si): Ditto.
10731 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10732 (aarch64_crypto_sha256su0v4si): Ditto.
10733 (aarch64_crypto_sha256su1v4si): Ditto.
10734 (doc/invoke.texi): Document new aes and sha2 options.
10736 2018-01-10 Martin Sebor <msebor@redhat.com>
10738 PR tree-optimization/83781
10739 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10742 2018-01-11 Martin Sebor <msebor@gmail.com>
10743 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10745 PR tree-optimization/83501
10746 PR tree-optimization/81703
10748 * tree-ssa-strlen.c (get_string_cst): Rename...
10749 (get_string_len): ...to this. Handle global constants.
10750 (handle_char_store): Adjust.
10752 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10753 Jim Wilson <jimw@sifive.com>
10755 * config/riscv/riscv-protos.h (riscv_output_return): New.
10756 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10757 (riscv_attribute_table, riscv_output_return),
10758 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10759 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10760 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10762 (riscv_expand_prologue): Add early return for naked function.
10763 (riscv_expand_epilogue): Likewise.
10764 (riscv_function_ok_for_sibcall): Return false for naked function.
10765 (riscv_set_current_function): New.
10766 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10767 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10768 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10769 * doc/extend.texi (RISC-V Function Attributes): New.
10771 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10773 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10774 check for 128-bit long double before checking TCmode.
10775 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10776 128-bit long doubles before checking TFmode or TCmode.
10777 (FLOAT128_IBM_P): Likewise.
10779 2018-01-10 Martin Sebor <msebor@redhat.com>
10781 PR tree-optimization/83671
10782 * builtins.c (c_strlen): Unconditionally return zero for the empty
10784 Use -Warray-bounds for warnings.
10785 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10786 for non-constant array indices with COMPONENT_REF, arrays of
10787 arrays, and pointers to arrays.
10788 (gimple_fold_builtin_strlen): Determine and set length range for
10789 non-constant character arrays.
10791 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10793 PR middle-end/81897
10794 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10797 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10799 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10801 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10804 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10805 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10806 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10807 indexed_or_indirect_operand predicate.
10808 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10809 (*vsx_le_perm_load_v8hi): Likewise.
10810 (*vsx_le_perm_load_v16qi): Likewise.
10811 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10812 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10813 (*vsx_le_perm_store_v8hi): Likewise.
10814 (*vsx_le_perm_store_v16qi): Likewise.
10815 (eight unnamed splitters): Likewise.
10817 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10819 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10820 * config/rs6000/emmintrin.h: Likewise.
10821 * config/rs6000/mmintrin.h: Likewise.
10822 * config/rs6000/xmmintrin.h: Likewise.
10824 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10827 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10829 * tree.c (tree_nop_conversion): Return true for location wrapper
10831 (maybe_wrap_with_location): New function.
10832 (selftest::check_strip_nops): New function.
10833 (selftest::test_location_wrappers): New function.
10834 (selftest::tree_c_tests): Call it.
10835 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10836 (maybe_wrap_with_location): New decl.
10837 (EXPR_LOCATION_WRAPPER_P): New macro.
10838 (location_wrapper_p): New inline function.
10839 (tree_strip_any_location_wrapper): New inline function.
10841 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10844 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10845 stack_realign_offset for the largest alignment of stack slot
10847 (ix86_find_max_used_stack_alignment): New function.
10848 (ix86_finalize_stack_frame_flags): Use it. Set
10849 max_used_stack_alignment if we don't realign stack.
10850 * config/i386/i386.h (machine_function): Add
10851 max_used_stack_alignment.
10853 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10855 * config/arm/arm.opt (-mbranch-cost): New option.
10856 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10859 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10862 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10863 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10865 2018-01-10 Richard Biener <rguenther@suse.de>
10868 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10869 early out so it also covers the case where we have a non-NULL
10872 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10874 PR tree-optimization/83753
10875 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10876 for non-strided grouped accesses if the number of elements is 1.
10878 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10881 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10882 * i386.h (TARGET_USE_GATHER): Define.
10883 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10885 2018-01-10 Martin Liska <mliska@suse.cz>
10888 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10889 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10891 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10892 CLEANUP_NO_PARTITIONING is not set.
10894 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10896 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10897 for vectors, as a partial revert of r254296.
10898 * rtl.h (const_vec_p): Delete.
10899 (const_vec_duplicate_p): Don't test for vector CONSTs.
10900 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10901 * expmed.c (make_tree): Likewise.
10904 * common.md (E, F): Use CONSTANT_P instead of checking for
10906 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10907 checking for CONST_VECTOR.
10909 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10911 PR middle-end/83575
10912 * predict.c (force_edge_cold): Handle in more sane way edges
10913 with no prediction.
10915 2018-01-09 Carl Love <cel@us.ibm.com>
10917 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10919 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10920 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10921 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10922 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10923 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10924 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10925 * config/rs6000/rs6000-protos.h: Add extern defition for
10926 rs6000_generate_float2_double_code.
10927 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10929 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10930 (float2_v2df): Add define_expand.
10932 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10935 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10936 op_mode in the force_to_mode call.
10938 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10940 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10941 instead of checking each element individually.
10942 (aarch64_evpc_uzp): Likewise.
10943 (aarch64_evpc_zip): Likewise.
10944 (aarch64_evpc_ext): Likewise.
10945 (aarch64_evpc_rev): Likewise.
10946 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10947 instead of checking each element individually. Return true without
10949 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10950 whether all selected elements come from the same input, instead of
10951 checking each element individually. Remove calls to gen_rtx_REG,
10952 start_sequence and end_sequence and instead assert that no rtl is
10955 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10957 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10958 order of HIGH and CONST checks.
10960 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10962 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10963 if the destination isn't an SSA_NAME.
10965 2018-01-09 Richard Biener <rguenther@suse.de>
10967 PR tree-optimization/83668
10968 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10970 (canonicalize_loop_form): ... here, renamed from ...
10971 (canonicalize_loop_closed_ssa_form): ... this and amended to
10972 swap successor edges for loop exit blocks to make us use
10973 the RPO order we need for initial schedule generation.
10975 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10977 PR tree-optimization/64811
10978 * match.pd: When optimizing comparisons with Inf, avoid
10979 introducing or losing exceptions from comparisons with NaN.
10981 2018-01-09 Martin Liska <mliska@suse.cz>
10984 * asan.c (shadow_mem_size): Add gcc_assert.
10986 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10988 Don't save registers in main().
10991 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10992 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10993 * config/avr/avr.c (avr_set_current_function): Don't error if
10994 naked, OS_task or OS_main are specified at the same time.
10995 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10997 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10999 * common/config/avr/avr-common.c (avr_option_optimization_table):
11000 Switch on -mmain-is-OS_task for optimizing compilations.
11002 2018-01-09 Richard Biener <rguenther@suse.de>
11004 PR tree-optimization/83572
11005 * graphite.c: Include cfganal.h.
11006 (graphite_transform_loops): Connect infinite loops to exit
11007 and remove fake edges at the end.
11009 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11011 * ipa-inline.c (edge_badness): Revert accidental checkin.
11013 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11016 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11017 symbols; not inline clones.
11019 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11022 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11023 hard registers. Formatting fixes.
11025 PR preprocessor/83722
11026 * gcc.c (try_generate_repro): Pass
11027 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11028 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11031 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11032 Kito Cheng <kito.cheng@gmail.com>
11034 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11035 (riscv_leaf_function_p): Delete.
11036 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11038 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11040 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11042 (do_ifelse): New function.
11043 (do_isel): New function.
11044 (do_sub3): New function.
11045 (do_add3): New function.
11046 (do_load_mask_compare): New function.
11047 (do_overlap_load_compare): New function.
11048 (expand_compare_loop): New function.
11049 (expand_block_compare): Call expand_compare_loop() when appropriate.
11050 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11051 option description.
11052 (-mblock-compare-inline-loop-limit): New option.
11054 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11057 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11058 Reverse order of second and third operands in first alternative.
11059 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11060 of first and second elements in UNSPEC_VPERMR vector.
11061 (altivec_expand_vec_perm_le): Likewise.
11063 2018-01-08 Jeff Law <law@redhat.com>
11065 PR rtl-optimizatin/81308
11066 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11067 (process_switch): If group_case_labels makes a change, then set
11069 (pass_convert_switch::execute): If a switch is converted, then
11070 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11072 PR rtl-optimization/81308
11073 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11076 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11078 PR target/83663 - Revert r255946
11079 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11080 generation for cases where splatting a value is not useful.
11081 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11082 across a vec_duplicate and a paradoxical subreg forming a vector
11083 mode to a vec_concat.
11085 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11087 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11088 -march=armv8.3-a variants.
11089 * config/arm/t-multilib: Likewise.
11090 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11092 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11094 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11096 (cceq_ior_compare_complement): Give it a name so I can use it, and
11097 change boolean_or_operator predicate to boolean_operator so it can
11098 be used to generate a crand.
11099 (eqne): New code iterator.
11100 (bd/bd_neg): New code_attrs.
11101 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11102 a single define_insn.
11103 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11104 decrement (bdnzt/bdnzf/bdzt/bdzf).
11105 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11106 with the new names of the branch decrement patterns, and added the
11107 names of the branch decrement conditional patterns.
11109 2018-01-08 Richard Biener <rguenther@suse.de>
11111 PR tree-optimization/83563
11112 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11115 2018-01-08 Richard Biener <rguenther@suse.de>
11117 PR middle-end/83713
11118 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11120 2018-01-08 Richard Biener <rguenther@suse.de>
11122 PR tree-optimization/83685
11123 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11124 references to abnormals.
11126 2018-01-08 Richard Biener <rguenther@suse.de>
11129 * dwarf2out.c (output_indirect_strings): Handle empty
11130 skeleton_debug_str_hash.
11131 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11133 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11135 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11136 (emit_store_direct): Likewise.
11137 (arc_trampoline_adjust_address): Likewise.
11138 (arc_asm_trampoline_template): New function.
11139 (arc_initialize_trampoline): Use asm_trampoline_template.
11140 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11141 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11142 * config/arc/arc.md (flush_icache): Delete pattern.
11144 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11146 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11147 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11150 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11153 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11154 by not USED_FOR_TARGET.
11155 (make_pass_resolve_sw_modes): Likewise.
11157 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11159 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11162 2018-01-08 Richard Biener <rguenther@suse.de>
11164 PR middle-end/83580
11165 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11167 2018-01-08 Richard Biener <rguenther@suse.de>
11169 PR middle-end/83517
11170 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11172 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11174 PR middle-end/81897
11175 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11176 basic blocks with a small number of successors.
11177 (convert_control_dep_chain_into_preds): Improve handling of
11179 (dump_predicates): Split apart into...
11180 (dump_pred_chain): ...here...
11181 (dump_pred_info): ...and here.
11182 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11183 (can_chain_union_be_invalidated_p): Improve check for invalidation
11185 (uninit_uses_cannot_happen): Avoid unnecessary if
11186 convert_control_dep_chain_into_preds yielded nothing.
11188 2018-01-06 Martin Sebor <msebor@redhat.com>
11190 PR tree-optimization/83640
11191 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11192 subtracting negative offset from size.
11193 (builtin_access::overlap): Adjust offset bounds of the access to fall
11194 within the size of the object if possible.
11196 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11198 PR rtl-optimization/83699
11199 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11200 extract_bit_field_as_subreg to cases in which the extracted
11201 value is also a vector.
11203 * lra-constraints.c (process_alt_operands): Test for the equivalence
11204 substitutions when detecting a possible reload cycle.
11206 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11209 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11210 by default if flag_selective_schedling{,2}. Formatting fixes.
11212 PR rtl-optimization/83682
11213 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11214 if it has non-VECTOR_MODE element mode.
11215 (vec_duplicate_p): Likewise.
11217 PR middle-end/83694
11218 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11219 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11221 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11224 * config/i386/i386-builtin.def
11225 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11226 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11227 Require also OPTION_MASK_ISA_AVX512F in addition to
11228 OPTION_MASK_ISA_GFNI.
11229 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11230 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11231 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11232 to OPTION_MASK_ISA_GFNI.
11233 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11234 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11235 OPTION_MASK_ISA_AVX512BW.
11236 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11237 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11238 addition to OPTION_MASK_ISA_GFNI.
11239 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11240 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11241 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11242 to OPTION_MASK_ISA_GFNI.
11243 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11244 a requirement for all ISAs rather than any of them with a few
11246 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11248 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11249 bitmasks to be enabled with 3 exceptions, instead of requiring any
11250 enabled ISA with lots of exceptions.
11251 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11252 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11253 Change avx512bw in isa attribute to avx512f.
11254 * config/i386/sgxintrin.h: Add license boilerplate.
11255 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11256 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11257 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11258 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11260 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11261 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11262 temporarily sse2 rather than sse if not enabled already.
11265 * config/i386/sse.md (VI248_VLBW): Rename to ...
11266 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11267 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11268 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11269 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11270 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11271 mode iterator instead of VI248_VLBW.
11273 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11275 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11276 (record_modified): Skip clobbers; add debug output.
11277 (param_change_prob): Use sreal frequencies.
11279 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11281 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11282 punt for user-aligned variables.
11284 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11286 * tree-chrec.c (chrec_contains_symbols): Return true for
11289 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11292 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11293 of (x|y) == x for BICS pattern.
11295 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11297 PR tree-optimization/83605
11298 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11299 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11302 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11304 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11305 * config/epiphany/rtems.h: New file.
11307 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11308 Uros Bizjak <ubizjak@gmail.com>
11311 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11312 QIreg_operand instead of register_operand predicate.
11313 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11314 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11315 comments instead of -fmitigate[-_]rop.
11317 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11320 * cgraphunit.c (symbol_table::compile): Switch to text_section
11321 before calling assembly_start debug hook.
11322 * run-rtl-passes.c (run_rtl_passes): Likewise.
11325 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11327 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11328 range_int_cst_p rather than !symbolic_range_p before calling
11329 extract_range_from_multiplicative_op_1.
11331 2018-01-04 Jeff Law <law@redhat.com>
11333 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11334 redundant test in assertion.
11336 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11338 * doc/rtl.texi: Document machine_mode wrapper classes.
11340 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11342 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11343 using tree_to_uhwi.
11345 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11347 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11348 the VEC_PERM_EXPR fold to fail.
11350 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11353 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11354 to switched_sections.
11356 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11359 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11360 test for d.testing.
11362 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11365 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11366 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11368 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11371 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11372 is BLKmode and bitpos not zero or mode change is needed.
11374 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11377 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11380 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11383 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11384 instead of MULT rtx. Update all corresponding splitters.
11385 (*saddl_se): Ditto.
11386 (*ssub<modesuffix>): Ditto.
11387 (*ssubl_se): Ditto.
11388 (*cmp_sadd_di): Update split patterns.
11389 (*cmp_sadd_si): Ditto.
11390 (*cmp_sadd_sidi): Ditto.
11391 (*cmp_ssub_di): Ditto.
11392 (*cmp_ssub_si): Ditto.
11393 (*cmp_ssub_sidi): Ditto.
11394 * config/alpha/predicates.md (const23_operand): New predicate.
11395 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11396 Look for ASHIFT, not MULT inner operand.
11397 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11399 2018-01-04 Martin Liska <mliska@suse.cz>
11401 PR gcov-profile/83669
11402 * gcov.c (output_intermediate_file): Add version to intermediate
11404 * doc/gcov.texi: Document new field 'version' in intermediate
11405 file format. Fix location of '-k' option of gcov command.
11407 2018-01-04 Martin Liska <mliska@suse.cz>
11410 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11412 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11414 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11416 2018-01-03 Martin Sebor <msebor@redhat.com>
11418 PR tree-optimization/83655
11419 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11420 checking calls with invalid arguments.
11422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11424 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11425 (vectorizable_mask_load_store): Delete.
11426 (vectorizable_call): Return false for masked loads and stores.
11427 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11428 instead of gimple_assign_rhs1.
11429 (vectorizable_load): Handle IFN_MASK_LOAD.
11430 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11434 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11436 (vectorizable_mask_load_store): ...here.
11437 (vectorizable_load): ...and here.
11439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11441 * tree-vect-stmts.c (vect_build_all_ones_mask)
11442 (vect_build_zero_merge_argument): New functions, split out from...
11443 (vectorizable_load): ...here.
11445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11447 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11449 (vectorizable_mask_load_store): ...here.
11450 (vectorizable_store): ...and here.
11452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11454 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11456 (vectorizable_mask_load_store): ...here.
11458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11460 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11461 (vect_model_store_cost): Take a vec_load_store_type instead of a
11463 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11464 (vect_model_store_cost): Take a vec_load_store_type instead of a
11466 (vectorizable_mask_load_store): Update accordingly.
11467 (vectorizable_store): Likewise.
11468 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11472 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11473 IFN_MASK_LOAD calls here rather than...
11474 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11477 Alan Hayward <alan.hayward@arm.com>
11478 David Sherwood <david.sherwood@arm.com>
11480 * expmed.c (extract_bit_field_1): For vector extracts,
11481 fall back to extract_bit_field_as_subreg if vec_extract
11484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11485 Alan Hayward <alan.hayward@arm.com>
11486 David Sherwood <david.sherwood@arm.com>
11488 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11489 they are variable or constant sized.
11490 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11491 slots for constant-sized data.
11493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11494 Alan Hayward <alan.hayward@arm.com>
11495 David Sherwood <david.sherwood@arm.com>
11497 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11498 handling COND_EXPRs with boolean comparisons, try to find a better
11499 basis for the mask type than the boolean itself.
11501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11503 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11504 is calculated and how it can be overridden.
11505 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11506 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11508 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11512 Alan Hayward <alan.hayward@arm.com>
11513 David Sherwood <david.sherwood@arm.com>
11515 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11516 Remove the mode argument.
11517 (aarch64_simd_valid_immediate): Remove the mode and inverse
11519 * config/aarch64/iterators.md (bitsize): New iterator.
11520 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11521 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11522 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11523 aarch64_simd_valid_immediate.
11524 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11525 (aarch64_reg_or_bic_imm): Likewise.
11526 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11527 with an insn_type enum and msl with a modifier_type enum.
11528 Replace element_width with a scalar_mode. Change the shift
11529 to unsigned int. Add constructors for scalar_float_mode and
11530 scalar_int_mode elements.
11531 (aarch64_vect_float_const_representable_p): Delete.
11532 (aarch64_can_const_movi_rtx_p)
11533 (aarch64_simd_scalar_immediate_valid_for_move)
11534 (aarch64_simd_make_constant): Update call to
11535 aarch64_simd_valid_immediate.
11536 (aarch64_advsimd_valid_immediate_hs): New function.
11537 (aarch64_advsimd_valid_immediate): Likewise.
11538 (aarch64_simd_valid_immediate): Remove mode and inverse
11539 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11540 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11541 and aarch64_float_const_representable_p on the result.
11542 (aarch64_output_simd_mov_immediate): Remove mode argument.
11543 Update call to aarch64_simd_valid_immediate and use of
11544 simd_immediate_info.
11545 (aarch64_output_scalar_simd_mov_immediate): Update call
11548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11549 Alan Hayward <alan.hayward@arm.com>
11550 David Sherwood <david.sherwood@arm.com>
11552 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11553 (mode_nunits): Likewise CONST_MODE_NUNITS.
11554 * machmode.def (ADJUST_NUNITS): Document.
11555 * genmodes.c (mode_data::need_nunits_adj): New field.
11556 (blank_mode): Update accordingly.
11557 (adj_nunits): New variable.
11558 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11560 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11561 listed in adj_nunits.
11562 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11563 listed in adj_nunits. Don't emit case statements for such modes.
11564 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11565 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11566 nothing if adj_nunits is nonnull.
11567 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11568 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11569 (emit_mode_fbit): Update use of print_maybe_const_decl.
11570 (emit_move_size): Likewise. Treat the array as non-const
11572 (emit_mode_adjustments): Handle adj_nunits.
11574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11576 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11577 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11578 (VECTOR_MODES): Use it.
11579 (make_vector_modes): Take the prefix as an argument.
11581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11582 Alan Hayward <alan.hayward@arm.com>
11583 David Sherwood <david.sherwood@arm.com>
11585 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11586 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11587 for MODE_VECTOR_BOOL.
11588 * machmode.def (VECTOR_BOOL_MODE): Document.
11589 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11590 (make_vector_bool_mode): New function.
11591 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11593 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11594 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11596 * stor-layout.c (int_mode_for_mode): Likewise.
11597 * tree.c (build_vector_type_for_mode): Likewise.
11598 * varasm.c (output_constant_pool_2): Likewise.
11599 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11600 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11601 for MODE_VECTOR_BOOL.
11602 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11603 of mode class checks.
11604 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11605 instead of a list of mode class checks.
11606 (expand_vector_scalar_condition): Likewise.
11607 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11610 Alan Hayward <alan.hayward@arm.com>
11611 David Sherwood <david.sherwood@arm.com>
11613 * machmode.h (mode_size): Change from unsigned short to
11615 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11616 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11617 or if measurement_type is not polynomial.
11618 (fixed_size_mode::includes_p): Check for constant-sized modes.
11619 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11620 return a poly_uint16 rather than an unsigned short.
11621 (emit_mode_size): Change the type of mode_size from unsigned short
11622 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11623 (emit_mode_adjustments): Cope with polynomial vector sizes.
11624 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11626 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11628 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11629 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11630 * caller-save.c (setup_save_areas): Likewise.
11631 (replace_reg_with_saved_mem): Likewise.
11632 * calls.c (emit_library_call_value_1): Likewise.
11633 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11634 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11635 (gen_lowpart_for_combine): Likewise.
11636 * convert.c (convert_to_integer_1): Likewise.
11637 * cse.c (equiv_constant, cse_insn): Likewise.
11638 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11639 (cselib_subst_to_values): Likewise.
11640 * dce.c (word_dce_process_block): Likewise.
11641 * df-problems.c (df_word_lr_mark_ref): Likewise.
11642 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11643 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11644 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11645 (rtl_for_decl_location): Likewise.
11646 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11647 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11648 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11649 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11650 (expand_expr_real_1): Likewise.
11651 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11652 (pad_below): Likewise.
11653 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11654 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11655 * ira.c (get_subreg_tracking_sizes): Likewise.
11656 * ira-build.c (ira_create_allocno_objects): Likewise.
11657 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11658 (ira_sort_regnos_for_alter_reg): Likewise.
11659 * ira-costs.c (record_operand_costs): Likewise.
11660 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11661 (resolve_simple_move): Likewise.
11662 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11663 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11664 (lra_constraints): Likewise.
11665 (CONST_POOL_OK_P): Reject variable-sized modes.
11666 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11667 (add_pseudo_to_slot, lra_spill): Likewise.
11668 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11669 * optabs-query.c (get_best_extraction_insn): Likewise.
11670 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11671 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11672 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11673 * recog.c (offsettable_address_addr_space_p): Likewise.
11674 * regcprop.c (maybe_mode_change): Likewise.
11675 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11676 * regrename.c (build_def_use): Likewise.
11677 * regstat.c (dump_reg_info): Likewise.
11678 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11679 (find_reloads, find_reloads_subreg_address): Likewise.
11680 * reload1.c (eliminate_regs_1): Likewise.
11681 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11682 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11683 (simplify_binary_operation_1, simplify_subreg): Likewise.
11684 * targhooks.c (default_function_arg_padding): Likewise.
11685 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11686 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11687 (verify_gimple_assign_ternary): Likewise.
11688 * tree-inline.c (estimate_move_cost): Likewise.
11689 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11690 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11691 (get_address_cost_ainc): Likewise.
11692 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11693 (vect_supportable_dr_alignment): Likewise.
11694 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11695 (vectorizable_reduction): Likewise.
11696 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11697 (vectorizable_operation, vectorizable_load): Likewise.
11698 * tree.c (build_same_sized_truth_vector_type): Likewise.
11699 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11700 * var-tracking.c (emit_note_insn_var_location): Likewise.
11701 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11702 (ADDR_VEC_ALIGN): Likewise.
11704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11705 Alan Hayward <alan.hayward@arm.com>
11706 David Sherwood <david.sherwood@arm.com>
11708 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11710 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11711 or if measurement_type is polynomial.
11712 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11713 * combine.c (make_extraction): Likewise.
11714 * dse.c (find_shift_sequence): Likewise.
11715 * dwarf2out.c (mem_loc_descriptor): Likewise.
11716 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11717 (extract_bit_field, extract_low_bits): Likewise.
11718 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11719 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11720 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11721 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11722 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11723 * reload.c (find_reloads): Likewise.
11724 * reload1.c (alter_reg): Likewise.
11725 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11726 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11727 * tree-if-conv.c (predicate_mem_writes): Likewise.
11728 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11729 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11730 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11731 * valtrack.c (dead_debug_insert_temp): Likewise.
11732 * varasm.c (mergeable_constant_section): Likewise.
11733 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11735 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11736 Alan Hayward <alan.hayward@arm.com>
11737 David Sherwood <david.sherwood@arm.com>
11739 * expr.c (expand_assignment): Cope with polynomial mode sizes
11740 when assigning to a CONCAT.
11742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11743 Alan Hayward <alan.hayward@arm.com>
11744 David Sherwood <david.sherwood@arm.com>
11746 * machmode.h (mode_precision): Change from unsigned short to
11748 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11750 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11751 or if measurement_type is not polynomial.
11752 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11753 in which the mode is already known to be a scalar_int_mode.
11754 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11755 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11757 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11758 for GET_MODE_PRECISION.
11759 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11760 for GET_MODE_PRECISION.
11761 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11763 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11764 (expand_field_assignment, make_extraction): Likewise.
11765 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11766 (get_last_value): Likewise.
11767 * convert.c (convert_to_integer_1): Likewise.
11768 * cse.c (cse_insn): Likewise.
11769 * expr.c (expand_expr_real_1): Likewise.
11770 * lra-constraints.c (simplify_operand_subreg): Likewise.
11771 * optabs-query.c (can_atomic_load_p): Likewise.
11772 * optabs.c (expand_atomic_load): Likewise.
11773 (expand_atomic_store): Likewise.
11774 * ree.c (combine_reaching_defs): Likewise.
11775 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11776 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11777 * tree.h (type_has_mode_precision_p): Likewise.
11778 * ubsan.c (instrument_si_overflow): Likewise.
11780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11781 Alan Hayward <alan.hayward@arm.com>
11782 David Sherwood <david.sherwood@arm.com>
11784 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11785 polynomial numbers of units.
11786 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11787 (valid_vector_subparts_p): New function.
11788 (build_vector_type): Remove temporary shim and take the number
11789 of units as a poly_uint64 rather than an int.
11790 (build_opaque_vector_type): Take the number of units as a
11791 poly_uint64 rather than an int.
11792 * tree.c (build_vector_from_ctor): Handle polynomial
11793 TYPE_VECTOR_SUBPARTS.
11794 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11795 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11796 (build_vector_from_val): If the number of units is variable,
11797 use build_vec_duplicate_cst for constant operands and
11798 VEC_DUPLICATE_EXPR otherwise.
11799 (make_vector_type): Remove temporary is_constant ().
11800 (build_vector_type, build_opaque_vector_type): Take the number of
11801 units as a poly_uint64 rather than an int.
11802 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11804 * cfgexpand.c (expand_debug_expr): Likewise.
11805 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11806 (store_constructor, expand_expr_real_1): Likewise.
11807 (const_scalar_mask_from_tree): Likewise.
11808 * fold-const-call.c (fold_const_reduction): Likewise.
11809 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11810 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11811 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11812 (fold_relational_const): Likewise.
11813 (native_interpret_vector): Likewise. Change the size from an
11814 int to an unsigned int.
11815 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11816 TYPE_VECTOR_SUBPARTS.
11817 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11818 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11819 duplicating a non-constant operand into a variable-length vector.
11820 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11821 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11822 * ipa-icf.c (sem_variable::equals): Likewise.
11823 * match.pd: Likewise.
11824 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11825 * print-tree.c (print_node): Likewise.
11826 * stor-layout.c (layout_type): Likewise.
11827 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11828 * tree-cfg.c (verify_gimple_comparison): Likewise.
11829 (verify_gimple_assign_binary): Likewise.
11830 (verify_gimple_assign_ternary): Likewise.
11831 (verify_gimple_assign_single): Likewise.
11832 * tree-pretty-print.c (dump_generic_node): Likewise.
11833 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11834 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11835 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11836 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11837 (vect_shift_permute_load_chain): Likewise.
11838 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11839 (expand_vector_condition, optimize_vector_constructor): Likewise.
11840 (lower_vec_perm, get_compute_type): Likewise.
11841 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11842 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11843 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11844 (vect_recog_mask_conversion_pattern): Likewise.
11845 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11846 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11847 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11848 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11849 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11850 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11851 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11852 (supportable_widening_operation): Likewise.
11853 (supportable_narrowing_operation): Likewise.
11854 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11856 * varasm.c (output_constant): Likewise.
11858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11859 Alan Hayward <alan.hayward@arm.com>
11860 David Sherwood <david.sherwood@arm.com>
11862 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11863 so that both the length == 3 and length != 3 cases set up their
11864 own permute vectors. Add comments explaining why we know the
11865 number of elements is constant.
11866 (vect_permute_load_chain): Likewise.
11868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11869 Alan Hayward <alan.hayward@arm.com>
11870 David Sherwood <david.sherwood@arm.com>
11872 * machmode.h (mode_nunits): Change from unsigned char to
11874 (ONLY_FIXED_SIZE_MODES): New macro.
11875 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11876 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11877 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11879 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11880 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11881 or if measurement_type is not polynomial.
11882 * genmodes.c (ZERO_COEFFS): New macro.
11883 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11885 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11886 Use ZERO_COEFFS when emitting initializers.
11887 * data-streamer.h (bp_pack_poly_value): New function.
11888 (bp_unpack_poly_value): Likewise.
11889 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11890 for GET_MODE_NUNITS.
11891 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11892 for GET_MODE_NUNITS.
11893 * tree.c (make_vector_type): Remove temporary shim and make
11894 the real function take the number of units as a poly_uint64
11895 rather than an int.
11896 (build_vector_type_for_mode): Handle polynomial nunits.
11897 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11898 * emit-rtl.c (const_vec_series_p_1): Likewise.
11899 (gen_rtx_CONST_VECTOR): Likewise.
11900 * fold-const.c (test_vec_duplicate_folding): Likewise.
11901 * genrecog.c (validate_pattern): Likewise.
11902 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11903 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11904 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11905 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11906 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11907 * rtlanal.c (subreg_get_info): Likewise.
11908 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11909 (vect_grouped_load_supported): Likewise.
11910 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11911 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11912 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11913 (simplify_const_unary_operation, simplify_binary_operation_1)
11914 (simplify_const_binary_operation, simplify_ternary_operation)
11915 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11916 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11917 instead of CONST_VECTOR_NUNITS.
11918 * varasm.c (output_constant_pool_2): Likewise.
11919 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11920 explicit-encoded elements in the XVEC for variable-length vectors.
11922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11924 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11927 Alan Hayward <alan.hayward@arm.com>
11928 David Sherwood <david.sherwood@arm.com>
11930 * coretypes.h (fixed_size_mode): Declare.
11931 (fixed_size_mode_pod): New typedef.
11932 * builtins.h (target_builtins::x_apply_args_mode)
11933 (target_builtins::x_apply_result_mode): Change type to
11934 fixed_size_mode_pod.
11935 * builtins.c (apply_args_size, apply_result_size, result_vector)
11936 (expand_builtin_apply_args_1, expand_builtin_apply)
11937 (expand_builtin_return): Update accordingly.
11939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11941 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11942 * cselib.c (cselib_hash_rtx): Likewise.
11943 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11944 CONST_VECTOR encoding.
11946 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11947 Jeff Law <law@redhat.com>
11950 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11951 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11952 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11953 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11956 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11957 explicitly probe *sp in a noreturn function if there were any callee
11958 register saves or frame pointer is needed.
11960 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11963 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11964 BLKmode for ternary, binary or unary expressions.
11967 * var-tracking.c (delete_vta_debug_insn): New inline function.
11968 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11969 insns from get_insns () to NULL instead of each bb separately.
11970 Use delete_vta_debug_insn. No longer static.
11971 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11972 delete_vta_debug_insns callers.
11973 * rtl.h (delete_vta_debug_insns): Declare.
11974 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11975 instead of variable_tracking_main.
11977 2018-01-03 Martin Sebor <msebor@redhat.com>
11979 PR tree-optimization/83603
11980 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11981 arguments past the endof the argument list in functions declared
11982 without a prototype.
11983 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11984 Avoid checking when arguments are null.
11986 2018-01-03 Martin Sebor <msebor@redhat.com>
11989 * doc/extend.texi (attribute const): Fix a typo.
11990 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11991 issuing -Wsuggest-attribute for void functions.
11993 2018-01-03 Martin Sebor <msebor@redhat.com>
11995 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11996 offset_int::from instead of wide_int::to_shwi.
11997 (maybe_diag_overlap): Remove assertion.
11998 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11999 * gimple-ssa-sprintf.c (format_directive): Same.
12000 (parse_directive): Same.
12001 (sprintf_dom_walker::compute_format_length): Same.
12002 (try_substitute_return_value): Same.
12004 2018-01-03 Jeff Law <law@redhat.com>
12006 PR middle-end/83654
12007 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12008 non-constant residual for zero at runtime and avoid probing in
12009 that case. Reorganize code for trailing problem to mirror handling
12012 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12014 PR tree-optimization/83501
12015 * tree-ssa-strlen.c (get_string_cst): New.
12016 (handle_char_store): Call get_string_cst.
12018 2018-01-03 Martin Liska <mliska@suse.cz>
12020 PR tree-optimization/83593
12021 * tree-ssa-strlen.c: Include tree-cfg.h.
12022 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12023 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12024 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12026 (strlen_dom_walker::before_dom_children): Call
12027 gimple_purge_dead_eh_edges. Dump tranformation with details
12029 (strlen_dom_walker::before_dom_children): Update call by adding
12030 new argument cleanup_eh.
12031 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12033 2018-01-03 Martin Liska <mliska@suse.cz>
12036 * cif-code.def (VARIADIC_THUNK): New enum value.
12037 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12040 2018-01-03 Jan Beulich <jbeulich@suse.com>
12042 * sse.md (mov<mode>_internal): Tighten condition for when to use
12043 vmovdqu<ssescalarsize> for TI and OI modes.
12045 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12047 Update copyright years.
12049 2018-01-03 Martin Liska <mliska@suse.cz>
12052 * ipa-visibility.c (function_and_variable_visibility): Skip
12053 functions with noipa attribure.
12055 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12057 * gcc.c (process_command): Update copyright notice dates.
12058 * gcov-dump.c (print_version): Ditto.
12059 * gcov.c (print_version): Ditto.
12060 * gcov-tool.c (print_version): Ditto.
12061 * gengtype.c (create_file): Ditto.
12062 * doc/cpp.texi: Bump @copying's copyright year.
12063 * doc/cppinternals.texi: Ditto.
12064 * doc/gcc.texi: Ditto.
12065 * doc/gccint.texi: Ditto.
12066 * doc/gcov.texi: Ditto.
12067 * doc/install.texi: Ditto.
12068 * doc/invoke.texi: Ditto.
12070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12072 * vector-builder.h (vector_builder::m_full_nelts): Change from
12073 unsigned int to poly_uint64.
12074 (vector_builder::full_nelts): Update prototype accordingly.
12075 (vector_builder::new_vector): Likewise.
12076 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12077 (vector_builder::operator ==): Likewise.
12078 (vector_builder::finalize): Likewise.
12079 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12080 Take the number of elements as a poly_uint64 rather than an
12082 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12083 from unsigned int to poly_uint64.
12084 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12085 (vec_perm_indices::new_vector): Likewise.
12086 (vec_perm_indices::length): Likewise.
12087 (vec_perm_indices::nelts_per_input): Likewise.
12088 (vec_perm_indices::input_nelts): Likewise.
12089 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12090 number of elements per input as a poly_uint64 rather than an
12091 unsigned int. Use the original encoding for variable-length
12092 vectors, rather than clamping each individual element.
12093 For the second and subsequent elements in each pattern,
12094 clamp the step and base before clamping their sum.
12095 (vec_perm_indices::series_p): Handle polynomial element counts.
12096 (vec_perm_indices::all_in_range_p): Likewise.
12097 (vec_perm_indices_to_tree): Likewise.
12098 (vec_perm_indices_to_rtx): Likewise.
12099 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12100 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12101 (tree_vector_builder::new_binary_operation): Handle polynomial
12102 element counts. Return false if we need to know the number
12103 of elements at compile time.
12104 * fold-const.c (fold_vec_perm): Punt if the number of elements
12105 isn't known at compile time.
12107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12109 * vec-perm-indices.h (vec_perm_builder): Change element type
12110 from HOST_WIDE_INT to poly_int64.
12111 (vec_perm_indices::element_type): Update accordingly.
12112 (vec_perm_indices::clamp): Handle polynomial element_types.
12113 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12114 (vec_perm_indices::all_in_range_p): Likewise.
12115 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12117 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12118 polynomial vec_perm_indices element types.
12119 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12120 * fold-const.c (fold_vec_perm): Likewise.
12121 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12122 * tree-vect-generic.c (lower_vec_perm): Likewise.
12123 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12124 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12125 element type to HOST_WIDE_INT.
12127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12128 Alan Hayward <alan.hayward@arm.com>
12129 David Sherwood <david.sherwood@arm.com>
12131 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12132 rather than an int. Use plus_constant.
12133 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12134 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12137 Alan Hayward <alan.hayward@arm.com>
12138 David Sherwood <david.sherwood@arm.com>
12140 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12141 a HOST_WIDE_INT to a poly_int64.
12143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12144 Alan Hayward <alan.hayward@arm.com>
12145 David Sherwood <david.sherwood@arm.com>
12147 * calls.c (load_register_parameters): Cope with polynomial
12148 mode sizes. Require a constant size for BLKmode parameters
12149 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12150 forces a parameter to be padded at the lsb end in order to
12151 fill a complete number of words, require the parameter size
12152 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
12159 from unsigned int to poly_uint64_pod.
12160 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
12167 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
12174 operands whose sizes aren't ordered.
12175 (match_reload): Refer to this check here.
12177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12178 Alan Hayward <alan.hayward@arm.com>
12179 David Sherwood <david.sherwood@arm.com>
12181 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12182 that the mode size is in the set {1, 2, 4, 8, 16}.
12184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12185 Alan Hayward <alan.hayward@arm.com>
12186 David Sherwood <david.sherwood@arm.com>
12188 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12189 Use plus_constant instead of gen_rtx_PLUS.
12191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12192 Alan Hayward <alan.hayward@arm.com>
12193 David Sherwood <david.sherwood@arm.com>
12195 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12196 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12197 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12198 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12199 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12200 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12201 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12202 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12203 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12204 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12206 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12207 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12208 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12209 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12210 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12211 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12212 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12213 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12214 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12215 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12217 * expr.c (emit_move_resolve_push): Treat the input and result
12218 of PUSH_ROUNDING as a poly_int64.
12219 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12220 (emit_push_insn): Likewise.
12221 * lra-eliminations.c (mark_not_eliminable): Likewise.
12222 * recog.c (push_operand): Likewise.
12223 * reload1.c (elimination_effects): Likewise.
12224 * rtlanal.c (nonzero_bits1): Likewise.
12225 * calls.c (store_one_arg): Likewise. Require the padding to be
12226 known at compile time.
12228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12229 Alan Hayward <alan.hayward@arm.com>
12230 David Sherwood <david.sherwood@arm.com>
12232 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12233 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12240 rather than an int.
12242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12243 Alan Hayward <alan.hayward@arm.com>
12244 David Sherwood <david.sherwood@arm.com>
12246 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12247 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12248 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
12255 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12256 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12257 as a poly_uint64 too.
12259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12260 Alan Hayward <alan.hayward@arm.com>
12261 David Sherwood <david.sherwood@arm.com>
12263 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12266 Alan Hayward <alan.hayward@arm.com>
12267 David Sherwood <david.sherwood@arm.com>
12269 * combine.c (can_change_dest_mode): Handle polynomial
12270 REGMODE_NATURAL_SIZE.
12271 * expmed.c (store_bit_field_1): Likewise.
12272 * expr.c (store_constructor): Likewise.
12273 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12274 and polynomial REGMODE_NATURAL_SIZE.
12275 (gen_lowpart_common): Likewise.
12276 * reginfo.c (record_subregs_of_mode): Likewise.
12277 * rtlanal.c (read_modify_subreg_p): Likewise.
12279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12280 Alan Hayward <alan.hayward@arm.com>
12281 David Sherwood <david.sherwood@arm.com>
12283 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12284 numbers of elements.
12286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12287 Alan Hayward <alan.hayward@arm.com>
12288 David Sherwood <david.sherwood@arm.com>
12290 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12297 in a POINTER_PLUS_EXPR.
12299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12300 Alan Hayward <alan.hayward@arm.com>
12301 David Sherwood <david.sherwood@arm.com>
12303 * omp-simd-clone.c (simd_clone_subparts): New function.
12304 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12305 (ipa_simd_modify_function_body): Likewise.
12307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12308 Alan Hayward <alan.hayward@arm.com>
12309 David Sherwood <david.sherwood@arm.com>
12311 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12312 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12313 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12314 (expand_vector_condition, vector_element): Likewise.
12315 (subparts_gt): New function.
12316 (get_compute_type): Use subparts_gt.
12317 (count_type_subparts): Delete.
12318 (expand_vector_operations_1): Use subparts_gt instead of
12319 count_type_subparts.
12321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12322 Alan Hayward <alan.hayward@arm.com>
12323 David Sherwood <david.sherwood@arm.com>
12325 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12326 (vect_compile_time_alias): ...this new function. Do the calculation
12327 on poly_ints rather than trees.
12328 (vect_prune_runtime_alias_test_list): Update call accordingly.
12330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12331 Alan Hayward <alan.hayward@arm.com>
12332 David Sherwood <david.sherwood@arm.com>
12334 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12336 (vect_schedule_slp_instance): Likewise.
12338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12339 Alan Hayward <alan.hayward@arm.com>
12340 David Sherwood <david.sherwood@arm.com>
12342 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12343 constant and extern definitions for variable-length vectors.
12344 (vect_get_constant_vectors): Note that the number of units
12345 is known to be constant.
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_conversion): Treat the number
12352 of units as polynomial. Choose between WIDE and NARROW based
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 (simd_clone_subparts): New function.
12360 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12363 Alan Hayward <alan.hayward@arm.com>
12364 David Sherwood <david.sherwood@arm.com>
12366 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12367 vectors as polynomial. Use build_index_vector for
12368 IFN_GOMP_SIMD_LANE.
12370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12371 Alan Hayward <alan.hayward@arm.com>
12372 David Sherwood <david.sherwood@arm.com>
12374 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12375 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12376 for variable-length vectors.
12377 (vectorizable_mask_load_store): Treat the number of units as
12378 polynomial, asserting that it is constant if the condition has
12379 already been enforced.
12380 (vectorizable_store, vectorizable_load): Likewise.
12382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12383 Alan Hayward <alan.hayward@arm.com>
12384 David Sherwood <david.sherwood@arm.com>
12386 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12387 of units as polynomial. Punt if we can't tell at compile time
12388 which vector contains the final result.
12390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12391 Alan Hayward <alan.hayward@arm.com>
12392 David Sherwood <david.sherwood@arm.com>
12394 * tree-vect-loop.c (vectorizable_induction): Treat the number
12395 of units as polynomial. Punt on SLP inductions. Use an integer
12396 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12397 cast of such a series for variable-length floating-point
12400 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12401 Alan Hayward <alan.hayward@arm.com>
12402 David Sherwood <david.sherwood@arm.com>
12404 * tree.h (build_index_vector): Declare.
12405 * tree.c (build_index_vector): New function.
12406 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12407 of units as polynomial, forcibly converting it to a constant if
12408 vectorizable_reduction has already enforced the condition.
12409 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12410 to create a {1,2,3,...} vector.
12411 (vectorizable_reduction): Treat the number of units as polynomial.
12412 Choose vectype_in based on the largest scalar element size rather
12413 than the smallest number of units. Enforce the restrictions
12416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12417 Alan Hayward <alan.hayward@arm.com>
12418 David Sherwood <david.sherwood@arm.com>
12420 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12421 number of units as polynomial.
12423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12424 Alan Hayward <alan.hayward@arm.com>
12425 David Sherwood <david.sherwood@arm.com>
12427 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12428 * target.def (autovectorize_vector_sizes): Return the vector sizes
12429 by pointer, using vector_sizes rather than a bitmask.
12430 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12431 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12432 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12434 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12435 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12436 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12437 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12438 * omp-general.c (omp_max_vf): Likewise.
12439 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12440 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12441 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12442 * tree-vect-slp.c (vect_slp_bb): Likewise.
12443 * doc/tm.texi: Regenerate.
12444 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12446 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12447 the vector size as a poly_uint64 rather than an unsigned int.
12448 (current_vector_size): Change from an unsigned int to a poly_uint64.
12449 (get_vectype_for_scalar_type): Update accordingly.
12450 * tree.h (build_truth_vector_type): Take the size and number of
12451 units as a poly_uint64 rather than an unsigned int.
12452 (build_vector_type): Add a temporary overload that takes
12453 the number of units as a poly_uint64 rather than an unsigned int.
12454 * tree.c (make_vector_type): Likewise.
12455 (build_truth_vector_type): Take the number of units as a poly_uint64
12456 rather than an unsigned int.
12458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12459 Alan Hayward <alan.hayward@arm.com>
12460 David Sherwood <david.sherwood@arm.com>
12462 * target.def (get_mask_mode): Take the number of units and length
12463 as poly_uint64s rather than unsigned ints.
12464 * targhooks.h (default_get_mask_mode): Update accordingly.
12465 * targhooks.c (default_get_mask_mode): Likewise.
12466 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12467 * doc/tm.texi: Regenerate.
12469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12470 Alan Hayward <alan.hayward@arm.com>
12471 David Sherwood <david.sherwood@arm.com>
12473 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12474 * omp-general.c (omp_max_vf): Likewise.
12475 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12476 (expand_omp_simd): Handle polynomial safelen.
12477 * omp-low.c (omplow_simd_context): Add a default constructor.
12478 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12479 (lower_rec_simd_input_clauses): Update accordingly.
12480 (lower_rec_input_clauses): Likewise.
12482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12483 Alan Hayward <alan.hayward@arm.com>
12484 David Sherwood <david.sherwood@arm.com>
12486 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12487 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12488 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12489 (vect_analyze_slp_cost): Likewise.
12490 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12491 (vect_model_load_cost): Likewise.
12493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12494 Alan Hayward <alan.hayward@arm.com>
12495 David Sherwood <david.sherwood@arm.com>
12497 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12498 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12499 from an unsigned int * to a poly_uint64_pod *.
12500 (calculate_unrolling_factor): New function.
12501 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12504 Alan Hayward <alan.hayward@arm.com>
12505 David Sherwood <david.sherwood@arm.com>
12507 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12508 from an unsigned int to a poly_uint64.
12509 (_loop_vec_info::slp_unrolling_factor): Likewise.
12510 (_loop_vec_info::vectorization_factor): Change from an int
12512 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12513 (vect_get_num_vectors): New function.
12514 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12515 (vect_get_num_copies): Use vect_get_num_vectors.
12516 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12517 to an unsigned int *.
12518 (vect_analyze_data_refs): Change min_vf from an int * to a
12520 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12521 than an unsigned HOST_WIDE_INT.
12522 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12523 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12524 to an unsigned int *.
12525 (vect_analyze_data_ref_dependences): Likewise.
12526 (vect_compute_data_ref_alignment): Handle polynomial vf.
12527 (vect_enhance_data_refs_alignment): Likewise.
12528 (vect_prune_runtime_alias_test_list): Likewise.
12529 (vect_shift_permute_load_chain): Likewise.
12530 (vect_supportable_dr_alignment): Likewise.
12531 (dependence_distance_ge_vf): Take the vectorization factor as a
12532 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12533 (vect_analyze_data_refs): Change min_vf from an int * to a
12535 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12536 vfm1 as a poly_uint64 rather than an int. Make the same change
12537 for the returned bound_scalar.
12538 (vect_gen_vector_loop_niters): Handle polynomial vf.
12539 (vect_do_peeling): Likewise. Update call to
12540 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12541 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12543 * tree-vect-loop.c (vect_determine_vectorization_factor)
12544 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12545 (vect_get_known_peeling_cost): Likewise.
12546 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12547 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12548 (vect_transform_loop): Likewise. Use the lowest possible VF when
12549 updating the upper bounds of the loop.
12550 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12551 rather than an int.
12552 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12553 polynomial unroll factors.
12554 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12555 (vect_make_slp_decision): Likewise.
12556 (vect_supported_load_permutation_p): Likewise, and polynomial
12558 (vect_analyze_slp_cost): Handle polynomial vf.
12559 (vect_slp_analyze_node_operations): Likewise.
12560 (vect_slp_analyze_bb_1): Likewise.
12561 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12562 than an unsigned HOST_WIDE_INT.
12563 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12564 (vectorizable_load): Handle polynomial vf.
12565 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12567 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12570 Alan Hayward <alan.hayward@arm.com>
12571 David Sherwood <david.sherwood@arm.com>
12573 * match.pd: Handle bit operations involving three constants
12574 and try to fold one pair.
12576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12578 * tree-vect-loop-manip.c: Include gimple-fold.h.
12579 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12580 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12581 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12582 Add a path that uses a step of VF instead of 1, but disable it
12584 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12585 and niters_no_overflow parameters. Update calls to
12586 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12587 Create a new SSA name if the latter choses to use a ste other
12588 than zero, and return it via niters_vector_mult_vf_var.
12589 * tree-vect-loop.c (vect_transform_loop): Update calls to
12590 vect_do_peeling, vect_gen_vector_loop_niters and
12591 slpeel_make_loop_iterate_ntimes.
12592 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12593 (vect_gen_vector_loop_niters): Update declarations after above changes.
12595 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12597 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12598 128-bit round to integer instructions.
12599 (ceil<mode>2): Likewise.
12600 (btrunc<mode>2): Likewise.
12601 (round<mode>2): Likewise.
12603 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12605 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12606 unaligned VSX load/store on P8/P9.
12607 (expand_block_clear): Allow the use of unaligned VSX
12608 load/store on P8/P9.
12610 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12612 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12614 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12615 swap associated with both a load and a store.
12617 2018-01-02 Andrew Waterman <andrew@sifive.com>
12619 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12620 * config/riscv/riscv.md (clear_cache): Use it.
12622 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12624 * web.c: Remove out-of-date comment.
12626 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12628 * expr.c (fixup_args_size_notes): Check that any existing
12629 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12630 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12631 (emit_single_push_insn): ...here.
12633 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12635 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12636 (const_vector_encoded_nelts): New function.
12637 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12638 (const_vector_int_elt, const_vector_elt): Declare.
12639 * emit-rtl.c (const_vector_int_elt_1): New function.
12640 (const_vector_elt): Likewise.
12641 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12642 of CONST_VECTOR_ELT.
12644 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12646 * expr.c: Include rtx-vector-builder.h.
12647 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12648 directly on the tree encoding.
12649 (const_vector_from_tree): Likewise.
12650 * optabs.c: Include rtx-vector-builder.h.
12651 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12652 sequence of "u" values.
12653 * vec-perm-indices.c: Include rtx-vector-builder.h.
12654 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12655 directly on the vec_perm_indices encoding.
12657 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12659 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12660 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12661 * rtx-vector-builder.h: New file.
12662 * rtx-vector-builder.c: Likewise.
12663 * rtl.h (rtx_def::u2): Add a const_vector field.
12664 (CONST_VECTOR_NPATTERNS): New macro.
12665 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12666 (CONST_VECTOR_DUPLICATE_P): Likewise.
12667 (CONST_VECTOR_STEPPED_P): Likewise.
12668 (CONST_VECTOR_ENCODED_ELT): Likewise.
12669 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12670 (unwrap_const_vec_duplicate): Likewise.
12671 (const_vec_series_p): Check for a non-duplicated vector encoding.
12672 Say that the function only returns true for integer vectors.
12673 * emit-rtl.c: Include rtx-vector-builder.h.
12674 (gen_const_vec_duplicate_1): Delete.
12675 (gen_const_vector): Call gen_const_vec_duplicate instead of
12676 gen_const_vec_duplicate_1.
12677 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12678 (gen_const_vec_duplicate): Use rtx_vector_builder.
12679 (gen_const_vec_series): Likewise.
12680 (gen_rtx_CONST_VECTOR): Likewise.
12681 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12682 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12683 Build a new vector rather than modifying a CONST_VECTOR in-place.
12684 (handle_special_swappables): Update call accordingly.
12685 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12686 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12687 Build a new vector rather than modifying a CONST_VECTOR in-place.
12688 (handle_special_swappables): Update call accordingly.
12690 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12692 * simplify-rtx.c (simplify_const_binary_operation): Use
12693 CONST_VECTOR_ELT instead of XVECEXP.
12695 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12697 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12698 the selector elements to be different from the data elements
12699 if the selector is a VECTOR_CST.
12700 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12701 ssizetype for the selector.
12703 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12705 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12706 before testing each element individually.
12707 * tree-vect-generic.c (lower_vec_perm): Likewise.
12709 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12711 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12712 * selftest-run-tests.c (selftest::run_tests): Call it.
12713 * vector-builder.h (vector_builder::operator ==): New function.
12714 (vector_builder::operator !=): Likewise.
12715 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12716 (vec_perm_indices::all_from_input_p): New function.
12717 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12718 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12719 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12720 instead of reading the VECTOR_CST directly. Detect whether both
12721 vector inputs are the same before constructing the vec_perm_indices,
12722 and update the number of inputs argument accordingly. Use the
12723 utility functions added above. Only construct sel2 if we need to.
12725 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12727 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12728 the broadcast of the low byte.
12729 (expand_mult_highpart): Use an explicit encoding for the permutes.
12730 * optabs-query.c (can_mult_highpart_p): Likewise.
12731 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12732 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12733 (vectorizable_bswap): Likewise.
12734 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12735 explicit encoding for the power-of-2 permutes.
12736 (vect_permute_store_chain): Likewise.
12737 (vect_grouped_load_supported): Likewise.
12738 (vect_permute_load_chain): Likewise.
12740 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12742 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12743 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12744 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12745 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12746 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12747 (vect_gen_perm_mask_any): Likewise.
12749 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12751 * int-vector-builder.h: New file.
12752 * vec-perm-indices.h: Include int-vector-builder.h.
12753 (vec_perm_indices): Redefine as an int_vector_builder.
12754 (auto_vec_perm_indices): Delete.
12755 (vec_perm_builder): Redefine as a stand-alone class.
12756 (vec_perm_indices::vec_perm_indices): New function.
12757 (vec_perm_indices::clamp): Likewise.
12758 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12759 (vec_perm_indices::new_vector): New function.
12760 (vec_perm_indices::new_expanded_vector): Update for new
12761 vec_perm_indices class.
12762 (vec_perm_indices::rotate_inputs): New function.
12763 (vec_perm_indices::all_in_range_p): Operate directly on the
12764 encoded form, without computing elided elements.
12765 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12766 encoding. Update for new vec_perm_indices class.
12767 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12768 the given vec_perm_builder.
12769 (expand_vec_perm_var): Update vec_perm_builder constructor.
12770 (expand_mult_highpart): Use vec_perm_builder instead of
12771 auto_vec_perm_indices.
12772 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12773 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12774 or double series encoding as appropriate.
12775 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12776 vec_perm_indices instead of auto_vec_perm_indices.
12777 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12778 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12779 (vect_permute_store_chain): Likewise.
12780 (vect_grouped_load_supported): Likewise.
12781 (vect_permute_load_chain): Likewise.
12782 (vect_shift_permute_load_chain): Likewise.
12783 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12784 (vect_transform_slp_perm_load): Likewise.
12785 (vect_schedule_slp_instance): Likewise.
12786 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12787 (vectorizable_mask_load_store): Likewise.
12788 (vectorizable_bswap): Likewise.
12789 (vectorizable_store): Likewise.
12790 (vectorizable_load): Likewise.
12791 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12792 vec_perm_indices instead of auto_vec_perm_indices. Use
12793 tree_to_vec_perm_builder to read the vector from a tree.
12794 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12795 vec_perm_builder instead of a vec_perm_indices.
12796 (have_whole_vector_shift): Use vec_perm_builder and
12797 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12798 truncation to calc_vec_perm_mask_for_shift.
12799 (vect_create_epilog_for_reduction): Likewise.
12800 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12801 from auto_vec_perm_indices to vec_perm_indices.
12802 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12803 instead of changing individual elements.
12804 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12805 the vector in d.perm.
12806 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12807 from auto_vec_perm_indices to vec_perm_indices.
12808 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12809 instead of changing individual elements.
12810 (arm_vectorize_vec_perm_const): Use new_vector to install
12811 the vector in d.perm.
12812 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12813 Update vec_perm_builder constructor.
12814 (rs6000_expand_interleave): Likewise.
12815 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12816 (rs6000_expand_interleave): Likewise.
12818 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12820 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12821 to qimode could truncate the indices.
12822 * optabs.c (expand_vec_perm_var): Likewise.
12824 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12826 * Makefile.in (OBJS): Add vec-perm-indices.o.
12827 * vec-perm-indices.h: New file.
12828 * vec-perm-indices.c: Likewise.
12829 * target.h (vec_perm_indices): Replace with a forward class
12831 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12832 * optabs.h: Include vec-perm-indices.h.
12833 (expand_vec_perm): Delete.
12834 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12835 (expand_vec_perm_const): Declare.
12836 * target.def (vec_perm_const_ok): Replace with...
12837 (vec_perm_const): ...this new hook.
12838 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12839 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12840 * doc/tm.texi: Regenerate.
12841 * optabs.def (vec_perm_const): Delete.
12842 * doc/md.texi (vec_perm_const): Likewise.
12843 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12844 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12845 expand_vec_perm for constant permutation vectors. Assert that
12846 the mode of variable permutation vectors is the integer equivalent
12847 of the mode that is being permuted.
12848 * optabs-query.h (selector_fits_mode_p): Declare.
12849 * optabs-query.c: Include vec-perm-indices.h.
12850 (selector_fits_mode_p): New function.
12851 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12852 is defined, instead of checking whether the vec_perm_const_optab
12853 exists. Use targetm.vectorize.vec_perm_const instead of
12854 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12855 fit in the vector mode before using a variable permute.
12856 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12857 vec_perm_indices instead of an rtx.
12858 (expand_vec_perm): Replace with...
12859 (expand_vec_perm_const): ...this new function. Take the selector
12860 as a vec_perm_indices rather than an rtx. Also take the mode of
12861 the selector. Update call to shift_amt_for_vec_perm_mask.
12862 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12863 Use vec_perm_indices::new_expanded_vector to expand the original
12864 selector into bytes. Check whether the indices fit in the vector
12865 mode before using a variable permute.
12866 (expand_vec_perm_var): Make global.
12867 (expand_mult_highpart): Use expand_vec_perm_const.
12868 * fold-const.c: Includes vec-perm-indices.h.
12869 * tree-ssa-forwprop.c: Likewise.
12870 * tree-vect-data-refs.c: Likewise.
12871 * tree-vect-generic.c: Likewise.
12872 * tree-vect-loop.c: Likewise.
12873 * tree-vect-slp.c: Likewise.
12874 * tree-vect-stmts.c: Likewise.
12875 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12877 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12878 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12879 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12880 (aarch64_vectorize_vec_perm_const): ...this new function.
12881 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12882 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12883 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12884 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12885 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12886 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12887 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12889 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12890 check for NEON modes.
12891 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12892 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12893 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12894 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12896 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12897 the old VEC_PERM_CONST conditions.
12898 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12899 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12900 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12901 (ia64_vectorize_vec_perm_const_ok): Merge into...
12902 (ia64_vectorize_vec_perm_const): ...this new function.
12903 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12904 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12905 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12906 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12907 * config/mips/mips.c (mips_expand_vec_perm_const)
12908 (mips_vectorize_vec_perm_const_ok): Merge into...
12909 (mips_vectorize_vec_perm_const): ...this new function.
12910 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12911 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12912 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12913 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12914 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12915 (rs6000_expand_vec_perm_const): Delete.
12916 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12918 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12919 (altivec_expand_vec_perm_const_le): Take each operand individually.
12920 Operate on constant selectors rather than rtxes.
12921 (altivec_expand_vec_perm_const): Likewise. Update call to
12922 altivec_expand_vec_perm_const_le.
12923 (rs6000_expand_vec_perm_const): Delete.
12924 (rs6000_vectorize_vec_perm_const_ok): Delete.
12925 (rs6000_vectorize_vec_perm_const): New function.
12926 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12927 an element count and rtx array.
12928 (rs6000_expand_extract_even): Update call accordingly.
12929 (rs6000_expand_interleave): Likewise.
12930 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12931 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12932 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12933 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12934 (rs6000_expand_vec_perm_const): Delete.
12935 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12936 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12937 (altivec_expand_vec_perm_const_le): Take each operand individually.
12938 Operate on constant selectors rather than rtxes.
12939 (altivec_expand_vec_perm_const): Likewise. Update call to
12940 altivec_expand_vec_perm_const_le.
12941 (rs6000_expand_vec_perm_const): Delete.
12942 (rs6000_vectorize_vec_perm_const_ok): Delete.
12943 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12944 reference to the SPE evmerge intructions.
12945 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12946 an element count and rtx array.
12947 (rs6000_expand_extract_even): Update call accordingly.
12948 (rs6000_expand_interleave): Likewise.
12949 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12950 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12952 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12954 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12956 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12957 vector mode and that that mode matches the mode of the data
12959 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12960 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12961 directly using expand_vec_perm_1 when forcing selectors into
12963 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12965 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12967 * optabs-query.h (can_vec_perm_p): Delete.
12968 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12969 * optabs-query.c (can_vec_perm_p): Split into...
12970 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12971 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12972 particular selector is valid.
12973 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12974 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12975 (vect_grouped_load_supported): Likewise.
12976 (vect_shift_permute_load_chain): Likewise.
12977 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12978 (vect_transform_slp_perm_load): Likewise.
12979 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12980 (vectorizable_bswap): Likewise.
12981 (vect_gen_perm_mask_checked): Likewise.
12982 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12983 implementations of variable permutation vectors into account
12984 when deciding which selector to use.
12985 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12986 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12987 with a false third argument.
12988 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12989 to test whether the constant selector is valid and can_vec_perm_var_p
12990 to test whether a variable selector is valid.
12992 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12994 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12995 * optabs-query.c (can_vec_perm_p): Likewise.
12996 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12997 instead of vec_perm_indices.
12998 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12999 (vect_gen_perm_mask_checked): Likewise,
13000 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13001 (vect_gen_perm_mask_checked): Likewise,
13003 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13005 * optabs-query.h (qimode_for_vec_perm): Declare.
13006 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13007 (qimode_for_vec_perm): ...this new function.
13008 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13010 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13012 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13013 does not have a conditional at the top.
13015 2018-01-02 Richard Biener <rguenther@suse.de>
13017 * ipa-inline.c (big_speedup_p): Fix expression.
13019 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13022 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13025 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13029 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13030 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13031 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13032 cond_taken_branch_cost 3->4.
13034 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13036 PR tree-optimization/83581
13037 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13038 TODO_cleanup_cfg if any changes have been made.
13040 PR middle-end/83608
13041 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13042 convert_modes if target mode has the right side, but different mode
13045 PR middle-end/83609
13046 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13047 last argument when extracting from CONCAT. If either from_real or
13048 from_imag is NULL, use expansion through memory. If result is not
13049 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13050 the parts directly to inner mode, if even that fails, use expansion
13053 PR middle-end/83623
13054 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13055 check for bswap in mode rather than HImode and use that in expand_unop
13058 Copyright (C) 2018 Free Software Foundation, Inc.
13060 Copying and distribution of this file, with or without modification,
13061 are permitted in any medium without royalty provided the copyright
13062 notice and this notice are preserved.