1 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
6 (ix86_gimple_fold_builtin): Likewise.
9 * config/i386/i386.c: Include tree-vector-builder.h.
10 (ix86_vector_shift_count): New function.
11 (ix86_fold_builtin): Fold shift builtins by scalar count.
12 (ix86_gimple_fold_builtin): Likewise.
14 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
15 _mm512_setzero): New intrinsics.
17 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
21 code generation for cases where splatting a value is not useful.
22 * simplify-rtx.c (simplify_ternary_operation): Simplify
23 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
26 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
28 * config.gcc: Support "goldmont-plus".
29 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
31 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
32 PROCESSOR_GOLDMONT_PLUS.
33 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
34 (processor_target_table): Add "goldmont-plus".
35 (PTA_GOLDMONT_PLUS): Define.
36 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
37 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
38 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
39 (fold_builtin_cpu): Add "goldmont-plus".
40 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
41 (ix86_option_override_internal): Add "goldmont-plus".
42 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
43 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
44 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
45 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
47 2018-05-17 Richard Biener <rguenther@suse.de>
49 PR tree-optimization/85757
50 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
51 remove defs that only feed that PHI from further processing.
53 2018-05-16 Jim Wilson <jimw@sifive.com>
55 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
57 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
59 2018-05-16 Mark Wielaard <mark@klomp.org>
61 * dwarf2out.c (count_index_strings): New function.
62 (output_indirect_strings): Call count_index_strings and generate
63 header for dwarf_version >= 5.
65 2018-05-16 Mark Wielaard <mark@klomp.org>
67 * dwarf2out.c (dwarf_FORM): New function.
68 (set_indirect_string): Use dwarf_FORM.
69 (reset_indirect_string): Likewise.
70 (size_of_die): Likewise.
71 (value_format): Likewise.
72 (output_die): Likewise.
73 (add_skeleton_AT_string): Likewise.
74 (output_macinfo_op): Likewise.
75 (index_string): Likewise.
76 (output_index_string_offset): Likewise.
77 (output_index_string): Likewise.
78 (count_index_strings): Likewise.
80 2018-05-16 Carl Love <cel@us.ibm.com>
82 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
83 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
85 2018-05-16 Martin Jambor <mjambor@suse.cz>
87 * ipa-prop.c (ipa_free_all_edge_args): Remove.
88 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
90 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
92 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
93 (fnma<mode>4): Likewise.
94 (fms<mode>4): Likewise.
95 (fnms<mode>4): Likewise.
96 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
97 (aarch64_fnma<mode>4): Likewise.
98 (aarch64_fms<mode>4): Likewise.
99 (aarch64_fnms<mode>4): Likewise.
100 (aarch64_fnmadd<mode>4): Likewise.
102 2018-05-16 Jason Merrill <jason@redhat.com>
104 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
106 2018-05-16 Richard Biener <rguenther@suse.de>
108 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
109 (dump_stmt_cost): Declare.
110 (add_stmt_cost): Dump cost we add.
111 (add_stmt_costs): New function.
112 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
114 (vect_analyze_stmt): Adjust prototype.
115 (vectorizable_condition): Likewise.
116 (vectorizable_live_operation): Likewise.
117 (vectorizable_reduction): Likewise.
118 (vectorizable_induction): Likewise.
119 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
120 cost vector to pass to vectorizable_ and record afterwards.
121 (vect_model_reduction_cost): Take cost vector argument and adjust.
122 (vect_model_induction_cost): Likewise.
123 (vectorizable_reduction): Likewise.
124 (vectorizable_induction): Likewise.
125 (vectorizable_live_operation): Likewise.
126 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
127 SLP_TREE_NUMBER_OF_VEC_STMTS.
128 (vect_analyze_slp_cost_1): Remove.
129 (vect_analyze_slp_cost): Likewise.
130 (vect_slp_analyze_node_operations): Take visited args and
131 a target cost vector. Avoid processing already visited stmt sets.
132 (vect_slp_analyze_operations): Use a local cost vector to gather
133 costs and register those of non-discarded instances.
134 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
135 (vect_schedule_slp_instance): Remove copying of
136 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
138 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
139 adding cost. Record cost entry location.
140 (vect_prologue_cost_for_slp_op): Function to compute cost of
141 a constant or invariant generated for SLP vect in the prologue,
142 split out from vect_analyze_slp_cost_1.
143 (vect_model_simple_cost): Make static. Adjust for SLP costing.
144 (vect_model_promotion_demotion_cost): Likewise.
145 (vect_model_store_cost): Likewise, make static.
146 (vect_model_load_cost): Likewise.
147 (vectorizable_bswap): Add cost vector arg and adjust.
148 (vectorizable_call): Likewise.
149 (vectorizable_simd_clone_call): Likewise.
150 (vectorizable_conversion): Likewise.
151 (vectorizable_assignment): Likewise.
152 (vectorizable_shift): Likewise.
153 (vectorizable_operation): Likewise.
154 (vectorizable_store): Likewise.
155 (vectorizable_load): Likewise.
156 (vectorizable_condition): Likewise.
157 (vectorizable_comparison): Likewise.
158 (can_vectorize_live_stmts): Likewise.
159 (vect_analyze_stmt): Likewise.
160 (vect_transform_stmt): Adjust calls to vectorizable_*.
161 * tree-vectorizer.c: Include gimple-pretty-print.h.
162 (dump_stmt_cost): New function.
164 2018-05-16 Richard Biener <rguenther@suse.de>
166 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
167 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
168 * tree-ssa-dse.c: Include tree-ssa-loop.h.
169 (check_name): New callback.
170 (dse_classify_store): Track cycles via a visited bitmap of PHI
171 defs and simplify handling of in-loop and across loop dead stores
172 and properly fail for loop-variant refs. Handle byte-tracking with
173 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
176 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
178 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
179 (vect_get_mask_type_for_stmt): Likewise.
180 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
182 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
183 to determine the statement's vector type and the vector type that
184 should be used for calculating nunits. Deal with cases in which
185 the type has to be deferred.
186 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
187 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
188 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
189 (vect_determine_vf_for_stmt): New functions, split out from...
190 (vect_determine_vectorization_factor): ...here.
191 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
192 (vect_get_mask_type_for_stmt): New functions, split out from
193 vect_determine_vectorization_factor.
195 2018-05-16 Richard Biener <rguenther@suse.de>
197 * tree-cfg.c (verify_gimple_assign_ternary): Properly
198 verify the [VEC_]COND_EXPR embedded comparison.
200 2018-05-15 Martin Sebor <msebor@redhat.com>
202 PR tree-optimization/85753
203 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
204 RECORD_TYPE in addition to ARRAY_TYPE.
206 2018-05-15 Martin Sebor <msebor@redhat.com>
209 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
211 2018-05-15 Richard Biener <rguenther@suse.de>
213 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
214 add by_clobber_p one. Change algorithm to collect all defs
215 representing uses we need to walk and try reducing them to
216 a single one before failing.
217 (dse_dom_walker::dse_optimize_stmt): Adjust.
219 2018-05-13 Mark Wielaard <mark@klomp.org>
221 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
222 (size_of_loc_descr): Likewise.
223 (output_loc_operands): Likewise.
224 (output_loc_operands_raw): Likewise.
225 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
226 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
227 (hash_loc_operands): Likewise.
228 (compare_loc_operands): Likewise.
230 2018-05-14 Mark Wielaard <mark@klomp.org>
232 * dwarf2out.c (count_index_addrs): New function.
233 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
235 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
237 PR tree-optimization/83648
238 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
239 return value as malloc candidate.
241 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
244 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
245 as true in call to suggest_attribute.
247 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
249 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
252 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
254 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
256 (aarch64_crypto_aesd_fused): Likewise.
258 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
260 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
261 (movsi_aarch64): Likewise.
262 (load_pairsi): Likewise.
263 (load_pairdi): Likewise.
264 (store_pairsi): Likewise.
265 (store_pairdi): Likewise.
266 (load_pairsf): Likewise.
267 (load_pairdf): Likewise.
268 (store_pairsf): Likewise.
269 (store_pairdf): Likewise.
270 (zero_extend): Likewise.
271 (trunc): Swap alternatives.
272 (fcvt_target): Add '?' to prefer w over r.
274 2018-05-14 Jakub Jelinek <jakub@redhat.com>
277 * config/i386/i386.md: Disallow non-commutative arithmetics in
278 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
279 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
280 in the peephole2 before it.
282 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
284 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
285 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
286 (ix86_handle_option): Handle -mcldemote.
287 * config.gcc: New header.
288 * config/i386/cldemoteintrin.h: New file.
289 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
290 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
292 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
293 OPTION_MASK_ISA_CLDEMOTE.
294 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
295 (ix86_valid_target_attribute_inner_p): Ditto.
296 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
297 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
298 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
299 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
300 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
302 * config/i386/i386.opt: Add -mcldemote.
303 * config/i386/x86intrin.h: New header.
304 * doc/invoke.texi: Add -mcldemote.
306 2018-05-14 Richard Biener <rguenther@suse.de>
308 * doc/match-and-simplify.texi: Adjust :s documentation.
310 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
312 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
313 intended memcpy size.
314 (REORDER_45): Likewise.
316 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
319 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
320 * vec.c (qsort_chk): Use gcc_qsort.
321 * Makefile.in (OBJS-libcommon): Add sort.o.
322 (build/sort.o): New target. Use it...
323 (BUILD_RTL): ... here, and...
324 (build/gencfn-macros): ... here, and...
325 (build/genmatch): ... here.
327 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
328 Chung-Ju Wu <jasonwucj@gmail.com>
330 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
331 * config/nds32/nds32-graywolf.md: New file.
332 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
333 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
335 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
336 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
337 * config/nds32/nds32.md (pipeline_model): Add graywolf.
338 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
339 * config/nds32/pipelines.md: Include n15 settings.
341 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
342 Chung-Ju Wu <jasonwucj@gmail.com>
344 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
345 * config/nds32/nds32-n13.md: New file.
346 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
347 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
349 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
350 * config/nds32/nds32.md (pipeline_model): Add n13.
351 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
352 * config/nds32/pipelines.md: Include n13 settings.
354 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
355 Chung-Ju Wu <jasonwucj@gmail.com>
357 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
358 * config/nds32/nds32-n10.md: New file.
359 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
360 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
362 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
363 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
364 * config/nds32/nds32.md (pipeline_model): Add n10.
365 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
366 * config/nds32/pipelines.md: Include n10 settings.
368 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
369 Kito Cheng <kito.cheng@gmail.com>
370 Chung-Ju Wu <jasonwucj@gmail.com>
372 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
373 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
374 Add enum values for DSP extension instructions.
375 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
377 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
378 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
380 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
381 * config/nds32/nds32-dspext.md: New file for DSP implementation.
382 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
383 * config/nds32/nds32-intrinsic.md: Likewise.
384 * config/nds32/nds32_intrinsic.h: Likewise.
385 * config/nds32/nds32-md-auxiliary.c: Likewise.
386 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
387 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
388 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
389 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
390 * config/nds32/nds32-protos.h: New declarations for DSP extension.
391 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
392 TYPE_DMAC in switch statement.
393 * config/nds32/nds32.c: New checking and implementation for DSP
394 extension instructions.
395 * config/nds32/nds32.h: Likewise.
396 * config/nds32/nds32.md: Likewise.
397 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
398 * config/nds32/predicates.md: Implement new predicates for DSP
401 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
403 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
404 Reformat alternatives and attributes so it is easier to identify
405 which constraints/attributes go with which instruction.
406 (mov<mode>_hardfloat32, FMOVE64): Likewise.
407 (mov<mode>_softfloat32, FMOVE64): Likewise.
408 (mov<mode>_hardfloat64, FMOVE64): Likewise.
409 (mov<mode>_softfloat64, FMOVE64): Likewise.
411 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
413 * doc/extend.texi (PowerPC Built-in Functions): Rename this
415 (Basic PowerPC Built-in Functions): The new name of the
416 subsection previously known as "PowerPC Built-in Functions".
417 (Basic PowerPC Built-in Functions Available on all Configurations):
419 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
420 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
421 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
422 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
424 2018-05-11 Martin Jambor <mjambor@suse.cz>
427 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
430 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
433 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
435 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
437 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
438 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
439 (ix86_handle_option): Handle -mwaitpkg.
440 * config.gcc: New header.
441 * config/i386/cpuid.h (bit_WAITPKG): New bit.
442 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
443 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
445 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
446 OPTION_MASK_ISA_WAITPKG.
447 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
448 (ix86_option_override_internal): Add PTA_WAITPKG.
449 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
450 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
452 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
453 __builtin_ia32_umwait and __builtin_ia32_tpause.
454 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
455 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
456 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
457 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
458 UNSPECV_TPAUSE): New.
459 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
460 * config/i386/i386.opt: Add -mwaitpkg.
461 * config/i386/waitpkgintrin.h: New file.
462 * config/i386/x86intrin.h: New header.
463 * doc/invoke.texi: Add -mwaitpkg.
465 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
468 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
470 (cortex-m0): Use armv6s-m isa.
471 (cortex-m0plus): Likewise.
472 (cortex-m1): Likewise.
473 (cortex-m0.small-multiply): Likewise.
474 (cortex-m0plus.small-multiply): Likewise.
475 (cortex-m1.small-multiply): Likewise.
477 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
478 Jakub Jelinek <jakub@redhat.com>
480 PR tree-optimization/85692
481 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
482 source permute as well.
484 2018-05-11 Martin Liska <mliska@suse.cz>
487 * doc/extend.texi: Document LLVM style format for no_sanitize
490 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
492 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
493 mode_supports_vsx_dform_quad to mode_supports_dq_form.
494 (mode_supports_vsx_dform_quad): Likewise.
495 (mode_supports_vmx_dform): Move these functions to be next to the
496 other mode_supports functions.
497 (mode_supports_dq_form): Likewise.
498 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
499 mode_supports_dq_form.
500 (reg_offset_addressing_ok_p): Likewise.
501 (offsettable_ok_by_alignment): Likewise.
502 (rs6000_legitimate_offset_address_p): Likewise.
503 (legitimate_lo_sum_address_p): Likewise.
504 (rs6000_legitimize_address): Likewise.
505 (rs6000_legitimize_reload_address): Likewise.
506 (rs6000_secondary_reload_inner): Likewise.
507 (rs6000_preferred_reload_class): Likewise.
508 (rs6000_output_move_128bit): Likewise.
510 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
512 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
513 Generate SImode target register for null target.
514 <case IX86_BUILTIN_XGETBV>: Ditto.
515 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
516 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
518 2018-05-10 Carl Love <cel@us.ibm.com>
520 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
521 dcbtt and dcbtstt if operands[2] is 0.
523 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
526 * config/i386/sse.md (usadv64qi): New expander.
528 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
530 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
531 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
532 -maltivec=be support.
533 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
534 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
535 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
536 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
537 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
538 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
539 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
540 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
541 altivec_vsumsws): Adjust.
542 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
543 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
544 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
546 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
547 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
548 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
549 (altivec_lve<VI_char>x): Delete expand.
550 (*altivec_lve<VI_char>x_internal): Rename to...
551 (altivec_lve<VI_char>x): ... this.
552 (altivec_lvxl_<mode>): Delete expand.
553 (*altivec_lvxl_<mode>_internal): Rename to ...
554 (altivec_lvxl_<mode>): ... this.
555 (altivec_stvxl_<mode>): Delete expand.
556 (*altivec_stvxl_<mode>_internal): Rename to ...
557 (altivec_stvxl_<mode>): ... this.
558 (altivec_stve<VI_char>x): Delete expand.
559 (*altivec_stve<VI_char>x_internal): Rename to ...
560 (altivec_stve<VI_char>x): ... this.
561 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
562 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
563 reduc_plus_scal_<mode>): Adjust.
564 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
566 (rs6000_cpu_cpp_builtins): Adjust.
567 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
568 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
569 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
570 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
571 -maltivec=be support.
572 (rs6000_split_vec_extract_var): Adjust.
573 (rs6000_split_v4si_init): Adjust.
574 (swap_selector_for_mode): Delete.
575 (altivec_expand_lvx_be, altivec_expand_stvx_be,
576 altivec_expand_stvex_be): Delete.
577 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
578 -maltivec=be support.
579 (rs6000_gimple_fold_builtin): Ditto.
580 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
582 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
583 (TARGET_DIRECT_MOVE_64BIT): Adjust.
584 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
585 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
586 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
587 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
588 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
589 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
590 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
591 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
592 anonymous split): Adjust.
593 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
594 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
596 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
598 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
599 when --with-gxx-include-dir is also specified.
600 * configure: Regenerate.
602 2018-05-09 Jim Wilson <jimw@sifive.com>
605 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
606 * config/riscv/t-withmultilib: New.
607 * config/riscv/withmultilib.h: New.
608 * doc/install.texi: Document RISC-V --with-multilib-list support.
610 2018-05-09 Richard Biener <rguenther@suse.de>
612 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
614 (vect_bb_vectorization_profitable_p): Adjust. Compute
615 actual scalar cost using the cost vector and the add_stmt_cost
618 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
620 PR rtl-optimization/85645
621 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
622 in the REG_CFA_REGISTER note for LR, don't leave it empty.
624 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
626 PR rtl-optimization/85645
627 * shrink-wrap.c (spread_components): Return a boolean saying if
628 anything was changed.
629 (try_shrink_wrapping_separate): Iterate spread_components until
630 nothing changes anymore.
632 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
634 PR rtl-optimization/85645
635 * regrename.c (build_def_use): Also kill the chains that include the
636 destination of a REG_CFA_REGISTER note.
638 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
640 PR rtl-optimization/85645
641 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
642 insn that has a REG_CFA_REGISTER note.
644 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
646 * cfgexpand.c (expand_clobber): New function.
647 (expand_gimple_stmt_1): Use it.
648 * tree-vect-stmts.c (vect_clobber_variable): New function,
650 (vectorizable_simd_clone_call): ...here.
651 (vectorizable_store): Emit a clobber either side of an
652 IFN_STORE_LANES sequence.
653 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
655 2018-05-09 Tom de Vries <tom@codesourcery.com>
658 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
659 (define_insn "trap_if_false"): Add exit after trap.
661 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
663 PR rtl-optimization/85638
664 * bb-reorder.c: Include common/common-target.h.
665 (create_forwarder_block): New function extracted from...
666 (fix_up_crossing_landing_pad): ...here. Rename into...
667 (dw2_fix_up_crossing_landing_pad): ...this.
668 (sjlj_fix_up_crossing_landing_pad): New function.
669 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
670 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
671 from both partitions and exit the loop after one iteration.
673 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
676 * doc/extend.texi (PowerPC Built-in Functions): Rename this
678 (Basic PowerPC Built-in Functions): The new name of the
679 subsection previously known as "PowerPC Built-in Functions".
680 (Basic PowerPC Built-in Functions Available on all Configurations):
682 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
684 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
686 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
688 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
691 2018-05-08 Jim Wilson <jimw@sifive.com>
693 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
694 (LD_EMUL_SUFFIX): New.
697 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
699 * doc/extend.texi (PowerPC Built-in Functions): Rename this
701 (Basic PowerPC Built-in Functions): The new name of the
702 subsection previously known as "PowerPC Built-in Functions".
703 (Basic PowerPC Built-in Functions Available on all Configurations):
705 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
707 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
709 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
711 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
714 2018-05-08 Jakub Jelinek <jakub@redhat.com>
717 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
718 after cmpelim optimization.
720 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
722 * config.gcc: Support "goldmont".
723 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
724 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
726 * config/i386/i386.c (m_GOLDMONT): Define.
727 (processor_target_table): Add "goldmont".
728 (PTA_GOLDMONT): Define.
729 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
730 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
731 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
732 (fold_builtin_cpu): Add "goldmont".
733 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
734 (ix86_option_override_internal): Add "goldmont".
735 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
736 (processor_type): Add PROCESSOR_GOLDMONT.
737 * config/i386/i386.md: Add CPU "glm".
738 * config/i386/glm.md: New file.
739 * config/i386/x86-tune.def: Add m_GOLDMONT.
740 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
742 2018-05-08 Jakub Jelinek <jakub@redhat.com>
745 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
747 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
748 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
749 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
752 * config/i386/i386.c (ix86_fold_builtin): Handle
753 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
756 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
757 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
759 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
762 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
763 (check_arch): Likewise.
764 (check_fpu): Return the result rather than printing it.
765 (end arch): Fix operator precedence.
767 (END): Print the result from check_fpu.
769 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
770 Alan Hayward <alan.hayward@arm.com>
771 David Sherwood <david.sherwood@arm.com>
773 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
774 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
775 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
776 (*fcmuo<mode>_and): New patterns.
778 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
780 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
781 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
782 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
783 (cmp_op, sve_imm_con): New code attributes.
784 (SVE_COND_INT_CMP, imm_con): Delete.
785 (cmp_op): Remove above unspecs from int attribute.
786 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
788 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
789 comparison-specific unspecs.
790 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
791 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
792 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
793 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
794 (*vec_fcm<cmp_op><mode>): Rename to...
795 (*fcm<cmp_op><mode>): ...this and adjust likewise.
796 (*vec_fcmuo<mode>): Rename to...
797 (*fcmuo<mode>): ...this and adjust likewise.
798 (*pred_fcm<cmp_op><mode>): New pattern.
799 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
800 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
802 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
804 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
805 (aarch64_emit_sve_predicated_cond): New function.
806 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
807 (aarch64_emit_unspec_cond_or): Replace with...
808 (aarch64_emit_sve_or_conds): ...this new function. Use
809 aarch64_emit_sve_ptrue_op for the individual comparisons and
810 aarch64_emit_binop to OR them together.
811 (aarch64_emit_inverted_unspec_cond): Replace with...
812 (aarch64_emit_sve_inverted_cond): ...this new function. Use
813 aarch64_emit_sve_ptrue_op for the comparison and
814 aarch64_emit_unop to invert the result.
815 (aarch64_expand_sve_vec_cmp_float): Update after the above
816 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
818 2018-05-07 Nathan Sidwell <nathan@acm.org>
820 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
821 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
822 (Backwards Compatibility): Likewise.
824 2018-05-07 Luis Machado <luis.machado@linaro.org>
828 2018-05-07 Luis Machado <luis.machado@linaro.org>
830 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
831 <prefetch_dynamic_strides>: New const bool field.
832 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
833 prefetch_dynamic_strides.
834 (exynosm1_prefetch_tune): Likewise.
835 (thunderxt88_prefetch_tune): Likewise.
836 (thunderx_prefetch_tune): Likewise.
837 (thunderx2t99_prefetch_tune): Likewise.
838 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
839 (aarch64_override_options_internal): Update to set
840 PARAM_PREFETCH_DYNAMIC_STRIDES.
841 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
842 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
843 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
844 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
845 prefetch-dynamic-strides setting.
847 2018-05-07 Luis Machado <luis.machado@linaro.org>
849 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
850 <minimum_stride>: New const int field.
851 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
852 minimum_stride field.
853 (exynosm1_prefetch_tune): Likewise.
854 (thunderxt88_prefetch_tune): Likewise.
855 (thunderx_prefetch_tune): Likewise.
856 (thunderx2t99_prefetch_tune): Likewise.
857 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
858 (aarch64_override_options_internal): Update to set
859 PARAM_PREFETCH_MINIMUM_STRIDE.
860 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
861 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
862 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
863 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
864 stride is constant and is below the minimum stride threshold.
866 2018-05-07 Luis Machado <luis.machado@linaro.org>
868 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
871 2018-05-07 Luis Machado <luis.machado@linaro.org>
873 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
874 <prefetch_dynamic_strides>: New const bool field.
875 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
876 prefetch_dynamic_strides.
877 (exynosm1_prefetch_tune): Likewise.
878 (thunderxt88_prefetch_tune): Likewise.
879 (thunderx_prefetch_tune): Likewise.
880 (thunderx2t99_prefetch_tune): Likewise.
881 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
882 (aarch64_override_options_internal): Update to set
883 PARAM_PREFETCH_DYNAMIC_STRIDES.
884 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
885 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
886 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
887 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
888 prefetch-dynamic-strides setting.
890 2018-05-07 Luis Machado <luis.machado@linaro.org>
892 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
893 <minimum_stride>: New const int field.
894 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
895 minimum_stride field.
896 (exynosm1_prefetch_tune): Likewise.
897 (thunderxt88_prefetch_tune): Likewise.
898 (thunderx_prefetch_tune): Likewise.
899 (thunderx2t99_prefetch_tune): Likewise.
900 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
901 (aarch64_override_options_internal): Update to set
902 PARAM_PREFETCH_MINIMUM_STRIDE.
903 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
904 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
905 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
906 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
907 stride is constant and is below the minimum stride threshold.
909 2018-05-06 Jakub Jelinek <jakub@redhat.com>
912 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
913 the type is addressable. Don't force op into register if it has
916 2018-05-05 Roland McGrath <mcgrathr@google.com>
919 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
920 any section for which we don't know a specific type it should have,
921 regardless of name. Previously this was done only for the exact
922 names ".init_array", ".fini_array", and ".preinit_array".
923 (default_elf_asm_named_section): Add comment about
924 relationship with default_section_type_flags and SECTION_NOTYPE.
925 (get_section): Don't consider it a type conflict if one side has
926 SECTION_NOTYPE and the other doesn't, as long as neither has the
927 SECTION_BSS et al used in the default_section_type_flags logic.
929 2018-05-05 Tom de Vries <tom@codesourcery.com>
932 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
933 (workaround_barsyncs): New function.
934 (nvptx_reorg): Use workaround_barsyncs.
935 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
936 (define_expand "nvptx_membar_cta"): New define_expand.
937 (define_insn "*nvptx_membar_cta"): New insn.
939 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
941 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
942 To improve optimization opportunities.
943 * builtin-types.def: The new needed builtin types for the above.
945 2018-05-04 Richard Biener <rguenther@suse.de>
947 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
948 * gimple-ssa-store-merging.c
949 (imm_store_chain_info::output_merged_store): Remove redundant create,
950 release split_store vector contents on failure.
951 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
952 scalar stmt vector on cache hit.
954 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
956 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
958 * config.gcc (powerpc-xilinx-eabi*): Remove.
959 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
961 (fusion_addis_mem_combo_load): Ditto.
962 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
964 (rs6000_cpu_cpp_builtins): Ditto.
965 * config/rs6000/rs6000-linux.c
966 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
967 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
968 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
970 (rs6000_setup_reg_addr_masks): Ditto.
971 (rs6000_init_hard_regno_mode_ok): Ditto.
972 (rs6000_option_override_internal): Ditto.
973 (legitimate_lo_sum_address_p): Ditto.
974 (rs6000_legitimize_address): Ditto.
975 (rs6000_legitimize_reload_address): Ditto.
976 (rs6000_legitimate_address_p): Ditto.
977 (abi_v4_pass_in_fpr): Ditto.
978 (setup_incoming_varargs): Ditto.
979 (rs6000_gimplify_va_arg): Ditto.
980 (rs6000_split_multireg_move): Ditto.
981 (rs6000_savres_strategy): Ditto.
982 (rs6000_emit_prologue_components): Ditto.
983 (rs6000_emit_epilogue_components): Ditto.
984 (rs6000_emit_prologue): Ditto.
985 (rs6000_emit_epilogue): Ditto.
986 (rs6000_elf_file_end): Ditto.
987 (rs6000_function_value): Ditto.
988 (rs6000_libcall_value): Ditto.
989 * config/rs6000/rs6000.h: Ditto.
990 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
991 (TARGET_MINMAX): ... this. New.
992 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
993 * config/rs6000/rs6000.md: Remove Xilinx FP support.
994 (*movsi_internal1_single): Delete.
995 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
996 mfpu=, mxilinx-fpu): Delete.
997 * config/rs6000/singlefp.h: Delete.
998 * config/rs6000/sysv4.h: Remove Xilinx FP support.
999 * config/rs6000/t-rs6000: Ditto.
1000 * config/rs6000/t-xilinx: Delete.
1001 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1002 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1003 (VStype_simple): Delete.
1004 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1005 * config/rs6000/xfpu.h: Delete.
1006 * config/rs6000/xfpu.md: Delete.
1007 * config/rs6000/xilinx.h: Delete.
1008 * config/rs6000/xilinx.opt: Delete.
1009 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1010 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1012 2018-05-04 Tom de Vries <tom@codesourcery.com>
1015 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1018 2018-05-04 Richard Biener <rguenther@suse.de>
1021 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1022 (expand_complex_div_wide): Likewise.
1023 (expand_complex_operations_1): Likewise.
1024 (expand_complex_libcall): Preserve EH info of the original stmt.
1025 (tree_lower_complex): Handle removed blocks.
1026 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1027 on complex multiplication and division libcall builtins.
1029 2018-05-04 Richard Biener <rguenther@suse.de>
1032 * fold-const.c (negate_expr_p): Restrict negation of operand
1033 zero of a division to when we know that can happen without
1035 (fold_negate_expr_1): Likewise.
1037 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1040 * real.h (real_nextafter): Declare.
1041 * real.c (real_nextafter): New function.
1042 * fold-const-call.c (fold_const_nextafter): New function.
1043 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1044 CASE_CFN_NEXTTOWARD.
1045 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1046 even when arg1_mode is different from arg0_mode.
1048 2018-05-03 Nathan Sidwell <nathan@acm.org>
1050 * doc/extend.texi (Deprecated Features): Remove
1052 (Backwards Compatibility): Likewise.
1053 * doc/invoke.texi (C++ Language Options): Likewise.
1054 (C++ Dialect Options): Likewise.
1056 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1059 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1060 _mm512_mask_mullox_epi64): New intrinsics.
1062 2018-05-03 Tom de Vries <tom@codesourcery.com>
1065 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1066 dump files): Add offload-tree.
1068 2018-05-03 Richard Biener <rguenther@suse.de>
1070 PR tree-optimization/85615
1071 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1072 to loops not nested in BBs loop father to avoid creating multi-entry
1075 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1077 PR tree-optimization/70291
1078 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1079 arguments. Change return type to tree. Emit libcall as a new
1080 statement rather than replacing existing one when inplace_p is true.
1081 (expand_complex_multiplication_components): New function.
1082 (expand_complex_multiplication): Expand floating-point complex
1083 multiplication using the above.
1084 (expand_complex_division): Rename inner_type parameter to type.
1085 Update expand_complex_libcall call-site.
1086 (expand_complex_operations_1): Update expand_complex_multiplication
1087 and expand_complex_division call-sites.
1089 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1092 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1093 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1094 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1095 the highest significant bit of the shift count mask is clear. In
1096 check whether and[sq]i3 is needed verify that all significant bits
1097 of the shift count other than the highest are set.
1099 2018-05-02 Tom de Vries <tom@codesourcery.com>
1102 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1103 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1104 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1105 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1106 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1107 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1108 __builtin_goacc_parlevel_size.
1110 2018-05-02 Richard Biener <rguenther@suse.de>
1112 PR tree-optimization/85597
1113 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1114 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1117 2018-05-02 Tom de Vries <tom@codesourcery.com>
1120 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1121 dump files): Add ltrans-tree.
1123 2018-05-02 Tom de Vries <tom@codesourcery.com>
1126 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1127 dump files): Add wpa-ipa.
1129 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1131 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1132 powerpc*-*-linux*paired* target.
1133 * config/rs6000/750cl.h: Delete.
1134 * config/rs6000/paired.h: Delete.
1135 * config/rs6000/paired.md: Delete.
1136 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1138 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1139 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1140 comment. Remove paired float support.
1141 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1142 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1144 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1145 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1147 * config/rs6000/rs6000.c: Remove paired float support.
1148 (paired_expand_vector_init, paired_expand_vector_move,
1149 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1150 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1151 paired_expand_builtin, paired_expand_predicate_builtin,
1152 paired_init_builtins): Delete.
1153 * config/rs6000/rs6000.h: Remove paired float support.
1154 * config/rs6000/rs6000.md: Remove paired float support.
1155 (move_from_CR_ov_bit): Delete.
1156 * config/rs6000/rs6000.opt (mpaired): Delete.
1157 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1158 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1160 2018-05-02 Richard Biener <rguenther@suse.de>
1163 * gimplify.c (gimplify_save_expr): When in SSA form allow
1164 SAVE_EXPRs to compute to SSA vars.
1166 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1169 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1170 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1171 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1172 clobber operands[2], instead use a new pseudo. Formatting fixes.
1174 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1176 PR tree-optimization/85586
1177 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1178 exit early for statements in the same group if the accesses are
1181 2018-05-02 Tom de Vries <tom@codesourcery.com>
1184 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1187 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1189 PR tree-optimization/85143
1190 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1192 2018-05-01 Tom de Vries <tom@codesourcery.com>
1195 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1196 not found" error message.
1198 2018-05-01 Tom de Vries <tom@codesourcery.com>
1201 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1202 * vec.c (test_ordered_remove_if): New function.
1203 (vec_c_tests): Call test_ordered_remove_if.
1204 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1205 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1206 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1207 VEC_ORDERED_REMOVE_IF.
1209 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1211 PR tree-optimization/82665
1212 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1213 pointer subtraction where arguments come from a memchr call.
1215 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1217 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1218 --push-state --as-needed and --pop-state instead of --as-needed and
1219 --no-as-needed if ld supports it.
1220 * configure: Regenerated.
1223 * doc/install.texi2html: Replace _002d with - and _002a with * in
1224 generated html files using sed.
1226 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1229 * gcc-rich-location.c (blank_line_before_p): New function.
1230 (use_new_line): New function.
1231 (gcc_rich_location::add_fixit_insert_formatted): New function.
1232 * gcc-rich-location.h
1233 (gcc_rich_location::add_fixit_insert_formatted): New function.
1235 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1237 * selftest.c (assert_streq): Rename "expected" and "actual" to
1238 "val1" and "val2". Extend NULL-handling to cover both inputs
1239 symmetrically, while still requiring both to be non-NULL for a pass.
1240 * selftest.h (assert_streq): Rename "expected" and "actual" to
1242 (ASSERT_EQ): Likewise.
1243 (ASSERT_EQ_AT): Likewise.
1244 (ASSERT_KNOWN_EQ): Likewise.
1245 (ASSERT_KNOWN_EQ_AT): Likewise.
1246 (ASSERT_NE): Likewise.
1247 (ASSERT_MAYBE_NE): Likewise.
1248 (ASSERT_MAYBE_NE_AT): Likewise.
1249 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1250 the assertion to pass.
1251 (ASSERT_STREQ_AT): Likewise.
1253 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1255 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1256 interaction with -pie.
1258 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1260 * selftest.h: Fix alphabetization of per-source-file selftest
1263 2018-04-30 Jason Merrill <jason@redhat.com>
1265 PR c++/61982 - dead stores to destroyed objects.
1266 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1269 2018-04-30 Jason Merrill <jason@redhat.com>
1271 * tree.c (build_clobber): New.
1272 * tree.h: Declare it.
1273 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1275 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1277 * diagnostic-show-locus.c (layout::layout): Update for
1278 location_get_source_line returning a char_span.
1279 (struct char_span): Move to input.h.
1280 (struct correction): Update for fields in char_span becoming
1282 (struct source_line): Update for location_get_source_line
1283 returning a char_span.
1284 (layout::print_line): Likewise.
1285 * edit-context.c (edited_file::print_content): Likewise.
1286 (edited_file::print_diff_hunk): Likewise.
1287 (edited_file::print_run_of_changed_lines): Likewise.
1288 (edited_file::get_num_lines): Likewise.
1289 (edited_line::edited_line): Likewise.
1290 * final.c (asm_show_source): Likewise.
1291 * input.c (location_get_source_line): Convert return type
1292 from const char * to char_span, losing the final "line_len"
1294 (dump_location_info): Update for the above.
1295 (get_substring_ranges_for_loc): Likewise. Use a char_span
1296 when handling the literal within the line.
1297 (test_reading_source_line): Update for location_get_source_line
1298 returning a char_span.
1299 * input.h (class char_span): Move here from
1300 diagnostic-show-locus.c, converting from a struct to a class.
1301 Make data members private.
1302 (char_span::operator bool): New.
1303 (char_span::length): New.
1304 (char_span::get_buffer): New.
1305 (char_span::operator[]): New.
1306 (char_span::subspan): Make const.
1307 (char_span::xstrdup): New.
1308 (location_get_source_line): Convert return type from const char *
1309 to char_span, losing the final "line_size" param.
1311 2018-04-30 Jan Hubicka <jh@suse.cz>
1313 * lto-wrapper.c (ltrans_priorities): New static var.
1314 (cmp_priority): New.
1315 (run_gcc): Read priorities and if doing parallel build order
1316 the Makefile by them.
1318 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1320 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1322 2018-04-30 Richard Biener <rguenther@suse.de>
1324 * tree-cfg.c (verify_address): Remove base argument, add
1325 flag whether to check TREE_ADDRESSABLE and do that.
1326 (verify_expr): Remove.
1327 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1328 (verify_gimple_assign_single): Likewise.
1329 (verify_gimple_switch): Likewise.
1330 (verify_expr_location_1): Dereference tp once. Add (disabled)
1331 piece from verify_expr.
1332 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1334 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1336 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1338 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1340 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1341 (small_data_pattern): Likewise.
1342 (arc_rewrite_small_data): Likewise.
1343 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1344 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1345 (get_symbol_alignment): New function.
1346 (legitimate_small_data_address_p): Likewise.
1347 (legitimate_scaled_address): Update, call
1348 legitimate_small_data_address_p.
1349 (output_sdata): New static variable.
1350 (arc_print_operand): Update how we handle small data operands.
1351 (arc_print_operand_address): Likewise.
1352 (arc_legitimate_address_p): Update, use
1353 legitimate_small_data_address_p.
1354 (arc_rewrite_small_data_p): Remove.
1355 (arc_rewrite_small_data_1): Likewise.
1356 (arc_rewrite_small_data): Likewise.
1357 (small_data_pattern): Likewise.
1358 (compact_sda_memory_operand): Update to use
1359 legitimate_small_data_address_p and get_symbol_alignment.
1360 (prepare_move_operands): Don't rewite sdata pattern.
1361 (prepare_extend_operands): Remove.
1362 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1364 (zero_extendqisi2): Likewise.
1365 (zero_extendhisi2): Likewise.
1366 (extendqihi2): Likewise.
1367 (extendqisi2): Likewise.
1368 (extendhisi2): Likewise.
1372 * config/arc/constraints.md (Usd): Change it to memory constraint.
1374 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1376 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1377 as source of std instructions.
1378 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1379 allow 6-bit constants as source for store instructions.
1380 (movdi_insn): Update instruction pattern to allow 6-bit constants
1381 as source for store instructions.
1383 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1385 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1387 2018-04-30 Nathan Sidwell <nathan@acm.org>
1388 Sandra Loosemore <sandra@codesourcery.com>
1390 * dumpfile.c (dump_open): Allow '-' for stdout.
1391 * doc/invoke.texi (Developer Options): Document dump filename
1392 determination early. Document stdin/stdout selection.
1394 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1396 Microblaze Target: PIC data text relative
1398 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1399 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1401 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1402 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1403 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1404 New addressing mode for data-text relative position indepenedent code.
1405 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1406 'ADDRESS_SYMBOLIC_TXT_REL'.
1407 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1408 (microblaze_legitimate_pic_operand): Exclude function calls from
1409 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1410 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1412 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1413 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1414 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1415 for 'address + offset'.
1416 (microblaze_expand_prologue): Add new function prologue call for
1418 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1419 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1420 table in case of TARGET_PIC_DATA_TEXT_REL.
1421 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1422 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1423 Add new macros 'UNSPEC_TEXT',
1424 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1425 + exclude function calls from 'UNSPEC_PLT' in case of data text
1427 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1428 new target hook for generating address diff vector tables in case of
1430 * doc/tm.texi : Regenerate.
1431 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1432 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1433 of addr diff vector generation.
1434 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1435 target hook definition.
1436 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1437 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1438 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1439 option for data text relative.
1441 2018-04-30 Richard Biener <rguenther@suse.de>
1443 * tree-chrec.h (evolution_function_is_constant_p): Remove
1445 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1447 2018-04-30 Richard Biener <rguenther@suse.de>
1450 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1452 2018-04-30 Richard Biener <rguenther@suse.de>
1454 PR tree-optimization/28364
1455 PR tree-optimization/85275
1456 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1457 copying first exit test.
1459 2018-04-28 Mark Wielaard <mark@klomp.org>
1461 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1463 (dwarf_AT): Handle DW_AT_addr_base.
1464 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1466 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1469 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1470 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1471 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1472 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1474 2018-04-28 Richard Biener <rguenther@suse.de>
1476 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1477 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1478 to reflect use. Only add interesting stmts.
1480 2018-04-27 Martin Jambor <mjambor@suse.cz>
1483 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1484 the jump function allows for passing through aggregate values.
1486 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1488 * input.h (in_system_header_at): Convert from macro to inline
1490 (from_macro_expansion_at): Likewise.
1491 (from_macro_definition_at): Likewise.
1493 2018-04-27 Jeff Law <law@redhat.com>
1495 * config.gcc: Mark tile* targets as deprecated/obsolete.
1497 2018-04-27 Richard Biener <rguenther@suse.de>
1499 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1502 2018-04-27 Richard Biener <rguenther@suse.de>
1504 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1506 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1508 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1509 with Yd constraint. Set "preferred_for_speed" attribute from
1510 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1512 (*movdi_internal): Ditto.
1513 (movti_interunit splitters): Remove
1514 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1515 (movdi_interunit splitters): Ditto.
1516 * config/i386/constraints.md (Ye): Remove.
1517 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1519 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1522 * config/aarch64/constraints.md (Usg): Limit to 31.
1525 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1527 PR tree-optimization/85529
1528 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1529 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1530 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1531 zero extension or masking of the MSB bit.
1532 (optimize_range_tests): Add FIRST_BB argument, pass it through
1533 to optimize_range_tests_var_bound.
1534 (maybe_optimize_range_tests, reassociate_bb): Adjust
1535 optimize_range_tests callers.
1537 2018-04-26 Richard Biener <rguenther@suse.de>
1538 Jakub Jelinek <jakub@redhat.com>
1540 * cgraph.h (symbol_table): Just declare debug method here.
1541 * symtab.c (symbol_table::debug): Define.
1543 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1545 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1547 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1549 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1550 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1551 (*movdi_internal): Substitute Yi and Yj constraint with x
1552 and Ym and Yn constraint with y constraint. Update "isa"
1553 attribute and set "preferred_for_speed" attribute from
1554 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1555 (*movsi_internal): Ditto.
1556 (*movdf_internal): Ditto.
1557 (*movsf_internal): Ditto.
1558 (*zero_extendsidi2): Ditto.
1559 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1560 (sse2_loadld): Ditto.
1561 (*vec_extract<ssevecmodelower>_0): Ditto.
1562 (*vec_extractv4si_0_zext_sse4): Ditto.
1563 (vec_concatv2di): Ditto.
1564 (*vec_dup<mode>): Ditto.
1565 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1566 * config/i386/constraints.md (Yi): Remove.
1571 2018-04-26 Nathan Sidwell <nathan@acm.org>
1573 * dumpfile.c (dump_open): New.
1574 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1575 (dump_finish): Detect stdio/stderr by value not name.
1577 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1579 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1581 2018-04-26 Tom de Vries <tom@codesourcery.com>
1584 * config/nvptx/nvptx.c (verify_neutering_jumps)
1585 (verify_neutering_labels): New function
1586 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1588 2018-04-26 Tom de Vries <tom@codesourcery.com>
1591 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1592 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1595 2018-04-26 Richard Biener <rguenther@suse.de>
1596 Tom de Vries <tom@codesourcery.com>
1599 * lto-streamer-out.c (output_function): Fixup loops if required to match
1600 discovery done in the reader.
1602 2018-04-26 Richard Biener <rguenther@suse.de>
1604 PR tree-optimization/85116
1605 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1606 have a loop exit from the single latch predecessor. Remove
1607 case of header with just condition.
1608 (ch_base::copy_headers): Exclude infinite loops from any
1610 (pass_ch::execute): Record exits.
1612 2018-04-26 Richard Biener <rguenther@suse.de>
1614 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1615 prologue cost vector and pass it to vect_get_load_cost.
1616 (vect_get_peeling_costs_all_drs): Likewise.
1617 (vect_peeling_hash_get_lowest_cost): Likewise.
1618 (vect_enhance_data_refs_alignment): Likewise.
1620 2018-04-26 Richard Biener <rguenther@suse.de>
1623 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1624 checking of integer<->pointer conversions.
1625 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1626 sign-/zero-extending pointer types.
1627 (expand_omp_for_static_chunk): Likewise.
1629 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1630 Jean Lee <xiaoyur347@gmail.com>
1632 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1633 (TARGET_ASAN_SHADOW_OFFSET): Define.
1634 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1635 true for -fsanitize=address.
1637 2018-04-25 Mark Wielaard <mark@klomp.org>
1639 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1642 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1644 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1645 than "alu", remove explicit "memory" and "imm_disp" attributes.
1646 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1649 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1650 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1651 gen_lowpart_no_emit.
1653 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1656 * config/i386/i386.c (ix86_expand_builtin): Change memory
1657 operand to XI, extend p0 to Pmode.
1658 * config/i386/i386.md: Change unspec volatile and operand
1659 1 mode to XI, change operand 0 mode to P.
1661 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1663 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1664 GET_MODE_MASK before any checking.
1665 (nds32_can_use_bset_p): Likewise.
1666 (nds32_can_use_btgl_p): Likewise.
1668 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1670 * config/nds32/nds32-doubleword.md: New define_split pattern for
1671 illegal register number.
1673 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1675 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1677 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1679 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1681 2018-04-25 Richard Biener <rguenther@suse.de>
1683 * lto-streamer.h (LTO_major_version): Bump to 8.
1685 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1687 * BASE-VER: Set to 9.0.0.
1689 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1691 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1692 in __abskf2 and __powikf2.
1694 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1697 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1698 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1699 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1700 Use the above on operand 2. Reindent.
1701 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1703 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1706 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1708 * config/i386/i386.opt (mcet): Removed.
1709 * doc/install.texi: Remove -mcet documentation.
1710 * doc/invoke.texi: Likewise.
1712 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1715 * doc/install.texi: Remove -mcet from bootstrap-cet.
1717 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1720 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1721 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1725 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1726 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1727 containing a CONST_VECTOR.
1729 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1731 * doc/install.texi: Update newlib dependency for nvptx.
1733 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1736 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1737 instead of INTVAL when shifting x left.
1739 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1741 PR tree-optimization/85478
1742 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1743 vect_grouped_store_supported for single element vectors.
1745 2018-04-24 Richard Biener <rguenther@suse.de>
1748 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1749 load cost increase to the case of non-constant step.
1751 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1754 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1755 destination if any_malformed_asm.
1757 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1760 * expr.c (store_field): In the bitfield case, if the value comes from
1761 a function call and is returned in registers by means of a PARALLEL,
1762 do not change the mode of the temporary unless BLKmode and VOIDmode.
1764 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1766 PR rtl-optimization/85423
1767 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1768 dependencies to debug insns when the previous insn is non-debug.
1770 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1772 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1773 enums into a single definition.
1774 (fls): Fix predicates and printing.
1777 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1779 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1780 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1781 and short u6 immediate.
1782 (check_if_valid_sleep_operand): Remove.
1783 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1785 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1787 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1788 flag_always_save_lp condition.
1789 * config/nds32/nds32.opt (malways-save-lp): New option.
1791 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1793 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1794 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1795 * config/nds32/nds32.h
1796 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1797 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1799 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1801 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1802 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1804 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1805 Chung-Ju Wu <jasonwucj@gmail.com>
1807 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1809 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1810 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1812 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1814 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1816 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1818 * config/nds32/nds32-protos.h (nds32_data_alignment,
1819 nds32_local_alignment): Declare.
1820 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1821 nds32_local_alignment): New functions.
1822 (TARGET_CONSTANT_ALIGNMENT): Define.
1823 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1825 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1827 * config/nds32/nds32.c
1828 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1829 (TARGET_MODES_TIEABLE_P): Likewise.
1831 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1833 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1836 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1837 Chung-Ju Wu <jasonwucj@gmail.com>
1839 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1840 for unaligned access.
1841 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1843 * config/nds32/nds32-intrinsic.md: Likewise.
1844 * config/nds32/nds32_intrinsic.h: Likewise.
1845 * config/nds32/nds32.h (nds32_builtins): Likewise.
1846 * config/nds32/nds32.opt (munaligned-access): New option.
1847 * config/nds32/nds32.c (nds32_asm_file_start): Display
1848 flag_unaligned_access status.
1850 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1852 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1853 -mno-relax is present.
1854 * config/riscv/linux.h (LINK_SPEC): Ditto.
1856 2018-04-20 Martin Sebor <msebor@redhat.com>
1859 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1861 (gimple_fold_builtin_stxcpy_chk): Same.
1862 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1864 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1867 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1868 __powikf2 when long double is IEEE 128-bit.
1870 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1872 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1873 step to make sure stack always aligned.
1875 2018-04-20 Carl Love <cel@us.ibm.com>
1878 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1879 size check for arg0.
1881 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1882 Tom de Vries <tom@codesourcery.com>
1885 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1886 Emit insns for calls too.
1887 (nvptx_find_par): Always look for worker-level predecessor insn.
1888 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1890 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1891 (nvptx_process_pars): Propagate frames for calls.
1893 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1896 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1898 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1899 (ix86_handle_option): Don't handle OPT_mibt.
1900 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1902 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1904 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1905 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1906 (ix86_target_macros): Define __CET__ with flag_cf_protection
1907 for -fcf-protection.
1908 * config/i386/i386.c (isa2_opts): Remove -mibt.
1909 * config/i386/i386.h (TARGET_IBT): Removed.
1910 (TARGET_IBT_P): Likewise.
1911 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1912 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1913 * config/i386/i386.opt (mcet): Update help message.
1916 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1917 -mcet as an alias for -mshstk.
1919 2018-04-20 Richard Biener <rguenther@suse.de>
1922 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1923 complexity by forcing a single use of the multiply operand.
1925 2018-04-20 Martin Jambor <mjambor@suse.cz>
1928 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1929 recursion dependency to only apply to non-clones.
1931 2018-04-20 Martin Jambor <mjambor@suse.cz>
1934 * ipa-cp.c (create_specialized_node): Check that clones of
1935 self-recursive edges exist during IPA-CP.
1937 2018-04-19 Toon Moene <toon@moene.org>
1939 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1942 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1944 PR tree-optimization/85467
1945 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1946 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1947 VECTOR_CST element to type.
1949 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1952 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1953 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1954 (builtin_longjmp): Likewise.
1955 (save_stack_nonlocal): New pattern.
1956 (restore_stack_nonlocal): Likewise.
1958 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1961 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1962 Replace ASM_OUTPUT_LABEL with fprintf.
1964 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1967 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1968 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1969 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1970 define __IBT__ and __SHSTK__ for -fcf-protection.
1971 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1973 (ix86_trampoline_init): Likewise.
1974 (x86_output_mi_thunk): Likewise.
1975 (ix86_notrack_prefixed_insn_p): Likewise.
1976 (ix86_option_override_internal): Don't disallow -fcf-protection.
1977 * config/i386/i386.md (rdssp<mode>): Also enable for
1979 (incssp<mode>): Likewise.
1980 (nop_endbr): Likewise.
1981 * config/i386/i386.opt (mcet): Change help message to built-in
1985 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1986 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1987 enable CET built-in functions.
1989 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1991 * common/config/i386/i386-common.c
1992 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1993 OPTION_MASK_ISA_MOVDIRI_UNSET,
1994 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1995 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1996 * config.gcc (movdirintrin.h): New header.
1997 * config/i386/cpuid.h (bit_MOVDIRI,
1998 bit_MOVDIR64B): New bits.
1999 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2001 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2002 (VOID, PVOID, PCVOID)): New function types.
2003 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2004 __builtin_ia32_directstoreu_u64,
2005 __builtin_ia32_movdir64b): New builtins.
2006 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2007 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2009 (ix86_valid_target_attribute_inner_p): Ditto.
2010 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2011 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2012 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2013 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2014 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2015 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2016 (movdiri<mode>, movdir64b_<mode>): New.
2017 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2018 * config/i386/immintrin.h: Include movdirintrin.h.
2019 * config/i386/movdirintrin.h: New file.
2020 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2022 2018-04-19 Richard Biener <rguenther@suse.de>
2025 * cfg.c (clear_bb_flags): When loop state says we have
2026 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2028 2018-04-19 Richard Biener <rguenther@suse.de>
2030 PR tree-optimization/84737
2031 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2032 copying restrict info.
2033 (vect_setup_realignment): Use it.
2034 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2035 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2036 the first DR to all generated stores.
2037 (vectorizable_load): Likewise for loads.
2039 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2041 PR tree-optimization/85446
2042 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2043 the integral and pointer types to have the same precision.
2045 * doc/install.texi: Document --disable-cet being the default and
2048 2018-04-18 Martin Liska <mliska@suse.cz>
2050 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2053 2018-04-18 Martin Liska <mliska@suse.cz>
2056 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2059 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2060 arguments if they are comparable.
2062 2018-04-18 Martin Liska <mliska@suse.cz>
2065 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2068 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2071 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2074 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2075 ENDBR after calling __morestack.
2077 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2080 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2081 by using gcc_base_ver to generate a gcc_driver_version, and use
2082 it when generating GCC_DRIVER_NAME.
2083 * configure: Regenerate.
2085 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2088 * config.gcc: Obsolete powerpc*-*-*spe*.
2090 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2093 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2094 (stabstr_D): Change type of unum from unsigned int to
2095 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2098 2018-04-17 Jim Wilson <jimw@sifive.com>
2101 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2102 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2103 Set arg_pointer_offset after using pretend_args_size.
2105 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2107 PR rtl-optimization/85431
2108 * dse.c (record_store): Ignore zero width stores.
2111 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2112 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2113 __builtin_stack_restore rather than after it.
2114 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2115 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2116 argument instead of virtual_dynamic_stack_rtx.
2118 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2120 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2122 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2123 Add note to error message to explain internal mapping of overloaded
2124 built-in function name to non-overloaded built-in function name.
2125 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2128 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2131 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2132 where the inputs overlap with the output.
2134 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2136 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2137 (=v, v) alternative and explicit "memory" attribute.
2138 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2139 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2141 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2142 "sselog1" type instead of "sselog".
2143 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2144 "sselog". Remove explicit "memory" attribute.
2145 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2146 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2148 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2149 "sselog1" type instead of "sselog". Remove explicit "memory"
2151 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2152 use "sselog1" type instead of "sselog". Remove explicit "memory"
2154 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2155 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2157 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2158 "sselog1" type instead of "sselog". Remove explicit "memory"
2160 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2161 use "sselog1" type instead of "sselog". Remove explicit "memory"
2165 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2168 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2171 2018-04-17 Martin Jambor <mjambor@suse.cz>
2174 * ipa-cp.c (create_specialized_node): Call
2175 expand_all_artificial_thunks if necessary.
2177 2018-04-17 Martin Liska <mliska@suse.cz>
2180 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2181 in message, remote space in between '_G' and '('.
2183 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2186 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2187 avx512f_vmcmp<mode>3<round_saeonly_name>,
2188 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2189 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2190 avx512f_rndscale<mode><round_saeonly_name>,
2191 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2192 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2193 Use %<iptr>2 instead of %2 for -masm=intel.
2194 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2195 avx512f_vcvttss2usi<round_saeonly_name>,
2196 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2198 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2199 avx512f_vcvttsd2usi<round_saeonly_name>,
2200 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2201 Use %q1 instead of %1 for -masm=intel.
2202 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2203 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2204 of %3 for -masm=intel.
2205 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2207 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2209 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2211 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2212 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2214 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2216 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2217 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2218 %g1 and one with %0 and %1.
2219 (avx512er_vmrcp28<mode><round_saeonly_name>,
2220 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2222 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2223 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2224 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2225 of %0 and %{%4%} for -masm=intel.
2226 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2227 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2228 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2229 order of %0 and %{%5%}%{z%} for -masm=intel.
2231 2018-04-17 Jan Hubicka <jh@suse.cz>
2234 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2236 2018-04-17 Martin Liska <mliska@suse.cz>
2239 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2240 for target_clone error message. Make default implementation
2241 clone to be a local declaration.
2242 (separate_attrs): Add new argument and check for an empty
2244 (expand_target_clones): Handle it.
2245 (ipa_target_clone): Make redirection just for target_clones
2248 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2249 Tom de Vries <tom@codesourcery.com>
2252 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2253 tiled basic blocks without omp continue statements.
2255 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2258 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2259 vec_extract expression as having side effects to make sure it gets
2262 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2265 * config/i386/i386.c (get_builtin_code_for_version): Check
2268 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2271 * gcc/config.gcc: Support "skylake".
2272 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2274 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2275 (processor_target_table): Add "skylake".
2276 (ix86_option_override_internal): Add "skylake".
2277 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2278 PROCESSOR_CANNONLAKE.
2279 (get_builtin_code_for_version): Fix priority for
2280 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2281 PROCESSOR_SKYLAKE-AVX512.
2282 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2283 (processor_type): Add PROCESSOR_SKYLAKE.
2285 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2286 Jason Merrill <jason@redhat.com>
2289 * convert.c (convert_to_integer_1): Use direct recursion for
2290 enumeral types and types with a precision less than the number
2291 of bits in their mode.
2293 2018-04-16 Julia Koval <julia.koval@intel.com>
2296 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2297 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2299 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2302 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2303 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2304 and -mno-direct-move.
2306 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2309 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2310 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2311 Ensure negative shifts result in {0}.
2313 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2315 PR rtl-optimization/79916
2316 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2317 regs (if any) to define how to gnerate SD moves when LRA is in
2320 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2322 PR rtl-optimization/85393
2323 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2324 * except.c (expand_dw2_landing_pad_for_region): Make static.
2325 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2326 a label and unconditional jump to old_bb, rather than
2327 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2330 PR rtl-optimization/85376
2331 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2332 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2333 instead of a specific value.
2335 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2336 Bin Cheng <bin.cheng@arm.com>
2338 PR tree-optimization/82965
2339 PR tree-optimization/83991
2340 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2341 by_profile_only parameter.
2342 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2343 information if the loop was predicted to iterate too many times.
2344 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2346 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2349 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2352 2018-04-13 Martin Liska <mliska@suse.cz>
2353 Jakub Jelinek <jakub@redhat.com>
2356 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2357 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2358 * builtins.c (expand_builtin_memory_copy_args): Use
2359 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2360 handle dest_addr == pc_rtx.
2362 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2365 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2366 asked to not generate direct moves.
2367 (fix_trunc<mode>si2_stfiwx): Similar.
2368 (fix_trunc<mode>si2_internal): Similar.
2370 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2373 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2374 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2375 lookup if dest in some wider mode is known to be const0_rtx and
2376 if so, record permanent equivalence for it to be ZERO_EXTEND of
2377 the narrower mode destination.
2379 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2381 * lto-streamer-out.c (output_function): Revert 259346.
2382 * omp-expand.c (expand_oacc_for): Likewise.
2384 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2386 PR rtl-optimization/85354
2387 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2388 * sel-sched.c (sel_global_init): ... here.
2390 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2393 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2394 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2395 mode for PE-COFF targets.
2396 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2397 (i386_pe_asm_lto_end): Likewise.
2398 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2399 (TARGET_ASM_LTO_END): Likewise.
2400 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2401 (i386_pe_asm_lto_start): New function.
2402 (i386_pe_asm_lto_end): Likewise.
2404 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2405 Richard Biener <rguenther@suse.de>
2408 * lto-streamer-out.c (output_function): Fix CFG loop state before
2410 * omp-expand.c (expand_oacc_for): Handle calls to internal
2411 functions like regular functions.
2413 2018-04-12 Richard Biener <rguenther@suse.de>
2416 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2417 for the early LTO debug to properly generate references to it
2418 during DIE emission. Do not re-use that for the skeleton for
2420 (dwarf2out_early_finish): Likewise.
2422 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2425 * config/i386/sse.md
2426 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2427 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2428 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2429 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2430 and output is a reg, avoid creating invalid lowpart subreg, but
2431 instead split into a 512-bit move. Don't split if not AVX512VL,
2432 input is xmm16+ reg and output is a mem.
2433 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2434 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2435 xmm16+ reg and output is a mem.
2437 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2439 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2440 also for flag_dwarf2_cfi_asm.
2442 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2444 PR rtl-optimization/85342
2445 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2446 a bool scalar var inside of the loop instead. Don't try to update
2447 recog_data.operand after failed apply_change_group.
2449 2018-04-12 Tom de Vries <tom@codesourcery.com>
2452 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2453 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2454 array with flexible array member as array without given dimension.
2455 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2456 argument for undefined param to true.
2458 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2461 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2462 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2463 from PowerPC section.
2464 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2465 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2466 help text that is too long.
2467 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2468 help text that is too long.
2469 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2470 help text that is too long.
2472 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2474 * config/alpha/alpha.md (stack_probe_internal): Rename
2475 from "probe_stack". Update all callers.
2477 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2479 PR rtl-optimization/84566
2480 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2481 sched_macro_fuse_insns.
2483 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2486 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2487 (compute_block_dependences): ... from here.
2489 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2491 PR tree-optimization/85331
2492 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2493 from int to HOST_WIDE_INT.
2495 2018-04-11 Martin Jambor <mjambor@suse.cz>
2498 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2499 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2500 not the same as the source val.
2501 (cgraph_edge_brings_value_p): New parameter.
2502 (gather_edges_for_value): Pass destination value to
2503 cgraph_edge_brings_value_p.
2504 (perhaps_add_new_callers): Likewise.
2505 (get_info_about_necessary_edges): Likewise and exclude values brought
2506 only by self-recursive edges.
2507 (create_specialized_node): Redirect only clones of self-calling edges.
2508 (+self_recursive_pass_through_p): New function.
2509 (find_more_scalar_values_for_callers_subset): Use it.
2510 (find_aggregate_values_for_callers_subset): Likewise.
2511 (known_aggs_to_agg_replacement_list): Removed.
2512 (decide_whether_version_node): Re-calculate known constants for all
2513 remaining context clones.
2515 2018-04-11 Richard Biener <rguenther@suse.de>
2518 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2519 from early DWARF output.
2520 (dwarf2out_early_finish): Output line info unconditionally into
2521 early DWARF and add reference to it.
2523 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2526 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2527 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2528 other than V2DFmode using iptr mode attribute.
2529 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2531 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2533 PR rtl-optimization/84659
2534 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2536 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2539 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2541 (output_loc_list): Pass address of a dummy size variable even in the
2542 locview handling loop.
2543 (index_location_lists): Add comment on why skip_loc_list_entry can't
2546 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2549 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2552 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2555 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2556 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2557 and -mstring-compare-inline-limit.
2559 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2562 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2563 for stack clash protection in a register whenever we need it to be in
2566 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2568 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2569 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2571 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2574 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2576 (mlong-double-): Ditto.
2577 * config/rs6000/sysv4.opt (msdata=): Ditto.
2578 (mtls-size=): Ditto.
2580 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2582 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2583 erroneous entries for
2584 "vector int vec_ldl (int, long int *)", and
2585 "vector unsigned int vec_ldl (int, unsigned long int *)".
2586 Add comments and entries for
2587 "vector bool char vec_ldl (int, bool char *)",
2588 "vector bool short vec_ldl (int, bool short *)",
2589 "vector bool int vec_ldl (int, bool int *)",
2590 "vector bool long long vec_ldl (int, bool long long *)",
2591 "vector pixel vec_ldl (int, pixel *)",
2592 "vector long long vec_ldl (int, long long *)",
2593 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2594 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2595 type tree bool_long_long_type_node and correct definition of
2596 bool_V2DI_type_node to make reference to this new type tree.
2597 (rs6000_mangle_type): Replace erroneous reference to
2598 bool_long_type_node with bool_long_long_type_node.
2599 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2600 comments to emphasize sign distinctions for char and int types and
2601 replace RS6000_BTI_bool_long constant with
2602 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2603 use of RS6000_BTI_pixel.
2604 (bool_long_type_node): Remove this macro definition.
2605 (bool_long_long_type_node): New macro definition
2607 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2609 PR rtl-optimization/85300
2610 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2611 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2612 simplify_unary_operation fails.
2614 2018-04-10 Martin Liska <mliska@suse.cz>
2616 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2617 cgraph_edge and ipa_ref.
2619 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2623 * config/i386/sse.md
2624 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2625 computation of the VEC_MERGE selector from mask.
2626 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2627 Fix decoding of the VEC_MERGE selector into mask.
2629 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2631 PR tree-optimization/85286
2632 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2634 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2636 * final.c (final_1): Set insn_last_address as well as
2637 insn_current_address.
2639 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2642 * explow.c (emit_stack_probe): Call validize_mem on memory location
2643 before passing it to gen_probe_stack. Create address operand and
2644 legitimize it for the probe_stack_address case.
2646 2018-04-09 Jan Hubicka <jh@suse.cz>
2649 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2650 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2651 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2653 (free_lang_data): Rebuild type inheritance graph.
2655 2018-04-09 Martin Sebor <msebor@redhat.com>
2657 * invoke.texi (-finline-small-functions): Mention other optimization
2659 (-findirect-inlining, -fpartial-inlining): Same.
2660 (-finline-functions-called-once): Same.
2661 (-freorder-blocks-and-partition): Same.
2663 2018-04-09 Jan Hubicka <jh@suse.cz>
2666 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2667 jumps; choose last target that matches the criteria (i.e.
2668 no partition changes for non-crossing jumps).
2669 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2670 support for redirecting crossing jumps to non-crossing.
2672 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2674 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2675 also for naked functions.
2677 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2679 * config/arc/arc.md (add_shift): New pattern.
2680 (add_shift2): Likewise.
2681 (sub_shift): Likewise.
2682 (sub_shift_cmp0_noout): Likewise.
2683 (compare_si_ashiftsi): Likewise.
2684 (xbfu_cmp0_noout): New combine pattern.
2685 (xbfu_cmp0"): Likewise.
2686 (movsi_set_cc_insn): Place the predicable variant first.
2687 (commutative_binary_cmp0_noout): Remove clobber.
2688 (commutative_binary_cmp0): New pattern.
2689 (noncommutative_binary_cmp0): Likewise.
2690 (noncommutative_binary_cmp0_noout): Likewise.
2691 (noncommutative_binary_comparison_result_used): Removed.
2692 (rsub_cmp0): New pattern.
2693 (rsub_cmp0_noout): Likewise.
2694 (extzvsi): Changed, keep only meaningful variants.
2695 (SQH, SEZ): New iterators.
2696 (SQH_postfix): New mode attribute.
2697 (SEZ_prefix): New code attribute.
2698 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2699 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2700 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2702 (noncommutative_operator): Check the availability of barrel
2705 2018-04-09 Richard Biener <rguenther@suse.de>
2707 PR tree-optimization/85284
2708 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2709 Only use the niter constraining form of simple_iv when the exit
2712 2018-04-09 Tom de Vries <tom@codesourcery.com>
2715 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2716 (define_expand "*memory_barrier"): New define_expand.
2717 (define_insn "memory_barrier"): New insn.
2719 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2721 PR rtl-optimization/80463
2722 PR rtl-optimization/83972
2723 PR rtl-optimization/83480
2725 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2726 correct producer for the insn.
2727 (tidy_control_flow): Fixup seqnos in case of debug insns.
2729 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2731 PR rtl-optimization/83913
2733 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2734 different sched-times when merging exprs.
2736 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2738 PR rtl-optimization/83962
2740 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2741 tidy_fallthru_edge and tidy_control_flow.
2743 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2745 PR rtl-optimization/83530
2747 * sel-sched.c (force_next_insn): New global variable.
2748 (remove_insn_for_debug): When force_next_insn is true, also leave only
2749 next insn in the ready list.
2750 (sel_sched_region): When the region wasn't scheduled, make another pass
2751 over it with force_next_insn set to 1.
2753 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2755 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2757 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2758 for interrupt control.
2759 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2760 functions for interrupt control.
2761 * config/nds32/nds32-intrinsic.md: Likewise.
2762 * config/nds32/nds32_intrinsic.h: Likewise.
2763 * config/nds32/nds32.h (nds32_builtins): Likewise.
2765 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2767 * config/nds32/nds32.c (nds32_init_machine_status,
2768 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2769 strict_aligned_p field.
2770 (nds32_expand_to_rtl_hook): New function.
2771 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2772 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2774 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2775 Chung-Ju Wu <jasonwucj@gmail.com>
2777 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2778 * config/nds32/nds32-n7.md: New file.
2779 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2780 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2782 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2783 * config/nds32/nds32.md (pipeline_model): Add n7.
2784 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2785 * config/nds32/pipelines.md: Include n7 settings.
2787 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2788 Chung-Ju Wu <jasonwucj@gmail.com>
2790 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2791 * config/nds32/nds32-e8.md: New file.
2792 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2793 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2795 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2796 * config/nds32/nds32.md (pipeline_model): Add e8.
2797 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2798 * config/nds32/pipelines.md: Include e8 settings.
2800 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2801 Chung-Ju Wu <jasonwucj@gmail.com>
2803 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2804 * config/nds32/nds32-n8.md: New file.
2805 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2806 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2808 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2809 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2810 * config/nds32/nds32.md (pipeline_model): Add n8.
2811 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2812 * config/nds32/pipelines.md: Include n8 settings.
2814 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2815 Chung-Ju Wu <jasonwucj@gmail.com>
2817 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2818 * config/nds32/nds32-n9-2r1w.md: New file.
2819 * config/nds32/nds32-n9-3r2w.md: New file.
2820 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2821 nds32_register_ports): New or modify for cpu n9.
2822 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2824 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2825 * config/nds32/nds32-utils.c: New file.
2826 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2827 TARGET_MUL_SLOW): Define.
2828 * config/nds32/nds32.md (pipeline_model): New attribute.
2829 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2830 New options that support cpu n9.
2831 * config/nds32/pipelines.md: Include n9 settings.
2832 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2834 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2836 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2837 information if necessary.
2838 (output_cond_branch_compare_zero): Likewise.
2839 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2840 (nds32_target_alignment): Refine for alignment.
2841 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2842 (FUNCTION_BOUNDARY): Modify.
2843 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2845 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2847 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2849 * config/nds32/constants.md (unspec_volatile_element): Add values for
2850 TLB operation and data prefetch.
2851 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2852 functions for TLB operation and data prefetch.
2853 * config/nds32/nds32-intrinsic.md: Likewise.
2854 * config/nds32/nds32_intrinsic.h: Likewise.
2855 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2856 (nds32_print_operand): Likewise.
2857 * config/nds32/nds32.h (nds32_builtins): Likewise.
2859 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2860 Andrew Pinski <pinsika@gcc.gnu.org>
2863 * match.pd: Use constant_boolean_node of correct type instead of
2864 boolean_true_node or boolean_false_node for simplifying
2865 pointer comparisons to zero.
2867 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2869 PR tree-optimization/80021
2870 * tree.c (verify_type_variant): Make error call in verify_variant_match
2871 translatable and remove final full stop.
2873 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2875 * config/nds32/constants.md (unspec_volatile_element): Add
2876 UNSPEC_VOLATILE_EH_RETURN.
2877 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2878 nds32_output_stack_pop): Support dwarf exception handling process.
2879 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2880 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2881 exception handling process.
2882 (nds32_compute_stack_frame): Likewise.
2883 (nds32_return_addr_rtx): Likewise.
2884 (nds32_initial_elimination_offset): Likewise.
2885 (nds32_expand_prologue): Likewise.
2886 (nds32_expand_epilogue): Likewise.
2887 (nds32_dynamic_chain_address): New function.
2888 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2890 (DYNAMIC_CHAIN_ADDRESS): Define.
2891 (EH_RETURN_DATA_REGNO): Define.
2892 (EH_RETURN_STACKADJ_RTX): Define.
2893 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2894 patterns for dwarf exception handling.
2896 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2898 * config/nds32/nds32.h: Clean up obsolete macros.
2900 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2902 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2903 Add enum values for particular instructions.
2904 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2905 particular intrinsic functions.
2906 * config/nds32/nds32-intrinsic.md: Likewise.
2907 * config/nds32/nds32_intrinsic.h: Likewise.
2908 * config/nds32/nds32.h (nds32_builtins): Likewise.
2909 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2910 (btst, ave): New patterns for particular instructions.
2912 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2914 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2915 Add enum values for atomic load/store and memory sync.
2916 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2918 * config/nds32/nds32-intrinsic.md: Likewise.
2919 * config/nds32/nds32_intrinsic.h: Likewise.
2920 * config/nds32/nds32.h (nds32_builtins): Likewise.
2922 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2924 PR tree-optimization/85257
2925 * fold-const.c (native_encode_vector): If not all elts could fit
2926 and off is -1, return 0 rather than offset.
2927 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2928 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2929 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2930 adjust buffer in native_interpret_expr call.
2932 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2934 * config/nds32/constants.md (unspec_volatile_element): Add cache
2935 control enum values.
2936 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2937 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2938 * config/nds32/nds32.c (nds32_cctl_names): New.
2939 (nds32_print_operand): Handle cache control register names.
2940 * config/nds32/nds32.h (nds32_builtins): New enum values.
2941 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2943 * config/nds32/nds32.md (type): Add mmu.
2944 * config/nds32/pipelines.md (simple_insn): Add mmu.
2946 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2948 * config/nds32/nds32.md (type): Remove call.
2949 * config/nds32/pipelines.md (simple_insn): Likewise.
2951 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2953 * config/nds32/constants.md (unspec_volatile_element): Add
2954 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2955 UNSPEC_VOLATILE_FMFCFG.
2956 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2957 description for fmfcfg and fmfcsr.
2958 (bdesc_1arg): Add fmtcsr.
2959 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2960 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2961 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2962 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2963 unspec_fmfcfg): New patterns.
2964 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2965 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2966 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2967 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2968 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2969 __nds32__fmfcfg): Define.
2971 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2973 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2974 intrinsic register names.
2975 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2976 intrinsic register enum values and macros.
2978 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2980 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2981 for load/store addressing form.
2982 (nds32_print_operand_address): Likewise.
2984 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2987 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2988 based on LABEL_REF. Remove useless assertion.
2989 (pic_address_needs_scratch): Fix formatting.
2990 (sparc_legitimize_pic_address): Minor tweaks.
2991 (sparc_delegitimize_address): Adjust assertion accordingly.
2992 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2993 into symbolic_operand.
2994 (movsi_high_pic_label_ref): Likewise.
2995 (movsi_lo_sum_pic_label_ref): Likewise.
2996 (movdi_pic_label_ref): Likewise.
2997 (movdi_high_pic_label_ref): Likewise.
2998 (movdi_lo_sum_pic_label_ref): Likewise.
3000 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3002 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3003 custom LIB_SPEC setup.
3005 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3006 Kito Cheng <kito.cheng@gmail.com>
3008 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3009 * config/riscv/freebsd.h: New.
3011 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3013 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3014 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3017 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3018 Kito Cheng <kito.cheng@gmail.com>
3020 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3021 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3022 * config/nds32/nds32-protos.h (nds32_output_call,
3023 nds32_output_return): Declare.
3024 * config/nds32/nds32.md: Refine all the call and return patterns.
3026 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3029 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3030 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3032 PR rtl-optimization/84872
3033 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3034 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3037 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3039 * expr.c (copy_blkmode_to_reg): Revert 254862.
3040 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3042 2018-04-06 Richard Biener <rguenther@suse.de>
3045 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3046 after seeing a component reference with an adjacent field. Treat
3047 refs to arrays at struct end of external decls similar to
3048 refs to unconstrained commons.
3050 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3053 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3054 look through SAVE_EXPRs with non-side-effects argument. Adjust
3056 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3059 2018-04-06 Richard Biener <rguenther@suse.de>
3062 * alias.c (find_base_term): New wrapper around find_base_term
3063 unwinding CSELIB_VAL_PTR changes.
3064 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3067 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3069 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3071 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3072 constant definitions.
3073 ("nop"): lr 0,0 -> nopr r0
3074 ("nop_lr0", "nop_lr1"): New insn definitions.
3076 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3078 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3079 NDS32_V3PUSH_AVAILABLE_P macro.
3081 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3082 Chung-Ju Wu <jasonwucj@gmail.com>
3084 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3085 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3086 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3087 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3088 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3089 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3090 * config/nds32/constraints.md: New constraints and checking for hard
3091 float configuration.
3092 * config/nds32/iterators.md: New mode iterator and attribute for hard
3093 float configuration.
3094 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3096 * config/nds32/nds32-fpu.md: New file.
3097 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3098 deal with hard float code generation.
3099 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3101 (abi_type, float_reg_number): New enum type.
3102 * config/nds32/nds32-predicates.c: New predicates for hard float.
3103 * config/nds32/nds32-protos.h: Declare functions for hard float.
3104 * config/nds32/nds32.c: Implementation for hard float configuration.
3105 * config/nds32/nds32.h: Definitions for hard float configuration.
3106 * config/nds32/nds32.md: Include hard float machine description and
3107 modify patterns for hard float configuration.
3108 * config/nds32/nds32.opt: New options for hard float configuration.
3109 * config/nds32/predicates.md: New predicates for hard float
3112 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3114 * common/config/nds32/nds32-common.c
3115 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3117 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3120 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3121 CONSTRUCTOR_ELT (ctor, ...)->value.
3123 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3126 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3128 2018-04-05 Tom de Vries <tom@codesourcery.com>
3131 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3134 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3135 Kito Cheng <kito.cheng@gmail.com>
3137 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3138 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3139 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3140 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3142 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3143 Kito Cheng <kito.cheng@gmail.com>
3145 * config/nds32/constraints.md (Ufe): New memory constraint.
3146 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3147 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3148 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3150 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3151 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3153 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3155 * config/nds32/nds32.md: Use optimize_size in the condition for
3156 alu-shift instructions.
3158 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3160 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3162 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3164 * config/nds32/nds32.md (negsi2): Refine pattern.
3166 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3167 Chung-Ju Wu <jasonwucj@gmail.com>
3169 * config/nds32/iterators.md (shift_rotate): New code iterator.
3170 (shift): New code attribute.
3171 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3172 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3173 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3174 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3175 bit-wise operations.
3176 (andsi3, *andsi3): Ditto.
3177 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3178 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3179 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3180 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3181 nds32_ior_operand, nds32_xor_operand): New predicates.
3183 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3185 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3186 (addsi3, subsi3): ... this.
3188 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3190 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3192 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3194 * config/nds32/nds32.md: Adjust indention.
3196 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3198 * config/nds32/nds32.md (feature): New attribute.
3200 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3202 * config/nds32/nds32.md (subtype): New attribute.
3204 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3207 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3208 expansion to perform a bitwise AND of the argument followed by a
3209 boolean negation of the result.
3211 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3213 PR rtl-optimization/84878
3214 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3215 the basic block. Assert the use reference is not artificial and that
3216 it has an associated insn.
3218 2018-04-04 Michael Matz <matz@suse.de>
3220 * builtins.c (compute_objsize): Pass correct operand
3221 to array_at_struct_end_p.
3223 2018-04-04 Richard Biener <rguenther@suse.de>
3226 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3227 from contexts for DINFO_LEVEL_TERSE and below.
3229 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3231 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3232 resiter_operand condition.
3233 * config/nds32/nds32.md (*move<mode>): Ditto.
3235 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3236 Monk Chiang <sh.chiang04@gmail.com>
3238 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3240 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3242 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3244 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3245 Kito Cheng <kito.cheng@gmail.com>
3247 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3248 nds32_cond_code_str, output_cond_branch,
3249 output_cond_branch_compare_zero, nds32_expand_cbranch,
3250 nds32_expand_cstore, nds32_expand_movcc,
3251 nds32_output_cbranchsi4_equality_zero,
3252 nds32_output_cbranchsi4_equality_reg,
3253 nds32_output_cbranchsi4_equality_reg_or_const_int,
3254 nds32_output_cbranchsi4_greater_less_zero: New functions.
3255 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3256 nds32_expand_cstore, nds32_expand_movcc,
3257 nds32_output_cbranchsi4_equality_zero,
3258 nds32_output_cbranchsi4_equality_reg,
3259 nds32_output_cbranchsi4_equality_reg_or_const_int,
3260 nds32_output_cbranchsi4_greater_less_zero): Declare.
3261 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3262 nds32_rimm11s_operand): New predicates.
3263 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3264 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3267 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3269 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3270 * config/nds32/nds32.md: Ditto.
3271 * config/nds32/pipelines.md: Ditto.
3273 2018-04-04 Richard Biener <rguenther@suse.de>
3275 PR tree-optimization/85168
3276 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3277 propagating abnormals.
3279 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3281 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3283 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3284 Kito Cheng <kito.cheng@gmail.com>
3286 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3287 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3288 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3289 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3290 * config/nds32/nds32.md (sibcall_internal): New.
3291 (sibcall_register): Remove.
3292 (sibcall_immediate): Remove.
3293 (sibcall_value_internal): New.
3294 (sibcall_value_register): Remove.
3295 (sibcall_value_immediate): Remove.
3296 * config/nds32/predicates.md (nds32_general_register_operand): New.
3297 (nds32_call_address_operand): New.
3299 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3301 PR rtl-optimization/85167
3302 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3303 bb_defs if *split_p, instead preinitialize it to NULL.
3305 PR tree-optimization/85156
3306 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3307 evaluating the argument multiple times.
3309 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3311 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3313 (_mm_cvtpd_ps): Likewise.
3314 (_mm_cvttpd_epi32): Likewise.
3315 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3316 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3317 vector, pixel, and bool following altivec.h include.
3319 2018-04-03 Martin Sebor <msebor@redhat.com>
3321 * doc/extend.texi (Common Function Attributes): Clarify.
3322 (const attribute): Likewise.
3323 (pure attribute): Likewise.
3325 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3328 * config/i386/i386.c (ix86_expand_vector_set): Use
3329 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3331 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3333 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3334 instructions when changing rounding bits to preserve precision bits
3335 in the x87 control word.
3337 2018-04-03 Martin Liska <mliska@suse.cz>
3339 PR tree-optimization/82491
3340 * rtl.h (strip_offset_and_add): Replace += suboffset with
3341 poly_uint64 () + suboffset.
3343 2018-03-29 Martin Liska <mliska@suse.cz>
3344 Martin Jambor <mjambor@suse.cz>
3347 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3348 param_type is not an integral or pointer type.
3350 2018-04-03 Richard Biener <rguenther@suse.de>
3352 * sese.h (recompute_all_dominators): Remove.
3354 2018-04-02 Martin Sebor <msebor@redhat.com>
3356 * doc/invoke.texi (-Wrestrict): Fix typos.
3358 2018-04-02 Jim Wilson <jimw@sifive.com>
3360 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3361 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3362 (<optab>di3, <optab>si3_extend): Likewise.
3363 (<optab>si3_mask, <optab>si3_mask_1): New.
3364 (<optab>di3_mask, <optab>di3_mask_1): New.
3365 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3366 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3367 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3369 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3371 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3374 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3376 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3377 (nds32_canonicalize_comparison): New function.
3379 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3380 Kito Cheng <kito.cheng@gmail.com>
3381 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3383 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3384 * config/nds32/constants.md (unspec_volatile_element): Add
3385 UNSPEC_VOLATILE_RELAX_GROUP.
3386 * config/nds32/nds32-relax-opt.c: New file.
3387 * config/nds32/nds32-predicates.c
3388 (nds32_symbol_load_store_p): New function.
3389 * config/nds32/nds32-protos.h
3390 (nds32_symbol_load_store_p): Declare function.
3391 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3392 * config/nds32/nds32.c
3393 (nds32_register_pass): New function to register pass.
3394 (nds32_register_passes): New function to register passes.
3395 * config/nds32/nds32.md (relax_group): New pattern.
3396 * config/nds32/nds32.opt (mrelax-hint): New option.
3397 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3399 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3401 * config/nds32/t-nds32: Modify files dependency.
3403 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3405 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3406 (PROFILE_HOOK): Define its implementation.
3408 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3410 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3411 type and 32-bit size.
3413 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3416 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3417 (V_128_256): New mode iterator.
3418 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3419 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3420 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3422 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3423 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3425 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3428 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3429 NaN inputs correctly.
3431 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3434 * config/rs6000/vsx.md (??r): New mode attribute.
3435 (*vsx_mov<mode>_64bit): Use it.
3436 (*vsx_mov<mode>_32bit): Likewise.
3438 2018-03-30 Martin Sebor <msebor@redhat.com>
3440 PR tree-optimization/84818
3441 * builtins.c (check_access): Use warning_n.
3443 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3446 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3448 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3451 2018-03-30 Julia Koval <julia.koval@intel.com>
3454 * x86-tune.def (movx, partial_reg_dependency): Enable for
3457 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3460 * lra-constraints.c (process_alt_operands): Move setting
3461 this_alternative_matches below.
3463 2018-03-29 Martin Liska <mliska@suse.cz>
3466 * doc/invoke.texi: Document how LTO works with debug info.
3467 Describe auto-load support of binutils. Mention 'x86-64'
3468 as valid option value of -march option.
3470 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3472 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3475 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3476 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3477 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3480 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3483 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3484 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3485 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3486 for RS6000_BTM_POWERPC64.
3487 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3488 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3489 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3494 2018-03-28 Carl Love <cel@us.ibm.com>
3497 2017-09-27 Carl Love <cel@us.ibm.com>
3499 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3500 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3501 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3504 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3506 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3507 instead of __vector bool.
3508 (_mm_max_pu8): Likewise.
3509 (_mm_min_pi16): Likewise.
3511 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3514 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3515 (DIVWEUO): Likewise.
3517 (DIVDEUO): Likewise.
3518 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3519 DIVWEUO and DIVDEUO.
3520 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3521 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3522 (div_extend): Likewise.
3523 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3525 (__builtin_divweuo): Likewise.
3526 (__builtin_divdeo): Likewise.
3527 (__builtin_divdeuo): Likewise.
3529 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3532 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3533 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3535 PR tree-optimization/82004
3536 * gimple-match-head.c (optimize_pow_to_exp): New function.
3537 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3538 Don't fold to exp if optimize_pow_to_exp is false.
3540 2018-03-28 Martin Liska <mliska@suse.cz>
3543 * calls.c (initialize_argument_information): Fix trailing space.
3544 * common.opt: Fix typo and provide better explanation for
3545 -fsanitize-coverage option.
3546 * config/i386/i386.opt: Fix typo.
3548 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3549 Martin Liska <mliska@suse.cz>
3552 * gimplify.c (asan_poison_variable): Don't do the check for
3553 gimplify_omp_ctxp here.
3554 (gimplify_decl_expr): Do it here.
3555 (gimplify_target_expr): Likewise.
3557 2018-03-28 Martin Liska <mliska@suse.cz>
3560 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3561 chkp_type_bounds_count if MPX is not enabled.
3563 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3565 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3567 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3570 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3571 function to create the function decl for complex long double
3572 multiply and divide for -mabi=ieeelongdouble.
3573 (init_float128_ieee): Call it.
3575 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3578 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3579 -fcf-protection=branch -mibt.
3580 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3582 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3585 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3587 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3590 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3591 extern array declarations.
3593 2018-03-27 Richard Biener <rguenther@suse.de>
3596 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3597 explicit single_use checks.
3599 2018-03-27 Richard Biener <rguenther@suse.de>
3601 PR tree-optimization/85082
3602 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3605 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3607 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3608 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3609 Turn on fasynchronous-unwind-tables and funwind-tables.
3611 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3614 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3615 (*bmi_blsr_<mode>_ccz): Ditto.
3617 2018-03-26 Tom de Vries <tom@codesourcery.com>
3619 PR tree-optimization/85063
3620 * omp-general.c (offloading_function_p): New function. Factor out
3622 * omp-offload.c (pass_omp_target_link::gate): ... here.
3623 * omp-general.h (offloading_function_p): Declare.
3624 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3625 with attribute omp declare target for offloading functions.
3627 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3629 PR tree-optimization/84005
3630 * tree-data-ref.h (get_base_for_alignment): Declare.
3631 * tree-data-ref.c (get_base_for_alignment_1): New function.
3632 (get_base_for_alignment): Likewise.
3633 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3634 get_base_for_alignment to find a suitable base object, instead
3635 of always using drb->base_address.
3637 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3640 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3641 known size by default.
3643 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3646 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3647 and non BLKmode operands.
3649 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3652 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3653 Clean up attributes.
3655 2018-03-23 Richard Biener <rguenther@suse.de>
3658 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3659 we are going to emit early debug for LTO.
3661 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3664 * function.c (match_asm_constraints_1): Don't optimize if input
3665 doesn't satisfy general_operand predicate for output's mode.
3668 * alias.c (write_dependence_p): Don't require for x_canonicalized
3669 non-VOIDmode if x has VOIDmode.
3672 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3673 just don't try to optimize it rather than assert it never happens.
3675 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3677 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3678 macro expansions for definition of ST_INTERNAL_<mode> and
3679 LD_INTERNAL_<mode> builtins.
3680 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3682 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3684 (altivec_expand_st_builtin): Likewise.
3685 (altivec_expand_builtin): Remove calls to deleted functions.
3686 (rs6000_address_for_altivec): Delete this function.
3687 * config/rs6000/vector.md: Remove expands for
3688 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3690 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3693 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3694 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3695 re-computing once computed.
3696 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3697 (arm_init_machine_status): Initialize
3698 machine->static_chain_stack_bytes.
3700 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3703 * doc/extend.texi: Add four new prototypes for vec_ld.
3704 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3705 definitions for more logical presentation.
3706 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3707 entries for V1TI variants of __builtin_altivec_ld builtin.
3708 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3709 handling of V1TI variant of LVX icode pattern.
3710 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3711 (rs6000_gimple_fold_builtin): Likewise.
3712 (altivec_init_builtins): Add code to define
3713 __builtin_altivec_lvx_v1ti function.
3715 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3718 * function.c (match_asm_constraints_1): Don't do the optimization
3719 if input isn't a REG, SUBREG, MEM or constant.
3721 2018-03-22 Tom de Vries <tom@codesourcery.com>
3723 PR tree-optimization/84956
3724 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3725 bb_has_abnormal_pred.
3727 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3730 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3731 DECL_INITIAL (decl) to decl at the end.
3732 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3735 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3737 * doc/extend.texi (__builtin_tgmath): Document when complex
3738 integer types are treated as _Complex _Float64.
3740 2018-03-21 Tom de Vries <tom@codesourcery.com>
3742 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3744 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3746 PR tree-optimization/84960
3747 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3748 if it is ENTRY block, move them into single succ of ENTRY in that case.
3750 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3752 PR tree-optimization/84811
3753 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3754 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3755 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3756 (known_subrange_p): Update accordingly. Cast each value involved
3757 in the size comparison, rather than casting the result of the
3760 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3762 PR tree-optimization/84982
3763 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3764 by flipping the least significant bit rather than all bits from
3765 bitpos to bitpos + bitsize - 1.
3767 2018-03-21 Nathan Sidwell <nathan@acm.org>
3769 * doc/extend.texi (Deprecated Features): Remove mention of
3770 long-deleted deprecations.
3772 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3775 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3776 * configure: Regenerate.
3778 2018-03-21 Tom de Vries <tom@codesourcery.com>
3780 PR tree-optimization/83126
3781 * tree-parloops.c (num_phis): New function.
3782 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3784 2018-03-21 Nathan Sidwell <nathan@acm.org>
3786 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3787 mention anon-struct/union members and trailing attributes.
3789 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3791 PR tree-optimization/84969
3792 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3793 builtin memset partitions if they set different rhs values.
3795 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3797 PR rtl-optimization/84989
3798 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3799 VEC_DUPLICATE with scalar result mode.
3801 2018-03-21 Martin Liska <mliska@suse.cz>
3804 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3805 not intended return statement.
3807 2018-03-21 Martin Liska <mliska@suse.cz>
3810 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3811 (chkp_find_bound_slots_1): Limit number of iterations.
3813 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3816 * Minor grammar fixes for x86 options.
3818 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3821 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3822 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3824 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3827 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3828 (altivec_lvx_<mode>_1op): Likewise.
3829 (altivec_stvx_<mode>_2op): Likewise.
3830 (altivec_stvx_<mode>_1op): Likewise.
3831 (altivec_lvx_<VM2:mode>): New define_expand.
3832 (altivec_stvx_<VM2:mode>): Likewise.
3833 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3834 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3835 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3836 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3837 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3838 (rs6000_gen_lvx): Likewise.
3839 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3840 (altivec_expand_stv_builtin): Likewise.
3841 (altivec_expand_builtin): Likewise.
3842 * config/rs6000/vector.md: Likewise.
3844 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3847 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3850 2018-03-20 Richard Biener <rguenther@suse.de>
3853 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3854 sign-conversions as zero, fall back to standard scalar_stmt
3857 2018-03-20 Martin Liska <mliska@suse.cz>
3860 * predict.c (rebuild_frequencies): Handle case when we have
3861 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3863 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3866 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3867 flag_section_anchors.
3868 * varasm.c (use_blocks_for_decl_p): Remove hack for
3869 dw2_force_const_mem.
3872 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3874 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3875 be created, use lowpart_subreg of operands[0] rather than operands[0]
3877 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3878 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3879 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3880 and n constraint instead of aarch64_shift_imm_di and Usd.
3881 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3882 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3884 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3887 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3888 to favor GPR over NEON registers.
3889 (<shift>di3_neon): Likewise.
3891 2018-03-20 Tom de Vries <tom@codesourcery.com>
3894 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3895 (nvptx_process_pars): Emit bar.sync asap and alap.
3897 2018-03-20 Tom de Vries <tom@codesourcery.com>
3900 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3901 seen_label if seen_label is already set.
3903 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3906 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3907 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3908 Use 1U instead of 1. Formatting fixes.
3911 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3912 instead of TREE_TYPE (s1) for the return value.
3914 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3916 PR tree-optimization/84946
3917 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3918 bitsize + bitsize in poly_uint64 rather than poly_int64.
3921 * dwarf2asm.c: Include fold-const.c.
3922 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3923 of decl rather than decl itself.
3925 PR rtl-optimization/84643
3926 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3928 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3931 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3932 calling assemble_variable.
3934 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3937 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3938 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3940 2018-03-19 Jim Wilson <jimw@sifive.com>
3943 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3944 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3945 (riscv_first_stack_step): Likewise.
3946 (riscv_option_override): Use STACK_BOUNDARY instead of
3948 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3950 (BIGGEST_ALIGNMENT): Set to 128.
3951 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3952 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3955 2018-03-19 Richard Biener <rguenther@suse.de>
3957 PR tree-optimization/84933
3958 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3959 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3961 2018-03-19 Richard Biener <rguenther@suse.de>
3963 PR tree-optimization/84859
3964 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3965 (cond_if_else_store_replacement): Perform sinking operation on
3966 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3967 Generalize what a BB with a single eligible store is.
3969 2018-03-19 Richard Biener <rguenther@suse.de>
3971 PR tree-optimization/84929
3972 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3973 chrec_is_positive against non-chrec arg.
3975 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3978 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3980 2018-03-18 Martin Liska <mliska@suse.cz>
3982 PR rtl-optimization/84635
3983 * regrename.c (build_def_use): Use matches_mode only when
3986 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3988 PR tree-optimization/84913
3989 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3990 vectorize chains of COND_EXPRs.
3992 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3994 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3996 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3998 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4000 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4002 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4004 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4005 Kito Cheng <kito.cheng@gmail.com>
4007 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4008 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4009 (nds32_adjust_reg_alloc_order): New function.
4010 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4012 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4014 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4015 nds32_print_operand, nds32_print_operand_address): Use
4016 HOST_WIDE_INT_PRINT_DEC instead.
4018 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4020 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4022 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4025 * config/i386/i386.c (initial_ix86_tune_features,
4026 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4028 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4029 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4030 rather than 1u << ix86_tune. Formatting fix.
4031 (ix86_option_override_internal): Change ix86_arch_mask from
4032 unsigned int to unsigned HOST_WIDE_INT, initialize to
4033 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4034 (ix86_function_specific_restore): Likewise.
4036 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4039 * postreload.c (reload_combine_recognize_pattern): Perform
4040 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4041 truncate_int_for_mode the result for the destination's mode.
4044 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4045 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4047 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4050 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4053 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4054 regno_allocno_class_array and sorted_pseudos.
4055 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4056 insns where regno is used.
4058 2018-03-16 Martin Liska <mliska@suse.cz>
4061 * multiple_target.c (create_dispatcher_calls): Redirect
4062 reference in the symbol table.
4064 2018-03-16 Martin Liska <mliska@suse.cz>
4067 * multiple_target.c (create_dispatcher_calls): Redirect also
4070 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4074 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4075 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4076 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4078 2018-03-16 Julia Koval <julia.koval@intel.com>
4080 * doc/invoke.texi (Skylake Server): Add CLWB.
4081 Cannonlake): Remove CLWB.
4083 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4085 PR tree-optimization/84841
4086 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4088 (FLOAT_ONE_CONST_TYPE): Define.
4089 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4090 (sort_by_operand_rank): Put entries with higher constant_type last
4091 rather than first to match comments.
4093 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4095 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4098 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4101 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4102 check and use address of target always.
4104 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4107 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4108 (indirect_thunk_bnd_needed): Likewise.
4109 (indirect_thunks_used): Likewise.
4110 (indirect_thunks_bnd_used): Likewise.
4111 (indirect_return_needed): New.
4112 (indirect_return_bnd_needed): Likewise.
4113 (output_indirect_thunk_function): Add a bool argument for
4115 (output_indirect_thunk_function): Don't generate alias for
4116 function return thunk.
4117 (ix86_code_end): Call output_indirect_thunk_function to generate
4118 function return thunks.
4119 (ix86_output_function_return): Set indirect_return_bnd_needed
4120 and indirect_return_needed instead of indirect_thunk_bnd_needed
4121 and indirect_thunk_needed.
4123 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4125 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4126 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4127 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4129 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4130 Paul Hua <paul.hua.gm@gmail.com>
4133 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4135 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4137 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4138 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4141 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4144 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4145 instead of GET_MODE_SIZE when comparing Units.
4147 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4150 * varasm.c (hash_section): Return an unchangeble hash value
4151 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4152 Return !aarch64_can_use_per_function_literal_pools_p ().
4154 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4157 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4158 rather than address of cmode as last argument to prepare_cmp_insn.
4160 2018-03-15 Julia Koval <julia.koval@intel.com>
4162 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4163 F_AVX512VNNI, F_AVX512BITALG): New.
4165 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4168 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4169 insn for floating-point loads and stores.
4171 2018-03-14 Carl Love <cel@us.ibm.com>
4173 * config/rs6000/rs6000-c.c: Add macro definitions for
4174 ALTIVEC_BUILTIN_VEC_PERMXOR.
4175 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4176 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4177 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4178 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4180 * config/doc/extend.texi: Add prototypes for vec_permxor.
4182 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4185 * diagnostic-show-locus.c (class layout_point): Convert m_line
4186 from int to linenum_type.
4187 (line_span::comparator): Use linenum "compare" function when
4188 comparing line numbers.
4189 (test_line_span): New function.
4190 (layout_range::contains_point): Convert param "row" from int to
4192 (layout_range::intersects_line_p): Likewise.
4193 (layout::will_show_line_p): Likewise.
4194 (layout::print_source_line): Likewise.
4195 (layout::should_print_annotation_line_p): Likewise.
4196 (layout::print_annotation_line): Likewise.
4197 (layout::print_leading_fixits): Likewise.
4198 (layout::annotation_line_showed_range_p): Likewise.
4199 (struct line_corrections): Likewise for field m_row.
4200 (line_corrections::line_corrections): Likewise for param "row".
4201 (layout::print_trailing_fixits): Likewise.
4202 (layout::get_state_at_point): Likewise.
4203 (layout::get_x_bound_for_row): Likewise.
4204 (layout::print_line): Likewise.
4205 (diagnostic_show_locus): Likewise for locals "last_line" and
4207 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4208 * input.c (selftest::test_linenum_comparisons): New function.
4209 (selftest::input_c_tests): Call it.
4210 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4211 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4212 * selftest.h (ASSERT_GT): New macro.
4213 (ASSERT_GT_AT): New macro.
4214 (ASSERT_LT): New macro.
4215 (ASSERT_LT_AT): New macro.
4217 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4219 PR rtl-optimization/84780
4220 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4222 2018-03-14 Martin Liska <mliska@suse.cz>
4224 * tree.c (record_node_allocation_statistics): Use
4225 get_stats_node_kind.
4226 (get_stats_node_kind): New function extracted from
4227 record_node_allocation_statistics.
4228 (free_node): Use get_stats_node_kind.
4230 2018-03-14 Richard Biener <rguenther@suse.de>
4232 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4233 that the value-set of ANTIC_IN doesn't grow.
4236 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4238 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4239 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4241 2018-03-14 Julia Koval <julia.koval@intel.com>
4243 * config.gcc (icelake-client, icelake-server): New.
4245 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4246 (initial_ix86_arch_features): Ditto.
4247 (PTA_SKYLAKE): Add SGX.
4248 (PTA_ICELAKE): Remove.
4249 (PTA_ICELAKE_CLIENT): New.
4250 (PTA_ICELAKE_SERVER): New.
4251 (ix86_option_override_internal): Split up icelake on icelake client and
4253 (get_builtin_code_for_version): Ditto.
4254 (fold_builtin_cpu): Ditto.
4255 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4256 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4257 * config/i386/i386.h (processor_type): Ditto.
4258 * doc/invoke.texi: Ditto.
4260 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4263 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4264 INTEGER_CST offset, add it together with bitpos / 8 and
4265 sign extend based on POINTER_SIZE.
4269 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4272 * config/i386/constraints.md (Yc): New register constraint.
4273 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4274 Use Yc constraint for alternative 2 of operand 0. Remove
4275 preferred_for_speed attribute.
4277 2018-03-14 Richard Biener <rguenther@suse.de>
4279 PR tree-optimization/84830
4280 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4281 with the old one to avoid oscillations.
4283 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4286 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4288 (assign_by_spills): Return a flag of reload assignment failure.
4289 Do not process the reload assignment failures. Do not spill other
4290 reload pseudos if they has the same reg class. Update n if
4292 (lra_assign): Add a return arg. Set up from the result of
4293 assign_by_spills call.
4294 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4295 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4296 usage_insns if it is not NULL.
4297 (spill_hard_reg_in_range): New function.
4298 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4299 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4300 function prototypes.
4301 (lra_assign): Change prototype.
4302 * lra.c (lra): Add code to deal with fails by splitting hard reg
4305 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4307 * config/riscv/riscv.opt (mrelax): New option.
4308 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4309 "norelax" when riscv_mrelax is disabled.
4310 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4312 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4315 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4316 reassociation for int modes.
4318 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4320 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4321 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4323 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4324 * config/aarch64/aarch64-sve.md
4325 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4326 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4327 (*extend<mode><Vwide>2): Rename to...
4328 (aarch64_sve_extend<mode><Vwide>2): ...this.
4329 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4330 renaming the old pattern to...
4331 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4333 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4334 define_expand, renaming the old pattern to...
4335 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4336 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4337 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4338 account when deciding which SVE instruction the optab should use.
4339 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4341 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4343 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4344 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4345 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4346 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4347 (tlsdesc_small_<mode>): Turn a define_expand and use
4348 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4349 (tlsdesc_small_advsimd_<mode>): ...this.
4350 (tlsdesc_small_sve_<mode>): New pattern.
4352 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4354 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4355 (UNSPEC_UMUL_HIGHPART): New constants.
4356 (MUL_HIGHPART): New int iteraor.
4357 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4358 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4360 (*<su>mul<mode>3_highpart): New define_insn.
4362 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4365 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4368 2018-03-13 Martin Liska <mliska@suse.cz>
4371 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4373 (sem_item_optimizer::~sem_item_optimizer): Release it.
4374 (sem_item_optimizer::merge_classes): Register variable aliases.
4375 (sem_item_optimizer::fixup_pt_set): New function.
4376 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4377 * ipa-icf.h: Declare new variables and functions.
4379 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4382 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4383 integer_pow2p@2 and test integer_pow2p in condition.
4384 (A < 0 ? C : 0): Similarly for @1.
4387 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4388 characters starting at p contain '\0' character, don't look beyond
4392 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4393 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4396 * reg-stack.c (change_stack): Change update_end var from int to
4397 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4398 also call set_block_for_insn on the newly added insns and rescan.
4401 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4402 on the last operand.
4405 * tree.c (stabilize_reference_1): Return save_expr (e) for
4406 STATEMENT_LIST even if it doesn't have side-effects.
4408 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4410 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4412 2018-03-12 Renlin Li <renlin.li@arm.com>
4414 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4415 aarch64_output_scalar_simd_mov_immediate.
4417 2018-03-12 Martin Sebor <msebor@redhat.com>
4419 PR tree-optimization/83456
4420 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4421 for perfectly overlapping calls to memcpy.
4422 (gimple_fold_builtin_memory_chk): Same.
4423 (gimple_fold_builtin_strcpy): Handle no-warning.
4424 (gimple_fold_builtin_stxcpy_chk): Same.
4425 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4427 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4429 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4430 parameter. Use it for SFmode.
4431 (rs6000_function_arg_advance_1): Adjust.
4432 (rs6000_function_arg): Adjust.
4433 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4435 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4437 PR rtl-optimization/84169
4438 PR rtl-optimization/84780
4439 * combine.c (can_combine_p): Check for a 2-insn combination whether
4440 the destination register is used between the two insns, too.
4442 2018-03-12 Richard Biener <rguenther@suse.de>
4444 PR tree-optimization/84803
4445 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4446 for refs DR analysis didn't process.
4448 2018-03-12 Richard Biener <rguenther@suse.de>
4450 PR tree-optimization/84777
4451 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4452 force-vectorize loops ignore whether we are optimizing for size.
4454 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4456 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4457 (TARGET_MD_ASM_ADJUST): Define.
4459 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4460 Kito Cheng <kito.cheng@gmail.com>
4461 Chung-Ju Wu <jasonwucj@gmail.com>
4463 * config/nds32/nds32.c (nds32_compute_stack_frame,
4464 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4465 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4466 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4467 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4468 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4469 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4470 * config/nds32/nds32.md (prologue, epilogue): Use macro
4471 NDS32_V3PUSH_AVAILABLE_P to do checking.
4473 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4476 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4477 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4478 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4479 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4480 addition of most attributes on !orig_type_die or the attribute not
4481 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4483 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4484 Chung-Ju Wu <jasonwucj@gmail.com>
4486 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4488 * config/nds32/nds32.opt (mvh): New option.
4490 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4491 Chung-Ju Wu <jasonwucj@gmail.com>
4493 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4495 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4496 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4499 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4500 Chung-Ju Wu <jasonwucj@gmail.com>
4502 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4504 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4505 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4507 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4508 Kito Cheng <kito.cheng@gmail.com>
4509 Chung-Ju Wu <jasonwucj@gmail.com>
4511 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4512 UNSPEC_FFMISM and UNSPEC_FLMISM.
4513 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4514 for ffb, ffmism and flmism.
4515 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4516 (unspec_ffmism): Ditto.
4517 (unspec_flmism): Ditto.
4518 (nds32_expand_builtin_impl): Check if string extension is available.
4519 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4520 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4522 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4525 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4528 * lra-assigns.c (assign_by_spills): Return a flag of reload
4529 assignment failure. Do not process the reload assignment
4530 failures. Do not spill other reload pseudos if they has the same
4532 (lra_assign): Add a return arg. Set up from the result of
4533 assign_by_spills call.
4534 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4535 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4536 usage_insns if it is not NULL.
4537 (spill_hard_reg_in_range): New function.
4538 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4539 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4540 function prototypes.
4541 (lra_assign): Change prototype.
4542 * lra.c (lra): Add code to deal with fails by splitting hard reg
4545 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4548 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4550 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4553 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4554 (dw_val_node): Add val_symbolic_view.
4555 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4556 (symview_upper_bound): New.
4557 (new_line_info_table): Initialize symviews_since_reset.
4558 (dwarf2out_source_line): Count symviews_since_reset and set
4559 symview_upper_bound.
4560 (dw_val_equal_p): Handle symview.
4561 (add_AT_symview): New.
4562 (print_dw_val): Handle symview.
4563 (attr_checksum, attr_checksum_ordered): Likewise.
4564 (same_dw_val_p, size_of_die): Likewise.
4565 (value_format, output_die): Likewise.
4566 (add_high_low_attributes): Use add_AT_symview for entry_view.
4567 (dwarf2out_finish): Reset symview_upper_bound, clear
4570 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4573 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4574 Add strict argument and use it.
4575 (rs6000_split_multireg_move): Update for new strict argument.
4576 (mem_operand_gpr): Disallow all non-offsettable addresses.
4577 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4579 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4582 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4583 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4584 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4587 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4588 decl, use remap_type if we want to use the type.
4590 2018-03-09 Martin Sebor <msebor@redhat.com>
4592 PR tree-optimization/84526
4593 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4595 (builtin_access::generic_overlap): Be prepared to handle non-array
4598 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4600 PR rtl-optimization/84682
4601 * lra-constraints.c (process_address_1): Check is_address flag
4602 for address constraints.
4603 (process_alt_operands): Likewise.
4604 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4605 preprocess_constraints.
4606 * recog.h (preprocess_constraints): Add oploc parameter.
4608 * recog.c (preprocess_constraints): Test address_operand for
4609 CT_ADDRESS constraints.
4611 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4614 * lra-assigns.c (assign_by_spills): Return a flag of reload
4615 assignment failure. Do not process the reload assignment
4616 failures. Do not spill other reload pseudos if they has the same
4618 (lra_assign): Add a return arg. Set up from the result of
4619 assign_by_spills call.
4620 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4621 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4622 usage_insns if it is not NULL.
4623 (spill_hard_reg_in_range): New function.
4624 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4625 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4626 function prototypes.
4627 (lra_assign): Change prototype.
4628 * lra.c (lra): Add code to deal with fails by splitting hard reg
4631 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4634 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4635 Accept complain bool parameter. Only emit errors if it is true.
4636 (arm_parse_cpu_option_name): Likewise.
4637 (arm_target_thumb_only): Adjust callers of the above.
4638 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4639 prototype to take a default true bool parameter.
4640 (arm_parse_arch_option_name): Likewise.
4642 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4643 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4647 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4648 * configure: Regenerate.
4649 * configure.ac ("linker --version-script option"): New.
4650 ("linker soname option"): New.
4652 2018-03-09 Richard Biener <rguenther@suse.de>
4654 PR tree-optimization/84775
4655 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4656 immediate uses of predicate stmts and mark them modified.
4659 PR tree-optimization/84178
4660 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4662 (version_loop_for_if_conversion): Delay update_ssa call.
4663 (tree_if_conversion): Delay update_ssa until after predicate
4666 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4669 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4670 when the function accesses prior frames.
4672 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4675 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4676 gen_llsym, otherwise call maybe_gen_llsym.
4679 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4680 has ',' character inside of it.
4682 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4685 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4686 as clobbering CC_REGNUM.
4688 2018-03-08 Richard Biener <rguenther@suse.de>
4691 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4692 (follow_copies_to_constant): Do not follow SSA names registered
4695 2018-03-08 Richard Biener <rguenther@suse.de>
4697 PR tree-optimization/84178
4698 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4700 (version_loop_for_if_conversion): Delay update_ssa call.
4701 (tree_if_conversion): Delay update_ssa until after predicate
4704 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4706 PR tree-optimization/84178
4707 * tree-if-conv.c (release_bb_predicate): Remove the
4708 the assertion that the stmts have NULL use_ops.
4709 Discard the statements, asserting that they haven't
4710 yet been added to a BB.
4712 2018-03-08 Richard Biener <rguenther@suse.de>
4714 PR tree-optimization/84746
4715 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4716 (phi_translate): Pass in destination ANTIC_OUT set.
4717 (phi_translate_1): Likewise. For a simplified result lookup
4718 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4719 (phi_translate_set): Adjust.
4720 (do_pre_regular_insertion): Likewise.
4721 (do_pre_partial_partial_insertion): Likewise.
4723 2018-03-08 Martin Liska <mliska@suse.cz>
4725 PR gcov-profile/84735
4726 * doc/gcov.texi: Document usage of profile files.
4727 * gcov-io.h: Document changes in the format.
4729 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4733 * dwarf2out.c (struct dw_line_info_table): Update comments for
4735 (FORCE_RESET_NEXT_VIEW): New.
4736 (FORCE_RESETTING_VIEW_P): New.
4737 (RESETTING_VIEW_P): Check for -1 too.
4738 (ZERO_VIEW_P): Likewise.
4739 (new_line_info_table): Force-reset next view.
4740 (dwarf2out_begin_function): Likewise.
4741 (dwarf2out_source_line): Simplify zero_view_p initialization.
4742 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4743 view directly. Omit view when omitting .loc at line 0.
4745 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4747 PR tree-optimization/84740
4748 * tree-switch-conversion.c (process_switch): Call build_constructors
4749 only if info.phi_count is non-zero.
4751 PR tree-optimization/84739
4752 * tree-tailcall.c (find_tail_calls): Check call arguments against
4753 DECL_ARGUMENTS (current_function_decl) rather than
4754 DECL_ARGUMENTS (func) when checking for tail recursion.
4756 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4758 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4759 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4760 Volker Reichelt's entry and add entries for people that perform
4761 GCC fuzzy testing and report numerous bugs.
4763 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4766 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4767 readonly data in sdata, if that is disabled.
4768 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4769 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4770 -mreadonly-in-sdata option.
4772 2018-03-07 Martin Sebor <msebor@redhat.com>
4774 PR tree-optimization/84468
4775 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4776 basic block when looking for nul assignment.
4778 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4781 * except.h (output_function_exception_table): Adjust prototype.
4782 * except.c (output_function_exception_table): Remove FNNAME parameter
4783 and add SECTION parameter. Ouput one part of the table at a time.
4784 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4785 the first part of the exception table and emit unwind directives.
4786 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4787 (i386_pe_seh_cold_init): Likewise.
4788 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4789 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4790 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4791 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4792 * config/i386/winnt.c: Include except.h.
4793 (struct seh_frame_state): Add reg_offset, after_prologue and
4794 in_cold_section fields.
4795 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4796 (i386_pe_seh_cold_init): New function.
4797 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4798 to seh->in_cold_section.
4799 (seh_emit_push): Record the offset of the push.
4800 (seh_emit_save): Record the offet of the save.
4801 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4802 Test seh->after_prologue to disregard the epilogue.
4803 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4804 (i386_pe_end_cold_function): New function.
4806 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4809 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4810 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4813 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4815 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4816 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4819 * multiple_target.c: Include tree-inline.h and intl.h.
4820 (expand_target_clones): Diagnose and fail if node->definition and
4821 !tree_versionable_function_p (node->decl).
4823 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4825 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4827 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4828 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4829 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4831 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4834 * combine.c (try_combine): Use reg_or_subregno instead of handling
4835 just paradoxical SUBREGs and REGs.
4837 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4839 * config/arc/arc.c (arc_finalize_pic): Remove function.
4840 (arc_must_save_register): We use single base PIC register, remove
4841 checks to save/restore the PIC register.
4842 (arc_expand_prologue): Likewise.
4843 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4845 (arc_verify_short): Likewise.
4846 (arc_attr_type): Likewise.
4847 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4848 (walk_stores): Likewise.
4849 (arc_address_cost): Make it static.
4850 (arc_verify_short): Likewise.
4851 (branch_dest): Likewise.
4852 (arc_attr_type): Likewise.
4853 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4854 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4855 (arc_final_prescan_insn): Remove inserting the nops due to
4856 hardware hazards. It is done in reorg step.
4857 (insn_length_variant_t): Remove.
4858 (insn_length_parameters_t): Likewise.
4859 (arc_insn_length_parameters): Likewise.
4860 (arc_get_insn_variants): Likewise.
4861 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4863 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4866 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4869 PR tree-optimization/84687
4870 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4872 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4874 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4876 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4877 Rename to ppc_speculation_barrier.
4878 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4879 __builtin_ppc_speculation_barrier.
4881 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4884 * combine.c (combine_simplify_rtx): Don't try to simplify if
4885 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4888 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4890 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4891 to 32 bytes when compiling for POWER9.
4893 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4896 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4897 regparm >= 3 with no arg reg available also for calls with
4898 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4901 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4903 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4905 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4908 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4910 2018-03-05 Richard Biener <rguenther@suse.de>
4912 PR tree-optimization/84486
4913 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4914 When inserting a __builtin_assume_aligned call set the LHS
4915 SSA name alignment info accordingly.
4917 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4919 PR tree-optimization/84114
4920 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4921 Avoid reassociation of FLOAT_MODE addition.
4923 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4925 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4926 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4927 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4928 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4929 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4930 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4931 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4933 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4934 __builtin_ia32_wbinvd): New builtins.
4935 (SPECIAL_ARGS2): New.
4936 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4937 (SPECIAL_ARGS2): New.
4938 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4939 (ix86_valid_target_attribute_inner_p): Ditto.
4940 (ix86_init_mmx_sse_builtins): Add special_args2.
4941 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4942 TARGET_WBNOINVD_P): New.
4943 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4944 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4945 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4946 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4947 * config/i386/pconfigintrin.h: New file.
4948 * config/i386/wbnoinvdintrin.h: Ditto.
4949 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4950 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4952 2018-03-05 Richard Biener <rguenther@suse.de>
4954 PR tree-optimization/84670
4955 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4957 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4958 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4959 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4960 doesn't grow if all successors (recursively) were visited at least
4963 2018-03-05 Richard Biener <rguenther@suse.de>
4965 PR tree-optimization/84650
4966 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4967 if executed in the loop pipeline.
4969 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4971 * doc/configfiles.texi (Configuration Files): Move info about
4972 conditionalizing $target-protos.h to...
4973 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4974 differs from $target-protos.h.
4976 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4977 Chung-Ju Wu <jasonwucj@gmail.com>
4979 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4980 * config/nds32/nds32-multiple.md (setmemsi): Define.
4981 * config/nds32/nds32-memory-manipulation.c
4982 (nds32_gen_dup_4_byte_to_word_value): New.
4983 (emit_setmem_word_loop): New.
4984 (emit_setmem_byte_loop): New.
4985 (nds32_expand_setmem_loop): New.
4986 (nds32_expand_setmem_loop_v3m): New.
4987 (nds32_expand_setmem_unroll): New.
4988 (nds32_expand_setmem): New.
4990 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4991 Chung-Ju Wu <jasonwucj@gmail.com>
4993 * config/nds32/nds32-memory-manipulation.c
4994 (nds32_emit_load_store): New.
4995 (nds32_emit_post_inc_load_store): New.
4996 (nds32_emit_mem_move): New.
4997 (nds32_emit_mem_move_block): New.
4998 (nds32_expand_movmemsi_loop_unknown_size): New.
4999 (nds32_expand_movmemsi_loop_known_size): New.
5000 (nds32_expand_movmemsi_loop): New.
5001 (nds32_expand_movmemsi_unroll): New.
5002 (nds32_expand_movmemqi): Rename ...
5003 (nds32_expand_movmemsi): ... to this.
5004 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5005 (movmemsi): ... to this.
5006 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5007 (nds32_expand_movmemsi): ... to this.
5009 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5010 Monk Chiang <sh.chiang04@gmail.com>
5011 Chung-Ju Wu <jasonwucj@gmail.com>
5013 * config/nds32/nds32-protos.h
5014 (nds32_expand_load_multiple): New arguments.
5015 (nds32_expand_store_multiple): Ditto.
5016 (nds32_valid_multiple_load_store): Rename ...
5017 (nds32_valid_multiple_load_store_p): ... to this.
5018 * config/nds32/nds32-memory-manipulation.c
5019 (nds32_expand_load_multiple): Refine implementation.
5020 (nds32_expand_store_multiple): Ditto.
5021 * config/nds32/nds32-multiple.md
5022 (load_multiple): Update nds32_expand_load_multiple interface.
5023 (store_multiple): Update nds32_expand_store_multiple interface.
5024 * config/nds32/nds32-predicates.c
5025 (nds32_valid_multiple_load_store): Rename ...
5026 (nds32_valid_multiple_load_store_p): ... to this and refine
5028 * config/nds32/predicates.md
5029 (nds32_load_multiple_and_update_address_operation): New predicate.
5030 (nds32_store_multiple_and_update_address_operation): New predicate.
5032 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5033 Chung-Ju Wu <jasonwucj@gmail.com>
5035 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5036 (combo): New attribute.
5037 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5039 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5041 * config/nds32/nds32.opt: Change -mcmodel= default value.
5043 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5044 Monk Chiang <sh.chiang04@gmail.com>
5045 Chung-Ju Wu <jasonwucj@gmail.com>
5047 * config/nds32/constants.md (unspec_element): New enum.
5048 * config/nds32/constraints.md (Umw): New constraint.
5049 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5050 * config/nds32/nds32-intrinsic.md: Likewise.
5051 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5052 (nds32_valid_smw_lwm_base_p): New.
5053 (nds32_output_smw_single_word): New.
5054 (nds32_output_lmw_single_word): New.
5055 (nds32_expand_unaligned_load): New.
5056 (nds32_expand_unaligned_store): New.
5057 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5058 (nds32_output_smw_single_word): Declare.
5059 (nds32_output_lmw_single_word): Declare.
5060 (nds32_expand_unaligned_load): Declare.
5061 (nds32_expand_unaligned_store): Declare.
5062 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5063 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5064 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5065 NDS32_BUILTIN_UASTORE_DW.
5066 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5069 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5070 Kito Cheng <kito.cheng@gmail.com>
5071 Chung-Ju Wu <jasonwucj@gmail.com>
5073 * config/nds32/nds32-intrinsic.c
5074 (nds32_expand_builtin_null_ftype_reg): Delete.
5075 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5076 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5077 (nds32_read_argument): New.
5078 (nds32_legitimize_target): Ditto.
5079 (nds32_legitimize_argument): Ditto.
5080 (nds32_check_constant_argument): Ditto.
5081 (nds32_expand_unop_builtin): Ditto.
5082 (nds32_expand_unopimm_builtin): Ditto.
5083 (nds32_expand_binop_builtin): Ditto.
5084 (nds32_builtin_decl_impl): Ditto.
5085 (builtin_description): Ditto.
5086 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5087 (nds32_init_builtins_impl): Ditto.
5088 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5089 (nds32_builtin_decl): New.
5090 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5091 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5093 2018-03-02 Jeff Law <law@redhat.com>
5095 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5096 (redundant_insn, fill_simple_delay_slots): Likewise.
5097 (fill_slots_from_thread): Likewise.
5098 * resource.c (mark_referenced_resources): Likewise.
5099 (mark_set_resources, find_dead_or_set_registers): Likewise.
5101 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5103 * substring-locations.h (format_warning_va): Formatting fix for
5105 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5107 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5108 * substring-locations.c: Include intl.h.
5109 (format_warning_va): Turned into small wrapper around
5110 format_warning_n_va, renamed to ...
5111 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5112 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5114 (format_warning_at_substring_n): New function.
5115 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5116 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5117 format_warning_at_substring with just a shorter name instead of
5118 const function pointer.
5119 (fmtwarn_n): New function.
5120 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5121 appropriate, get rid of all the fmtstr temporaries, move conditionals
5122 with G_() wrapped string literals directly into fmtwarn arguments,
5123 cast dir.len to (int), formatting fixes.
5125 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5127 * doc/invoke.texi: Remove "Cilk Plus" references.
5129 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5130 Richard Biener <rguenther@suse.de>
5133 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5134 for error or warning attributes if CALL_FROM_THUNK_P is set.
5137 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5140 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5141 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5144 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5145 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5147 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5148 instead of -1U in last predictors element's probability member.
5150 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5153 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5154 arguments if they are comparable.
5156 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5158 PR tree-optimization/84634
5159 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5160 masks and masked_loop_p with a single loop_masks, making sure it's
5161 null for bb vectorization.
5163 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5165 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5166 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5167 loop->force_vectorize to check whether there is no alias.
5169 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5172 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5174 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5176 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5177 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5178 instead of a loop around prev_real_insn.
5179 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5183 * config/i386/i386.c (ix86_print_operand): Use conditional
5184 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5187 2018-03-02 Richard Biener <rguenther@suse.de>
5189 PR tree-optimization/84427
5190 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5191 (bitmap_set_subtract_values): Rewrite to handle multiple
5194 (prune_clobbered_mems): Likewise.
5195 (phi_translate): Take edge instead of pred/phiblock.
5196 (phi_translate_1): Likewise.
5197 (phi_translate_set): Likewise. Insert all translated
5198 exprs for a value into the set, keeping possibly multiple
5199 expressions per value.
5200 (compute_antic_aux): Adjust for phi_translate changes.
5201 When intersecting union the expressions and prune those
5202 not in the final value set, keeping possibly multiple
5203 expressions per value. Do not use value-insertion
5204 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5205 all expressions. Add verification that the value-sets
5206 only shrink during iteration.
5207 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5208 (do_pre_regular_insertion): Likewise.
5209 (do_pre_partial_partial_insertion): Likewise.
5211 2018-03-02 Richard Biener <rguenther@suse.de>
5214 * config/darwin.c (saved_debug_info_level): New static global.
5215 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5216 (darwin_asm_lto_end): Restore debug info generation settings.
5218 2018-03-01 Martin Liska <mliska@suse.cz>
5221 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5224 2018-03-01 Richard Biener <rguenther@suse.de>
5227 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5228 type annotation with LTO when debug was disabled at compile-time.
5230 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5232 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5234 (mips_final_postscan_insn): Likewise.
5236 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5238 PR rtl-optimization/84528
5239 * alias.c (init_alias_target): Add commentary.
5240 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5241 a unique base value if the frame pointer is not eliminated
5242 to the stack pointer.
5244 2018-03-01 Tom de Vries <tom@codesourcery.com>
5246 PR rtl-optimization/83327
5247 * lra-int.h (hard_regs_spilled_into): Declare.
5248 * lra.c (hard_regs_spilled_into): Define.
5249 (init_reg_info): Init hard_regs_spilled_into.
5250 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5251 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5252 (process_bb_lives): Handle hard_regs_spilled_into.
5253 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5254 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5256 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5258 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5259 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5260 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5261 * config/rs6000/aix72.h: New file.
5263 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5265 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5266 instead of warning_at with conditional singular and plural messages
5270 * stor-layout.c (update_alignment_for_field): For
5271 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5272 && !DECL_PACKED (field), do the alignment update, just use
5273 only desired_align instead of MAX (type_align, desired_align)
5275 (place_field): Don't do known_align < desired_align handling
5276 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5277 is non-NULL, instead do it after rli->prev_field handling and
5278 only if not within a bitfield word. For DECL_PACKED (field)
5279 use type_align of BITS_PER_UNIT.
5281 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5283 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5284 superfluous parentheses and trailing spaces.
5286 2018-02-28 Richard Biener <rguenther@suse.de>
5288 PR tree-optimization/84584
5289 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5290 SCoPs with fake exit edge.
5292 2018-02-28 Martin Liska <mliska@suse.cz>
5295 * timevar.c (timer::print): Fix format to properly print 100%
5298 2018-02-28 Richard Biener <rguenther@suse.de>
5301 * genmatch.c (capture_info::walk_match): Do not mark
5302 captured expressions without operands as expr_p given
5303 they act more like predicates and should be subject to
5304 "lost tail" side-effect preserving.
5306 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5308 PR rtl-optimization/81611
5309 * auto-inc-dec.c (attempt_change): Move dead note from
5310 mem_insn if it's the next use of regno
5311 (find_address): Take address use of reg holding
5312 non-incremented value. Add parm to limit search to the named
5314 (merge_in_block): Attempt to use a mem insn that is the next
5315 use of the original regno.
5317 2018-02-27 Martin Sebor <msebor@redhat.com>
5320 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5321 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5323 2018-02-27 Martin Sebor <msebor@redhat.com>
5325 PR translation/84207
5326 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5327 n argument to unsigned HOST_WIDE_INT.
5328 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5329 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5330 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5331 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5333 2018-02-27 Richard Biener <rguenther@suse.de>
5335 PR tree-optimization/84512
5336 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5337 Do not use the estimate returned from record_stmt_cost for
5338 the scalar iteration cost but sum properly using add_stmt_cost.
5340 2018-02-27 Richard Biener <rguenther@suse.de>
5342 PR tree-optimization/84466
5343 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5344 Adjust last change to less strictly validate use operands.
5346 2018-02-27 Martin Liska <mliska@suse.cz>
5348 PR gcov-profile/84548
5349 * gcov.c (process_file): Allow partial overlap and consider it
5350 also as group functions.
5351 (output_lines): Properly calculate range of lines for a group.
5353 2018-02-27 Martin Liska <mliska@suse.cz>
5355 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5356 'ggc' suffixes. Change first column width.
5357 (timer::print): Fix formatting of the column.
5359 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5361 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5362 preserve inline entry blocks for the sake of debug inline
5363 entry point markers alone.
5364 (remove_unused_locals): Suggest in comments a better place to
5365 force the preservation of inline entry blocks that are
5366 otherwise unused, but do not preserve them.
5368 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5370 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5372 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5375 * config/i386/constraints.md (Bs): Replace
5376 ix86_indirect_branch_register with
5377 TARGET_INDIRECT_BRANCH_REGISTER.
5379 * config/i386/i386.md (indirect_jump): Likewise.
5380 (tablejump): Likewise.
5381 (*sibcall_memory): Likewise.
5382 (*sibcall_value_memory): Likewise.
5383 Peepholes of indirect call and jump via memory: Likewise.
5384 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5385 (*sibcall_value_GOT_32): Likewise.
5386 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5387 (GOT_memory_operand): Likewise.
5388 (call_insn_operand): Likewise.
5389 (sibcall_insn_operand): Likewise.
5390 (GOT32_symbol_operand): Likewise.
5391 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5393 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5395 PR rtl-optimization/83496
5396 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5397 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5398 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5399 redundant insn, if any.
5400 (relax_delay_slots): Likewise.
5401 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5403 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5405 PR tree-optimization/83965
5406 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5407 that grouped statements are part of a reduction chain. Return
5408 true if the statement is not marked as a reduction itself but
5410 (vect_recog_dot_prod_pattern): Don't check whether the statement
5411 is part of a group here.
5412 (vect_recog_sad_pattern): Likewise.
5413 (vect_recog_widen_sum_pattern): Likewise.
5415 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5418 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5420 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5423 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5425 (ix86_output_indirect_function_return): New prototype.
5426 (ix86_split_simple_return_pop_internal): Likewise.
5427 * config/i386/i386.c (indirect_return_via_cx): New.
5428 (indirect_return_via_cx_bnd): Likewise.
5429 (indirect_thunk_name): Handle return va CX_REG.
5430 (output_indirect_thunk_function): Create alias for
5431 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5432 (ix86_output_indirect_jmp): Remove the bool argument.
5433 (ix86_output_indirect_function_return): New function.
5434 (ix86_split_simple_return_pop_internal): Likewise.
5435 * config/i386/i386.md (*indirect_jump): Don't pass false
5436 to ix86_output_indirect_jmp.
5437 (*tablejump_1): Likewise.
5438 (simple_return_pop_internal): Change it to define_insn_and_split.
5439 Call ix86_split_simple_return_pop_internal to split it for
5441 (simple_return_indirect_internal): Call
5442 ix86_output_indirect_function_return instead of
5443 ix86_output_indirect_jmp.
5445 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5448 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5449 memset and value initialization afterwards.
5451 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5453 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5455 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5458 * common/config/aarch64/aarch64-common.c
5459 (aarch_option_optimization_table[]): Switch
5460 off fomit-frame-pointer
5462 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5463 Chung-Ju Wu <jasonwucj@gmail.com>
5465 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5467 (store_multiple): Ditto.
5469 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5471 * config.gcc: Add --with-cpu support for nds32 target.
5472 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5473 * config/nds32/nds32.opt: Add -mcpu= option.
5475 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5477 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5478 isel=yes): Warn for these deprecated options.
5480 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5482 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5483 ISA_2_5_MASKS_EMBEDDED.
5485 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5487 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5488 p->max as pointers rather than using iterative_hash_expr.
5490 2018-02-23 Carl Love <cel@us.ibm.com>
5492 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5493 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5495 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5496 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5497 P8V_BUILTIN_VEC_VUNSIGNED2.
5499 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5502 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5503 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5505 * lra-constraints.c (curr_insn_transform): Set up
5506 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5508 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5509 * lra-eliminations.c (spill_pseudos): Ditto.
5510 (process_insn_for_elimination): Ditto.
5511 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5512 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5513 LRA_NON_CLOBBERED_ALT.
5515 2018-02-22 Martin Sebor <msebor@redhat.com>
5517 PR tree-optimization/84480
5518 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5519 to maybe_diag_stxncpy_trunc. Call it.
5520 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5521 from gimple_fold_builtin_strcpy. Print inlining stack.
5522 (handle_builtin_stxncpy): Print inlining stack.
5523 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5525 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5528 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5529 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5530 and -fcheck-pointer-bounds are used together.
5531 (indirect_thunk_prefix): New enum.
5532 (indirect_thunk_need_prefix): New function.
5533 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5534 "_nt" instead of "_bnd" for NOTRACK prefix.
5535 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5536 (output_indirect_thunk_function): Likewise.
5538 (ix86_code_end): Update output_indirect_thunk_function calls.
5539 (ix86_output_indirect_branch_via_reg): Replace
5540 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5541 (ix86_output_indirect_branch_via_push): Likewise.
5542 (ix86_output_function_return): Likewise.
5543 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5544 incompatible with -fcf-protection=branch and
5545 -fcheck-pointer-bounds.
5547 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5550 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5551 Change gcc_assert call to output_operand_lossage.
5553 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5555 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5557 2018-02-22 DJ Delorie <dj@redhat.com>
5558 Sebastian Perta <sebastian.perta@renesas.com>
5559 Oleg Endo <olegendo@gcc.gnu.org>
5561 * config/rx/rx.c (rx_rtx_costs): New function.
5562 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5564 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5566 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5568 2018-02-22 Martin Liska <mliska@suse.cz>
5571 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5572 Add "native" as a possible value.
5574 2018-02-22 Martin Liska <mliska@suse.cz>
5577 * config/i386/i386.c (ix86_option_override_internal):
5578 Add "native" as a possible value for -march and -mtune.
5580 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5583 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5584 to all type variants.
5586 PR tree-optimization/84503
5587 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5588 width as info->bitpos + info->bitsize - start.
5589 (merged_store_group::merge_overlapping): Simplify width computation.
5590 (check_no_overlap): New function.
5591 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5592 start + width and last_order of the group, fail if check_no_overlap
5594 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5595 to group if check_no_overlap fails.
5597 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5599 * config/rs6000/altivec.md: Delete contraint arguments to
5600 define_expand, define_split, and define_peephole2, and in
5601 define_insn_and_split if always unused.
5602 * config/rs6000/darwin.md: Ditto.
5603 * config/rs6000/dfp.md: Ditto.
5604 * config/rs6000/rs6000.md: Ditto.
5605 * config/rs6000/sync.md: Ditto.
5606 * config/rs6000/vector.md: Ditto.
5607 * config/rs6000/vsx.md: Ditto.
5609 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5611 * config/rs6000/altivec.md: Write output control strings as braced
5612 blocks instead of double-quoted strings.
5613 * config/rs6000/darwin.md: Ditto.
5614 * config/rs6000/rs6000.md: Ditto.
5615 * config/rs6000/vector.md: Ditto.
5616 * config/rs6000/vsx.md: Ditto.
5618 2018-02-21 Jason Merrill <jason@redhat.com>
5620 PR c++/84314 - ICE with templates and fastcall attribute.
5621 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5623 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5625 * ipa-cp.c (determine_versionability): Fix comment typos.
5627 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5630 * ipa-cp.c (determine_versionability): Do not version functions caling
5633 2018-02-21 Martin Liska <mliska@suse.cz>
5636 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5637 Add "native" as a possible value.
5638 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5639 the macro when native cpu detection is available.
5641 2018-02-21 Martin Liska <mliska@suse.cz>
5644 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5645 Add "native" as a possible value.
5646 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5647 when native cpu detection is available.
5649 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5650 Martin Sebor <msebor@redhat.com>
5652 PR tree-optimization/84478
5653 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5655 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5656 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5657 support which is conservatively correct, for 2 only stay conservative
5658 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5659 argument to the 2 argument get_range_strlen, adjust 6 arg
5660 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5662 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5663 (gimple_fold_builtin_strlen): Pass true as last argument to
5666 2018-02-20 Martin Sebor <msebor@redhat.com>
5669 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5670 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5671 (builtin_memref::builtin_memref): Factor out parts into
5672 set_base_and_offset and call it.
5674 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5677 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5678 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5679 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5680 search at the associated MODE_INT.
5682 2018-02-20 Jeff Law <law@redhat.com>
5685 PR tree-optimization/81592
5687 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5688 for range data rather than using global data.
5689 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5690 range data rather than using global data.
5691 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5692 pass it to children as needed.
5693 (struct directive::fmtresult): Similarly.
5694 (struct directive::set_width): Similarly.
5695 (struct directive::set_precision): Similarly.
5696 (format_integer, format_directive, parse_directive): Similarly.
5697 (format_none): Accept unnamed vr_values parameter.
5698 (format_percent, format_floating, format_character): Similarly.
5699 (format_string, format_plain): Similarly.
5700 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5701 the EVRP range analyzer for range data rather than using global data.
5702 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5703 gimple-ssa-evrp-analyze.h
5704 (class sprintf_dom_walker): Add after_dom_children member function.
5705 Add evrp_range_analyzer member.
5706 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5707 range analyzer as needed.
5708 (sprintf_dom_walker::after_dom_children): New member function.
5709 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5711 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5712 (evrp_range_analyzer::pop_to_marker): Likewise.
5714 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5716 PR tree-optimization/84419
5717 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5718 with the required type if its current type is compatible but
5721 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5724 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5725 after vectorization.
5727 2018-02-20 Martin Liska <mliska@suse.cz>
5730 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5731 possible values if we don't have a hint.
5733 2018-02-20 Martin Liska <mliska@suse.cz>
5737 * final.c (shorten_branches): Build align_tab array with one
5739 * opts.c (finish_options): Add alignment option limit check.
5740 (MAX_CODE_ALIGN): Likewise.
5741 (MAX_CODE_ALIGN_VALUE): Likewise.
5742 * doc/invoke.texi: Document maximum allowed option value for
5743 all -falign-* options.
5745 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5748 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5749 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5750 * var-tracking.c (emit_note_insn_var_location): Remove all references
5751 to NOTE_INSN_CALL_ARG_LOCATION.
5752 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5753 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5754 Use copy_rtx_if_shared.
5755 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5756 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5757 (dwarf2out_var_location): Remove handling of
5758 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5760 * final.c (final_scan_insn): Remove all references to
5761 NOTE_INSN_CALL_ARG_LOCATION.
5762 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5763 before dumping final insns.
5764 * except.c (emit_note_eh_region_end): Remove all references to
5765 NOTE_INSN_CALL_ARG_LOCATION.
5766 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5767 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5768 * config/arc/arc.c (hwloop_optimize): Likewise.
5769 * config/arm/arm.c (create_fix_barrier): Likewise.
5770 * config/s390/s390.c (s390_chunkify_start): Likewise.
5771 * config/sh/sh.c (find_barrier): Likewise.
5772 * config/i386/i386.c (rest_of_insert_endbranch,
5773 ix86_seh_fixup_eh_fallthru): Likewise.
5774 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5775 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5776 * config/frv/frv.c (frv_function_prologue): Likewise.
5777 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5779 (note_outside_basic_block_p): Remove all references to
5780 NOTE_INSN_CALL_ARG_LOCATION.
5781 * gengtype.c (adjust_field_rtx_def): Likewise.
5782 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5784 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5785 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5788 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5791 PR tree-optimization/84452
5792 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5793 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5796 2018-02-19 Martin Liska <mliska@suse.cz>
5799 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5801 2018-02-19 Martin Liska <mliska@suse.cz>
5802 Richard Sandiford <richard.sandiford@linaro.org>
5804 PR tree-optimization/82491
5805 * gimple-fold.c (get_base_constructor): Make earlier bail out
5808 2018-02-19 Carl Love <cel@us.ibm.com>
5810 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5811 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5813 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5814 P8V_BUILTIN_VEC_NEG.
5816 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5818 * config/rl78/rl78.md (movdf): New define expand.
5820 2018-02-19 Martin Liska <mliska@suse.cz>
5823 * doc/invoke.texi: Fix typo.
5824 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5826 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5828 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5829 handle rs6000_single_float and rs6000_double_float specially for
5832 2018-02-16 Jeff Law <law@redhat.com>
5834 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5837 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5840 * ipa-inline.c (inline_small_functions): Fix a typo.
5842 2018-02-16 Nathan Sidwell <nathan@acm.org>
5844 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5846 2018-02-16 Carl Love <cel@us.ibm.com>
5848 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5849 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5850 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5851 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5852 expansion to P8V_BUILTIN_VEC_FLOAT2.
5854 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5856 PR rtl-optimization/70023
5857 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5858 src_regno into account.
5860 2018-02-16 Carl Love <cel@us.ibm.com>
5862 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5863 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5864 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5865 * config/rs6000/rs6000.c: Remove case statements for
5866 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5867 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5868 and P9V_BUILTIN_VEC_VINSERT4B.
5869 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5870 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5871 * config/rs6000/vsx.md:
5872 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5875 2018-02-16 Carl Love <cel@us.ibm.com>
5877 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5879 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5881 * config/rs6000/rs6000-c.c: Add the definitions for
5882 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5883 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5884 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5885 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5886 definition for insert4b and define insn *insert3b_internal.
5887 * doc/extend.texi: Add documentation for vec_extract4b.
5889 2018-02-16 Nathan Sidwell <nathan@acm.org>
5891 * doc/extend.texi (Backwards Compatibility): Mention friend
5892 injection. Note for-scope is deprecated.
5893 * doc/invoke.texi (-ffriend-injection): Deprecate.
5895 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5897 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5898 that moved to I2, also allow destinations that are a paradoxical
5899 subreg (instead of a normal reg).
5901 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5904 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5907 2018-02-16 Richard Biener <rguenther@suse.de>
5909 PR tree-optimization/84037
5910 PR tree-optimization/84016
5912 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5913 Adjust vec_construct for the fact we need additional higher latency
5914 128bit inserts for AVX256 and AVX512 vector builds.
5915 (ix86_add_stmt_cost): Scale vector construction cost for
5918 2018-02-16 Richard Biener <rguenther@suse.de>
5920 PR tree-optimization/84417
5921 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5922 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5923 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5925 2018-02-16 Martin Liska <mliska@suse.cz>
5928 * internal-fn.def (ASAN_CHECK): Set proper flags.
5929 (ASAN_MARK): Likewise.
5931 2018-02-16 Julia Koval <julia.koval@intel.com>
5933 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5934 from PTA_CANNONLAKE.
5936 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5939 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5940 Use ++iter rather than iter++ for std::list iterators.
5941 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5942 defer deleting them until all nodes in the forest are processed. Do
5943 free even leaf nodes. Change to_process into auto_vec.
5946 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5947 * vec.h (vec_default_construct): Use memset instead of placement new
5948 if BROKEN_VALUE_INITIALIZATION is defined.
5949 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5950 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5953 PR rtl-optimization/83723
5954 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5955 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5956 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5958 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5960 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5962 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5964 PR rtl-optimization/81443
5965 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5966 from inner REGs to paradoxical SUBREGs.
5968 2018-02-16 Richard Biener <rguenther@suse.de>
5970 PR tree-optimization/84399
5971 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5972 For operands we can analyze at their definition make sure we can
5973 analyze them at each use as well.
5975 2018-02-16 Richard Biener <rguenther@suse.de>
5977 PR tree-optimization/84190
5978 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5979 volatile accesses if the decl isn't volatile.
5981 2018-02-15 Jason Merrill <jason@redhat.com>
5983 PR c++/84314 - ICE with templates and fastcall attribute.
5984 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5985 TYPE_CANONICAL on an existing type.
5987 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5989 PR tree-optimization/84383
5990 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5991 dstoff nor call operand_equal_p if dstbase is NULL.
5993 PR tree-optimization/84334
5994 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5995 also a CONSTANT_CLASS_P, punt.
5997 2018-02-14 Jim Wilson <jimw@sifive.com>
5999 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6000 first SMALL_OPERAND check. New local min_second_step. Move assert
6001 to where locals are set. Add TARGET_RVC support.
6002 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6004 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6006 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6008 2018-02-14 Martin Sebor <msebor@redhat.com>
6010 PR tree-optimization/83698
6011 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6012 arrays constrain the offset range to their bounds.
6013 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6014 (builtin_access::overlap): Avoid setting the size of overlap if it's
6016 (maybe_diag_overlap): Also consider arrays when deciding what values
6017 of offsets to include in diagnostics.
6019 2018-02-14 Martin Sebor <msebor@redhat.com>
6022 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6023 that correspond to the kind of a declaration.
6025 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6028 * config/pa/pa.md: Load address of PIC label using the linkage table
6029 if the label is nonlocal.
6031 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6033 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6034 warning message if user requests -maltivec=be.
6035 * doc/invoke.texi: Document deprecation of -maltivec=be.
6037 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6040 * config/rs6000/rs6000-c.c: Update definitions for
6041 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6042 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6044 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6047 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6048 add _get_ssp intrinsics. Remove argument from
6049 __builtin_ia32_rdssp[d|q].
6050 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6051 * config/i386/i386-builtin.def: Remove argument from
6052 __builtin_ia32_rdssp[d|q].
6053 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6054 ix86_expand_special_args_builtin for _rdssp[d|q].
6055 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6056 Clear register before usage.
6057 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6058 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6060 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6062 PR tree-optimization/84357
6063 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6064 operand 1 of an ARRAY_REF too.
6066 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6069 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6070 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6072 (set_of_reg): New struct.
6073 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6074 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6075 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6077 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6078 Split into bitclr, bitset, bitinvert patterns if appropriate.
6079 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6080 use rx_fuse_in_memory_bitop.
6081 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6082 to named insn, correct maximum insn length.
6084 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6087 * machmode.def: Define a complex mode for PARTIAL_INT.
6088 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6090 * doc/rtl.texi: Document CSPImode.
6091 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6093 (msp430_hard_regno_nregs_with_padding): Likewise.
6095 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6098 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6100 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6102 PR rtl-optimization/84169
6103 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6104 we generated a parallel as new i3 and we split that to new i2 and i3
6105 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6106 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6107 those to i2, not i1. Partially rewrite this scan code.
6109 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6112 * stor-layout.c (place_field): For variable length fields, adjust
6113 offset_align afterwards not just based on the field's alignment,
6114 but also on the size.
6117 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6118 of exps and logs in the use_exp2 case.
6120 2018-02-13 Jeff Law <law@redhat.com>
6122 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6125 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6128 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6132 * common.opt (gas-loc-support, gas-locview-support): New.
6133 (ginline-points, ginternal-reset-location-views): New.
6134 * doc/invoke.texi: Document them. Use @itemx where intended.
6135 (gvariable-location-views): Adjust.
6136 * target.def (reset_location_view): New.
6137 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6138 (TARGET_RESET_LOCATION_VIEW): New.
6139 * doc/tm.texi: Rebuilt.
6140 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6141 (dwarf2out_default_as_locview_support): New.
6142 (output_asm_line_debug_info): Use option variables.
6143 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6144 (output_loc_list): Likewise.
6145 (add_high_low_attributes): Check option variables.
6146 Don't output entry view attribute in strict mode.
6147 (gen_inlined_subroutine_die): Check option variables.
6148 (dwarf2out_inline_entry): Likewise.
6149 (init_sections_and_labels): Likewise.
6150 (dwarf2out_early_finish): Likewise.
6151 (maybe_reset_location_view): New, from...
6152 (dwarf2out_var_location): ... here. Call it.
6153 * debug.h (dwarf2out_default_as_loc_support): Declare.
6154 (dwarf2out_default_as_locview_support): Declare.
6155 * hooks.c (hook_int_rtx_insn_0): New.
6156 * hooks.h (hook_int_rtx_insn_0): Declare.
6157 * toplev.c (process_options): Take -gas-loc-support and
6158 -gas-locview-support from dwarf2out. Enable
6159 -gvariable-location-views by default only with locview
6160 assembler support. Enable -ginternal-reset-location-views by
6161 default only if the target defines the corresponding hook.
6162 Enable -ginline-points by default if location views are
6163 enabled; force it disabled if statement frontiers are
6165 * tree-inline.c (expand_call_inline): Check option variables.
6166 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6168 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6170 PR tree-optimization/84321
6171 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6172 handling. Also check whether the anti-range contains any values
6173 that satisfy the mask; switch to a VR_RANGE if not.
6175 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6178 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6180 2018-02-13 Martin Jambor <mjambor@suse.cz>
6183 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6184 of call statements, also set location of a load to a temporary.
6186 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6188 * config/rl78/rl78.c (add_vector_labels): New function.
6189 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6190 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6191 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6192 which checks that no arguments are passed.
6193 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6194 * doc/extend.texi: Documentation for the new attribute.
6196 2018-02-13 Andreas Schwab <schwab@suse.de>
6198 * config/riscv/linux.h (CPP_SPEC): Define.
6200 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6203 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6204 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6205 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6206 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6207 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6208 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6209 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6210 temporarily for AES and PCLMUL builtins.
6212 PR tree-optimization/84339
6213 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6214 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6218 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6219 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6220 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6222 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6224 * omp-simd-clone.h: New file.
6225 * omp-simd-clone.c: Include omp-simd-clone.h.
6226 (expand_simd_clones): No longer static.
6227 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6228 cgraph.h and omp-simd-clone.h.
6229 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6230 (vect_recog_widen_shift_pattern): Formatting fix.
6231 (vect_pattern_recog_1): Don't check optab for calls.
6234 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6235 operands[2] into a REG before using gen_lowpart on it.
6237 2018-02-12 Jeff Law <law@redhat.com>
6240 * config/sh/sh.c (find_barrier): Consider a sibling call
6243 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6244 successfully back substituting a reg.
6246 2018-02-12 Richard Biener <rguenther@suse.de>
6248 PR tree-optimization/84037
6249 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6250 parameter, move visited init to caller.
6251 (vect_slp_analyze_operations): Separate cost from validity
6252 check, initialize visited once for all instances.
6253 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6255 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6257 (vect_model_promotion_demotion_cost): Likewise.
6258 (vectorizable_bswap): Guard cost modeling with !slp_node
6259 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6261 (vectorizable_call): Likewise.
6262 (vectorizable_conversion): Likewise.
6263 (vectorizable_assignment): Likewise.
6264 (vectorizable_shift): Likewise.
6265 (vectorizable_operation): Likewise.
6266 (vectorizable_store): Likewise.
6267 (vectorizable_load): Likewise.
6268 (vectorizable_condition): Likewise.
6269 (vectorizable_comparison): Likewise.
6271 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6274 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6275 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6276 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6278 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6281 * params.def (inline-min-speedup): Increase from 8 to 15.
6282 (max-inline-insns-auto): Decrease from 40 to 30.
6283 * ipa-split.c (consider_split): Add some buffer for function to
6284 be considered inlining candidate.
6285 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6288 2018-02-12 Richard Biener <rguenther@suse.de>
6290 PR tree-optimization/84037
6291 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6292 matched stmts if we cannot swap the non-matched ones.
6294 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6296 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6297 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6298 _mm_maskz_scalef_round_ss): New intrinsics.
6299 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6300 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6301 __builtin_ia32_scalefss_round): Remove.
6302 (__builtin_ia32_scalefsd_mask_round,
6303 __builtin_ia32_scalefss_mask_round): New intrinsics.
6304 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6305 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6306 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6307 "<round_constraint>")): Changed to ...
6308 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6309 "<round_scalar_constraint>")): ... this.
6310 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6311 %0, %1, %2<round_op3>}"): Changed to ...
6312 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6313 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6314 %2<round_scalar_mask_op3>}"): ... this.
6315 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6317 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6319 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6320 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6321 (_mm_maskz_sqrt_round_ss): New intrinsics.
6322 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6323 (__builtin_ia32_sqrtsd_mask_round)
6324 (__builtin_ia32_sqrtss_mask_round): New builtins.
6325 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6326 (__builtin_ia32_sqrtss_round): Remove.
6327 (__builtin_ia32_sqrtsd_mask_round)
6328 (__builtin_ia32_sqrtss_mask_round): New builtins.
6329 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6330 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6331 ((match_operand:VF_128 1 "vector_operand"
6332 "xBm,<round_constraint>")): Changed to ...
6333 ((match_operand:VF_128 1 "vector_operand"
6334 "xBm,<round_scalar_constraint>")): ... this.
6335 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6336 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6337 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6338 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6339 %<iptr>1<round_scalar_mask_op3>}): ... this.
6340 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6341 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6343 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6346 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6347 Cast vec_cmpeq result to correct type.
6348 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6349 Cast vec_cmpgt result to correct type.
6351 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6353 * final.c (final_scan_insn_1): Renamed from...
6354 (final_scan_insn): ... this. New wrapper, to recover
6355 seen from the outermost call in recursive ones.
6356 * config/sparc/sparc.c (output_return): Drop seen from call.
6357 (output_sibcall): Likewise.
6358 * config/visium/visium.c (output_branch): Likewise.
6360 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6362 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6365 2018-02-10 Alan Modra <amodra@gmail.com>
6368 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6369 Specify LR as an input.
6371 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6374 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6375 remove_member_access_dummy_vars): New functions.
6376 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6377 lower_omp_1, execute_lower_omp): Use them.
6379 PR rtl-optimization/84308
6380 * shrink-wrap.c (spread_components): Release todo vector.
6382 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6384 PR rtl-optimization/57193
6385 * ira-color.c (struct allocno_color_data): Add member
6386 conflict_allocno_hard_prefs.
6387 (update_conflict_allocno_hard_prefs): New.
6388 (bucket_allocno_compare_func): Add a preference based on
6389 conflict_allocno_hard_prefs.
6390 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6391 (color_allocnos): Remove a dead code. Initiate
6392 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6394 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6397 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6398 constraint from =wa to wa. Avoid a subreg on the output operand,
6399 instead use a pseudo and subreg it in a move.
6400 (p9_xxbrd_<mode>): Changed to ...
6401 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6402 (p9_xxbrd_v2df): New expander.
6403 (p9_xxbrw_<mode>): Changed to ...
6404 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6405 (p9_xxbrw_v4sf): New expander.
6407 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6409 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6411 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6414 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6415 multiply in 32-bit mode.
6416 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6417 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6420 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6422 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6423 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6424 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6425 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6427 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6430 * dwarf2out.c (is_trivial_indirect_ref): New function.
6431 (dwarf2out_late_global_decl): Do not generate a location
6432 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6433 and that are not defined in the current unit.
6435 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6437 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6438 instead of a libcall for UNORDERED.
6440 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6443 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6444 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6446 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6449 * config/s390/s390.c (s390_set_current_function): Invoke
6450 s390_indirect_branch_settings also if fndecl didn't change.
6452 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6454 * config/rs6000/rs6000.md (blockage): Set length to zero.
6456 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6458 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6460 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6463 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6464 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6468 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6469 PARALLEL incoming that failed vt_get_decl_and_offset check.
6472 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6473 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6475 (get_variable_section): For decls in named .bss* sections pass true as
6476 second argument to bss_initializer_p.
6478 2018-02-09 Marek Polacek <polacek@redhat.com>
6479 Jakub Jelinek <jakub@redhat.com>
6482 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6483 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6484 Sync some changes from cxx_fold_indirect_ref.
6486 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6488 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6490 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6491 (BLOCK_INLINE_ENTRY_LABEL): New.
6492 (dwarf2out_var_location): Disregard inline entry markers.
6493 (inline_entry_data): New struct.
6494 (inline_entry_data_hasher): New hashtable type.
6495 (inline_entry_data_hasher::hash): New.
6496 (inline_entry_data_hasher::equal): New.
6497 (inline_entry_data_table): New variable.
6498 (add_high_low_attributes): Add DW_AT_entry_pc and
6499 DW_AT_GNU_entry_view attributes if a pending entry is found
6500 in inline_entry_data_table. Add old entry_pc attribute only
6501 if debug nonbinding markers are disabled.
6502 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6503 markers are enabled.
6504 (block_within_block_p, dwarf2out_inline_entry): New.
6505 (dwarf2out_finish): Check that no entries remained in
6506 inline_entry_data_table.
6507 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6508 (final_scan_insn, notice_source_line): Likewise.
6509 (rest_of_clean_state): Skip inline entry markers.
6510 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6512 * gimple.c (gimple_build_debug_inline_entry): New.
6513 * gimple.h (enum gimple_debug_subcode): Add
6514 GIMPLE_DEBUG_INLINE_ENTRY.
6515 (gimple_build_debug_inline_entry): Declare.
6516 (gimple_debug_inline_entry_p): New.
6517 (gimple_debug_nonbind_marker_p): Adjust.
6518 * insn-notes.def (INLINE_ENTRY): New.
6519 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6520 inline entry marker notes.
6521 (print_insn): Likewise.
6522 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6523 (INSN_DEBUG_MARKER_KIND): Likewise.
6524 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6525 * tree-inline.c (expand_call_inline): Build and insert
6526 debug_inline_entry stmt.
6527 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6528 inline entry blocks early, if nonbind markers are enabled.
6529 (dump_scope_block): Dump fragment info.
6530 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6531 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6532 (gimple_build_debug_inline_entry): New.
6533 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6534 Enable/disable inline entry points too.
6535 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6536 (DEBUG_INSN): Describe inline entry markers.
6538 * common.opt (gvariable-location-views): New.
6539 (gvariable-location-views=incompat5): New.
6540 * config.in: Rebuilt.
6541 * configure: Rebuilt.
6542 * configure.ac: Test assembler for view support.
6543 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6544 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6545 * dwarf2out.c (var_loc_view): New typedef.
6546 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6547 (dwarf2out_locviews_in_attribute): New.
6548 (dwarf2out_locviews_in_loclist): New.
6549 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6550 (enum dw_line_info_opcode): Add LI_adv_address.
6551 (struct dw_line_info_table): Add view.
6552 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6553 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6554 (zero_view_p): New variable.
6555 (ZERO_VIEW_P): New macro.
6556 (output_asm_line_debug_info): New.
6557 (struct var_loc_node): Add view.
6558 (add_AT_view_list, AT_loc_list): New.
6559 (add_var_loc_to_decl): Add view param. Test it against last.
6560 (new_loc_list): Add view params. Record them.
6561 (AT_loc_list_ptr): Handle loc and view lists.
6562 (view_list_to_loc_list_val_node): New.
6563 (print_dw_val): Handle dw_val_class_view_list.
6564 (size_of_die): Likewise.
6565 (value_format): Likewise.
6566 (loc_list_has_views): New.
6567 (gen_llsym): Set vl_symbol too.
6568 (maybe_gen_llsym, skip_loc_list_entry): New.
6569 (dwarf2out_maybe_output_loclist_view_pair): New.
6570 (output_loc_list): Output view list or entries too.
6571 (output_view_list_offset): New.
6572 (output_die): Handle dw_val_class_view_list.
6573 (output_dwarf_version): New.
6574 (output_compilation_unit_header): Use it.
6575 (output_skeleton_debug_sections): Likewise.
6576 (output_rnglists, output_line_info): Likewise.
6577 (output_pubnames, output_aranges): Update version comments.
6578 (output_one_line_info_table): Output view numbers in asm comments.
6579 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6580 Call maybe_gen_llsym.
6581 (loc_list_from_tree_1): Adjust.
6582 (add_AT_location_description): Create view list attribute if
6583 needed, check it's absent otherwise.
6584 (convert_cfa_to_fb_loc_list): Adjust.
6585 (maybe_emit_file): Call output_asm_line_debug_info for test.
6586 (dwarf2out_var_location): Reset views as needed. Precompute
6587 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6588 attribute. Set view.
6589 (new_line_info_table): Reset next view.
6590 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6591 (dwarf2out_source_line): Likewise. Output view resets and labels to
6592 the assembler, or select appropriate line info opcodes.
6593 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6594 (optimize_string_length): Catch it. Adjust.
6595 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6596 dw_val_class_view_list, and remove it if no longer needed.
6597 (hash_loc_list): Hash view numbers.
6598 (loc_list_hasher::equal): Compare them.
6599 (optimize_location_lists): Check whether a view list symbol is
6600 needed, and whether the locview attribute is present, and
6601 whether they match. Remove the locview attribute if no longer
6603 (index_location_lists): Call skip_loc_list_entry for test.
6604 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6605 Use output_dwarf_version.
6606 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6607 (struct dw_val_node): Add val_view_list.
6608 * final.c (SEEN_NEXT_VIEW): New.
6609 (set_next_view_needed): New.
6610 (clear_next_view_needed): New.
6611 (maybe_output_next_view): New.
6612 (final_start_function): Rename to...
6613 (final_start_function_1): ... this. Take pointer to FIRST,
6614 add SEEN parameter. Emit param bindings in the initial view.
6615 (final_start_function): Reintroduce SEEN-less interface.
6616 (final): Rename to...
6617 (final_1): ... this. Take SEEN parameter. Output final pending
6618 next view at the end.
6619 (final): Reintroduce seen-less interface.
6620 (final_scan_insn): Output pending next view before switching
6621 sections or ending a block. Mark the next view as needed when
6622 outputting variable locations. Notify debug backend of section
6623 changes, and of location view changes.
6624 (rest_of_handle_final): Adjust.
6625 * toplev.c (process_options): Autodetect value for debug variable
6626 location views option. Warn on incompat5 without -gdwarf-5.
6627 * doc/invoke.texi (gvariable-location-views): New.
6628 (gvariable-location-views=incompat5): New.
6629 (gno-variable-location-views): New.
6631 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6633 PR tree-optimization/84136
6634 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6635 that the result of find_edge is non-NULL.
6637 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6640 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6641 storing integer register in SImode. Fix cost of 256 and 512
6642 byte aligned SSE register store.
6644 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6646 * config/i386/i386.c (ix86_multiplication_cost): Fix
6647 multiplication cost for TARGET_AVX512DQ.
6649 2018-02-08 Marek Polacek <polacek@redhat.com>
6651 PR tree-optimization/84238
6652 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6655 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6657 PR tree-optimization/84265
6658 * tree-vect-stmts.c (vectorizable_store): Don't treat
6659 VMAT_CONTIGUOUS accesses as grouped.
6660 (vectorizable_load): Likewise.
6662 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6664 PR tree-optimization/81635
6665 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6666 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6667 (test_round_for_mask): New functions.
6668 (wide_int_cc_tests): Call test_round_for_mask.
6669 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6670 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6671 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6672 range returned by get_range_info.
6674 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6677 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6678 * symtab.c: Include builtins.h
6679 (symtab_node::output_to_lto_symbol_table_p): Move here
6680 from lto-streamer-out.c:output_symbol_p.
6681 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6682 (output_symbol_p): Move all logic to symtab.c
6683 (produce_symtab): Update.
6685 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6687 * config/s390/s390-opts.h (enum indirect_branch): Define.
6688 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6689 (s390_indirect_branch_via_thunk)
6690 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6691 (enum s390_indirect_branch_type): Define.
6692 * config/s390/s390.c (struct s390_frame_layout, struct
6693 machine_function): Remove.
6694 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6695 (indirect_branch_table_label_no, indirect_branch_table_name):
6697 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6698 (enum s390_indirect_branch_option): Define.
6699 (s390_return_addr_from_memory): New function.
6700 (s390_handle_string_attribute): New function.
6701 (s390_attribute_table): Add new attribute handler.
6702 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6703 (s390_indirect_branch_via_thunk): New function.
6704 (s390_indirect_branch_via_inline_thunk): New function.
6705 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6706 sibling call optimization for non z10 compiles.
6707 (s390_emit_call): Force indirect branch target to be a single
6708 register. Add r1 clobber for non-z10 compiles.
6709 (s390_emit_epilogue): Emit return jump via return_use expander.
6710 (s390_reorg): Handle JUMP_INSNs as execute targets.
6711 (s390_option_override_internal): Perform validity checks for the
6712 new command line options.
6713 (s390_indirect_branch_attrvalue): New function.
6714 (s390_indirect_branch_settings): New function.
6715 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6716 (s390_output_indirect_thunk_function): New function.
6717 (s390_code_end): Implement target hook.
6718 (s390_case_values_threshold): Implement target hook.
6719 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6721 * config/s390/s390.h (struct s390_frame_layout)
6722 (struct machine_function): Move here from s390.c.
6723 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6724 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6725 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6726 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6727 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6728 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6729 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6730 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6731 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6732 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6733 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6734 (mnemonic attribute): Add values which aren't recognized
6736 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6737 pattern for branch conversion. Fix mnemonic attribute.
6738 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6739 indirect branch via thunk if requested.
6740 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6741 ("*indirect_jump"): Disable for branch conversion using out of
6743 ("indirect_jump_via_thunk<mode>_z10")
6744 ("indirect_jump_via_thunk<mode>")
6745 ("indirect_jump_via_inlinethunk<mode>_z10")
6746 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6747 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6748 ("casesi_jump_via_inlinethunk<mode>_z10")
6749 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6750 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6751 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6752 ("*indirect2_jump"): Disable for branch conversion.
6753 ("casesi_jump"): Turn into expander and expand patterns for branch
6755 ("return_use"): New expander.
6756 ("*return"): Emit return via thunk and rename it to ...
6757 ("*return<mode>"): ... this one.
6758 * config/s390/s390.opt: Add new options and and enum for the
6761 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6763 * lra-constraints.c (match_reload): Unconditionally use
6764 gen_lowpart_SUBREG, rather than selecting between that
6765 and equivalent gen_rtx_SUBREG code.
6767 2018-02-08 Richard Biener <rguenther@suse.de>
6769 PR tree-optimization/84233
6770 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6771 changed flag instead of boguously re-using phi_inserted.
6773 2018-02-08 Martin Jambor <mjambor@suse.cz>
6775 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6776 static local variables.
6778 2018-02-08 Richard Biener <rguenther@suse.de>
6780 PR tree-optimization/84278
6781 * tree-vect-stmts.c (vectorizable_store): When looking for
6782 smaller vector types to perform grouped strided loads/stores
6783 make sure the mode is supported by the target.
6784 (vectorizable_load): Likewise.
6786 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6788 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6789 Increase LDP/STP opportunities by adding adjacent callee-saves.
6791 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6793 PR rtl-optimization/84068
6794 PR rtl-optimization/83459
6795 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6797 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6799 PR tree-optimization/84224
6800 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6801 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6804 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6807 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6808 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6809 count, remove one USE.
6811 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6813 * doc/install.texi (Configuration): Document the
6814 --with-long-double-format={ibm,ieee} PowerPC configuration
6818 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6819 Convert from define_expand to be define_insn_and_split. Rework
6820 float/double/_Float128 conversions to QI/HI/SImode to work with
6821 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6822 conversions to QI/HImode types did a store and then a load to
6823 truncate the value. For conversions to VSX registers, don't split
6824 the insn, instead emit the code directly. Use the code iterator
6825 any_fix to combine signed and unsigned conversions.
6826 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6827 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6828 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6829 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6830 (fix_<mode>di2_hw): Likewise.
6831 (fixuns_<mode>di2_hw): Likewise.
6832 (fix_<mode>si2_hw): Likewise.
6833 (fixuns_<mode>si2_hw): Likewise.
6834 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6835 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6836 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6837 fix<uns>_trunc<SFDF:mode>si2_p8.
6838 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6840 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6841 (fix<uns>_<mode>_mem): Likewise.
6842 (fctiw<u>z_<mode>_mem): Likewise.
6843 (fix<uns>_<mode>_mem): Likewise.
6844 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6845 the register allocator from doing a direct move to the GPRs to do
6846 a store, and instead use the ISA 3.0 store byte/half-word from
6847 vector register instruction. For IEEE 128-bit floating point,
6848 also optimize stores of 32-bit ints.
6849 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6851 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6853 * genextract.c (push_pathstr_operand): New function to support
6855 (walk_rtx): Call push_pathstr_operand.
6856 (print_path): Support [a-zA-Z].
6858 2018-02-07 Richard Biener <rguenther@suse.de>
6860 PR tree-optimization/84037
6861 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6862 (cse_and_gimplify_to_preheader): Declare.
6863 (vect_get_place_in_interleaving_chain): Likewise.
6864 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6866 (_loop_vec_info::~_loop_vec_info): Delete it.
6867 (cse_and_gimplify_to_preheader): New function.
6868 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6869 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6870 (vectorizable_load): Likewise. For grouped stores always base
6871 the IV on the first element.
6872 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6873 condition before gimplifying.
6875 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6877 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6878 *DIV_EXPR and *MOD_EXPR.
6880 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6883 * config/i386/i386.c (ix86_option_override_internal): Mask out
6884 the CF_SET bit when checking -fcf-protection.
6886 2018-02-07 Tom de Vries <tom@codesourcery.com>
6889 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6892 2018-02-07 Richard Biener <rguenther@suse.de>
6894 PR tree-optimization/84204
6895 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6898 PR tree-optimization/84205
6899 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6900 special-case isl_ast_op_zdiv_r.
6902 PR tree-optimization/84223
6903 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6904 Only add conditions from within the region.
6905 (gather_bbs::after_dom_children): Adjust.
6907 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6910 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6911 * config/avr/avr.md: Only post-reload split REG-REG moves if
6912 either register is GENERAL_REG_P.
6914 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6916 PR tree-optimization/84235
6917 * tree-ssa-scopedtables.c
6918 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6919 if the subtraction is performed in floating point type where NaNs are
6920 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6921 build 1. Formatting fix.
6923 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6926 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6927 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6928 and skip it regardless of bb boundaries. Use CALL_P macro,
6929 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6930 unnecessarily, formatting fix.
6932 2018-02-06 Michael Collison <michael.collison@arm.com>
6934 * config/arm/thumb2.md:
6935 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6936 (*thumb_mov_notscc): Ditto.
6938 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6941 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6942 unsigned_fix, not "s".
6944 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6946 * configure.ac (gcc_fn_eh_frame_ro): New function.
6947 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6948 correct .eh_frame permissions.
6949 * configure: Regenerate.
6951 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6953 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6956 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6958 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6959 Display warning message for -mno-speculate-indirect-jumps.
6961 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6963 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6965 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6967 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6969 PR tree-optimization/84225
6970 * tree-eh.c (find_trapping_overflow): Only call
6971 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6973 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6976 * config/i386/i386.c: Reimplement the check of possible options
6977 -mibt/-mshstk conbination. Change error messages.
6978 * doc/invoke.texi: Fix a typo: remove extra '='.
6980 2018-02-06 Marek Polacek <polacek@redhat.com>
6982 PR tree-optimization/84228
6983 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6985 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6988 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6989 emitted arch directives.
6990 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6991 __ARM_FEATURE_COPROC before changing architectures.
6993 2018-02-06 Richard Biener <rguenther@suse.de>
6995 * config/i386/i386.c (print_reg): Fix typo.
6996 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6998 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7000 * configure: Regenerate.
7002 2018-02-05 Martin Sebor <msebor@redhat.com>
7004 PR tree-optimization/83369
7005 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7008 2018-02-05 Martin Liska <mliska@suse.cz>
7010 * doc/invoke.texi: Cherry-pick upstream r323995.
7012 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7014 * ira.c (ira_init_register_move_cost): Adjust comment.
7016 2018-02-05 Martin Liska <mliska@suse.cz>
7018 PR gcov-profile/84137
7019 * doc/gcov.texi: Fix typo in documentation.
7021 2018-02-05 Martin Liska <mliska@suse.cz>
7023 PR gcov-profile/83879
7024 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7025 using dlopen functionality.
7027 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7029 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7030 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7031 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7032 _mm_maskz_range_round_ss): New intrinsics.
7033 (__builtin_ia32_rangesd128_round)
7034 (__builtin_ia32_rangess128_round): Remove.
7035 (__builtin_ia32_rangesd128_mask_round,
7036 __builtin_ia32_rangess128_mask_round): New builtins.
7037 * config/i386/i386-builtin.def (__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/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7042 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7043 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7044 "<round_saeonly_constraint>")): Changed to ...
7045 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7046 "<round_saeonly_scalar_constraint>")): ... this.
7047 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7048 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7049 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7050 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7051 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7053 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7055 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7057 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7058 Remove all values except native, 8540 and 8548.
7060 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7062 * config/i386/i386.c (ix86_output_function_return): Pass
7063 INVALID_REGNUM, instead of -1, as invalid register number to
7064 indirect_thunk_name and output_indirect_thunk.
7066 2018-02-02 Julia Koval <julia.koval@intel.com>
7068 * config.gcc: Add -march=icelake.
7069 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7070 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7071 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7072 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7073 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7074 (processor_target_table): Add icelake.
7075 (ix86_option_override_internal): Handle new PTAs.
7076 (get_builtin_code_for_version): Handle icelake.
7077 (M_INTEL_COREI7_ICELAKE): New.
7078 (fold_builtin_cpu): Handle icelake.
7079 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7080 * doc/invoke.texi: Add -march=icelake.
7082 2018-02-02 Julia Koval <julia.koval@intel.com>
7084 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7085 to wide_int_bitmask.
7086 * wide-int-bitmask.h: New.
7088 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7091 * config/i386/i386.md: Replace Pmode with word_mode in
7092 builtin_setjmp_setup and builtin_longjmp to support x32.
7094 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7098 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7100 (rs6000_supported_cpu_names): New static variable.
7101 (linux_cpu_translation_table): Likewise.
7102 (elf_platform) <cpu>: Define new static variable and use it.
7103 Translate kernel AT_PLATFORM name to canonical name if needed.
7104 Error if platform name is unknown.
7106 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7109 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7111 2018-02-01 Jeff Law <law@redhat.com>
7114 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7115 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7116 the scratch if RELEASE_VIA_POP is false.
7117 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7118 If we have to save a temporary register, decrement SIZE appropriately.
7119 Pass new arguments to release_scratch_register_on_entry.
7120 (ix86_adjust_stack_and_probe): Likewise.
7121 (ix86_emit_probe_stack_range): Pass new arguments to
7122 release_scratch_register_on_entry.
7124 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7126 PR rtl-optimization/84157
7127 * combine.c (change_zero_ext): Use REG_P predicate in
7128 front of HARD_REGISTER_P predicate.
7130 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7132 * config/avr/avr.c (avr_option_override): Move disabling of
7133 -fdelete-null-pointer-checks to...
7134 * common/config/avr/avr-common.c (avr_option_optimization_table):
7137 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7139 PR tree-optimization/81635
7140 * tree-data-ref.c (split_constant_offset_1): For types that
7141 wrap on overflow, try to use range info to prove that wrapping
7144 2018-02-01 Renlin Li <renlin.li@arm.com>
7147 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7149 (aarch64_register_move_cost): Likewise.
7150 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7152 (REG_CLASS_NAMES): Likewise.
7153 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7154 TAILCALL_ADDR_REGS. Remove IP registers.
7155 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7157 2018-02-01 Richard Biener <rguenther@suse.de>
7159 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7160 for specifying RPO order and allow NULL for that.
7161 * domwalk.c (dom_walker::dom_walker): Likewise.
7162 (dom_walker::walk): Handle NULL RPO order.
7163 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7165 (rewrite_update_dom_walker): Likewise.
7166 (mark_def_dom_walker): Likewise.
7168 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7170 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7171 (aarch64_maybe_expand_sve_subreg_move): Declare.
7172 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7173 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7175 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7176 that are semantically a reverse operation.
7177 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7178 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7179 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7181 (aarch64_can_change_mode_class): For big-endian, forbid changes
7182 between two SVE modes if they have different element sizes.
7184 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7186 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7187 the TImode handling for big-endian targets.
7189 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7191 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7192 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7194 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7195 Remove BSWAP handing for big-endian targets and use the form of
7196 LD1RQ appropariate for the mode.
7198 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7200 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7201 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7204 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7207 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7208 check for operands that need to go through aarch64_sve_reload_be.
7210 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7212 PR tree-optimization/81661
7213 PR tree-optimization/84117
7214 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7215 * tree-eh.c: Include gimplify.h.
7216 (find_trapping_overflow, replace_trapping_overflow,
7217 rewrite_to_non_trapping_overflow): New functions.
7218 * tree-vect-loop.c: Include tree-eh.h.
7219 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7220 * tree-data-ref.c: Include tree-eh.h.
7221 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7223 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7225 PR rtl-optimization/84123
7226 * combine.c (change_zero_ext): Check if hard register satisfies
7227 can_change_dest_mode before calling gen_lowpart_SUBREG.
7229 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7232 * ira.c (ira_init_register_move_cost): Remove assert.
7234 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7236 PR rtl-optimization/84071
7237 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7238 * doc/tm.texi: Regenerate.
7240 2018-01-31 Richard Biener <rguenther@suse.de>
7242 PR tree-optimization/84132
7243 * tree-data-ref.c (analyze_miv_subscript): Properly
7244 check whether evolution_function_is_affine_multivariate_p
7245 before calling gcd_of_steps_may_divide_p.
7247 2018-01-31 Julia Koval <julia.koval@intel.com>
7250 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7251 * config/i386/i386.md (rdpid_rex64) New.
7252 (rdpid): Make 32bit only.
7254 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7257 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7258 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7260 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7263 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7265 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7267 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7269 PR rtl-optimization/84071
7270 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7271 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7273 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7275 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7276 (arc_attribute_table): Add 'aux' attribute.
7277 (arc_in_small_data_p): Consider aux like variables.
7278 (arc_is_aux_reg_p): New function.
7279 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7280 (arc_get_aux_arg): New function.
7281 (prepare_move_operands): Handle aux-register access.
7282 (arc_handle_aux_attribute): New function.
7283 * doc/extend.texi (ARC Variable attributes): Add subsection.
7285 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7287 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7288 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7289 (arc_attribute_table): Add 'uncached' attribute.
7290 (arc_print_operand): Print '.di' flag for uncached memory
7292 (arc_in_small_data_p): Do not consider for small data the uncached
7294 (arc_is_uncached_mem_p): New function.
7295 * config/arc/predicates.md (compact_store_memory_operand): Check
7296 for uncached memory accesses.
7297 (nonvol_nonimm_operand): Likewise.
7298 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7300 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7303 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7304 falign-loops=): Add Optimization flag.
7306 2018-01-30 Jeff Law <law@redhat.com>
7309 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7310 INT_REGISTERS_SAVED. Check it prior to calling
7311 get_scratch_register_on_entry.
7312 (ix86_adjust_stack_and_probe): Similarly.
7313 (ix86_emit_probe_stack_range): Similarly.
7314 (ix86_expand_prologue): Corresponding changes.
7316 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7319 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7320 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7322 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7325 * lra-constraints.c (curr_insn_transform): Process AND in the
7328 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7330 PR rtl-optimization/83986
7331 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7332 dependence against last_pending_memory_flush in addition to
7335 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7337 PR tree-optimization/81611
7338 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7341 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7344 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7347 2018-01-30 Richard Biener <rguenther@suse.de>
7348 Jakub Jelinek <jakub@redhat.com>
7350 PR tree-optimization/84111
7351 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7352 inner loops added during recursion, as they don't have up-to-date
7355 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7358 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7359 (can_inline_edge_by_limits_p): ... here.
7360 (can_early_inline_edge_p, check_callers,
7361 update_caller_keys, update_callee_keys, recursive_inlining,
7362 add_new_edges_to_heap, speculation_useful_p,
7363 inline_small_functions,
7364 inline_small_functions, flatten_function,
7365 inline_to_all_callers_1): Update.
7367 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7369 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7372 2018-01-30 Richard Biener <rguenther@suse.de>
7374 PR tree-optimization/83008
7375 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7376 invariant and constant vector uses in stmts when they need
7379 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7382 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7383 * configure: Regenerate.
7385 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7387 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7389 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7390 Use gen_rtx_REG rather than gen_lowpart.
7392 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7394 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7395 rather than 0 when creating partial subregs.
7397 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7399 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7402 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7405 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7406 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7407 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7408 flags. This restores the settings used before the 2017-07-24.
7409 Turning off pre increment/decrement/modify allows IVOPTS to
7410 optimize DF/SF loops where the index is an int.
7412 2018-01-29 Richard Biener <rguenther@suse.de>
7413 Kelvin Nilsen <kelvin@gcc.gnu.org>
7416 * tree-vect-stmts.c (vectorizable_call): Don't call
7417 targetm.vectorize_builtin_md_vectorized_function if callee is
7420 2018-01-22 Carl Love <cel@us.ibm.com>
7422 * doc/extend.tex: Fix typo in second arg in
7423 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7425 2018-01-29 Richard Biener <rguenther@suse.de>
7427 PR tree-optimization/84086
7428 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7429 (flush_ssaname_freelist): When SSA names were released reset
7430 the SCEV hash table.
7432 2018-01-29 Richard Biener <rguenther@suse.de>
7434 PR tree-optimization/84057
7435 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7436 removed paths when removing edges.
7438 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7440 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7441 -mfunction-return=@var{choice}.
7443 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7446 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7448 (layout::print_source_line): Likewise.
7449 (test_get_line_width_without_trailing_whitespace): Add test cases.
7451 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7454 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7457 2018-01-26 Jim Wilson <jimw@sifive.com>
7459 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7461 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7464 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7466 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7467 and CMP + SUB-immediate -> SUBS.
7469 2018-01-26 Martin Sebor <msebor@redhat.com>
7471 PR tree-optimization/83896
7472 * tree-ssa-strlen.c (get_string_len): Rename...
7473 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7474 Avoid assuming length is constant.
7475 (handle_char_store): Use HOST_WIDE_INT for string length.
7477 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7480 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7481 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7483 2018-01-26 Richard Biener <rguenther@suse.de>
7485 PR rtl-optimization/84003
7486 * dse.c (record_store): Only record redundant stores when
7487 the earlier store aliases at least all accesses the later one does.
7489 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7491 PR rtl-optimization/83985
7492 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7493 REG_CFA_RESTORE insns.
7494 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7495 REG_CFA_RESTORE insns here.
7498 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7499 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7501 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7503 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7504 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7505 (arc_init): Likewise.
7506 (arc_override_options): Likewise.
7507 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7509 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7511 * config/arc/arc.h (TARGET_DBNZ): Define.
7512 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7513 properly set the tune attribute.
7514 (dbnz): Use TARGET_DBNZ guard.
7515 * config/arc/arc.opt (mtune): Add core3 option.
7517 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7519 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7520 recognize new pic like addresses.
7521 (arc_delegitimize_address): Clean up.
7523 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7525 * config/arc/arc-arches.def: Option mrf16 valid for all
7527 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7528 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7529 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7530 * config/arc/arc-tables.opt: Regenerate.
7531 * config/arc/arc.c (arc_conditional_register_usage): Handle
7532 reduced register file case.
7533 (arc_file_start): Set must have build attributes.
7534 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7536 * config/arc/arc.opt (mrf16): Add new option.
7537 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7538 * config/arc/genmultilib.awk: Handle new mrf16 option.
7539 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7540 * config/arc/t-multilib: Regenerate.
7541 * doc/invoke.texi (ARC Options): Document mrf16 option.
7543 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7545 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7546 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7547 (arc_attribute_table): Add 'secure_call' attribute.
7548 (arc_print_operand): Print secure call operand.
7549 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7551 (arc_is_secure_call_p): New function. * config/arc/arc.md
7552 (call_i): Add support for sjli instruction.
7553 (call_value_i): Likewise.
7554 * config/arc/constraints.md (Csc): New constraint.
7556 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7557 John Eric Martin <John.Martin@emmicro-us.com>
7559 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7560 * config/arc/arc.c (_arc_jli_section): New struct.
7561 (arc_jli_section): New type.
7562 (rc_jli_sections): New static variable.
7563 (arc_handle_jli_attribute): New function.
7564 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7565 (arc_file_end): New function.
7566 (TARGET_ASM_FILE_END): Define.
7567 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7568 (arc_add_jli_section): New function.
7569 (jli_call_scan): Likewise.
7570 (arc_reorg): Call jli_call_scan.
7571 (arc_output_addsi): Remove 'S' from printing asm operand.
7572 (arc_is_jli_call_p): New function.
7573 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7575 (movhi_insn): Likewise.
7576 (movsi_insn): Likewise.
7577 (movsi_set_cc_insn): Likewise.
7578 (loadqi_update): Likewise.
7579 (load_zeroextendqisi_update): Likewise.
7580 (load_signextendqisi_update): Likewise.
7581 (loadhi_update): Likewise.
7582 (load_zeroextendhisi_update): Likewise.
7583 (load_signextendhisi_update): Likewise.
7584 (loadsi_update): Likewise.
7585 (loadsf_update): Likewise.
7586 (movsicc_insn): Likewise.
7587 (bset_insn): Likewise.
7588 (bxor_insn): Likewise.
7589 (bclr_insn): Likewise.
7590 (bmsk_insn): Likewise.
7591 (bicsi3_insn): Likewise.
7592 (cmpsi_cc_c_insn): Likewise.
7593 (movsi_ne): Likewise.
7594 (movsi_cond_exec): Likewise.
7595 (clrsbsi2): Likewise.
7606 (call_i): Remove 'S' asm letter, add jli instruction.
7607 (call_value_i): Likewise.
7608 * config/arc/arc.op (mjli-always): New option.
7609 * config/arc/constraints.md (Cji): New constraint.
7610 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7612 (subsf3_fpx): Likewise.
7613 (mulsf3_fpx): Likewise.
7614 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7616 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7618 * doc/invoke.texi (ARC): Document mjli-always option.
7620 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7622 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7623 avoid addition with 0 and use incw and decw where possible.
7625 2018-01-26 Richard Biener <rguenther@suse.de>
7627 PR tree-optimization/81082
7628 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7629 association if it requires casting to unsigned.
7630 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7631 from fold_plusminus_mult_expr to catch important cases late when
7632 range info is available.
7634 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7636 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7637 * configure.ac (hidden_linkonce): New test.
7638 * configure: Regenerate.
7639 * config.in: Regenerate.
7641 2018-01-26 Julia Koval <julia.koval@intel.com>
7643 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7644 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7645 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7646 _mm_mask_bitshuffle_epi64_mask): Fix type.
7647 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7648 USI_FTYPE_V4DI_V4DI_USI): Remove.
7649 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7650 __builtin_ia32_vpshufbitqmb256_mask,
7651 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7652 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7653 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7655 2018-01-26 Alan Modra <amodra@gmail.com>
7658 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7659 UNSPEC_VBPERMQ. Sort other unspecs.
7661 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7663 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7665 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7668 * predict.c (drop_profile): Do not push/pop cfun; update also
7670 (handle_missing_profiles): Fix logic looking for zero profiles.
7672 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7675 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7676 on functions with #pragma omp declare simd or functions with simd
7678 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7679 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7680 Remove trailing \n from warning_at calls.
7682 2018-01-25 Tom de Vries <tom@codesourcery.com>
7685 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7686 for neutered workers.
7688 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7691 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7692 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7694 2018-01-24 Jeff Law <law@redhat.com>
7697 * i386.c (get_probe_interval): Move to earlier point.
7698 (ix86_compute_frame_layout): If -fstack-clash-protection and
7699 the frame is larger than the probe interval, then use pushes
7700 to save registers rather than reg->mem moves.
7701 (ix86_expand_prologue): Remove conditional for int_registers_saved
7704 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7707 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7708 min/max for never referenced object.
7710 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7713 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7715 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7716 attributes from DECL_ATTRIBUTES (decl) without affecting
7717 DECL_ATTRIBUTES (current_function_decl).
7718 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7719 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7721 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7723 PR tree-optimization/83979
7724 * fold-const.c (fold_comparison): Use constant_boolean_node
7725 instead of boolean_{true,false}_node.
7727 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7729 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7732 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7734 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7735 Simplify the clause that sets the length attribute.
7736 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7737 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7738 clause that sets the length attribute.
7739 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7741 2018-01-24 Tom de Vries <tom@codesourcery.com>
7744 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7745 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7746 Add strict parameter.
7747 (prevent_branch_around_nothing): Insert dummy insn between branch to
7748 label and label with no ptx insn inbetween.
7749 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7751 2018-01-24 Tom de Vries <tom@codesourcery.com>
7754 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7755 for neutered threads in warp.
7756 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7758 2018-01-24 Richard Biener <rguenther@suse.de>
7760 PR tree-optimization/83176
7761 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7764 2018-01-24 Richard Biener <rguenther@suse.de>
7766 PR tree-optimization/82819
7767 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7768 code generating pluses that are no-ops in the target precision.
7770 2018-01-24 Richard Biener <rguenther@suse.de>
7773 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7775 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7777 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7778 to merge probabilities.
7779 * predict.c (probably_never_executed): Also mark as cold functions
7780 with global 0 profile and guessed local profile.
7781 * profile-count.c (profile_probability::combine_with_count): New
7783 * profile-count.h (profile_probability::operator*,
7784 profile_probability::operator*=, profile_probability::operator/,
7785 profile_probability::operator/=): Reduce precision to adjusted
7786 and set value to guessed on contradictory divisions.
7787 (profile_probability::combine_with_freq): Remove.
7788 (profile_probability::combine_wiht_count): Declare.
7789 (profile_count::force_nonzero):: Set to adjusted.
7790 (profile_count::probability_in):: Set quality to adjusted.
7791 * tree-ssa-tail-merge.c (replace_block_by): Use
7794 2018-01-23 Andrew Waterman <andrew@sifive.com>
7795 Jim Wilson <jimw@sifive.com>
7797 * config/riscv/riscv.c (riscv_stack_boundary): New.
7798 (riscv_option_override): Set riscv_stack_boundary. Handle
7799 riscv_preferred_stack_boundary_arg.
7800 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7801 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7802 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7803 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7804 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7805 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7807 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7810 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7811 of struct ix86_frame.
7812 (ix86_expand_epilogue): Likewise. Add a local variable for
7813 the reg_save_offset field in struct ix86_frame.
7815 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7817 PR tree-optimization/82604
7818 * tree-loop-distribution.c (enum partition_kind): New enum item
7819 PKIND_PARTIAL_MEMSET.
7820 (partition_builtin_p): Support above new enum item.
7821 (generate_code_for_partition): Ditto.
7822 (compute_access_range): Differentiate cases that equality can be
7823 proven at all loops, the innermost loops or no loops.
7824 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7825 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7826 (finalize_partitions, distribute_loop): Don't fuse partition of
7827 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7828 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7831 2018-01-23 Martin Liska <mliska@suse.cz>
7833 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7834 order to ignore the predictor.
7835 (PRED_POLYMORPHIC_CALL): Likewise.
7836 (PRED_RECURSIVE_CALL): Likewise.
7838 2018-01-23 Martin Liska <mliska@suse.cz>
7840 * tree-profile.c (tree_profiling): Print function header to
7841 aware reader which function we are working on.
7842 * value-prof.c (gimple_find_values_to_profile): Do not print
7843 not interesting value histograms.
7845 2018-01-23 Martin Liska <mliska@suse.cz>
7847 * profile-count.h (enum profile_quality): Add
7848 profile_uninitialized as the first value. Do not number values
7849 as they are zero based.
7850 (profile_count::verify): Update sanity check.
7851 (profile_probability::verify): Likewise.
7853 2018-01-23 Nathan Sidwell <nathan@acm.org>
7855 * doc/invoke.texi (ffor-scope): Deprecate.
7857 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7859 PR tree-optimization/83510
7860 * domwalk.c (set_all_edges_as_executable): New function.
7861 (dom_walker::dom_walker): Convert bool param
7862 "skip_unreachable_blocks" to enum reachability. Move setup of
7863 edge flags to set_all_edges_as_executable and only do it when
7864 reachability is REACHABLE_BLOCKS.
7865 * domwalk.h (enum dom_walker::reachability): New enum.
7866 (dom_walker::dom_walker): Convert bool param
7867 "skip_unreachable_blocks" to enum reachability.
7868 (set_all_edges_as_executable): New decl.
7869 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7870 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7872 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7873 but converting true to REACHABLE_BLOCKS.
7874 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7876 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7877 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7878 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7880 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7881 if check_all_array_refs will be called.
7883 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7885 * tree.c (selftest::test_location_wrappers): Add more test
7888 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7890 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7891 (selftest::test_bit_in_range): Likewise.
7893 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7896 * doc/sourcebuild.texi (vect_float): Say that the selector
7897 only describes the situation when -funsafe-math-optimizations is on.
7898 (vect_float_strict): Document.
7900 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7902 PR tree-optimization/83965
7903 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7904 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7905 instead of checking only for a reduction.
7906 (vect_recog_widen_sum_pattern): Likewise.
7908 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7910 * predict.c (probably_never_executed): Only use precise profile info.
7911 (compute_function_frequency): Skip after inlining hack since we now
7912 have quality checking.
7914 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7916 * profile-count.h (profile_probability::very_unlikely,
7917 profile_probability::unlikely, profile_probability::even): Set
7918 precision to guessed.
7920 2018-01-23 Richard Biener <rguenther@suse.de>
7922 PR tree-optimization/83963
7923 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7924 Properly terminate dominator walk when crossing the exit edge not
7925 when visiting its source block.
7927 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7930 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7931 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7933 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7935 PR tree-optimization/83957
7936 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7937 semicolon after for body surrounded by braces.
7939 PR tree-optimization/83081
7940 * profile-count.h (profile_probability::split): New method.
7941 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7942 Use profile_probability::split.
7943 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7944 when splitting a single conditional jump into 2.
7946 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7948 PR tree-optimization/69452
7949 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7952 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7954 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7955 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7956 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7958 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7960 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7961 * config/rl78/rl78.md (movdi): New define_expand.
7962 * config/rl78/rl78.c (rl78_split_movdi): New function.
7964 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7967 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7969 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7970 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7971 128-bit to produce an UNSPEC move to get the double word with the
7972 signbit and then a shift directly to do signbit.
7973 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7974 implementation with a new version that just does either a direct
7975 move or a regular move. Move memory interface to separate insns.
7976 Move insns so they are next to the expander.
7977 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7978 with signbit move. Split big and little endian case.
7979 (signbit<mode>2_dm_mem_le): Likewise.
7980 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7981 (signbit<mode>2_dm2): Likewise.
7983 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7985 * config/rl78/rl78.md (anddi3): New define_expand.
7987 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7989 * config/rl78/rl78.md (umindi3): New define_expand.
7991 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7993 * config/rl78/rl78.md (smindi3): New define_expand.
7995 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7997 * config/rl78/rl78.md (smaxdi3): New define_expand.
7999 2018-01-22 Carl Love <cel@us.ibm.com>
8001 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8002 LVX_V1TI): Add macro expansion.
8003 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8004 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8005 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8006 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8007 Change check to determine if the instruction is a byte reversing
8008 entry. Fix typo in comment.
8009 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8010 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8011 Add def_builtin calls for new builtins.
8012 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8013 Add define_insn expansion.
8015 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8017 * config/rl78/rl78.md (umaxdi3): New define_expand.
8019 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8021 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8022 for non-QImode registers.
8024 2018-01-22 Richard Biener <rguenther@suse.de>
8026 PR tree-optimization/83963
8027 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8028 including the loop exit block.
8029 (scop_detection::merge_sese): Likewise.
8030 (scop_detection::add_scop): Do it here instead.
8032 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8034 * doc/sourcebuild.texi (arm_softfloat): Document.
8036 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8039 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8040 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8041 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8043 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8044 David Edelsohn <dje.gcc@gmail.com>
8047 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8048 Change "crset eq" to "crset 2".
8049 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8050 (*call_indirect_aix<mode>_nospec): Likewise.
8051 (*call_value_indirect_aix<mode>_nospec): Likewise.
8052 (*call_indirect_elfv2<mode>_nospec): Likewise.
8053 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8054 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8055 change assembly output from . to $.
8056 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8057 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8058 (*tablejump<mode>_internal1_nospec): Likewise.
8060 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8063 * config/sh/sh_optimize_sett_clrt.cc:
8064 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8066 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8068 PR tree-optimization/83940
8069 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8070 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8071 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8072 use it to pass back the definition type.
8073 (vect_check_store_rhs): Likewise rhs_dt_out.
8074 (vect_build_gather_load_calls): Add a mask_dt argument and use
8075 it instead of a call to vect_is_simple_use.
8076 (vectorizable_store): Update calls to vect_check_load_store_mask
8077 and vect_check_store_rhs. Use the dt returned by the latter instead
8078 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8079 instead of calls to vect_is_simple_use. Pass the scalar rather
8080 than the vector operand to vect_is_simple_use when handling
8081 second and subsequent copies of an rhs value.
8082 (vectorizable_load): Update calls to vect_check_load_store_mask
8083 and vect_build_gather_load_calls. Use the cached mask_dt and
8084 gs_info.offset_dt instead of calls to vect_is_simple_use.
8086 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8089 * tree-emutls.c: Include gimplify.h.
8090 (lower_emutls_2): New function.
8091 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8092 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8093 it before further processing.
8096 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8097 UINTVAL (trueop1) instead of INTVAL (op1).
8099 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8103 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8104 INCOMING_FRAME_SP_OFFSET if not defined.
8105 (scan_trace): Add ENTRY argument. If true and
8106 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8107 emit a note to adjust the CFA offset.
8108 (create_cfi_notes): Adjust scan_trace callers.
8109 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8110 INCOMING_FRAME_SP_OFFSET in the CIE.
8111 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8112 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8114 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8115 * doc/tm.texi: Regenerated.
8117 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8119 PR rtl-optimization/83147
8120 * lra-constraints.c (remove_inheritance_pseudos): Use
8121 lra_substitute_pseudo_within_insn.
8123 2018-01-19 Tom de Vries <tom@codesourcery.com>
8124 Cesar Philippidis <cesar@codesourcery.com>
8127 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8129 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8132 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8133 spaces for function labels.
8135 2018-01-19 Martin Liska <mliska@suse.cz>
8137 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8138 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8139 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8140 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8141 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8142 (PRED_CONST_RETURN): Change from 69 to 65.
8143 (PRED_NULL_RETURN): Change from 91 to 71.
8144 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8145 (PRED_LOOP_GUARD): Change from 66 to 73.
8147 2018-01-19 Martin Liska <mliska@suse.cz>
8149 * predict.c (predict_insn_def): Add new assert.
8150 (struct branch_predictor): Change type to signed integer.
8151 (test_prediction_value_range): Amend test to cover
8153 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8154 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8155 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8156 (PRED_LOOP_IV_COMPARE): Likewise.
8157 * predict.h (PROB_UNINITIALIZED): Define new constant.
8159 2018-01-19 Martin Liska <mliska@suse.cz>
8161 * predict.c (dump_prediction): Add new format for
8162 analyze_brprob.py script which is enabled with -details
8164 * profile-count.h (precise_p): New function.
8166 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8168 PR tree-optimization/83922
8169 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8170 there are no statements that need masking.
8171 (vect_active_double_reduction_p): New function.
8172 (vect_analyze_loop_operations): Use it when handling phis that
8173 are not in the loop header.
8175 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8177 PR tree-optimization/83914
8178 * tree-vect-loop.c (vectorizable_induction): Don't convert
8179 init_expr or apply the peeling adjustment for inductions
8180 that are nested within the vectorized loop.
8182 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8184 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8187 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8191 * function.h (gimplify_parameters): Add gimple_seq * argument.
8192 * function.c: Include gimple.h and options.h.
8193 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8194 for the added local temporaries if needed.
8195 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8196 if there are any parameter cleanups, wrap whole body into a
8197 try/finally with the cleanups.
8199 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8202 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8203 Use GET_MODE_CLASS for scalar floating point.
8205 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8209 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8210 Fix call of call_cgraph_insertion_hooks.
8212 2018-01-18 Martin Sebor <msebor@redhat.com>
8214 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8216 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8219 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8222 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8225 * common.opt: (-ffile-prefix-map): New option.
8226 * opts.c (common_handle_option): Defer it.
8227 * opts-global.c (handle_common_deferred_options): Handle it.
8228 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8229 * file-prefix-map.h: New file.
8230 (remap_debug_filename, add_debug_prefix_map): ...here.
8231 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8232 * final.c (debug_prefix_map, add_debug_prefix_map
8233 remap_debug_filename): Move to...
8234 * file-prefix-map.c: New file.
8235 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8236 generalize, get rid of alloca(), use strrchr() instead of strchr().
8237 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8238 Implement in terms of add_prefix_map().
8239 (remap_macro_filename, remap_debug_filename): Implement in term of
8241 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8242 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8243 * dbxout.c: Include file-prefix-map.h.
8244 * varasm.c: Likewise.
8245 * vmsdbgout.c: Likewise.
8246 * xcoffout.c: Likewise.
8247 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8248 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8249 * doc/invoke.texi (-ffile-prefix-map): Document.
8250 (-fdebug-prefix-map): Update description.
8252 2018-01-18 Martin Liska <mliska@suse.cz>
8254 * config/i386/i386.c (indirect_thunk_name): Document that also
8256 (output_indirect_thunk): Document why both instructions
8257 (pause and lfence) are generated.
8259 2018-01-18 Richard Biener <rguenther@suse.de>
8261 PR tree-optimization/83887
8262 * graphite-scop-detection.c
8263 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8264 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8265 (scop_detection::merge_sese): Re-implement with a flood-fill
8266 algorithm that properly finds a SESE region if it exists.
8268 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8271 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8272 pointer_diff optimizations use view_convert instead of convert.
8274 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8276 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8277 Generate different code for -mno-speculate-indirect-jumps.
8278 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8279 (*call_indirect_aix<mode>): Disable for
8280 -mno-speculate-indirect-jumps.
8281 (*call_indirect_aix<mode>_nospec): New define_insn.
8282 (*call_value_indirect_aix<mode>): Disable for
8283 -mno-speculate-indirect-jumps.
8284 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8285 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8286 -mno-speculate-indirect-jumps.
8287 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8289 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8291 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8292 long double type, set the flags for noting the default long double
8293 type, even if we don't pass or return a long double type.
8295 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8298 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8301 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8303 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8304 support for merge[hl].
8305 (fold_mergehl_helper): New helper function.
8306 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8307 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8308 (altivec_vmrglw_direct): Add xxmrglw insn.
8310 2018-01-17 Andrew Waterman <andrew@sifive.com>
8312 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8313 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8315 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8318 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8319 call the lto_location_cache before reading the
8320 DECL_SOURCE_LOCATION of the types.
8322 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8323 Richard Sandiford <richard.sandiford@linaro.org>
8325 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8326 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8327 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8328 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8329 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8331 * config/aarch64/constraints.md (aarch64_movti_operand):
8333 * config/aarch64/predicates.md (Uti): Add new constraint.
8335 2018-01-17 Carl Love <cel@us.ibm.com>
8337 * config/rs6000/vsx.md (define_expand xl_len_r,
8338 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8339 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8341 (define_expand, define_insn): Move the shift left from the
8342 define_insn to the define_expand for lxvl and stxvl instructions.
8343 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8344 and XL_LEN_R definitions to PURE.
8346 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8348 * config/i386/i386.c (indirect_thunk_name): Declare regno
8349 as unsigned int. Compare regno with INVALID_REGNUM.
8350 (output_indirect_thunk): Ditto.
8351 (output_indirect_thunk_function): Ditto.
8352 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8353 in the call to output_indirect_thunk_function.
8355 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8358 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8359 rather than the size of inner_type to determine the stack slot size
8360 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8362 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8365 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8368 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8370 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8371 endian Linux systems to optionally enable multilibs for selecting
8372 the long double type if the user configured an explicit type.
8373 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8374 have no long double multilibs if not defined.
8375 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8376 warn if the user used -mabi={ieee,ibm}longdouble and we built
8377 multilibs for long double.
8378 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8379 appropriate multilib option.
8380 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8382 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8383 for building long double multilibs.
8384 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8386 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8388 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8391 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8393 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8396 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8399 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8402 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8404 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8405 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8407 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8409 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8410 different rtl trees depending on TARGET_64BIT.
8411 (rs6000_gen_lvx): Likewise.
8413 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8415 * config/visium/visium.md (nop): Tweak comment.
8416 (hazard_nop): Likewise.
8418 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8420 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8421 -mspeculate-indirect-jumps.
8422 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8423 for -mno-speculate-indirect-jumps.
8424 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8425 (*call_value_indirect_elfv2<mode>): Disable for
8426 -mno-speculate-indirect-jumps.
8427 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8428 (indirect_jump): Emit different RTL for
8429 -mno-speculate-indirect-jumps.
8430 (*indirect_jump<mode>): Disable for
8431 -mno-speculate-indirect-jumps.
8432 (*indirect_jump<mode>_nospec): New define_insn.
8433 (tablejump): Emit different RTL for
8434 -mno-speculate-indirect-jumps.
8435 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8436 (tablejumpsi_nospec): New define_expand.
8437 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8438 (tablejumpdi_nospec): New define_expand.
8439 (*tablejump<mode>_internal1): Disable for
8440 -mno-speculate-indirect-jumps.
8441 (*tablejump<mode>_internal1_nospec): New define_insn.
8442 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8445 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8447 * caller-save.c (insert_save): Drop unnecessary parameter. All
8450 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8451 Richard Biener <rguenth@suse.de>
8454 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8455 return early, inline manually is_gimple_sizepos. Make sure if we
8456 call gimplify_expr we don't end up with a gimple constant.
8457 * tree.c (variably_modified_type_p): Don't return true for
8458 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8459 * gimplify.h (is_gimple_sizepos): Remove.
8461 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8463 PR tree-optimization/83857
8464 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8465 vectorizable_live_operation for pure SLP statements.
8466 (vectorizable_live_operation): Handle PHIs.
8468 2018-01-16 Richard Biener <rguenther@suse.de>
8470 PR tree-optimization/83867
8471 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8472 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8474 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8477 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8478 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8479 If off is not INTEGER_CST, issue a may not be aligned warning
8480 rather than isn't aligned. Use isn%'t rather than isn't.
8481 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8483 <case MULT_EXPR>: Improve the case when bottom and one of the
8484 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8485 operand, in that case check if the other operand is multiple of
8486 bottom divided by the INTEGER_CST operand.
8488 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8491 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8492 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8493 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8494 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8495 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8496 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8497 (pa_function_arg_size): New function.
8499 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8501 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8502 in a separate statement.
8504 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8506 PR tree-optimization/83847
8507 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8508 group gathers and scatters.
8510 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8512 PR rtl-optimization/86620
8513 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8515 PR rtl-optimization/83213
8516 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8517 to last if both are JUMP_INSNs.
8519 PR tree-optimization/83843
8520 * gimple-ssa-store-merging.c
8521 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8522 store_immediate_info for bswap/nop orig_stores.
8524 2018-01-15 Andrew Waterman <andrew@sifive.com>
8526 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8528 <UDIV>: Increase cost if !TARGET_DIV.
8530 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8532 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8533 (define_attr "cr_logical_3op"): New.
8534 (cceq_ior_compare): Adjust.
8535 (cceq_ior_compare_complement): Adjust.
8536 (*cceq_rev_compare): Adjust.
8537 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8538 (is_cracked_insn): Adjust.
8539 (insn_must_be_first_in_group): Adjust.
8540 * config/rs6000/40x.md: Adjust.
8541 * config/rs6000/440.md: Adjust.
8542 * config/rs6000/476.md: Adjust.
8543 * config/rs6000/601.md: Adjust.
8544 * config/rs6000/603.md: Adjust.
8545 * config/rs6000/6xx.md: Adjust.
8546 * config/rs6000/7450.md: Adjust.
8547 * config/rs6000/7xx.md: Adjust.
8548 * config/rs6000/8540.md: Adjust.
8549 * config/rs6000/cell.md: Adjust.
8550 * config/rs6000/e300c2c3.md: Adjust.
8551 * config/rs6000/e500mc.md: Adjust.
8552 * config/rs6000/e500mc64.md: Adjust.
8553 * config/rs6000/e5500.md: Adjust.
8554 * config/rs6000/e6500.md: Adjust.
8555 * config/rs6000/mpc.md: Adjust.
8556 * config/rs6000/power4.md: Adjust.
8557 * config/rs6000/power5.md: Adjust.
8558 * config/rs6000/power6.md: Adjust.
8559 * config/rs6000/power7.md: Adjust.
8560 * config/rs6000/power8.md: Adjust.
8561 * config/rs6000/power9.md: Adjust.
8562 * config/rs6000/rs64.md: Adjust.
8563 * config/rs6000/titan.md: Adjust.
8565 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8567 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8568 ix86_indirect_branch_register logic.
8570 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8572 * config/i386/constraints.md (Bs): Update
8573 ix86_indirect_branch_register check. Don't check
8574 ix86_indirect_branch_register with GOT_memory_operand.
8576 * config/i386/predicates.md (GOT_memory_operand): Don't check
8577 ix86_indirect_branch_register here.
8578 (GOT32_symbol_operand): Likewise.
8580 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8582 * config/i386/predicates.md (constant_call_address_operand):
8583 Rewrite ix86_indirect_branch_register logic.
8584 (sibcall_insn_operand): Likewise.
8586 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8588 * config/i386/constraints.md (Bs): Replace
8589 ix86_indirect_branch_thunk_register with
8590 ix86_indirect_branch_register.
8592 * config/i386/i386.md (indirect_jump): Likewise.
8593 (tablejump): Likewise.
8594 (*sibcall_memory): Likewise.
8595 (*sibcall_value_memory): Likewise.
8596 Peepholes of indirect call and jump via memory: Likewise.
8597 * config/i386/i386.opt: Likewise.
8598 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8599 (GOT_memory_operand): Likewise.
8600 (call_insn_operand): Likewise.
8601 (sibcall_insn_operand): Likewise.
8602 (GOT32_symbol_operand): Likewise.
8604 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8607 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8608 type rather than type addr's type points to.
8609 (expand_omp_atomic_mutex): Likewise.
8610 (expand_omp_atomic): Likewise.
8612 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8615 * config/i386/i386.c (output_indirect_thunk_function): Use
8616 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8617 for __x86_return_thunk.
8619 2018-01-15 Richard Biener <rguenther@suse.de>
8622 * expmed.c (extract_bit_field_1): Fix typo.
8624 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8627 * config/arm/iterators.md (VF): New mode iterator.
8628 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8629 Remove integer-related logic from pattern.
8630 (neon_vabd<mode>_3): Likewise.
8632 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8635 * common.opt (fstrict-overflow): No longer an alias.
8636 (fwrapv-pointer): New option.
8637 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8638 also for pointer types based on flag_wrapv_pointer.
8639 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8640 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8641 opts->x_flag_wrapv got set.
8642 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8643 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8644 POINTER_TYPE_OVERFLOW_UNDEFINED.
8645 * match.pd: Likewise in address comparison pattern.
8646 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8648 2018-01-15 Richard Biener <rguenther@suse.de>
8651 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8652 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8653 Reset type names to their identifier if their TYPE_DECL doesn't
8654 have linkage (and thus is used for ODR and devirt).
8655 (save_debug_info_for_decl): Remove.
8656 (save_debug_info_for_type): Likewise.
8657 (add_tree_to_fld_list): Adjust.
8658 * tree-pretty-print.c (dump_generic_node): Make dumping of
8659 type names more robust.
8661 2018-01-15 Richard Biener <rguenther@suse.de>
8663 * BASE-VER: Bump to 8.0.1.
8665 2018-01-14 Martin Sebor <msebor@redhat.com>
8668 * builtins.c (check_access): Avoid warning when the no-warning bit
8671 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8673 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8674 * ira-color (allocno_hard_regs_compare): Likewise.
8676 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8679 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8680 Use .pushsection/.popsection.
8682 2018-01-14 Martin Sebor <msebor@redhat.com>
8685 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8687 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8689 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8690 entry from extra_headers.
8691 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8692 extra_headers, make the list bitwise identical to the i?86-*-* one.
8694 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8696 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8697 -mcmodel=large with -mindirect-branch=thunk,
8698 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8699 -mfunction-return=thunk-extern.
8700 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8701 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8702 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8704 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8706 * config/i386/i386.c (print_reg): Print the name of the full
8707 integer register without '%'.
8708 (ix86_print_operand): Handle 'V'.
8709 * doc/extend.texi: Document 'V' modifier.
8711 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8713 * config/i386/constraints.md (Bs): Disallow memory operand for
8714 -mindirect-branch-register.
8716 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8717 (GOT_memory_operand): Likewise.
8718 (call_insn_operand): Likewise.
8719 (sibcall_insn_operand): Likewise.
8720 (GOT32_symbol_operand): Likewise.
8721 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8722 for -mindirect-branch-register.
8723 (tablejump): Likewise.
8724 (*sibcall_memory): Likewise.
8725 (*sibcall_value_memory): Likewise.
8726 Disallow peepholes of indirect call and jump via memory for
8727 -mindirect-branch-register.
8728 (*call_pop): Replace m with Bw.
8729 (*call_value_pop): Likewise.
8730 (*sibcall_pop_memory): Replace m with Bs.
8731 * config/i386/i386.opt (mindirect-branch-register): New option.
8732 * doc/invoke.texi: Document -mindirect-branch-register option.
8734 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8736 * config/i386/i386-protos.h (ix86_output_function_return): New.
8737 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8738 set function_return_type.
8739 (indirect_thunk_name): Add ret_p to indicate thunk for function
8741 (output_indirect_thunk_function): Pass false to
8742 indirect_thunk_name.
8743 (ix86_output_indirect_branch_via_reg): Likewise.
8744 (ix86_output_indirect_branch_via_push): Likewise.
8745 (output_indirect_thunk_function): Create alias for function
8746 return thunk if regno < 0.
8747 (ix86_output_function_return): New function.
8748 (ix86_handle_fndecl_attribute): Handle function_return.
8749 (ix86_attribute_table): Add function_return.
8750 * config/i386/i386.h (machine_function): Add
8751 function_return_type.
8752 * config/i386/i386.md (simple_return_internal): Use
8753 ix86_output_function_return.
8754 (simple_return_internal_long): Likewise.
8755 * config/i386/i386.opt (mfunction-return=): New option.
8756 (indirect_branch): Mention -mfunction-return=.
8757 * doc/extend.texi: Document function_return function attribute.
8758 * doc/invoke.texi: Document -mfunction-return= option.
8760 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8762 * config/i386/i386-opts.h (indirect_branch): New.
8763 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8764 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8765 with local indirect jump when converting indirect call and jump.
8766 (ix86_set_indirect_branch_type): New.
8767 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8768 (indirectlabelno): New.
8769 (indirect_thunk_needed): Likewise.
8770 (indirect_thunk_bnd_needed): Likewise.
8771 (indirect_thunks_used): Likewise.
8772 (indirect_thunks_bnd_used): Likewise.
8773 (INDIRECT_LABEL): Likewise.
8774 (indirect_thunk_name): Likewise.
8775 (output_indirect_thunk): Likewise.
8776 (output_indirect_thunk_function): Likewise.
8777 (ix86_output_indirect_branch_via_reg): Likewise.
8778 (ix86_output_indirect_branch_via_push): Likewise.
8779 (ix86_output_indirect_branch): Likewise.
8780 (ix86_output_indirect_jmp): Likewise.
8781 (ix86_code_end): Call output_indirect_thunk_function if needed.
8782 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8784 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8785 (ix86_attribute_table): Add indirect_branch.
8786 * config/i386/i386.h (machine_function): Add indirect_branch_type
8787 and has_local_indirect_jump.
8788 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8790 (tablejump): Likewise.
8791 (*indirect_jump): Use ix86_output_indirect_jmp.
8792 (*tablejump_1): Likewise.
8793 (simple_return_indirect_internal): Likewise.
8794 * config/i386/i386.opt (mindirect-branch=): New option.
8795 (indirect_branch): New.
8798 (thunk-inline): Likewise.
8799 (thunk-extern): Likewise.
8800 * doc/extend.texi: Document indirect_branch function attribute.
8801 * doc/invoke.texi: Document -mindirect-branch= option.
8803 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8806 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8808 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8810 * ipa-inline.c (want_inline_small_function_p): Return false if
8811 inlining has already failed with CIF_FINAL_ERROR.
8812 (update_caller_keys): Call want_inline_small_function_p before
8814 (update_callee_keys): Likewise.
8816 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8818 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8820 (rs6000_quadword_masked_address_p): Likewise.
8821 (quad_aligned_load_p): Likewise.
8822 (quad_aligned_store_p): Likewise.
8823 (const_load_sequence_p): Add comment to describe the outer-most loop.
8824 (mimic_memory_attributes_and_flags): New function.
8825 (rs6000_gen_stvx): Likewise.
8826 (replace_swapped_aligned_store): Likewise.
8827 (rs6000_gen_lvx): Likewise.
8828 (replace_swapped_aligned_load): Likewise.
8829 (replace_swapped_load_constant): Capitalize argument name in
8830 comment describing this function.
8831 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8832 and stores that access quad-word aligned addresses and replace
8833 with stvx or lvx instructions when appropriate.
8834 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8835 New function prototype.
8836 (rs6000_quadword_masked_address_p): Likewise.
8837 (rs6000_gen_lvx): Likewise.
8838 (rs6000_gen_stvx): Likewise.
8839 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8840 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8841 when memory address is aligned.
8842 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8843 this split to select lvx instruction when memory address is aligned.
8844 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8845 instruction when memory address is aligned.
8846 (*vsx_le_perm_load_v16qi): Likewise.
8847 (four unnamed splitters): Modify to select the stvx instruction
8848 when memory is aligned.
8850 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8852 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8853 which appears in the queue multiple times.
8855 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8856 Alan Hayward <alan.hayward@arm.com>
8857 David Sherwood <david.sherwood@arm.com>
8859 * tree-vectorizer.h (vec_lower_bound): New structure.
8860 (_loop_vec_info): Add check_nonzero and lower_bounds.
8861 (LOOP_VINFO_CHECK_NONZERO): New macro.
8862 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8863 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8864 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8865 fields. Make seg_len the distance travelled, not including the
8867 (dr_direction_indicator): Declare.
8868 (dr_zero_step_indicator): Likewise.
8869 (dr_known_forward_stride_p): Likewise.
8870 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8872 (runtime_alias_check_p): Allow runtime alias checks with
8874 (operator ==): Compare access_size and align.
8875 (prune_runtime_alias_test_list): Rework for new distinction between
8876 the access_size and seg_len.
8877 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8879 (get_segment_min_max): New function.
8880 (create_intersect_range_checks): Use it.
8881 (dr_step_indicator): New function.
8882 (dr_direction_indicator): Likewise.
8883 (dr_zero_step_indicator): Likewise.
8884 (dr_known_forward_stride_p): Likewise.
8885 * tree-loop-distribution.c (data_ref_segment_size): Return
8886 DR_STEP * (niters - 1).
8887 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8889 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8890 (vect_preserves_scalar_order_p): New function, split out from...
8891 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8892 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8893 (vect_vfa_access_size): New function.
8894 (vect_vfa_align): Likewise.
8895 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8896 (dump_lower_bound): New function.
8897 (vect_check_lower_bound): Likewise.
8898 (vect_small_gap_p): Likewise.
8899 (vectorizable_with_step_bound_p): Likewise.
8900 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8901 depencies if the vectorization factor is 1. Convert the checks
8902 for nonzero steps into checks on the bounds of DR_STEP. Try using
8903 a bunds check for variable steps if the minimum required step is
8904 relatively small. Update calls to the dr_with_seg_len
8905 constructor and to vect_compile_time_alias.
8906 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8908 (vect_loop_versioning): Call it.
8909 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8911 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8913 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8914 Alan Hayward <alan.hayward@arm.com>
8915 David Sherwood <david.sherwood@arm.com>
8917 * doc/sourcebuild.texi (vect_scatter_store): Document.
8918 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8920 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8922 * genopinit.c (main): Add supports_vec_scatter_store and
8923 supports_vec_scatter_store_cached to target_optabs.
8924 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8925 IFN_MASK_SCATTER_STORE.
8926 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8928 * internal-fn.h (internal_store_fn_p): Declare.
8929 (internal_fn_stored_value_index): Likewise.
8930 * internal-fn.c (scatter_store_direct): New macro.
8931 (expand_scatter_store_optab_fn): New function.
8932 (direct_scatter_store_optab_supported_p): New macro.
8933 (internal_store_fn_p): New function.
8934 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8935 IFN_MASK_SCATTER_STORE.
8936 (internal_fn_mask_index): Likewise.
8937 (internal_fn_stored_value_index): New function.
8938 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8940 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8941 * optabs-query.c (supports_vec_scatter_store_p): New function.
8942 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8943 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8944 true for scatter stores.
8945 (vect_gather_scatter_fn_p): Handle scatter stores too.
8946 (vect_check_gather_scatter): Consider using scatter stores if
8947 supports_vec_scatter_store_p.
8948 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8950 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8951 internal_fn_stored_value_index.
8952 (check_load_store_masking): Handle scatter stores too.
8953 (vect_get_store_rhs): Make public.
8954 (vectorizable_call): Use internal_store_fn_p.
8955 (vectorizable_store): Handle scatter store internal functions.
8956 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8957 when deciding whether the end of the group has been reached.
8958 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8959 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8960 (mask_scatter_store<mode>): New insns.
8962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8963 Alan Hayward <alan.hayward@arm.com>
8964 David Sherwood <david.sherwood@arm.com>
8966 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8967 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8968 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8970 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8971 Use vect_truncate_gather_scatter_offset if we can't treat the
8972 operation as a normal gather load or scatter store.
8973 (get_group_load_store_type): Take the gather_scatter_info
8974 as argument. Try using a gather load or scatter store for
8975 single-element groups.
8976 (get_load_store_type): Update calls to get_group_load_store_type
8977 and vect_use_strided_gather_scatters_p.
8979 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8980 Alan Hayward <alan.hayward@arm.com>
8981 David Sherwood <david.sherwood@arm.com>
8983 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8984 optional tree argument.
8985 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8987 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8988 but continue to use the current value as a fallback.
8989 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8990 to compare the updates.
8991 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8992 (get_load_store_type): Use it when handling a strided access.
8993 (vect_get_strided_load_store_ops): New function.
8994 (vect_get_data_ptr_increment): Likewise.
8995 (vectorizable_load): Handle strided gather loads. Always pass
8996 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8998 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8999 Alan Hayward <alan.hayward@arm.com>
9000 David Sherwood <david.sherwood@arm.com>
9002 * doc/md.texi (gather_load@var{m}): Document.
9003 (mask_gather_load@var{m}): Likewise.
9004 * genopinit.c (main): Add supports_vec_gather_load and
9005 supports_vec_gather_load_cached to target_optabs.
9006 * optabs-tree.c (init_tree_optimization_optabs): Use
9007 ggc_cleared_alloc to allocate target_optabs.
9008 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9009 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9011 * internal-fn.h (internal_load_fn_p): Declare.
9012 (internal_gather_scatter_fn_p): Likewise.
9013 (internal_fn_mask_index): Likewise.
9014 (internal_gather_scatter_fn_supported_p): Likewise.
9015 * internal-fn.c (gather_load_direct): New macro.
9016 (expand_gather_load_optab_fn): New function.
9017 (direct_gather_load_optab_supported_p): New macro.
9018 (direct_internal_fn_optab): New function.
9019 (internal_load_fn_p): Likewise.
9020 (internal_gather_scatter_fn_p): Likewise.
9021 (internal_fn_mask_index): Likewise.
9022 (internal_gather_scatter_fn_supported_p): Likewise.
9023 * optabs-query.c (supports_at_least_one_mode_p): New function.
9024 (supports_vec_gather_load_p): Likewise.
9025 * optabs-query.h (supports_vec_gather_load_p): Declare.
9026 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9027 and memory_type field.
9028 (NUM_PATTERNS): Bump to 15.
9029 * tree-vect-data-refs.c: Include internal-fn.h.
9030 (vect_gather_scatter_fn_p): New function.
9031 (vect_describe_gather_scatter_call): Likewise.
9032 (vect_check_gather_scatter): Try using internal functions for
9033 gather loads. Recognize existing calls to a gather load function.
9034 (vect_analyze_data_refs): Consider using gather loads if
9035 supports_vec_gather_load_p.
9036 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9037 (vect_get_gather_scatter_offset_type): Likewise.
9038 (vect_convert_mask_for_vectype): Likewise.
9039 (vect_add_conversion_to_patterm): Likewise.
9040 (vect_try_gather_scatter_pattern): Likewise.
9041 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9042 (vect_vect_recog_func_ptrs): Add it.
9043 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9044 internal_fn_mask_index and internal_gather_scatter_fn_p.
9045 (check_load_store_masking): Take the gather_scatter_info as an
9046 argument and handle gather loads.
9047 (vect_get_gather_scatter_ops): New function.
9048 (vectorizable_call): Check internal_load_fn_p.
9049 (vectorizable_load): Likewise. Handle gather load internal
9051 (vectorizable_store): Update call to check_load_store_masking.
9052 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9053 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9054 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9055 (aarch64_gather_scale_operand_d): New predicates.
9056 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9057 (mask_gather_load<mode>): New insns.
9059 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9060 Alan Hayward <alan.hayward@arm.com>
9061 David Sherwood <david.sherwood@arm.com>
9063 * optabs.def (fold_left_plus_optab): New optab.
9064 * doc/md.texi (fold_left_plus_@var{m}): Document.
9065 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9066 * internal-fn.c (fold_left_direct): Define.
9067 (expand_fold_left_optab_fn): Likewise.
9068 (direct_fold_left_optab_supported_p): Likewise.
9069 * fold-const-call.c (fold_const_fold_left): New function.
9070 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9071 * tree-parloops.c (valid_reduction_p): New function.
9072 (gather_scalar_reductions): Use it.
9073 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9074 (vect_finish_replace_stmt): Declare.
9075 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9076 (needs_fold_left_reduction_p): New function, split out from...
9077 (vect_is_simple_reduction): ...here. Accept reductions that
9078 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9079 (vect_force_simple_reduction): Also store the reduction type in
9080 the assignment's STMT_VINFO_REDUC_TYPE.
9081 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9082 (merge_with_identity): New function.
9083 (vect_expand_fold_left): Likewise.
9084 (vectorize_fold_left_reduction): Likewise.
9085 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9086 scalar phi in place for it. Check for target support and reject
9087 cases that would reassociate the operation. Defer the transform
9088 phase to vectorize_fold_left_reduction.
9089 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9090 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9091 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9093 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9095 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9096 call to ifc_temp_var.
9098 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9099 Alan Hayward <alan.hayward@arm.com>
9100 David Sherwood <david.sherwood@arm.com>
9102 * target.def (legitimize_address_displacement): Take the original
9103 offset as a poly_int.
9104 * targhooks.h (default_legitimize_address_displacement): Update
9106 * targhooks.c (default_legitimize_address_displacement): Likewise.
9107 * doc/tm.texi: Regenerate.
9108 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9109 as an argument, moving assert of ad->disp == ad->disp_term to...
9110 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9111 Try calling targetm.legitimize_address_displacement before expanding
9112 the address rather than afterwards, and adjust for the new interface.
9113 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9114 Match the new hook interface. Handle SVE addresses.
9115 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9118 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9120 * Makefile.in (OBJS): Add early-remat.o.
9121 * target.def (select_early_remat_modes): New hook.
9122 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9123 * doc/tm.texi: Regenerate.
9124 * targhooks.h (default_select_early_remat_modes): Declare.
9125 * targhooks.c (default_select_early_remat_modes): New function.
9126 * timevar.def (TV_EARLY_REMAT): New timevar.
9127 * passes.def (pass_early_remat): New pass.
9128 * tree-pass.h (make_pass_early_remat): Declare.
9129 * early-remat.c: New file.
9130 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9132 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9135 Alan Hayward <alan.hayward@arm.com>
9136 David Sherwood <david.sherwood@arm.com>
9138 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9139 vfm1 with a bound_epilog parameter.
9140 (vect_do_peeling): Update calls accordingly, and move the prologue
9141 call earlier in the function. Treat the base bound_epilog as 0 for
9142 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9143 this base when peeling for gaps.
9144 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9145 with fully-masked loops.
9146 (vect_estimate_min_profitable_iters): Handle the single peeled
9147 iteration in that case.
9149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9150 Alan Hayward <alan.hayward@arm.com>
9151 David Sherwood <david.sherwood@arm.com>
9153 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9154 single-element interleaving even if the size is not a power of 2.
9155 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9156 accesses for single-element interleaving if the group size is
9159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9160 Alan Hayward <alan.hayward@arm.com>
9161 David Sherwood <david.sherwood@arm.com>
9163 * doc/md.texi (fold_extract_last_@var{m}): Document.
9164 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9165 * optabs.def (fold_extract_last_optab): New optab.
9166 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9167 * internal-fn.c (fold_extract_direct): New macro.
9168 (expand_fold_extract_optab_fn): Likewise.
9169 (direct_fold_extract_optab_supported_p): Likewise.
9170 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9171 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9172 EXTRACT_LAST_REDUCTION.
9173 (get_initial_def_for_reduction): Do not create an initial vector
9174 for EXTRACT_LAST_REDUCTION reductions.
9175 (vectorizable_reduction): Leave the scalar phi in place for
9176 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9177 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9178 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9179 transform phase to vectorizable_condition.
9180 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9182 (vect_finish_stmt_generation): ...here.
9183 (vect_finish_replace_stmt): New function.
9184 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9185 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9187 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9189 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9190 Alan Hayward <alan.hayward@arm.com>
9191 David Sherwood <david.sherwood@arm.com>
9193 * doc/md.texi (extract_last_@var{m}): Document.
9194 * optabs.def (extract_last_optab): New optab.
9195 * internal-fn.def (EXTRACT_LAST): New internal function.
9196 * internal-fn.c (cond_unary_direct): New macro.
9197 (expand_cond_unary_optab_fn): Likewise.
9198 (direct_cond_unary_optab_supported_p): Likewise.
9199 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9200 loops using EXTRACT_LAST.
9201 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9202 (extract_last_<mode>): ...this optab.
9203 (vec_extract<mode><Vel>): Update accordingly.
9205 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9206 Alan Hayward <alan.hayward@arm.com>
9207 David Sherwood <david.sherwood@arm.com>
9209 * target.def (empty_mask_is_expensive): New hook.
9210 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9211 * doc/tm.texi: Regenerate.
9212 * targhooks.h (default_empty_mask_is_expensive): Declare.
9213 * targhooks.c (default_empty_mask_is_expensive): New function.
9214 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9215 if the target says that empty masks are expensive.
9216 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9218 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9220 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9221 Alan Hayward <alan.hayward@arm.com>
9222 David Sherwood <david.sherwood@arm.com>
9224 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9225 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9226 (vect_use_loop_mask_for_alignment_p): New function.
9227 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9228 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9229 niters_skip argument. Make sure that the first niters_skip elements
9230 of the first iteration are inactive.
9231 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9232 Update call to vect_set_loop_masks_directly.
9233 (get_misalign_in_elems): New function, split out from...
9234 (vect_gen_prolog_loop_niters): ...here.
9235 (vect_update_init_of_dr): Take a code argument that specifies whether
9236 the adjustment should be added or subtracted.
9237 (vect_update_init_of_drs): Likewise.
9238 (vect_prepare_for_masked_peels): New function.
9239 (vect_do_peeling): Skip prologue peeling if we're using a mask
9240 instead. Update call to vect_update_inits_of_drs.
9241 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9243 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9244 alignment. Do not include the number of peeled iterations in
9245 the minimum threshold in that case.
9246 (vectorizable_induction): Adjust the start value down by
9247 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9248 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9249 Take the number of skipped iterations into account when calculating
9251 * tree-vect-stmts.c (vect_gen_while_not): New function.
9253 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9254 Alan Hayward <alan.hayward@arm.com>
9255 David Sherwood <david.sherwood@arm.com>
9257 * doc/sourcebuild.texi (vect_fully_masked): Document.
9258 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9260 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9262 (vect_analyze_loop_2): ...here. Don't check the vectorization
9263 factor against the number of loop iterations if the loop is
9266 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9267 Alan Hayward <alan.hayward@arm.com>
9268 David Sherwood <david.sherwood@arm.com>
9270 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9271 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9272 (dump_groups): Update accordingly.
9273 (iv_use::mem_type): New member variable.
9274 (address_p): New function.
9275 (record_use): Add a mem_type argument and initialize the new
9277 (record_group_use): Add a mem_type argument. Use address_p.
9278 Remove obsolete null checks of base_object. Update call to record_use.
9279 (find_interesting_uses_op): Update call to record_group_use.
9280 (find_interesting_uses_cond): Likewise.
9281 (find_interesting_uses_address): Likewise.
9282 (get_mem_type_for_internal_fn): New function.
9283 (find_address_like_use): Likewise.
9284 (find_interesting_uses_stmt): Try find_address_like_use before
9285 calling find_interesting_uses_op.
9286 (addr_offset_valid_p): Use the iv mem_type field as the type
9287 of the addressed memory.
9288 (add_autoinc_candidates): Likewise.
9289 (get_address_cost): Likewise.
9290 (split_small_address_groups_p): Use address_p.
9291 (split_address_groups): Likewise.
9292 (add_iv_candidate_for_use): Likewise.
9293 (autoinc_possible_for_pair): Likewise.
9294 (rewrite_groups): Likewise.
9295 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9296 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9297 (get_alias_ptr_type_for_ptr_address): New function.
9298 (rewrite_use_address): Rewrite address uses in calls that were
9299 identified by find_address_like_use.
9301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9302 Alan Hayward <alan.hayward@arm.com>
9303 David Sherwood <david.sherwood@arm.com>
9305 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9307 * gimple-expr.h (is_gimple_addressable: Likewise.
9308 * gimple-expr.c (is_gimple_address): Likewise.
9309 * internal-fn.c (expand_call_mem_ref): New function.
9310 (expand_mask_load_optab_fn): Use it.
9311 (expand_mask_store_optab_fn): Likewise.
9313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9314 Alan Hayward <alan.hayward@arm.com>
9315 David Sherwood <david.sherwood@arm.com>
9317 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9318 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9319 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9320 (cond_umax@var{mode}): Document.
9321 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9322 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9323 (cond_umin_optab, cond_umax_optab): New optabs.
9324 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9325 (COND_IOR, COND_XOR): New internal functions.
9326 * internal-fn.h (get_conditional_internal_fn): Declare.
9327 * internal-fn.c (cond_binary_direct): New macro.
9328 (expand_cond_binary_optab_fn): Likewise.
9329 (direct_cond_binary_optab_supported_p): Likewise.
9330 (get_conditional_internal_fn): New function.
9331 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9332 Cope with reduction statements that are vectorized as calls rather
9334 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9335 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9336 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9337 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9338 (UNSPEC_COND_EOR): New unspecs.
9339 (optab): Add mappings for them.
9340 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9341 (sve_int_op, sve_fp_op): New int attributes.
9343 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9344 Alan Hayward <alan.hayward@arm.com>
9345 David Sherwood <david.sherwood@arm.com>
9347 * optabs.def (while_ult_optab): New optab.
9348 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9349 * internal-fn.def (WHILE_ULT): New internal function.
9350 * internal-fn.h (direct_internal_fn_supported_p): New override
9351 that takes two types as argument.
9352 * internal-fn.c (while_direct): New macro.
9353 (expand_while_optab_fn): New function.
9354 (convert_optab_supported_p): Likewise.
9355 (direct_while_optab_supported_p): New macro.
9356 * wide-int.h (wi::udiv_ceil): New function.
9357 * tree-vectorizer.h (rgroup_masks): New structure.
9358 (vec_loop_masks): New typedef.
9359 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9361 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9362 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9363 (vect_max_vf): New function.
9364 (slpeel_make_loop_iterate_ntimes): Delete.
9365 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9366 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9367 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9368 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9369 internal-fn.h, stor-layout.h and optabs-query.h.
9370 (vect_set_loop_mask): New function.
9371 (add_preheader_seq): Likewise.
9372 (add_header_seq): Likewise.
9373 (interleave_supported_p): Likewise.
9374 (vect_maybe_permute_loop_masks): Likewise.
9375 (vect_set_loop_masks_directly): Likewise.
9376 (vect_set_loop_condition_masked): Likewise.
9377 (vect_set_loop_condition_unmasked): New function, split out from
9378 slpeel_make_loop_iterate_ntimes.
9379 (slpeel_make_loop_iterate_ntimes): Rename to..
9380 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9381 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9382 (vect_do_peeling): Update call accordingly.
9383 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9385 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9386 mask_compare_type, can_fully_mask_p and fully_masked_p.
9387 (release_vec_loop_masks): New function.
9388 (_loop_vec_info): Use it to free the loop masks.
9389 (can_produce_all_loop_masks_p): New function.
9390 (vect_get_max_nscalars_per_iter): Likewise.
9391 (vect_verify_full_masking): Likewise.
9392 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9393 retries, and free the mask rgroups before retrying. Check loop-wide
9394 reasons for disallowing fully-masked loops. Make the final decision
9395 about whether use a fully-masked loop or not.
9396 (vect_estimate_min_profitable_iters): Do not assume that peeling
9397 for the number of iterations will be needed for fully-masked loops.
9398 (vectorizable_reduction): Disable fully-masked loops.
9399 (vectorizable_live_operation): Likewise.
9400 (vect_halve_mask_nunits): New function.
9401 (vect_double_mask_nunits): Likewise.
9402 (vect_record_loop_mask): Likewise.
9403 (vect_get_loop_mask): Likewise.
9404 (vect_transform_loop): Handle the case in which the final loop
9405 iteration might handle a partial vector. Call vect_set_loop_condition
9406 instead of slpeel_make_loop_iterate_ntimes.
9407 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9408 (check_load_store_masking): New function.
9409 (prepare_load_store_mask): Likewise.
9410 (vectorizable_store): Handle fully-masked loops.
9411 (vectorizable_load): Likewise.
9412 (supportable_widening_operation): Use vect_halve_mask_nunits for
9414 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9415 (vect_gen_while): New function.
9416 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9417 (aarch64_uqdec<mode>): New insn.
9419 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9420 Alan Hayward <alan.hayward@arm.com>
9421 David Sherwood <david.sherwood@arm.com>
9423 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9424 (reduc_xor_scal_optab): New optabs.
9425 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9426 (reduc_xor_scal_@var{m}): Document.
9427 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9428 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9430 * fold-const-call.c (fold_const_call): Handle them.
9431 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9432 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9433 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9434 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9435 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9436 (UNSPEC_XORV): New unspecs.
9437 (optab): Add entries for them.
9438 (BITWISEV): New int iterator.
9439 (bit_reduc_op): New int attributes.
9441 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9442 Alan Hayward <alan.hayward@arm.com>
9443 David Sherwood <david.sherwood@arm.com>
9445 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9446 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9447 * optabs.def (vec_shl_insert_optab): New optab.
9448 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9449 (duplicate_and_interleave): Likewise.
9450 * tree-vect-loop.c: Include internal-fn.h.
9451 (neutral_op_for_slp_reduction): New function, split out from
9452 get_initial_defs_for_reduction.
9453 (get_initial_def_for_reduction): Handle option 2 for variable-length
9454 vectors by loading the neutral value into a vector and then shifting
9455 the initial value into element 0.
9456 (get_initial_defs_for_reduction): Replace the code argument with
9457 the neutral value calculated by neutral_op_for_slp_reduction.
9458 Use gimple_build_vector for constant-length vectors.
9459 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9460 but the first group_size elements have a neutral value.
9461 Use duplicate_and_interleave otherwise.
9462 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9463 Update call to get_initial_defs_for_reduction. Handle SLP
9464 reductions for variable-length vectors by creating one vector
9465 result for each scalar result, with the elements associated
9466 with other scalar results stubbed out with the neutral value.
9467 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9468 Require IFN_VEC_SHL_INSERT for double reductions on
9469 variable-length vectors, or SLP reductions that have
9470 a neutral value. Require can_duplicate_and_interleave_p
9471 support for variable-length unchained SLP reductions if there
9472 is no neutral value, such as for MIN/MAX reductions. Also require
9473 the number of vector elements to be a multiple of the number of
9474 SLP statements when doing variable-length unchained SLP reductions.
9475 Update call to vect_create_epilog_for_reduction.
9476 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9477 and remove initial values.
9478 (duplicate_and_interleave): Make public.
9479 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9480 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9482 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9483 Alan Hayward <alan.hayward@arm.com>
9484 David Sherwood <david.sherwood@arm.com>
9486 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9487 (can_duplicate_and_interleave_p): New function.
9488 (vect_get_and_check_slp_defs): Take the vector of statements
9489 rather than just the current one. Remove excess parentheses.
9490 Restriction rejectinon of vect_constant_def and vect_external_def
9491 for variable-length vectors to boolean types, or types for which
9492 can_duplicate_and_interleave_p is false.
9493 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9494 (duplicate_and_interleave): New function.
9495 (vect_get_constant_vectors): Use gimple_build_vector for
9496 constant-length vectors and suitable variable-length constant
9497 vectors. Use duplicate_and_interleave for other variable-length
9498 vectors. Don't defer the update when inserting new statements.
9500 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9501 Alan Hayward <alan.hayward@arm.com>
9502 David Sherwood <david.sherwood@arm.com>
9504 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9505 min_profitable_iters doesn't go negative.
9507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9508 Alan Hayward <alan.hayward@arm.com>
9509 David Sherwood <david.sherwood@arm.com>
9511 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9512 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9513 * optabs.def (vec_mask_load_lanes_optab): New optab.
9514 (vec_mask_store_lanes_optab): Likewise.
9515 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9516 (MASK_STORE_LANES): Likewise.
9517 * internal-fn.c (mask_load_lanes_direct): New macro.
9518 (mask_store_lanes_direct): Likewise.
9519 (expand_mask_load_optab_fn): Handle masked operations.
9520 (expand_mask_load_lanes_optab_fn): New macro.
9521 (expand_mask_store_optab_fn): Handle masked operations.
9522 (expand_mask_store_lanes_optab_fn): New macro.
9523 (direct_mask_load_lanes_optab_supported_p): Likewise.
9524 (direct_mask_store_lanes_optab_supported_p): Likewise.
9525 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9527 (vect_load_lanes_supported): Likewise.
9528 * tree-vect-data-refs.c (strip_conversion): New function.
9529 (can_group_stmts_p): Likewise.
9530 (vect_analyze_data_ref_accesses): Use it instead of checking
9531 for a pair of assignments.
9532 (vect_store_lanes_supported): Take a masked_p parameter.
9533 (vect_load_lanes_supported): Likewise.
9534 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9535 vect_store_lanes_supported and vect_load_lanes_supported.
9536 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9537 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9538 parameter. Don't allow gaps for masked accesses.
9539 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9540 and vect_load_lanes_supported.
9541 (get_load_store_type): Take a masked_p parameter and update
9542 call to get_group_load_store_type.
9543 (vectorizable_store): Update call to get_load_store_type.
9544 Handle IFN_MASK_STORE_LANES.
9545 (vectorizable_load): Update call to get_load_store_type.
9546 Handle IFN_MASK_LOAD_LANES.
9548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9549 Alan Hayward <alan.hayward@arm.com>
9550 David Sherwood <david.sherwood@arm.com>
9552 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9554 * config/aarch64/aarch64-protos.h
9555 (aarch64_sve_struct_memory_operand_p): Declare.
9556 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9557 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9558 (VPRED, vpred): Handle SVE structure modes.
9559 * config/aarch64/constraints.md (Utx): New constraint.
9560 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9561 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9562 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9563 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9564 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9565 structure modes. Split into pieces after RA.
9566 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9567 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9569 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9570 SVE structure modes.
9571 (aarch64_classify_address): Likewise.
9572 (sizetochar): Move earlier in file.
9573 (aarch64_print_operand): Handle SVE register lists.
9574 (aarch64_array_mode): New function.
9575 (aarch64_sve_struct_memory_operand_p): Likewise.
9576 (TARGET_ARRAY_MODE): Redefine.
9578 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9579 Alan Hayward <alan.hayward@arm.com>
9580 David Sherwood <david.sherwood@arm.com>
9582 * target.def (array_mode): New target hook.
9583 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9584 * doc/tm.texi: Regenerate.
9585 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9586 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9587 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9589 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9592 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9593 Alan Hayward <alan.hayward@arm.com>
9594 David Sherwood <david.sherwood@arm.com>
9596 * fold-const.c (fold_binary_loc): Check the argument types
9597 rather than the result type when testing for a vector operation.
9599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9601 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9602 * doc/tm.texi: Regenerate.
9604 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9605 Alan Hayward <alan.hayward@arm.com>
9606 David Sherwood <david.sherwood@arm.com>
9608 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9609 (sve): Document new AArch64 extension.
9610 * doc/md.texi (w): Extend the description of the AArch64
9611 constraint to include SVE vectors.
9612 (Upl, Upa): Document new AArch64 predicate constraints.
9613 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9615 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9616 (msve-vector-bits=): New option.
9617 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9618 SVE when these are disabled.
9619 (sve): New extension.
9620 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9621 modes. Adjust their number of units based on aarch64_sve_vg.
9622 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9623 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9624 aarch64_addr_query_type.
9625 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9626 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9627 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9628 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9629 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9630 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9631 (aarch64_simd_imm_zero_p): Delete.
9632 (aarch64_check_zero_based_sve_index_immediate): Declare.
9633 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9634 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9635 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9636 (aarch64_sve_float_mul_immediate_p): Likewise.
9637 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9639 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9640 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9641 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9642 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9643 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9644 (aarch64_regmode_natural_size): Likewise.
9645 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9646 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9648 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9649 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9650 for VG and the SVE predicate registers.
9651 (V_ALIASES): Add a "z"-prefixed alias.
9652 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9653 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9654 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9655 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9656 (REG_CLASS_NAMES): Add entries for them.
9657 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9658 and the predicate registers.
9659 (aarch64_sve_vg): Declare.
9660 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9661 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9662 (REGMODE_NATURAL_SIZE): Define.
9663 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9665 * config/aarch64/aarch64.c: Include cfgrtl.h.
9666 (simd_immediate_info): Add a constructor for series vectors,
9667 and an associated step field.
9668 (aarch64_sve_vg): New variable.
9669 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9670 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9671 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9672 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9673 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9674 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9675 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9676 (aarch64_get_mask_mode): New functions.
9677 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9678 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9679 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9680 predicate modes and predicate registers. Explicitly restrict
9681 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9682 to store a vector mode if it is recognized by
9683 aarch64_classify_vector_mode.
9684 (aarch64_regmode_natural_size): New function.
9685 (aarch64_hard_regno_caller_save_mode): Return the original mode
9687 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9688 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9689 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9690 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9692 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9693 does not overlap dest if the function is frame-related. Handle
9695 (aarch64_split_add_offset): New function.
9696 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9697 them aarch64_add_offset.
9698 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9699 and update call to aarch64_sub_sp.
9700 (aarch64_add_cfa_expression): New function.
9701 (aarch64_expand_prologue): Pass extra temporary registers to the
9702 functions above. Handle the case in which we need to emit new
9703 DW_CFA_expressions for registers that were originally saved
9704 relative to the stack pointer, but now have to be expressed
9705 relative to the frame pointer.
9706 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9708 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9709 IP0 and IP1 values for SVE frames.
9710 (aarch64_expand_vec_series): New function.
9711 (aarch64_expand_sve_widened_duplicate): Likewise.
9712 (aarch64_expand_sve_const_vector): Likewise.
9713 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9714 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9715 into the register, rather than emitting a SET directly.
9716 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9717 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9718 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9719 (offset_9bit_signed_scaled_p): New functions.
9720 (aarch64_replicate_bitmask_imm): New function.
9721 (aarch64_bitmask_imm): Use it.
9722 (aarch64_cannot_force_const_mem): Reject expressions involving
9723 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9724 (aarch64_classify_index): Handle SVE indices, by requiring
9725 a plain register index with a scale that matches the element size.
9726 (aarch64_classify_address): Handle SVE addresses. Assert that
9727 the mode of the address is VOIDmode or an integer mode.
9728 Update call to aarch64_classify_symbol.
9729 (aarch64_classify_symbolic_expression): Update call to
9730 aarch64_classify_symbol.
9731 (aarch64_const_vec_all_in_range_p): New function.
9732 (aarch64_print_vector_float_operand): Likewise.
9733 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9734 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9735 and the FP immediates 1.0 and 0.5.
9736 (aarch64_print_address_internal): Handle SVE addresses.
9737 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9738 (aarch64_regno_regclass): Handle predicate registers.
9739 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9741 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9742 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9743 (aarch64_convert_sve_vector_bits): New function.
9744 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9745 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9747 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9748 Handle SVE vector and predicate modes. Accept VL-based constants
9749 that need only one temporary register, and VL offsets that require
9750 no temporary registers.
9751 (aarch64_conditional_register_usage): Mark the predicate registers
9752 as fixed if SVE isn't available.
9753 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9754 Return true for SVE vector and predicate modes.
9755 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9756 rather than an unsigned int. Handle SVE modes.
9757 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9759 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9761 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9762 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9763 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9764 (aarch64_sve_float_mul_immediate_p): New functions.
9765 (aarch64_sve_valid_immediate): New function.
9766 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9767 Explicitly reject structure modes. Check for INDEX constants.
9768 Handle PTRUE and PFALSE constants.
9769 (aarch64_check_zero_based_sve_index_immediate): New function.
9770 (aarch64_simd_imm_zero_p): Delete.
9771 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9772 vector modes. Accept constants in the range of CNT[BHWD].
9773 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9774 ask for an Advanced SIMD mode.
9775 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9776 (aarch64_simd_vector_alignment): Handle SVE predicates.
9777 (aarch64_vectorize_preferred_vector_alignment): New function.
9778 (aarch64_simd_vector_alignment_reachable): Use it instead of
9780 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9781 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9783 (MAX_VECT_LEN): Delete.
9784 (expand_vec_perm_d): Add a vec_flags field.
9785 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9786 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9787 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9789 (aarch64_evpc_rev): Rename to...
9790 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9791 (aarch64_evpc_rev_global): New function.
9792 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9793 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9795 (aarch64_evpc_sve_tbl): New function.
9796 (aarch64_expand_vec_perm_const_1): Update after rename of
9797 aarch64_evpc_rev. Handle SVE permutes too, trying
9798 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9799 than aarch64_evpc_tbl.
9800 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9801 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9802 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9803 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9804 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9805 (aarch64_expand_sve_vcond): New functions.
9806 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9807 of aarch64_vector_mode_p.
9808 (aarch64_dwarf_poly_indeterminate_value): New function.
9809 (aarch64_compute_pressure_classes): Likewise.
9810 (aarch64_can_change_mode_class): Likewise.
9811 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9812 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9813 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9814 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9815 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9816 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9817 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9818 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9820 (Dn, Dl, Dr): Accept const as well as const_vector.
9821 (Dz): Likewise. Compare against CONST0_RTX.
9822 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9823 of "vector" where appropriate.
9824 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9825 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9826 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9827 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9828 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9829 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9830 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9831 (v_int_equiv): Extend to SVE modes.
9832 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9834 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9835 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9836 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9837 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9838 (SVE_COND_FP_CMP): New int iterators.
9839 (perm_hilo): Handle the new unpack unspecs.
9840 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9842 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9843 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9844 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9845 (aarch64_equality_operator, aarch64_constant_vector_operand)
9846 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9847 (aarch64_sve_nonimmediate_operand): Likewise.
9848 (aarch64_sve_general_operand): Likewise.
9849 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9850 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9851 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9852 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9853 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9854 (aarch64_sve_float_arith_immediate): Likewise.
9855 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9856 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9857 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9858 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9859 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9860 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9861 (aarch64_sve_float_arith_operand): Likewise.
9862 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9863 (aarch64_sve_float_mul_operand): Likewise.
9864 (aarch64_sve_vec_perm_operand): Likewise.
9865 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9866 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9867 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9868 as well as const_vector.
9869 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9870 in file. Use CONST0_RTX and CONSTM1_RTX.
9871 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9872 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9873 Use aarch64_simd_imm_zero.
9874 * config/aarch64/aarch64-sve.md: New file.
9875 * config/aarch64/aarch64.md: Include it.
9876 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9877 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9878 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9879 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9880 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9881 (sve): New attribute.
9882 (enabled): Disable instructions with the sve attribute unless
9884 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9885 aarch64_expand_mov_immediate.
9886 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9887 CNT[BHSD] immediates.
9888 (movti): Split CONST_POLY_INT moves into two halves.
9889 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9890 Split additions that need a temporary here if the destination
9891 is the stack pointer.
9892 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9893 (*add<mode>3_poly_1): New instruction.
9894 (set_clobber_cc): New expander.
9896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9898 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9899 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9900 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9901 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9902 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9903 Change innermode from fixed_mode_size to machine_mode.
9904 (simplify_subreg): Update call accordingly. Handle a constant-sized
9905 subreg of a variable-length CONST_VECTOR.
9907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9908 Alan Hayward <alan.hayward@arm.com>
9909 David Sherwood <david.sherwood@arm.com>
9911 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9912 (add_offset_to_base): New function, split out from...
9913 (create_mem_ref): ...here. When handling a scale other than 1,
9914 check first whether the address is valid without the offset.
9915 Add it into the base if so, leaving the index and scale as-is.
9917 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9920 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9921 fold_for_warn before checking if arg2 is INTEGER_CST.
9923 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9925 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9926 (store_multiple_operation): Delete.
9927 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9928 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9929 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9930 guarded by TARGET_STRING.
9931 (rs6000_output_load_multiple): Delete.
9932 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9933 OPTION_MASK_STRING / TARGET_STRING handling.
9934 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9935 (const rs6000_opt_masks) <"string">: Change mask to 0.
9936 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9937 (MASK_STRING): Delete.
9938 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9940 (load_multiple): Delete.
9947 (store_multiple): Delete.
9954 (movmemsi_8reg): Delete.
9955 (corresponding unnamed define_insn): Delete.
9956 (movmemsi_6reg): Delete.
9957 (corresponding unnamed define_insn): Delete.
9958 (movmemsi_4reg): Delete.
9959 (corresponding unnamed define_insn): Delete.
9960 (movmemsi_2reg): Delete.
9961 (corresponding unnamed define_insn): Delete.
9962 (movmemsi_1reg): Delete.
9963 (corresponding unnamed define_insn): Delete.
9964 * config/rs6000/rs6000.opt (mno-string): New.
9965 (mstring): Replace by deprecation warning stub.
9966 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9968 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9970 * regrename.c (regrename_do_replace): If replacing the same
9971 reg multiple times, try to reuse last created gen_raw_REG.
9974 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9975 main to workaround a bug in GDB.
9977 2018-01-12 Tom de Vries <tom@codesourcery.com>
9980 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9982 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9984 PR rtl-optimization/80481
9985 * ira-color.c (get_cap_member): New function.
9986 (allocnos_conflict_by_live_ranges_p): Use it.
9987 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9988 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9990 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9993 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9994 (*saddl_se_1): Ditto.
9996 (*ssubl_se_1): Ditto.
9998 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10000 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10001 rather than wi::to_widest for DR_INITs.
10002 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10003 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10004 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10006 (vect_analyze_group_access_1): Note that here.
10008 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10010 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10011 polynomial type sizes.
10013 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10015 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10016 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10017 (gimple_add_tmp_var): Likewise.
10019 2018-01-12 Martin Liska <mliska@suse.cz>
10021 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10022 (gimple_alloc_sizes): Likewise.
10023 (dump_gimple_statistics): Use PRIu64 in printf format.
10024 * gimple.h: Change uint64_t to int.
10026 2018-01-12 Martin Liska <mliska@suse.cz>
10028 * tree-core.h: Use uint64_t instead of int.
10029 * tree.c (tree_node_counts): Likewise.
10030 (tree_node_sizes): Likewise.
10031 (dump_tree_statistics): Use PRIu64 in printf format.
10033 2018-01-12 Martin Liska <mliska@suse.cz>
10035 * Makefile.in: As qsort_chk is implemented in vec.c, add
10036 vec.o to linkage of gencfn-macros.
10037 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10038 passing the info to record_node_allocation_statistics.
10039 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10041 * ggc-common.c (struct ggc_usage): Add operator== and use
10042 it in operator< and compare function.
10043 * mem-stats.h (struct mem_usage): Likewise.
10044 * vec.c (struct vec_usage): Remove operator< and compare
10045 function. Can be simply inherited.
10047 2018-01-12 Martin Jambor <mjambor@suse.cz>
10050 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10051 * tree-ssa-math-opts.c: Include domwalk.h.
10052 (convert_mult_to_fma_1): New function.
10053 (fma_transformation_info): New type.
10054 (fma_deferring_state): Likewise.
10055 (cancel_fma_deferring): New function.
10056 (result_of_phi): Likewise.
10057 (last_fma_candidate_feeds_initial_phi): Likewise.
10058 (convert_mult_to_fma): Added deferring logic, split actual
10059 transformation to convert_mult_to_fma_1.
10060 (math_opts_dom_walker): New type.
10061 (math_opts_dom_walker::after_dom_children): New method, body moved
10062 here from pass_optimize_widening_mul::execute, added deferring logic
10064 (pass_optimize_widening_mul::execute): Moved most of code to
10065 math_opts_dom_walker::after_dom_children.
10066 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10067 * config/i386/i386.c (ix86_option_override_internal): Added
10068 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10070 2018-01-12 Richard Biener <rguenther@suse.de>
10073 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10074 inline instance vars.
10076 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10079 * config/rx/rx.c (rx_is_restricted_memory_address):
10080 Handle SUBREG case.
10082 2018-01-12 Richard Biener <rguenther@suse.de>
10084 PR tree-optimization/80846
10085 * target.def (split_reduction): New target hook.
10086 * targhooks.c (default_split_reduction): New function.
10087 * targhooks.h (default_split_reduction): Declare.
10088 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10089 target requests first reduce vectors by combining low and high
10091 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10092 (get_vectype_for_scalar_type_and_size): Export.
10093 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10094 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10095 * doc/tm.texi: Regenerate.
10096 * config/i386/i386.c (ix86_split_reduction): Implement
10097 TARGET_VECTORIZE_SPLIT_REDUCTION.
10099 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10102 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10103 in PIC mode except for TARGET_VXWORKS_RTP.
10104 * config/sparc/sparc.c: Include cfgrtl.h.
10105 (TARGET_INIT_PIC_REG): Define.
10106 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10107 (sparc_pic_register_p): New predicate.
10108 (sparc_legitimate_address_p): Use it.
10109 (sparc_legitimize_pic_address): Likewise.
10110 (sparc_delegitimize_address): Likewise.
10111 (sparc_mode_dependent_address_p): Likewise.
10112 (gen_load_pcrel_sym): Remove 4th parameter.
10113 (load_got_register): Adjust call to above. Remove obsolete stuff.
10114 (sparc_expand_prologue): Do not call load_got_register here.
10115 (sparc_flat_expand_prologue): Likewise.
10116 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10117 (sparc_use_pseudo_pic_reg): New function.
10118 (sparc_init_pic_reg): Likewise.
10119 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10120 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10122 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10124 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10125 Add item for branch_cost.
10127 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10129 PR rtl-optimization/83565
10130 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10131 not extend the result to a larger mode for rotate operations.
10132 (num_sign_bit_copies1): Likewise.
10134 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10137 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10139 Use values-Xc.o for -pedantic.
10140 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10142 2018-01-12 Martin Liska <mliska@suse.cz>
10145 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10147 (possible_polymorphic_call_targets): Use it.
10148 (ipa_devirt): Likewise.
10150 2018-01-12 Martin Liska <mliska@suse.cz>
10152 * profile-count.h (enum profile_quality): Use 0 as invalid
10153 enum value of profile_quality.
10155 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10157 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10158 -mext-string options.
10160 2018-01-12 Richard Biener <rguenther@suse.de>
10162 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10163 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10164 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10166 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10168 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10170 * configure.ac (--with-long-double-format): Add support for the
10171 configuration option to change the default long double format on
10173 * config.gcc (powerpc*-linux*-*): Likewise.
10174 * configure: Regenerate.
10175 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10176 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10177 used without modification.
10179 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10181 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10182 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10183 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10184 MISC_BUILTIN_SPEC_BARRIER.
10185 (rs6000_init_builtins): Likewise.
10186 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10188 (speculation_barrier): New define_insn.
10189 * doc/extend.texi: Document __builtin_speculation_barrier.
10191 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10194 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10195 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10196 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10198 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10199 integral modes instead of "ss" and "sd".
10200 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10201 vectors with 32-bit and 64-bit elements.
10202 (vecdupssescalarmodesuffix): New mode attribute.
10203 (vec_dup<mode>): Use it.
10205 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10208 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10209 frame if argument is passed on stack.
10211 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10214 * ree.c (combine_reaching_defs): Optimize also
10215 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10216 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10218 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10220 PR middle-end/83189
10221 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10223 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10225 PR middle-end/83718
10226 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10227 after they are computed.
10229 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10231 PR tree-optimization/83695
10232 * gimple-loop-linterchange.cc
10233 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10234 reset cached scev information after interchange.
10235 (pass_linterchange::execute): Remove call to scev_reset_htab.
10237 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10239 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10240 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10241 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10242 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10243 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10244 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10245 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10246 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10247 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10248 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10249 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10250 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10251 (V_lane_reg): Likewise.
10252 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10254 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10255 (vfmal_lane_low<mode>_intrinsic,
10256 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10257 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10258 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10259 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10260 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10261 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10263 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10265 * config/arm/arm-cpus.in (fp16fml): New feature.
10266 (ALL_SIMD): Add fp16fml.
10267 (armv8.2-a): Add fp16fml as an option.
10268 (armv8.3-a): Likewise.
10269 (armv8.4-a): Add fp16fml as part of fp16.
10270 * config/arm/arm.h (TARGET_FP16FML): Define.
10271 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10273 * config/arm/arm-modes.def (V2HF): Define.
10274 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10275 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10276 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10277 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10278 vfmsl_low, vfmsl_high): New set of builtins.
10279 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10280 (vfml_op): New code attribute.
10281 (VFMLHALVES): New int iterator.
10282 (VFML, VFMLSEL): New mode attributes.
10283 (V_reg): Define mapping for V2HF.
10284 (V_hi, V_lo): New mode attributes.
10285 (VF_constraint): Likewise.
10286 (vfml_half, vfml_half_selector): New int attributes.
10287 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10289 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10290 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10292 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10293 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10294 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10295 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10297 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10298 Document new effective target and option set.
10300 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10302 * config/arm/arm-cpus.in (armv8_4): New feature.
10303 (ARMv8_4a): New fgroup.
10304 (armv8.4-a): New arch.
10305 * config/arm/arm-tables.opt: Regenerate.
10306 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10307 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10308 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10309 Add matching rules for -march=armv8.4-a and extensions.
10310 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10312 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10315 * config/rx/rx.md (BW): New mode attribute.
10316 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10318 2018-01-11 Richard Biener <rguenther@suse.de>
10320 PR tree-optimization/83435
10321 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10322 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10323 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10325 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10326 Alan Hayward <alan.hayward@arm.com>
10327 David Sherwood <david.sherwood@arm.com>
10329 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10331 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10332 (aarch64_print_address_internal): Use it to check for a zero offset.
10334 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10335 Alan Hayward <alan.hayward@arm.com>
10336 David Sherwood <david.sherwood@arm.com>
10338 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10339 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10340 Return a poly_int64 rather than a HOST_WIDE_INT.
10341 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10342 rather than a HOST_WIDE_INT.
10343 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10344 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10345 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10346 final_offset from HOST_WIDE_INT to poly_int64.
10347 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10348 to_constant when getting the number of units in an Advanced SIMD
10350 (aarch64_builtin_vectorized_function): Check for a constant number
10352 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10354 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10355 attribute instead of GET_MODE_NUNITS.
10356 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10357 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10358 GET_MODE_SIZE for fixed-size registers.
10359 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10360 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10361 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10362 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10363 (aarch64_print_operand, aarch64_print_address_internal)
10364 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10365 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10366 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10367 Handle polynomial GET_MODE_SIZE.
10368 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10369 wider than SImode without modification.
10370 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10371 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10372 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10373 passing and returning SVE modes.
10374 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10375 rather than GEN_INT.
10376 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10377 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10378 (aarch64_allocate_and_probe_stack_space): Likewise.
10379 (aarch64_layout_frame): Cope with polynomial offsets.
10380 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10381 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10382 polynomial offsets.
10383 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10384 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10385 poly_int64 rather than a HOST_WIDE_INT.
10386 (aarch64_get_separate_components, aarch64_process_components)
10387 (aarch64_expand_prologue, aarch64_expand_epilogue)
10388 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10389 (aarch64_anchor_offset): New function, split out from...
10390 (aarch64_legitimize_address): ...here.
10391 (aarch64_builtin_vectorization_cost): Handle polynomial
10392 TYPE_VECTOR_SUBPARTS.
10393 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10395 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10396 number of elements from the PARALLEL rather than the mode.
10397 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10398 rather than GET_MODE_BITSIZE.
10399 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10400 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10401 (aarch64_expand_vec_perm_const_1): Handle polynomial
10402 d->perm.length () and d->perm elements.
10403 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10404 Apply to_constant to d->perm elements.
10405 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10406 polynomial CONST_VECTOR_NUNITS.
10407 (aarch64_move_pointer): Take amount as a poly_int64 rather
10409 (aarch64_progress_pointer): Avoid temporary variable.
10410 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10411 the mode attribute instead of GET_MODE.
10413 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10414 Alan Hayward <alan.hayward@arm.com>
10415 David Sherwood <david.sherwood@arm.com>
10417 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10418 x exists before using it.
10419 (aarch64_add_constant_internal): Rename to...
10420 (aarch64_add_offset_1): ...this. Replace regnum with separate
10421 src and dest rtxes. Handle the case in which they're different,
10422 including when the offset is zero. Replace scratchreg with an rtx.
10423 Use 2 additions if there is no spare register into which we can
10424 move a 16-bit constant.
10425 (aarch64_add_constant): Delete.
10426 (aarch64_add_offset): Replace reg with separate src and dest
10427 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10428 Use aarch64_add_offset_1.
10429 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10430 an rtx rather than an int. Take the delta as a poly_int64
10431 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10432 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10433 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10434 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10435 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10436 and aarch64_add_sp.
10437 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10438 aarch64_add_constant.
10440 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10442 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10443 Use scalar_float_mode.
10445 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10447 * config/aarch64/aarch64-simd.md
10448 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10449 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10450 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10451 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10452 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10453 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10454 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10455 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10456 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10457 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10459 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10462 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10463 targ_options->x_arm_arch_string is non NULL.
10465 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10467 * config/aarch64/aarch64.h
10468 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10470 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10473 * expmed.c (emit_store_flag_force): Swap if const op0
10474 and change VOIDmode to mode of op0.
10476 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10478 PR rtl-optimization/83761
10479 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10480 than bytes to mode_for_size.
10482 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10484 PR middle-end/83189
10485 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10486 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10489 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10491 PR middle-end/83575
10492 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10493 when in layout mode.
10494 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10495 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10498 2018-01-10 Michael Collison <michael.collison@arm.com>
10500 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10501 * config/aarch64/aarch64-option-extension.def: Add
10502 AARCH64_OPT_EXTENSION of 'fp16fml'.
10503 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10504 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10505 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10506 * config/aarch64/constraints.md (Ui7): New constraint.
10507 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10508 (VFMLA_SEL_W): Ditto.
10511 (VFMLA16_LOW): New int iterator.
10512 (VFMLA16_HIGH): Ditto.
10513 (UNSPEC_FMLAL): New unspec.
10514 (UNSPEC_FMLSL): Ditto.
10515 (UNSPEC_FMLAL2): Ditto.
10516 (UNSPEC_FMLSL2): Ditto.
10517 (f16mac): New code attribute.
10518 * config/aarch64/aarch64-simd-builtins.def
10519 (aarch64_fmlal_lowv2sf): Ditto.
10520 (aarch64_fmlsl_lowv2sf): Ditto.
10521 (aarch64_fmlalq_lowv4sf): Ditto.
10522 (aarch64_fmlslq_lowv4sf): Ditto.
10523 (aarch64_fmlal_highv2sf): Ditto.
10524 (aarch64_fmlsl_highv2sf): Ditto.
10525 (aarch64_fmlalq_highv4sf): Ditto.
10526 (aarch64_fmlslq_highv4sf): Ditto.
10527 (aarch64_fmlal_lane_lowv2sf): Ditto.
10528 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10529 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10530 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10531 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10532 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10533 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10534 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10535 (aarch64_fmlal_lane_highv2sf): Ditto.
10536 (aarch64_fmlsl_lane_highv2sf): Ditto.
10537 (aarch64_fmlal_laneq_highv2sf): Ditto.
10538 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10539 (aarch64_fmlalq_lane_highv4sf): Ditto.
10540 (aarch64_fmlsl_lane_highv4sf): Ditto.
10541 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10542 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10543 * config/aarch64/aarch64-simd.md:
10544 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10545 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10546 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10547 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10548 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10549 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10550 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10551 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10552 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10553 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10554 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10555 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10556 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10557 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10558 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10559 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10560 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10561 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10562 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10563 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10564 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10565 (vfmlsl_low_u32): Ditto.
10566 (vfmlalq_low_u32): Ditto.
10567 (vfmlslq_low_u32): Ditto.
10568 (vfmlal_high_u32): Ditto.
10569 (vfmlsl_high_u32): Ditto.
10570 (vfmlalq_high_u32): Ditto.
10571 (vfmlslq_high_u32): Ditto.
10572 (vfmlal_lane_low_u32): Ditto.
10573 (vfmlsl_lane_low_u32): Ditto.
10574 (vfmlal_laneq_low_u32): Ditto.
10575 (vfmlsl_laneq_low_u32): Ditto.
10576 (vfmlalq_lane_low_u32): Ditto.
10577 (vfmlslq_lane_low_u32): Ditto.
10578 (vfmlalq_laneq_low_u32): Ditto.
10579 (vfmlslq_laneq_low_u32): Ditto.
10580 (vfmlal_lane_high_u32): Ditto.
10581 (vfmlsl_lane_high_u32): Ditto.
10582 (vfmlal_laneq_high_u32): Ditto.
10583 (vfmlsl_laneq_high_u32): Ditto.
10584 (vfmlalq_lane_high_u32): Ditto.
10585 (vfmlslq_lane_high_u32): Ditto.
10586 (vfmlalq_laneq_high_u32): Ditto.
10587 (vfmlslq_laneq_high_u32): Ditto.
10588 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10589 (AARCH64_FL_FOR_ARCH8_4): New.
10590 (AARCH64_ISA_F16FML): New ISA flag.
10591 (TARGET_F16FML): New feature flag for fp16fml.
10592 (doc/invoke.texi): Document new fp16fml option.
10594 2018-01-10 Michael Collison <michael.collison@arm.com>
10596 * config/aarch64/aarch64-builtins.c:
10597 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10598 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10599 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10600 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10601 (AARCH64_ISA_SHA3): New ISA flag.
10602 (TARGET_SHA3): New feature flag for sha3.
10603 * config/aarch64/iterators.md (sha512_op): New int attribute.
10604 (CRYPTO_SHA512): New int iterator.
10605 (UNSPEC_SHA512H): New unspec.
10606 (UNSPEC_SHA512H2): Ditto.
10607 (UNSPEC_SHA512SU0): Ditto.
10608 (UNSPEC_SHA512SU1): Ditto.
10609 * config/aarch64/aarch64-simd-builtins.def
10610 (aarch64_crypto_sha512hqv2di): New builtin.
10611 (aarch64_crypto_sha512h2qv2di): Ditto.
10612 (aarch64_crypto_sha512su0qv2di): Ditto.
10613 (aarch64_crypto_sha512su1qv2di): Ditto.
10614 (aarch64_eor3qv8hi): Ditto.
10615 (aarch64_rax1qv2di): Ditto.
10616 (aarch64_xarqv2di): Ditto.
10617 (aarch64_bcaxqv8hi): Ditto.
10618 * config/aarch64/aarch64-simd.md:
10619 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10620 (aarch64_crypto_sha512su0qv2di): Ditto.
10621 (aarch64_crypto_sha512su1qv2di): Ditto.
10622 (aarch64_eor3qv8hi): Ditto.
10623 (aarch64_rax1qv2di): Ditto.
10624 (aarch64_xarqv2di): Ditto.
10625 (aarch64_bcaxqv8hi): Ditto.
10626 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10627 (vsha512h2q_u64): Ditto.
10628 (vsha512su0q_u64): Ditto.
10629 (vsha512su1q_u64): Ditto.
10630 (veor3q_u16): Ditto.
10631 (vrax1q_u64): Ditto.
10632 (vxarq_u64): Ditto.
10633 (vbcaxq_u16): Ditto.
10634 * config/arm/types.md (crypto_sha512): New type attribute.
10635 (crypto_sha3): Ditto.
10636 (doc/invoke.texi): Document new sha3 option.
10638 2018-01-10 Michael Collison <michael.collison@arm.com>
10640 * config/aarch64/aarch64-builtins.c:
10641 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10642 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10643 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10644 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10645 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10646 (AARCH64_ISA_SM4): New ISA flag.
10647 (TARGET_SM4): New feature flag for sm4.
10648 * config/aarch64/aarch64-simd-builtins.def
10649 (aarch64_sm3ss1qv4si): Ditto.
10650 (aarch64_sm3tt1aq4si): Ditto.
10651 (aarch64_sm3tt1bq4si): Ditto.
10652 (aarch64_sm3tt2aq4si): Ditto.
10653 (aarch64_sm3tt2bq4si): Ditto.
10654 (aarch64_sm3partw1qv4si): Ditto.
10655 (aarch64_sm3partw2qv4si): Ditto.
10656 (aarch64_sm4eqv4si): Ditto.
10657 (aarch64_sm4ekeyqv4si): Ditto.
10658 * config/aarch64/aarch64-simd.md:
10659 (aarch64_sm3ss1qv4si): Ditto.
10660 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10661 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10662 (aarch64_sm4eqv4si): Ditto.
10663 (aarch64_sm4ekeyqv4si): Ditto.
10664 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10665 (sm3part_op): Ditto.
10666 (CRYPTO_SM3TT): Ditto.
10667 (CRYPTO_SM3PART): Ditto.
10668 (UNSPEC_SM3SS1): New unspec.
10669 (UNSPEC_SM3TT1A): Ditto.
10670 (UNSPEC_SM3TT1B): Ditto.
10671 (UNSPEC_SM3TT2A): Ditto.
10672 (UNSPEC_SM3TT2B): Ditto.
10673 (UNSPEC_SM3PARTW1): Ditto.
10674 (UNSPEC_SM3PARTW2): Ditto.
10675 (UNSPEC_SM4E): Ditto.
10676 (UNSPEC_SM4EKEY): Ditto.
10677 * config/aarch64/constraints.md (Ui2): New constraint.
10678 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10679 * config/arm/types.md (crypto_sm3): New type attribute.
10680 (crypto_sm4): Ditto.
10681 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10682 (vsm3tt1aq_u32): Ditto.
10683 (vsm3tt1bq_u32): Ditto.
10684 (vsm3tt2aq_u32): Ditto.
10685 (vsm3tt2bq_u32): Ditto.
10686 (vsm3partw1q_u32): Ditto.
10687 (vsm3partw2q_u32): Ditto.
10688 (vsm4eq_u32): Ditto.
10689 (vsm4ekeyq_u32): Ditto.
10690 (doc/invoke.texi): Document new sm4 option.
10692 2018-01-10 Michael Collison <michael.collison@arm.com>
10694 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10695 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10696 (AARCH64_FL_FOR_ARCH8_4): New.
10697 (AARCH64_FL_V8_4): New flag.
10698 (doc/invoke.texi): Document new armv8.4-a option.
10700 2018-01-10 Michael Collison <michael.collison@arm.com>
10702 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10703 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10704 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10705 * config/aarch64/aarch64-option-extension.def: Add
10706 AARCH64_OPT_EXTENSION of 'sha2'.
10707 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10708 (crypto): Disable sha2 and aes if crypto disabled.
10709 (crypto): Enable aes and sha2 if enabled.
10710 (simd): Disable sha2 and aes if simd disabled.
10711 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10713 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10714 (TARGET_SHA2): New feature flag for sha2.
10715 (TARGET_AES): New feature flag for aes.
10716 * config/aarch64/aarch64-simd.md:
10717 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10718 conditional on TARGET_AES.
10719 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10720 (aarch64_crypto_sha1hsi): Make pattern conditional
10722 (aarch64_crypto_sha1hv4si): Ditto.
10723 (aarch64_be_crypto_sha1hv4si): Ditto.
10724 (aarch64_crypto_sha1su1v4si): Ditto.
10725 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10726 (aarch64_crypto_sha1su0v4si): Ditto.
10727 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10728 (aarch64_crypto_sha256su0v4si): Ditto.
10729 (aarch64_crypto_sha256su1v4si): Ditto.
10730 (doc/invoke.texi): Document new aes and sha2 options.
10732 2018-01-10 Martin Sebor <msebor@redhat.com>
10734 PR tree-optimization/83781
10735 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10738 2018-01-11 Martin Sebor <msebor@gmail.com>
10739 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10741 PR tree-optimization/83501
10742 PR tree-optimization/81703
10744 * tree-ssa-strlen.c (get_string_cst): Rename...
10745 (get_string_len): ...to this. Handle global constants.
10746 (handle_char_store): Adjust.
10748 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10749 Jim Wilson <jimw@sifive.com>
10751 * config/riscv/riscv-protos.h (riscv_output_return): New.
10752 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10753 (riscv_attribute_table, riscv_output_return),
10754 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10755 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10756 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10758 (riscv_expand_prologue): Add early return for naked function.
10759 (riscv_expand_epilogue): Likewise.
10760 (riscv_function_ok_for_sibcall): Return false for naked function.
10761 (riscv_set_current_function): New.
10762 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10763 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10764 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10765 * doc/extend.texi (RISC-V Function Attributes): New.
10767 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10769 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10770 check for 128-bit long double before checking TCmode.
10771 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10772 128-bit long doubles before checking TFmode or TCmode.
10773 (FLOAT128_IBM_P): Likewise.
10775 2018-01-10 Martin Sebor <msebor@redhat.com>
10777 PR tree-optimization/83671
10778 * builtins.c (c_strlen): Unconditionally return zero for the empty
10780 Use -Warray-bounds for warnings.
10781 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10782 for non-constant array indices with COMPONENT_REF, arrays of
10783 arrays, and pointers to arrays.
10784 (gimple_fold_builtin_strlen): Determine and set length range for
10785 non-constant character arrays.
10787 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10789 PR middle-end/81897
10790 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10793 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10795 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10797 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10800 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10801 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10802 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10803 indexed_or_indirect_operand predicate.
10804 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10805 (*vsx_le_perm_load_v8hi): Likewise.
10806 (*vsx_le_perm_load_v16qi): Likewise.
10807 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10808 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10809 (*vsx_le_perm_store_v8hi): Likewise.
10810 (*vsx_le_perm_store_v16qi): Likewise.
10811 (eight unnamed splitters): Likewise.
10813 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10815 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10816 * config/rs6000/emmintrin.h: Likewise.
10817 * config/rs6000/mmintrin.h: Likewise.
10818 * config/rs6000/xmmintrin.h: Likewise.
10820 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10823 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10825 * tree.c (tree_nop_conversion): Return true for location wrapper
10827 (maybe_wrap_with_location): New function.
10828 (selftest::check_strip_nops): New function.
10829 (selftest::test_location_wrappers): New function.
10830 (selftest::tree_c_tests): Call it.
10831 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10832 (maybe_wrap_with_location): New decl.
10833 (EXPR_LOCATION_WRAPPER_P): New macro.
10834 (location_wrapper_p): New inline function.
10835 (tree_strip_any_location_wrapper): New inline function.
10837 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10840 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10841 stack_realign_offset for the largest alignment of stack slot
10843 (ix86_find_max_used_stack_alignment): New function.
10844 (ix86_finalize_stack_frame_flags): Use it. Set
10845 max_used_stack_alignment if we don't realign stack.
10846 * config/i386/i386.h (machine_function): Add
10847 max_used_stack_alignment.
10849 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10851 * config/arm/arm.opt (-mbranch-cost): New option.
10852 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10855 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10858 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10859 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10861 2018-01-10 Richard Biener <rguenther@suse.de>
10864 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10865 early out so it also covers the case where we have a non-NULL
10868 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10870 PR tree-optimization/83753
10871 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10872 for non-strided grouped accesses if the number of elements is 1.
10874 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10877 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10878 * i386.h (TARGET_USE_GATHER): Define.
10879 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10881 2018-01-10 Martin Liska <mliska@suse.cz>
10884 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10885 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10887 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10888 CLEANUP_NO_PARTITIONING is not set.
10890 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10892 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10893 for vectors, as a partial revert of r254296.
10894 * rtl.h (const_vec_p): Delete.
10895 (const_vec_duplicate_p): Don't test for vector CONSTs.
10896 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10897 * expmed.c (make_tree): Likewise.
10900 * common.md (E, F): Use CONSTANT_P instead of checking for
10902 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10903 checking for CONST_VECTOR.
10905 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10907 PR middle-end/83575
10908 * predict.c (force_edge_cold): Handle in more sane way edges
10909 with no prediction.
10911 2018-01-09 Carl Love <cel@us.ibm.com>
10913 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10915 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10916 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10917 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10918 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10919 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10920 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10921 * config/rs6000/rs6000-protos.h: Add extern defition for
10922 rs6000_generate_float2_double_code.
10923 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10925 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10926 (float2_v2df): Add define_expand.
10928 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10931 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10932 op_mode in the force_to_mode call.
10934 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10936 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10937 instead of checking each element individually.
10938 (aarch64_evpc_uzp): Likewise.
10939 (aarch64_evpc_zip): Likewise.
10940 (aarch64_evpc_ext): Likewise.
10941 (aarch64_evpc_rev): Likewise.
10942 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10943 instead of checking each element individually. Return true without
10945 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10946 whether all selected elements come from the same input, instead of
10947 checking each element individually. Remove calls to gen_rtx_REG,
10948 start_sequence and end_sequence and instead assert that no rtl is
10951 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10953 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10954 order of HIGH and CONST checks.
10956 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10958 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10959 if the destination isn't an SSA_NAME.
10961 2018-01-09 Richard Biener <rguenther@suse.de>
10963 PR tree-optimization/83668
10964 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10966 (canonicalize_loop_form): ... here, renamed from ...
10967 (canonicalize_loop_closed_ssa_form): ... this and amended to
10968 swap successor edges for loop exit blocks to make us use
10969 the RPO order we need for initial schedule generation.
10971 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10973 PR tree-optimization/64811
10974 * match.pd: When optimizing comparisons with Inf, avoid
10975 introducing or losing exceptions from comparisons with NaN.
10977 2018-01-09 Martin Liska <mliska@suse.cz>
10980 * asan.c (shadow_mem_size): Add gcc_assert.
10982 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10984 Don't save registers in main().
10987 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10988 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10989 * config/avr/avr.c (avr_set_current_function): Don't error if
10990 naked, OS_task or OS_main are specified at the same time.
10991 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10993 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10995 * common/config/avr/avr-common.c (avr_option_optimization_table):
10996 Switch on -mmain-is-OS_task for optimizing compilations.
10998 2018-01-09 Richard Biener <rguenther@suse.de>
11000 PR tree-optimization/83572
11001 * graphite.c: Include cfganal.h.
11002 (graphite_transform_loops): Connect infinite loops to exit
11003 and remove fake edges at the end.
11005 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11007 * ipa-inline.c (edge_badness): Revert accidental checkin.
11009 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11012 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11013 symbols; not inline clones.
11015 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11018 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11019 hard registers. Formatting fixes.
11021 PR preprocessor/83722
11022 * gcc.c (try_generate_repro): Pass
11023 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11024 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11027 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11028 Kito Cheng <kito.cheng@gmail.com>
11030 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11031 (riscv_leaf_function_p): Delete.
11032 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11034 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11036 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11038 (do_ifelse): New function.
11039 (do_isel): New function.
11040 (do_sub3): New function.
11041 (do_add3): New function.
11042 (do_load_mask_compare): New function.
11043 (do_overlap_load_compare): New function.
11044 (expand_compare_loop): New function.
11045 (expand_block_compare): Call expand_compare_loop() when appropriate.
11046 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11047 option description.
11048 (-mblock-compare-inline-loop-limit): New option.
11050 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11053 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11054 Reverse order of second and third operands in first alternative.
11055 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11056 of first and second elements in UNSPEC_VPERMR vector.
11057 (altivec_expand_vec_perm_le): Likewise.
11059 2018-01-08 Jeff Law <law@redhat.com>
11061 PR rtl-optimizatin/81308
11062 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11063 (process_switch): If group_case_labels makes a change, then set
11065 (pass_convert_switch::execute): If a switch is converted, then
11066 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11068 PR rtl-optimization/81308
11069 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11072 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11074 PR target/83663 - Revert r255946
11075 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11076 generation for cases where splatting a value is not useful.
11077 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11078 across a vec_duplicate and a paradoxical subreg forming a vector
11079 mode to a vec_concat.
11081 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11083 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11084 -march=armv8.3-a variants.
11085 * config/arm/t-multilib: Likewise.
11086 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11088 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11090 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11092 (cceq_ior_compare_complement): Give it a name so I can use it, and
11093 change boolean_or_operator predicate to boolean_operator so it can
11094 be used to generate a crand.
11095 (eqne): New code iterator.
11096 (bd/bd_neg): New code_attrs.
11097 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11098 a single define_insn.
11099 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11100 decrement (bdnzt/bdnzf/bdzt/bdzf).
11101 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11102 with the new names of the branch decrement patterns, and added the
11103 names of the branch decrement conditional patterns.
11105 2018-01-08 Richard Biener <rguenther@suse.de>
11107 PR tree-optimization/83563
11108 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11111 2018-01-08 Richard Biener <rguenther@suse.de>
11113 PR middle-end/83713
11114 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11116 2018-01-08 Richard Biener <rguenther@suse.de>
11118 PR tree-optimization/83685
11119 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11120 references to abnormals.
11122 2018-01-08 Richard Biener <rguenther@suse.de>
11125 * dwarf2out.c (output_indirect_strings): Handle empty
11126 skeleton_debug_str_hash.
11127 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11129 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11131 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11132 (emit_store_direct): Likewise.
11133 (arc_trampoline_adjust_address): Likewise.
11134 (arc_asm_trampoline_template): New function.
11135 (arc_initialize_trampoline): Use asm_trampoline_template.
11136 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11137 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11138 * config/arc/arc.md (flush_icache): Delete pattern.
11140 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11142 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11143 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11146 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11149 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11150 by not USED_FOR_TARGET.
11151 (make_pass_resolve_sw_modes): Likewise.
11153 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11155 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11158 2018-01-08 Richard Biener <rguenther@suse.de>
11160 PR middle-end/83580
11161 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11163 2018-01-08 Richard Biener <rguenther@suse.de>
11165 PR middle-end/83517
11166 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11168 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11170 PR middle-end/81897
11171 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11172 basic blocks with a small number of successors.
11173 (convert_control_dep_chain_into_preds): Improve handling of
11175 (dump_predicates): Split apart into...
11176 (dump_pred_chain): ...here...
11177 (dump_pred_info): ...and here.
11178 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11179 (can_chain_union_be_invalidated_p): Improve check for invalidation
11181 (uninit_uses_cannot_happen): Avoid unnecessary if
11182 convert_control_dep_chain_into_preds yielded nothing.
11184 2018-01-06 Martin Sebor <msebor@redhat.com>
11186 PR tree-optimization/83640
11187 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11188 subtracting negative offset from size.
11189 (builtin_access::overlap): Adjust offset bounds of the access to fall
11190 within the size of the object if possible.
11192 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11194 PR rtl-optimization/83699
11195 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11196 extract_bit_field_as_subreg to cases in which the extracted
11197 value is also a vector.
11199 * lra-constraints.c (process_alt_operands): Test for the equivalence
11200 substitutions when detecting a possible reload cycle.
11202 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11205 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11206 by default if flag_selective_schedling{,2}. Formatting fixes.
11208 PR rtl-optimization/83682
11209 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11210 if it has non-VECTOR_MODE element mode.
11211 (vec_duplicate_p): Likewise.
11213 PR middle-end/83694
11214 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11215 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11217 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11220 * config/i386/i386-builtin.def
11221 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11222 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11223 Require also OPTION_MASK_ISA_AVX512F in addition to
11224 OPTION_MASK_ISA_GFNI.
11225 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11226 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11227 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11228 to OPTION_MASK_ISA_GFNI.
11229 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11230 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11231 OPTION_MASK_ISA_AVX512BW.
11232 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11233 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11234 addition to OPTION_MASK_ISA_GFNI.
11235 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11236 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11237 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11238 to OPTION_MASK_ISA_GFNI.
11239 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11240 a requirement for all ISAs rather than any of them with a few
11242 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11244 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11245 bitmasks to be enabled with 3 exceptions, instead of requiring any
11246 enabled ISA with lots of exceptions.
11247 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11248 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11249 Change avx512bw in isa attribute to avx512f.
11250 * config/i386/sgxintrin.h: Add license boilerplate.
11251 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11252 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11253 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11254 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11256 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11257 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11258 temporarily sse2 rather than sse if not enabled already.
11261 * config/i386/sse.md (VI248_VLBW): Rename to ...
11262 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11263 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11264 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11265 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11266 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11267 mode iterator instead of VI248_VLBW.
11269 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11271 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11272 (record_modified): Skip clobbers; add debug output.
11273 (param_change_prob): Use sreal frequencies.
11275 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11277 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11278 punt for user-aligned variables.
11280 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11282 * tree-chrec.c (chrec_contains_symbols): Return true for
11285 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11288 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11289 of (x|y) == x for BICS pattern.
11291 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11293 PR tree-optimization/83605
11294 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11295 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11298 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11300 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11301 * config/epiphany/rtems.h: New file.
11303 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11304 Uros Bizjak <ubizjak@gmail.com>
11307 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11308 QIreg_operand instead of register_operand predicate.
11309 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11310 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11311 comments instead of -fmitigate[-_]rop.
11313 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11316 * cgraphunit.c (symbol_table::compile): Switch to text_section
11317 before calling assembly_start debug hook.
11318 * run-rtl-passes.c (run_rtl_passes): Likewise.
11321 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11323 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11324 range_int_cst_p rather than !symbolic_range_p before calling
11325 extract_range_from_multiplicative_op_1.
11327 2018-01-04 Jeff Law <law@redhat.com>
11329 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11330 redundant test in assertion.
11332 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11334 * doc/rtl.texi: Document machine_mode wrapper classes.
11336 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11338 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11339 using tree_to_uhwi.
11341 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11343 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11344 the VEC_PERM_EXPR fold to fail.
11346 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11349 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11350 to switched_sections.
11352 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11355 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11356 test for d.testing.
11358 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11361 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11362 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11364 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11367 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11368 is BLKmode and bitpos not zero or mode change is needed.
11370 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11373 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11376 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11379 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11380 instead of MULT rtx. Update all corresponding splitters.
11381 (*saddl_se): Ditto.
11382 (*ssub<modesuffix>): Ditto.
11383 (*ssubl_se): Ditto.
11384 (*cmp_sadd_di): Update split patterns.
11385 (*cmp_sadd_si): Ditto.
11386 (*cmp_sadd_sidi): Ditto.
11387 (*cmp_ssub_di): Ditto.
11388 (*cmp_ssub_si): Ditto.
11389 (*cmp_ssub_sidi): Ditto.
11390 * config/alpha/predicates.md (const23_operand): New predicate.
11391 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11392 Look for ASHIFT, not MULT inner operand.
11393 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11395 2018-01-04 Martin Liska <mliska@suse.cz>
11397 PR gcov-profile/83669
11398 * gcov.c (output_intermediate_file): Add version to intermediate
11400 * doc/gcov.texi: Document new field 'version' in intermediate
11401 file format. Fix location of '-k' option of gcov command.
11403 2018-01-04 Martin Liska <mliska@suse.cz>
11406 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11408 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11410 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11412 2018-01-03 Martin Sebor <msebor@redhat.com>
11414 PR tree-optimization/83655
11415 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11416 checking calls with invalid arguments.
11418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11420 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11421 (vectorizable_mask_load_store): Delete.
11422 (vectorizable_call): Return false for masked loads and stores.
11423 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11424 instead of gimple_assign_rhs1.
11425 (vectorizable_load): Handle IFN_MASK_LOAD.
11426 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11430 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11432 (vectorizable_mask_load_store): ...here.
11433 (vectorizable_load): ...and here.
11435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11437 * tree-vect-stmts.c (vect_build_all_ones_mask)
11438 (vect_build_zero_merge_argument): New functions, split out from...
11439 (vectorizable_load): ...here.
11441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11443 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11445 (vectorizable_mask_load_store): ...here.
11446 (vectorizable_store): ...and here.
11448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11450 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11452 (vectorizable_mask_load_store): ...here.
11454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11456 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11457 (vect_model_store_cost): Take a vec_load_store_type instead of a
11459 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11460 (vect_model_store_cost): Take a vec_load_store_type instead of a
11462 (vectorizable_mask_load_store): Update accordingly.
11463 (vectorizable_store): Likewise.
11464 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11468 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11469 IFN_MASK_LOAD calls here rather than...
11470 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11473 Alan Hayward <alan.hayward@arm.com>
11474 David Sherwood <david.sherwood@arm.com>
11476 * expmed.c (extract_bit_field_1): For vector extracts,
11477 fall back to extract_bit_field_as_subreg if vec_extract
11480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11481 Alan Hayward <alan.hayward@arm.com>
11482 David Sherwood <david.sherwood@arm.com>
11484 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11485 they are variable or constant sized.
11486 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11487 slots for constant-sized data.
11489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11490 Alan Hayward <alan.hayward@arm.com>
11491 David Sherwood <david.sherwood@arm.com>
11493 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11494 handling COND_EXPRs with boolean comparisons, try to find a better
11495 basis for the mask type than the boolean itself.
11497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11499 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11500 is calculated and how it can be overridden.
11501 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11502 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11504 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11508 Alan Hayward <alan.hayward@arm.com>
11509 David Sherwood <david.sherwood@arm.com>
11511 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11512 Remove the mode argument.
11513 (aarch64_simd_valid_immediate): Remove the mode and inverse
11515 * config/aarch64/iterators.md (bitsize): New iterator.
11516 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11517 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11518 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11519 aarch64_simd_valid_immediate.
11520 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11521 (aarch64_reg_or_bic_imm): Likewise.
11522 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11523 with an insn_type enum and msl with a modifier_type enum.
11524 Replace element_width with a scalar_mode. Change the shift
11525 to unsigned int. Add constructors for scalar_float_mode and
11526 scalar_int_mode elements.
11527 (aarch64_vect_float_const_representable_p): Delete.
11528 (aarch64_can_const_movi_rtx_p)
11529 (aarch64_simd_scalar_immediate_valid_for_move)
11530 (aarch64_simd_make_constant): Update call to
11531 aarch64_simd_valid_immediate.
11532 (aarch64_advsimd_valid_immediate_hs): New function.
11533 (aarch64_advsimd_valid_immediate): Likewise.
11534 (aarch64_simd_valid_immediate): Remove mode and inverse
11535 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11536 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11537 and aarch64_float_const_representable_p on the result.
11538 (aarch64_output_simd_mov_immediate): Remove mode argument.
11539 Update call to aarch64_simd_valid_immediate and use of
11540 simd_immediate_info.
11541 (aarch64_output_scalar_simd_mov_immediate): Update call
11544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11545 Alan Hayward <alan.hayward@arm.com>
11546 David Sherwood <david.sherwood@arm.com>
11548 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11549 (mode_nunits): Likewise CONST_MODE_NUNITS.
11550 * machmode.def (ADJUST_NUNITS): Document.
11551 * genmodes.c (mode_data::need_nunits_adj): New field.
11552 (blank_mode): Update accordingly.
11553 (adj_nunits): New variable.
11554 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11556 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11557 listed in adj_nunits.
11558 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11559 listed in adj_nunits. Don't emit case statements for such modes.
11560 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11561 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11562 nothing if adj_nunits is nonnull.
11563 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11564 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11565 (emit_mode_fbit): Update use of print_maybe_const_decl.
11566 (emit_move_size): Likewise. Treat the array as non-const
11568 (emit_mode_adjustments): Handle adj_nunits.
11570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11572 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11573 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11574 (VECTOR_MODES): Use it.
11575 (make_vector_modes): Take the prefix as an argument.
11577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11578 Alan Hayward <alan.hayward@arm.com>
11579 David Sherwood <david.sherwood@arm.com>
11581 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11582 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11583 for MODE_VECTOR_BOOL.
11584 * machmode.def (VECTOR_BOOL_MODE): Document.
11585 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11586 (make_vector_bool_mode): New function.
11587 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11589 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11590 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11592 * stor-layout.c (int_mode_for_mode): Likewise.
11593 * tree.c (build_vector_type_for_mode): Likewise.
11594 * varasm.c (output_constant_pool_2): Likewise.
11595 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11596 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11597 for MODE_VECTOR_BOOL.
11598 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11599 of mode class checks.
11600 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11601 instead of a list of mode class checks.
11602 (expand_vector_scalar_condition): Likewise.
11603 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11606 Alan Hayward <alan.hayward@arm.com>
11607 David Sherwood <david.sherwood@arm.com>
11609 * machmode.h (mode_size): Change from unsigned short to
11611 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11612 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11613 or if measurement_type is not polynomial.
11614 (fixed_size_mode::includes_p): Check for constant-sized modes.
11615 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11616 return a poly_uint16 rather than an unsigned short.
11617 (emit_mode_size): Change the type of mode_size from unsigned short
11618 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11619 (emit_mode_adjustments): Cope with polynomial vector sizes.
11620 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11622 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11624 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11625 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11626 * caller-save.c (setup_save_areas): Likewise.
11627 (replace_reg_with_saved_mem): Likewise.
11628 * calls.c (emit_library_call_value_1): Likewise.
11629 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11630 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11631 (gen_lowpart_for_combine): Likewise.
11632 * convert.c (convert_to_integer_1): Likewise.
11633 * cse.c (equiv_constant, cse_insn): Likewise.
11634 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11635 (cselib_subst_to_values): Likewise.
11636 * dce.c (word_dce_process_block): Likewise.
11637 * df-problems.c (df_word_lr_mark_ref): Likewise.
11638 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11639 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11640 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11641 (rtl_for_decl_location): Likewise.
11642 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11643 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11644 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11645 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11646 (expand_expr_real_1): Likewise.
11647 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11648 (pad_below): Likewise.
11649 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11650 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11651 * ira.c (get_subreg_tracking_sizes): Likewise.
11652 * ira-build.c (ira_create_allocno_objects): Likewise.
11653 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11654 (ira_sort_regnos_for_alter_reg): Likewise.
11655 * ira-costs.c (record_operand_costs): Likewise.
11656 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11657 (resolve_simple_move): Likewise.
11658 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11659 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11660 (lra_constraints): Likewise.
11661 (CONST_POOL_OK_P): Reject variable-sized modes.
11662 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11663 (add_pseudo_to_slot, lra_spill): Likewise.
11664 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11665 * optabs-query.c (get_best_extraction_insn): Likewise.
11666 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11667 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11668 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11669 * recog.c (offsettable_address_addr_space_p): Likewise.
11670 * regcprop.c (maybe_mode_change): Likewise.
11671 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11672 * regrename.c (build_def_use): Likewise.
11673 * regstat.c (dump_reg_info): Likewise.
11674 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11675 (find_reloads, find_reloads_subreg_address): Likewise.
11676 * reload1.c (eliminate_regs_1): Likewise.
11677 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11678 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11679 (simplify_binary_operation_1, simplify_subreg): Likewise.
11680 * targhooks.c (default_function_arg_padding): Likewise.
11681 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11682 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11683 (verify_gimple_assign_ternary): Likewise.
11684 * tree-inline.c (estimate_move_cost): Likewise.
11685 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11686 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11687 (get_address_cost_ainc): Likewise.
11688 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11689 (vect_supportable_dr_alignment): Likewise.
11690 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11691 (vectorizable_reduction): Likewise.
11692 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11693 (vectorizable_operation, vectorizable_load): Likewise.
11694 * tree.c (build_same_sized_truth_vector_type): Likewise.
11695 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11696 * var-tracking.c (emit_note_insn_var_location): Likewise.
11697 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11698 (ADDR_VEC_ALIGN): Likewise.
11700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11701 Alan Hayward <alan.hayward@arm.com>
11702 David Sherwood <david.sherwood@arm.com>
11704 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11706 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11707 or if measurement_type is polynomial.
11708 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11709 * combine.c (make_extraction): Likewise.
11710 * dse.c (find_shift_sequence): Likewise.
11711 * dwarf2out.c (mem_loc_descriptor): Likewise.
11712 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11713 (extract_bit_field, extract_low_bits): Likewise.
11714 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11715 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11716 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11717 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11718 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11719 * reload.c (find_reloads): Likewise.
11720 * reload1.c (alter_reg): Likewise.
11721 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11722 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11723 * tree-if-conv.c (predicate_mem_writes): Likewise.
11724 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11725 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11726 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11727 * valtrack.c (dead_debug_insert_temp): Likewise.
11728 * varasm.c (mergeable_constant_section): Likewise.
11729 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11732 Alan Hayward <alan.hayward@arm.com>
11733 David Sherwood <david.sherwood@arm.com>
11735 * expr.c (expand_assignment): Cope with polynomial mode sizes
11736 when assigning to a CONCAT.
11738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11739 Alan Hayward <alan.hayward@arm.com>
11740 David Sherwood <david.sherwood@arm.com>
11742 * machmode.h (mode_precision): Change from unsigned short to
11744 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11746 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11747 or if measurement_type is not polynomial.
11748 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11749 in which the mode is already known to be a scalar_int_mode.
11750 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11751 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11753 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11754 for GET_MODE_PRECISION.
11755 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11756 for GET_MODE_PRECISION.
11757 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11759 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11760 (expand_field_assignment, make_extraction): Likewise.
11761 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11762 (get_last_value): Likewise.
11763 * convert.c (convert_to_integer_1): Likewise.
11764 * cse.c (cse_insn): Likewise.
11765 * expr.c (expand_expr_real_1): Likewise.
11766 * lra-constraints.c (simplify_operand_subreg): Likewise.
11767 * optabs-query.c (can_atomic_load_p): Likewise.
11768 * optabs.c (expand_atomic_load): Likewise.
11769 (expand_atomic_store): Likewise.
11770 * ree.c (combine_reaching_defs): Likewise.
11771 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11772 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11773 * tree.h (type_has_mode_precision_p): Likewise.
11774 * ubsan.c (instrument_si_overflow): Likewise.
11776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11777 Alan Hayward <alan.hayward@arm.com>
11778 David Sherwood <david.sherwood@arm.com>
11780 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11781 polynomial numbers of units.
11782 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11783 (valid_vector_subparts_p): New function.
11784 (build_vector_type): Remove temporary shim and take the number
11785 of units as a poly_uint64 rather than an int.
11786 (build_opaque_vector_type): Take the number of units as a
11787 poly_uint64 rather than an int.
11788 * tree.c (build_vector_from_ctor): Handle polynomial
11789 TYPE_VECTOR_SUBPARTS.
11790 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11791 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11792 (build_vector_from_val): If the number of units is variable,
11793 use build_vec_duplicate_cst for constant operands and
11794 VEC_DUPLICATE_EXPR otherwise.
11795 (make_vector_type): Remove temporary is_constant ().
11796 (build_vector_type, build_opaque_vector_type): Take the number of
11797 units as a poly_uint64 rather than an int.
11798 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11800 * cfgexpand.c (expand_debug_expr): Likewise.
11801 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11802 (store_constructor, expand_expr_real_1): Likewise.
11803 (const_scalar_mask_from_tree): Likewise.
11804 * fold-const-call.c (fold_const_reduction): Likewise.
11805 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11806 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11807 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11808 (fold_relational_const): Likewise.
11809 (native_interpret_vector): Likewise. Change the size from an
11810 int to an unsigned int.
11811 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11812 TYPE_VECTOR_SUBPARTS.
11813 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11814 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11815 duplicating a non-constant operand into a variable-length vector.
11816 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11817 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11818 * ipa-icf.c (sem_variable::equals): Likewise.
11819 * match.pd: Likewise.
11820 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11821 * print-tree.c (print_node): Likewise.
11822 * stor-layout.c (layout_type): Likewise.
11823 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11824 * tree-cfg.c (verify_gimple_comparison): Likewise.
11825 (verify_gimple_assign_binary): Likewise.
11826 (verify_gimple_assign_ternary): Likewise.
11827 (verify_gimple_assign_single): Likewise.
11828 * tree-pretty-print.c (dump_generic_node): Likewise.
11829 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11830 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11831 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11832 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11833 (vect_shift_permute_load_chain): Likewise.
11834 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11835 (expand_vector_condition, optimize_vector_constructor): Likewise.
11836 (lower_vec_perm, get_compute_type): Likewise.
11837 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11838 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11839 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11840 (vect_recog_mask_conversion_pattern): Likewise.
11841 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11842 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11843 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11844 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11845 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11846 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11847 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11848 (supportable_widening_operation): Likewise.
11849 (supportable_narrowing_operation): Likewise.
11850 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11852 * varasm.c (output_constant): Likewise.
11854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11855 Alan Hayward <alan.hayward@arm.com>
11856 David Sherwood <david.sherwood@arm.com>
11858 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11859 so that both the length == 3 and length != 3 cases set up their
11860 own permute vectors. Add comments explaining why we know the
11861 number of elements is constant.
11862 (vect_permute_load_chain): Likewise.
11864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11865 Alan Hayward <alan.hayward@arm.com>
11866 David Sherwood <david.sherwood@arm.com>
11868 * machmode.h (mode_nunits): Change from unsigned char to
11870 (ONLY_FIXED_SIZE_MODES): New macro.
11871 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11872 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11873 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11875 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11876 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11877 or if measurement_type is not polynomial.
11878 * genmodes.c (ZERO_COEFFS): New macro.
11879 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11881 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11882 Use ZERO_COEFFS when emitting initializers.
11883 * data-streamer.h (bp_pack_poly_value): New function.
11884 (bp_unpack_poly_value): Likewise.
11885 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11886 for GET_MODE_NUNITS.
11887 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11888 for GET_MODE_NUNITS.
11889 * tree.c (make_vector_type): Remove temporary shim and make
11890 the real function take the number of units as a poly_uint64
11891 rather than an int.
11892 (build_vector_type_for_mode): Handle polynomial nunits.
11893 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11894 * emit-rtl.c (const_vec_series_p_1): Likewise.
11895 (gen_rtx_CONST_VECTOR): Likewise.
11896 * fold-const.c (test_vec_duplicate_folding): Likewise.
11897 * genrecog.c (validate_pattern): Likewise.
11898 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11899 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11900 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11901 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11902 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11903 * rtlanal.c (subreg_get_info): Likewise.
11904 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11905 (vect_grouped_load_supported): Likewise.
11906 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11907 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11908 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11909 (simplify_const_unary_operation, simplify_binary_operation_1)
11910 (simplify_const_binary_operation, simplify_ternary_operation)
11911 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11912 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11913 instead of CONST_VECTOR_NUNITS.
11914 * varasm.c (output_constant_pool_2): Likewise.
11915 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11916 explicit-encoded elements in the XVEC for variable-length vectors.
11918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11920 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11923 Alan Hayward <alan.hayward@arm.com>
11924 David Sherwood <david.sherwood@arm.com>
11926 * coretypes.h (fixed_size_mode): Declare.
11927 (fixed_size_mode_pod): New typedef.
11928 * builtins.h (target_builtins::x_apply_args_mode)
11929 (target_builtins::x_apply_result_mode): Change type to
11930 fixed_size_mode_pod.
11931 * builtins.c (apply_args_size, apply_result_size, result_vector)
11932 (expand_builtin_apply_args_1, expand_builtin_apply)
11933 (expand_builtin_return): Update accordingly.
11935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11937 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11938 * cselib.c (cselib_hash_rtx): Likewise.
11939 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11940 CONST_VECTOR encoding.
11942 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11943 Jeff Law <law@redhat.com>
11946 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11947 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11948 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11949 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11952 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11953 explicitly probe *sp in a noreturn function if there were any callee
11954 register saves or frame pointer is needed.
11956 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11959 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11960 BLKmode for ternary, binary or unary expressions.
11963 * var-tracking.c (delete_vta_debug_insn): New inline function.
11964 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11965 insns from get_insns () to NULL instead of each bb separately.
11966 Use delete_vta_debug_insn. No longer static.
11967 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11968 delete_vta_debug_insns callers.
11969 * rtl.h (delete_vta_debug_insns): Declare.
11970 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11971 instead of variable_tracking_main.
11973 2018-01-03 Martin Sebor <msebor@redhat.com>
11975 PR tree-optimization/83603
11976 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11977 arguments past the endof the argument list in functions declared
11978 without a prototype.
11979 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11980 Avoid checking when arguments are null.
11982 2018-01-03 Martin Sebor <msebor@redhat.com>
11985 * doc/extend.texi (attribute const): Fix a typo.
11986 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11987 issuing -Wsuggest-attribute for void functions.
11989 2018-01-03 Martin Sebor <msebor@redhat.com>
11991 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11992 offset_int::from instead of wide_int::to_shwi.
11993 (maybe_diag_overlap): Remove assertion.
11994 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11995 * gimple-ssa-sprintf.c (format_directive): Same.
11996 (parse_directive): Same.
11997 (sprintf_dom_walker::compute_format_length): Same.
11998 (try_substitute_return_value): Same.
12000 2018-01-03 Jeff Law <law@redhat.com>
12002 PR middle-end/83654
12003 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12004 non-constant residual for zero at runtime and avoid probing in
12005 that case. Reorganize code for trailing problem to mirror handling
12008 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12010 PR tree-optimization/83501
12011 * tree-ssa-strlen.c (get_string_cst): New.
12012 (handle_char_store): Call get_string_cst.
12014 2018-01-03 Martin Liska <mliska@suse.cz>
12016 PR tree-optimization/83593
12017 * tree-ssa-strlen.c: Include tree-cfg.h.
12018 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12019 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12020 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12022 (strlen_dom_walker::before_dom_children): Call
12023 gimple_purge_dead_eh_edges. Dump tranformation with details
12025 (strlen_dom_walker::before_dom_children): Update call by adding
12026 new argument cleanup_eh.
12027 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12029 2018-01-03 Martin Liska <mliska@suse.cz>
12032 * cif-code.def (VARIADIC_THUNK): New enum value.
12033 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12036 2018-01-03 Jan Beulich <jbeulich@suse.com>
12038 * sse.md (mov<mode>_internal): Tighten condition for when to use
12039 vmovdqu<ssescalarsize> for TI and OI modes.
12041 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12043 Update copyright years.
12045 2018-01-03 Martin Liska <mliska@suse.cz>
12048 * ipa-visibility.c (function_and_variable_visibility): Skip
12049 functions with noipa attribure.
12051 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12053 * gcc.c (process_command): Update copyright notice dates.
12054 * gcov-dump.c (print_version): Ditto.
12055 * gcov.c (print_version): Ditto.
12056 * gcov-tool.c (print_version): Ditto.
12057 * gengtype.c (create_file): Ditto.
12058 * doc/cpp.texi: Bump @copying's copyright year.
12059 * doc/cppinternals.texi: Ditto.
12060 * doc/gcc.texi: Ditto.
12061 * doc/gccint.texi: Ditto.
12062 * doc/gcov.texi: Ditto.
12063 * doc/install.texi: Ditto.
12064 * doc/invoke.texi: Ditto.
12066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12068 * vector-builder.h (vector_builder::m_full_nelts): Change from
12069 unsigned int to poly_uint64.
12070 (vector_builder::full_nelts): Update prototype accordingly.
12071 (vector_builder::new_vector): Likewise.
12072 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12073 (vector_builder::operator ==): Likewise.
12074 (vector_builder::finalize): Likewise.
12075 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12076 Take the number of elements as a poly_uint64 rather than an
12078 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12079 from unsigned int to poly_uint64.
12080 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12081 (vec_perm_indices::new_vector): Likewise.
12082 (vec_perm_indices::length): Likewise.
12083 (vec_perm_indices::nelts_per_input): Likewise.
12084 (vec_perm_indices::input_nelts): Likewise.
12085 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12086 number of elements per input as a poly_uint64 rather than an
12087 unsigned int. Use the original encoding for variable-length
12088 vectors, rather than clamping each individual element.
12089 For the second and subsequent elements in each pattern,
12090 clamp the step and base before clamping their sum.
12091 (vec_perm_indices::series_p): Handle polynomial element counts.
12092 (vec_perm_indices::all_in_range_p): Likewise.
12093 (vec_perm_indices_to_tree): Likewise.
12094 (vec_perm_indices_to_rtx): Likewise.
12095 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12096 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12097 (tree_vector_builder::new_binary_operation): Handle polynomial
12098 element counts. Return false if we need to know the number
12099 of elements at compile time.
12100 * fold-const.c (fold_vec_perm): Punt if the number of elements
12101 isn't known at compile time.
12103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12105 * vec-perm-indices.h (vec_perm_builder): Change element type
12106 from HOST_WIDE_INT to poly_int64.
12107 (vec_perm_indices::element_type): Update accordingly.
12108 (vec_perm_indices::clamp): Handle polynomial element_types.
12109 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12110 (vec_perm_indices::all_in_range_p): Likewise.
12111 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12113 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12114 polynomial vec_perm_indices element types.
12115 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12116 * fold-const.c (fold_vec_perm): Likewise.
12117 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12118 * tree-vect-generic.c (lower_vec_perm): Likewise.
12119 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12120 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12121 element type to HOST_WIDE_INT.
12123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12124 Alan Hayward <alan.hayward@arm.com>
12125 David Sherwood <david.sherwood@arm.com>
12127 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12128 rather than an int. Use plus_constant.
12129 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12130 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12133 Alan Hayward <alan.hayward@arm.com>
12134 David Sherwood <david.sherwood@arm.com>
12136 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12137 a HOST_WIDE_INT to a poly_int64.
12139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12140 Alan Hayward <alan.hayward@arm.com>
12141 David Sherwood <david.sherwood@arm.com>
12143 * calls.c (load_register_parameters): Cope with polynomial
12144 mode sizes. Require a constant size for BLKmode parameters
12145 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12146 forces a parameter to be padded at the lsb end in order to
12147 fill a complete number of words, require the parameter size
12148 to be ordered wrt UNITS_PER_WORD.
12150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12151 Alan Hayward <alan.hayward@arm.com>
12152 David Sherwood <david.sherwood@arm.com>
12154 * reload1.c (spill_stack_slot_width): Change element type
12155 from unsigned int to poly_uint64_pod.
12156 (alter_reg): Treat mode sizes as polynomial.
12158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12159 Alan Hayward <alan.hayward@arm.com>
12160 David Sherwood <david.sherwood@arm.com>
12162 * reload.c (complex_word_subreg_p): New function.
12163 (reload_inner_reg_of_subreg, push_reload): Use it.
12165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12166 Alan Hayward <alan.hayward@arm.com>
12167 David Sherwood <david.sherwood@arm.com>
12169 * lra-constraints.c (process_alt_operands): Reject matched
12170 operands whose sizes aren't ordered.
12171 (match_reload): Refer to this check here.
12173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12174 Alan Hayward <alan.hayward@arm.com>
12175 David Sherwood <david.sherwood@arm.com>
12177 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12178 that the mode size is in the set {1, 2, 4, 8, 16}.
12180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12181 Alan Hayward <alan.hayward@arm.com>
12182 David Sherwood <david.sherwood@arm.com>
12184 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12185 Use plus_constant instead of gen_rtx_PLUS.
12187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12188 Alan Hayward <alan.hayward@arm.com>
12189 David Sherwood <david.sherwood@arm.com>
12191 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12192 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12193 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12194 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12195 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12196 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12197 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12198 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12199 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12200 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12202 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12203 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12204 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12205 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12206 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12207 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12208 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12209 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12210 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12211 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12213 * expr.c (emit_move_resolve_push): Treat the input and result
12214 of PUSH_ROUNDING as a poly_int64.
12215 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12216 (emit_push_insn): Likewise.
12217 * lra-eliminations.c (mark_not_eliminable): Likewise.
12218 * recog.c (push_operand): Likewise.
12219 * reload1.c (elimination_effects): Likewise.
12220 * rtlanal.c (nonzero_bits1): Likewise.
12221 * calls.c (store_one_arg): Likewise. Require the padding to be
12222 known at compile time.
12224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12225 Alan Hayward <alan.hayward@arm.com>
12226 David Sherwood <david.sherwood@arm.com>
12228 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12229 Use plus_constant instead of gen_rtx_PLUS.
12231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12232 Alan Hayward <alan.hayward@arm.com>
12233 David Sherwood <david.sherwood@arm.com>
12235 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12236 rather than an int.
12238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12239 Alan Hayward <alan.hayward@arm.com>
12240 David Sherwood <david.sherwood@arm.com>
12242 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12243 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12244 via stack temporaries. Treat the mode size as polynomial too.
12246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12247 Alan Hayward <alan.hayward@arm.com>
12248 David Sherwood <david.sherwood@arm.com>
12250 * expr.c (expand_expr_real_2): When handling conversions involving
12251 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12252 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12253 as a poly_uint64 too.
12255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12256 Alan Hayward <alan.hayward@arm.com>
12257 David Sherwood <david.sherwood@arm.com>
12259 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12262 Alan Hayward <alan.hayward@arm.com>
12263 David Sherwood <david.sherwood@arm.com>
12265 * combine.c (can_change_dest_mode): Handle polynomial
12266 REGMODE_NATURAL_SIZE.
12267 * expmed.c (store_bit_field_1): Likewise.
12268 * expr.c (store_constructor): Likewise.
12269 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12270 and polynomial REGMODE_NATURAL_SIZE.
12271 (gen_lowpart_common): Likewise.
12272 * reginfo.c (record_subregs_of_mode): Likewise.
12273 * rtlanal.c (read_modify_subreg_p): Likewise.
12275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12276 Alan Hayward <alan.hayward@arm.com>
12277 David Sherwood <david.sherwood@arm.com>
12279 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12280 numbers of elements.
12282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12283 Alan Hayward <alan.hayward@arm.com>
12284 David Sherwood <david.sherwood@arm.com>
12286 * match.pd: Cope with polynomial numbers of vector elements.
12288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12289 Alan Hayward <alan.hayward@arm.com>
12290 David Sherwood <david.sherwood@arm.com>
12292 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12293 in a POINTER_PLUS_EXPR.
12295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12296 Alan Hayward <alan.hayward@arm.com>
12297 David Sherwood <david.sherwood@arm.com>
12299 * omp-simd-clone.c (simd_clone_subparts): New function.
12300 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12301 (ipa_simd_modify_function_body): Likewise.
12303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12304 Alan Hayward <alan.hayward@arm.com>
12305 David Sherwood <david.sherwood@arm.com>
12307 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12308 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12309 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12310 (expand_vector_condition, vector_element): Likewise.
12311 (subparts_gt): New function.
12312 (get_compute_type): Use subparts_gt.
12313 (count_type_subparts): Delete.
12314 (expand_vector_operations_1): Use subparts_gt instead of
12315 count_type_subparts.
12317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12318 Alan Hayward <alan.hayward@arm.com>
12319 David Sherwood <david.sherwood@arm.com>
12321 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12322 (vect_compile_time_alias): ...this new function. Do the calculation
12323 on poly_ints rather than trees.
12324 (vect_prune_runtime_alias_test_list): Update call accordingly.
12326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12327 Alan Hayward <alan.hayward@arm.com>
12328 David Sherwood <david.sherwood@arm.com>
12330 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12332 (vect_schedule_slp_instance): Likewise.
12334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12335 Alan Hayward <alan.hayward@arm.com>
12336 David Sherwood <david.sherwood@arm.com>
12338 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12339 constant and extern definitions for variable-length vectors.
12340 (vect_get_constant_vectors): Note that the number of units
12341 is known to be constant.
12343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12344 Alan Hayward <alan.hayward@arm.com>
12345 David Sherwood <david.sherwood@arm.com>
12347 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12348 of units as polynomial. Choose between WIDE and NARROW based
12351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12352 Alan Hayward <alan.hayward@arm.com>
12353 David Sherwood <david.sherwood@arm.com>
12355 * tree-vect-stmts.c (simd_clone_subparts): New function.
12356 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12359 Alan Hayward <alan.hayward@arm.com>
12360 David Sherwood <david.sherwood@arm.com>
12362 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12363 vectors as polynomial. Use build_index_vector for
12364 IFN_GOMP_SIMD_LANE.
12366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12367 Alan Hayward <alan.hayward@arm.com>
12368 David Sherwood <david.sherwood@arm.com>
12370 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12371 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12372 for variable-length vectors.
12373 (vectorizable_mask_load_store): Treat the number of units as
12374 polynomial, asserting that it is constant if the condition has
12375 already been enforced.
12376 (vectorizable_store, vectorizable_load): Likewise.
12378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12379 Alan Hayward <alan.hayward@arm.com>
12380 David Sherwood <david.sherwood@arm.com>
12382 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12383 of units as polynomial. Punt if we can't tell at compile time
12384 which vector contains the final result.
12386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12387 Alan Hayward <alan.hayward@arm.com>
12388 David Sherwood <david.sherwood@arm.com>
12390 * tree-vect-loop.c (vectorizable_induction): Treat the number
12391 of units as polynomial. Punt on SLP inductions. Use an integer
12392 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12393 cast of such a series for variable-length floating-point
12396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12397 Alan Hayward <alan.hayward@arm.com>
12398 David Sherwood <david.sherwood@arm.com>
12400 * tree.h (build_index_vector): Declare.
12401 * tree.c (build_index_vector): New function.
12402 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12403 of units as polynomial, forcibly converting it to a constant if
12404 vectorizable_reduction has already enforced the condition.
12405 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12406 to create a {1,2,3,...} vector.
12407 (vectorizable_reduction): Treat the number of units as polynomial.
12408 Choose vectype_in based on the largest scalar element size rather
12409 than the smallest number of units. Enforce the restrictions
12412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12413 Alan Hayward <alan.hayward@arm.com>
12414 David Sherwood <david.sherwood@arm.com>
12416 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12417 number of units as polynomial.
12419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12420 Alan Hayward <alan.hayward@arm.com>
12421 David Sherwood <david.sherwood@arm.com>
12423 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12424 * target.def (autovectorize_vector_sizes): Return the vector sizes
12425 by pointer, using vector_sizes rather than a bitmask.
12426 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12427 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12428 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12430 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12431 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12432 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12433 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12434 * omp-general.c (omp_max_vf): Likewise.
12435 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12436 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12437 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12438 * tree-vect-slp.c (vect_slp_bb): Likewise.
12439 * doc/tm.texi: Regenerate.
12440 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12442 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12443 the vector size as a poly_uint64 rather than an unsigned int.
12444 (current_vector_size): Change from an unsigned int to a poly_uint64.
12445 (get_vectype_for_scalar_type): Update accordingly.
12446 * tree.h (build_truth_vector_type): Take the size and number of
12447 units as a poly_uint64 rather than an unsigned int.
12448 (build_vector_type): Add a temporary overload that takes
12449 the number of units as a poly_uint64 rather than an unsigned int.
12450 * tree.c (make_vector_type): Likewise.
12451 (build_truth_vector_type): Take the number of units as a poly_uint64
12452 rather than an unsigned int.
12454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12455 Alan Hayward <alan.hayward@arm.com>
12456 David Sherwood <david.sherwood@arm.com>
12458 * target.def (get_mask_mode): Take the number of units and length
12459 as poly_uint64s rather than unsigned ints.
12460 * targhooks.h (default_get_mask_mode): Update accordingly.
12461 * targhooks.c (default_get_mask_mode): Likewise.
12462 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12463 * doc/tm.texi: Regenerate.
12465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12466 Alan Hayward <alan.hayward@arm.com>
12467 David Sherwood <david.sherwood@arm.com>
12469 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12470 * omp-general.c (omp_max_vf): Likewise.
12471 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12472 (expand_omp_simd): Handle polynomial safelen.
12473 * omp-low.c (omplow_simd_context): Add a default constructor.
12474 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12475 (lower_rec_simd_input_clauses): Update accordingly.
12476 (lower_rec_input_clauses): Likewise.
12478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12479 Alan Hayward <alan.hayward@arm.com>
12480 David Sherwood <david.sherwood@arm.com>
12482 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12483 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12484 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12485 (vect_analyze_slp_cost): Likewise.
12486 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12487 (vect_model_load_cost): Likewise.
12489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12490 Alan Hayward <alan.hayward@arm.com>
12491 David Sherwood <david.sherwood@arm.com>
12493 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12494 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12495 from an unsigned int * to a poly_uint64_pod *.
12496 (calculate_unrolling_factor): New function.
12497 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12500 Alan Hayward <alan.hayward@arm.com>
12501 David Sherwood <david.sherwood@arm.com>
12503 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12504 from an unsigned int to a poly_uint64.
12505 (_loop_vec_info::slp_unrolling_factor): Likewise.
12506 (_loop_vec_info::vectorization_factor): Change from an int
12508 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12509 (vect_get_num_vectors): New function.
12510 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12511 (vect_get_num_copies): Use vect_get_num_vectors.
12512 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12513 to an unsigned int *.
12514 (vect_analyze_data_refs): Change min_vf from an int * to a
12516 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12517 than an unsigned HOST_WIDE_INT.
12518 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12519 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12520 to an unsigned int *.
12521 (vect_analyze_data_ref_dependences): Likewise.
12522 (vect_compute_data_ref_alignment): Handle polynomial vf.
12523 (vect_enhance_data_refs_alignment): Likewise.
12524 (vect_prune_runtime_alias_test_list): Likewise.
12525 (vect_shift_permute_load_chain): Likewise.
12526 (vect_supportable_dr_alignment): Likewise.
12527 (dependence_distance_ge_vf): Take the vectorization factor as a
12528 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12529 (vect_analyze_data_refs): Change min_vf from an int * to a
12531 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12532 vfm1 as a poly_uint64 rather than an int. Make the same change
12533 for the returned bound_scalar.
12534 (vect_gen_vector_loop_niters): Handle polynomial vf.
12535 (vect_do_peeling): Likewise. Update call to
12536 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12537 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12539 * tree-vect-loop.c (vect_determine_vectorization_factor)
12540 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12541 (vect_get_known_peeling_cost): Likewise.
12542 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12543 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12544 (vect_transform_loop): Likewise. Use the lowest possible VF when
12545 updating the upper bounds of the loop.
12546 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12547 rather than an int.
12548 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12549 polynomial unroll factors.
12550 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12551 (vect_make_slp_decision): Likewise.
12552 (vect_supported_load_permutation_p): Likewise, and polynomial
12554 (vect_analyze_slp_cost): Handle polynomial vf.
12555 (vect_slp_analyze_node_operations): Likewise.
12556 (vect_slp_analyze_bb_1): Likewise.
12557 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12558 than an unsigned HOST_WIDE_INT.
12559 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12560 (vectorizable_load): Handle polynomial vf.
12561 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12563 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12566 Alan Hayward <alan.hayward@arm.com>
12567 David Sherwood <david.sherwood@arm.com>
12569 * match.pd: Handle bit operations involving three constants
12570 and try to fold one pair.
12572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12574 * tree-vect-loop-manip.c: Include gimple-fold.h.
12575 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12576 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12577 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12578 Add a path that uses a step of VF instead of 1, but disable it
12580 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12581 and niters_no_overflow parameters. Update calls to
12582 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12583 Create a new SSA name if the latter choses to use a ste other
12584 than zero, and return it via niters_vector_mult_vf_var.
12585 * tree-vect-loop.c (vect_transform_loop): Update calls to
12586 vect_do_peeling, vect_gen_vector_loop_niters and
12587 slpeel_make_loop_iterate_ntimes.
12588 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12589 (vect_gen_vector_loop_niters): Update declarations after above changes.
12591 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12593 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12594 128-bit round to integer instructions.
12595 (ceil<mode>2): Likewise.
12596 (btrunc<mode>2): Likewise.
12597 (round<mode>2): Likewise.
12599 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12601 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12602 unaligned VSX load/store on P8/P9.
12603 (expand_block_clear): Allow the use of unaligned VSX
12604 load/store on P8/P9.
12606 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12608 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12610 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12611 swap associated with both a load and a store.
12613 2018-01-02 Andrew Waterman <andrew@sifive.com>
12615 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12616 * config/riscv/riscv.md (clear_cache): Use it.
12618 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12620 * web.c: Remove out-of-date comment.
12622 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12624 * expr.c (fixup_args_size_notes): Check that any existing
12625 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12626 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12627 (emit_single_push_insn): ...here.
12629 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12631 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12632 (const_vector_encoded_nelts): New function.
12633 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12634 (const_vector_int_elt, const_vector_elt): Declare.
12635 * emit-rtl.c (const_vector_int_elt_1): New function.
12636 (const_vector_elt): Likewise.
12637 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12638 of CONST_VECTOR_ELT.
12640 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12642 * expr.c: Include rtx-vector-builder.h.
12643 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12644 directly on the tree encoding.
12645 (const_vector_from_tree): Likewise.
12646 * optabs.c: Include rtx-vector-builder.h.
12647 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12648 sequence of "u" values.
12649 * vec-perm-indices.c: Include rtx-vector-builder.h.
12650 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12651 directly on the vec_perm_indices encoding.
12653 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12655 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12656 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12657 * rtx-vector-builder.h: New file.
12658 * rtx-vector-builder.c: Likewise.
12659 * rtl.h (rtx_def::u2): Add a const_vector field.
12660 (CONST_VECTOR_NPATTERNS): New macro.
12661 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12662 (CONST_VECTOR_DUPLICATE_P): Likewise.
12663 (CONST_VECTOR_STEPPED_P): Likewise.
12664 (CONST_VECTOR_ENCODED_ELT): Likewise.
12665 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12666 (unwrap_const_vec_duplicate): Likewise.
12667 (const_vec_series_p): Check for a non-duplicated vector encoding.
12668 Say that the function only returns true for integer vectors.
12669 * emit-rtl.c: Include rtx-vector-builder.h.
12670 (gen_const_vec_duplicate_1): Delete.
12671 (gen_const_vector): Call gen_const_vec_duplicate instead of
12672 gen_const_vec_duplicate_1.
12673 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12674 (gen_const_vec_duplicate): Use rtx_vector_builder.
12675 (gen_const_vec_series): Likewise.
12676 (gen_rtx_CONST_VECTOR): Likewise.
12677 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12678 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12679 Build a new vector rather than modifying a CONST_VECTOR in-place.
12680 (handle_special_swappables): Update call accordingly.
12681 * config/rs6000/rs6000-p8swap.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.
12686 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12688 * simplify-rtx.c (simplify_const_binary_operation): Use
12689 CONST_VECTOR_ELT instead of XVECEXP.
12691 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12693 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12694 the selector elements to be different from the data elements
12695 if the selector is a VECTOR_CST.
12696 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12697 ssizetype for the selector.
12699 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12701 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12702 before testing each element individually.
12703 * tree-vect-generic.c (lower_vec_perm): Likewise.
12705 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12707 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12708 * selftest-run-tests.c (selftest::run_tests): Call it.
12709 * vector-builder.h (vector_builder::operator ==): New function.
12710 (vector_builder::operator !=): Likewise.
12711 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12712 (vec_perm_indices::all_from_input_p): New function.
12713 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12714 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12715 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12716 instead of reading the VECTOR_CST directly. Detect whether both
12717 vector inputs are the same before constructing the vec_perm_indices,
12718 and update the number of inputs argument accordingly. Use the
12719 utility functions added above. Only construct sel2 if we need to.
12721 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12723 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12724 the broadcast of the low byte.
12725 (expand_mult_highpart): Use an explicit encoding for the permutes.
12726 * optabs-query.c (can_mult_highpart_p): Likewise.
12727 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12728 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12729 (vectorizable_bswap): Likewise.
12730 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12731 explicit encoding for the power-of-2 permutes.
12732 (vect_permute_store_chain): Likewise.
12733 (vect_grouped_load_supported): Likewise.
12734 (vect_permute_load_chain): Likewise.
12736 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12738 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12739 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12740 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12741 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12742 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12743 (vect_gen_perm_mask_any): Likewise.
12745 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12747 * int-vector-builder.h: New file.
12748 * vec-perm-indices.h: Include int-vector-builder.h.
12749 (vec_perm_indices): Redefine as an int_vector_builder.
12750 (auto_vec_perm_indices): Delete.
12751 (vec_perm_builder): Redefine as a stand-alone class.
12752 (vec_perm_indices::vec_perm_indices): New function.
12753 (vec_perm_indices::clamp): Likewise.
12754 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12755 (vec_perm_indices::new_vector): New function.
12756 (vec_perm_indices::new_expanded_vector): Update for new
12757 vec_perm_indices class.
12758 (vec_perm_indices::rotate_inputs): New function.
12759 (vec_perm_indices::all_in_range_p): Operate directly on the
12760 encoded form, without computing elided elements.
12761 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12762 encoding. Update for new vec_perm_indices class.
12763 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12764 the given vec_perm_builder.
12765 (expand_vec_perm_var): Update vec_perm_builder constructor.
12766 (expand_mult_highpart): Use vec_perm_builder instead of
12767 auto_vec_perm_indices.
12768 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12769 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12770 or double series encoding as appropriate.
12771 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12772 vec_perm_indices instead of auto_vec_perm_indices.
12773 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12774 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12775 (vect_permute_store_chain): Likewise.
12776 (vect_grouped_load_supported): Likewise.
12777 (vect_permute_load_chain): Likewise.
12778 (vect_shift_permute_load_chain): Likewise.
12779 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12780 (vect_transform_slp_perm_load): Likewise.
12781 (vect_schedule_slp_instance): Likewise.
12782 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12783 (vectorizable_mask_load_store): Likewise.
12784 (vectorizable_bswap): Likewise.
12785 (vectorizable_store): Likewise.
12786 (vectorizable_load): Likewise.
12787 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12788 vec_perm_indices instead of auto_vec_perm_indices. Use
12789 tree_to_vec_perm_builder to read the vector from a tree.
12790 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12791 vec_perm_builder instead of a vec_perm_indices.
12792 (have_whole_vector_shift): Use vec_perm_builder and
12793 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12794 truncation to calc_vec_perm_mask_for_shift.
12795 (vect_create_epilog_for_reduction): Likewise.
12796 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12797 from auto_vec_perm_indices to vec_perm_indices.
12798 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12799 instead of changing individual elements.
12800 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12801 the vector in d.perm.
12802 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12803 from auto_vec_perm_indices to vec_perm_indices.
12804 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12805 instead of changing individual elements.
12806 (arm_vectorize_vec_perm_const): Use new_vector to install
12807 the vector in d.perm.
12808 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12809 Update vec_perm_builder constructor.
12810 (rs6000_expand_interleave): Likewise.
12811 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12812 (rs6000_expand_interleave): Likewise.
12814 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12816 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12817 to qimode could truncate the indices.
12818 * optabs.c (expand_vec_perm_var): Likewise.
12820 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12822 * Makefile.in (OBJS): Add vec-perm-indices.o.
12823 * vec-perm-indices.h: New file.
12824 * vec-perm-indices.c: Likewise.
12825 * target.h (vec_perm_indices): Replace with a forward class
12827 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12828 * optabs.h: Include vec-perm-indices.h.
12829 (expand_vec_perm): Delete.
12830 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12831 (expand_vec_perm_const): Declare.
12832 * target.def (vec_perm_const_ok): Replace with...
12833 (vec_perm_const): ...this new hook.
12834 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12835 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12836 * doc/tm.texi: Regenerate.
12837 * optabs.def (vec_perm_const): Delete.
12838 * doc/md.texi (vec_perm_const): Likewise.
12839 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12840 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12841 expand_vec_perm for constant permutation vectors. Assert that
12842 the mode of variable permutation vectors is the integer equivalent
12843 of the mode that is being permuted.
12844 * optabs-query.h (selector_fits_mode_p): Declare.
12845 * optabs-query.c: Include vec-perm-indices.h.
12846 (selector_fits_mode_p): New function.
12847 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12848 is defined, instead of checking whether the vec_perm_const_optab
12849 exists. Use targetm.vectorize.vec_perm_const instead of
12850 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12851 fit in the vector mode before using a variable permute.
12852 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12853 vec_perm_indices instead of an rtx.
12854 (expand_vec_perm): Replace with...
12855 (expand_vec_perm_const): ...this new function. Take the selector
12856 as a vec_perm_indices rather than an rtx. Also take the mode of
12857 the selector. Update call to shift_amt_for_vec_perm_mask.
12858 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12859 Use vec_perm_indices::new_expanded_vector to expand the original
12860 selector into bytes. Check whether the indices fit in the vector
12861 mode before using a variable permute.
12862 (expand_vec_perm_var): Make global.
12863 (expand_mult_highpart): Use expand_vec_perm_const.
12864 * fold-const.c: Includes vec-perm-indices.h.
12865 * tree-ssa-forwprop.c: Likewise.
12866 * tree-vect-data-refs.c: Likewise.
12867 * tree-vect-generic.c: Likewise.
12868 * tree-vect-loop.c: Likewise.
12869 * tree-vect-slp.c: Likewise.
12870 * tree-vect-stmts.c: Likewise.
12871 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12873 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12874 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12875 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12876 (aarch64_vectorize_vec_perm_const): ...this new function.
12877 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12878 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12879 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12880 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12881 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12882 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12883 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12885 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12886 check for NEON modes.
12887 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12888 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12889 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12890 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12892 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12893 the old VEC_PERM_CONST conditions.
12894 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12895 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12896 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12897 (ia64_vectorize_vec_perm_const_ok): Merge into...
12898 (ia64_vectorize_vec_perm_const): ...this new function.
12899 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12900 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12901 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12902 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12903 * config/mips/mips.c (mips_expand_vec_perm_const)
12904 (mips_vectorize_vec_perm_const_ok): Merge into...
12905 (mips_vectorize_vec_perm_const): ...this new function.
12906 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12907 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12908 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12909 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12910 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12911 (rs6000_expand_vec_perm_const): Delete.
12912 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12914 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12915 (altivec_expand_vec_perm_const_le): Take each operand individually.
12916 Operate on constant selectors rather than rtxes.
12917 (altivec_expand_vec_perm_const): Likewise. Update call to
12918 altivec_expand_vec_perm_const_le.
12919 (rs6000_expand_vec_perm_const): Delete.
12920 (rs6000_vectorize_vec_perm_const_ok): Delete.
12921 (rs6000_vectorize_vec_perm_const): New function.
12922 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12923 an element count and rtx array.
12924 (rs6000_expand_extract_even): Update call accordingly.
12925 (rs6000_expand_interleave): Likewise.
12926 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12927 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12928 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12929 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12930 (rs6000_expand_vec_perm_const): Delete.
12931 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12932 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12933 (altivec_expand_vec_perm_const_le): Take each operand individually.
12934 Operate on constant selectors rather than rtxes.
12935 (altivec_expand_vec_perm_const): Likewise. Update call to
12936 altivec_expand_vec_perm_const_le.
12937 (rs6000_expand_vec_perm_const): Delete.
12938 (rs6000_vectorize_vec_perm_const_ok): Delete.
12939 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12940 reference to the SPE evmerge intructions.
12941 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12942 an element count and rtx array.
12943 (rs6000_expand_extract_even): Update call accordingly.
12944 (rs6000_expand_interleave): Likewise.
12945 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12946 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12948 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12950 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12952 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12953 vector mode and that that mode matches the mode of the data
12955 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12956 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12957 directly using expand_vec_perm_1 when forcing selectors into
12959 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12961 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12963 * optabs-query.h (can_vec_perm_p): Delete.
12964 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12965 * optabs-query.c (can_vec_perm_p): Split into...
12966 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12967 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12968 particular selector is valid.
12969 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12970 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12971 (vect_grouped_load_supported): Likewise.
12972 (vect_shift_permute_load_chain): Likewise.
12973 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12974 (vect_transform_slp_perm_load): Likewise.
12975 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12976 (vectorizable_bswap): Likewise.
12977 (vect_gen_perm_mask_checked): Likewise.
12978 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12979 implementations of variable permutation vectors into account
12980 when deciding which selector to use.
12981 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12982 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12983 with a false third argument.
12984 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12985 to test whether the constant selector is valid and can_vec_perm_var_p
12986 to test whether a variable selector is valid.
12988 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12990 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12991 * optabs-query.c (can_vec_perm_p): Likewise.
12992 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12993 instead of vec_perm_indices.
12994 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12995 (vect_gen_perm_mask_checked): Likewise,
12996 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12997 (vect_gen_perm_mask_checked): Likewise,
12999 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13001 * optabs-query.h (qimode_for_vec_perm): Declare.
13002 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13003 (qimode_for_vec_perm): ...this new function.
13004 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13006 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13008 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13009 does not have a conditional at the top.
13011 2018-01-02 Richard Biener <rguenther@suse.de>
13013 * ipa-inline.c (big_speedup_p): Fix expression.
13015 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13018 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13021 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13025 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13026 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13027 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13028 cond_taken_branch_cost 3->4.
13030 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13032 PR tree-optimization/83581
13033 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13034 TODO_cleanup_cfg if any changes have been made.
13036 PR middle-end/83608
13037 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13038 convert_modes if target mode has the right side, but different mode
13041 PR middle-end/83609
13042 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13043 last argument when extracting from CONCAT. If either from_real or
13044 from_imag is NULL, use expansion through memory. If result is not
13045 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13046 the parts directly to inner mode, if even that fails, use expansion
13049 PR middle-end/83623
13050 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13051 check for bswap in mode rather than HImode and use that in expand_unop
13054 Copyright (C) 2018 Free Software Foundation, Inc.
13056 Copying and distribution of this file, with or without modification,
13057 are permitted in any medium without royalty provided the copyright
13058 notice and this notice are preserved.