1 2018-05-16 Mark Wielaard <mark@klomp.org>
3 * dwarf2out.c (dwarf_FORM): New function.
4 (set_indirect_string): Use dwarf_FORM.
5 (reset_indirect_string): Likewise.
6 (size_of_die): Likewise.
7 (value_format): Likewise.
8 (output_die): Likewise.
9 (add_skeleton_AT_string): Likewise.
10 (output_macinfo_op): Likewise.
11 (index_string): Likewise.
12 (output_index_string_offset): Likewise.
13 (output_index_string): Likewise.
14 (count_index_strings): Likewise.
16 2018-05-16 Carl Love <cel@us.ibm.com>
18 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
21 2018-05-16 Martin Jambor <mjambor@suse.cz>
23 * ipa-prop.c (ipa_free_all_edge_args): Remove.
24 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
26 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
28 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
29 (fnma<mode>4): Likewise.
30 (fms<mode>4): Likewise.
31 (fnms<mode>4): Likewise.
32 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
33 (aarch64_fnma<mode>4): Likewise.
34 (aarch64_fms<mode>4): Likewise.
35 (aarch64_fnms<mode>4): Likewise.
36 (aarch64_fnmadd<mode>4): Likewise.
38 2018-05-16 Jason Merrill <jason@redhat.com>
40 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
42 2018-05-16 Richard Biener <rguenther@suse.de>
44 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
45 (dump_stmt_cost): Declare.
46 (add_stmt_cost): Dump cost we add.
47 (add_stmt_costs): New function.
48 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
50 (vect_analyze_stmt): Adjust prototype.
51 (vectorizable_condition): Likewise.
52 (vectorizable_live_operation): Likewise.
53 (vectorizable_reduction): Likewise.
54 (vectorizable_induction): Likewise.
55 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
56 cost vector to pass to vectorizable_ and record afterwards.
57 (vect_model_reduction_cost): Take cost vector argument and adjust.
58 (vect_model_induction_cost): Likewise.
59 (vectorizable_reduction): Likewise.
60 (vectorizable_induction): Likewise.
61 (vectorizable_live_operation): Likewise.
62 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
63 SLP_TREE_NUMBER_OF_VEC_STMTS.
64 (vect_analyze_slp_cost_1): Remove.
65 (vect_analyze_slp_cost): Likewise.
66 (vect_slp_analyze_node_operations): Take visited args and
67 a target cost vector. Avoid processing already visited stmt sets.
68 (vect_slp_analyze_operations): Use a local cost vector to gather
69 costs and register those of non-discarded instances.
70 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
71 (vect_schedule_slp_instance): Remove copying of
72 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
74 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
75 adding cost. Record cost entry location.
76 (vect_prologue_cost_for_slp_op): Function to compute cost of
77 a constant or invariant generated for SLP vect in the prologue,
78 split out from vect_analyze_slp_cost_1.
79 (vect_model_simple_cost): Make static. Adjust for SLP costing.
80 (vect_model_promotion_demotion_cost): Likewise.
81 (vect_model_store_cost): Likewise, make static.
82 (vect_model_load_cost): Likewise.
83 (vectorizable_bswap): Add cost vector arg and adjust.
84 (vectorizable_call): Likewise.
85 (vectorizable_simd_clone_call): Likewise.
86 (vectorizable_conversion): Likewise.
87 (vectorizable_assignment): Likewise.
88 (vectorizable_shift): Likewise.
89 (vectorizable_operation): Likewise.
90 (vectorizable_store): Likewise.
91 (vectorizable_load): Likewise.
92 (vectorizable_condition): Likewise.
93 (vectorizable_comparison): Likewise.
94 (can_vectorize_live_stmts): Likewise.
95 (vect_analyze_stmt): Likewise.
96 (vect_transform_stmt): Adjust calls to vectorizable_*.
97 * tree-vectorizer.c: Include gimple-pretty-print.h.
98 (dump_stmt_cost): New function.
100 2018-05-16 Richard Biener <rguenther@suse.de>
102 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
103 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
104 * tree-ssa-dse.c: Include tree-ssa-loop.h.
105 (check_name): New callback.
106 (dse_classify_store): Track cycles via a visited bitmap of PHI
107 defs and simplify handling of in-loop and across loop dead stores
108 and properly fail for loop-variant refs. Handle byte-tracking with
109 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
112 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
114 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
115 (vect_get_mask_type_for_stmt): Likewise.
116 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
118 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
119 to determine the statement's vector type and the vector type that
120 should be used for calculating nunits. Deal with cases in which
121 the type has to be deferred.
122 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
123 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
124 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
125 (vect_determine_vf_for_stmt): New functions, split out from...
126 (vect_determine_vectorization_factor): ...here.
127 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
128 (vect_get_mask_type_for_stmt): New functions, split out from
129 vect_determine_vectorization_factor.
131 2018-05-16 Richard Biener <rguenther@suse.de>
133 * tree-cfg.c (verify_gimple_assign_ternary): Properly
134 verify the [VEC_]COND_EXPR embedded comparison.
136 2018-05-15 Martin Sebor <msebor@redhat.com>
138 PR tree-optimization/85753
139 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
140 RECORD_TYPE in addition to ARRAY_TYPE.
142 2018-05-15 Martin Sebor <msebor@redhat.com>
145 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
147 2018-05-15 Richard Biener <rguenther@suse.de>
149 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
150 add by_clobber_p one. Change algorithm to collect all defs
151 representing uses we need to walk and try reducing them to
152 a single one before failing.
153 (dse_dom_walker::dse_optimize_stmt): Adjust.
155 2018-05-13 Mark Wielaard <mark@klomp.org>
157 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
158 (size_of_loc_descr): Likewise.
159 (output_loc_operands): Likewise.
160 (output_loc_operands_raw): Likewise.
161 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
162 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
163 (hash_loc_operands): Likewise.
164 (compare_loc_operands): Likewise.
166 2018-05-14 Mark Wielaard <mark@klomp.org>
168 * dwarf2out.c (count_index_addrs): New function.
169 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
171 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
173 PR tree-optimization/83648
174 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
175 return value as malloc candidate.
177 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
180 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
181 as true in call to suggest_attribute.
183 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
185 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
188 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
190 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
192 (aarch64_crypto_aesd_fused): Likewise.
194 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
196 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
197 (movsi_aarch64): Likewise.
198 (load_pairsi): Likewise.
199 (load_pairdi): Likewise.
200 (store_pairsi): Likewise.
201 (store_pairdi): Likewise.
202 (load_pairsf): Likewise.
203 (load_pairdf): Likewise.
204 (store_pairsf): Likewise.
205 (store_pairdf): Likewise.
206 (zero_extend): Likewise.
207 (trunc): Swap alternatives.
208 (fcvt_target): Add '?' to prefer w over r.
210 2018-05-14 Jakub Jelinek <jakub@redhat.com>
213 * config/i386/i386.md: Disallow non-commutative arithmetics in
214 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
215 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
216 in the peephole2 before it.
218 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
220 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
221 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
222 (ix86_handle_option): Handle -mcldemote.
223 * config.gcc: New header.
224 * config/i386/cldemoteintrin.h: New file.
225 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
226 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
228 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
229 OPTION_MASK_ISA_CLDEMOTE.
230 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
231 (ix86_valid_target_attribute_inner_p): Ditto.
232 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
233 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
234 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
235 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
236 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
238 * config/i386/i386.opt: Add -mcldemote.
239 * config/i386/x86intrin.h: New header.
240 * doc/invoke.texi: Add -mcldemote.
242 2018-05-14 Richard Biener <rguenther@suse.de>
244 * doc/match-and-simplify.texi: Adjust :s documentation.
246 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
248 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
249 intended memcpy size.
250 (REORDER_45): Likewise.
252 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
255 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
256 * vec.c (qsort_chk): Use gcc_qsort.
257 * Makefile.in (OBJS-libcommon): Add sort.o.
258 (build/sort.o): New target. Use it...
259 (BUILD_RTL): ... here, and...
260 (build/gencfn-macros): ... here, and...
261 (build/genmatch): ... here.
263 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
264 Chung-Ju Wu <jasonwucj@gmail.com>
266 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
267 * config/nds32/nds32-graywolf.md: New file.
268 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
269 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
271 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
272 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
273 * config/nds32/nds32.md (pipeline_model): Add graywolf.
274 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
275 * config/nds32/pipelines.md: Include n15 settings.
277 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
278 Chung-Ju Wu <jasonwucj@gmail.com>
280 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
281 * config/nds32/nds32-n13.md: New file.
282 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
283 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
285 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
286 * config/nds32/nds32.md (pipeline_model): Add n13.
287 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
288 * config/nds32/pipelines.md: Include n13 settings.
290 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
291 Chung-Ju Wu <jasonwucj@gmail.com>
293 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
294 * config/nds32/nds32-n10.md: New file.
295 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
296 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
298 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
299 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
300 * config/nds32/nds32.md (pipeline_model): Add n10.
301 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
302 * config/nds32/pipelines.md: Include n10 settings.
304 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
305 Kito Cheng <kito.cheng@gmail.com>
306 Chung-Ju Wu <jasonwucj@gmail.com>
308 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
309 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
310 Add enum values for DSP extension instructions.
311 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
313 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
314 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
316 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
317 * config/nds32/nds32-dspext.md: New file for DSP implementation.
318 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
319 * config/nds32/nds32-intrinsic.md: Likewise.
320 * config/nds32/nds32_intrinsic.h: Likewise.
321 * config/nds32/nds32-md-auxiliary.c: Likewise.
322 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
323 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
324 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
325 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
326 * config/nds32/nds32-protos.h: New declarations for DSP extension.
327 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
328 TYPE_DMAC in switch statement.
329 * config/nds32/nds32.c: New checking and implementation for DSP
330 extension instructions.
331 * config/nds32/nds32.h: Likewise.
332 * config/nds32/nds32.md: Likewise.
333 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
334 * config/nds32/predicates.md: Implement new predicates for DSP
337 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
339 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
340 Reformat alternatives and attributes so it is easier to identify
341 which constraints/attributes go with which instruction.
342 (mov<mode>_hardfloat32, FMOVE64): Likewise.
343 (mov<mode>_softfloat32, FMOVE64): Likewise.
344 (mov<mode>_hardfloat64, FMOVE64): Likewise.
345 (mov<mode>_softfloat64, FMOVE64): Likewise.
347 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
349 * doc/extend.texi (PowerPC Built-in Functions): Rename this
351 (Basic PowerPC Built-in Functions): The new name of the
352 subsection previously known as "PowerPC Built-in Functions".
353 (Basic PowerPC Built-in Functions Available on all Configurations):
355 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
356 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
357 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
358 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
360 2018-05-11 Martin Jambor <mjambor@suse.cz>
363 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
366 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
369 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
371 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
373 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
374 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
375 (ix86_handle_option): Handle -mwaitpkg.
376 * config.gcc: New header.
377 * config/i386/cpuid.h (bit_WAITPKG): New bit.
378 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
379 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
381 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
382 OPTION_MASK_ISA_WAITPKG.
383 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
384 (ix86_option_override_internal): Add PTA_WAITPKG.
385 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
386 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
388 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
389 __builtin_ia32_umwait and __builtin_ia32_tpause.
390 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
391 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
392 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
393 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
394 UNSPECV_TPAUSE): New.
395 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
396 * config/i386/i386.opt: Add -mwaitpkg.
397 * config/i386/waitpkgintrin.h: New file.
398 * config/i386/x86intrin.h: New header.
399 * doc/invoke.texi: Add -mwaitpkg.
401 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
404 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
406 (cortex-m0): Use armv6s-m isa.
407 (cortex-m0plus): Likewise.
408 (cortex-m1): Likewise.
409 (cortex-m0.small-multiply): Likewise.
410 (cortex-m0plus.small-multiply): Likewise.
411 (cortex-m1.small-multiply): Likewise.
413 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
414 Jakub Jelinek <jakub@redhat.com>
416 PR tree-optimization/85692
417 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
418 source permute as well.
420 2018-05-11 Martin Liska <mliska@suse.cz>
423 * doc/extend.texi: Document LLVM style format for no_sanitize
426 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
428 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
429 mode_supports_vsx_dform_quad to mode_supports_dq_form.
430 (mode_supports_vsx_dform_quad): Likewise.
431 (mode_supports_vmx_dform): Move these functions to be next to the
432 other mode_supports functions.
433 (mode_supports_dq_form): Likewise.
434 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
435 mode_supports_dq_form.
436 (reg_offset_addressing_ok_p): Likewise.
437 (offsettable_ok_by_alignment): Likewise.
438 (rs6000_legitimate_offset_address_p): Likewise.
439 (legitimate_lo_sum_address_p): Likewise.
440 (rs6000_legitimize_address): Likewise.
441 (rs6000_legitimize_reload_address): Likewise.
442 (rs6000_secondary_reload_inner): Likewise.
443 (rs6000_preferred_reload_class): Likewise.
444 (rs6000_output_move_128bit): Likewise.
446 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
448 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
449 Generate SImode target register for null target.
450 <case IX86_BUILTIN_XGETBV>: Ditto.
451 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
452 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
454 2018-05-10 Carl Love <cel@us.ibm.com>
456 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
457 dcbtt and dcbtstt if operands[2] is 0.
459 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
462 * config/i386/sse.md (usadv64qi): New expander.
464 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
466 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
467 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
468 -maltivec=be support.
469 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
470 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
471 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
472 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
473 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
474 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
475 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
476 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
477 altivec_vsumsws): Adjust.
478 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
479 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
480 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
482 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
483 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
484 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
485 (altivec_lve<VI_char>x): Delete expand.
486 (*altivec_lve<VI_char>x_internal): Rename to...
487 (altivec_lve<VI_char>x): ... this.
488 (altivec_lvxl_<mode>): Delete expand.
489 (*altivec_lvxl_<mode>_internal): Rename to ...
490 (altivec_lvxl_<mode>): ... this.
491 (altivec_stvxl_<mode>): Delete expand.
492 (*altivec_stvxl_<mode>_internal): Rename to ...
493 (altivec_stvxl_<mode>): ... this.
494 (altivec_stve<VI_char>x): Delete expand.
495 (*altivec_stve<VI_char>x_internal): Rename to ...
496 (altivec_stve<VI_char>x): ... this.
497 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
498 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
499 reduc_plus_scal_<mode>): Adjust.
500 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
502 (rs6000_cpu_cpp_builtins): Adjust.
503 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
504 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
505 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
506 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
507 -maltivec=be support.
508 (rs6000_split_vec_extract_var): Adjust.
509 (rs6000_split_v4si_init): Adjust.
510 (swap_selector_for_mode): Delete.
511 (altivec_expand_lvx_be, altivec_expand_stvx_be,
512 altivec_expand_stvex_be): Delete.
513 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
514 -maltivec=be support.
515 (rs6000_gimple_fold_builtin): Ditto.
516 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
518 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
519 (TARGET_DIRECT_MOVE_64BIT): Adjust.
520 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
521 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
522 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
523 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
524 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
525 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
526 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
527 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
528 anonymous split): Adjust.
529 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
530 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
532 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
534 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
535 when --with-gxx-include-dir is also specified.
536 * configure: Regenerate.
538 2018-05-09 Jim Wilson <jimw@sifive.com>
541 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
542 * config/riscv/t-withmultilib: New.
543 * config/riscv/withmultilib.h: New.
544 * doc/install.texi: Document RISC-V --with-multilib-list support.
546 2018-05-09 Richard Biener <rguenther@suse.de>
548 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
550 (vect_bb_vectorization_profitable_p): Adjust. Compute
551 actual scalar cost using the cost vector and the add_stmt_cost
554 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
556 PR rtl-optimization/85645
557 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
558 in the REG_CFA_REGISTER note for LR, don't leave it empty.
560 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
562 PR rtl-optimization/85645
563 * shrink-wrap.c (spread_components): Return a boolean saying if
564 anything was changed.
565 (try_shrink_wrapping_separate): Iterate spread_components until
566 nothing changes anymore.
568 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
570 PR rtl-optimization/85645
571 * regrename.c (build_def_use): Also kill the chains that include the
572 destination of a REG_CFA_REGISTER note.
574 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
576 PR rtl-optimization/85645
577 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
578 insn that has a REG_CFA_REGISTER note.
580 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
582 * cfgexpand.c (expand_clobber): New function.
583 (expand_gimple_stmt_1): Use it.
584 * tree-vect-stmts.c (vect_clobber_variable): New function,
586 (vectorizable_simd_clone_call): ...here.
587 (vectorizable_store): Emit a clobber either side of an
588 IFN_STORE_LANES sequence.
589 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
591 2018-05-09 Tom de Vries <tom@codesourcery.com>
594 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
595 (define_insn "trap_if_false"): Add exit after trap.
597 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
599 PR rtl-optimization/85638
600 * bb-reorder.c: Include common/common-target.h.
601 (create_forwarder_block): New function extracted from...
602 (fix_up_crossing_landing_pad): ...here. Rename into...
603 (dw2_fix_up_crossing_landing_pad): ...this.
604 (sjlj_fix_up_crossing_landing_pad): New function.
605 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
606 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
607 from both partitions and exit the loop after one iteration.
609 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
612 * doc/extend.texi (PowerPC Built-in Functions): Rename this
614 (Basic PowerPC Built-in Functions): The new name of the
615 subsection previously known as "PowerPC Built-in Functions".
616 (Basic PowerPC Built-in Functions Available on all Configurations):
618 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
620 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
622 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
624 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
627 2018-05-08 Jim Wilson <jimw@sifive.com>
629 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
630 (LD_EMUL_SUFFIX): New.
633 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
635 * doc/extend.texi (PowerPC Built-in Functions): Rename this
637 (Basic PowerPC Built-in Functions): The new name of the
638 subsection previously known as "PowerPC Built-in Functions".
639 (Basic PowerPC Built-in Functions Available on all Configurations):
641 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
643 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
645 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
647 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
650 2018-05-08 Jakub Jelinek <jakub@redhat.com>
653 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
654 after cmpelim optimization.
656 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
658 * config.gcc: Support "goldmont".
659 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
660 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
662 * config/i386/i386.c (m_GOLDMONT): Define.
663 (processor_target_table): Add "goldmont".
664 (PTA_GOLDMONT): Define.
665 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
666 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
667 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
668 (fold_builtin_cpu): Add "goldmont".
669 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
670 (ix86_option_override_internal): Add "goldmont".
671 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
672 (processor_type): Add PROCESSOR_GOLDMONT.
673 * config/i386/i386.md: Add CPU "glm".
674 * config/i386/glm.md: New file.
675 * config/i386/x86-tune.def: Add m_GOLDMONT.
676 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
678 2018-05-08 Jakub Jelinek <jakub@redhat.com>
681 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
683 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
684 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
685 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
688 * config/i386/i386.c (ix86_fold_builtin): Handle
689 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
692 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
693 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
695 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
698 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
699 (check_arch): Likewise.
700 (check_fpu): Return the result rather than printing it.
701 (end arch): Fix operator precedence.
703 (END): Print the result from check_fpu.
705 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
706 Alan Hayward <alan.hayward@arm.com>
707 David Sherwood <david.sherwood@arm.com>
709 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
710 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
711 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
712 (*fcmuo<mode>_and): New patterns.
714 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
716 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
717 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
718 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
719 (cmp_op, sve_imm_con): New code attributes.
720 (SVE_COND_INT_CMP, imm_con): Delete.
721 (cmp_op): Remove above unspecs from int attribute.
722 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
724 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
725 comparison-specific unspecs.
726 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
727 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
728 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
729 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
730 (*vec_fcm<cmp_op><mode>): Rename to...
731 (*fcm<cmp_op><mode>): ...this and adjust likewise.
732 (*vec_fcmuo<mode>): Rename to...
733 (*fcmuo<mode>): ...this and adjust likewise.
734 (*pred_fcm<cmp_op><mode>): New pattern.
735 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
736 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
738 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
740 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
741 (aarch64_emit_sve_predicated_cond): New function.
742 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
743 (aarch64_emit_unspec_cond_or): Replace with...
744 (aarch64_emit_sve_or_conds): ...this new function. Use
745 aarch64_emit_sve_ptrue_op for the individual comparisons and
746 aarch64_emit_binop to OR them together.
747 (aarch64_emit_inverted_unspec_cond): Replace with...
748 (aarch64_emit_sve_inverted_cond): ...this new function. Use
749 aarch64_emit_sve_ptrue_op for the comparison and
750 aarch64_emit_unop to invert the result.
751 (aarch64_expand_sve_vec_cmp_float): Update after the above
752 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
754 2018-05-07 Nathan Sidwell <nathan@acm.org>
756 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
757 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
758 (Backwards Compatibility): Likewise.
760 2018-05-07 Luis Machado <luis.machado@linaro.org>
764 2018-05-07 Luis Machado <luis.machado@linaro.org>
766 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
767 <prefetch_dynamic_strides>: New const bool field.
768 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
769 prefetch_dynamic_strides.
770 (exynosm1_prefetch_tune): Likewise.
771 (thunderxt88_prefetch_tune): Likewise.
772 (thunderx_prefetch_tune): Likewise.
773 (thunderx2t99_prefetch_tune): Likewise.
774 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
775 (aarch64_override_options_internal): Update to set
776 PARAM_PREFETCH_DYNAMIC_STRIDES.
777 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
778 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
779 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
780 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
781 prefetch-dynamic-strides setting.
783 2018-05-07 Luis Machado <luis.machado@linaro.org>
785 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
786 <minimum_stride>: New const int field.
787 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
788 minimum_stride field.
789 (exynosm1_prefetch_tune): Likewise.
790 (thunderxt88_prefetch_tune): Likewise.
791 (thunderx_prefetch_tune): Likewise.
792 (thunderx2t99_prefetch_tune): Likewise.
793 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
794 (aarch64_override_options_internal): Update to set
795 PARAM_PREFETCH_MINIMUM_STRIDE.
796 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
797 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
798 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
799 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
800 stride is constant and is below the minimum stride threshold.
802 2018-05-07 Luis Machado <luis.machado@linaro.org>
804 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
807 2018-05-07 Luis Machado <luis.machado@linaro.org>
809 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
810 <prefetch_dynamic_strides>: New const bool field.
811 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
812 prefetch_dynamic_strides.
813 (exynosm1_prefetch_tune): Likewise.
814 (thunderxt88_prefetch_tune): Likewise.
815 (thunderx_prefetch_tune): Likewise.
816 (thunderx2t99_prefetch_tune): Likewise.
817 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
818 (aarch64_override_options_internal): Update to set
819 PARAM_PREFETCH_DYNAMIC_STRIDES.
820 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
821 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
822 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
823 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
824 prefetch-dynamic-strides setting.
826 2018-05-07 Luis Machado <luis.machado@linaro.org>
828 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
829 <minimum_stride>: New const int field.
830 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
831 minimum_stride field.
832 (exynosm1_prefetch_tune): Likewise.
833 (thunderxt88_prefetch_tune): Likewise.
834 (thunderx_prefetch_tune): Likewise.
835 (thunderx2t99_prefetch_tune): Likewise.
836 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
837 (aarch64_override_options_internal): Update to set
838 PARAM_PREFETCH_MINIMUM_STRIDE.
839 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
840 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
841 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
842 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
843 stride is constant and is below the minimum stride threshold.
845 2018-05-06 Jakub Jelinek <jakub@redhat.com>
848 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
849 the type is addressable. Don't force op into register if it has
852 2018-05-05 Roland McGrath <mcgrathr@google.com>
855 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
856 any section for which we don't know a specific type it should have,
857 regardless of name. Previously this was done only for the exact
858 names ".init_array", ".fini_array", and ".preinit_array".
859 (default_elf_asm_named_section): Add comment about
860 relationship with default_section_type_flags and SECTION_NOTYPE.
861 (get_section): Don't consider it a type conflict if one side has
862 SECTION_NOTYPE and the other doesn't, as long as neither has the
863 SECTION_BSS et al used in the default_section_type_flags logic.
865 2018-05-05 Tom de Vries <tom@codesourcery.com>
868 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
869 (workaround_barsyncs): New function.
870 (nvptx_reorg): Use workaround_barsyncs.
871 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
872 (define_expand "nvptx_membar_cta"): New define_expand.
873 (define_insn "*nvptx_membar_cta"): New insn.
875 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
877 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
878 To improve optimization opportunities.
879 * builtin-types.def: The new needed builtin types for the above.
881 2018-05-04 Richard Biener <rguenther@suse.de>
883 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
884 * gimple-ssa-store-merging.c
885 (imm_store_chain_info::output_merged_store): Remove redundant create,
886 release split_store vector contents on failure.
887 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
888 scalar stmt vector on cache hit.
890 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
892 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
894 * config.gcc (powerpc-xilinx-eabi*): Remove.
895 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
897 (fusion_addis_mem_combo_load): Ditto.
898 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
900 (rs6000_cpu_cpp_builtins): Ditto.
901 * config/rs6000/rs6000-linux.c
902 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
903 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
904 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
906 (rs6000_setup_reg_addr_masks): Ditto.
907 (rs6000_init_hard_regno_mode_ok): Ditto.
908 (rs6000_option_override_internal): Ditto.
909 (legitimate_lo_sum_address_p): Ditto.
910 (rs6000_legitimize_address): Ditto.
911 (rs6000_legitimize_reload_address): Ditto.
912 (rs6000_legitimate_address_p): Ditto.
913 (abi_v4_pass_in_fpr): Ditto.
914 (setup_incoming_varargs): Ditto.
915 (rs6000_gimplify_va_arg): Ditto.
916 (rs6000_split_multireg_move): Ditto.
917 (rs6000_savres_strategy): Ditto.
918 (rs6000_emit_prologue_components): Ditto.
919 (rs6000_emit_epilogue_components): Ditto.
920 (rs6000_emit_prologue): Ditto.
921 (rs6000_emit_epilogue): Ditto.
922 (rs6000_elf_file_end): Ditto.
923 (rs6000_function_value): Ditto.
924 (rs6000_libcall_value): Ditto.
925 * config/rs6000/rs6000.h: Ditto.
926 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
927 (TARGET_MINMAX): ... this. New.
928 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
929 * config/rs6000/rs6000.md: Remove Xilinx FP support.
930 (*movsi_internal1_single): Delete.
931 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
932 mfpu=, mxilinx-fpu): Delete.
933 * config/rs6000/singlefp.h: Delete.
934 * config/rs6000/sysv4.h: Remove Xilinx FP support.
935 * config/rs6000/t-rs6000: Ditto.
936 * config/rs6000/t-xilinx: Delete.
937 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
938 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
939 (VStype_simple): Delete.
940 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
941 * config/rs6000/xfpu.h: Delete.
942 * config/rs6000/xfpu.md: Delete.
943 * config/rs6000/xilinx.h: Delete.
944 * config/rs6000/xilinx.opt: Delete.
945 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
946 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
948 2018-05-04 Tom de Vries <tom@codesourcery.com>
951 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
954 2018-05-04 Richard Biener <rguenther@suse.de>
957 * tree-complex.c (update_complex_assignment): We are always in SSA form.
958 (expand_complex_div_wide): Likewise.
959 (expand_complex_operations_1): Likewise.
960 (expand_complex_libcall): Preserve EH info of the original stmt.
961 (tree_lower_complex): Handle removed blocks.
962 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
963 on complex multiplication and division libcall builtins.
965 2018-05-04 Richard Biener <rguenther@suse.de>
968 * fold-const.c (negate_expr_p): Restrict negation of operand
969 zero of a division to when we know that can happen without
971 (fold_negate_expr_1): Likewise.
973 2018-05-04 Jakub Jelinek <jakub@redhat.com>
976 * real.h (real_nextafter): Declare.
977 * real.c (real_nextafter): New function.
978 * fold-const-call.c (fold_const_nextafter): New function.
979 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
981 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
982 even when arg1_mode is different from arg0_mode.
984 2018-05-03 Nathan Sidwell <nathan@acm.org>
986 * doc/extend.texi (Deprecated Features): Remove
988 (Backwards Compatibility): Likewise.
989 * doc/invoke.texi (C++ Language Options): Likewise.
990 (C++ Dialect Options): Likewise.
992 2018-05-03 Jakub Jelinek <jakub@redhat.com>
995 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
996 _mm512_mask_mullox_epi64): New intrinsics.
998 2018-05-03 Tom de Vries <tom@codesourcery.com>
1001 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1002 dump files): Add offload-tree.
1004 2018-05-03 Richard Biener <rguenther@suse.de>
1006 PR tree-optimization/85615
1007 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1008 to loops not nested in BBs loop father to avoid creating multi-entry
1011 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1013 PR tree-optimization/70291
1014 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1015 arguments. Change return type to tree. Emit libcall as a new
1016 statement rather than replacing existing one when inplace_p is true.
1017 (expand_complex_multiplication_components): New function.
1018 (expand_complex_multiplication): Expand floating-point complex
1019 multiplication using the above.
1020 (expand_complex_division): Rename inner_type parameter to type.
1021 Update expand_complex_libcall call-site.
1022 (expand_complex_operations_1): Update expand_complex_multiplication
1023 and expand_complex_division call-sites.
1025 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1028 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1029 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1030 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1031 the highest significant bit of the shift count mask is clear. In
1032 check whether and[sq]i3 is needed verify that all significant bits
1033 of the shift count other than the highest are set.
1035 2018-05-02 Tom de Vries <tom@codesourcery.com>
1038 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1039 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1040 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1041 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1042 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1043 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1044 __builtin_goacc_parlevel_size.
1046 2018-05-02 Richard Biener <rguenther@suse.de>
1048 PR tree-optimization/85597
1049 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1050 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1053 2018-05-02 Tom de Vries <tom@codesourcery.com>
1056 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1057 dump files): Add ltrans-tree.
1059 2018-05-02 Tom de Vries <tom@codesourcery.com>
1062 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1063 dump files): Add wpa-ipa.
1065 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1067 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1068 powerpc*-*-linux*paired* target.
1069 * config/rs6000/750cl.h: Delete.
1070 * config/rs6000/paired.h: Delete.
1071 * config/rs6000/paired.md: Delete.
1072 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1074 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1075 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1076 comment. Remove paired float support.
1077 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1078 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1080 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1081 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1083 * config/rs6000/rs6000.c: Remove paired float support.
1084 (paired_expand_vector_init, paired_expand_vector_move,
1085 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1086 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1087 paired_expand_builtin, paired_expand_predicate_builtin,
1088 paired_init_builtins): Delete.
1089 * config/rs6000/rs6000.h: Remove paired float support.
1090 * config/rs6000/rs6000.md: Remove paired float support.
1091 (move_from_CR_ov_bit): Delete.
1092 * config/rs6000/rs6000.opt (mpaired): Delete.
1093 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1094 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1096 2018-05-02 Richard Biener <rguenther@suse.de>
1099 * gimplify.c (gimplify_save_expr): When in SSA form allow
1100 SAVE_EXPRs to compute to SSA vars.
1102 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1105 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1106 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1107 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1108 clobber operands[2], instead use a new pseudo. Formatting fixes.
1110 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1112 PR tree-optimization/85586
1113 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1114 exit early for statements in the same group if the accesses are
1117 2018-05-02 Tom de Vries <tom@codesourcery.com>
1120 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1123 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1125 PR tree-optimization/85143
1126 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1128 2018-05-01 Tom de Vries <tom@codesourcery.com>
1131 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1132 not found" error message.
1134 2018-05-01 Tom de Vries <tom@codesourcery.com>
1137 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1138 * vec.c (test_ordered_remove_if): New function.
1139 (vec_c_tests): Call test_ordered_remove_if.
1140 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1141 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1142 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1143 VEC_ORDERED_REMOVE_IF.
1145 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1147 PR tree-optimization/82665
1148 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1149 pointer subtraction where arguments come from a memchr call.
1151 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1153 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1154 --push-state --as-needed and --pop-state instead of --as-needed and
1155 --no-as-needed if ld supports it.
1156 * configure: Regenerated.
1159 * doc/install.texi2html: Replace _002d with - and _002a with * in
1160 generated html files using sed.
1162 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1165 * gcc-rich-location.c (blank_line_before_p): New function.
1166 (use_new_line): New function.
1167 (gcc_rich_location::add_fixit_insert_formatted): New function.
1168 * gcc-rich-location.h
1169 (gcc_rich_location::add_fixit_insert_formatted): New function.
1171 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1173 * selftest.c (assert_streq): Rename "expected" and "actual" to
1174 "val1" and "val2". Extend NULL-handling to cover both inputs
1175 symmetrically, while still requiring both to be non-NULL for a pass.
1176 * selftest.h (assert_streq): Rename "expected" and "actual" to
1178 (ASSERT_EQ): Likewise.
1179 (ASSERT_EQ_AT): Likewise.
1180 (ASSERT_KNOWN_EQ): Likewise.
1181 (ASSERT_KNOWN_EQ_AT): Likewise.
1182 (ASSERT_NE): Likewise.
1183 (ASSERT_MAYBE_NE): Likewise.
1184 (ASSERT_MAYBE_NE_AT): Likewise.
1185 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1186 the assertion to pass.
1187 (ASSERT_STREQ_AT): Likewise.
1189 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1191 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1192 interaction with -pie.
1194 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1196 * selftest.h: Fix alphabetization of per-source-file selftest
1199 2018-04-30 Jason Merrill <jason@redhat.com>
1201 PR c++/61982 - dead stores to destroyed objects.
1202 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1205 2018-04-30 Jason Merrill <jason@redhat.com>
1207 * tree.c (build_clobber): New.
1208 * tree.h: Declare it.
1209 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1211 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1213 * diagnostic-show-locus.c (layout::layout): Update for
1214 location_get_source_line returning a char_span.
1215 (struct char_span): Move to input.h.
1216 (struct correction): Update for fields in char_span becoming
1218 (struct source_line): Update for location_get_source_line
1219 returning a char_span.
1220 (layout::print_line): Likewise.
1221 * edit-context.c (edited_file::print_content): Likewise.
1222 (edited_file::print_diff_hunk): Likewise.
1223 (edited_file::print_run_of_changed_lines): Likewise.
1224 (edited_file::get_num_lines): Likewise.
1225 (edited_line::edited_line): Likewise.
1226 * final.c (asm_show_source): Likewise.
1227 * input.c (location_get_source_line): Convert return type
1228 from const char * to char_span, losing the final "line_len"
1230 (dump_location_info): Update for the above.
1231 (get_substring_ranges_for_loc): Likewise. Use a char_span
1232 when handling the literal within the line.
1233 (test_reading_source_line): Update for location_get_source_line
1234 returning a char_span.
1235 * input.h (class char_span): Move here from
1236 diagnostic-show-locus.c, converting from a struct to a class.
1237 Make data members private.
1238 (char_span::operator bool): New.
1239 (char_span::length): New.
1240 (char_span::get_buffer): New.
1241 (char_span::operator[]): New.
1242 (char_span::subspan): Make const.
1243 (char_span::xstrdup): New.
1244 (location_get_source_line): Convert return type from const char *
1245 to char_span, losing the final "line_size" param.
1247 2018-04-30 Jan Hubicka <jh@suse.cz>
1249 * lto-wrapper.c (ltrans_priorities): New static var.
1250 (cmp_priority): New.
1251 (run_gcc): Read priorities and if doing parallel build order
1252 the Makefile by them.
1254 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1256 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1258 2018-04-30 Richard Biener <rguenther@suse.de>
1260 * tree-cfg.c (verify_address): Remove base argument, add
1261 flag whether to check TREE_ADDRESSABLE and do that.
1262 (verify_expr): Remove.
1263 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1264 (verify_gimple_assign_single): Likewise.
1265 (verify_gimple_switch): Likewise.
1266 (verify_expr_location_1): Dereference tp once. Add (disabled)
1267 piece from verify_expr.
1268 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1270 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1272 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1274 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1276 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1277 (small_data_pattern): Likewise.
1278 (arc_rewrite_small_data): Likewise.
1279 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1280 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1281 (get_symbol_alignment): New function.
1282 (legitimate_small_data_address_p): Likewise.
1283 (legitimate_scaled_address): Update, call
1284 legitimate_small_data_address_p.
1285 (output_sdata): New static variable.
1286 (arc_print_operand): Update how we handle small data operands.
1287 (arc_print_operand_address): Likewise.
1288 (arc_legitimate_address_p): Update, use
1289 legitimate_small_data_address_p.
1290 (arc_rewrite_small_data_p): Remove.
1291 (arc_rewrite_small_data_1): Likewise.
1292 (arc_rewrite_small_data): Likewise.
1293 (small_data_pattern): Likewise.
1294 (compact_sda_memory_operand): Update to use
1295 legitimate_small_data_address_p and get_symbol_alignment.
1296 (prepare_move_operands): Don't rewite sdata pattern.
1297 (prepare_extend_operands): Remove.
1298 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1300 (zero_extendqisi2): Likewise.
1301 (zero_extendhisi2): Likewise.
1302 (extendqihi2): Likewise.
1303 (extendqisi2): Likewise.
1304 (extendhisi2): Likewise.
1308 * config/arc/constraints.md (Usd): Change it to memory constraint.
1310 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1312 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1313 as source of std instructions.
1314 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1315 allow 6-bit constants as source for store instructions.
1316 (movdi_insn): Update instruction pattern to allow 6-bit constants
1317 as source for store instructions.
1319 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1321 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1323 2018-04-30 Nathan Sidwell <nathan@acm.org>
1324 Sandra Loosemore <sandra@codesourcery.com>
1326 * dumpfile.c (dump_open): Allow '-' for stdout.
1327 * doc/invoke.texi (Developer Options): Document dump filename
1328 determination early. Document stdin/stdout selection.
1330 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1332 Microblaze Target: PIC data text relative
1334 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1335 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1337 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1338 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1339 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1340 New addressing mode for data-text relative position indepenedent code.
1341 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1342 'ADDRESS_SYMBOLIC_TXT_REL'.
1343 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1344 (microblaze_legitimate_pic_operand): Exclude function calls from
1345 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1346 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1348 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1349 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1350 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1351 for 'address + offset'.
1352 (microblaze_expand_prologue): Add new function prologue call for
1354 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1355 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1356 table in case of TARGET_PIC_DATA_TEXT_REL.
1357 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1358 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1359 Add new macros 'UNSPEC_TEXT',
1360 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1361 + exclude function calls from 'UNSPEC_PLT' in case of data text
1363 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1364 new target hook for generating address diff vector tables in case of
1366 * doc/tm.texi : Regenerate.
1367 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1368 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1369 of addr diff vector generation.
1370 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1371 target hook definition.
1372 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1373 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1374 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1375 option for data text relative.
1377 2018-04-30 Richard Biener <rguenther@suse.de>
1379 * tree-chrec.h (evolution_function_is_constant_p): Remove
1381 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1383 2018-04-30 Richard Biener <rguenther@suse.de>
1386 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1388 2018-04-30 Richard Biener <rguenther@suse.de>
1390 PR tree-optimization/28364
1391 PR tree-optimization/85275
1392 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1393 copying first exit test.
1395 2018-04-28 Mark Wielaard <mark@klomp.org>
1397 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1399 (dwarf_AT): Handle DW_AT_addr_base.
1400 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1402 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1405 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1406 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1407 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1408 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1410 2018-04-28 Richard Biener <rguenther@suse.de>
1412 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1413 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1414 to reflect use. Only add interesting stmts.
1416 2018-04-27 Martin Jambor <mjambor@suse.cz>
1419 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1420 the jump function allows for passing through aggregate values.
1422 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1424 * input.h (in_system_header_at): Convert from macro to inline
1426 (from_macro_expansion_at): Likewise.
1427 (from_macro_definition_at): Likewise.
1429 2018-04-27 Jeff Law <law@redhat.com>
1431 * config.gcc: Mark tile* targets as deprecated/obsolete.
1433 2018-04-27 Richard Biener <rguenther@suse.de>
1435 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1438 2018-04-27 Richard Biener <rguenther@suse.de>
1440 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1442 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1444 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1445 with Yd constraint. Set "preferred_for_speed" attribute from
1446 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1448 (*movdi_internal): Ditto.
1449 (movti_interunit splitters): Remove
1450 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1451 (movdi_interunit splitters): Ditto.
1452 * config/i386/constraints.md (Ye): Remove.
1453 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1455 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1458 * config/aarch64/constraints.md (Usg): Limit to 31.
1461 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1463 PR tree-optimization/85529
1464 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1465 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1466 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1467 zero extension or masking of the MSB bit.
1468 (optimize_range_tests): Add FIRST_BB argument, pass it through
1469 to optimize_range_tests_var_bound.
1470 (maybe_optimize_range_tests, reassociate_bb): Adjust
1471 optimize_range_tests callers.
1473 2018-04-26 Richard Biener <rguenther@suse.de>
1474 Jakub Jelinek <jakub@redhat.com>
1476 * cgraph.h (symbol_table): Just declare debug method here.
1477 * symtab.c (symbol_table::debug): Define.
1479 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1481 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1483 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1485 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1486 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1487 (*movdi_internal): Substitute Yi and Yj constraint with x
1488 and Ym and Yn constraint with y constraint. Update "isa"
1489 attribute and set "preferred_for_speed" attribute from
1490 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1491 (*movsi_internal): Ditto.
1492 (*movdf_internal): Ditto.
1493 (*movsf_internal): Ditto.
1494 (*zero_extendsidi2): Ditto.
1495 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1496 (sse2_loadld): Ditto.
1497 (*vec_extract<ssevecmodelower>_0): Ditto.
1498 (*vec_extractv4si_0_zext_sse4): Ditto.
1499 (vec_concatv2di): Ditto.
1500 (*vec_dup<mode>): Ditto.
1501 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1502 * config/i386/constraints.md (Yi): Remove.
1507 2018-04-26 Nathan Sidwell <nathan@acm.org>
1509 * dumpfile.c (dump_open): New.
1510 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1511 (dump_finish): Detect stdio/stderr by value not name.
1513 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1515 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1517 2018-04-26 Tom de Vries <tom@codesourcery.com>
1520 * config/nvptx/nvptx.c (verify_neutering_jumps)
1521 (verify_neutering_labels): New function
1522 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1524 2018-04-26 Tom de Vries <tom@codesourcery.com>
1527 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1528 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1531 2018-04-26 Richard Biener <rguenther@suse.de>
1532 Tom de Vries <tom@codesourcery.com>
1535 * lto-streamer-out.c (output_function): Fixup loops if required to match
1536 discovery done in the reader.
1538 2018-04-26 Richard Biener <rguenther@suse.de>
1540 PR tree-optimization/85116
1541 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1542 have a loop exit from the single latch predecessor. Remove
1543 case of header with just condition.
1544 (ch_base::copy_headers): Exclude infinite loops from any
1546 (pass_ch::execute): Record exits.
1548 2018-04-26 Richard Biener <rguenther@suse.de>
1550 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1551 prologue cost vector and pass it to vect_get_load_cost.
1552 (vect_get_peeling_costs_all_drs): Likewise.
1553 (vect_peeling_hash_get_lowest_cost): Likewise.
1554 (vect_enhance_data_refs_alignment): Likewise.
1556 2018-04-26 Richard Biener <rguenther@suse.de>
1559 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1560 checking of integer<->pointer conversions.
1561 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1562 sign-/zero-extending pointer types.
1563 (expand_omp_for_static_chunk): Likewise.
1565 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1566 Jean Lee <xiaoyur347@gmail.com>
1568 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1569 (TARGET_ASAN_SHADOW_OFFSET): Define.
1570 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1571 true for -fsanitize=address.
1573 2018-04-25 Mark Wielaard <mark@klomp.org>
1575 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1578 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1580 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1581 than "alu", remove explicit "memory" and "imm_disp" attributes.
1582 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1585 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1586 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1587 gen_lowpart_no_emit.
1589 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1592 * config/i386/i386.c (ix86_expand_builtin): Change memory
1593 operand to XI, extend p0 to Pmode.
1594 * config/i386/i386.md: Change unspec volatile and operand
1595 1 mode to XI, change operand 0 mode to P.
1597 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1599 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1600 GET_MODE_MASK before any checking.
1601 (nds32_can_use_bset_p): Likewise.
1602 (nds32_can_use_btgl_p): Likewise.
1604 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1606 * config/nds32/nds32-doubleword.md: New define_split pattern for
1607 illegal register number.
1609 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1611 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1613 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1615 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1617 2018-04-25 Richard Biener <rguenther@suse.de>
1619 * lto-streamer.h (LTO_major_version): Bump to 8.
1621 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1623 * BASE-VER: Set to 9.0.0.
1625 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1627 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1628 in __abskf2 and __powikf2.
1630 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1633 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1634 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1635 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1636 Use the above on operand 2. Reindent.
1637 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1639 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1642 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1644 * config/i386/i386.opt (mcet): Removed.
1645 * doc/install.texi: Remove -mcet documentation.
1646 * doc/invoke.texi: Likewise.
1648 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1651 * doc/install.texi: Remove -mcet from bootstrap-cet.
1653 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1656 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1657 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1661 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1662 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1663 containing a CONST_VECTOR.
1665 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1667 * doc/install.texi: Update newlib dependency for nvptx.
1669 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1672 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1673 instead of INTVAL when shifting x left.
1675 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1677 PR tree-optimization/85478
1678 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1679 vect_grouped_store_supported for single element vectors.
1681 2018-04-24 Richard Biener <rguenther@suse.de>
1684 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1685 load cost increase to the case of non-constant step.
1687 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1690 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1691 destination if any_malformed_asm.
1693 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1696 * expr.c (store_field): In the bitfield case, if the value comes from
1697 a function call and is returned in registers by means of a PARALLEL,
1698 do not change the mode of the temporary unless BLKmode and VOIDmode.
1700 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1702 PR rtl-optimization/85423
1703 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1704 dependencies to debug insns when the previous insn is non-debug.
1706 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1708 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1709 enums into a single definition.
1710 (fls): Fix predicates and printing.
1713 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1715 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1716 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1717 and short u6 immediate.
1718 (check_if_valid_sleep_operand): Remove.
1719 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1721 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1723 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1724 flag_always_save_lp condition.
1725 * config/nds32/nds32.opt (malways-save-lp): New option.
1727 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1729 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1730 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1731 * config/nds32/nds32.h
1732 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1733 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1735 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1737 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1738 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1740 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1741 Chung-Ju Wu <jasonwucj@gmail.com>
1743 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1745 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1746 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1748 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1750 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1752 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1754 * config/nds32/nds32-protos.h (nds32_data_alignment,
1755 nds32_local_alignment): Declare.
1756 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1757 nds32_local_alignment): New functions.
1758 (TARGET_CONSTANT_ALIGNMENT): Define.
1759 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1761 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1763 * config/nds32/nds32.c
1764 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1765 (TARGET_MODES_TIEABLE_P): Likewise.
1767 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1769 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1772 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1773 Chung-Ju Wu <jasonwucj@gmail.com>
1775 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1776 for unaligned access.
1777 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1779 * config/nds32/nds32-intrinsic.md: Likewise.
1780 * config/nds32/nds32_intrinsic.h: Likewise.
1781 * config/nds32/nds32.h (nds32_builtins): Likewise.
1782 * config/nds32/nds32.opt (munaligned-access): New option.
1783 * config/nds32/nds32.c (nds32_asm_file_start): Display
1784 flag_unaligned_access status.
1786 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1788 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1789 -mno-relax is present.
1790 * config/riscv/linux.h (LINK_SPEC): Ditto.
1792 2018-04-20 Martin Sebor <msebor@redhat.com>
1795 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1797 (gimple_fold_builtin_stxcpy_chk): Same.
1798 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1800 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1803 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1804 __powikf2 when long double is IEEE 128-bit.
1806 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1808 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1809 step to make sure stack always aligned.
1811 2018-04-20 Carl Love <cel@us.ibm.com>
1814 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1815 size check for arg0.
1817 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1818 Tom de Vries <tom@codesourcery.com>
1821 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1822 Emit insns for calls too.
1823 (nvptx_find_par): Always look for worker-level predecessor insn.
1824 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1826 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1827 (nvptx_process_pars): Propagate frames for calls.
1829 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1832 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1834 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1835 (ix86_handle_option): Don't handle OPT_mibt.
1836 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1838 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1840 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1841 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1842 (ix86_target_macros): Define __CET__ with flag_cf_protection
1843 for -fcf-protection.
1844 * config/i386/i386.c (isa2_opts): Remove -mibt.
1845 * config/i386/i386.h (TARGET_IBT): Removed.
1846 (TARGET_IBT_P): Likewise.
1847 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1848 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1849 * config/i386/i386.opt (mcet): Update help message.
1852 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1853 -mcet as an alias for -mshstk.
1855 2018-04-20 Richard Biener <rguenther@suse.de>
1858 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1859 complexity by forcing a single use of the multiply operand.
1861 2018-04-20 Martin Jambor <mjambor@suse.cz>
1864 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1865 recursion dependency to only apply to non-clones.
1867 2018-04-20 Martin Jambor <mjambor@suse.cz>
1870 * ipa-cp.c (create_specialized_node): Check that clones of
1871 self-recursive edges exist during IPA-CP.
1873 2018-04-19 Toon Moene <toon@moene.org>
1875 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1878 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1880 PR tree-optimization/85467
1881 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1882 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1883 VECTOR_CST element to type.
1885 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1888 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1889 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1890 (builtin_longjmp): Likewise.
1891 (save_stack_nonlocal): New pattern.
1892 (restore_stack_nonlocal): Likewise.
1894 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1897 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1898 Replace ASM_OUTPUT_LABEL with fprintf.
1900 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1903 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1904 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1905 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1906 define __IBT__ and __SHSTK__ for -fcf-protection.
1907 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1909 (ix86_trampoline_init): Likewise.
1910 (x86_output_mi_thunk): Likewise.
1911 (ix86_notrack_prefixed_insn_p): Likewise.
1912 (ix86_option_override_internal): Don't disallow -fcf-protection.
1913 * config/i386/i386.md (rdssp<mode>): Also enable for
1915 (incssp<mode>): Likewise.
1916 (nop_endbr): Likewise.
1917 * config/i386/i386.opt (mcet): Change help message to built-in
1921 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1922 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1923 enable CET built-in functions.
1925 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1927 * common/config/i386/i386-common.c
1928 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1929 OPTION_MASK_ISA_MOVDIRI_UNSET,
1930 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1931 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1932 * config.gcc (movdirintrin.h): New header.
1933 * config/i386/cpuid.h (bit_MOVDIRI,
1934 bit_MOVDIR64B): New bits.
1935 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1937 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1938 (VOID, PVOID, PCVOID)): New function types.
1939 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1940 __builtin_ia32_directstoreu_u64,
1941 __builtin_ia32_movdir64b): New builtins.
1942 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1943 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1945 (ix86_valid_target_attribute_inner_p): Ditto.
1946 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1947 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1948 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1949 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1950 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1951 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1952 (movdiri<mode>, movdir64b_<mode>): New.
1953 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1954 * config/i386/immintrin.h: Include movdirintrin.h.
1955 * config/i386/movdirintrin.h: New file.
1956 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1958 2018-04-19 Richard Biener <rguenther@suse.de>
1961 * cfg.c (clear_bb_flags): When loop state says we have
1962 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1964 2018-04-19 Richard Biener <rguenther@suse.de>
1966 PR tree-optimization/84737
1967 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1968 copying restrict info.
1969 (vect_setup_realignment): Use it.
1970 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1971 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1972 the first DR to all generated stores.
1973 (vectorizable_load): Likewise for loads.
1975 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1977 PR tree-optimization/85446
1978 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1979 the integral and pointer types to have the same precision.
1981 * doc/install.texi: Document --disable-cet being the default and
1984 2018-04-18 Martin Liska <mliska@suse.cz>
1986 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
1989 2018-04-18 Martin Liska <mliska@suse.cz>
1992 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1995 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1996 arguments if they are comparable.
1998 2018-04-18 Martin Liska <mliska@suse.cz>
2001 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2004 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2007 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2010 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2011 ENDBR after calling __morestack.
2013 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2016 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2017 by using gcc_base_ver to generate a gcc_driver_version, and use
2018 it when generating GCC_DRIVER_NAME.
2019 * configure: Regenerate.
2021 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2024 * config.gcc: Obsolete powerpc*-*-*spe*.
2026 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2029 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2030 (stabstr_D): Change type of unum from unsigned int to
2031 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2034 2018-04-17 Jim Wilson <jimw@sifive.com>
2037 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2038 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2039 Set arg_pointer_offset after using pretend_args_size.
2041 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2043 PR rtl-optimization/85431
2044 * dse.c (record_store): Ignore zero width stores.
2047 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2048 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2049 __builtin_stack_restore rather than after it.
2050 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2051 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2052 argument instead of virtual_dynamic_stack_rtx.
2054 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2056 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2058 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2059 Add note to error message to explain internal mapping of overloaded
2060 built-in function name to non-overloaded built-in function name.
2061 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2064 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2067 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2068 where the inputs overlap with the output.
2070 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2072 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2073 (=v, v) alternative and explicit "memory" attribute.
2074 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2075 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2077 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2078 "sselog1" type instead of "sselog".
2079 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2080 "sselog". Remove explicit "memory" attribute.
2081 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2082 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2084 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2085 "sselog1" type instead of "sselog". Remove explicit "memory"
2087 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2088 use "sselog1" type instead of "sselog". Remove explicit "memory"
2090 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2091 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2093 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2094 "sselog1" type instead of "sselog". Remove explicit "memory"
2096 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2097 use "sselog1" type instead of "sselog". Remove explicit "memory"
2101 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2104 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2107 2018-04-17 Martin Jambor <mjambor@suse.cz>
2110 * ipa-cp.c (create_specialized_node): Call
2111 expand_all_artificial_thunks if necessary.
2113 2018-04-17 Martin Liska <mliska@suse.cz>
2116 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2117 in message, remote space in between '_G' and '('.
2119 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2122 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2123 avx512f_vmcmp<mode>3<round_saeonly_name>,
2124 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2125 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2126 avx512f_rndscale<mode><round_saeonly_name>,
2127 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2128 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2129 Use %<iptr>2 instead of %2 for -masm=intel.
2130 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2131 avx512f_vcvttss2usi<round_saeonly_name>,
2132 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2134 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2135 avx512f_vcvttsd2usi<round_saeonly_name>,
2136 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2137 Use %q1 instead of %1 for -masm=intel.
2138 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2139 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2140 of %3 for -masm=intel.
2141 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2143 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2145 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2147 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2148 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2150 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2152 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2153 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2154 %g1 and one with %0 and %1.
2155 (avx512er_vmrcp28<mode><round_saeonly_name>,
2156 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2158 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2159 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2160 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2161 of %0 and %{%4%} for -masm=intel.
2162 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2163 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2164 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2165 order of %0 and %{%5%}%{z%} for -masm=intel.
2167 2018-04-17 Jan Hubicka <jh@suse.cz>
2170 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2172 2018-04-17 Martin Liska <mliska@suse.cz>
2175 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2176 for target_clone error message. Make default implementation
2177 clone to be a local declaration.
2178 (separate_attrs): Add new argument and check for an empty
2180 (expand_target_clones): Handle it.
2181 (ipa_target_clone): Make redirection just for target_clones
2184 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2185 Tom de Vries <tom@codesourcery.com>
2188 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2189 tiled basic blocks without omp continue statements.
2191 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2194 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2195 vec_extract expression as having side effects to make sure it gets
2198 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2201 * config/i386/i386.c (get_builtin_code_for_version): Check
2204 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2207 * gcc/config.gcc: Support "skylake".
2208 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2210 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2211 (processor_target_table): Add "skylake".
2212 (ix86_option_override_internal): Add "skylake".
2213 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2214 PROCESSOR_CANNONLAKE.
2215 (get_builtin_code_for_version): Fix priority for
2216 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2217 PROCESSOR_SKYLAKE-AVX512.
2218 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2219 (processor_type): Add PROCESSOR_SKYLAKE.
2221 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2222 Jason Merrill <jason@redhat.com>
2225 * convert.c (convert_to_integer_1): Use direct recursion for
2226 enumeral types and types with a precision less than the number
2227 of bits in their mode.
2229 2018-04-16 Julia Koval <julia.koval@intel.com>
2232 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2233 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2235 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2238 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2239 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2240 and -mno-direct-move.
2242 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2245 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2246 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2247 Ensure negative shifts result in {0}.
2249 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2251 PR rtl-optimization/79916
2252 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2253 regs (if any) to define how to gnerate SD moves when LRA is in
2256 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2258 PR rtl-optimization/85393
2259 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2260 * except.c (expand_dw2_landing_pad_for_region): Make static.
2261 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2262 a label and unconditional jump to old_bb, rather than
2263 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2266 PR rtl-optimization/85376
2267 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2268 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2269 instead of a specific value.
2271 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2272 Bin Cheng <bin.cheng@arm.com>
2274 PR tree-optimization/82965
2275 PR tree-optimization/83991
2276 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2277 by_profile_only parameter.
2278 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2279 information if the loop was predicted to iterate too many times.
2280 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2282 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2285 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2288 2018-04-13 Martin Liska <mliska@suse.cz>
2289 Jakub Jelinek <jakub@redhat.com>
2292 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2293 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2294 * builtins.c (expand_builtin_memory_copy_args): Use
2295 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2296 handle dest_addr == pc_rtx.
2298 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2301 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2302 asked to not generate direct moves.
2303 (fix_trunc<mode>si2_stfiwx): Similar.
2304 (fix_trunc<mode>si2_internal): Similar.
2306 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2309 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2310 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2311 lookup if dest in some wider mode is known to be const0_rtx and
2312 if so, record permanent equivalence for it to be ZERO_EXTEND of
2313 the narrower mode destination.
2315 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2317 * lto-streamer-out.c (output_function): Revert 259346.
2318 * omp-expand.c (expand_oacc_for): Likewise.
2320 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2322 PR rtl-optimization/85354
2323 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2324 * sel-sched.c (sel_global_init): ... here.
2326 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2329 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2330 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2331 mode for PE-COFF targets.
2332 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2333 (i386_pe_asm_lto_end): Likewise.
2334 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2335 (TARGET_ASM_LTO_END): Likewise.
2336 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2337 (i386_pe_asm_lto_start): New function.
2338 (i386_pe_asm_lto_end): Likewise.
2340 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2341 Richard Biener <rguenther@suse.de>
2344 * lto-streamer-out.c (output_function): Fix CFG loop state before
2346 * omp-expand.c (expand_oacc_for): Handle calls to internal
2347 functions like regular functions.
2349 2018-04-12 Richard Biener <rguenther@suse.de>
2352 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2353 for the early LTO debug to properly generate references to it
2354 during DIE emission. Do not re-use that for the skeleton for
2356 (dwarf2out_early_finish): Likewise.
2358 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2361 * config/i386/sse.md
2362 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2363 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2364 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2365 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2366 and output is a reg, avoid creating invalid lowpart subreg, but
2367 instead split into a 512-bit move. Don't split if not AVX512VL,
2368 input is xmm16+ reg and output is a mem.
2369 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2370 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2371 xmm16+ reg and output is a mem.
2373 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2375 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2376 also for flag_dwarf2_cfi_asm.
2378 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2380 PR rtl-optimization/85342
2381 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2382 a bool scalar var inside of the loop instead. Don't try to update
2383 recog_data.operand after failed apply_change_group.
2385 2018-04-12 Tom de Vries <tom@codesourcery.com>
2388 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2389 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2390 array with flexible array member as array without given dimension.
2391 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2392 argument for undefined param to true.
2394 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2397 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2398 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2399 from PowerPC section.
2400 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2401 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2402 help text that is too long.
2403 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2404 help text that is too long.
2405 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2406 help text that is too long.
2408 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2410 * config/alpha/alpha.md (stack_probe_internal): Rename
2411 from "probe_stack". Update all callers.
2413 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2415 PR rtl-optimization/84566
2416 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2417 sched_macro_fuse_insns.
2419 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2422 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2423 (compute_block_dependences): ... from here.
2425 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2427 PR tree-optimization/85331
2428 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2429 from int to HOST_WIDE_INT.
2431 2018-04-11 Martin Jambor <mjambor@suse.cz>
2434 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2435 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2436 not the same as the source val.
2437 (cgraph_edge_brings_value_p): New parameter.
2438 (gather_edges_for_value): Pass destination value to
2439 cgraph_edge_brings_value_p.
2440 (perhaps_add_new_callers): Likewise.
2441 (get_info_about_necessary_edges): Likewise and exclude values brought
2442 only by self-recursive edges.
2443 (create_specialized_node): Redirect only clones of self-calling edges.
2444 (+self_recursive_pass_through_p): New function.
2445 (find_more_scalar_values_for_callers_subset): Use it.
2446 (find_aggregate_values_for_callers_subset): Likewise.
2447 (known_aggs_to_agg_replacement_list): Removed.
2448 (decide_whether_version_node): Re-calculate known constants for all
2449 remaining context clones.
2451 2018-04-11 Richard Biener <rguenther@suse.de>
2454 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2455 from early DWARF output.
2456 (dwarf2out_early_finish): Output line info unconditionally into
2457 early DWARF and add reference to it.
2459 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2462 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2463 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2464 other than V2DFmode using iptr mode attribute.
2465 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2467 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2469 PR rtl-optimization/84659
2470 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2472 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2475 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2477 (output_loc_list): Pass address of a dummy size variable even in the
2478 locview handling loop.
2479 (index_location_lists): Add comment on why skip_loc_list_entry can't
2482 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2485 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2488 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2491 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2492 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2493 and -mstring-compare-inline-limit.
2495 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2498 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2499 for stack clash protection in a register whenever we need it to be in
2502 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2504 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2505 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2507 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2510 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2512 (mlong-double-): Ditto.
2513 * config/rs6000/sysv4.opt (msdata=): Ditto.
2514 (mtls-size=): Ditto.
2516 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2518 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2519 erroneous entries for
2520 "vector int vec_ldl (int, long int *)", and
2521 "vector unsigned int vec_ldl (int, unsigned long int *)".
2522 Add comments and entries for
2523 "vector bool char vec_ldl (int, bool char *)",
2524 "vector bool short vec_ldl (int, bool short *)",
2525 "vector bool int vec_ldl (int, bool int *)",
2526 "vector bool long long vec_ldl (int, bool long long *)",
2527 "vector pixel vec_ldl (int, pixel *)",
2528 "vector long long vec_ldl (int, long long *)",
2529 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2530 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2531 type tree bool_long_long_type_node and correct definition of
2532 bool_V2DI_type_node to make reference to this new type tree.
2533 (rs6000_mangle_type): Replace erroneous reference to
2534 bool_long_type_node with bool_long_long_type_node.
2535 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2536 comments to emphasize sign distinctions for char and int types and
2537 replace RS6000_BTI_bool_long constant with
2538 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2539 use of RS6000_BTI_pixel.
2540 (bool_long_type_node): Remove this macro definition.
2541 (bool_long_long_type_node): New macro definition
2543 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2545 PR rtl-optimization/85300
2546 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2547 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2548 simplify_unary_operation fails.
2550 2018-04-10 Martin Liska <mliska@suse.cz>
2552 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2553 cgraph_edge and ipa_ref.
2555 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2559 * config/i386/sse.md
2560 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2561 computation of the VEC_MERGE selector from mask.
2562 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2563 Fix decoding of the VEC_MERGE selector into mask.
2565 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2567 PR tree-optimization/85286
2568 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2570 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2572 * final.c (final_1): Set insn_last_address as well as
2573 insn_current_address.
2575 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2578 * explow.c (emit_stack_probe): Call validize_mem on memory location
2579 before passing it to gen_probe_stack. Create address operand and
2580 legitimize it for the probe_stack_address case.
2582 2018-04-09 Jan Hubicka <jh@suse.cz>
2585 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2586 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2587 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2589 (free_lang_data): Rebuild type inheritance graph.
2591 2018-04-09 Martin Sebor <msebor@redhat.com>
2593 * invoke.texi (-finline-small-functions): Mention other optimization
2595 (-findirect-inlining, -fpartial-inlining): Same.
2596 (-finline-functions-called-once): Same.
2597 (-freorder-blocks-and-partition): Same.
2599 2018-04-09 Jan Hubicka <jh@suse.cz>
2602 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2603 jumps; choose last target that matches the criteria (i.e.
2604 no partition changes for non-crossing jumps).
2605 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2606 support for redirecting crossing jumps to non-crossing.
2608 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2610 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2611 also for naked functions.
2613 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2615 * config/arc/arc.md (add_shift): New pattern.
2616 (add_shift2): Likewise.
2617 (sub_shift): Likewise.
2618 (sub_shift_cmp0_noout): Likewise.
2619 (compare_si_ashiftsi): Likewise.
2620 (xbfu_cmp0_noout): New combine pattern.
2621 (xbfu_cmp0"): Likewise.
2622 (movsi_set_cc_insn): Place the predicable variant first.
2623 (commutative_binary_cmp0_noout): Remove clobber.
2624 (commutative_binary_cmp0): New pattern.
2625 (noncommutative_binary_cmp0): Likewise.
2626 (noncommutative_binary_cmp0_noout): Likewise.
2627 (noncommutative_binary_comparison_result_used): Removed.
2628 (rsub_cmp0): New pattern.
2629 (rsub_cmp0_noout): Likewise.
2630 (extzvsi): Changed, keep only meaningful variants.
2631 (SQH, SEZ): New iterators.
2632 (SQH_postfix): New mode attribute.
2633 (SEZ_prefix): New code attribute.
2634 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2635 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2636 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2638 (noncommutative_operator): Check the availability of barrel
2641 2018-04-09 Richard Biener <rguenther@suse.de>
2643 PR tree-optimization/85284
2644 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2645 Only use the niter constraining form of simple_iv when the exit
2648 2018-04-09 Tom de Vries <tom@codesourcery.com>
2651 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2652 (define_expand "*memory_barrier"): New define_expand.
2653 (define_insn "memory_barrier"): New insn.
2655 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2657 PR rtl-optimization/80463
2658 PR rtl-optimization/83972
2659 PR rtl-optimization/83480
2661 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2662 correct producer for the insn.
2663 (tidy_control_flow): Fixup seqnos in case of debug insns.
2665 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2667 PR rtl-optimization/83913
2669 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2670 different sched-times when merging exprs.
2672 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2674 PR rtl-optimization/83962
2676 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2677 tidy_fallthru_edge and tidy_control_flow.
2679 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2681 PR rtl-optimization/83530
2683 * sel-sched.c (force_next_insn): New global variable.
2684 (remove_insn_for_debug): When force_next_insn is true, also leave only
2685 next insn in the ready list.
2686 (sel_sched_region): When the region wasn't scheduled, make another pass
2687 over it with force_next_insn set to 1.
2689 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2691 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2693 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2694 for interrupt control.
2695 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2696 functions for interrupt control.
2697 * config/nds32/nds32-intrinsic.md: Likewise.
2698 * config/nds32/nds32_intrinsic.h: Likewise.
2699 * config/nds32/nds32.h (nds32_builtins): Likewise.
2701 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2703 * config/nds32/nds32.c (nds32_init_machine_status,
2704 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2705 strict_aligned_p field.
2706 (nds32_expand_to_rtl_hook): New function.
2707 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2708 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2710 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2711 Chung-Ju Wu <jasonwucj@gmail.com>
2713 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2714 * config/nds32/nds32-n7.md: New file.
2715 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2716 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2718 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2719 * config/nds32/nds32.md (pipeline_model): Add n7.
2720 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2721 * config/nds32/pipelines.md: Include n7 settings.
2723 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2724 Chung-Ju Wu <jasonwucj@gmail.com>
2726 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2727 * config/nds32/nds32-e8.md: New file.
2728 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2729 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2731 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2732 * config/nds32/nds32.md (pipeline_model): Add e8.
2733 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2734 * config/nds32/pipelines.md: Include e8 settings.
2736 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2737 Chung-Ju Wu <jasonwucj@gmail.com>
2739 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2740 * config/nds32/nds32-n8.md: New file.
2741 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2742 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2744 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2745 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2746 * config/nds32/nds32.md (pipeline_model): Add n8.
2747 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2748 * config/nds32/pipelines.md: Include n8 settings.
2750 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2751 Chung-Ju Wu <jasonwucj@gmail.com>
2753 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2754 * config/nds32/nds32-n9-2r1w.md: New file.
2755 * config/nds32/nds32-n9-3r2w.md: New file.
2756 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2757 nds32_register_ports): New or modify for cpu n9.
2758 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2760 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2761 * config/nds32/nds32-utils.c: New file.
2762 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2763 TARGET_MUL_SLOW): Define.
2764 * config/nds32/nds32.md (pipeline_model): New attribute.
2765 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2766 New options that support cpu n9.
2767 * config/nds32/pipelines.md: Include n9 settings.
2768 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2770 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2772 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2773 information if necessary.
2774 (output_cond_branch_compare_zero): Likewise.
2775 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2776 (nds32_target_alignment): Refine for alignment.
2777 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2778 (FUNCTION_BOUNDARY): Modify.
2779 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2781 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2783 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2785 * config/nds32/constants.md (unspec_volatile_element): Add values for
2786 TLB operation and data prefetch.
2787 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2788 functions for TLB operation and data prefetch.
2789 * config/nds32/nds32-intrinsic.md: Likewise.
2790 * config/nds32/nds32_intrinsic.h: Likewise.
2791 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2792 (nds32_print_operand): Likewise.
2793 * config/nds32/nds32.h (nds32_builtins): Likewise.
2795 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2796 Andrew Pinski <pinsika@gcc.gnu.org>
2799 * match.pd: Use constant_boolean_node of correct type instead of
2800 boolean_true_node or boolean_false_node for simplifying
2801 pointer comparisons to zero.
2803 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2805 PR tree-optimization/80021
2806 * tree.c (verify_type_variant): Make error call in verify_variant_match
2807 translatable and remove final full stop.
2809 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2811 * config/nds32/constants.md (unspec_volatile_element): Add
2812 UNSPEC_VOLATILE_EH_RETURN.
2813 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2814 nds32_output_stack_pop): Support dwarf exception handling process.
2815 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2816 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2817 exception handling process.
2818 (nds32_compute_stack_frame): Likewise.
2819 (nds32_return_addr_rtx): Likewise.
2820 (nds32_initial_elimination_offset): Likewise.
2821 (nds32_expand_prologue): Likewise.
2822 (nds32_expand_epilogue): Likewise.
2823 (nds32_dynamic_chain_address): New function.
2824 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2826 (DYNAMIC_CHAIN_ADDRESS): Define.
2827 (EH_RETURN_DATA_REGNO): Define.
2828 (EH_RETURN_STACKADJ_RTX): Define.
2829 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2830 patterns for dwarf exception handling.
2832 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2834 * config/nds32/nds32.h: Clean up obsolete macros.
2836 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2838 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2839 Add enum values for particular instructions.
2840 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2841 particular intrinsic functions.
2842 * config/nds32/nds32-intrinsic.md: Likewise.
2843 * config/nds32/nds32_intrinsic.h: Likewise.
2844 * config/nds32/nds32.h (nds32_builtins): Likewise.
2845 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2846 (btst, ave): New patterns for particular instructions.
2848 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2850 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2851 Add enum values for atomic load/store and memory sync.
2852 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2854 * config/nds32/nds32-intrinsic.md: Likewise.
2855 * config/nds32/nds32_intrinsic.h: Likewise.
2856 * config/nds32/nds32.h (nds32_builtins): Likewise.
2858 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2860 PR tree-optimization/85257
2861 * fold-const.c (native_encode_vector): If not all elts could fit
2862 and off is -1, return 0 rather than offset.
2863 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2864 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2865 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2866 adjust buffer in native_interpret_expr call.
2868 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2870 * config/nds32/constants.md (unspec_volatile_element): Add cache
2871 control enum values.
2872 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2873 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2874 * config/nds32/nds32.c (nds32_cctl_names): New.
2875 (nds32_print_operand): Handle cache control register names.
2876 * config/nds32/nds32.h (nds32_builtins): New enum values.
2877 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2879 * config/nds32/nds32.md (type): Add mmu.
2880 * config/nds32/pipelines.md (simple_insn): Add mmu.
2882 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2884 * config/nds32/nds32.md (type): Remove call.
2885 * config/nds32/pipelines.md (simple_insn): Likewise.
2887 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2889 * config/nds32/constants.md (unspec_volatile_element): Add
2890 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2891 UNSPEC_VOLATILE_FMFCFG.
2892 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2893 description for fmfcfg and fmfcsr.
2894 (bdesc_1arg): Add fmtcsr.
2895 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2896 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2897 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2898 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2899 unspec_fmfcfg): New patterns.
2900 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2901 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2902 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2903 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2904 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2905 __nds32__fmfcfg): Define.
2907 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2909 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2910 intrinsic register names.
2911 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2912 intrinsic register enum values and macros.
2914 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2916 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2917 for load/store addressing form.
2918 (nds32_print_operand_address): Likewise.
2920 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2923 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2924 based on LABEL_REF. Remove useless assertion.
2925 (pic_address_needs_scratch): Fix formatting.
2926 (sparc_legitimize_pic_address): Minor tweaks.
2927 (sparc_delegitimize_address): Adjust assertion accordingly.
2928 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2929 into symbolic_operand.
2930 (movsi_high_pic_label_ref): Likewise.
2931 (movsi_lo_sum_pic_label_ref): Likewise.
2932 (movdi_pic_label_ref): Likewise.
2933 (movdi_high_pic_label_ref): Likewise.
2934 (movdi_lo_sum_pic_label_ref): Likewise.
2936 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2938 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2939 custom LIB_SPEC setup.
2941 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2942 Kito Cheng <kito.cheng@gmail.com>
2944 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2945 * config/riscv/freebsd.h: New.
2947 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2949 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2950 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2953 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2954 Kito Cheng <kito.cheng@gmail.com>
2956 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2957 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2958 * config/nds32/nds32-protos.h (nds32_output_call,
2959 nds32_output_return): Declare.
2960 * config/nds32/nds32.md: Refine all the call and return patterns.
2962 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2965 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2966 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2968 PR rtl-optimization/84872
2969 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2970 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2973 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2975 * expr.c (copy_blkmode_to_reg): Revert 254862.
2976 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2978 2018-04-06 Richard Biener <rguenther@suse.de>
2981 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2982 after seeing a component reference with an adjacent field. Treat
2983 refs to arrays at struct end of external decls similar to
2984 refs to unconstrained commons.
2986 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2989 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
2990 look through SAVE_EXPRs with non-side-effects argument. Adjust
2992 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2995 2018-04-06 Richard Biener <rguenther@suse.de>
2998 * alias.c (find_base_term): New wrapper around find_base_term
2999 unwinding CSELIB_VAL_PTR changes.
3000 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3003 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3005 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3007 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3008 constant definitions.
3009 ("nop"): lr 0,0 -> nopr r0
3010 ("nop_lr0", "nop_lr1"): New insn definitions.
3012 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3014 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3015 NDS32_V3PUSH_AVAILABLE_P macro.
3017 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3018 Chung-Ju Wu <jasonwucj@gmail.com>
3020 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3021 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3022 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3023 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3024 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3025 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3026 * config/nds32/constraints.md: New constraints and checking for hard
3027 float configuration.
3028 * config/nds32/iterators.md: New mode iterator and attribute for hard
3029 float configuration.
3030 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3032 * config/nds32/nds32-fpu.md: New file.
3033 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3034 deal with hard float code generation.
3035 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3037 (abi_type, float_reg_number): New enum type.
3038 * config/nds32/nds32-predicates.c: New predicates for hard float.
3039 * config/nds32/nds32-protos.h: Declare functions for hard float.
3040 * config/nds32/nds32.c: Implementation for hard float configuration.
3041 * config/nds32/nds32.h: Definitions for hard float configuration.
3042 * config/nds32/nds32.md: Include hard float machine description and
3043 modify patterns for hard float configuration.
3044 * config/nds32/nds32.opt: New options for hard float configuration.
3045 * config/nds32/predicates.md: New predicates for hard float
3048 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3050 * common/config/nds32/nds32-common.c
3051 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3053 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3056 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3057 CONSTRUCTOR_ELT (ctor, ...)->value.
3059 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3062 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3064 2018-04-05 Tom de Vries <tom@codesourcery.com>
3067 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3070 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3071 Kito Cheng <kito.cheng@gmail.com>
3073 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3074 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3075 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3076 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3078 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3079 Kito Cheng <kito.cheng@gmail.com>
3081 * config/nds32/constraints.md (Ufe): New memory constraint.
3082 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3083 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3084 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3086 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3087 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3089 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3091 * config/nds32/nds32.md: Use optimize_size in the condition for
3092 alu-shift instructions.
3094 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3096 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3098 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3100 * config/nds32/nds32.md (negsi2): Refine pattern.
3102 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3103 Chung-Ju Wu <jasonwucj@gmail.com>
3105 * config/nds32/iterators.md (shift_rotate): New code iterator.
3106 (shift): New code attribute.
3107 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3108 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3109 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3110 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3111 bit-wise operations.
3112 (andsi3, *andsi3): Ditto.
3113 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3114 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3115 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3116 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3117 nds32_ior_operand, nds32_xor_operand): New predicates.
3119 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3121 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3122 (addsi3, subsi3): ... this.
3124 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3126 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3128 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3130 * config/nds32/nds32.md: Adjust indention.
3132 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3134 * config/nds32/nds32.md (feature): New attribute.
3136 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3138 * config/nds32/nds32.md (subtype): New attribute.
3140 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3143 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3144 expansion to perform a bitwise AND of the argument followed by a
3145 boolean negation of the result.
3147 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3149 PR rtl-optimization/84878
3150 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3151 the basic block. Assert the use reference is not artificial and that
3152 it has an associated insn.
3154 2018-04-04 Michael Matz <matz@suse.de>
3156 * builtins.c (compute_objsize): Pass correct operand
3157 to array_at_struct_end_p.
3159 2018-04-04 Richard Biener <rguenther@suse.de>
3162 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3163 from contexts for DINFO_LEVEL_TERSE and below.
3165 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3167 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3168 resiter_operand condition.
3169 * config/nds32/nds32.md (*move<mode>): Ditto.
3171 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3172 Monk Chiang <sh.chiang04@gmail.com>
3174 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3176 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3178 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3180 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3181 Kito Cheng <kito.cheng@gmail.com>
3183 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3184 nds32_cond_code_str, output_cond_branch,
3185 output_cond_branch_compare_zero, nds32_expand_cbranch,
3186 nds32_expand_cstore, nds32_expand_movcc,
3187 nds32_output_cbranchsi4_equality_zero,
3188 nds32_output_cbranchsi4_equality_reg,
3189 nds32_output_cbranchsi4_equality_reg_or_const_int,
3190 nds32_output_cbranchsi4_greater_less_zero: New functions.
3191 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3192 nds32_expand_cstore, nds32_expand_movcc,
3193 nds32_output_cbranchsi4_equality_zero,
3194 nds32_output_cbranchsi4_equality_reg,
3195 nds32_output_cbranchsi4_equality_reg_or_const_int,
3196 nds32_output_cbranchsi4_greater_less_zero): Declare.
3197 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3198 nds32_rimm11s_operand): New predicates.
3199 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3200 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3203 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3205 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3206 * config/nds32/nds32.md: Ditto.
3207 * config/nds32/pipelines.md: Ditto.
3209 2018-04-04 Richard Biener <rguenther@suse.de>
3211 PR tree-optimization/85168
3212 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3213 propagating abnormals.
3215 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3217 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3219 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3220 Kito Cheng <kito.cheng@gmail.com>
3222 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3223 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3224 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3225 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3226 * config/nds32/nds32.md (sibcall_internal): New.
3227 (sibcall_register): Remove.
3228 (sibcall_immediate): Remove.
3229 (sibcall_value_internal): New.
3230 (sibcall_value_register): Remove.
3231 (sibcall_value_immediate): Remove.
3232 * config/nds32/predicates.md (nds32_general_register_operand): New.
3233 (nds32_call_address_operand): New.
3235 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3237 PR rtl-optimization/85167
3238 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3239 bb_defs if *split_p, instead preinitialize it to NULL.
3241 PR tree-optimization/85156
3242 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3243 evaluating the argument multiple times.
3245 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3247 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3249 (_mm_cvtpd_ps): Likewise.
3250 (_mm_cvttpd_epi32): Likewise.
3251 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3252 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3253 vector, pixel, and bool following altivec.h include.
3255 2018-04-03 Martin Sebor <msebor@redhat.com>
3257 * doc/extend.texi (Common Function Attributes): Clarify.
3258 (const attribute): Likewise.
3259 (pure attribute): Likewise.
3261 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3264 * config/i386/i386.c (ix86_expand_vector_set): Use
3265 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3267 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3269 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3270 instructions when changing rounding bits to preserve precision bits
3271 in the x87 control word.
3273 2018-04-03 Martin Liska <mliska@suse.cz>
3275 PR tree-optimization/82491
3276 * rtl.h (strip_offset_and_add): Replace += suboffset with
3277 poly_uint64 () + suboffset.
3279 2018-03-29 Martin Liska <mliska@suse.cz>
3280 Martin Jambor <mjambor@suse.cz>
3283 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3284 param_type is not an integral or pointer type.
3286 2018-04-03 Richard Biener <rguenther@suse.de>
3288 * sese.h (recompute_all_dominators): Remove.
3290 2018-04-02 Martin Sebor <msebor@redhat.com>
3292 * doc/invoke.texi (-Wrestrict): Fix typos.
3294 2018-04-02 Jim Wilson <jimw@sifive.com>
3296 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3297 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3298 (<optab>di3, <optab>si3_extend): Likewise.
3299 (<optab>si3_mask, <optab>si3_mask_1): New.
3300 (<optab>di3_mask, <optab>di3_mask_1): New.
3301 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3302 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3303 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3305 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3307 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3310 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3312 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3313 (nds32_canonicalize_comparison): New function.
3315 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3316 Kito Cheng <kito.cheng@gmail.com>
3317 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3319 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3320 * config/nds32/constants.md (unspec_volatile_element): Add
3321 UNSPEC_VOLATILE_RELAX_GROUP.
3322 * config/nds32/nds32-relax-opt.c: New file.
3323 * config/nds32/nds32-predicates.c
3324 (nds32_symbol_load_store_p): New function.
3325 * config/nds32/nds32-protos.h
3326 (nds32_symbol_load_store_p): Declare function.
3327 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3328 * config/nds32/nds32.c
3329 (nds32_register_pass): New function to register pass.
3330 (nds32_register_passes): New function to register passes.
3331 * config/nds32/nds32.md (relax_group): New pattern.
3332 * config/nds32/nds32.opt (mrelax-hint): New option.
3333 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3335 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3337 * config/nds32/t-nds32: Modify files dependency.
3339 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3341 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3342 (PROFILE_HOOK): Define its implementation.
3344 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3346 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3347 type and 32-bit size.
3349 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3352 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3353 (V_128_256): New mode iterator.
3354 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3355 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3356 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3358 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3359 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3361 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3364 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3365 NaN inputs correctly.
3367 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3370 * config/rs6000/vsx.md (??r): New mode attribute.
3371 (*vsx_mov<mode>_64bit): Use it.
3372 (*vsx_mov<mode>_32bit): Likewise.
3374 2018-03-30 Martin Sebor <msebor@redhat.com>
3376 PR tree-optimization/84818
3377 * builtins.c (check_access): Use warning_n.
3379 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3382 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3384 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3387 2018-03-30 Julia Koval <julia.koval@intel.com>
3390 * x86-tune.def (movx, partial_reg_dependency): Enable for
3393 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3396 * lra-constraints.c (process_alt_operands): Move setting
3397 this_alternative_matches below.
3399 2018-03-29 Martin Liska <mliska@suse.cz>
3402 * doc/invoke.texi: Document how LTO works with debug info.
3403 Describe auto-load support of binutils. Mention 'x86-64'
3404 as valid option value of -march option.
3406 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3408 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3411 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3412 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3413 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3416 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3419 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3420 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3421 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3422 for RS6000_BTM_POWERPC64.
3423 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3424 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3425 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3430 2018-03-28 Carl Love <cel@us.ibm.com>
3433 2017-09-27 Carl Love <cel@us.ibm.com>
3435 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3436 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3437 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3440 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3442 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3443 instead of __vector bool.
3444 (_mm_max_pu8): Likewise.
3445 (_mm_min_pi16): Likewise.
3447 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3450 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3451 (DIVWEUO): Likewise.
3453 (DIVDEUO): Likewise.
3454 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3455 DIVWEUO and DIVDEUO.
3456 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3457 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3458 (div_extend): Likewise.
3459 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3461 (__builtin_divweuo): Likewise.
3462 (__builtin_divdeo): Likewise.
3463 (__builtin_divdeuo): Likewise.
3465 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3468 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3469 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3471 PR tree-optimization/82004
3472 * gimple-match-head.c (optimize_pow_to_exp): New function.
3473 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3474 Don't fold to exp if optimize_pow_to_exp is false.
3476 2018-03-28 Martin Liska <mliska@suse.cz>
3479 * calls.c (initialize_argument_information): Fix trailing space.
3480 * common.opt: Fix typo and provide better explanation for
3481 -fsanitize-coverage option.
3482 * config/i386/i386.opt: Fix typo.
3484 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3485 Martin Liska <mliska@suse.cz>
3488 * gimplify.c (asan_poison_variable): Don't do the check for
3489 gimplify_omp_ctxp here.
3490 (gimplify_decl_expr): Do it here.
3491 (gimplify_target_expr): Likewise.
3493 2018-03-28 Martin Liska <mliska@suse.cz>
3496 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3497 chkp_type_bounds_count if MPX is not enabled.
3499 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3501 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3503 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3506 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3507 function to create the function decl for complex long double
3508 multiply and divide for -mabi=ieeelongdouble.
3509 (init_float128_ieee): Call it.
3511 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3514 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3515 -fcf-protection=branch -mibt.
3516 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3518 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3521 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3523 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3526 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3527 extern array declarations.
3529 2018-03-27 Richard Biener <rguenther@suse.de>
3532 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3533 explicit single_use checks.
3535 2018-03-27 Richard Biener <rguenther@suse.de>
3537 PR tree-optimization/85082
3538 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3541 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3543 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3544 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3545 Turn on fasynchronous-unwind-tables and funwind-tables.
3547 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3550 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3551 (*bmi_blsr_<mode>_ccz): Ditto.
3553 2018-03-26 Tom de Vries <tom@codesourcery.com>
3555 PR tree-optimization/85063
3556 * omp-general.c (offloading_function_p): New function. Factor out
3558 * omp-offload.c (pass_omp_target_link::gate): ... here.
3559 * omp-general.h (offloading_function_p): Declare.
3560 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3561 with attribute omp declare target for offloading functions.
3563 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3565 PR tree-optimization/84005
3566 * tree-data-ref.h (get_base_for_alignment): Declare.
3567 * tree-data-ref.c (get_base_for_alignment_1): New function.
3568 (get_base_for_alignment): Likewise.
3569 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3570 get_base_for_alignment to find a suitable base object, instead
3571 of always using drb->base_address.
3573 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3576 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3577 known size by default.
3579 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3582 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3583 and non BLKmode operands.
3585 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3588 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3589 Clean up attributes.
3591 2018-03-23 Richard Biener <rguenther@suse.de>
3594 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3595 we are going to emit early debug for LTO.
3597 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3600 * function.c (match_asm_constraints_1): Don't optimize if input
3601 doesn't satisfy general_operand predicate for output's mode.
3604 * alias.c (write_dependence_p): Don't require for x_canonicalized
3605 non-VOIDmode if x has VOIDmode.
3608 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3609 just don't try to optimize it rather than assert it never happens.
3611 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3613 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3614 macro expansions for definition of ST_INTERNAL_<mode> and
3615 LD_INTERNAL_<mode> builtins.
3616 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3618 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3620 (altivec_expand_st_builtin): Likewise.
3621 (altivec_expand_builtin): Remove calls to deleted functions.
3622 (rs6000_address_for_altivec): Delete this function.
3623 * config/rs6000/vector.md: Remove expands for
3624 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3626 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3629 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3630 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3631 re-computing once computed.
3632 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3633 (arm_init_machine_status): Initialize
3634 machine->static_chain_stack_bytes.
3636 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3639 * doc/extend.texi: Add four new prototypes for vec_ld.
3640 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3641 definitions for more logical presentation.
3642 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3643 entries for V1TI variants of __builtin_altivec_ld builtin.
3644 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3645 handling of V1TI variant of LVX icode pattern.
3646 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3647 (rs6000_gimple_fold_builtin): Likewise.
3648 (altivec_init_builtins): Add code to define
3649 __builtin_altivec_lvx_v1ti function.
3651 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3654 * function.c (match_asm_constraints_1): Don't do the optimization
3655 if input isn't a REG, SUBREG, MEM or constant.
3657 2018-03-22 Tom de Vries <tom@codesourcery.com>
3659 PR tree-optimization/84956
3660 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3661 bb_has_abnormal_pred.
3663 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3666 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3667 DECL_INITIAL (decl) to decl at the end.
3668 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3671 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3673 * doc/extend.texi (__builtin_tgmath): Document when complex
3674 integer types are treated as _Complex _Float64.
3676 2018-03-21 Tom de Vries <tom@codesourcery.com>
3678 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3680 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3682 PR tree-optimization/84960
3683 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3684 if it is ENTRY block, move them into single succ of ENTRY in that case.
3686 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3688 PR tree-optimization/84811
3689 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3690 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3691 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3692 (known_subrange_p): Update accordingly. Cast each value involved
3693 in the size comparison, rather than casting the result of the
3696 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3698 PR tree-optimization/84982
3699 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3700 by flipping the least significant bit rather than all bits from
3701 bitpos to bitpos + bitsize - 1.
3703 2018-03-21 Nathan Sidwell <nathan@acm.org>
3705 * doc/extend.texi (Deprecated Features): Remove mention of
3706 long-deleted deprecations.
3708 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3711 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3712 * configure: Regenerate.
3714 2018-03-21 Tom de Vries <tom@codesourcery.com>
3716 PR tree-optimization/83126
3717 * tree-parloops.c (num_phis): New function.
3718 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3720 2018-03-21 Nathan Sidwell <nathan@acm.org>
3722 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3723 mention anon-struct/union members and trailing attributes.
3725 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3727 PR tree-optimization/84969
3728 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3729 builtin memset partitions if they set different rhs values.
3731 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3733 PR rtl-optimization/84989
3734 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3735 VEC_DUPLICATE with scalar result mode.
3737 2018-03-21 Martin Liska <mliska@suse.cz>
3740 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3741 not intended return statement.
3743 2018-03-21 Martin Liska <mliska@suse.cz>
3746 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3747 (chkp_find_bound_slots_1): Limit number of iterations.
3749 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3752 * Minor grammar fixes for x86 options.
3754 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3757 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3758 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3760 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3763 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3764 (altivec_lvx_<mode>_1op): Likewise.
3765 (altivec_stvx_<mode>_2op): Likewise.
3766 (altivec_stvx_<mode>_1op): Likewise.
3767 (altivec_lvx_<VM2:mode>): New define_expand.
3768 (altivec_stvx_<VM2:mode>): Likewise.
3769 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3770 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3771 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3772 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3773 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3774 (rs6000_gen_lvx): Likewise.
3775 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3776 (altivec_expand_stv_builtin): Likewise.
3777 (altivec_expand_builtin): Likewise.
3778 * config/rs6000/vector.md: Likewise.
3780 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3783 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3786 2018-03-20 Richard Biener <rguenther@suse.de>
3789 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3790 sign-conversions as zero, fall back to standard scalar_stmt
3793 2018-03-20 Martin Liska <mliska@suse.cz>
3796 * predict.c (rebuild_frequencies): Handle case when we have
3797 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3799 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3802 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3803 flag_section_anchors.
3804 * varasm.c (use_blocks_for_decl_p): Remove hack for
3805 dw2_force_const_mem.
3808 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3810 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3811 be created, use lowpart_subreg of operands[0] rather than operands[0]
3813 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3814 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3815 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3816 and n constraint instead of aarch64_shift_imm_di and Usd.
3817 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3818 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3820 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3823 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3824 to favor GPR over NEON registers.
3825 (<shift>di3_neon): Likewise.
3827 2018-03-20 Tom de Vries <tom@codesourcery.com>
3830 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3831 (nvptx_process_pars): Emit bar.sync asap and alap.
3833 2018-03-20 Tom de Vries <tom@codesourcery.com>
3836 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3837 seen_label if seen_label is already set.
3839 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3842 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3843 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3844 Use 1U instead of 1. Formatting fixes.
3847 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3848 instead of TREE_TYPE (s1) for the return value.
3850 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3852 PR tree-optimization/84946
3853 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3854 bitsize + bitsize in poly_uint64 rather than poly_int64.
3857 * dwarf2asm.c: Include fold-const.c.
3858 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3859 of decl rather than decl itself.
3861 PR rtl-optimization/84643
3862 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3864 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3867 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3868 calling assemble_variable.
3870 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3873 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3874 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3876 2018-03-19 Jim Wilson <jimw@sifive.com>
3879 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3880 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3881 (riscv_first_stack_step): Likewise.
3882 (riscv_option_override): Use STACK_BOUNDARY instead of
3884 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3886 (BIGGEST_ALIGNMENT): Set to 128.
3887 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3888 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3891 2018-03-19 Richard Biener <rguenther@suse.de>
3893 PR tree-optimization/84933
3894 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3895 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3897 2018-03-19 Richard Biener <rguenther@suse.de>
3899 PR tree-optimization/84859
3900 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3901 (cond_if_else_store_replacement): Perform sinking operation on
3902 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3903 Generalize what a BB with a single eligible store is.
3905 2018-03-19 Richard Biener <rguenther@suse.de>
3907 PR tree-optimization/84929
3908 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3909 chrec_is_positive against non-chrec arg.
3911 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3914 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3916 2018-03-18 Martin Liska <mliska@suse.cz>
3918 PR rtl-optimization/84635
3919 * regrename.c (build_def_use): Use matches_mode only when
3922 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3924 PR tree-optimization/84913
3925 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3926 vectorize chains of COND_EXPRs.
3928 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3930 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3932 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3934 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3936 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3938 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3940 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3941 Kito Cheng <kito.cheng@gmail.com>
3943 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3944 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3945 (nds32_adjust_reg_alloc_order): New function.
3946 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3948 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3950 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3951 nds32_print_operand, nds32_print_operand_address): Use
3952 HOST_WIDE_INT_PRINT_DEC instead.
3954 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3956 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3958 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3961 * config/i386/i386.c (initial_ix86_tune_features,
3962 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3964 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3965 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3966 rather than 1u << ix86_tune. Formatting fix.
3967 (ix86_option_override_internal): Change ix86_arch_mask from
3968 unsigned int to unsigned HOST_WIDE_INT, initialize to
3969 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3970 (ix86_function_specific_restore): Likewise.
3972 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3975 * postreload.c (reload_combine_recognize_pattern): Perform
3976 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3977 truncate_int_for_mode the result for the destination's mode.
3980 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3981 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3983 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3986 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
3989 * lra-assigns.c (lra_split_hard_reg_for): Don't use
3990 regno_allocno_class_array and sorted_pseudos.
3991 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3992 insns where regno is used.
3994 2018-03-16 Martin Liska <mliska@suse.cz>
3997 * multiple_target.c (create_dispatcher_calls): Redirect
3998 reference in the symbol table.
4000 2018-03-16 Martin Liska <mliska@suse.cz>
4003 * multiple_target.c (create_dispatcher_calls): Redirect also
4006 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4010 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4011 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4012 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4014 2018-03-16 Julia Koval <julia.koval@intel.com>
4016 * doc/invoke.texi (Skylake Server): Add CLWB.
4017 Cannonlake): Remove CLWB.
4019 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4021 PR tree-optimization/84841
4022 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4024 (FLOAT_ONE_CONST_TYPE): Define.
4025 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4026 (sort_by_operand_rank): Put entries with higher constant_type last
4027 rather than first to match comments.
4029 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4031 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4034 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4037 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4038 check and use address of target always.
4040 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4043 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4044 (indirect_thunk_bnd_needed): Likewise.
4045 (indirect_thunks_used): Likewise.
4046 (indirect_thunks_bnd_used): Likewise.
4047 (indirect_return_needed): New.
4048 (indirect_return_bnd_needed): Likewise.
4049 (output_indirect_thunk_function): Add a bool argument for
4051 (output_indirect_thunk_function): Don't generate alias for
4052 function return thunk.
4053 (ix86_code_end): Call output_indirect_thunk_function to generate
4054 function return thunks.
4055 (ix86_output_function_return): Set indirect_return_bnd_needed
4056 and indirect_return_needed instead of indirect_thunk_bnd_needed
4057 and indirect_thunk_needed.
4059 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4061 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4062 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4063 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4065 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4066 Paul Hua <paul.hua.gm@gmail.com>
4069 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4071 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4073 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4074 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4077 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4080 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4081 instead of GET_MODE_SIZE when comparing Units.
4083 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4086 * varasm.c (hash_section): Return an unchangeble hash value
4087 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4088 Return !aarch64_can_use_per_function_literal_pools_p ().
4090 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4093 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4094 rather than address of cmode as last argument to prepare_cmp_insn.
4096 2018-03-15 Julia Koval <julia.koval@intel.com>
4098 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4099 F_AVX512VNNI, F_AVX512BITALG): New.
4101 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4104 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4105 insn for floating-point loads and stores.
4107 2018-03-14 Carl Love <cel@us.ibm.com>
4109 * config/rs6000/rs6000-c.c: Add macro definitions for
4110 ALTIVEC_BUILTIN_VEC_PERMXOR.
4111 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4112 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4113 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4114 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4116 * config/doc/extend.texi: Add prototypes for vec_permxor.
4118 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4121 * diagnostic-show-locus.c (class layout_point): Convert m_line
4122 from int to linenum_type.
4123 (line_span::comparator): Use linenum "compare" function when
4124 comparing line numbers.
4125 (test_line_span): New function.
4126 (layout_range::contains_point): Convert param "row" from int to
4128 (layout_range::intersects_line_p): Likewise.
4129 (layout::will_show_line_p): Likewise.
4130 (layout::print_source_line): Likewise.
4131 (layout::should_print_annotation_line_p): Likewise.
4132 (layout::print_annotation_line): Likewise.
4133 (layout::print_leading_fixits): Likewise.
4134 (layout::annotation_line_showed_range_p): Likewise.
4135 (struct line_corrections): Likewise for field m_row.
4136 (line_corrections::line_corrections): Likewise for param "row".
4137 (layout::print_trailing_fixits): Likewise.
4138 (layout::get_state_at_point): Likewise.
4139 (layout::get_x_bound_for_row): Likewise.
4140 (layout::print_line): Likewise.
4141 (diagnostic_show_locus): Likewise for locals "last_line" and
4143 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4144 * input.c (selftest::test_linenum_comparisons): New function.
4145 (selftest::input_c_tests): Call it.
4146 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4147 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4148 * selftest.h (ASSERT_GT): New macro.
4149 (ASSERT_GT_AT): New macro.
4150 (ASSERT_LT): New macro.
4151 (ASSERT_LT_AT): New macro.
4153 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4155 PR rtl-optimization/84780
4156 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4158 2018-03-14 Martin Liska <mliska@suse.cz>
4160 * tree.c (record_node_allocation_statistics): Use
4161 get_stats_node_kind.
4162 (get_stats_node_kind): New function extracted from
4163 record_node_allocation_statistics.
4164 (free_node): Use get_stats_node_kind.
4166 2018-03-14 Richard Biener <rguenther@suse.de>
4168 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4169 that the value-set of ANTIC_IN doesn't grow.
4172 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4174 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4175 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4177 2018-03-14 Julia Koval <julia.koval@intel.com>
4179 * config.gcc (icelake-client, icelake-server): New.
4181 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4182 (initial_ix86_arch_features): Ditto.
4183 (PTA_SKYLAKE): Add SGX.
4184 (PTA_ICELAKE): Remove.
4185 (PTA_ICELAKE_CLIENT): New.
4186 (PTA_ICELAKE_SERVER): New.
4187 (ix86_option_override_internal): Split up icelake on icelake client and
4189 (get_builtin_code_for_version): Ditto.
4190 (fold_builtin_cpu): Ditto.
4191 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4192 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4193 * config/i386/i386.h (processor_type): Ditto.
4194 * doc/invoke.texi: Ditto.
4196 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4199 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4200 INTEGER_CST offset, add it together with bitpos / 8 and
4201 sign extend based on POINTER_SIZE.
4205 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4208 * config/i386/constraints.md (Yc): New register constraint.
4209 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4210 Use Yc constraint for alternative 2 of operand 0. Remove
4211 preferred_for_speed attribute.
4213 2018-03-14 Richard Biener <rguenther@suse.de>
4215 PR tree-optimization/84830
4216 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4217 with the old one to avoid oscillations.
4219 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4222 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4224 (assign_by_spills): Return a flag of reload assignment failure.
4225 Do not process the reload assignment failures. Do not spill other
4226 reload pseudos if they has the same reg class. Update n if
4228 (lra_assign): Add a return arg. Set up from the result of
4229 assign_by_spills call.
4230 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4231 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4232 usage_insns if it is not NULL.
4233 (spill_hard_reg_in_range): New function.
4234 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4235 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4236 function prototypes.
4237 (lra_assign): Change prototype.
4238 * lra.c (lra): Add code to deal with fails by splitting hard reg
4241 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4243 * config/riscv/riscv.opt (mrelax): New option.
4244 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4245 "norelax" when riscv_mrelax is disabled.
4246 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4248 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4251 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4252 reassociation for int modes.
4254 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4256 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4257 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4259 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4260 * config/aarch64/aarch64-sve.md
4261 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4262 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4263 (*extend<mode><Vwide>2): Rename to...
4264 (aarch64_sve_extend<mode><Vwide>2): ...this.
4265 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4266 renaming the old pattern to...
4267 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4269 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4270 define_expand, renaming the old pattern to...
4271 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4272 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4273 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4274 account when deciding which SVE instruction the optab should use.
4275 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4277 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4279 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4280 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4281 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4282 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4283 (tlsdesc_small_<mode>): Turn a define_expand and use
4284 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4285 (tlsdesc_small_advsimd_<mode>): ...this.
4286 (tlsdesc_small_sve_<mode>): New pattern.
4288 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4290 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4291 (UNSPEC_UMUL_HIGHPART): New constants.
4292 (MUL_HIGHPART): New int iteraor.
4293 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4294 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4296 (*<su>mul<mode>3_highpart): New define_insn.
4298 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4301 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4304 2018-03-13 Martin Liska <mliska@suse.cz>
4307 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4309 (sem_item_optimizer::~sem_item_optimizer): Release it.
4310 (sem_item_optimizer::merge_classes): Register variable aliases.
4311 (sem_item_optimizer::fixup_pt_set): New function.
4312 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4313 * ipa-icf.h: Declare new variables and functions.
4315 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4318 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4319 integer_pow2p@2 and test integer_pow2p in condition.
4320 (A < 0 ? C : 0): Similarly for @1.
4323 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4324 characters starting at p contain '\0' character, don't look beyond
4328 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4329 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4332 * reg-stack.c (change_stack): Change update_end var from int to
4333 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4334 also call set_block_for_insn on the newly added insns and rescan.
4337 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4338 on the last operand.
4341 * tree.c (stabilize_reference_1): Return save_expr (e) for
4342 STATEMENT_LIST even if it doesn't have side-effects.
4344 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4346 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4348 2018-03-12 Renlin Li <renlin.li@arm.com>
4350 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4351 aarch64_output_scalar_simd_mov_immediate.
4353 2018-03-12 Martin Sebor <msebor@redhat.com>
4355 PR tree-optimization/83456
4356 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4357 for perfectly overlapping calls to memcpy.
4358 (gimple_fold_builtin_memory_chk): Same.
4359 (gimple_fold_builtin_strcpy): Handle no-warning.
4360 (gimple_fold_builtin_stxcpy_chk): Same.
4361 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4363 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4365 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4366 parameter. Use it for SFmode.
4367 (rs6000_function_arg_advance_1): Adjust.
4368 (rs6000_function_arg): Adjust.
4369 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4371 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4373 PR rtl-optimization/84169
4374 PR rtl-optimization/84780
4375 * combine.c (can_combine_p): Check for a 2-insn combination whether
4376 the destination register is used between the two insns, too.
4378 2018-03-12 Richard Biener <rguenther@suse.de>
4380 PR tree-optimization/84803
4381 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4382 for refs DR analysis didn't process.
4384 2018-03-12 Richard Biener <rguenther@suse.de>
4386 PR tree-optimization/84777
4387 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4388 force-vectorize loops ignore whether we are optimizing for size.
4390 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4392 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4393 (TARGET_MD_ASM_ADJUST): Define.
4395 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4396 Kito Cheng <kito.cheng@gmail.com>
4397 Chung-Ju Wu <jasonwucj@gmail.com>
4399 * config/nds32/nds32.c (nds32_compute_stack_frame,
4400 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4401 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4402 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4403 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4404 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4405 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4406 * config/nds32/nds32.md (prologue, epilogue): Use macro
4407 NDS32_V3PUSH_AVAILABLE_P to do checking.
4409 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4412 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4413 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4414 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4415 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4416 addition of most attributes on !orig_type_die or the attribute not
4417 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4419 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4420 Chung-Ju Wu <jasonwucj@gmail.com>
4422 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4424 * config/nds32/nds32.opt (mvh): New option.
4426 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4427 Chung-Ju Wu <jasonwucj@gmail.com>
4429 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4431 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4432 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4435 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4436 Chung-Ju Wu <jasonwucj@gmail.com>
4438 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4440 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4441 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4443 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4444 Kito Cheng <kito.cheng@gmail.com>
4445 Chung-Ju Wu <jasonwucj@gmail.com>
4447 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4448 UNSPEC_FFMISM and UNSPEC_FLMISM.
4449 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4450 for ffb, ffmism and flmism.
4451 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4452 (unspec_ffmism): Ditto.
4453 (unspec_flmism): Ditto.
4454 (nds32_expand_builtin_impl): Check if string extension is available.
4455 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4456 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4458 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4461 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4464 * lra-assigns.c (assign_by_spills): Return a flag of reload
4465 assignment failure. Do not process the reload assignment
4466 failures. Do not spill other reload pseudos if they has the same
4468 (lra_assign): Add a return arg. Set up from the result of
4469 assign_by_spills call.
4470 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4471 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4472 usage_insns if it is not NULL.
4473 (spill_hard_reg_in_range): New function.
4474 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4475 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4476 function prototypes.
4477 (lra_assign): Change prototype.
4478 * lra.c (lra): Add code to deal with fails by splitting hard reg
4481 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4484 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4486 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4489 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4490 (dw_val_node): Add val_symbolic_view.
4491 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4492 (symview_upper_bound): New.
4493 (new_line_info_table): Initialize symviews_since_reset.
4494 (dwarf2out_source_line): Count symviews_since_reset and set
4495 symview_upper_bound.
4496 (dw_val_equal_p): Handle symview.
4497 (add_AT_symview): New.
4498 (print_dw_val): Handle symview.
4499 (attr_checksum, attr_checksum_ordered): Likewise.
4500 (same_dw_val_p, size_of_die): Likewise.
4501 (value_format, output_die): Likewise.
4502 (add_high_low_attributes): Use add_AT_symview for entry_view.
4503 (dwarf2out_finish): Reset symview_upper_bound, clear
4506 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4509 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4510 Add strict argument and use it.
4511 (rs6000_split_multireg_move): Update for new strict argument.
4512 (mem_operand_gpr): Disallow all non-offsettable addresses.
4513 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4515 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4518 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4519 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4520 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4523 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4524 decl, use remap_type if we want to use the type.
4526 2018-03-09 Martin Sebor <msebor@redhat.com>
4528 PR tree-optimization/84526
4529 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4531 (builtin_access::generic_overlap): Be prepared to handle non-array
4534 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4536 PR rtl-optimization/84682
4537 * lra-constraints.c (process_address_1): Check is_address flag
4538 for address constraints.
4539 (process_alt_operands): Likewise.
4540 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4541 preprocess_constraints.
4542 * recog.h (preprocess_constraints): Add oploc parameter.
4544 * recog.c (preprocess_constraints): Test address_operand for
4545 CT_ADDRESS constraints.
4547 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4550 * lra-assigns.c (assign_by_spills): Return a flag of reload
4551 assignment failure. Do not process the reload assignment
4552 failures. Do not spill other reload pseudos if they has the same
4554 (lra_assign): Add a return arg. Set up from the result of
4555 assign_by_spills call.
4556 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4557 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4558 usage_insns if it is not NULL.
4559 (spill_hard_reg_in_range): New function.
4560 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4561 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4562 function prototypes.
4563 (lra_assign): Change prototype.
4564 * lra.c (lra): Add code to deal with fails by splitting hard reg
4567 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4570 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4571 Accept complain bool parameter. Only emit errors if it is true.
4572 (arm_parse_cpu_option_name): Likewise.
4573 (arm_target_thumb_only): Adjust callers of the above.
4574 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4575 prototype to take a default true bool parameter.
4576 (arm_parse_arch_option_name): Likewise.
4578 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4579 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4583 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4584 * configure: Regenerate.
4585 * configure.ac ("linker --version-script option"): New.
4586 ("linker soname option"): New.
4588 2018-03-09 Richard Biener <rguenther@suse.de>
4590 PR tree-optimization/84775
4591 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4592 immediate uses of predicate stmts and mark them modified.
4595 PR tree-optimization/84178
4596 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4598 (version_loop_for_if_conversion): Delay update_ssa call.
4599 (tree_if_conversion): Delay update_ssa until after predicate
4602 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4605 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4606 when the function accesses prior frames.
4608 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4611 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4612 gen_llsym, otherwise call maybe_gen_llsym.
4615 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4616 has ',' character inside of it.
4618 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4621 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4622 as clobbering CC_REGNUM.
4624 2018-03-08 Richard Biener <rguenther@suse.de>
4627 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4628 (follow_copies_to_constant): Do not follow SSA names registered
4631 2018-03-08 Richard Biener <rguenther@suse.de>
4633 PR tree-optimization/84178
4634 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4636 (version_loop_for_if_conversion): Delay update_ssa call.
4637 (tree_if_conversion): Delay update_ssa until after predicate
4640 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4642 PR tree-optimization/84178
4643 * tree-if-conv.c (release_bb_predicate): Remove the
4644 the assertion that the stmts have NULL use_ops.
4645 Discard the statements, asserting that they haven't
4646 yet been added to a BB.
4648 2018-03-08 Richard Biener <rguenther@suse.de>
4650 PR tree-optimization/84746
4651 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4652 (phi_translate): Pass in destination ANTIC_OUT set.
4653 (phi_translate_1): Likewise. For a simplified result lookup
4654 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4655 (phi_translate_set): Adjust.
4656 (do_pre_regular_insertion): Likewise.
4657 (do_pre_partial_partial_insertion): Likewise.
4659 2018-03-08 Martin Liska <mliska@suse.cz>
4661 PR gcov-profile/84735
4662 * doc/gcov.texi: Document usage of profile files.
4663 * gcov-io.h: Document changes in the format.
4665 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4669 * dwarf2out.c (struct dw_line_info_table): Update comments for
4671 (FORCE_RESET_NEXT_VIEW): New.
4672 (FORCE_RESETTING_VIEW_P): New.
4673 (RESETTING_VIEW_P): Check for -1 too.
4674 (ZERO_VIEW_P): Likewise.
4675 (new_line_info_table): Force-reset next view.
4676 (dwarf2out_begin_function): Likewise.
4677 (dwarf2out_source_line): Simplify zero_view_p initialization.
4678 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4679 view directly. Omit view when omitting .loc at line 0.
4681 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4683 PR tree-optimization/84740
4684 * tree-switch-conversion.c (process_switch): Call build_constructors
4685 only if info.phi_count is non-zero.
4687 PR tree-optimization/84739
4688 * tree-tailcall.c (find_tail_calls): Check call arguments against
4689 DECL_ARGUMENTS (current_function_decl) rather than
4690 DECL_ARGUMENTS (func) when checking for tail recursion.
4692 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4694 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4695 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4696 Volker Reichelt's entry and add entries for people that perform
4697 GCC fuzzy testing and report numerous bugs.
4699 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4702 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4703 readonly data in sdata, if that is disabled.
4704 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4705 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4706 -mreadonly-in-sdata option.
4708 2018-03-07 Martin Sebor <msebor@redhat.com>
4710 PR tree-optimization/84468
4711 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4712 basic block when looking for nul assignment.
4714 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4717 * except.h (output_function_exception_table): Adjust prototype.
4718 * except.c (output_function_exception_table): Remove FNNAME parameter
4719 and add SECTION parameter. Ouput one part of the table at a time.
4720 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4721 the first part of the exception table and emit unwind directives.
4722 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4723 (i386_pe_seh_cold_init): Likewise.
4724 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4725 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4726 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4727 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4728 * config/i386/winnt.c: Include except.h.
4729 (struct seh_frame_state): Add reg_offset, after_prologue and
4730 in_cold_section fields.
4731 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4732 (i386_pe_seh_cold_init): New function.
4733 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4734 to seh->in_cold_section.
4735 (seh_emit_push): Record the offset of the push.
4736 (seh_emit_save): Record the offet of the save.
4737 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4738 Test seh->after_prologue to disregard the epilogue.
4739 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4740 (i386_pe_end_cold_function): New function.
4742 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4745 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4746 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4749 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4751 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4752 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4755 * multiple_target.c: Include tree-inline.h and intl.h.
4756 (expand_target_clones): Diagnose and fail if node->definition and
4757 !tree_versionable_function_p (node->decl).
4759 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4761 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4763 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4764 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4765 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4767 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4770 * combine.c (try_combine): Use reg_or_subregno instead of handling
4771 just paradoxical SUBREGs and REGs.
4773 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4775 * config/arc/arc.c (arc_finalize_pic): Remove function.
4776 (arc_must_save_register): We use single base PIC register, remove
4777 checks to save/restore the PIC register.
4778 (arc_expand_prologue): Likewise.
4779 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4781 (arc_verify_short): Likewise.
4782 (arc_attr_type): Likewise.
4783 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4784 (walk_stores): Likewise.
4785 (arc_address_cost): Make it static.
4786 (arc_verify_short): Likewise.
4787 (branch_dest): Likewise.
4788 (arc_attr_type): Likewise.
4789 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4790 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4791 (arc_final_prescan_insn): Remove inserting the nops due to
4792 hardware hazards. It is done in reorg step.
4793 (insn_length_variant_t): Remove.
4794 (insn_length_parameters_t): Likewise.
4795 (arc_insn_length_parameters): Likewise.
4796 (arc_get_insn_variants): Likewise.
4797 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4799 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4802 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4805 PR tree-optimization/84687
4806 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4808 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4810 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4812 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4813 Rename to ppc_speculation_barrier.
4814 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4815 __builtin_ppc_speculation_barrier.
4817 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4820 * combine.c (combine_simplify_rtx): Don't try to simplify if
4821 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4824 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4826 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4827 to 32 bytes when compiling for POWER9.
4829 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4832 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4833 regparm >= 3 with no arg reg available also for calls with
4834 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4837 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4839 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4841 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4844 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4846 2018-03-05 Richard Biener <rguenther@suse.de>
4848 PR tree-optimization/84486
4849 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4850 When inserting a __builtin_assume_aligned call set the LHS
4851 SSA name alignment info accordingly.
4853 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4855 PR tree-optimization/84114
4856 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4857 Avoid reassociation of FLOAT_MODE addition.
4859 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4861 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4862 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4863 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4864 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4865 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4866 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4867 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4869 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4870 __builtin_ia32_wbinvd): New builtins.
4871 (SPECIAL_ARGS2): New.
4872 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4873 (SPECIAL_ARGS2): New.
4874 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4875 (ix86_valid_target_attribute_inner_p): Ditto.
4876 (ix86_init_mmx_sse_builtins): Add special_args2.
4877 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4878 TARGET_WBNOINVD_P): New.
4879 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4880 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4881 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4882 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4883 * config/i386/pconfigintrin.h: New file.
4884 * config/i386/wbnoinvdintrin.h: Ditto.
4885 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4886 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4888 2018-03-05 Richard Biener <rguenther@suse.de>
4890 PR tree-optimization/84670
4891 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4893 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4894 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4895 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4896 doesn't grow if all successors (recursively) were visited at least
4899 2018-03-05 Richard Biener <rguenther@suse.de>
4901 PR tree-optimization/84650
4902 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4903 if executed in the loop pipeline.
4905 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4907 * doc/configfiles.texi (Configuration Files): Move info about
4908 conditionalizing $target-protos.h to...
4909 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4910 differs from $target-protos.h.
4912 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4913 Chung-Ju Wu <jasonwucj@gmail.com>
4915 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4916 * config/nds32/nds32-multiple.md (setmemsi): Define.
4917 * config/nds32/nds32-memory-manipulation.c
4918 (nds32_gen_dup_4_byte_to_word_value): New.
4919 (emit_setmem_word_loop): New.
4920 (emit_setmem_byte_loop): New.
4921 (nds32_expand_setmem_loop): New.
4922 (nds32_expand_setmem_loop_v3m): New.
4923 (nds32_expand_setmem_unroll): New.
4924 (nds32_expand_setmem): New.
4926 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4927 Chung-Ju Wu <jasonwucj@gmail.com>
4929 * config/nds32/nds32-memory-manipulation.c
4930 (nds32_emit_load_store): New.
4931 (nds32_emit_post_inc_load_store): New.
4932 (nds32_emit_mem_move): New.
4933 (nds32_emit_mem_move_block): New.
4934 (nds32_expand_movmemsi_loop_unknown_size): New.
4935 (nds32_expand_movmemsi_loop_known_size): New.
4936 (nds32_expand_movmemsi_loop): New.
4937 (nds32_expand_movmemsi_unroll): New.
4938 (nds32_expand_movmemqi): Rename ...
4939 (nds32_expand_movmemsi): ... to this.
4940 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4941 (movmemsi): ... to this.
4942 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4943 (nds32_expand_movmemsi): ... to this.
4945 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4946 Monk Chiang <sh.chiang04@gmail.com>
4947 Chung-Ju Wu <jasonwucj@gmail.com>
4949 * config/nds32/nds32-protos.h
4950 (nds32_expand_load_multiple): New arguments.
4951 (nds32_expand_store_multiple): Ditto.
4952 (nds32_valid_multiple_load_store): Rename ...
4953 (nds32_valid_multiple_load_store_p): ... to this.
4954 * config/nds32/nds32-memory-manipulation.c
4955 (nds32_expand_load_multiple): Refine implementation.
4956 (nds32_expand_store_multiple): Ditto.
4957 * config/nds32/nds32-multiple.md
4958 (load_multiple): Update nds32_expand_load_multiple interface.
4959 (store_multiple): Update nds32_expand_store_multiple interface.
4960 * config/nds32/nds32-predicates.c
4961 (nds32_valid_multiple_load_store): Rename ...
4962 (nds32_valid_multiple_load_store_p): ... to this and refine
4964 * config/nds32/predicates.md
4965 (nds32_load_multiple_and_update_address_operation): New predicate.
4966 (nds32_store_multiple_and_update_address_operation): New predicate.
4968 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4969 Chung-Ju Wu <jasonwucj@gmail.com>
4971 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4972 (combo): New attribute.
4973 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4975 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4977 * config/nds32/nds32.opt: Change -mcmodel= default value.
4979 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4980 Monk Chiang <sh.chiang04@gmail.com>
4981 Chung-Ju Wu <jasonwucj@gmail.com>
4983 * config/nds32/constants.md (unspec_element): New enum.
4984 * config/nds32/constraints.md (Umw): New constraint.
4985 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4986 * config/nds32/nds32-intrinsic.md: Likewise.
4987 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
4988 (nds32_valid_smw_lwm_base_p): New.
4989 (nds32_output_smw_single_word): New.
4990 (nds32_output_lmw_single_word): New.
4991 (nds32_expand_unaligned_load): New.
4992 (nds32_expand_unaligned_store): New.
4993 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4994 (nds32_output_smw_single_word): Declare.
4995 (nds32_output_lmw_single_word): Declare.
4996 (nds32_expand_unaligned_load): Declare.
4997 (nds32_expand_unaligned_store): Declare.
4998 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4999 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5000 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5001 NDS32_BUILTIN_UASTORE_DW.
5002 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5005 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5006 Kito Cheng <kito.cheng@gmail.com>
5007 Chung-Ju Wu <jasonwucj@gmail.com>
5009 * config/nds32/nds32-intrinsic.c
5010 (nds32_expand_builtin_null_ftype_reg): Delete.
5011 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5012 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5013 (nds32_read_argument): New.
5014 (nds32_legitimize_target): Ditto.
5015 (nds32_legitimize_argument): Ditto.
5016 (nds32_check_constant_argument): Ditto.
5017 (nds32_expand_unop_builtin): Ditto.
5018 (nds32_expand_unopimm_builtin): Ditto.
5019 (nds32_expand_binop_builtin): Ditto.
5020 (nds32_builtin_decl_impl): Ditto.
5021 (builtin_description): Ditto.
5022 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5023 (nds32_init_builtins_impl): Ditto.
5024 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5025 (nds32_builtin_decl): New.
5026 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5027 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5029 2018-03-02 Jeff Law <law@redhat.com>
5031 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5032 (redundant_insn, fill_simple_delay_slots): Likewise.
5033 (fill_slots_from_thread): Likewise.
5034 * resource.c (mark_referenced_resources): Likewise.
5035 (mark_set_resources, find_dead_or_set_registers): Likewise.
5037 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5039 * substring-locations.h (format_warning_va): Formatting fix for
5041 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5043 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5044 * substring-locations.c: Include intl.h.
5045 (format_warning_va): Turned into small wrapper around
5046 format_warning_n_va, renamed to ...
5047 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5048 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5050 (format_warning_at_substring_n): New function.
5051 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5052 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5053 format_warning_at_substring with just a shorter name instead of
5054 const function pointer.
5055 (fmtwarn_n): New function.
5056 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5057 appropriate, get rid of all the fmtstr temporaries, move conditionals
5058 with G_() wrapped string literals directly into fmtwarn arguments,
5059 cast dir.len to (int), formatting fixes.
5061 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5063 * doc/invoke.texi: Remove "Cilk Plus" references.
5065 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5066 Richard Biener <rguenther@suse.de>
5069 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5070 for error or warning attributes if CALL_FROM_THUNK_P is set.
5073 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5076 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5077 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5080 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5081 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5083 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5084 instead of -1U in last predictors element's probability member.
5086 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5089 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5090 arguments if they are comparable.
5092 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5094 PR tree-optimization/84634
5095 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5096 masks and masked_loop_p with a single loop_masks, making sure it's
5097 null for bb vectorization.
5099 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5101 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5102 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5103 loop->force_vectorize to check whether there is no alias.
5105 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5108 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5110 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5112 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5113 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5114 instead of a loop around prev_real_insn.
5115 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5119 * config/i386/i386.c (ix86_print_operand): Use conditional
5120 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5123 2018-03-02 Richard Biener <rguenther@suse.de>
5125 PR tree-optimization/84427
5126 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5127 (bitmap_set_subtract_values): Rewrite to handle multiple
5130 (prune_clobbered_mems): Likewise.
5131 (phi_translate): Take edge instead of pred/phiblock.
5132 (phi_translate_1): Likewise.
5133 (phi_translate_set): Likewise. Insert all translated
5134 exprs for a value into the set, keeping possibly multiple
5135 expressions per value.
5136 (compute_antic_aux): Adjust for phi_translate changes.
5137 When intersecting union the expressions and prune those
5138 not in the final value set, keeping possibly multiple
5139 expressions per value. Do not use value-insertion
5140 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5141 all expressions. Add verification that the value-sets
5142 only shrink during iteration.
5143 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5144 (do_pre_regular_insertion): Likewise.
5145 (do_pre_partial_partial_insertion): Likewise.
5147 2018-03-02 Richard Biener <rguenther@suse.de>
5150 * config/darwin.c (saved_debug_info_level): New static global.
5151 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5152 (darwin_asm_lto_end): Restore debug info generation settings.
5154 2018-03-01 Martin Liska <mliska@suse.cz>
5157 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5160 2018-03-01 Richard Biener <rguenther@suse.de>
5163 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5164 type annotation with LTO when debug was disabled at compile-time.
5166 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5168 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5170 (mips_final_postscan_insn): Likewise.
5172 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5174 PR rtl-optimization/84528
5175 * alias.c (init_alias_target): Add commentary.
5176 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5177 a unique base value if the frame pointer is not eliminated
5178 to the stack pointer.
5180 2018-03-01 Tom de Vries <tom@codesourcery.com>
5182 PR rtl-optimization/83327
5183 * lra-int.h (hard_regs_spilled_into): Declare.
5184 * lra.c (hard_regs_spilled_into): Define.
5185 (init_reg_info): Init hard_regs_spilled_into.
5186 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5187 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5188 (process_bb_lives): Handle hard_regs_spilled_into.
5189 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5190 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5192 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5194 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5195 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5196 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5197 * config/rs6000/aix72.h: New file.
5199 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5201 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5202 instead of warning_at with conditional singular and plural messages
5206 * stor-layout.c (update_alignment_for_field): For
5207 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5208 && !DECL_PACKED (field), do the alignment update, just use
5209 only desired_align instead of MAX (type_align, desired_align)
5211 (place_field): Don't do known_align < desired_align handling
5212 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5213 is non-NULL, instead do it after rli->prev_field handling and
5214 only if not within a bitfield word. For DECL_PACKED (field)
5215 use type_align of BITS_PER_UNIT.
5217 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5219 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5220 superfluous parentheses and trailing spaces.
5222 2018-02-28 Richard Biener <rguenther@suse.de>
5224 PR tree-optimization/84584
5225 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5226 SCoPs with fake exit edge.
5228 2018-02-28 Martin Liska <mliska@suse.cz>
5231 * timevar.c (timer::print): Fix format to properly print 100%
5234 2018-02-28 Richard Biener <rguenther@suse.de>
5237 * genmatch.c (capture_info::walk_match): Do not mark
5238 captured expressions without operands as expr_p given
5239 they act more like predicates and should be subject to
5240 "lost tail" side-effect preserving.
5242 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5244 PR rtl-optimization/81611
5245 * auto-inc-dec.c (attempt_change): Move dead note from
5246 mem_insn if it's the next use of regno
5247 (find_address): Take address use of reg holding
5248 non-incremented value. Add parm to limit search to the named
5250 (merge_in_block): Attempt to use a mem insn that is the next
5251 use of the original regno.
5253 2018-02-27 Martin Sebor <msebor@redhat.com>
5256 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5257 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5259 2018-02-27 Martin Sebor <msebor@redhat.com>
5261 PR translation/84207
5262 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5263 n argument to unsigned HOST_WIDE_INT.
5264 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5265 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5266 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5267 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5269 2018-02-27 Richard Biener <rguenther@suse.de>
5271 PR tree-optimization/84512
5272 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5273 Do not use the estimate returned from record_stmt_cost for
5274 the scalar iteration cost but sum properly using add_stmt_cost.
5276 2018-02-27 Richard Biener <rguenther@suse.de>
5278 PR tree-optimization/84466
5279 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5280 Adjust last change to less strictly validate use operands.
5282 2018-02-27 Martin Liska <mliska@suse.cz>
5284 PR gcov-profile/84548
5285 * gcov.c (process_file): Allow partial overlap and consider it
5286 also as group functions.
5287 (output_lines): Properly calculate range of lines for a group.
5289 2018-02-27 Martin Liska <mliska@suse.cz>
5291 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5292 'ggc' suffixes. Change first column width.
5293 (timer::print): Fix formatting of the column.
5295 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5297 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5298 preserve inline entry blocks for the sake of debug inline
5299 entry point markers alone.
5300 (remove_unused_locals): Suggest in comments a better place to
5301 force the preservation of inline entry blocks that are
5302 otherwise unused, but do not preserve them.
5304 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5306 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5308 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5311 * config/i386/constraints.md (Bs): Replace
5312 ix86_indirect_branch_register with
5313 TARGET_INDIRECT_BRANCH_REGISTER.
5315 * config/i386/i386.md (indirect_jump): Likewise.
5316 (tablejump): Likewise.
5317 (*sibcall_memory): Likewise.
5318 (*sibcall_value_memory): Likewise.
5319 Peepholes of indirect call and jump via memory: Likewise.
5320 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5321 (*sibcall_value_GOT_32): Likewise.
5322 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5323 (GOT_memory_operand): Likewise.
5324 (call_insn_operand): Likewise.
5325 (sibcall_insn_operand): Likewise.
5326 (GOT32_symbol_operand): Likewise.
5327 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5329 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5331 PR rtl-optimization/83496
5332 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5333 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5334 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5335 redundant insn, if any.
5336 (relax_delay_slots): Likewise.
5337 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5339 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5341 PR tree-optimization/83965
5342 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5343 that grouped statements are part of a reduction chain. Return
5344 true if the statement is not marked as a reduction itself but
5346 (vect_recog_dot_prod_pattern): Don't check whether the statement
5347 is part of a group here.
5348 (vect_recog_sad_pattern): Likewise.
5349 (vect_recog_widen_sum_pattern): Likewise.
5351 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5354 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5356 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5359 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5361 (ix86_output_indirect_function_return): New prototype.
5362 (ix86_split_simple_return_pop_internal): Likewise.
5363 * config/i386/i386.c (indirect_return_via_cx): New.
5364 (indirect_return_via_cx_bnd): Likewise.
5365 (indirect_thunk_name): Handle return va CX_REG.
5366 (output_indirect_thunk_function): Create alias for
5367 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5368 (ix86_output_indirect_jmp): Remove the bool argument.
5369 (ix86_output_indirect_function_return): New function.
5370 (ix86_split_simple_return_pop_internal): Likewise.
5371 * config/i386/i386.md (*indirect_jump): Don't pass false
5372 to ix86_output_indirect_jmp.
5373 (*tablejump_1): Likewise.
5374 (simple_return_pop_internal): Change it to define_insn_and_split.
5375 Call ix86_split_simple_return_pop_internal to split it for
5377 (simple_return_indirect_internal): Call
5378 ix86_output_indirect_function_return instead of
5379 ix86_output_indirect_jmp.
5381 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5384 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5385 memset and value initialization afterwards.
5387 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5389 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5391 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5394 * common/config/aarch64/aarch64-common.c
5395 (aarch_option_optimization_table[]): Switch
5396 off fomit-frame-pointer
5398 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5399 Chung-Ju Wu <jasonwucj@gmail.com>
5401 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5403 (store_multiple): Ditto.
5405 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5407 * config.gcc: Add --with-cpu support for nds32 target.
5408 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5409 * config/nds32/nds32.opt: Add -mcpu= option.
5411 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5413 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5414 isel=yes): Warn for these deprecated options.
5416 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5418 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5419 ISA_2_5_MASKS_EMBEDDED.
5421 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5423 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5424 p->max as pointers rather than using iterative_hash_expr.
5426 2018-02-23 Carl Love <cel@us.ibm.com>
5428 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5429 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5431 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5432 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5433 P8V_BUILTIN_VEC_VUNSIGNED2.
5435 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5438 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5439 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5441 * lra-constraints.c (curr_insn_transform): Set up
5442 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5444 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5445 * lra-eliminations.c (spill_pseudos): Ditto.
5446 (process_insn_for_elimination): Ditto.
5447 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5448 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5449 LRA_NON_CLOBBERED_ALT.
5451 2018-02-22 Martin Sebor <msebor@redhat.com>
5453 PR tree-optimization/84480
5454 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5455 to maybe_diag_stxncpy_trunc. Call it.
5456 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5457 from gimple_fold_builtin_strcpy. Print inlining stack.
5458 (handle_builtin_stxncpy): Print inlining stack.
5459 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5461 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5464 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5465 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5466 and -fcheck-pointer-bounds are used together.
5467 (indirect_thunk_prefix): New enum.
5468 (indirect_thunk_need_prefix): New function.
5469 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5470 "_nt" instead of "_bnd" for NOTRACK prefix.
5471 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5472 (output_indirect_thunk_function): Likewise.
5474 (ix86_code_end): Update output_indirect_thunk_function calls.
5475 (ix86_output_indirect_branch_via_reg): Replace
5476 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5477 (ix86_output_indirect_branch_via_push): Likewise.
5478 (ix86_output_function_return): Likewise.
5479 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5480 incompatible with -fcf-protection=branch and
5481 -fcheck-pointer-bounds.
5483 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5486 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5487 Change gcc_assert call to output_operand_lossage.
5489 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5491 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5493 2018-02-22 DJ Delorie <dj@redhat.com>
5494 Sebastian Perta <sebastian.perta@renesas.com>
5495 Oleg Endo <olegendo@gcc.gnu.org>
5497 * config/rx/rx.c (rx_rtx_costs): New function.
5498 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5500 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5502 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5504 2018-02-22 Martin Liska <mliska@suse.cz>
5507 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5508 Add "native" as a possible value.
5510 2018-02-22 Martin Liska <mliska@suse.cz>
5513 * config/i386/i386.c (ix86_option_override_internal):
5514 Add "native" as a possible value for -march and -mtune.
5516 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5519 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5520 to all type variants.
5522 PR tree-optimization/84503
5523 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5524 width as info->bitpos + info->bitsize - start.
5525 (merged_store_group::merge_overlapping): Simplify width computation.
5526 (check_no_overlap): New function.
5527 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5528 start + width and last_order of the group, fail if check_no_overlap
5530 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5531 to group if check_no_overlap fails.
5533 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5535 * config/rs6000/altivec.md: Delete contraint arguments to
5536 define_expand, define_split, and define_peephole2, and in
5537 define_insn_and_split if always unused.
5538 * config/rs6000/darwin.md: Ditto.
5539 * config/rs6000/dfp.md: Ditto.
5540 * config/rs6000/rs6000.md: Ditto.
5541 * config/rs6000/sync.md: Ditto.
5542 * config/rs6000/vector.md: Ditto.
5543 * config/rs6000/vsx.md: Ditto.
5545 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5547 * config/rs6000/altivec.md: Write output control strings as braced
5548 blocks instead of double-quoted strings.
5549 * config/rs6000/darwin.md: Ditto.
5550 * config/rs6000/rs6000.md: Ditto.
5551 * config/rs6000/vector.md: Ditto.
5552 * config/rs6000/vsx.md: Ditto.
5554 2018-02-21 Jason Merrill <jason@redhat.com>
5556 PR c++/84314 - ICE with templates and fastcall attribute.
5557 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5559 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5561 * ipa-cp.c (determine_versionability): Fix comment typos.
5563 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5566 * ipa-cp.c (determine_versionability): Do not version functions caling
5569 2018-02-21 Martin Liska <mliska@suse.cz>
5572 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5573 Add "native" as a possible value.
5574 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5575 the macro when native cpu detection is available.
5577 2018-02-21 Martin Liska <mliska@suse.cz>
5580 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5581 Add "native" as a possible value.
5582 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5583 when native cpu detection is available.
5585 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5586 Martin Sebor <msebor@redhat.com>
5588 PR tree-optimization/84478
5589 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5591 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5592 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5593 support which is conservatively correct, for 2 only stay conservative
5594 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5595 argument to the 2 argument get_range_strlen, adjust 6 arg
5596 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5598 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5599 (gimple_fold_builtin_strlen): Pass true as last argument to
5602 2018-02-20 Martin Sebor <msebor@redhat.com>
5605 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5606 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5607 (builtin_memref::builtin_memref): Factor out parts into
5608 set_base_and_offset and call it.
5610 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5613 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5614 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5615 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5616 search at the associated MODE_INT.
5618 2018-02-20 Jeff Law <law@redhat.com>
5621 PR tree-optimization/81592
5623 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5624 for range data rather than using global data.
5625 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5626 range data rather than using global data.
5627 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5628 pass it to children as needed.
5629 (struct directive::fmtresult): Similarly.
5630 (struct directive::set_width): Similarly.
5631 (struct directive::set_precision): Similarly.
5632 (format_integer, format_directive, parse_directive): Similarly.
5633 (format_none): Accept unnamed vr_values parameter.
5634 (format_percent, format_floating, format_character): Similarly.
5635 (format_string, format_plain): Similarly.
5636 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5637 the EVRP range analyzer for range data rather than using global data.
5638 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5639 gimple-ssa-evrp-analyze.h
5640 (class sprintf_dom_walker): Add after_dom_children member function.
5641 Add evrp_range_analyzer member.
5642 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5643 range analyzer as needed.
5644 (sprintf_dom_walker::after_dom_children): New member function.
5645 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5647 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5648 (evrp_range_analyzer::pop_to_marker): Likewise.
5650 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5652 PR tree-optimization/84419
5653 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5654 with the required type if its current type is compatible but
5657 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5660 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5661 after vectorization.
5663 2018-02-20 Martin Liska <mliska@suse.cz>
5666 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5667 possible values if we don't have a hint.
5669 2018-02-20 Martin Liska <mliska@suse.cz>
5673 * final.c (shorten_branches): Build align_tab array with one
5675 * opts.c (finish_options): Add alignment option limit check.
5676 (MAX_CODE_ALIGN): Likewise.
5677 (MAX_CODE_ALIGN_VALUE): Likewise.
5678 * doc/invoke.texi: Document maximum allowed option value for
5679 all -falign-* options.
5681 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5684 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5685 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5686 * var-tracking.c (emit_note_insn_var_location): Remove all references
5687 to NOTE_INSN_CALL_ARG_LOCATION.
5688 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5689 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5690 Use copy_rtx_if_shared.
5691 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5692 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5693 (dwarf2out_var_location): Remove handling of
5694 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5696 * final.c (final_scan_insn): Remove all references to
5697 NOTE_INSN_CALL_ARG_LOCATION.
5698 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5699 before dumping final insns.
5700 * except.c (emit_note_eh_region_end): Remove all references to
5701 NOTE_INSN_CALL_ARG_LOCATION.
5702 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5703 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5704 * config/arc/arc.c (hwloop_optimize): Likewise.
5705 * config/arm/arm.c (create_fix_barrier): Likewise.
5706 * config/s390/s390.c (s390_chunkify_start): Likewise.
5707 * config/sh/sh.c (find_barrier): Likewise.
5708 * config/i386/i386.c (rest_of_insert_endbranch,
5709 ix86_seh_fixup_eh_fallthru): Likewise.
5710 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5711 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5712 * config/frv/frv.c (frv_function_prologue): Likewise.
5713 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5715 (note_outside_basic_block_p): Remove all references to
5716 NOTE_INSN_CALL_ARG_LOCATION.
5717 * gengtype.c (adjust_field_rtx_def): Likewise.
5718 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5720 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5721 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5724 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5727 PR tree-optimization/84452
5728 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5729 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5732 2018-02-19 Martin Liska <mliska@suse.cz>
5735 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5737 2018-02-19 Martin Liska <mliska@suse.cz>
5738 Richard Sandiford <richard.sandiford@linaro.org>
5740 PR tree-optimization/82491
5741 * gimple-fold.c (get_base_constructor): Make earlier bail out
5744 2018-02-19 Carl Love <cel@us.ibm.com>
5746 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5747 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5749 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5750 P8V_BUILTIN_VEC_NEG.
5752 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5754 * config/rl78/rl78.md (movdf): New define expand.
5756 2018-02-19 Martin Liska <mliska@suse.cz>
5759 * doc/invoke.texi: Fix typo.
5760 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5762 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5764 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5765 handle rs6000_single_float and rs6000_double_float specially for
5768 2018-02-16 Jeff Law <law@redhat.com>
5770 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5773 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5776 * ipa-inline.c (inline_small_functions): Fix a typo.
5778 2018-02-16 Nathan Sidwell <nathan@acm.org>
5780 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5782 2018-02-16 Carl Love <cel@us.ibm.com>
5784 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5785 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5786 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5787 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5788 expansion to P8V_BUILTIN_VEC_FLOAT2.
5790 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5792 PR rtl-optimization/70023
5793 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5794 src_regno into account.
5796 2018-02-16 Carl Love <cel@us.ibm.com>
5798 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5799 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5800 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5801 * config/rs6000/rs6000.c: Remove case statements for
5802 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5803 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5804 and P9V_BUILTIN_VEC_VINSERT4B.
5805 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5806 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5807 * config/rs6000/vsx.md:
5808 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5811 2018-02-16 Carl Love <cel@us.ibm.com>
5813 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5815 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5817 * config/rs6000/rs6000-c.c: Add the definitions for
5818 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5819 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5820 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5821 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5822 definition for insert4b and define insn *insert3b_internal.
5823 * doc/extend.texi: Add documentation for vec_extract4b.
5825 2018-02-16 Nathan Sidwell <nathan@acm.org>
5827 * doc/extend.texi (Backwards Compatibility): Mention friend
5828 injection. Note for-scope is deprecated.
5829 * doc/invoke.texi (-ffriend-injection): Deprecate.
5831 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5833 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5834 that moved to I2, also allow destinations that are a paradoxical
5835 subreg (instead of a normal reg).
5837 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5840 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5843 2018-02-16 Richard Biener <rguenther@suse.de>
5845 PR tree-optimization/84037
5846 PR tree-optimization/84016
5848 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5849 Adjust vec_construct for the fact we need additional higher latency
5850 128bit inserts for AVX256 and AVX512 vector builds.
5851 (ix86_add_stmt_cost): Scale vector construction cost for
5854 2018-02-16 Richard Biener <rguenther@suse.de>
5856 PR tree-optimization/84417
5857 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5858 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5859 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5861 2018-02-16 Martin Liska <mliska@suse.cz>
5864 * internal-fn.def (ASAN_CHECK): Set proper flags.
5865 (ASAN_MARK): Likewise.
5867 2018-02-16 Julia Koval <julia.koval@intel.com>
5869 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5870 from PTA_CANNONLAKE.
5872 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5875 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5876 Use ++iter rather than iter++ for std::list iterators.
5877 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5878 defer deleting them until all nodes in the forest are processed. Do
5879 free even leaf nodes. Change to_process into auto_vec.
5882 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5883 * vec.h (vec_default_construct): Use memset instead of placement new
5884 if BROKEN_VALUE_INITIALIZATION is defined.
5885 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5886 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5889 PR rtl-optimization/83723
5890 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5891 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5892 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5894 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5896 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5898 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5900 PR rtl-optimization/81443
5901 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5902 from inner REGs to paradoxical SUBREGs.
5904 2018-02-16 Richard Biener <rguenther@suse.de>
5906 PR tree-optimization/84399
5907 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5908 For operands we can analyze at their definition make sure we can
5909 analyze them at each use as well.
5911 2018-02-16 Richard Biener <rguenther@suse.de>
5913 PR tree-optimization/84190
5914 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5915 volatile accesses if the decl isn't volatile.
5917 2018-02-15 Jason Merrill <jason@redhat.com>
5919 PR c++/84314 - ICE with templates and fastcall attribute.
5920 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5921 TYPE_CANONICAL on an existing type.
5923 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5925 PR tree-optimization/84383
5926 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5927 dstoff nor call operand_equal_p if dstbase is NULL.
5929 PR tree-optimization/84334
5930 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5931 also a CONSTANT_CLASS_P, punt.
5933 2018-02-14 Jim Wilson <jimw@sifive.com>
5935 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5936 first SMALL_OPERAND check. New local min_second_step. Move assert
5937 to where locals are set. Add TARGET_RVC support.
5938 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5940 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5942 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5944 2018-02-14 Martin Sebor <msebor@redhat.com>
5946 PR tree-optimization/83698
5947 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5948 arrays constrain the offset range to their bounds.
5949 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5950 (builtin_access::overlap): Avoid setting the size of overlap if it's
5952 (maybe_diag_overlap): Also consider arrays when deciding what values
5953 of offsets to include in diagnostics.
5955 2018-02-14 Martin Sebor <msebor@redhat.com>
5958 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5959 that correspond to the kind of a declaration.
5961 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5964 * config/pa/pa.md: Load address of PIC label using the linkage table
5965 if the label is nonlocal.
5967 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5969 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5970 warning message if user requests -maltivec=be.
5971 * doc/invoke.texi: Document deprecation of -maltivec=be.
5973 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5976 * config/rs6000/rs6000-c.c: Update definitions for
5977 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5978 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5980 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5983 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5984 add _get_ssp intrinsics. Remove argument from
5985 __builtin_ia32_rdssp[d|q].
5986 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5987 * config/i386/i386-builtin.def: Remove argument from
5988 __builtin_ia32_rdssp[d|q].
5989 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
5990 ix86_expand_special_args_builtin for _rdssp[d|q].
5991 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5992 Clear register before usage.
5993 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5994 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5996 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5998 PR tree-optimization/84357
5999 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6000 operand 1 of an ARRAY_REF too.
6002 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6005 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6006 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6008 (set_of_reg): New struct.
6009 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6010 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6011 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6013 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6014 Split into bitclr, bitset, bitinvert patterns if appropriate.
6015 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6016 use rx_fuse_in_memory_bitop.
6017 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6018 to named insn, correct maximum insn length.
6020 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6023 * machmode.def: Define a complex mode for PARTIAL_INT.
6024 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6026 * doc/rtl.texi: Document CSPImode.
6027 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6029 (msp430_hard_regno_nregs_with_padding): Likewise.
6031 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6034 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6036 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6038 PR rtl-optimization/84169
6039 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6040 we generated a parallel as new i3 and we split that to new i2 and i3
6041 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6042 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6043 those to i2, not i1. Partially rewrite this scan code.
6045 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6048 * stor-layout.c (place_field): For variable length fields, adjust
6049 offset_align afterwards not just based on the field's alignment,
6050 but also on the size.
6053 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6054 of exps and logs in the use_exp2 case.
6056 2018-02-13 Jeff Law <law@redhat.com>
6058 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6061 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6064 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6068 * common.opt (gas-loc-support, gas-locview-support): New.
6069 (ginline-points, ginternal-reset-location-views): New.
6070 * doc/invoke.texi: Document them. Use @itemx where intended.
6071 (gvariable-location-views): Adjust.
6072 * target.def (reset_location_view): New.
6073 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6074 (TARGET_RESET_LOCATION_VIEW): New.
6075 * doc/tm.texi: Rebuilt.
6076 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6077 (dwarf2out_default_as_locview_support): New.
6078 (output_asm_line_debug_info): Use option variables.
6079 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6080 (output_loc_list): Likewise.
6081 (add_high_low_attributes): Check option variables.
6082 Don't output entry view attribute in strict mode.
6083 (gen_inlined_subroutine_die): Check option variables.
6084 (dwarf2out_inline_entry): Likewise.
6085 (init_sections_and_labels): Likewise.
6086 (dwarf2out_early_finish): Likewise.
6087 (maybe_reset_location_view): New, from...
6088 (dwarf2out_var_location): ... here. Call it.
6089 * debug.h (dwarf2out_default_as_loc_support): Declare.
6090 (dwarf2out_default_as_locview_support): Declare.
6091 * hooks.c (hook_int_rtx_insn_0): New.
6092 * hooks.h (hook_int_rtx_insn_0): Declare.
6093 * toplev.c (process_options): Take -gas-loc-support and
6094 -gas-locview-support from dwarf2out. Enable
6095 -gvariable-location-views by default only with locview
6096 assembler support. Enable -ginternal-reset-location-views by
6097 default only if the target defines the corresponding hook.
6098 Enable -ginline-points by default if location views are
6099 enabled; force it disabled if statement frontiers are
6101 * tree-inline.c (expand_call_inline): Check option variables.
6102 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6104 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6106 PR tree-optimization/84321
6107 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6108 handling. Also check whether the anti-range contains any values
6109 that satisfy the mask; switch to a VR_RANGE if not.
6111 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6114 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6116 2018-02-13 Martin Jambor <mjambor@suse.cz>
6119 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6120 of call statements, also set location of a load to a temporary.
6122 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6124 * config/rl78/rl78.c (add_vector_labels): New function.
6125 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6126 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6127 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6128 which checks that no arguments are passed.
6129 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6130 * doc/extend.texi: Documentation for the new attribute.
6132 2018-02-13 Andreas Schwab <schwab@suse.de>
6134 * config/riscv/linux.h (CPP_SPEC): Define.
6136 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6139 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6140 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6141 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6142 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6143 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6144 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6145 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6146 temporarily for AES and PCLMUL builtins.
6148 PR tree-optimization/84339
6149 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6150 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6154 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6155 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6156 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6158 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6160 * omp-simd-clone.h: New file.
6161 * omp-simd-clone.c: Include omp-simd-clone.h.
6162 (expand_simd_clones): No longer static.
6163 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6164 cgraph.h and omp-simd-clone.h.
6165 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6166 (vect_recog_widen_shift_pattern): Formatting fix.
6167 (vect_pattern_recog_1): Don't check optab for calls.
6170 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6171 operands[2] into a REG before using gen_lowpart on it.
6173 2018-02-12 Jeff Law <law@redhat.com>
6176 * config/sh/sh.c (find_barrier): Consider a sibling call
6179 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6180 successfully back substituting a reg.
6182 2018-02-12 Richard Biener <rguenther@suse.de>
6184 PR tree-optimization/84037
6185 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6186 parameter, move visited init to caller.
6187 (vect_slp_analyze_operations): Separate cost from validity
6188 check, initialize visited once for all instances.
6189 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6191 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6193 (vect_model_promotion_demotion_cost): Likewise.
6194 (vectorizable_bswap): Guard cost modeling with !slp_node
6195 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6197 (vectorizable_call): Likewise.
6198 (vectorizable_conversion): Likewise.
6199 (vectorizable_assignment): Likewise.
6200 (vectorizable_shift): Likewise.
6201 (vectorizable_operation): Likewise.
6202 (vectorizable_store): Likewise.
6203 (vectorizable_load): Likewise.
6204 (vectorizable_condition): Likewise.
6205 (vectorizable_comparison): Likewise.
6207 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6210 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6211 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6212 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6214 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6217 * params.def (inline-min-speedup): Increase from 8 to 15.
6218 (max-inline-insns-auto): Decrease from 40 to 30.
6219 * ipa-split.c (consider_split): Add some buffer for function to
6220 be considered inlining candidate.
6221 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6224 2018-02-12 Richard Biener <rguenther@suse.de>
6226 PR tree-optimization/84037
6227 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6228 matched stmts if we cannot swap the non-matched ones.
6230 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6232 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6233 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6234 _mm_maskz_scalef_round_ss): New intrinsics.
6235 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6236 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6237 __builtin_ia32_scalefss_round): Remove.
6238 (__builtin_ia32_scalefsd_mask_round,
6239 __builtin_ia32_scalefss_mask_round): New intrinsics.
6240 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6241 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6242 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6243 "<round_constraint>")): Changed to ...
6244 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6245 "<round_scalar_constraint>")): ... this.
6246 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6247 %0, %1, %2<round_op3>}"): Changed to ...
6248 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6249 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6250 %2<round_scalar_mask_op3>}"): ... this.
6251 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6253 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6255 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6256 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6257 (_mm_maskz_sqrt_round_ss): New intrinsics.
6258 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6259 (__builtin_ia32_sqrtsd_mask_round)
6260 (__builtin_ia32_sqrtss_mask_round): New builtins.
6261 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6262 (__builtin_ia32_sqrtss_round): Remove.
6263 (__builtin_ia32_sqrtsd_mask_round)
6264 (__builtin_ia32_sqrtss_mask_round): New builtins.
6265 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6266 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6267 ((match_operand:VF_128 1 "vector_operand"
6268 "xBm,<round_constraint>")): Changed to ...
6269 ((match_operand:VF_128 1 "vector_operand"
6270 "xBm,<round_scalar_constraint>")): ... this.
6271 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6272 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6273 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6274 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6275 %<iptr>1<round_scalar_mask_op3>}): ... this.
6276 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6277 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6279 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6282 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6283 Cast vec_cmpeq result to correct type.
6284 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6285 Cast vec_cmpgt result to correct type.
6287 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6289 * final.c (final_scan_insn_1): Renamed from...
6290 (final_scan_insn): ... this. New wrapper, to recover
6291 seen from the outermost call in recursive ones.
6292 * config/sparc/sparc.c (output_return): Drop seen from call.
6293 (output_sibcall): Likewise.
6294 * config/visium/visium.c (output_branch): Likewise.
6296 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6298 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6301 2018-02-10 Alan Modra <amodra@gmail.com>
6304 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6305 Specify LR as an input.
6307 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6310 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6311 remove_member_access_dummy_vars): New functions.
6312 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6313 lower_omp_1, execute_lower_omp): Use them.
6315 PR rtl-optimization/84308
6316 * shrink-wrap.c (spread_components): Release todo vector.
6318 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6320 PR rtl-optimization/57193
6321 * ira-color.c (struct allocno_color_data): Add member
6322 conflict_allocno_hard_prefs.
6323 (update_conflict_allocno_hard_prefs): New.
6324 (bucket_allocno_compare_func): Add a preference based on
6325 conflict_allocno_hard_prefs.
6326 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6327 (color_allocnos): Remove a dead code. Initiate
6328 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6330 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6333 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6334 constraint from =wa to wa. Avoid a subreg on the output operand,
6335 instead use a pseudo and subreg it in a move.
6336 (p9_xxbrd_<mode>): Changed to ...
6337 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6338 (p9_xxbrd_v2df): New expander.
6339 (p9_xxbrw_<mode>): Changed to ...
6340 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6341 (p9_xxbrw_v4sf): New expander.
6343 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6345 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6347 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6350 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6351 multiply in 32-bit mode.
6352 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6353 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6356 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6358 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6359 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6360 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6361 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6363 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6366 * dwarf2out.c (is_trivial_indirect_ref): New function.
6367 (dwarf2out_late_global_decl): Do not generate a location
6368 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6369 and that are not defined in the current unit.
6371 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6373 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6374 instead of a libcall for UNORDERED.
6376 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6379 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6380 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6382 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6385 * config/s390/s390.c (s390_set_current_function): Invoke
6386 s390_indirect_branch_settings also if fndecl didn't change.
6388 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6390 * config/rs6000/rs6000.md (blockage): Set length to zero.
6392 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6394 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6396 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6399 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6400 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6404 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6405 PARALLEL incoming that failed vt_get_decl_and_offset check.
6408 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6409 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6411 (get_variable_section): For decls in named .bss* sections pass true as
6412 second argument to bss_initializer_p.
6414 2018-02-09 Marek Polacek <polacek@redhat.com>
6415 Jakub Jelinek <jakub@redhat.com>
6418 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6419 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6420 Sync some changes from cxx_fold_indirect_ref.
6422 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6424 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6426 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6427 (BLOCK_INLINE_ENTRY_LABEL): New.
6428 (dwarf2out_var_location): Disregard inline entry markers.
6429 (inline_entry_data): New struct.
6430 (inline_entry_data_hasher): New hashtable type.
6431 (inline_entry_data_hasher::hash): New.
6432 (inline_entry_data_hasher::equal): New.
6433 (inline_entry_data_table): New variable.
6434 (add_high_low_attributes): Add DW_AT_entry_pc and
6435 DW_AT_GNU_entry_view attributes if a pending entry is found
6436 in inline_entry_data_table. Add old entry_pc attribute only
6437 if debug nonbinding markers are disabled.
6438 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6439 markers are enabled.
6440 (block_within_block_p, dwarf2out_inline_entry): New.
6441 (dwarf2out_finish): Check that no entries remained in
6442 inline_entry_data_table.
6443 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6444 (final_scan_insn, notice_source_line): Likewise.
6445 (rest_of_clean_state): Skip inline entry markers.
6446 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6448 * gimple.c (gimple_build_debug_inline_entry): New.
6449 * gimple.h (enum gimple_debug_subcode): Add
6450 GIMPLE_DEBUG_INLINE_ENTRY.
6451 (gimple_build_debug_inline_entry): Declare.
6452 (gimple_debug_inline_entry_p): New.
6453 (gimple_debug_nonbind_marker_p): Adjust.
6454 * insn-notes.def (INLINE_ENTRY): New.
6455 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6456 inline entry marker notes.
6457 (print_insn): Likewise.
6458 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6459 (INSN_DEBUG_MARKER_KIND): Likewise.
6460 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6461 * tree-inline.c (expand_call_inline): Build and insert
6462 debug_inline_entry stmt.
6463 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6464 inline entry blocks early, if nonbind markers are enabled.
6465 (dump_scope_block): Dump fragment info.
6466 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6467 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6468 (gimple_build_debug_inline_entry): New.
6469 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6470 Enable/disable inline entry points too.
6471 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6472 (DEBUG_INSN): Describe inline entry markers.
6474 * common.opt (gvariable-location-views): New.
6475 (gvariable-location-views=incompat5): New.
6476 * config.in: Rebuilt.
6477 * configure: Rebuilt.
6478 * configure.ac: Test assembler for view support.
6479 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6480 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6481 * dwarf2out.c (var_loc_view): New typedef.
6482 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6483 (dwarf2out_locviews_in_attribute): New.
6484 (dwarf2out_locviews_in_loclist): New.
6485 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6486 (enum dw_line_info_opcode): Add LI_adv_address.
6487 (struct dw_line_info_table): Add view.
6488 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6489 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6490 (zero_view_p): New variable.
6491 (ZERO_VIEW_P): New macro.
6492 (output_asm_line_debug_info): New.
6493 (struct var_loc_node): Add view.
6494 (add_AT_view_list, AT_loc_list): New.
6495 (add_var_loc_to_decl): Add view param. Test it against last.
6496 (new_loc_list): Add view params. Record them.
6497 (AT_loc_list_ptr): Handle loc and view lists.
6498 (view_list_to_loc_list_val_node): New.
6499 (print_dw_val): Handle dw_val_class_view_list.
6500 (size_of_die): Likewise.
6501 (value_format): Likewise.
6502 (loc_list_has_views): New.
6503 (gen_llsym): Set vl_symbol too.
6504 (maybe_gen_llsym, skip_loc_list_entry): New.
6505 (dwarf2out_maybe_output_loclist_view_pair): New.
6506 (output_loc_list): Output view list or entries too.
6507 (output_view_list_offset): New.
6508 (output_die): Handle dw_val_class_view_list.
6509 (output_dwarf_version): New.
6510 (output_compilation_unit_header): Use it.
6511 (output_skeleton_debug_sections): Likewise.
6512 (output_rnglists, output_line_info): Likewise.
6513 (output_pubnames, output_aranges): Update version comments.
6514 (output_one_line_info_table): Output view numbers in asm comments.
6515 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6516 Call maybe_gen_llsym.
6517 (loc_list_from_tree_1): Adjust.
6518 (add_AT_location_description): Create view list attribute if
6519 needed, check it's absent otherwise.
6520 (convert_cfa_to_fb_loc_list): Adjust.
6521 (maybe_emit_file): Call output_asm_line_debug_info for test.
6522 (dwarf2out_var_location): Reset views as needed. Precompute
6523 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6524 attribute. Set view.
6525 (new_line_info_table): Reset next view.
6526 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6527 (dwarf2out_source_line): Likewise. Output view resets and labels to
6528 the assembler, or select appropriate line info opcodes.
6529 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6530 (optimize_string_length): Catch it. Adjust.
6531 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6532 dw_val_class_view_list, and remove it if no longer needed.
6533 (hash_loc_list): Hash view numbers.
6534 (loc_list_hasher::equal): Compare them.
6535 (optimize_location_lists): Check whether a view list symbol is
6536 needed, and whether the locview attribute is present, and
6537 whether they match. Remove the locview attribute if no longer
6539 (index_location_lists): Call skip_loc_list_entry for test.
6540 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6541 Use output_dwarf_version.
6542 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6543 (struct dw_val_node): Add val_view_list.
6544 * final.c (SEEN_NEXT_VIEW): New.
6545 (set_next_view_needed): New.
6546 (clear_next_view_needed): New.
6547 (maybe_output_next_view): New.
6548 (final_start_function): Rename to...
6549 (final_start_function_1): ... this. Take pointer to FIRST,
6550 add SEEN parameter. Emit param bindings in the initial view.
6551 (final_start_function): Reintroduce SEEN-less interface.
6552 (final): Rename to...
6553 (final_1): ... this. Take SEEN parameter. Output final pending
6554 next view at the end.
6555 (final): Reintroduce seen-less interface.
6556 (final_scan_insn): Output pending next view before switching
6557 sections or ending a block. Mark the next view as needed when
6558 outputting variable locations. Notify debug backend of section
6559 changes, and of location view changes.
6560 (rest_of_handle_final): Adjust.
6561 * toplev.c (process_options): Autodetect value for debug variable
6562 location views option. Warn on incompat5 without -gdwarf-5.
6563 * doc/invoke.texi (gvariable-location-views): New.
6564 (gvariable-location-views=incompat5): New.
6565 (gno-variable-location-views): New.
6567 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6569 PR tree-optimization/84136
6570 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6571 that the result of find_edge is non-NULL.
6573 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6576 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6577 storing integer register in SImode. Fix cost of 256 and 512
6578 byte aligned SSE register store.
6580 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6582 * config/i386/i386.c (ix86_multiplication_cost): Fix
6583 multiplication cost for TARGET_AVX512DQ.
6585 2018-02-08 Marek Polacek <polacek@redhat.com>
6587 PR tree-optimization/84238
6588 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6591 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6593 PR tree-optimization/84265
6594 * tree-vect-stmts.c (vectorizable_store): Don't treat
6595 VMAT_CONTIGUOUS accesses as grouped.
6596 (vectorizable_load): Likewise.
6598 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6600 PR tree-optimization/81635
6601 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6602 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6603 (test_round_for_mask): New functions.
6604 (wide_int_cc_tests): Call test_round_for_mask.
6605 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6606 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6607 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6608 range returned by get_range_info.
6610 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6613 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6614 * symtab.c: Include builtins.h
6615 (symtab_node::output_to_lto_symbol_table_p): Move here
6616 from lto-streamer-out.c:output_symbol_p.
6617 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6618 (output_symbol_p): Move all logic to symtab.c
6619 (produce_symtab): Update.
6621 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6623 * config/s390/s390-opts.h (enum indirect_branch): Define.
6624 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6625 (s390_indirect_branch_via_thunk)
6626 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6627 (enum s390_indirect_branch_type): Define.
6628 * config/s390/s390.c (struct s390_frame_layout, struct
6629 machine_function): Remove.
6630 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6631 (indirect_branch_table_label_no, indirect_branch_table_name):
6633 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6634 (enum s390_indirect_branch_option): Define.
6635 (s390_return_addr_from_memory): New function.
6636 (s390_handle_string_attribute): New function.
6637 (s390_attribute_table): Add new attribute handler.
6638 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6639 (s390_indirect_branch_via_thunk): New function.
6640 (s390_indirect_branch_via_inline_thunk): New function.
6641 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6642 sibling call optimization for non z10 compiles.
6643 (s390_emit_call): Force indirect branch target to be a single
6644 register. Add r1 clobber for non-z10 compiles.
6645 (s390_emit_epilogue): Emit return jump via return_use expander.
6646 (s390_reorg): Handle JUMP_INSNs as execute targets.
6647 (s390_option_override_internal): Perform validity checks for the
6648 new command line options.
6649 (s390_indirect_branch_attrvalue): New function.
6650 (s390_indirect_branch_settings): New function.
6651 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6652 (s390_output_indirect_thunk_function): New function.
6653 (s390_code_end): Implement target hook.
6654 (s390_case_values_threshold): Implement target hook.
6655 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6657 * config/s390/s390.h (struct s390_frame_layout)
6658 (struct machine_function): Move here from s390.c.
6659 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6660 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6661 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6662 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6663 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6664 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6665 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6666 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6667 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6668 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6669 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6670 (mnemonic attribute): Add values which aren't recognized
6672 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6673 pattern for branch conversion. Fix mnemonic attribute.
6674 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6675 indirect branch via thunk if requested.
6676 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6677 ("*indirect_jump"): Disable for branch conversion using out of
6679 ("indirect_jump_via_thunk<mode>_z10")
6680 ("indirect_jump_via_thunk<mode>")
6681 ("indirect_jump_via_inlinethunk<mode>_z10")
6682 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6683 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6684 ("casesi_jump_via_inlinethunk<mode>_z10")
6685 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6686 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6687 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6688 ("*indirect2_jump"): Disable for branch conversion.
6689 ("casesi_jump"): Turn into expander and expand patterns for branch
6691 ("return_use"): New expander.
6692 ("*return"): Emit return via thunk and rename it to ...
6693 ("*return<mode>"): ... this one.
6694 * config/s390/s390.opt: Add new options and and enum for the
6697 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6699 * lra-constraints.c (match_reload): Unconditionally use
6700 gen_lowpart_SUBREG, rather than selecting between that
6701 and equivalent gen_rtx_SUBREG code.
6703 2018-02-08 Richard Biener <rguenther@suse.de>
6705 PR tree-optimization/84233
6706 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6707 changed flag instead of boguously re-using phi_inserted.
6709 2018-02-08 Martin Jambor <mjambor@suse.cz>
6711 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6712 static local variables.
6714 2018-02-08 Richard Biener <rguenther@suse.de>
6716 PR tree-optimization/84278
6717 * tree-vect-stmts.c (vectorizable_store): When looking for
6718 smaller vector types to perform grouped strided loads/stores
6719 make sure the mode is supported by the target.
6720 (vectorizable_load): Likewise.
6722 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6724 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6725 Increase LDP/STP opportunities by adding adjacent callee-saves.
6727 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6729 PR rtl-optimization/84068
6730 PR rtl-optimization/83459
6731 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6733 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6735 PR tree-optimization/84224
6736 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6737 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6740 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6743 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6744 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6745 count, remove one USE.
6747 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6749 * doc/install.texi (Configuration): Document the
6750 --with-long-double-format={ibm,ieee} PowerPC configuration
6754 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6755 Convert from define_expand to be define_insn_and_split. Rework
6756 float/double/_Float128 conversions to QI/HI/SImode to work with
6757 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6758 conversions to QI/HImode types did a store and then a load to
6759 truncate the value. For conversions to VSX registers, don't split
6760 the insn, instead emit the code directly. Use the code iterator
6761 any_fix to combine signed and unsigned conversions.
6762 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6763 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6764 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6765 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6766 (fix_<mode>di2_hw): Likewise.
6767 (fixuns_<mode>di2_hw): Likewise.
6768 (fix_<mode>si2_hw): Likewise.
6769 (fixuns_<mode>si2_hw): Likewise.
6770 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6771 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6772 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6773 fix<uns>_trunc<SFDF:mode>si2_p8.
6774 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6776 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6777 (fix<uns>_<mode>_mem): Likewise.
6778 (fctiw<u>z_<mode>_mem): Likewise.
6779 (fix<uns>_<mode>_mem): Likewise.
6780 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6781 the register allocator from doing a direct move to the GPRs to do
6782 a store, and instead use the ISA 3.0 store byte/half-word from
6783 vector register instruction. For IEEE 128-bit floating point,
6784 also optimize stores of 32-bit ints.
6785 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6787 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6789 * genextract.c (push_pathstr_operand): New function to support
6791 (walk_rtx): Call push_pathstr_operand.
6792 (print_path): Support [a-zA-Z].
6794 2018-02-07 Richard Biener <rguenther@suse.de>
6796 PR tree-optimization/84037
6797 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6798 (cse_and_gimplify_to_preheader): Declare.
6799 (vect_get_place_in_interleaving_chain): Likewise.
6800 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6802 (_loop_vec_info::~_loop_vec_info): Delete it.
6803 (cse_and_gimplify_to_preheader): New function.
6804 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6805 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6806 (vectorizable_load): Likewise. For grouped stores always base
6807 the IV on the first element.
6808 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6809 condition before gimplifying.
6811 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6813 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6814 *DIV_EXPR and *MOD_EXPR.
6816 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6819 * config/i386/i386.c (ix86_option_override_internal): Mask out
6820 the CF_SET bit when checking -fcf-protection.
6822 2018-02-07 Tom de Vries <tom@codesourcery.com>
6825 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6828 2018-02-07 Richard Biener <rguenther@suse.de>
6830 PR tree-optimization/84204
6831 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6834 PR tree-optimization/84205
6835 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6836 special-case isl_ast_op_zdiv_r.
6838 PR tree-optimization/84223
6839 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6840 Only add conditions from within the region.
6841 (gather_bbs::after_dom_children): Adjust.
6843 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6846 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6847 * config/avr/avr.md: Only post-reload split REG-REG moves if
6848 either register is GENERAL_REG_P.
6850 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6852 PR tree-optimization/84235
6853 * tree-ssa-scopedtables.c
6854 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6855 if the subtraction is performed in floating point type where NaNs are
6856 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6857 build 1. Formatting fix.
6859 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6862 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6863 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6864 and skip it regardless of bb boundaries. Use CALL_P macro,
6865 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6866 unnecessarily, formatting fix.
6868 2018-02-06 Michael Collison <michael.collison@arm.com>
6870 * config/arm/thumb2.md:
6871 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6872 (*thumb_mov_notscc): Ditto.
6874 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6877 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6878 unsigned_fix, not "s".
6880 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6882 * configure.ac (gcc_fn_eh_frame_ro): New function.
6883 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6884 correct .eh_frame permissions.
6885 * configure: Regenerate.
6887 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6889 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6892 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6894 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6895 Display warning message for -mno-speculate-indirect-jumps.
6897 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6899 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6901 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6903 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6905 PR tree-optimization/84225
6906 * tree-eh.c (find_trapping_overflow): Only call
6907 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6909 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6912 * config/i386/i386.c: Reimplement the check of possible options
6913 -mibt/-mshstk conbination. Change error messages.
6914 * doc/invoke.texi: Fix a typo: remove extra '='.
6916 2018-02-06 Marek Polacek <polacek@redhat.com>
6918 PR tree-optimization/84228
6919 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6921 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6924 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6925 emitted arch directives.
6926 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6927 __ARM_FEATURE_COPROC before changing architectures.
6929 2018-02-06 Richard Biener <rguenther@suse.de>
6931 * config/i386/i386.c (print_reg): Fix typo.
6932 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6934 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6936 * configure: Regenerate.
6938 2018-02-05 Martin Sebor <msebor@redhat.com>
6940 PR tree-optimization/83369
6941 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6944 2018-02-05 Martin Liska <mliska@suse.cz>
6946 * doc/invoke.texi: Cherry-pick upstream r323995.
6948 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6950 * ira.c (ira_init_register_move_cost): Adjust comment.
6952 2018-02-05 Martin Liska <mliska@suse.cz>
6954 PR gcov-profile/84137
6955 * doc/gcov.texi: Fix typo in documentation.
6957 2018-02-05 Martin Liska <mliska@suse.cz>
6959 PR gcov-profile/83879
6960 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6961 using dlopen functionality.
6963 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6965 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6966 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6967 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6968 _mm_maskz_range_round_ss): New intrinsics.
6969 (__builtin_ia32_rangesd128_round)
6970 (__builtin_ia32_rangess128_round): Remove.
6971 (__builtin_ia32_rangesd128_mask_round,
6972 __builtin_ia32_rangess128_mask_round): New builtins.
6973 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6974 __builtin_ia32_rangess128_round): Remove.
6975 (__builtin_ia32_rangesd128_mask_round,
6976 __builtin_ia32_rangess128_mask_round): New builtins.
6977 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6978 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6979 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6980 "<round_saeonly_constraint>")): Changed to ...
6981 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6982 "<round_saeonly_scalar_constraint>")): ... this.
6983 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6984 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6985 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6986 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6987 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
6989 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6991 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6993 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6994 Remove all values except native, 8540 and 8548.
6996 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6998 * config/i386/i386.c (ix86_output_function_return): Pass
6999 INVALID_REGNUM, instead of -1, as invalid register number to
7000 indirect_thunk_name and output_indirect_thunk.
7002 2018-02-02 Julia Koval <julia.koval@intel.com>
7004 * config.gcc: Add -march=icelake.
7005 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7006 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7007 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7008 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7009 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7010 (processor_target_table): Add icelake.
7011 (ix86_option_override_internal): Handle new PTAs.
7012 (get_builtin_code_for_version): Handle icelake.
7013 (M_INTEL_COREI7_ICELAKE): New.
7014 (fold_builtin_cpu): Handle icelake.
7015 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7016 * doc/invoke.texi: Add -march=icelake.
7018 2018-02-02 Julia Koval <julia.koval@intel.com>
7020 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7021 to wide_int_bitmask.
7022 * wide-int-bitmask.h: New.
7024 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7027 * config/i386/i386.md: Replace Pmode with word_mode in
7028 builtin_setjmp_setup and builtin_longjmp to support x32.
7030 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7034 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7036 (rs6000_supported_cpu_names): New static variable.
7037 (linux_cpu_translation_table): Likewise.
7038 (elf_platform) <cpu>: Define new static variable and use it.
7039 Translate kernel AT_PLATFORM name to canonical name if needed.
7040 Error if platform name is unknown.
7042 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7045 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7047 2018-02-01 Jeff Law <law@redhat.com>
7050 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7051 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7052 the scratch if RELEASE_VIA_POP is false.
7053 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7054 If we have to save a temporary register, decrement SIZE appropriately.
7055 Pass new arguments to release_scratch_register_on_entry.
7056 (ix86_adjust_stack_and_probe): Likewise.
7057 (ix86_emit_probe_stack_range): Pass new arguments to
7058 release_scratch_register_on_entry.
7060 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7062 PR rtl-optimization/84157
7063 * combine.c (change_zero_ext): Use REG_P predicate in
7064 front of HARD_REGISTER_P predicate.
7066 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7068 * config/avr/avr.c (avr_option_override): Move disabling of
7069 -fdelete-null-pointer-checks to...
7070 * common/config/avr/avr-common.c (avr_option_optimization_table):
7073 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7075 PR tree-optimization/81635
7076 * tree-data-ref.c (split_constant_offset_1): For types that
7077 wrap on overflow, try to use range info to prove that wrapping
7080 2018-02-01 Renlin Li <renlin.li@arm.com>
7083 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7085 (aarch64_register_move_cost): Likewise.
7086 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7088 (REG_CLASS_NAMES): Likewise.
7089 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7090 TAILCALL_ADDR_REGS. Remove IP registers.
7091 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7093 2018-02-01 Richard Biener <rguenther@suse.de>
7095 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7096 for specifying RPO order and allow NULL for that.
7097 * domwalk.c (dom_walker::dom_walker): Likewise.
7098 (dom_walker::walk): Handle NULL RPO order.
7099 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7101 (rewrite_update_dom_walker): Likewise.
7102 (mark_def_dom_walker): Likewise.
7104 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7106 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7107 (aarch64_maybe_expand_sve_subreg_move): Declare.
7108 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7109 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7111 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7112 that are semantically a reverse operation.
7113 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7114 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7115 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7117 (aarch64_can_change_mode_class): For big-endian, forbid changes
7118 between two SVE modes if they have different element sizes.
7120 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7122 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7123 the TImode handling for big-endian targets.
7125 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7127 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7128 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7130 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7131 Remove BSWAP handing for big-endian targets and use the form of
7132 LD1RQ appropariate for the mode.
7134 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7136 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7137 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7140 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7143 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7144 check for operands that need to go through aarch64_sve_reload_be.
7146 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7148 PR tree-optimization/81661
7149 PR tree-optimization/84117
7150 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7151 * tree-eh.c: Include gimplify.h.
7152 (find_trapping_overflow, replace_trapping_overflow,
7153 rewrite_to_non_trapping_overflow): New functions.
7154 * tree-vect-loop.c: Include tree-eh.h.
7155 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7156 * tree-data-ref.c: Include tree-eh.h.
7157 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7159 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7161 PR rtl-optimization/84123
7162 * combine.c (change_zero_ext): Check if hard register satisfies
7163 can_change_dest_mode before calling gen_lowpart_SUBREG.
7165 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7168 * ira.c (ira_init_register_move_cost): Remove assert.
7170 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7172 PR rtl-optimization/84071
7173 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7174 * doc/tm.texi: Regenerate.
7176 2018-01-31 Richard Biener <rguenther@suse.de>
7178 PR tree-optimization/84132
7179 * tree-data-ref.c (analyze_miv_subscript): Properly
7180 check whether evolution_function_is_affine_multivariate_p
7181 before calling gcd_of_steps_may_divide_p.
7183 2018-01-31 Julia Koval <julia.koval@intel.com>
7186 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7187 * config/i386/i386.md (rdpid_rex64) New.
7188 (rdpid): Make 32bit only.
7190 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7193 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7194 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7196 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7199 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7201 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7203 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7205 PR rtl-optimization/84071
7206 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7207 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7209 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7211 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7212 (arc_attribute_table): Add 'aux' attribute.
7213 (arc_in_small_data_p): Consider aux like variables.
7214 (arc_is_aux_reg_p): New function.
7215 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7216 (arc_get_aux_arg): New function.
7217 (prepare_move_operands): Handle aux-register access.
7218 (arc_handle_aux_attribute): New function.
7219 * doc/extend.texi (ARC Variable attributes): Add subsection.
7221 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7223 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7224 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7225 (arc_attribute_table): Add 'uncached' attribute.
7226 (arc_print_operand): Print '.di' flag for uncached memory
7228 (arc_in_small_data_p): Do not consider for small data the uncached
7230 (arc_is_uncached_mem_p): New function.
7231 * config/arc/predicates.md (compact_store_memory_operand): Check
7232 for uncached memory accesses.
7233 (nonvol_nonimm_operand): Likewise.
7234 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7236 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7239 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7240 falign-loops=): Add Optimization flag.
7242 2018-01-30 Jeff Law <law@redhat.com>
7245 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7246 INT_REGISTERS_SAVED. Check it prior to calling
7247 get_scratch_register_on_entry.
7248 (ix86_adjust_stack_and_probe): Similarly.
7249 (ix86_emit_probe_stack_range): Similarly.
7250 (ix86_expand_prologue): Corresponding changes.
7252 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7255 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7256 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7258 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7261 * lra-constraints.c (curr_insn_transform): Process AND in the
7264 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7266 PR rtl-optimization/83986
7267 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7268 dependence against last_pending_memory_flush in addition to
7271 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7273 PR tree-optimization/81611
7274 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7277 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7280 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7283 2018-01-30 Richard Biener <rguenther@suse.de>
7284 Jakub Jelinek <jakub@redhat.com>
7286 PR tree-optimization/84111
7287 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7288 inner loops added during recursion, as they don't have up-to-date
7291 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7294 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7295 (can_inline_edge_by_limits_p): ... here.
7296 (can_early_inline_edge_p, check_callers,
7297 update_caller_keys, update_callee_keys, recursive_inlining,
7298 add_new_edges_to_heap, speculation_useful_p,
7299 inline_small_functions,
7300 inline_small_functions, flatten_function,
7301 inline_to_all_callers_1): Update.
7303 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7305 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7308 2018-01-30 Richard Biener <rguenther@suse.de>
7310 PR tree-optimization/83008
7311 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7312 invariant and constant vector uses in stmts when they need
7315 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7318 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7319 * configure: Regenerate.
7321 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7323 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7325 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7326 Use gen_rtx_REG rather than gen_lowpart.
7328 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7330 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7331 rather than 0 when creating partial subregs.
7333 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7335 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7338 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7341 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7342 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7343 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7344 flags. This restores the settings used before the 2017-07-24.
7345 Turning off pre increment/decrement/modify allows IVOPTS to
7346 optimize DF/SF loops where the index is an int.
7348 2018-01-29 Richard Biener <rguenther@suse.de>
7349 Kelvin Nilsen <kelvin@gcc.gnu.org>
7352 * tree-vect-stmts.c (vectorizable_call): Don't call
7353 targetm.vectorize_builtin_md_vectorized_function if callee is
7356 2018-01-22 Carl Love <cel@us.ibm.com>
7358 * doc/extend.tex: Fix typo in second arg in
7359 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7361 2018-01-29 Richard Biener <rguenther@suse.de>
7363 PR tree-optimization/84086
7364 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7365 (flush_ssaname_freelist): When SSA names were released reset
7366 the SCEV hash table.
7368 2018-01-29 Richard Biener <rguenther@suse.de>
7370 PR tree-optimization/84057
7371 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7372 removed paths when removing edges.
7374 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7376 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7377 -mfunction-return=@var{choice}.
7379 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7382 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7384 (layout::print_source_line): Likewise.
7385 (test_get_line_width_without_trailing_whitespace): Add test cases.
7387 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7390 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7393 2018-01-26 Jim Wilson <jimw@sifive.com>
7395 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7397 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7400 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7402 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7403 and CMP + SUB-immediate -> SUBS.
7405 2018-01-26 Martin Sebor <msebor@redhat.com>
7407 PR tree-optimization/83896
7408 * tree-ssa-strlen.c (get_string_len): Rename...
7409 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7410 Avoid assuming length is constant.
7411 (handle_char_store): Use HOST_WIDE_INT for string length.
7413 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7416 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7417 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7419 2018-01-26 Richard Biener <rguenther@suse.de>
7421 PR rtl-optimization/84003
7422 * dse.c (record_store): Only record redundant stores when
7423 the earlier store aliases at least all accesses the later one does.
7425 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7427 PR rtl-optimization/83985
7428 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7429 REG_CFA_RESTORE insns.
7430 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7431 REG_CFA_RESTORE insns here.
7434 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7435 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7437 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7439 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7440 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7441 (arc_init): Likewise.
7442 (arc_override_options): Likewise.
7443 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7445 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7447 * config/arc/arc.h (TARGET_DBNZ): Define.
7448 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7449 properly set the tune attribute.
7450 (dbnz): Use TARGET_DBNZ guard.
7451 * config/arc/arc.opt (mtune): Add core3 option.
7453 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7455 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7456 recognize new pic like addresses.
7457 (arc_delegitimize_address): Clean up.
7459 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7461 * config/arc/arc-arches.def: Option mrf16 valid for all
7463 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7464 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7465 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7466 * config/arc/arc-tables.opt: Regenerate.
7467 * config/arc/arc.c (arc_conditional_register_usage): Handle
7468 reduced register file case.
7469 (arc_file_start): Set must have build attributes.
7470 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7472 * config/arc/arc.opt (mrf16): Add new option.
7473 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7474 * config/arc/genmultilib.awk: Handle new mrf16 option.
7475 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7476 * config/arc/t-multilib: Regenerate.
7477 * doc/invoke.texi (ARC Options): Document mrf16 option.
7479 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7481 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7482 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7483 (arc_attribute_table): Add 'secure_call' attribute.
7484 (arc_print_operand): Print secure call operand.
7485 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7487 (arc_is_secure_call_p): New function. * config/arc/arc.md
7488 (call_i): Add support for sjli instruction.
7489 (call_value_i): Likewise.
7490 * config/arc/constraints.md (Csc): New constraint.
7492 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7493 John Eric Martin <John.Martin@emmicro-us.com>
7495 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7496 * config/arc/arc.c (_arc_jli_section): New struct.
7497 (arc_jli_section): New type.
7498 (rc_jli_sections): New static variable.
7499 (arc_handle_jli_attribute): New function.
7500 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7501 (arc_file_end): New function.
7502 (TARGET_ASM_FILE_END): Define.
7503 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7504 (arc_add_jli_section): New function.
7505 (jli_call_scan): Likewise.
7506 (arc_reorg): Call jli_call_scan.
7507 (arc_output_addsi): Remove 'S' from printing asm operand.
7508 (arc_is_jli_call_p): New function.
7509 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7511 (movhi_insn): Likewise.
7512 (movsi_insn): Likewise.
7513 (movsi_set_cc_insn): Likewise.
7514 (loadqi_update): Likewise.
7515 (load_zeroextendqisi_update): Likewise.
7516 (load_signextendqisi_update): Likewise.
7517 (loadhi_update): Likewise.
7518 (load_zeroextendhisi_update): Likewise.
7519 (load_signextendhisi_update): Likewise.
7520 (loadsi_update): Likewise.
7521 (loadsf_update): Likewise.
7522 (movsicc_insn): Likewise.
7523 (bset_insn): Likewise.
7524 (bxor_insn): Likewise.
7525 (bclr_insn): Likewise.
7526 (bmsk_insn): Likewise.
7527 (bicsi3_insn): Likewise.
7528 (cmpsi_cc_c_insn): Likewise.
7529 (movsi_ne): Likewise.
7530 (movsi_cond_exec): Likewise.
7531 (clrsbsi2): Likewise.
7542 (call_i): Remove 'S' asm letter, add jli instruction.
7543 (call_value_i): Likewise.
7544 * config/arc/arc.op (mjli-always): New option.
7545 * config/arc/constraints.md (Cji): New constraint.
7546 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7548 (subsf3_fpx): Likewise.
7549 (mulsf3_fpx): Likewise.
7550 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7552 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7554 * doc/invoke.texi (ARC): Document mjli-always option.
7556 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7558 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7559 avoid addition with 0 and use incw and decw where possible.
7561 2018-01-26 Richard Biener <rguenther@suse.de>
7563 PR tree-optimization/81082
7564 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7565 association if it requires casting to unsigned.
7566 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7567 from fold_plusminus_mult_expr to catch important cases late when
7568 range info is available.
7570 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7572 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7573 * configure.ac (hidden_linkonce): New test.
7574 * configure: Regenerate.
7575 * config.in: Regenerate.
7577 2018-01-26 Julia Koval <julia.koval@intel.com>
7579 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7580 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7581 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7582 _mm_mask_bitshuffle_epi64_mask): Fix type.
7583 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7584 USI_FTYPE_V4DI_V4DI_USI): Remove.
7585 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7586 __builtin_ia32_vpshufbitqmb256_mask,
7587 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7588 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7589 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7591 2018-01-26 Alan Modra <amodra@gmail.com>
7594 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7595 UNSPEC_VBPERMQ. Sort other unspecs.
7597 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7599 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7601 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7604 * predict.c (drop_profile): Do not push/pop cfun; update also
7606 (handle_missing_profiles): Fix logic looking for zero profiles.
7608 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7611 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7612 on functions with #pragma omp declare simd or functions with simd
7614 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7615 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7616 Remove trailing \n from warning_at calls.
7618 2018-01-25 Tom de Vries <tom@codesourcery.com>
7621 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7622 for neutered workers.
7624 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7627 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7628 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7630 2018-01-24 Jeff Law <law@redhat.com>
7633 * i386.c (get_probe_interval): Move to earlier point.
7634 (ix86_compute_frame_layout): If -fstack-clash-protection and
7635 the frame is larger than the probe interval, then use pushes
7636 to save registers rather than reg->mem moves.
7637 (ix86_expand_prologue): Remove conditional for int_registers_saved
7640 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7643 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7644 min/max for never referenced object.
7646 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7649 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7651 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7652 attributes from DECL_ATTRIBUTES (decl) without affecting
7653 DECL_ATTRIBUTES (current_function_decl).
7654 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7655 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7657 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7659 PR tree-optimization/83979
7660 * fold-const.c (fold_comparison): Use constant_boolean_node
7661 instead of boolean_{true,false}_node.
7663 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7665 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7668 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7670 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7671 Simplify the clause that sets the length attribute.
7672 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7673 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7674 clause that sets the length attribute.
7675 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7677 2018-01-24 Tom de Vries <tom@codesourcery.com>
7680 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7681 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7682 Add strict parameter.
7683 (prevent_branch_around_nothing): Insert dummy insn between branch to
7684 label and label with no ptx insn inbetween.
7685 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7687 2018-01-24 Tom de Vries <tom@codesourcery.com>
7690 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7691 for neutered threads in warp.
7692 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7694 2018-01-24 Richard Biener <rguenther@suse.de>
7696 PR tree-optimization/83176
7697 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7700 2018-01-24 Richard Biener <rguenther@suse.de>
7702 PR tree-optimization/82819
7703 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7704 code generating pluses that are no-ops in the target precision.
7706 2018-01-24 Richard Biener <rguenther@suse.de>
7709 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7711 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7713 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7714 to merge probabilities.
7715 * predict.c (probably_never_executed): Also mark as cold functions
7716 with global 0 profile and guessed local profile.
7717 * profile-count.c (profile_probability::combine_with_count): New
7719 * profile-count.h (profile_probability::operator*,
7720 profile_probability::operator*=, profile_probability::operator/,
7721 profile_probability::operator/=): Reduce precision to adjusted
7722 and set value to guessed on contradictory divisions.
7723 (profile_probability::combine_with_freq): Remove.
7724 (profile_probability::combine_wiht_count): Declare.
7725 (profile_count::force_nonzero):: Set to adjusted.
7726 (profile_count::probability_in):: Set quality to adjusted.
7727 * tree-ssa-tail-merge.c (replace_block_by): Use
7730 2018-01-23 Andrew Waterman <andrew@sifive.com>
7731 Jim Wilson <jimw@sifive.com>
7733 * config/riscv/riscv.c (riscv_stack_boundary): New.
7734 (riscv_option_override): Set riscv_stack_boundary. Handle
7735 riscv_preferred_stack_boundary_arg.
7736 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7737 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7738 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7739 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7740 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7741 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7743 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7746 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7747 of struct ix86_frame.
7748 (ix86_expand_epilogue): Likewise. Add a local variable for
7749 the reg_save_offset field in struct ix86_frame.
7751 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7753 PR tree-optimization/82604
7754 * tree-loop-distribution.c (enum partition_kind): New enum item
7755 PKIND_PARTIAL_MEMSET.
7756 (partition_builtin_p): Support above new enum item.
7757 (generate_code_for_partition): Ditto.
7758 (compute_access_range): Differentiate cases that equality can be
7759 proven at all loops, the innermost loops or no loops.
7760 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7761 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7762 (finalize_partitions, distribute_loop): Don't fuse partition of
7763 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7764 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7767 2018-01-23 Martin Liska <mliska@suse.cz>
7769 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7770 order to ignore the predictor.
7771 (PRED_POLYMORPHIC_CALL): Likewise.
7772 (PRED_RECURSIVE_CALL): Likewise.
7774 2018-01-23 Martin Liska <mliska@suse.cz>
7776 * tree-profile.c (tree_profiling): Print function header to
7777 aware reader which function we are working on.
7778 * value-prof.c (gimple_find_values_to_profile): Do not print
7779 not interesting value histograms.
7781 2018-01-23 Martin Liska <mliska@suse.cz>
7783 * profile-count.h (enum profile_quality): Add
7784 profile_uninitialized as the first value. Do not number values
7785 as they are zero based.
7786 (profile_count::verify): Update sanity check.
7787 (profile_probability::verify): Likewise.
7789 2018-01-23 Nathan Sidwell <nathan@acm.org>
7791 * doc/invoke.texi (ffor-scope): Deprecate.
7793 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7795 PR tree-optimization/83510
7796 * domwalk.c (set_all_edges_as_executable): New function.
7797 (dom_walker::dom_walker): Convert bool param
7798 "skip_unreachable_blocks" to enum reachability. Move setup of
7799 edge flags to set_all_edges_as_executable and only do it when
7800 reachability is REACHABLE_BLOCKS.
7801 * domwalk.h (enum dom_walker::reachability): New enum.
7802 (dom_walker::dom_walker): Convert bool param
7803 "skip_unreachable_blocks" to enum reachability.
7804 (set_all_edges_as_executable): New decl.
7805 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7806 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7808 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7809 but converting true to REACHABLE_BLOCKS.
7810 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7812 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7813 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7814 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7816 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7817 if check_all_array_refs will be called.
7819 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7821 * tree.c (selftest::test_location_wrappers): Add more test
7824 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7826 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7827 (selftest::test_bit_in_range): Likewise.
7829 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7832 * doc/sourcebuild.texi (vect_float): Say that the selector
7833 only describes the situation when -funsafe-math-optimizations is on.
7834 (vect_float_strict): Document.
7836 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7838 PR tree-optimization/83965
7839 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7840 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7841 instead of checking only for a reduction.
7842 (vect_recog_widen_sum_pattern): Likewise.
7844 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7846 * predict.c (probably_never_executed): Only use precise profile info.
7847 (compute_function_frequency): Skip after inlining hack since we now
7848 have quality checking.
7850 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7852 * profile-count.h (profile_probability::very_unlikely,
7853 profile_probability::unlikely, profile_probability::even): Set
7854 precision to guessed.
7856 2018-01-23 Richard Biener <rguenther@suse.de>
7858 PR tree-optimization/83963
7859 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7860 Properly terminate dominator walk when crossing the exit edge not
7861 when visiting its source block.
7863 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7866 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7867 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7869 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7871 PR tree-optimization/83957
7872 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7873 semicolon after for body surrounded by braces.
7875 PR tree-optimization/83081
7876 * profile-count.h (profile_probability::split): New method.
7877 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7878 Use profile_probability::split.
7879 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7880 when splitting a single conditional jump into 2.
7882 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7884 PR tree-optimization/69452
7885 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7888 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7890 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7891 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7892 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7894 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7896 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7897 * config/rl78/rl78.md (movdi): New define_expand.
7898 * config/rl78/rl78.c (rl78_split_movdi): New function.
7900 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7903 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7905 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7906 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7907 128-bit to produce an UNSPEC move to get the double word with the
7908 signbit and then a shift directly to do signbit.
7909 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7910 implementation with a new version that just does either a direct
7911 move or a regular move. Move memory interface to separate insns.
7912 Move insns so they are next to the expander.
7913 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7914 with signbit move. Split big and little endian case.
7915 (signbit<mode>2_dm_mem_le): Likewise.
7916 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7917 (signbit<mode>2_dm2): Likewise.
7919 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7921 * config/rl78/rl78.md (anddi3): New define_expand.
7923 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7925 * config/rl78/rl78.md (umindi3): New define_expand.
7927 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7929 * config/rl78/rl78.md (smindi3): New define_expand.
7931 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7933 * config/rl78/rl78.md (smaxdi3): New define_expand.
7935 2018-01-22 Carl Love <cel@us.ibm.com>
7937 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7938 LVX_V1TI): Add macro expansion.
7939 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7940 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7941 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7942 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7943 Change check to determine if the instruction is a byte reversing
7944 entry. Fix typo in comment.
7945 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7946 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7947 Add def_builtin calls for new builtins.
7948 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7949 Add define_insn expansion.
7951 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7953 * config/rl78/rl78.md (umaxdi3): New define_expand.
7955 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7957 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7958 for non-QImode registers.
7960 2018-01-22 Richard Biener <rguenther@suse.de>
7962 PR tree-optimization/83963
7963 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7964 including the loop exit block.
7965 (scop_detection::merge_sese): Likewise.
7966 (scop_detection::add_scop): Do it here instead.
7968 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7970 * doc/sourcebuild.texi (arm_softfloat): Document.
7972 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7975 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7976 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7977 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7979 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7980 David Edelsohn <dje.gcc@gmail.com>
7983 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7984 Change "crset eq" to "crset 2".
7985 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7986 (*call_indirect_aix<mode>_nospec): Likewise.
7987 (*call_value_indirect_aix<mode>_nospec): Likewise.
7988 (*call_indirect_elfv2<mode>_nospec): Likewise.
7989 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
7990 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7991 change assembly output from . to $.
7992 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7993 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7994 (*tablejump<mode>_internal1_nospec): Likewise.
7996 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7999 * config/sh/sh_optimize_sett_clrt.cc:
8000 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8002 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8004 PR tree-optimization/83940
8005 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8006 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8007 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8008 use it to pass back the definition type.
8009 (vect_check_store_rhs): Likewise rhs_dt_out.
8010 (vect_build_gather_load_calls): Add a mask_dt argument and use
8011 it instead of a call to vect_is_simple_use.
8012 (vectorizable_store): Update calls to vect_check_load_store_mask
8013 and vect_check_store_rhs. Use the dt returned by the latter instead
8014 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8015 instead of calls to vect_is_simple_use. Pass the scalar rather
8016 than the vector operand to vect_is_simple_use when handling
8017 second and subsequent copies of an rhs value.
8018 (vectorizable_load): Update calls to vect_check_load_store_mask
8019 and vect_build_gather_load_calls. Use the cached mask_dt and
8020 gs_info.offset_dt instead of calls to vect_is_simple_use.
8022 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8025 * tree-emutls.c: Include gimplify.h.
8026 (lower_emutls_2): New function.
8027 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8028 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8029 it before further processing.
8032 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8033 UINTVAL (trueop1) instead of INTVAL (op1).
8035 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8039 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8040 INCOMING_FRAME_SP_OFFSET if not defined.
8041 (scan_trace): Add ENTRY argument. If true and
8042 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8043 emit a note to adjust the CFA offset.
8044 (create_cfi_notes): Adjust scan_trace callers.
8045 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8046 INCOMING_FRAME_SP_OFFSET in the CIE.
8047 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8048 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8050 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8051 * doc/tm.texi: Regenerated.
8053 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8055 PR rtl-optimization/83147
8056 * lra-constraints.c (remove_inheritance_pseudos): Use
8057 lra_substitute_pseudo_within_insn.
8059 2018-01-19 Tom de Vries <tom@codesourcery.com>
8060 Cesar Philippidis <cesar@codesourcery.com>
8063 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8065 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8068 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8069 spaces for function labels.
8071 2018-01-19 Martin Liska <mliska@suse.cz>
8073 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8074 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8075 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8076 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8077 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8078 (PRED_CONST_RETURN): Change from 69 to 65.
8079 (PRED_NULL_RETURN): Change from 91 to 71.
8080 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8081 (PRED_LOOP_GUARD): Change from 66 to 73.
8083 2018-01-19 Martin Liska <mliska@suse.cz>
8085 * predict.c (predict_insn_def): Add new assert.
8086 (struct branch_predictor): Change type to signed integer.
8087 (test_prediction_value_range): Amend test to cover
8089 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8090 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8091 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8092 (PRED_LOOP_IV_COMPARE): Likewise.
8093 * predict.h (PROB_UNINITIALIZED): Define new constant.
8095 2018-01-19 Martin Liska <mliska@suse.cz>
8097 * predict.c (dump_prediction): Add new format for
8098 analyze_brprob.py script which is enabled with -details
8100 * profile-count.h (precise_p): New function.
8102 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8104 PR tree-optimization/83922
8105 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8106 there are no statements that need masking.
8107 (vect_active_double_reduction_p): New function.
8108 (vect_analyze_loop_operations): Use it when handling phis that
8109 are not in the loop header.
8111 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8113 PR tree-optimization/83914
8114 * tree-vect-loop.c (vectorizable_induction): Don't convert
8115 init_expr or apply the peeling adjustment for inductions
8116 that are nested within the vectorized loop.
8118 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8120 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8123 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8127 * function.h (gimplify_parameters): Add gimple_seq * argument.
8128 * function.c: Include gimple.h and options.h.
8129 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8130 for the added local temporaries if needed.
8131 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8132 if there are any parameter cleanups, wrap whole body into a
8133 try/finally with the cleanups.
8135 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8138 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8139 Use GET_MODE_CLASS for scalar floating point.
8141 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8145 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8146 Fix call of call_cgraph_insertion_hooks.
8148 2018-01-18 Martin Sebor <msebor@redhat.com>
8150 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8152 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8155 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8158 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8161 * common.opt: (-ffile-prefix-map): New option.
8162 * opts.c (common_handle_option): Defer it.
8163 * opts-global.c (handle_common_deferred_options): Handle it.
8164 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8165 * file-prefix-map.h: New file.
8166 (remap_debug_filename, add_debug_prefix_map): ...here.
8167 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8168 * final.c (debug_prefix_map, add_debug_prefix_map
8169 remap_debug_filename): Move to...
8170 * file-prefix-map.c: New file.
8171 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8172 generalize, get rid of alloca(), use strrchr() instead of strchr().
8173 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8174 Implement in terms of add_prefix_map().
8175 (remap_macro_filename, remap_debug_filename): Implement in term of
8177 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8178 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8179 * dbxout.c: Include file-prefix-map.h.
8180 * varasm.c: Likewise.
8181 * vmsdbgout.c: Likewise.
8182 * xcoffout.c: Likewise.
8183 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8184 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8185 * doc/invoke.texi (-ffile-prefix-map): Document.
8186 (-fdebug-prefix-map): Update description.
8188 2018-01-18 Martin Liska <mliska@suse.cz>
8190 * config/i386/i386.c (indirect_thunk_name): Document that also
8192 (output_indirect_thunk): Document why both instructions
8193 (pause and lfence) are generated.
8195 2018-01-18 Richard Biener <rguenther@suse.de>
8197 PR tree-optimization/83887
8198 * graphite-scop-detection.c
8199 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8200 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8201 (scop_detection::merge_sese): Re-implement with a flood-fill
8202 algorithm that properly finds a SESE region if it exists.
8204 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8207 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8208 pointer_diff optimizations use view_convert instead of convert.
8210 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8212 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8213 Generate different code for -mno-speculate-indirect-jumps.
8214 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8215 (*call_indirect_aix<mode>): Disable for
8216 -mno-speculate-indirect-jumps.
8217 (*call_indirect_aix<mode>_nospec): New define_insn.
8218 (*call_value_indirect_aix<mode>): Disable for
8219 -mno-speculate-indirect-jumps.
8220 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8221 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8222 -mno-speculate-indirect-jumps.
8223 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8225 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8227 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8228 long double type, set the flags for noting the default long double
8229 type, even if we don't pass or return a long double type.
8231 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8234 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8237 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8239 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8240 support for merge[hl].
8241 (fold_mergehl_helper): New helper function.
8242 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8243 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8244 (altivec_vmrglw_direct): Add xxmrglw insn.
8246 2018-01-17 Andrew Waterman <andrew@sifive.com>
8248 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8249 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8251 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8254 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8255 call the lto_location_cache before reading the
8256 DECL_SOURCE_LOCATION of the types.
8258 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8259 Richard Sandiford <richard.sandiford@linaro.org>
8261 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8262 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8263 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8264 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8265 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8267 * config/aarch64/constraints.md (aarch64_movti_operand):
8269 * config/aarch64/predicates.md (Uti): Add new constraint.
8271 2018-01-17 Carl Love <cel@us.ibm.com>
8273 * config/rs6000/vsx.md (define_expand xl_len_r,
8274 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8275 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8277 (define_expand, define_insn): Move the shift left from the
8278 define_insn to the define_expand for lxvl and stxvl instructions.
8279 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8280 and XL_LEN_R definitions to PURE.
8282 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8284 * config/i386/i386.c (indirect_thunk_name): Declare regno
8285 as unsigned int. Compare regno with INVALID_REGNUM.
8286 (output_indirect_thunk): Ditto.
8287 (output_indirect_thunk_function): Ditto.
8288 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8289 in the call to output_indirect_thunk_function.
8291 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8294 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8295 rather than the size of inner_type to determine the stack slot size
8296 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8298 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8301 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8304 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8306 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8307 endian Linux systems to optionally enable multilibs for selecting
8308 the long double type if the user configured an explicit type.
8309 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8310 have no long double multilibs if not defined.
8311 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8312 warn if the user used -mabi={ieee,ibm}longdouble and we built
8313 multilibs for long double.
8314 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8315 appropriate multilib option.
8316 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8318 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8319 for building long double multilibs.
8320 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8322 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8324 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8327 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8329 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8332 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8335 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8338 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8340 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8341 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8343 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8345 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8346 different rtl trees depending on TARGET_64BIT.
8347 (rs6000_gen_lvx): Likewise.
8349 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8351 * config/visium/visium.md (nop): Tweak comment.
8352 (hazard_nop): Likewise.
8354 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8356 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8357 -mspeculate-indirect-jumps.
8358 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8359 for -mno-speculate-indirect-jumps.
8360 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8361 (*call_value_indirect_elfv2<mode>): Disable for
8362 -mno-speculate-indirect-jumps.
8363 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8364 (indirect_jump): Emit different RTL for
8365 -mno-speculate-indirect-jumps.
8366 (*indirect_jump<mode>): Disable for
8367 -mno-speculate-indirect-jumps.
8368 (*indirect_jump<mode>_nospec): New define_insn.
8369 (tablejump): Emit different RTL for
8370 -mno-speculate-indirect-jumps.
8371 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8372 (tablejumpsi_nospec): New define_expand.
8373 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8374 (tablejumpdi_nospec): New define_expand.
8375 (*tablejump<mode>_internal1): Disable for
8376 -mno-speculate-indirect-jumps.
8377 (*tablejump<mode>_internal1_nospec): New define_insn.
8378 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8381 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8383 * caller-save.c (insert_save): Drop unnecessary parameter. All
8386 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8387 Richard Biener <rguenth@suse.de>
8390 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8391 return early, inline manually is_gimple_sizepos. Make sure if we
8392 call gimplify_expr we don't end up with a gimple constant.
8393 * tree.c (variably_modified_type_p): Don't return true for
8394 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8395 * gimplify.h (is_gimple_sizepos): Remove.
8397 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8399 PR tree-optimization/83857
8400 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8401 vectorizable_live_operation for pure SLP statements.
8402 (vectorizable_live_operation): Handle PHIs.
8404 2018-01-16 Richard Biener <rguenther@suse.de>
8406 PR tree-optimization/83867
8407 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8408 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8410 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8413 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8414 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8415 If off is not INTEGER_CST, issue a may not be aligned warning
8416 rather than isn't aligned. Use isn%'t rather than isn't.
8417 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8419 <case MULT_EXPR>: Improve the case when bottom and one of the
8420 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8421 operand, in that case check if the other operand is multiple of
8422 bottom divided by the INTEGER_CST operand.
8424 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8427 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8428 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8429 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8430 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8431 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8432 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8433 (pa_function_arg_size): New function.
8435 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8437 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8438 in a separate statement.
8440 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8442 PR tree-optimization/83847
8443 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8444 group gathers and scatters.
8446 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8448 PR rtl-optimization/86620
8449 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8451 PR rtl-optimization/83213
8452 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8453 to last if both are JUMP_INSNs.
8455 PR tree-optimization/83843
8456 * gimple-ssa-store-merging.c
8457 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8458 store_immediate_info for bswap/nop orig_stores.
8460 2018-01-15 Andrew Waterman <andrew@sifive.com>
8462 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8464 <UDIV>: Increase cost if !TARGET_DIV.
8466 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8468 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8469 (define_attr "cr_logical_3op"): New.
8470 (cceq_ior_compare): Adjust.
8471 (cceq_ior_compare_complement): Adjust.
8472 (*cceq_rev_compare): Adjust.
8473 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8474 (is_cracked_insn): Adjust.
8475 (insn_must_be_first_in_group): Adjust.
8476 * config/rs6000/40x.md: Adjust.
8477 * config/rs6000/440.md: Adjust.
8478 * config/rs6000/476.md: Adjust.
8479 * config/rs6000/601.md: Adjust.
8480 * config/rs6000/603.md: Adjust.
8481 * config/rs6000/6xx.md: Adjust.
8482 * config/rs6000/7450.md: Adjust.
8483 * config/rs6000/7xx.md: Adjust.
8484 * config/rs6000/8540.md: Adjust.
8485 * config/rs6000/cell.md: Adjust.
8486 * config/rs6000/e300c2c3.md: Adjust.
8487 * config/rs6000/e500mc.md: Adjust.
8488 * config/rs6000/e500mc64.md: Adjust.
8489 * config/rs6000/e5500.md: Adjust.
8490 * config/rs6000/e6500.md: Adjust.
8491 * config/rs6000/mpc.md: Adjust.
8492 * config/rs6000/power4.md: Adjust.
8493 * config/rs6000/power5.md: Adjust.
8494 * config/rs6000/power6.md: Adjust.
8495 * config/rs6000/power7.md: Adjust.
8496 * config/rs6000/power8.md: Adjust.
8497 * config/rs6000/power9.md: Adjust.
8498 * config/rs6000/rs64.md: Adjust.
8499 * config/rs6000/titan.md: Adjust.
8501 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8503 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8504 ix86_indirect_branch_register logic.
8506 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8508 * config/i386/constraints.md (Bs): Update
8509 ix86_indirect_branch_register check. Don't check
8510 ix86_indirect_branch_register with GOT_memory_operand.
8512 * config/i386/predicates.md (GOT_memory_operand): Don't check
8513 ix86_indirect_branch_register here.
8514 (GOT32_symbol_operand): Likewise.
8516 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8518 * config/i386/predicates.md (constant_call_address_operand):
8519 Rewrite ix86_indirect_branch_register logic.
8520 (sibcall_insn_operand): Likewise.
8522 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8524 * config/i386/constraints.md (Bs): Replace
8525 ix86_indirect_branch_thunk_register with
8526 ix86_indirect_branch_register.
8528 * config/i386/i386.md (indirect_jump): Likewise.
8529 (tablejump): Likewise.
8530 (*sibcall_memory): Likewise.
8531 (*sibcall_value_memory): Likewise.
8532 Peepholes of indirect call and jump via memory: Likewise.
8533 * config/i386/i386.opt: Likewise.
8534 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8535 (GOT_memory_operand): Likewise.
8536 (call_insn_operand): Likewise.
8537 (sibcall_insn_operand): Likewise.
8538 (GOT32_symbol_operand): Likewise.
8540 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8543 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8544 type rather than type addr's type points to.
8545 (expand_omp_atomic_mutex): Likewise.
8546 (expand_omp_atomic): Likewise.
8548 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8551 * config/i386/i386.c (output_indirect_thunk_function): Use
8552 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8553 for __x86_return_thunk.
8555 2018-01-15 Richard Biener <rguenther@suse.de>
8558 * expmed.c (extract_bit_field_1): Fix typo.
8560 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8563 * config/arm/iterators.md (VF): New mode iterator.
8564 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8565 Remove integer-related logic from pattern.
8566 (neon_vabd<mode>_3): Likewise.
8568 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8571 * common.opt (fstrict-overflow): No longer an alias.
8572 (fwrapv-pointer): New option.
8573 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8574 also for pointer types based on flag_wrapv_pointer.
8575 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8576 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8577 opts->x_flag_wrapv got set.
8578 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8579 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8580 POINTER_TYPE_OVERFLOW_UNDEFINED.
8581 * match.pd: Likewise in address comparison pattern.
8582 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8584 2018-01-15 Richard Biener <rguenther@suse.de>
8587 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8588 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8589 Reset type names to their identifier if their TYPE_DECL doesn't
8590 have linkage (and thus is used for ODR and devirt).
8591 (save_debug_info_for_decl): Remove.
8592 (save_debug_info_for_type): Likewise.
8593 (add_tree_to_fld_list): Adjust.
8594 * tree-pretty-print.c (dump_generic_node): Make dumping of
8595 type names more robust.
8597 2018-01-15 Richard Biener <rguenther@suse.de>
8599 * BASE-VER: Bump to 8.0.1.
8601 2018-01-14 Martin Sebor <msebor@redhat.com>
8604 * builtins.c (check_access): Avoid warning when the no-warning bit
8607 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8609 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8610 * ira-color (allocno_hard_regs_compare): Likewise.
8612 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8615 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8616 Use .pushsection/.popsection.
8618 2018-01-14 Martin Sebor <msebor@redhat.com>
8621 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8623 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8625 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8626 entry from extra_headers.
8627 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8628 extra_headers, make the list bitwise identical to the i?86-*-* one.
8630 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8632 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8633 -mcmodel=large with -mindirect-branch=thunk,
8634 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8635 -mfunction-return=thunk-extern.
8636 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8637 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8638 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8640 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8642 * config/i386/i386.c (print_reg): Print the name of the full
8643 integer register without '%'.
8644 (ix86_print_operand): Handle 'V'.
8645 * doc/extend.texi: Document 'V' modifier.
8647 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8649 * config/i386/constraints.md (Bs): Disallow memory operand for
8650 -mindirect-branch-register.
8652 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8653 (GOT_memory_operand): Likewise.
8654 (call_insn_operand): Likewise.
8655 (sibcall_insn_operand): Likewise.
8656 (GOT32_symbol_operand): Likewise.
8657 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8658 for -mindirect-branch-register.
8659 (tablejump): Likewise.
8660 (*sibcall_memory): Likewise.
8661 (*sibcall_value_memory): Likewise.
8662 Disallow peepholes of indirect call and jump via memory for
8663 -mindirect-branch-register.
8664 (*call_pop): Replace m with Bw.
8665 (*call_value_pop): Likewise.
8666 (*sibcall_pop_memory): Replace m with Bs.
8667 * config/i386/i386.opt (mindirect-branch-register): New option.
8668 * doc/invoke.texi: Document -mindirect-branch-register option.
8670 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8672 * config/i386/i386-protos.h (ix86_output_function_return): New.
8673 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8674 set function_return_type.
8675 (indirect_thunk_name): Add ret_p to indicate thunk for function
8677 (output_indirect_thunk_function): Pass false to
8678 indirect_thunk_name.
8679 (ix86_output_indirect_branch_via_reg): Likewise.
8680 (ix86_output_indirect_branch_via_push): Likewise.
8681 (output_indirect_thunk_function): Create alias for function
8682 return thunk if regno < 0.
8683 (ix86_output_function_return): New function.
8684 (ix86_handle_fndecl_attribute): Handle function_return.
8685 (ix86_attribute_table): Add function_return.
8686 * config/i386/i386.h (machine_function): Add
8687 function_return_type.
8688 * config/i386/i386.md (simple_return_internal): Use
8689 ix86_output_function_return.
8690 (simple_return_internal_long): Likewise.
8691 * config/i386/i386.opt (mfunction-return=): New option.
8692 (indirect_branch): Mention -mfunction-return=.
8693 * doc/extend.texi: Document function_return function attribute.
8694 * doc/invoke.texi: Document -mfunction-return= option.
8696 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8698 * config/i386/i386-opts.h (indirect_branch): New.
8699 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8700 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8701 with local indirect jump when converting indirect call and jump.
8702 (ix86_set_indirect_branch_type): New.
8703 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8704 (indirectlabelno): New.
8705 (indirect_thunk_needed): Likewise.
8706 (indirect_thunk_bnd_needed): Likewise.
8707 (indirect_thunks_used): Likewise.
8708 (indirect_thunks_bnd_used): Likewise.
8709 (INDIRECT_LABEL): Likewise.
8710 (indirect_thunk_name): Likewise.
8711 (output_indirect_thunk): Likewise.
8712 (output_indirect_thunk_function): Likewise.
8713 (ix86_output_indirect_branch_via_reg): Likewise.
8714 (ix86_output_indirect_branch_via_push): Likewise.
8715 (ix86_output_indirect_branch): Likewise.
8716 (ix86_output_indirect_jmp): Likewise.
8717 (ix86_code_end): Call output_indirect_thunk_function if needed.
8718 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8720 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8721 (ix86_attribute_table): Add indirect_branch.
8722 * config/i386/i386.h (machine_function): Add indirect_branch_type
8723 and has_local_indirect_jump.
8724 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8726 (tablejump): Likewise.
8727 (*indirect_jump): Use ix86_output_indirect_jmp.
8728 (*tablejump_1): Likewise.
8729 (simple_return_indirect_internal): Likewise.
8730 * config/i386/i386.opt (mindirect-branch=): New option.
8731 (indirect_branch): New.
8734 (thunk-inline): Likewise.
8735 (thunk-extern): Likewise.
8736 * doc/extend.texi: Document indirect_branch function attribute.
8737 * doc/invoke.texi: Document -mindirect-branch= option.
8739 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8742 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8744 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8746 * ipa-inline.c (want_inline_small_function_p): Return false if
8747 inlining has already failed with CIF_FINAL_ERROR.
8748 (update_caller_keys): Call want_inline_small_function_p before
8750 (update_callee_keys): Likewise.
8752 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8754 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8756 (rs6000_quadword_masked_address_p): Likewise.
8757 (quad_aligned_load_p): Likewise.
8758 (quad_aligned_store_p): Likewise.
8759 (const_load_sequence_p): Add comment to describe the outer-most loop.
8760 (mimic_memory_attributes_and_flags): New function.
8761 (rs6000_gen_stvx): Likewise.
8762 (replace_swapped_aligned_store): Likewise.
8763 (rs6000_gen_lvx): Likewise.
8764 (replace_swapped_aligned_load): Likewise.
8765 (replace_swapped_load_constant): Capitalize argument name in
8766 comment describing this function.
8767 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8768 and stores that access quad-word aligned addresses and replace
8769 with stvx or lvx instructions when appropriate.
8770 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8771 New function prototype.
8772 (rs6000_quadword_masked_address_p): Likewise.
8773 (rs6000_gen_lvx): Likewise.
8774 (rs6000_gen_stvx): Likewise.
8775 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8776 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8777 when memory address is aligned.
8778 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8779 this split to select lvx instruction when memory address is aligned.
8780 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8781 instruction when memory address is aligned.
8782 (*vsx_le_perm_load_v16qi): Likewise.
8783 (four unnamed splitters): Modify to select the stvx instruction
8784 when memory is aligned.
8786 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8788 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8789 which appears in the queue multiple times.
8791 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8792 Alan Hayward <alan.hayward@arm.com>
8793 David Sherwood <david.sherwood@arm.com>
8795 * tree-vectorizer.h (vec_lower_bound): New structure.
8796 (_loop_vec_info): Add check_nonzero and lower_bounds.
8797 (LOOP_VINFO_CHECK_NONZERO): New macro.
8798 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8799 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8800 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8801 fields. Make seg_len the distance travelled, not including the
8803 (dr_direction_indicator): Declare.
8804 (dr_zero_step_indicator): Likewise.
8805 (dr_known_forward_stride_p): Likewise.
8806 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8808 (runtime_alias_check_p): Allow runtime alias checks with
8810 (operator ==): Compare access_size and align.
8811 (prune_runtime_alias_test_list): Rework for new distinction between
8812 the access_size and seg_len.
8813 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8815 (get_segment_min_max): New function.
8816 (create_intersect_range_checks): Use it.
8817 (dr_step_indicator): New function.
8818 (dr_direction_indicator): Likewise.
8819 (dr_zero_step_indicator): Likewise.
8820 (dr_known_forward_stride_p): Likewise.
8821 * tree-loop-distribution.c (data_ref_segment_size): Return
8822 DR_STEP * (niters - 1).
8823 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8825 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8826 (vect_preserves_scalar_order_p): New function, split out from...
8827 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8828 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8829 (vect_vfa_access_size): New function.
8830 (vect_vfa_align): Likewise.
8831 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8832 (dump_lower_bound): New function.
8833 (vect_check_lower_bound): Likewise.
8834 (vect_small_gap_p): Likewise.
8835 (vectorizable_with_step_bound_p): Likewise.
8836 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8837 depencies if the vectorization factor is 1. Convert the checks
8838 for nonzero steps into checks on the bounds of DR_STEP. Try using
8839 a bunds check for variable steps if the minimum required step is
8840 relatively small. Update calls to the dr_with_seg_len
8841 constructor and to vect_compile_time_alias.
8842 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8844 (vect_loop_versioning): Call it.
8845 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8847 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8849 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8850 Alan Hayward <alan.hayward@arm.com>
8851 David Sherwood <david.sherwood@arm.com>
8853 * doc/sourcebuild.texi (vect_scatter_store): Document.
8854 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8856 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8858 * genopinit.c (main): Add supports_vec_scatter_store and
8859 supports_vec_scatter_store_cached to target_optabs.
8860 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8861 IFN_MASK_SCATTER_STORE.
8862 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8864 * internal-fn.h (internal_store_fn_p): Declare.
8865 (internal_fn_stored_value_index): Likewise.
8866 * internal-fn.c (scatter_store_direct): New macro.
8867 (expand_scatter_store_optab_fn): New function.
8868 (direct_scatter_store_optab_supported_p): New macro.
8869 (internal_store_fn_p): New function.
8870 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8871 IFN_MASK_SCATTER_STORE.
8872 (internal_fn_mask_index): Likewise.
8873 (internal_fn_stored_value_index): New function.
8874 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8876 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8877 * optabs-query.c (supports_vec_scatter_store_p): New function.
8878 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8879 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8880 true for scatter stores.
8881 (vect_gather_scatter_fn_p): Handle scatter stores too.
8882 (vect_check_gather_scatter): Consider using scatter stores if
8883 supports_vec_scatter_store_p.
8884 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8886 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8887 internal_fn_stored_value_index.
8888 (check_load_store_masking): Handle scatter stores too.
8889 (vect_get_store_rhs): Make public.
8890 (vectorizable_call): Use internal_store_fn_p.
8891 (vectorizable_store): Handle scatter store internal functions.
8892 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8893 when deciding whether the end of the group has been reached.
8894 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8895 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8896 (mask_scatter_store<mode>): New insns.
8898 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8899 Alan Hayward <alan.hayward@arm.com>
8900 David Sherwood <david.sherwood@arm.com>
8902 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8903 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8904 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8906 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8907 Use vect_truncate_gather_scatter_offset if we can't treat the
8908 operation as a normal gather load or scatter store.
8909 (get_group_load_store_type): Take the gather_scatter_info
8910 as argument. Try using a gather load or scatter store for
8911 single-element groups.
8912 (get_load_store_type): Update calls to get_group_load_store_type
8913 and vect_use_strided_gather_scatters_p.
8915 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8916 Alan Hayward <alan.hayward@arm.com>
8917 David Sherwood <david.sherwood@arm.com>
8919 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8920 optional tree argument.
8921 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8923 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8924 but continue to use the current value as a fallback.
8925 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8926 to compare the updates.
8927 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8928 (get_load_store_type): Use it when handling a strided access.
8929 (vect_get_strided_load_store_ops): New function.
8930 (vect_get_data_ptr_increment): Likewise.
8931 (vectorizable_load): Handle strided gather loads. Always pass
8932 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8934 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8935 Alan Hayward <alan.hayward@arm.com>
8936 David Sherwood <david.sherwood@arm.com>
8938 * doc/md.texi (gather_load@var{m}): Document.
8939 (mask_gather_load@var{m}): Likewise.
8940 * genopinit.c (main): Add supports_vec_gather_load and
8941 supports_vec_gather_load_cached to target_optabs.
8942 * optabs-tree.c (init_tree_optimization_optabs): Use
8943 ggc_cleared_alloc to allocate target_optabs.
8944 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8945 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8947 * internal-fn.h (internal_load_fn_p): Declare.
8948 (internal_gather_scatter_fn_p): Likewise.
8949 (internal_fn_mask_index): Likewise.
8950 (internal_gather_scatter_fn_supported_p): Likewise.
8951 * internal-fn.c (gather_load_direct): New macro.
8952 (expand_gather_load_optab_fn): New function.
8953 (direct_gather_load_optab_supported_p): New macro.
8954 (direct_internal_fn_optab): New function.
8955 (internal_load_fn_p): Likewise.
8956 (internal_gather_scatter_fn_p): Likewise.
8957 (internal_fn_mask_index): Likewise.
8958 (internal_gather_scatter_fn_supported_p): Likewise.
8959 * optabs-query.c (supports_at_least_one_mode_p): New function.
8960 (supports_vec_gather_load_p): Likewise.
8961 * optabs-query.h (supports_vec_gather_load_p): Declare.
8962 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8963 and memory_type field.
8964 (NUM_PATTERNS): Bump to 15.
8965 * tree-vect-data-refs.c: Include internal-fn.h.
8966 (vect_gather_scatter_fn_p): New function.
8967 (vect_describe_gather_scatter_call): Likewise.
8968 (vect_check_gather_scatter): Try using internal functions for
8969 gather loads. Recognize existing calls to a gather load function.
8970 (vect_analyze_data_refs): Consider using gather loads if
8971 supports_vec_gather_load_p.
8972 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8973 (vect_get_gather_scatter_offset_type): Likewise.
8974 (vect_convert_mask_for_vectype): Likewise.
8975 (vect_add_conversion_to_patterm): Likewise.
8976 (vect_try_gather_scatter_pattern): Likewise.
8977 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8978 (vect_vect_recog_func_ptrs): Add it.
8979 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8980 internal_fn_mask_index and internal_gather_scatter_fn_p.
8981 (check_load_store_masking): Take the gather_scatter_info as an
8982 argument and handle gather loads.
8983 (vect_get_gather_scatter_ops): New function.
8984 (vectorizable_call): Check internal_load_fn_p.
8985 (vectorizable_load): Likewise. Handle gather load internal
8987 (vectorizable_store): Update call to check_load_store_masking.
8988 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
8989 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
8990 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8991 (aarch64_gather_scale_operand_d): New predicates.
8992 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8993 (mask_gather_load<mode>): New insns.
8995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8996 Alan Hayward <alan.hayward@arm.com>
8997 David Sherwood <david.sherwood@arm.com>
8999 * optabs.def (fold_left_plus_optab): New optab.
9000 * doc/md.texi (fold_left_plus_@var{m}): Document.
9001 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9002 * internal-fn.c (fold_left_direct): Define.
9003 (expand_fold_left_optab_fn): Likewise.
9004 (direct_fold_left_optab_supported_p): Likewise.
9005 * fold-const-call.c (fold_const_fold_left): New function.
9006 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9007 * tree-parloops.c (valid_reduction_p): New function.
9008 (gather_scalar_reductions): Use it.
9009 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9010 (vect_finish_replace_stmt): Declare.
9011 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9012 (needs_fold_left_reduction_p): New function, split out from...
9013 (vect_is_simple_reduction): ...here. Accept reductions that
9014 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9015 (vect_force_simple_reduction): Also store the reduction type in
9016 the assignment's STMT_VINFO_REDUC_TYPE.
9017 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9018 (merge_with_identity): New function.
9019 (vect_expand_fold_left): Likewise.
9020 (vectorize_fold_left_reduction): Likewise.
9021 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9022 scalar phi in place for it. Check for target support and reject
9023 cases that would reassociate the operation. Defer the transform
9024 phase to vectorize_fold_left_reduction.
9025 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9026 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9027 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9029 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9031 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9032 call to ifc_temp_var.
9034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9035 Alan Hayward <alan.hayward@arm.com>
9036 David Sherwood <david.sherwood@arm.com>
9038 * target.def (legitimize_address_displacement): Take the original
9039 offset as a poly_int.
9040 * targhooks.h (default_legitimize_address_displacement): Update
9042 * targhooks.c (default_legitimize_address_displacement): Likewise.
9043 * doc/tm.texi: Regenerate.
9044 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9045 as an argument, moving assert of ad->disp == ad->disp_term to...
9046 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9047 Try calling targetm.legitimize_address_displacement before expanding
9048 the address rather than afterwards, and adjust for the new interface.
9049 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9050 Match the new hook interface. Handle SVE addresses.
9051 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9056 * Makefile.in (OBJS): Add early-remat.o.
9057 * target.def (select_early_remat_modes): New hook.
9058 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9059 * doc/tm.texi: Regenerate.
9060 * targhooks.h (default_select_early_remat_modes): Declare.
9061 * targhooks.c (default_select_early_remat_modes): New function.
9062 * timevar.def (TV_EARLY_REMAT): New timevar.
9063 * passes.def (pass_early_remat): New pass.
9064 * tree-pass.h (make_pass_early_remat): Declare.
9065 * early-remat.c: New file.
9066 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9068 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9070 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9071 Alan Hayward <alan.hayward@arm.com>
9072 David Sherwood <david.sherwood@arm.com>
9074 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9075 vfm1 with a bound_epilog parameter.
9076 (vect_do_peeling): Update calls accordingly, and move the prologue
9077 call earlier in the function. Treat the base bound_epilog as 0 for
9078 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9079 this base when peeling for gaps.
9080 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9081 with fully-masked loops.
9082 (vect_estimate_min_profitable_iters): Handle the single peeled
9083 iteration in that case.
9085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9086 Alan Hayward <alan.hayward@arm.com>
9087 David Sherwood <david.sherwood@arm.com>
9089 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9090 single-element interleaving even if the size is not a power of 2.
9091 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9092 accesses for single-element interleaving if the group size is
9095 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9096 Alan Hayward <alan.hayward@arm.com>
9097 David Sherwood <david.sherwood@arm.com>
9099 * doc/md.texi (fold_extract_last_@var{m}): Document.
9100 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9101 * optabs.def (fold_extract_last_optab): New optab.
9102 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9103 * internal-fn.c (fold_extract_direct): New macro.
9104 (expand_fold_extract_optab_fn): Likewise.
9105 (direct_fold_extract_optab_supported_p): Likewise.
9106 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9107 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9108 EXTRACT_LAST_REDUCTION.
9109 (get_initial_def_for_reduction): Do not create an initial vector
9110 for EXTRACT_LAST_REDUCTION reductions.
9111 (vectorizable_reduction): Leave the scalar phi in place for
9112 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9113 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9114 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9115 transform phase to vectorizable_condition.
9116 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9118 (vect_finish_stmt_generation): ...here.
9119 (vect_finish_replace_stmt): New function.
9120 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9121 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9123 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9125 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9126 Alan Hayward <alan.hayward@arm.com>
9127 David Sherwood <david.sherwood@arm.com>
9129 * doc/md.texi (extract_last_@var{m}): Document.
9130 * optabs.def (extract_last_optab): New optab.
9131 * internal-fn.def (EXTRACT_LAST): New internal function.
9132 * internal-fn.c (cond_unary_direct): New macro.
9133 (expand_cond_unary_optab_fn): Likewise.
9134 (direct_cond_unary_optab_supported_p): Likewise.
9135 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9136 loops using EXTRACT_LAST.
9137 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9138 (extract_last_<mode>): ...this optab.
9139 (vec_extract<mode><Vel>): Update accordingly.
9141 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9142 Alan Hayward <alan.hayward@arm.com>
9143 David Sherwood <david.sherwood@arm.com>
9145 * target.def (empty_mask_is_expensive): New hook.
9146 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9147 * doc/tm.texi: Regenerate.
9148 * targhooks.h (default_empty_mask_is_expensive): Declare.
9149 * targhooks.c (default_empty_mask_is_expensive): New function.
9150 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9151 if the target says that empty masks are expensive.
9152 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9154 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9156 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9157 Alan Hayward <alan.hayward@arm.com>
9158 David Sherwood <david.sherwood@arm.com>
9160 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9161 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9162 (vect_use_loop_mask_for_alignment_p): New function.
9163 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9164 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9165 niters_skip argument. Make sure that the first niters_skip elements
9166 of the first iteration are inactive.
9167 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9168 Update call to vect_set_loop_masks_directly.
9169 (get_misalign_in_elems): New function, split out from...
9170 (vect_gen_prolog_loop_niters): ...here.
9171 (vect_update_init_of_dr): Take a code argument that specifies whether
9172 the adjustment should be added or subtracted.
9173 (vect_update_init_of_drs): Likewise.
9174 (vect_prepare_for_masked_peels): New function.
9175 (vect_do_peeling): Skip prologue peeling if we're using a mask
9176 instead. Update call to vect_update_inits_of_drs.
9177 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9179 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9180 alignment. Do not include the number of peeled iterations in
9181 the minimum threshold in that case.
9182 (vectorizable_induction): Adjust the start value down by
9183 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9184 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9185 Take the number of skipped iterations into account when calculating
9187 * tree-vect-stmts.c (vect_gen_while_not): New function.
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/sourcebuild.texi (vect_fully_masked): Document.
9194 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9196 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9198 (vect_analyze_loop_2): ...here. Don't check the vectorization
9199 factor against the number of loop iterations if the loop is
9202 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9203 Alan Hayward <alan.hayward@arm.com>
9204 David Sherwood <david.sherwood@arm.com>
9206 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9207 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9208 (dump_groups): Update accordingly.
9209 (iv_use::mem_type): New member variable.
9210 (address_p): New function.
9211 (record_use): Add a mem_type argument and initialize the new
9213 (record_group_use): Add a mem_type argument. Use address_p.
9214 Remove obsolete null checks of base_object. Update call to record_use.
9215 (find_interesting_uses_op): Update call to record_group_use.
9216 (find_interesting_uses_cond): Likewise.
9217 (find_interesting_uses_address): Likewise.
9218 (get_mem_type_for_internal_fn): New function.
9219 (find_address_like_use): Likewise.
9220 (find_interesting_uses_stmt): Try find_address_like_use before
9221 calling find_interesting_uses_op.
9222 (addr_offset_valid_p): Use the iv mem_type field as the type
9223 of the addressed memory.
9224 (add_autoinc_candidates): Likewise.
9225 (get_address_cost): Likewise.
9226 (split_small_address_groups_p): Use address_p.
9227 (split_address_groups): Likewise.
9228 (add_iv_candidate_for_use): Likewise.
9229 (autoinc_possible_for_pair): Likewise.
9230 (rewrite_groups): Likewise.
9231 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9232 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9233 (get_alias_ptr_type_for_ptr_address): New function.
9234 (rewrite_use_address): Rewrite address uses in calls that were
9235 identified by find_address_like_use.
9237 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9238 Alan Hayward <alan.hayward@arm.com>
9239 David Sherwood <david.sherwood@arm.com>
9241 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9243 * gimple-expr.h (is_gimple_addressable: Likewise.
9244 * gimple-expr.c (is_gimple_address): Likewise.
9245 * internal-fn.c (expand_call_mem_ref): New function.
9246 (expand_mask_load_optab_fn): Use it.
9247 (expand_mask_store_optab_fn): Likewise.
9249 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9250 Alan Hayward <alan.hayward@arm.com>
9251 David Sherwood <david.sherwood@arm.com>
9253 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9254 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9255 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9256 (cond_umax@var{mode}): Document.
9257 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9258 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9259 (cond_umin_optab, cond_umax_optab): New optabs.
9260 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9261 (COND_IOR, COND_XOR): New internal functions.
9262 * internal-fn.h (get_conditional_internal_fn): Declare.
9263 * internal-fn.c (cond_binary_direct): New macro.
9264 (expand_cond_binary_optab_fn): Likewise.
9265 (direct_cond_binary_optab_supported_p): Likewise.
9266 (get_conditional_internal_fn): New function.
9267 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9268 Cope with reduction statements that are vectorized as calls rather
9270 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9271 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9272 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9273 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9274 (UNSPEC_COND_EOR): New unspecs.
9275 (optab): Add mappings for them.
9276 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9277 (sve_int_op, sve_fp_op): New int attributes.
9279 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9280 Alan Hayward <alan.hayward@arm.com>
9281 David Sherwood <david.sherwood@arm.com>
9283 * optabs.def (while_ult_optab): New optab.
9284 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9285 * internal-fn.def (WHILE_ULT): New internal function.
9286 * internal-fn.h (direct_internal_fn_supported_p): New override
9287 that takes two types as argument.
9288 * internal-fn.c (while_direct): New macro.
9289 (expand_while_optab_fn): New function.
9290 (convert_optab_supported_p): Likewise.
9291 (direct_while_optab_supported_p): New macro.
9292 * wide-int.h (wi::udiv_ceil): New function.
9293 * tree-vectorizer.h (rgroup_masks): New structure.
9294 (vec_loop_masks): New typedef.
9295 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9297 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9298 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9299 (vect_max_vf): New function.
9300 (slpeel_make_loop_iterate_ntimes): Delete.
9301 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9302 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9303 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9304 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9305 internal-fn.h, stor-layout.h and optabs-query.h.
9306 (vect_set_loop_mask): New function.
9307 (add_preheader_seq): Likewise.
9308 (add_header_seq): Likewise.
9309 (interleave_supported_p): Likewise.
9310 (vect_maybe_permute_loop_masks): Likewise.
9311 (vect_set_loop_masks_directly): Likewise.
9312 (vect_set_loop_condition_masked): Likewise.
9313 (vect_set_loop_condition_unmasked): New function, split out from
9314 slpeel_make_loop_iterate_ntimes.
9315 (slpeel_make_loop_iterate_ntimes): Rename to..
9316 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9317 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9318 (vect_do_peeling): Update call accordingly.
9319 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9321 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9322 mask_compare_type, can_fully_mask_p and fully_masked_p.
9323 (release_vec_loop_masks): New function.
9324 (_loop_vec_info): Use it to free the loop masks.
9325 (can_produce_all_loop_masks_p): New function.
9326 (vect_get_max_nscalars_per_iter): Likewise.
9327 (vect_verify_full_masking): Likewise.
9328 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9329 retries, and free the mask rgroups before retrying. Check loop-wide
9330 reasons for disallowing fully-masked loops. Make the final decision
9331 about whether use a fully-masked loop or not.
9332 (vect_estimate_min_profitable_iters): Do not assume that peeling
9333 for the number of iterations will be needed for fully-masked loops.
9334 (vectorizable_reduction): Disable fully-masked loops.
9335 (vectorizable_live_operation): Likewise.
9336 (vect_halve_mask_nunits): New function.
9337 (vect_double_mask_nunits): Likewise.
9338 (vect_record_loop_mask): Likewise.
9339 (vect_get_loop_mask): Likewise.
9340 (vect_transform_loop): Handle the case in which the final loop
9341 iteration might handle a partial vector. Call vect_set_loop_condition
9342 instead of slpeel_make_loop_iterate_ntimes.
9343 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9344 (check_load_store_masking): New function.
9345 (prepare_load_store_mask): Likewise.
9346 (vectorizable_store): Handle fully-masked loops.
9347 (vectorizable_load): Likewise.
9348 (supportable_widening_operation): Use vect_halve_mask_nunits for
9350 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9351 (vect_gen_while): New function.
9352 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9353 (aarch64_uqdec<mode>): New insn.
9355 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9356 Alan Hayward <alan.hayward@arm.com>
9357 David Sherwood <david.sherwood@arm.com>
9359 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9360 (reduc_xor_scal_optab): New optabs.
9361 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9362 (reduc_xor_scal_@var{m}): Document.
9363 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9364 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9366 * fold-const-call.c (fold_const_call): Handle them.
9367 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9368 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9369 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9370 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9371 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9372 (UNSPEC_XORV): New unspecs.
9373 (optab): Add entries for them.
9374 (BITWISEV): New int iterator.
9375 (bit_reduc_op): New int attributes.
9377 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9378 Alan Hayward <alan.hayward@arm.com>
9379 David Sherwood <david.sherwood@arm.com>
9381 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9382 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9383 * optabs.def (vec_shl_insert_optab): New optab.
9384 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9385 (duplicate_and_interleave): Likewise.
9386 * tree-vect-loop.c: Include internal-fn.h.
9387 (neutral_op_for_slp_reduction): New function, split out from
9388 get_initial_defs_for_reduction.
9389 (get_initial_def_for_reduction): Handle option 2 for variable-length
9390 vectors by loading the neutral value into a vector and then shifting
9391 the initial value into element 0.
9392 (get_initial_defs_for_reduction): Replace the code argument with
9393 the neutral value calculated by neutral_op_for_slp_reduction.
9394 Use gimple_build_vector for constant-length vectors.
9395 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9396 but the first group_size elements have a neutral value.
9397 Use duplicate_and_interleave otherwise.
9398 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9399 Update call to get_initial_defs_for_reduction. Handle SLP
9400 reductions for variable-length vectors by creating one vector
9401 result for each scalar result, with the elements associated
9402 with other scalar results stubbed out with the neutral value.
9403 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9404 Require IFN_VEC_SHL_INSERT for double reductions on
9405 variable-length vectors, or SLP reductions that have
9406 a neutral value. Require can_duplicate_and_interleave_p
9407 support for variable-length unchained SLP reductions if there
9408 is no neutral value, such as for MIN/MAX reductions. Also require
9409 the number of vector elements to be a multiple of the number of
9410 SLP statements when doing variable-length unchained SLP reductions.
9411 Update call to vect_create_epilog_for_reduction.
9412 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9413 and remove initial values.
9414 (duplicate_and_interleave): Make public.
9415 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9416 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9418 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9419 Alan Hayward <alan.hayward@arm.com>
9420 David Sherwood <david.sherwood@arm.com>
9422 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9423 (can_duplicate_and_interleave_p): New function.
9424 (vect_get_and_check_slp_defs): Take the vector of statements
9425 rather than just the current one. Remove excess parentheses.
9426 Restriction rejectinon of vect_constant_def and vect_external_def
9427 for variable-length vectors to boolean types, or types for which
9428 can_duplicate_and_interleave_p is false.
9429 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9430 (duplicate_and_interleave): New function.
9431 (vect_get_constant_vectors): Use gimple_build_vector for
9432 constant-length vectors and suitable variable-length constant
9433 vectors. Use duplicate_and_interleave for other variable-length
9434 vectors. Don't defer the update when inserting new statements.
9436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9437 Alan Hayward <alan.hayward@arm.com>
9438 David Sherwood <david.sherwood@arm.com>
9440 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9441 min_profitable_iters doesn't go negative.
9443 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9444 Alan Hayward <alan.hayward@arm.com>
9445 David Sherwood <david.sherwood@arm.com>
9447 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9448 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9449 * optabs.def (vec_mask_load_lanes_optab): New optab.
9450 (vec_mask_store_lanes_optab): Likewise.
9451 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9452 (MASK_STORE_LANES): Likewise.
9453 * internal-fn.c (mask_load_lanes_direct): New macro.
9454 (mask_store_lanes_direct): Likewise.
9455 (expand_mask_load_optab_fn): Handle masked operations.
9456 (expand_mask_load_lanes_optab_fn): New macro.
9457 (expand_mask_store_optab_fn): Handle masked operations.
9458 (expand_mask_store_lanes_optab_fn): New macro.
9459 (direct_mask_load_lanes_optab_supported_p): Likewise.
9460 (direct_mask_store_lanes_optab_supported_p): Likewise.
9461 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9463 (vect_load_lanes_supported): Likewise.
9464 * tree-vect-data-refs.c (strip_conversion): New function.
9465 (can_group_stmts_p): Likewise.
9466 (vect_analyze_data_ref_accesses): Use it instead of checking
9467 for a pair of assignments.
9468 (vect_store_lanes_supported): Take a masked_p parameter.
9469 (vect_load_lanes_supported): Likewise.
9470 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9471 vect_store_lanes_supported and vect_load_lanes_supported.
9472 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9473 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9474 parameter. Don't allow gaps for masked accesses.
9475 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9476 and vect_load_lanes_supported.
9477 (get_load_store_type): Take a masked_p parameter and update
9478 call to get_group_load_store_type.
9479 (vectorizable_store): Update call to get_load_store_type.
9480 Handle IFN_MASK_STORE_LANES.
9481 (vectorizable_load): Update call to get_load_store_type.
9482 Handle IFN_MASK_LOAD_LANES.
9484 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9485 Alan Hayward <alan.hayward@arm.com>
9486 David Sherwood <david.sherwood@arm.com>
9488 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9490 * config/aarch64/aarch64-protos.h
9491 (aarch64_sve_struct_memory_operand_p): Declare.
9492 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9493 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9494 (VPRED, vpred): Handle SVE structure modes.
9495 * config/aarch64/constraints.md (Utx): New constraint.
9496 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9497 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9498 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9499 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9500 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9501 structure modes. Split into pieces after RA.
9502 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9503 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9505 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9506 SVE structure modes.
9507 (aarch64_classify_address): Likewise.
9508 (sizetochar): Move earlier in file.
9509 (aarch64_print_operand): Handle SVE register lists.
9510 (aarch64_array_mode): New function.
9511 (aarch64_sve_struct_memory_operand_p): Likewise.
9512 (TARGET_ARRAY_MODE): Redefine.
9514 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9515 Alan Hayward <alan.hayward@arm.com>
9516 David Sherwood <david.sherwood@arm.com>
9518 * target.def (array_mode): New target hook.
9519 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9520 * doc/tm.texi: Regenerate.
9521 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9522 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9523 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9525 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9528 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9529 Alan Hayward <alan.hayward@arm.com>
9530 David Sherwood <david.sherwood@arm.com>
9532 * fold-const.c (fold_binary_loc): Check the argument types
9533 rather than the result type when testing for a vector operation.
9535 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9537 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9538 * doc/tm.texi: Regenerate.
9540 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9541 Alan Hayward <alan.hayward@arm.com>
9542 David Sherwood <david.sherwood@arm.com>
9544 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9545 (sve): Document new AArch64 extension.
9546 * doc/md.texi (w): Extend the description of the AArch64
9547 constraint to include SVE vectors.
9548 (Upl, Upa): Document new AArch64 predicate constraints.
9549 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9551 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9552 (msve-vector-bits=): New option.
9553 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9554 SVE when these are disabled.
9555 (sve): New extension.
9556 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9557 modes. Adjust their number of units based on aarch64_sve_vg.
9558 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9559 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9560 aarch64_addr_query_type.
9561 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9562 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9563 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9564 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9565 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9566 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9567 (aarch64_simd_imm_zero_p): Delete.
9568 (aarch64_check_zero_based_sve_index_immediate): Declare.
9569 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9570 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9571 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9572 (aarch64_sve_float_mul_immediate_p): Likewise.
9573 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9575 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9576 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9577 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9578 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9579 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9580 (aarch64_regmode_natural_size): Likewise.
9581 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9582 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9584 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9585 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9586 for VG and the SVE predicate registers.
9587 (V_ALIASES): Add a "z"-prefixed alias.
9588 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9589 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9590 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9591 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9592 (REG_CLASS_NAMES): Add entries for them.
9593 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9594 and the predicate registers.
9595 (aarch64_sve_vg): Declare.
9596 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9597 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9598 (REGMODE_NATURAL_SIZE): Define.
9599 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9601 * config/aarch64/aarch64.c: Include cfgrtl.h.
9602 (simd_immediate_info): Add a constructor for series vectors,
9603 and an associated step field.
9604 (aarch64_sve_vg): New variable.
9605 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9606 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9607 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9608 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9609 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9610 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9611 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9612 (aarch64_get_mask_mode): New functions.
9613 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9614 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9615 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9616 predicate modes and predicate registers. Explicitly restrict
9617 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9618 to store a vector mode if it is recognized by
9619 aarch64_classify_vector_mode.
9620 (aarch64_regmode_natural_size): New function.
9621 (aarch64_hard_regno_caller_save_mode): Return the original mode
9623 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9624 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9625 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9626 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9628 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9629 does not overlap dest if the function is frame-related. Handle
9631 (aarch64_split_add_offset): New function.
9632 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9633 them aarch64_add_offset.
9634 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9635 and update call to aarch64_sub_sp.
9636 (aarch64_add_cfa_expression): New function.
9637 (aarch64_expand_prologue): Pass extra temporary registers to the
9638 functions above. Handle the case in which we need to emit new
9639 DW_CFA_expressions for registers that were originally saved
9640 relative to the stack pointer, but now have to be expressed
9641 relative to the frame pointer.
9642 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9644 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9645 IP0 and IP1 values for SVE frames.
9646 (aarch64_expand_vec_series): New function.
9647 (aarch64_expand_sve_widened_duplicate): Likewise.
9648 (aarch64_expand_sve_const_vector): Likewise.
9649 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9650 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9651 into the register, rather than emitting a SET directly.
9652 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9653 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9654 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9655 (offset_9bit_signed_scaled_p): New functions.
9656 (aarch64_replicate_bitmask_imm): New function.
9657 (aarch64_bitmask_imm): Use it.
9658 (aarch64_cannot_force_const_mem): Reject expressions involving
9659 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9660 (aarch64_classify_index): Handle SVE indices, by requiring
9661 a plain register index with a scale that matches the element size.
9662 (aarch64_classify_address): Handle SVE addresses. Assert that
9663 the mode of the address is VOIDmode or an integer mode.
9664 Update call to aarch64_classify_symbol.
9665 (aarch64_classify_symbolic_expression): Update call to
9666 aarch64_classify_symbol.
9667 (aarch64_const_vec_all_in_range_p): New function.
9668 (aarch64_print_vector_float_operand): Likewise.
9669 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9670 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9671 and the FP immediates 1.0 and 0.5.
9672 (aarch64_print_address_internal): Handle SVE addresses.
9673 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9674 (aarch64_regno_regclass): Handle predicate registers.
9675 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9677 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9678 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9679 (aarch64_convert_sve_vector_bits): New function.
9680 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9681 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9683 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9684 Handle SVE vector and predicate modes. Accept VL-based constants
9685 that need only one temporary register, and VL offsets that require
9686 no temporary registers.
9687 (aarch64_conditional_register_usage): Mark the predicate registers
9688 as fixed if SVE isn't available.
9689 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9690 Return true for SVE vector and predicate modes.
9691 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9692 rather than an unsigned int. Handle SVE modes.
9693 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9695 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9697 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9698 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9699 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9700 (aarch64_sve_float_mul_immediate_p): New functions.
9701 (aarch64_sve_valid_immediate): New function.
9702 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9703 Explicitly reject structure modes. Check for INDEX constants.
9704 Handle PTRUE and PFALSE constants.
9705 (aarch64_check_zero_based_sve_index_immediate): New function.
9706 (aarch64_simd_imm_zero_p): Delete.
9707 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9708 vector modes. Accept constants in the range of CNT[BHWD].
9709 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9710 ask for an Advanced SIMD mode.
9711 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9712 (aarch64_simd_vector_alignment): Handle SVE predicates.
9713 (aarch64_vectorize_preferred_vector_alignment): New function.
9714 (aarch64_simd_vector_alignment_reachable): Use it instead of
9716 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9717 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9719 (MAX_VECT_LEN): Delete.
9720 (expand_vec_perm_d): Add a vec_flags field.
9721 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9722 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9723 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9725 (aarch64_evpc_rev): Rename to...
9726 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9727 (aarch64_evpc_rev_global): New function.
9728 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9729 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9731 (aarch64_evpc_sve_tbl): New function.
9732 (aarch64_expand_vec_perm_const_1): Update after rename of
9733 aarch64_evpc_rev. Handle SVE permutes too, trying
9734 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9735 than aarch64_evpc_tbl.
9736 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9737 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9738 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9739 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9740 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9741 (aarch64_expand_sve_vcond): New functions.
9742 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9743 of aarch64_vector_mode_p.
9744 (aarch64_dwarf_poly_indeterminate_value): New function.
9745 (aarch64_compute_pressure_classes): Likewise.
9746 (aarch64_can_change_mode_class): Likewise.
9747 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9748 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9749 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9750 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9751 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9752 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9753 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9754 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9756 (Dn, Dl, Dr): Accept const as well as const_vector.
9757 (Dz): Likewise. Compare against CONST0_RTX.
9758 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9759 of "vector" where appropriate.
9760 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9761 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9762 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9763 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9764 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9765 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9766 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9767 (v_int_equiv): Extend to SVE modes.
9768 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9770 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9771 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9772 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9773 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9774 (SVE_COND_FP_CMP): New int iterators.
9775 (perm_hilo): Handle the new unpack unspecs.
9776 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9778 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9779 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9780 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9781 (aarch64_equality_operator, aarch64_constant_vector_operand)
9782 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9783 (aarch64_sve_nonimmediate_operand): Likewise.
9784 (aarch64_sve_general_operand): Likewise.
9785 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9786 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9787 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9788 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9789 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9790 (aarch64_sve_float_arith_immediate): Likewise.
9791 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9792 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9793 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9794 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9795 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9796 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9797 (aarch64_sve_float_arith_operand): Likewise.
9798 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9799 (aarch64_sve_float_mul_operand): Likewise.
9800 (aarch64_sve_vec_perm_operand): Likewise.
9801 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9802 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9803 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9804 as well as const_vector.
9805 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9806 in file. Use CONST0_RTX and CONSTM1_RTX.
9807 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9808 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9809 Use aarch64_simd_imm_zero.
9810 * config/aarch64/aarch64-sve.md: New file.
9811 * config/aarch64/aarch64.md: Include it.
9812 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9813 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9814 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9815 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9816 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9817 (sve): New attribute.
9818 (enabled): Disable instructions with the sve attribute unless
9820 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9821 aarch64_expand_mov_immediate.
9822 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9823 CNT[BHSD] immediates.
9824 (movti): Split CONST_POLY_INT moves into two halves.
9825 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9826 Split additions that need a temporary here if the destination
9827 is the stack pointer.
9828 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9829 (*add<mode>3_poly_1): New instruction.
9830 (set_clobber_cc): New expander.
9832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9834 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9835 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9836 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9837 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9838 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9839 Change innermode from fixed_mode_size to machine_mode.
9840 (simplify_subreg): Update call accordingly. Handle a constant-sized
9841 subreg of a variable-length CONST_VECTOR.
9843 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9844 Alan Hayward <alan.hayward@arm.com>
9845 David Sherwood <david.sherwood@arm.com>
9847 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9848 (add_offset_to_base): New function, split out from...
9849 (create_mem_ref): ...here. When handling a scale other than 1,
9850 check first whether the address is valid without the offset.
9851 Add it into the base if so, leaving the index and scale as-is.
9853 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9856 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9857 fold_for_warn before checking if arg2 is INTEGER_CST.
9859 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9861 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9862 (store_multiple_operation): Delete.
9863 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9864 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9865 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9866 guarded by TARGET_STRING.
9867 (rs6000_output_load_multiple): Delete.
9868 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9869 OPTION_MASK_STRING / TARGET_STRING handling.
9870 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9871 (const rs6000_opt_masks) <"string">: Change mask to 0.
9872 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9873 (MASK_STRING): Delete.
9874 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9876 (load_multiple): Delete.
9883 (store_multiple): Delete.
9890 (movmemsi_8reg): Delete.
9891 (corresponding unnamed define_insn): Delete.
9892 (movmemsi_6reg): Delete.
9893 (corresponding unnamed define_insn): Delete.
9894 (movmemsi_4reg): Delete.
9895 (corresponding unnamed define_insn): Delete.
9896 (movmemsi_2reg): Delete.
9897 (corresponding unnamed define_insn): Delete.
9898 (movmemsi_1reg): Delete.
9899 (corresponding unnamed define_insn): Delete.
9900 * config/rs6000/rs6000.opt (mno-string): New.
9901 (mstring): Replace by deprecation warning stub.
9902 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9904 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9906 * regrename.c (regrename_do_replace): If replacing the same
9907 reg multiple times, try to reuse last created gen_raw_REG.
9910 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9911 main to workaround a bug in GDB.
9913 2018-01-12 Tom de Vries <tom@codesourcery.com>
9916 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9918 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9920 PR rtl-optimization/80481
9921 * ira-color.c (get_cap_member): New function.
9922 (allocnos_conflict_by_live_ranges_p): Use it.
9923 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9924 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9926 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9929 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9930 (*saddl_se_1): Ditto.
9932 (*ssubl_se_1): Ditto.
9934 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9936 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9937 rather than wi::to_widest for DR_INITs.
9938 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9939 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9940 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9942 (vect_analyze_group_access_1): Note that here.
9944 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9946 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9947 polynomial type sizes.
9949 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9951 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9952 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9953 (gimple_add_tmp_var): Likewise.
9955 2018-01-12 Martin Liska <mliska@suse.cz>
9957 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9958 (gimple_alloc_sizes): Likewise.
9959 (dump_gimple_statistics): Use PRIu64 in printf format.
9960 * gimple.h: Change uint64_t to int.
9962 2018-01-12 Martin Liska <mliska@suse.cz>
9964 * tree-core.h: Use uint64_t instead of int.
9965 * tree.c (tree_node_counts): Likewise.
9966 (tree_node_sizes): Likewise.
9967 (dump_tree_statistics): Use PRIu64 in printf format.
9969 2018-01-12 Martin Liska <mliska@suse.cz>
9971 * Makefile.in: As qsort_chk is implemented in vec.c, add
9972 vec.o to linkage of gencfn-macros.
9973 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9974 passing the info to record_node_allocation_statistics.
9975 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9977 * ggc-common.c (struct ggc_usage): Add operator== and use
9978 it in operator< and compare function.
9979 * mem-stats.h (struct mem_usage): Likewise.
9980 * vec.c (struct vec_usage): Remove operator< and compare
9981 function. Can be simply inherited.
9983 2018-01-12 Martin Jambor <mjambor@suse.cz>
9986 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9987 * tree-ssa-math-opts.c: Include domwalk.h.
9988 (convert_mult_to_fma_1): New function.
9989 (fma_transformation_info): New type.
9990 (fma_deferring_state): Likewise.
9991 (cancel_fma_deferring): New function.
9992 (result_of_phi): Likewise.
9993 (last_fma_candidate_feeds_initial_phi): Likewise.
9994 (convert_mult_to_fma): Added deferring logic, split actual
9995 transformation to convert_mult_to_fma_1.
9996 (math_opts_dom_walker): New type.
9997 (math_opts_dom_walker::after_dom_children): New method, body moved
9998 here from pass_optimize_widening_mul::execute, added deferring logic
10000 (pass_optimize_widening_mul::execute): Moved most of code to
10001 math_opts_dom_walker::after_dom_children.
10002 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10003 * config/i386/i386.c (ix86_option_override_internal): Added
10004 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10006 2018-01-12 Richard Biener <rguenther@suse.de>
10009 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10010 inline instance vars.
10012 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10015 * config/rx/rx.c (rx_is_restricted_memory_address):
10016 Handle SUBREG case.
10018 2018-01-12 Richard Biener <rguenther@suse.de>
10020 PR tree-optimization/80846
10021 * target.def (split_reduction): New target hook.
10022 * targhooks.c (default_split_reduction): New function.
10023 * targhooks.h (default_split_reduction): Declare.
10024 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10025 target requests first reduce vectors by combining low and high
10027 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10028 (get_vectype_for_scalar_type_and_size): Export.
10029 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10030 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10031 * doc/tm.texi: Regenerate.
10032 * config/i386/i386.c (ix86_split_reduction): Implement
10033 TARGET_VECTORIZE_SPLIT_REDUCTION.
10035 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10038 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10039 in PIC mode except for TARGET_VXWORKS_RTP.
10040 * config/sparc/sparc.c: Include cfgrtl.h.
10041 (TARGET_INIT_PIC_REG): Define.
10042 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10043 (sparc_pic_register_p): New predicate.
10044 (sparc_legitimate_address_p): Use it.
10045 (sparc_legitimize_pic_address): Likewise.
10046 (sparc_delegitimize_address): Likewise.
10047 (sparc_mode_dependent_address_p): Likewise.
10048 (gen_load_pcrel_sym): Remove 4th parameter.
10049 (load_got_register): Adjust call to above. Remove obsolete stuff.
10050 (sparc_expand_prologue): Do not call load_got_register here.
10051 (sparc_flat_expand_prologue): Likewise.
10052 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10053 (sparc_use_pseudo_pic_reg): New function.
10054 (sparc_init_pic_reg): Likewise.
10055 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10056 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10058 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10060 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10061 Add item for branch_cost.
10063 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10065 PR rtl-optimization/83565
10066 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10067 not extend the result to a larger mode for rotate operations.
10068 (num_sign_bit_copies1): Likewise.
10070 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10073 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10075 Use values-Xc.o for -pedantic.
10076 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10078 2018-01-12 Martin Liska <mliska@suse.cz>
10081 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10083 (possible_polymorphic_call_targets): Use it.
10084 (ipa_devirt): Likewise.
10086 2018-01-12 Martin Liska <mliska@suse.cz>
10088 * profile-count.h (enum profile_quality): Use 0 as invalid
10089 enum value of profile_quality.
10091 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10093 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10094 -mext-string options.
10096 2018-01-12 Richard Biener <rguenther@suse.de>
10098 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10099 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10100 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10102 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10104 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10106 * configure.ac (--with-long-double-format): Add support for the
10107 configuration option to change the default long double format on
10109 * config.gcc (powerpc*-linux*-*): Likewise.
10110 * configure: Regenerate.
10111 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10112 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10113 used without modification.
10115 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10117 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10118 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10119 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10120 MISC_BUILTIN_SPEC_BARRIER.
10121 (rs6000_init_builtins): Likewise.
10122 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10124 (speculation_barrier): New define_insn.
10125 * doc/extend.texi: Document __builtin_speculation_barrier.
10127 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10130 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10131 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10132 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10134 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10135 integral modes instead of "ss" and "sd".
10136 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10137 vectors with 32-bit and 64-bit elements.
10138 (vecdupssescalarmodesuffix): New mode attribute.
10139 (vec_dup<mode>): Use it.
10141 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10144 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10145 frame if argument is passed on stack.
10147 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10150 * ree.c (combine_reaching_defs): Optimize also
10151 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10152 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10154 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10156 PR middle-end/83189
10157 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10159 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10161 PR middle-end/83718
10162 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10163 after they are computed.
10165 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10167 PR tree-optimization/83695
10168 * gimple-loop-linterchange.cc
10169 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10170 reset cached scev information after interchange.
10171 (pass_linterchange::execute): Remove call to scev_reset_htab.
10173 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10175 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10176 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10177 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10178 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10179 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10180 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10181 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10182 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10183 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10184 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10185 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10186 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10187 (V_lane_reg): Likewise.
10188 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10190 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10191 (vfmal_lane_low<mode>_intrinsic,
10192 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10193 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10194 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10195 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10196 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10197 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10199 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10201 * config/arm/arm-cpus.in (fp16fml): New feature.
10202 (ALL_SIMD): Add fp16fml.
10203 (armv8.2-a): Add fp16fml as an option.
10204 (armv8.3-a): Likewise.
10205 (armv8.4-a): Add fp16fml as part of fp16.
10206 * config/arm/arm.h (TARGET_FP16FML): Define.
10207 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10209 * config/arm/arm-modes.def (V2HF): Define.
10210 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10211 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10212 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10213 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10214 vfmsl_low, vfmsl_high): New set of builtins.
10215 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10216 (vfml_op): New code attribute.
10217 (VFMLHALVES): New int iterator.
10218 (VFML, VFMLSEL): New mode attributes.
10219 (V_reg): Define mapping for V2HF.
10220 (V_hi, V_lo): New mode attributes.
10221 (VF_constraint): Likewise.
10222 (vfml_half, vfml_half_selector): New int attributes.
10223 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10225 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10226 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10228 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10229 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10230 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10231 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10233 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10234 Document new effective target and option set.
10236 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10238 * config/arm/arm-cpus.in (armv8_4): New feature.
10239 (ARMv8_4a): New fgroup.
10240 (armv8.4-a): New arch.
10241 * config/arm/arm-tables.opt: Regenerate.
10242 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10243 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10244 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10245 Add matching rules for -march=armv8.4-a and extensions.
10246 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10248 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10251 * config/rx/rx.md (BW): New mode attribute.
10252 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10254 2018-01-11 Richard Biener <rguenther@suse.de>
10256 PR tree-optimization/83435
10257 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10258 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10259 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10261 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10262 Alan Hayward <alan.hayward@arm.com>
10263 David Sherwood <david.sherwood@arm.com>
10265 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10267 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10268 (aarch64_print_address_internal): Use it to check for a zero offset.
10270 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10271 Alan Hayward <alan.hayward@arm.com>
10272 David Sherwood <david.sherwood@arm.com>
10274 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10275 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10276 Return a poly_int64 rather than a HOST_WIDE_INT.
10277 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10278 rather than a HOST_WIDE_INT.
10279 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10280 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10281 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10282 final_offset from HOST_WIDE_INT to poly_int64.
10283 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10284 to_constant when getting the number of units in an Advanced SIMD
10286 (aarch64_builtin_vectorized_function): Check for a constant number
10288 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10290 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10291 attribute instead of GET_MODE_NUNITS.
10292 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10293 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10294 GET_MODE_SIZE for fixed-size registers.
10295 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10296 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10297 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10298 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10299 (aarch64_print_operand, aarch64_print_address_internal)
10300 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10301 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10302 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10303 Handle polynomial GET_MODE_SIZE.
10304 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10305 wider than SImode without modification.
10306 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10307 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10308 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10309 passing and returning SVE modes.
10310 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10311 rather than GEN_INT.
10312 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10313 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10314 (aarch64_allocate_and_probe_stack_space): Likewise.
10315 (aarch64_layout_frame): Cope with polynomial offsets.
10316 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10317 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10318 polynomial offsets.
10319 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10320 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10321 poly_int64 rather than a HOST_WIDE_INT.
10322 (aarch64_get_separate_components, aarch64_process_components)
10323 (aarch64_expand_prologue, aarch64_expand_epilogue)
10324 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10325 (aarch64_anchor_offset): New function, split out from...
10326 (aarch64_legitimize_address): ...here.
10327 (aarch64_builtin_vectorization_cost): Handle polynomial
10328 TYPE_VECTOR_SUBPARTS.
10329 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10331 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10332 number of elements from the PARALLEL rather than the mode.
10333 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10334 rather than GET_MODE_BITSIZE.
10335 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10336 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10337 (aarch64_expand_vec_perm_const_1): Handle polynomial
10338 d->perm.length () and d->perm elements.
10339 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10340 Apply to_constant to d->perm elements.
10341 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10342 polynomial CONST_VECTOR_NUNITS.
10343 (aarch64_move_pointer): Take amount as a poly_int64 rather
10345 (aarch64_progress_pointer): Avoid temporary variable.
10346 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10347 the mode attribute instead of GET_MODE.
10349 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10350 Alan Hayward <alan.hayward@arm.com>
10351 David Sherwood <david.sherwood@arm.com>
10353 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10354 x exists before using it.
10355 (aarch64_add_constant_internal): Rename to...
10356 (aarch64_add_offset_1): ...this. Replace regnum with separate
10357 src and dest rtxes. Handle the case in which they're different,
10358 including when the offset is zero. Replace scratchreg with an rtx.
10359 Use 2 additions if there is no spare register into which we can
10360 move a 16-bit constant.
10361 (aarch64_add_constant): Delete.
10362 (aarch64_add_offset): Replace reg with separate src and dest
10363 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10364 Use aarch64_add_offset_1.
10365 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10366 an rtx rather than an int. Take the delta as a poly_int64
10367 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10368 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10369 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10370 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10371 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10372 and aarch64_add_sp.
10373 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10374 aarch64_add_constant.
10376 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10378 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10379 Use scalar_float_mode.
10381 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10383 * config/aarch64/aarch64-simd.md
10384 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10385 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10386 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10387 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10388 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10389 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10390 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10391 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10392 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10393 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10395 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10398 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10399 targ_options->x_arm_arch_string is non NULL.
10401 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10403 * config/aarch64/aarch64.h
10404 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10406 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10409 * expmed.c (emit_store_flag_force): Swap if const op0
10410 and change VOIDmode to mode of op0.
10412 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10414 PR rtl-optimization/83761
10415 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10416 than bytes to mode_for_size.
10418 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10420 PR middle-end/83189
10421 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10422 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10425 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10427 PR middle-end/83575
10428 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10429 when in layout mode.
10430 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10431 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10434 2018-01-10 Michael Collison <michael.collison@arm.com>
10436 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10437 * config/aarch64/aarch64-option-extension.def: Add
10438 AARCH64_OPT_EXTENSION of 'fp16fml'.
10439 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10440 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10441 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10442 * config/aarch64/constraints.md (Ui7): New constraint.
10443 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10444 (VFMLA_SEL_W): Ditto.
10447 (VFMLA16_LOW): New int iterator.
10448 (VFMLA16_HIGH): Ditto.
10449 (UNSPEC_FMLAL): New unspec.
10450 (UNSPEC_FMLSL): Ditto.
10451 (UNSPEC_FMLAL2): Ditto.
10452 (UNSPEC_FMLSL2): Ditto.
10453 (f16mac): New code attribute.
10454 * config/aarch64/aarch64-simd-builtins.def
10455 (aarch64_fmlal_lowv2sf): Ditto.
10456 (aarch64_fmlsl_lowv2sf): Ditto.
10457 (aarch64_fmlalq_lowv4sf): Ditto.
10458 (aarch64_fmlslq_lowv4sf): Ditto.
10459 (aarch64_fmlal_highv2sf): Ditto.
10460 (aarch64_fmlsl_highv2sf): Ditto.
10461 (aarch64_fmlalq_highv4sf): Ditto.
10462 (aarch64_fmlslq_highv4sf): Ditto.
10463 (aarch64_fmlal_lane_lowv2sf): Ditto.
10464 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10465 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10466 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10467 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10468 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10469 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10470 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10471 (aarch64_fmlal_lane_highv2sf): Ditto.
10472 (aarch64_fmlsl_lane_highv2sf): Ditto.
10473 (aarch64_fmlal_laneq_highv2sf): Ditto.
10474 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10475 (aarch64_fmlalq_lane_highv4sf): Ditto.
10476 (aarch64_fmlsl_lane_highv4sf): Ditto.
10477 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10478 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10479 * config/aarch64/aarch64-simd.md:
10480 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10481 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10482 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10483 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10484 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10485 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10486 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10487 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10488 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10489 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10490 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10491 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10492 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10493 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10494 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10495 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10496 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10497 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10498 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10499 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10500 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10501 (vfmlsl_low_u32): Ditto.
10502 (vfmlalq_low_u32): Ditto.
10503 (vfmlslq_low_u32): Ditto.
10504 (vfmlal_high_u32): Ditto.
10505 (vfmlsl_high_u32): Ditto.
10506 (vfmlalq_high_u32): Ditto.
10507 (vfmlslq_high_u32): Ditto.
10508 (vfmlal_lane_low_u32): Ditto.
10509 (vfmlsl_lane_low_u32): Ditto.
10510 (vfmlal_laneq_low_u32): Ditto.
10511 (vfmlsl_laneq_low_u32): Ditto.
10512 (vfmlalq_lane_low_u32): Ditto.
10513 (vfmlslq_lane_low_u32): Ditto.
10514 (vfmlalq_laneq_low_u32): Ditto.
10515 (vfmlslq_laneq_low_u32): Ditto.
10516 (vfmlal_lane_high_u32): Ditto.
10517 (vfmlsl_lane_high_u32): Ditto.
10518 (vfmlal_laneq_high_u32): Ditto.
10519 (vfmlsl_laneq_high_u32): Ditto.
10520 (vfmlalq_lane_high_u32): Ditto.
10521 (vfmlslq_lane_high_u32): Ditto.
10522 (vfmlalq_laneq_high_u32): Ditto.
10523 (vfmlslq_laneq_high_u32): Ditto.
10524 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10525 (AARCH64_FL_FOR_ARCH8_4): New.
10526 (AARCH64_ISA_F16FML): New ISA flag.
10527 (TARGET_F16FML): New feature flag for fp16fml.
10528 (doc/invoke.texi): Document new fp16fml option.
10530 2018-01-10 Michael Collison <michael.collison@arm.com>
10532 * config/aarch64/aarch64-builtins.c:
10533 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10534 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10535 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10536 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10537 (AARCH64_ISA_SHA3): New ISA flag.
10538 (TARGET_SHA3): New feature flag for sha3.
10539 * config/aarch64/iterators.md (sha512_op): New int attribute.
10540 (CRYPTO_SHA512): New int iterator.
10541 (UNSPEC_SHA512H): New unspec.
10542 (UNSPEC_SHA512H2): Ditto.
10543 (UNSPEC_SHA512SU0): Ditto.
10544 (UNSPEC_SHA512SU1): Ditto.
10545 * config/aarch64/aarch64-simd-builtins.def
10546 (aarch64_crypto_sha512hqv2di): New builtin.
10547 (aarch64_crypto_sha512h2qv2di): Ditto.
10548 (aarch64_crypto_sha512su0qv2di): Ditto.
10549 (aarch64_crypto_sha512su1qv2di): Ditto.
10550 (aarch64_eor3qv8hi): Ditto.
10551 (aarch64_rax1qv2di): Ditto.
10552 (aarch64_xarqv2di): Ditto.
10553 (aarch64_bcaxqv8hi): Ditto.
10554 * config/aarch64/aarch64-simd.md:
10555 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10556 (aarch64_crypto_sha512su0qv2di): Ditto.
10557 (aarch64_crypto_sha512su1qv2di): Ditto.
10558 (aarch64_eor3qv8hi): Ditto.
10559 (aarch64_rax1qv2di): Ditto.
10560 (aarch64_xarqv2di): Ditto.
10561 (aarch64_bcaxqv8hi): Ditto.
10562 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10563 (vsha512h2q_u64): Ditto.
10564 (vsha512su0q_u64): Ditto.
10565 (vsha512su1q_u64): Ditto.
10566 (veor3q_u16): Ditto.
10567 (vrax1q_u64): Ditto.
10568 (vxarq_u64): Ditto.
10569 (vbcaxq_u16): Ditto.
10570 * config/arm/types.md (crypto_sha512): New type attribute.
10571 (crypto_sha3): Ditto.
10572 (doc/invoke.texi): Document new sha3 option.
10574 2018-01-10 Michael Collison <michael.collison@arm.com>
10576 * config/aarch64/aarch64-builtins.c:
10577 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10578 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10579 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10580 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10581 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10582 (AARCH64_ISA_SM4): New ISA flag.
10583 (TARGET_SM4): New feature flag for sm4.
10584 * config/aarch64/aarch64-simd-builtins.def
10585 (aarch64_sm3ss1qv4si): Ditto.
10586 (aarch64_sm3tt1aq4si): Ditto.
10587 (aarch64_sm3tt1bq4si): Ditto.
10588 (aarch64_sm3tt2aq4si): Ditto.
10589 (aarch64_sm3tt2bq4si): Ditto.
10590 (aarch64_sm3partw1qv4si): Ditto.
10591 (aarch64_sm3partw2qv4si): Ditto.
10592 (aarch64_sm4eqv4si): Ditto.
10593 (aarch64_sm4ekeyqv4si): Ditto.
10594 * config/aarch64/aarch64-simd.md:
10595 (aarch64_sm3ss1qv4si): Ditto.
10596 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10597 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10598 (aarch64_sm4eqv4si): Ditto.
10599 (aarch64_sm4ekeyqv4si): Ditto.
10600 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10601 (sm3part_op): Ditto.
10602 (CRYPTO_SM3TT): Ditto.
10603 (CRYPTO_SM3PART): Ditto.
10604 (UNSPEC_SM3SS1): New unspec.
10605 (UNSPEC_SM3TT1A): Ditto.
10606 (UNSPEC_SM3TT1B): Ditto.
10607 (UNSPEC_SM3TT2A): Ditto.
10608 (UNSPEC_SM3TT2B): Ditto.
10609 (UNSPEC_SM3PARTW1): Ditto.
10610 (UNSPEC_SM3PARTW2): Ditto.
10611 (UNSPEC_SM4E): Ditto.
10612 (UNSPEC_SM4EKEY): Ditto.
10613 * config/aarch64/constraints.md (Ui2): New constraint.
10614 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10615 * config/arm/types.md (crypto_sm3): New type attribute.
10616 (crypto_sm4): Ditto.
10617 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10618 (vsm3tt1aq_u32): Ditto.
10619 (vsm3tt1bq_u32): Ditto.
10620 (vsm3tt2aq_u32): Ditto.
10621 (vsm3tt2bq_u32): Ditto.
10622 (vsm3partw1q_u32): Ditto.
10623 (vsm3partw2q_u32): Ditto.
10624 (vsm4eq_u32): Ditto.
10625 (vsm4ekeyq_u32): Ditto.
10626 (doc/invoke.texi): Document new sm4 option.
10628 2018-01-10 Michael Collison <michael.collison@arm.com>
10630 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10631 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10632 (AARCH64_FL_FOR_ARCH8_4): New.
10633 (AARCH64_FL_V8_4): New flag.
10634 (doc/invoke.texi): Document new armv8.4-a option.
10636 2018-01-10 Michael Collison <michael.collison@arm.com>
10638 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10639 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10640 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10641 * config/aarch64/aarch64-option-extension.def: Add
10642 AARCH64_OPT_EXTENSION of 'sha2'.
10643 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10644 (crypto): Disable sha2 and aes if crypto disabled.
10645 (crypto): Enable aes and sha2 if enabled.
10646 (simd): Disable sha2 and aes if simd disabled.
10647 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10649 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10650 (TARGET_SHA2): New feature flag for sha2.
10651 (TARGET_AES): New feature flag for aes.
10652 * config/aarch64/aarch64-simd.md:
10653 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10654 conditional on TARGET_AES.
10655 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10656 (aarch64_crypto_sha1hsi): Make pattern conditional
10658 (aarch64_crypto_sha1hv4si): Ditto.
10659 (aarch64_be_crypto_sha1hv4si): Ditto.
10660 (aarch64_crypto_sha1su1v4si): Ditto.
10661 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10662 (aarch64_crypto_sha1su0v4si): Ditto.
10663 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10664 (aarch64_crypto_sha256su0v4si): Ditto.
10665 (aarch64_crypto_sha256su1v4si): Ditto.
10666 (doc/invoke.texi): Document new aes and sha2 options.
10668 2018-01-10 Martin Sebor <msebor@redhat.com>
10670 PR tree-optimization/83781
10671 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10674 2018-01-11 Martin Sebor <msebor@gmail.com>
10675 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10677 PR tree-optimization/83501
10678 PR tree-optimization/81703
10680 * tree-ssa-strlen.c (get_string_cst): Rename...
10681 (get_string_len): ...to this. Handle global constants.
10682 (handle_char_store): Adjust.
10684 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10685 Jim Wilson <jimw@sifive.com>
10687 * config/riscv/riscv-protos.h (riscv_output_return): New.
10688 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10689 (riscv_attribute_table, riscv_output_return),
10690 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10691 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10692 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10694 (riscv_expand_prologue): Add early return for naked function.
10695 (riscv_expand_epilogue): Likewise.
10696 (riscv_function_ok_for_sibcall): Return false for naked function.
10697 (riscv_set_current_function): New.
10698 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10699 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10700 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10701 * doc/extend.texi (RISC-V Function Attributes): New.
10703 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10705 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10706 check for 128-bit long double before checking TCmode.
10707 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10708 128-bit long doubles before checking TFmode or TCmode.
10709 (FLOAT128_IBM_P): Likewise.
10711 2018-01-10 Martin Sebor <msebor@redhat.com>
10713 PR tree-optimization/83671
10714 * builtins.c (c_strlen): Unconditionally return zero for the empty
10716 Use -Warray-bounds for warnings.
10717 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10718 for non-constant array indices with COMPONENT_REF, arrays of
10719 arrays, and pointers to arrays.
10720 (gimple_fold_builtin_strlen): Determine and set length range for
10721 non-constant character arrays.
10723 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10725 PR middle-end/81897
10726 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10729 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10731 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10733 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10736 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10737 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10738 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10739 indexed_or_indirect_operand predicate.
10740 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10741 (*vsx_le_perm_load_v8hi): Likewise.
10742 (*vsx_le_perm_load_v16qi): Likewise.
10743 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10744 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10745 (*vsx_le_perm_store_v8hi): Likewise.
10746 (*vsx_le_perm_store_v16qi): Likewise.
10747 (eight unnamed splitters): Likewise.
10749 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10751 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10752 * config/rs6000/emmintrin.h: Likewise.
10753 * config/rs6000/mmintrin.h: Likewise.
10754 * config/rs6000/xmmintrin.h: Likewise.
10756 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10759 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10761 * tree.c (tree_nop_conversion): Return true for location wrapper
10763 (maybe_wrap_with_location): New function.
10764 (selftest::check_strip_nops): New function.
10765 (selftest::test_location_wrappers): New function.
10766 (selftest::tree_c_tests): Call it.
10767 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10768 (maybe_wrap_with_location): New decl.
10769 (EXPR_LOCATION_WRAPPER_P): New macro.
10770 (location_wrapper_p): New inline function.
10771 (tree_strip_any_location_wrapper): New inline function.
10773 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10776 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10777 stack_realign_offset for the largest alignment of stack slot
10779 (ix86_find_max_used_stack_alignment): New function.
10780 (ix86_finalize_stack_frame_flags): Use it. Set
10781 max_used_stack_alignment if we don't realign stack.
10782 * config/i386/i386.h (machine_function): Add
10783 max_used_stack_alignment.
10785 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10787 * config/arm/arm.opt (-mbranch-cost): New option.
10788 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10791 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10794 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10795 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10797 2018-01-10 Richard Biener <rguenther@suse.de>
10800 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10801 early out so it also covers the case where we have a non-NULL
10804 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10806 PR tree-optimization/83753
10807 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10808 for non-strided grouped accesses if the number of elements is 1.
10810 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10813 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10814 * i386.h (TARGET_USE_GATHER): Define.
10815 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10817 2018-01-10 Martin Liska <mliska@suse.cz>
10820 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10821 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10823 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10824 CLEANUP_NO_PARTITIONING is not set.
10826 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10828 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10829 for vectors, as a partial revert of r254296.
10830 * rtl.h (const_vec_p): Delete.
10831 (const_vec_duplicate_p): Don't test for vector CONSTs.
10832 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10833 * expmed.c (make_tree): Likewise.
10836 * common.md (E, F): Use CONSTANT_P instead of checking for
10838 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10839 checking for CONST_VECTOR.
10841 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10843 PR middle-end/83575
10844 * predict.c (force_edge_cold): Handle in more sane way edges
10845 with no prediction.
10847 2018-01-09 Carl Love <cel@us.ibm.com>
10849 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10851 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10852 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10853 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10854 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10855 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10856 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10857 * config/rs6000/rs6000-protos.h: Add extern defition for
10858 rs6000_generate_float2_double_code.
10859 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10861 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10862 (float2_v2df): Add define_expand.
10864 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10867 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10868 op_mode in the force_to_mode call.
10870 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10872 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10873 instead of checking each element individually.
10874 (aarch64_evpc_uzp): Likewise.
10875 (aarch64_evpc_zip): Likewise.
10876 (aarch64_evpc_ext): Likewise.
10877 (aarch64_evpc_rev): Likewise.
10878 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10879 instead of checking each element individually. Return true without
10881 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10882 whether all selected elements come from the same input, instead of
10883 checking each element individually. Remove calls to gen_rtx_REG,
10884 start_sequence and end_sequence and instead assert that no rtl is
10887 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10889 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10890 order of HIGH and CONST checks.
10892 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10894 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10895 if the destination isn't an SSA_NAME.
10897 2018-01-09 Richard Biener <rguenther@suse.de>
10899 PR tree-optimization/83668
10900 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10902 (canonicalize_loop_form): ... here, renamed from ...
10903 (canonicalize_loop_closed_ssa_form): ... this and amended to
10904 swap successor edges for loop exit blocks to make us use
10905 the RPO order we need for initial schedule generation.
10907 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10909 PR tree-optimization/64811
10910 * match.pd: When optimizing comparisons with Inf, avoid
10911 introducing or losing exceptions from comparisons with NaN.
10913 2018-01-09 Martin Liska <mliska@suse.cz>
10916 * asan.c (shadow_mem_size): Add gcc_assert.
10918 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10920 Don't save registers in main().
10923 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10924 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10925 * config/avr/avr.c (avr_set_current_function): Don't error if
10926 naked, OS_task or OS_main are specified at the same time.
10927 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10929 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10931 * common/config/avr/avr-common.c (avr_option_optimization_table):
10932 Switch on -mmain-is-OS_task for optimizing compilations.
10934 2018-01-09 Richard Biener <rguenther@suse.de>
10936 PR tree-optimization/83572
10937 * graphite.c: Include cfganal.h.
10938 (graphite_transform_loops): Connect infinite loops to exit
10939 and remove fake edges at the end.
10941 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10943 * ipa-inline.c (edge_badness): Revert accidental checkin.
10945 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10948 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10949 symbols; not inline clones.
10951 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10954 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10955 hard registers. Formatting fixes.
10957 PR preprocessor/83722
10958 * gcc.c (try_generate_repro): Pass
10959 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10960 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10963 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10964 Kito Cheng <kito.cheng@gmail.com>
10966 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10967 (riscv_leaf_function_p): Delete.
10968 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10970 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10972 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10974 (do_ifelse): New function.
10975 (do_isel): New function.
10976 (do_sub3): New function.
10977 (do_add3): New function.
10978 (do_load_mask_compare): New function.
10979 (do_overlap_load_compare): New function.
10980 (expand_compare_loop): New function.
10981 (expand_block_compare): Call expand_compare_loop() when appropriate.
10982 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10983 option description.
10984 (-mblock-compare-inline-loop-limit): New option.
10986 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10989 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
10990 Reverse order of second and third operands in first alternative.
10991 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10992 of first and second elements in UNSPEC_VPERMR vector.
10993 (altivec_expand_vec_perm_le): Likewise.
10995 2018-01-08 Jeff Law <law@redhat.com>
10997 PR rtl-optimizatin/81308
10998 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10999 (process_switch): If group_case_labels makes a change, then set
11001 (pass_convert_switch::execute): If a switch is converted, then
11002 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11004 PR rtl-optimization/81308
11005 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11008 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11010 PR target/83663 - Revert r255946
11011 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11012 generation for cases where splatting a value is not useful.
11013 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11014 across a vec_duplicate and a paradoxical subreg forming a vector
11015 mode to a vec_concat.
11017 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11019 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11020 -march=armv8.3-a variants.
11021 * config/arm/t-multilib: Likewise.
11022 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11024 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11026 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11028 (cceq_ior_compare_complement): Give it a name so I can use it, and
11029 change boolean_or_operator predicate to boolean_operator so it can
11030 be used to generate a crand.
11031 (eqne): New code iterator.
11032 (bd/bd_neg): New code_attrs.
11033 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11034 a single define_insn.
11035 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11036 decrement (bdnzt/bdnzf/bdzt/bdzf).
11037 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11038 with the new names of the branch decrement patterns, and added the
11039 names of the branch decrement conditional patterns.
11041 2018-01-08 Richard Biener <rguenther@suse.de>
11043 PR tree-optimization/83563
11044 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11047 2018-01-08 Richard Biener <rguenther@suse.de>
11049 PR middle-end/83713
11050 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11052 2018-01-08 Richard Biener <rguenther@suse.de>
11054 PR tree-optimization/83685
11055 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11056 references to abnormals.
11058 2018-01-08 Richard Biener <rguenther@suse.de>
11061 * dwarf2out.c (output_indirect_strings): Handle empty
11062 skeleton_debug_str_hash.
11063 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11065 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11067 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11068 (emit_store_direct): Likewise.
11069 (arc_trampoline_adjust_address): Likewise.
11070 (arc_asm_trampoline_template): New function.
11071 (arc_initialize_trampoline): Use asm_trampoline_template.
11072 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11073 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11074 * config/arc/arc.md (flush_icache): Delete pattern.
11076 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11078 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11079 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11082 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11085 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11086 by not USED_FOR_TARGET.
11087 (make_pass_resolve_sw_modes): Likewise.
11089 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11091 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11094 2018-01-08 Richard Biener <rguenther@suse.de>
11096 PR middle-end/83580
11097 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11099 2018-01-08 Richard Biener <rguenther@suse.de>
11101 PR middle-end/83517
11102 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11104 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11106 PR middle-end/81897
11107 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11108 basic blocks with a small number of successors.
11109 (convert_control_dep_chain_into_preds): Improve handling of
11111 (dump_predicates): Split apart into...
11112 (dump_pred_chain): ...here...
11113 (dump_pred_info): ...and here.
11114 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11115 (can_chain_union_be_invalidated_p): Improve check for invalidation
11117 (uninit_uses_cannot_happen): Avoid unnecessary if
11118 convert_control_dep_chain_into_preds yielded nothing.
11120 2018-01-06 Martin Sebor <msebor@redhat.com>
11122 PR tree-optimization/83640
11123 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11124 subtracting negative offset from size.
11125 (builtin_access::overlap): Adjust offset bounds of the access to fall
11126 within the size of the object if possible.
11128 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11130 PR rtl-optimization/83699
11131 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11132 extract_bit_field_as_subreg to cases in which the extracted
11133 value is also a vector.
11135 * lra-constraints.c (process_alt_operands): Test for the equivalence
11136 substitutions when detecting a possible reload cycle.
11138 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11141 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11142 by default if flag_selective_schedling{,2}. Formatting fixes.
11144 PR rtl-optimization/83682
11145 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11146 if it has non-VECTOR_MODE element mode.
11147 (vec_duplicate_p): Likewise.
11149 PR middle-end/83694
11150 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11151 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11153 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11156 * config/i386/i386-builtin.def
11157 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11158 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11159 Require also OPTION_MASK_ISA_AVX512F in addition to
11160 OPTION_MASK_ISA_GFNI.
11161 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11162 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11163 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11164 to OPTION_MASK_ISA_GFNI.
11165 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11166 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11167 OPTION_MASK_ISA_AVX512BW.
11168 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11169 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11170 addition to OPTION_MASK_ISA_GFNI.
11171 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11172 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11173 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11174 to OPTION_MASK_ISA_GFNI.
11175 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11176 a requirement for all ISAs rather than any of them with a few
11178 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11180 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11181 bitmasks to be enabled with 3 exceptions, instead of requiring any
11182 enabled ISA with lots of exceptions.
11183 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11184 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11185 Change avx512bw in isa attribute to avx512f.
11186 * config/i386/sgxintrin.h: Add license boilerplate.
11187 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11188 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11189 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11190 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11192 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11193 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11194 temporarily sse2 rather than sse if not enabled already.
11197 * config/i386/sse.md (VI248_VLBW): Rename to ...
11198 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11199 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11200 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11201 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11202 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11203 mode iterator instead of VI248_VLBW.
11205 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11207 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11208 (record_modified): Skip clobbers; add debug output.
11209 (param_change_prob): Use sreal frequencies.
11211 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11213 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11214 punt for user-aligned variables.
11216 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11218 * tree-chrec.c (chrec_contains_symbols): Return true for
11221 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11224 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11225 of (x|y) == x for BICS pattern.
11227 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11229 PR tree-optimization/83605
11230 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11231 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11234 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11236 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11237 * config/epiphany/rtems.h: New file.
11239 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11240 Uros Bizjak <ubizjak@gmail.com>
11243 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11244 QIreg_operand instead of register_operand predicate.
11245 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11246 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11247 comments instead of -fmitigate[-_]rop.
11249 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11252 * cgraphunit.c (symbol_table::compile): Switch to text_section
11253 before calling assembly_start debug hook.
11254 * run-rtl-passes.c (run_rtl_passes): Likewise.
11257 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11259 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11260 range_int_cst_p rather than !symbolic_range_p before calling
11261 extract_range_from_multiplicative_op_1.
11263 2018-01-04 Jeff Law <law@redhat.com>
11265 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11266 redundant test in assertion.
11268 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11270 * doc/rtl.texi: Document machine_mode wrapper classes.
11272 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11274 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11275 using tree_to_uhwi.
11277 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11279 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11280 the VEC_PERM_EXPR fold to fail.
11282 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11285 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11286 to switched_sections.
11288 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11291 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11292 test for d.testing.
11294 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11297 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11298 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11300 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11303 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11304 is BLKmode and bitpos not zero or mode change is needed.
11306 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11309 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11312 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11315 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11316 instead of MULT rtx. Update all corresponding splitters.
11317 (*saddl_se): Ditto.
11318 (*ssub<modesuffix>): Ditto.
11319 (*ssubl_se): Ditto.
11320 (*cmp_sadd_di): Update split patterns.
11321 (*cmp_sadd_si): Ditto.
11322 (*cmp_sadd_sidi): Ditto.
11323 (*cmp_ssub_di): Ditto.
11324 (*cmp_ssub_si): Ditto.
11325 (*cmp_ssub_sidi): Ditto.
11326 * config/alpha/predicates.md (const23_operand): New predicate.
11327 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11328 Look for ASHIFT, not MULT inner operand.
11329 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11331 2018-01-04 Martin Liska <mliska@suse.cz>
11333 PR gcov-profile/83669
11334 * gcov.c (output_intermediate_file): Add version to intermediate
11336 * doc/gcov.texi: Document new field 'version' in intermediate
11337 file format. Fix location of '-k' option of gcov command.
11339 2018-01-04 Martin Liska <mliska@suse.cz>
11342 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11344 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11346 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11348 2018-01-03 Martin Sebor <msebor@redhat.com>
11350 PR tree-optimization/83655
11351 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11352 checking calls with invalid arguments.
11354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11356 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11357 (vectorizable_mask_load_store): Delete.
11358 (vectorizable_call): Return false for masked loads and stores.
11359 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11360 instead of gimple_assign_rhs1.
11361 (vectorizable_load): Handle IFN_MASK_LOAD.
11362 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11366 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11368 (vectorizable_mask_load_store): ...here.
11369 (vectorizable_load): ...and here.
11371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11373 * tree-vect-stmts.c (vect_build_all_ones_mask)
11374 (vect_build_zero_merge_argument): New functions, split out from...
11375 (vectorizable_load): ...here.
11377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11379 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11381 (vectorizable_mask_load_store): ...here.
11382 (vectorizable_store): ...and here.
11384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11386 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11388 (vectorizable_mask_load_store): ...here.
11390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11392 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11393 (vect_model_store_cost): Take a vec_load_store_type instead of a
11395 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11396 (vect_model_store_cost): Take a vec_load_store_type instead of a
11398 (vectorizable_mask_load_store): Update accordingly.
11399 (vectorizable_store): Likewise.
11400 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11404 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11405 IFN_MASK_LOAD calls here rather than...
11406 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11409 Alan Hayward <alan.hayward@arm.com>
11410 David Sherwood <david.sherwood@arm.com>
11412 * expmed.c (extract_bit_field_1): For vector extracts,
11413 fall back to extract_bit_field_as_subreg if vec_extract
11416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11417 Alan Hayward <alan.hayward@arm.com>
11418 David Sherwood <david.sherwood@arm.com>
11420 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11421 they are variable or constant sized.
11422 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11423 slots for constant-sized data.
11425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11426 Alan Hayward <alan.hayward@arm.com>
11427 David Sherwood <david.sherwood@arm.com>
11429 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11430 handling COND_EXPRs with boolean comparisons, try to find a better
11431 basis for the mask type than the boolean itself.
11433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11435 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11436 is calculated and how it can be overridden.
11437 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11438 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11440 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11444 Alan Hayward <alan.hayward@arm.com>
11445 David Sherwood <david.sherwood@arm.com>
11447 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11448 Remove the mode argument.
11449 (aarch64_simd_valid_immediate): Remove the mode and inverse
11451 * config/aarch64/iterators.md (bitsize): New iterator.
11452 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11453 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11454 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11455 aarch64_simd_valid_immediate.
11456 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11457 (aarch64_reg_or_bic_imm): Likewise.
11458 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11459 with an insn_type enum and msl with a modifier_type enum.
11460 Replace element_width with a scalar_mode. Change the shift
11461 to unsigned int. Add constructors for scalar_float_mode and
11462 scalar_int_mode elements.
11463 (aarch64_vect_float_const_representable_p): Delete.
11464 (aarch64_can_const_movi_rtx_p)
11465 (aarch64_simd_scalar_immediate_valid_for_move)
11466 (aarch64_simd_make_constant): Update call to
11467 aarch64_simd_valid_immediate.
11468 (aarch64_advsimd_valid_immediate_hs): New function.
11469 (aarch64_advsimd_valid_immediate): Likewise.
11470 (aarch64_simd_valid_immediate): Remove mode and inverse
11471 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11472 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11473 and aarch64_float_const_representable_p on the result.
11474 (aarch64_output_simd_mov_immediate): Remove mode argument.
11475 Update call to aarch64_simd_valid_immediate and use of
11476 simd_immediate_info.
11477 (aarch64_output_scalar_simd_mov_immediate): Update call
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 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11485 (mode_nunits): Likewise CONST_MODE_NUNITS.
11486 * machmode.def (ADJUST_NUNITS): Document.
11487 * genmodes.c (mode_data::need_nunits_adj): New field.
11488 (blank_mode): Update accordingly.
11489 (adj_nunits): New variable.
11490 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11492 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11493 listed in adj_nunits.
11494 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11495 listed in adj_nunits. Don't emit case statements for such modes.
11496 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11497 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11498 nothing if adj_nunits is nonnull.
11499 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11500 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11501 (emit_mode_fbit): Update use of print_maybe_const_decl.
11502 (emit_move_size): Likewise. Treat the array as non-const
11504 (emit_mode_adjustments): Handle adj_nunits.
11506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11508 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11509 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11510 (VECTOR_MODES): Use it.
11511 (make_vector_modes): Take the prefix as an argument.
11513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11514 Alan Hayward <alan.hayward@arm.com>
11515 David Sherwood <david.sherwood@arm.com>
11517 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11518 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11519 for MODE_VECTOR_BOOL.
11520 * machmode.def (VECTOR_BOOL_MODE): Document.
11521 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11522 (make_vector_bool_mode): New function.
11523 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11525 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11526 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11528 * stor-layout.c (int_mode_for_mode): Likewise.
11529 * tree.c (build_vector_type_for_mode): Likewise.
11530 * varasm.c (output_constant_pool_2): Likewise.
11531 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11532 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11533 for MODE_VECTOR_BOOL.
11534 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11535 of mode class checks.
11536 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11537 instead of a list of mode class checks.
11538 (expand_vector_scalar_condition): Likewise.
11539 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11542 Alan Hayward <alan.hayward@arm.com>
11543 David Sherwood <david.sherwood@arm.com>
11545 * machmode.h (mode_size): Change from unsigned short to
11547 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11548 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11549 or if measurement_type is not polynomial.
11550 (fixed_size_mode::includes_p): Check for constant-sized modes.
11551 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11552 return a poly_uint16 rather than an unsigned short.
11553 (emit_mode_size): Change the type of mode_size from unsigned short
11554 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11555 (emit_mode_adjustments): Cope with polynomial vector sizes.
11556 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11558 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11560 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11561 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11562 * caller-save.c (setup_save_areas): Likewise.
11563 (replace_reg_with_saved_mem): Likewise.
11564 * calls.c (emit_library_call_value_1): Likewise.
11565 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11566 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11567 (gen_lowpart_for_combine): Likewise.
11568 * convert.c (convert_to_integer_1): Likewise.
11569 * cse.c (equiv_constant, cse_insn): Likewise.
11570 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11571 (cselib_subst_to_values): Likewise.
11572 * dce.c (word_dce_process_block): Likewise.
11573 * df-problems.c (df_word_lr_mark_ref): Likewise.
11574 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11575 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11576 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11577 (rtl_for_decl_location): Likewise.
11578 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11579 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11580 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11581 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11582 (expand_expr_real_1): Likewise.
11583 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11584 (pad_below): Likewise.
11585 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11586 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11587 * ira.c (get_subreg_tracking_sizes): Likewise.
11588 * ira-build.c (ira_create_allocno_objects): Likewise.
11589 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11590 (ira_sort_regnos_for_alter_reg): Likewise.
11591 * ira-costs.c (record_operand_costs): Likewise.
11592 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11593 (resolve_simple_move): Likewise.
11594 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11595 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11596 (lra_constraints): Likewise.
11597 (CONST_POOL_OK_P): Reject variable-sized modes.
11598 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11599 (add_pseudo_to_slot, lra_spill): Likewise.
11600 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11601 * optabs-query.c (get_best_extraction_insn): Likewise.
11602 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11603 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11604 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11605 * recog.c (offsettable_address_addr_space_p): Likewise.
11606 * regcprop.c (maybe_mode_change): Likewise.
11607 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11608 * regrename.c (build_def_use): Likewise.
11609 * regstat.c (dump_reg_info): Likewise.
11610 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11611 (find_reloads, find_reloads_subreg_address): Likewise.
11612 * reload1.c (eliminate_regs_1): Likewise.
11613 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11614 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11615 (simplify_binary_operation_1, simplify_subreg): Likewise.
11616 * targhooks.c (default_function_arg_padding): Likewise.
11617 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11618 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11619 (verify_gimple_assign_ternary): Likewise.
11620 * tree-inline.c (estimate_move_cost): Likewise.
11621 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11622 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11623 (get_address_cost_ainc): Likewise.
11624 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11625 (vect_supportable_dr_alignment): Likewise.
11626 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11627 (vectorizable_reduction): Likewise.
11628 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11629 (vectorizable_operation, vectorizable_load): Likewise.
11630 * tree.c (build_same_sized_truth_vector_type): Likewise.
11631 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11632 * var-tracking.c (emit_note_insn_var_location): Likewise.
11633 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11634 (ADDR_VEC_ALIGN): Likewise.
11636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11637 Alan Hayward <alan.hayward@arm.com>
11638 David Sherwood <david.sherwood@arm.com>
11640 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11642 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11643 or if measurement_type is polynomial.
11644 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11645 * combine.c (make_extraction): Likewise.
11646 * dse.c (find_shift_sequence): Likewise.
11647 * dwarf2out.c (mem_loc_descriptor): Likewise.
11648 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11649 (extract_bit_field, extract_low_bits): Likewise.
11650 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11651 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11652 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11653 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11654 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11655 * reload.c (find_reloads): Likewise.
11656 * reload1.c (alter_reg): Likewise.
11657 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11658 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11659 * tree-if-conv.c (predicate_mem_writes): Likewise.
11660 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11661 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11662 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11663 * valtrack.c (dead_debug_insert_temp): Likewise.
11664 * varasm.c (mergeable_constant_section): Likewise.
11665 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11668 Alan Hayward <alan.hayward@arm.com>
11669 David Sherwood <david.sherwood@arm.com>
11671 * expr.c (expand_assignment): Cope with polynomial mode sizes
11672 when assigning to a CONCAT.
11674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11675 Alan Hayward <alan.hayward@arm.com>
11676 David Sherwood <david.sherwood@arm.com>
11678 * machmode.h (mode_precision): Change from unsigned short to
11680 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11682 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11683 or if measurement_type is not polynomial.
11684 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11685 in which the mode is already known to be a scalar_int_mode.
11686 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11687 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11689 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11690 for GET_MODE_PRECISION.
11691 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11692 for GET_MODE_PRECISION.
11693 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11695 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11696 (expand_field_assignment, make_extraction): Likewise.
11697 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11698 (get_last_value): Likewise.
11699 * convert.c (convert_to_integer_1): Likewise.
11700 * cse.c (cse_insn): Likewise.
11701 * expr.c (expand_expr_real_1): Likewise.
11702 * lra-constraints.c (simplify_operand_subreg): Likewise.
11703 * optabs-query.c (can_atomic_load_p): Likewise.
11704 * optabs.c (expand_atomic_load): Likewise.
11705 (expand_atomic_store): Likewise.
11706 * ree.c (combine_reaching_defs): Likewise.
11707 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11708 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11709 * tree.h (type_has_mode_precision_p): Likewise.
11710 * ubsan.c (instrument_si_overflow): Likewise.
11712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11713 Alan Hayward <alan.hayward@arm.com>
11714 David Sherwood <david.sherwood@arm.com>
11716 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11717 polynomial numbers of units.
11718 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11719 (valid_vector_subparts_p): New function.
11720 (build_vector_type): Remove temporary shim and take the number
11721 of units as a poly_uint64 rather than an int.
11722 (build_opaque_vector_type): Take the number of units as a
11723 poly_uint64 rather than an int.
11724 * tree.c (build_vector_from_ctor): Handle polynomial
11725 TYPE_VECTOR_SUBPARTS.
11726 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11727 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11728 (build_vector_from_val): If the number of units is variable,
11729 use build_vec_duplicate_cst for constant operands and
11730 VEC_DUPLICATE_EXPR otherwise.
11731 (make_vector_type): Remove temporary is_constant ().
11732 (build_vector_type, build_opaque_vector_type): Take the number of
11733 units as a poly_uint64 rather than an int.
11734 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11736 * cfgexpand.c (expand_debug_expr): Likewise.
11737 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11738 (store_constructor, expand_expr_real_1): Likewise.
11739 (const_scalar_mask_from_tree): Likewise.
11740 * fold-const-call.c (fold_const_reduction): Likewise.
11741 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11742 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11743 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11744 (fold_relational_const): Likewise.
11745 (native_interpret_vector): Likewise. Change the size from an
11746 int to an unsigned int.
11747 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11748 TYPE_VECTOR_SUBPARTS.
11749 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11750 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11751 duplicating a non-constant operand into a variable-length vector.
11752 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11753 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11754 * ipa-icf.c (sem_variable::equals): Likewise.
11755 * match.pd: Likewise.
11756 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11757 * print-tree.c (print_node): Likewise.
11758 * stor-layout.c (layout_type): Likewise.
11759 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11760 * tree-cfg.c (verify_gimple_comparison): Likewise.
11761 (verify_gimple_assign_binary): Likewise.
11762 (verify_gimple_assign_ternary): Likewise.
11763 (verify_gimple_assign_single): Likewise.
11764 * tree-pretty-print.c (dump_generic_node): Likewise.
11765 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11766 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11767 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11768 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11769 (vect_shift_permute_load_chain): Likewise.
11770 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11771 (expand_vector_condition, optimize_vector_constructor): Likewise.
11772 (lower_vec_perm, get_compute_type): Likewise.
11773 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11774 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11775 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11776 (vect_recog_mask_conversion_pattern): Likewise.
11777 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11778 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11779 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11780 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11781 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11782 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11783 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11784 (supportable_widening_operation): Likewise.
11785 (supportable_narrowing_operation): Likewise.
11786 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11788 * varasm.c (output_constant): Likewise.
11790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11791 Alan Hayward <alan.hayward@arm.com>
11792 David Sherwood <david.sherwood@arm.com>
11794 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11795 so that both the length == 3 and length != 3 cases set up their
11796 own permute vectors. Add comments explaining why we know the
11797 number of elements is constant.
11798 (vect_permute_load_chain): Likewise.
11800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11801 Alan Hayward <alan.hayward@arm.com>
11802 David Sherwood <david.sherwood@arm.com>
11804 * machmode.h (mode_nunits): Change from unsigned char to
11806 (ONLY_FIXED_SIZE_MODES): New macro.
11807 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11808 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11809 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11811 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11812 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11813 or if measurement_type is not polynomial.
11814 * genmodes.c (ZERO_COEFFS): New macro.
11815 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11817 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11818 Use ZERO_COEFFS when emitting initializers.
11819 * data-streamer.h (bp_pack_poly_value): New function.
11820 (bp_unpack_poly_value): Likewise.
11821 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11822 for GET_MODE_NUNITS.
11823 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11824 for GET_MODE_NUNITS.
11825 * tree.c (make_vector_type): Remove temporary shim and make
11826 the real function take the number of units as a poly_uint64
11827 rather than an int.
11828 (build_vector_type_for_mode): Handle polynomial nunits.
11829 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11830 * emit-rtl.c (const_vec_series_p_1): Likewise.
11831 (gen_rtx_CONST_VECTOR): Likewise.
11832 * fold-const.c (test_vec_duplicate_folding): Likewise.
11833 * genrecog.c (validate_pattern): Likewise.
11834 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11835 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11836 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11837 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11838 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11839 * rtlanal.c (subreg_get_info): Likewise.
11840 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11841 (vect_grouped_load_supported): Likewise.
11842 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11843 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11844 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11845 (simplify_const_unary_operation, simplify_binary_operation_1)
11846 (simplify_const_binary_operation, simplify_ternary_operation)
11847 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11848 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11849 instead of CONST_VECTOR_NUNITS.
11850 * varasm.c (output_constant_pool_2): Likewise.
11851 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11852 explicit-encoded elements in the XVEC for variable-length vectors.
11854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11856 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11859 Alan Hayward <alan.hayward@arm.com>
11860 David Sherwood <david.sherwood@arm.com>
11862 * coretypes.h (fixed_size_mode): Declare.
11863 (fixed_size_mode_pod): New typedef.
11864 * builtins.h (target_builtins::x_apply_args_mode)
11865 (target_builtins::x_apply_result_mode): Change type to
11866 fixed_size_mode_pod.
11867 * builtins.c (apply_args_size, apply_result_size, result_vector)
11868 (expand_builtin_apply_args_1, expand_builtin_apply)
11869 (expand_builtin_return): Update accordingly.
11871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11873 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11874 * cselib.c (cselib_hash_rtx): Likewise.
11875 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11876 CONST_VECTOR encoding.
11878 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11879 Jeff Law <law@redhat.com>
11882 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11883 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11884 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11885 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11888 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11889 explicitly probe *sp in a noreturn function if there were any callee
11890 register saves or frame pointer is needed.
11892 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11895 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11896 BLKmode for ternary, binary or unary expressions.
11899 * var-tracking.c (delete_vta_debug_insn): New inline function.
11900 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11901 insns from get_insns () to NULL instead of each bb separately.
11902 Use delete_vta_debug_insn. No longer static.
11903 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11904 delete_vta_debug_insns callers.
11905 * rtl.h (delete_vta_debug_insns): Declare.
11906 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11907 instead of variable_tracking_main.
11909 2018-01-03 Martin Sebor <msebor@redhat.com>
11911 PR tree-optimization/83603
11912 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11913 arguments past the endof the argument list in functions declared
11914 without a prototype.
11915 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11916 Avoid checking when arguments are null.
11918 2018-01-03 Martin Sebor <msebor@redhat.com>
11921 * doc/extend.texi (attribute const): Fix a typo.
11922 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11923 issuing -Wsuggest-attribute for void functions.
11925 2018-01-03 Martin Sebor <msebor@redhat.com>
11927 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11928 offset_int::from instead of wide_int::to_shwi.
11929 (maybe_diag_overlap): Remove assertion.
11930 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11931 * gimple-ssa-sprintf.c (format_directive): Same.
11932 (parse_directive): Same.
11933 (sprintf_dom_walker::compute_format_length): Same.
11934 (try_substitute_return_value): Same.
11936 2018-01-03 Jeff Law <law@redhat.com>
11938 PR middle-end/83654
11939 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11940 non-constant residual for zero at runtime and avoid probing in
11941 that case. Reorganize code for trailing problem to mirror handling
11944 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11946 PR tree-optimization/83501
11947 * tree-ssa-strlen.c (get_string_cst): New.
11948 (handle_char_store): Call get_string_cst.
11950 2018-01-03 Martin Liska <mliska@suse.cz>
11952 PR tree-optimization/83593
11953 * tree-ssa-strlen.c: Include tree-cfg.h.
11954 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11955 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11956 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11958 (strlen_dom_walker::before_dom_children): Call
11959 gimple_purge_dead_eh_edges. Dump tranformation with details
11961 (strlen_dom_walker::before_dom_children): Update call by adding
11962 new argument cleanup_eh.
11963 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11965 2018-01-03 Martin Liska <mliska@suse.cz>
11968 * cif-code.def (VARIADIC_THUNK): New enum value.
11969 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11972 2018-01-03 Jan Beulich <jbeulich@suse.com>
11974 * sse.md (mov<mode>_internal): Tighten condition for when to use
11975 vmovdqu<ssescalarsize> for TI and OI modes.
11977 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11979 Update copyright years.
11981 2018-01-03 Martin Liska <mliska@suse.cz>
11984 * ipa-visibility.c (function_and_variable_visibility): Skip
11985 functions with noipa attribure.
11987 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11989 * gcc.c (process_command): Update copyright notice dates.
11990 * gcov-dump.c (print_version): Ditto.
11991 * gcov.c (print_version): Ditto.
11992 * gcov-tool.c (print_version): Ditto.
11993 * gengtype.c (create_file): Ditto.
11994 * doc/cpp.texi: Bump @copying's copyright year.
11995 * doc/cppinternals.texi: Ditto.
11996 * doc/gcc.texi: Ditto.
11997 * doc/gccint.texi: Ditto.
11998 * doc/gcov.texi: Ditto.
11999 * doc/install.texi: Ditto.
12000 * doc/invoke.texi: Ditto.
12002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12004 * vector-builder.h (vector_builder::m_full_nelts): Change from
12005 unsigned int to poly_uint64.
12006 (vector_builder::full_nelts): Update prototype accordingly.
12007 (vector_builder::new_vector): Likewise.
12008 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12009 (vector_builder::operator ==): Likewise.
12010 (vector_builder::finalize): Likewise.
12011 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12012 Take the number of elements as a poly_uint64 rather than an
12014 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12015 from unsigned int to poly_uint64.
12016 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12017 (vec_perm_indices::new_vector): Likewise.
12018 (vec_perm_indices::length): Likewise.
12019 (vec_perm_indices::nelts_per_input): Likewise.
12020 (vec_perm_indices::input_nelts): Likewise.
12021 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12022 number of elements per input as a poly_uint64 rather than an
12023 unsigned int. Use the original encoding for variable-length
12024 vectors, rather than clamping each individual element.
12025 For the second and subsequent elements in each pattern,
12026 clamp the step and base before clamping their sum.
12027 (vec_perm_indices::series_p): Handle polynomial element counts.
12028 (vec_perm_indices::all_in_range_p): Likewise.
12029 (vec_perm_indices_to_tree): Likewise.
12030 (vec_perm_indices_to_rtx): Likewise.
12031 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12032 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12033 (tree_vector_builder::new_binary_operation): Handle polynomial
12034 element counts. Return false if we need to know the number
12035 of elements at compile time.
12036 * fold-const.c (fold_vec_perm): Punt if the number of elements
12037 isn't known at compile time.
12039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12041 * vec-perm-indices.h (vec_perm_builder): Change element type
12042 from HOST_WIDE_INT to poly_int64.
12043 (vec_perm_indices::element_type): Update accordingly.
12044 (vec_perm_indices::clamp): Handle polynomial element_types.
12045 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12046 (vec_perm_indices::all_in_range_p): Likewise.
12047 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12049 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12050 polynomial vec_perm_indices element types.
12051 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12052 * fold-const.c (fold_vec_perm): Likewise.
12053 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12054 * tree-vect-generic.c (lower_vec_perm): Likewise.
12055 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12056 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12057 element type to HOST_WIDE_INT.
12059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12060 Alan Hayward <alan.hayward@arm.com>
12061 David Sherwood <david.sherwood@arm.com>
12063 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12064 rather than an int. Use plus_constant.
12065 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12066 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12069 Alan Hayward <alan.hayward@arm.com>
12070 David Sherwood <david.sherwood@arm.com>
12072 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12073 a HOST_WIDE_INT to a poly_int64.
12075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12076 Alan Hayward <alan.hayward@arm.com>
12077 David Sherwood <david.sherwood@arm.com>
12079 * calls.c (load_register_parameters): Cope with polynomial
12080 mode sizes. Require a constant size for BLKmode parameters
12081 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12082 forces a parameter to be padded at the lsb end in order to
12083 fill a complete number of words, require the parameter size
12084 to be ordered wrt UNITS_PER_WORD.
12086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12087 Alan Hayward <alan.hayward@arm.com>
12088 David Sherwood <david.sherwood@arm.com>
12090 * reload1.c (spill_stack_slot_width): Change element type
12091 from unsigned int to poly_uint64_pod.
12092 (alter_reg): Treat mode sizes as polynomial.
12094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12095 Alan Hayward <alan.hayward@arm.com>
12096 David Sherwood <david.sherwood@arm.com>
12098 * reload.c (complex_word_subreg_p): New function.
12099 (reload_inner_reg_of_subreg, push_reload): Use it.
12101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12102 Alan Hayward <alan.hayward@arm.com>
12103 David Sherwood <david.sherwood@arm.com>
12105 * lra-constraints.c (process_alt_operands): Reject matched
12106 operands whose sizes aren't ordered.
12107 (match_reload): Refer to this check here.
12109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12110 Alan Hayward <alan.hayward@arm.com>
12111 David Sherwood <david.sherwood@arm.com>
12113 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12114 that the mode size is in the set {1, 2, 4, 8, 16}.
12116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12117 Alan Hayward <alan.hayward@arm.com>
12118 David Sherwood <david.sherwood@arm.com>
12120 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12121 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12128 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12129 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12130 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12131 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12132 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12133 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12134 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12135 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12136 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12138 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12139 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12140 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12141 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12142 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12143 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12144 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12145 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12146 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12147 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12149 * expr.c (emit_move_resolve_push): Treat the input and result
12150 of PUSH_ROUNDING as a poly_int64.
12151 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12152 (emit_push_insn): Likewise.
12153 * lra-eliminations.c (mark_not_eliminable): Likewise.
12154 * recog.c (push_operand): Likewise.
12155 * reload1.c (elimination_effects): Likewise.
12156 * rtlanal.c (nonzero_bits1): Likewise.
12157 * calls.c (store_one_arg): Likewise. Require the padding to be
12158 known at compile time.
12160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12161 Alan Hayward <alan.hayward@arm.com>
12162 David Sherwood <david.sherwood@arm.com>
12164 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12165 Use plus_constant instead of gen_rtx_PLUS.
12167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12168 Alan Hayward <alan.hayward@arm.com>
12169 David Sherwood <david.sherwood@arm.com>
12171 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12172 rather than an int.
12174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12175 Alan Hayward <alan.hayward@arm.com>
12176 David Sherwood <david.sherwood@arm.com>
12178 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12179 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12180 via stack temporaries. Treat the mode size as polynomial too.
12182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12183 Alan Hayward <alan.hayward@arm.com>
12184 David Sherwood <david.sherwood@arm.com>
12186 * expr.c (expand_expr_real_2): When handling conversions involving
12187 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12188 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12189 as a poly_uint64 too.
12191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12192 Alan Hayward <alan.hayward@arm.com>
12193 David Sherwood <david.sherwood@arm.com>
12195 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12198 Alan Hayward <alan.hayward@arm.com>
12199 David Sherwood <david.sherwood@arm.com>
12201 * combine.c (can_change_dest_mode): Handle polynomial
12202 REGMODE_NATURAL_SIZE.
12203 * expmed.c (store_bit_field_1): Likewise.
12204 * expr.c (store_constructor): Likewise.
12205 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12206 and polynomial REGMODE_NATURAL_SIZE.
12207 (gen_lowpart_common): Likewise.
12208 * reginfo.c (record_subregs_of_mode): Likewise.
12209 * rtlanal.c (read_modify_subreg_p): Likewise.
12211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12212 Alan Hayward <alan.hayward@arm.com>
12213 David Sherwood <david.sherwood@arm.com>
12215 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12216 numbers of elements.
12218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12219 Alan Hayward <alan.hayward@arm.com>
12220 David Sherwood <david.sherwood@arm.com>
12222 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12229 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
12236 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12237 (ipa_simd_modify_function_body): Likewise.
12239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12240 Alan Hayward <alan.hayward@arm.com>
12241 David Sherwood <david.sherwood@arm.com>
12243 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12244 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12245 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12246 (expand_vector_condition, vector_element): Likewise.
12247 (subparts_gt): New function.
12248 (get_compute_type): Use subparts_gt.
12249 (count_type_subparts): Delete.
12250 (expand_vector_operations_1): Use subparts_gt instead of
12251 count_type_subparts.
12253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12254 Alan Hayward <alan.hayward@arm.com>
12255 David Sherwood <david.sherwood@arm.com>
12257 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12258 (vect_compile_time_alias): ...this new function. Do the calculation
12259 on poly_ints rather than trees.
12260 (vect_prune_runtime_alias_test_list): Update call accordingly.
12262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12263 Alan Hayward <alan.hayward@arm.com>
12264 David Sherwood <david.sherwood@arm.com>
12266 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12268 (vect_schedule_slp_instance): Likewise.
12270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12271 Alan Hayward <alan.hayward@arm.com>
12272 David Sherwood <david.sherwood@arm.com>
12274 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12275 constant and extern definitions for variable-length vectors.
12276 (vect_get_constant_vectors): Note that the number of units
12277 is known to be constant.
12279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12280 Alan Hayward <alan.hayward@arm.com>
12281 David Sherwood <david.sherwood@arm.com>
12283 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12284 of units as polynomial. Choose between WIDE and NARROW based
12287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12288 Alan Hayward <alan.hayward@arm.com>
12289 David Sherwood <david.sherwood@arm.com>
12291 * tree-vect-stmts.c (simd_clone_subparts): New function.
12292 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12295 Alan Hayward <alan.hayward@arm.com>
12296 David Sherwood <david.sherwood@arm.com>
12298 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12299 vectors as polynomial. Use build_index_vector for
12300 IFN_GOMP_SIMD_LANE.
12302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12303 Alan Hayward <alan.hayward@arm.com>
12304 David Sherwood <david.sherwood@arm.com>
12306 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12307 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12308 for variable-length vectors.
12309 (vectorizable_mask_load_store): Treat the number of units as
12310 polynomial, asserting that it is constant if the condition has
12311 already been enforced.
12312 (vectorizable_store, vectorizable_load): Likewise.
12314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12315 Alan Hayward <alan.hayward@arm.com>
12316 David Sherwood <david.sherwood@arm.com>
12318 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12319 of units as polynomial. Punt if we can't tell at compile time
12320 which vector contains the final result.
12322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12323 Alan Hayward <alan.hayward@arm.com>
12324 David Sherwood <david.sherwood@arm.com>
12326 * tree-vect-loop.c (vectorizable_induction): Treat the number
12327 of units as polynomial. Punt on SLP inductions. Use an integer
12328 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12329 cast of such a series for variable-length floating-point
12332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12333 Alan Hayward <alan.hayward@arm.com>
12334 David Sherwood <david.sherwood@arm.com>
12336 * tree.h (build_index_vector): Declare.
12337 * tree.c (build_index_vector): New function.
12338 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12339 of units as polynomial, forcibly converting it to a constant if
12340 vectorizable_reduction has already enforced the condition.
12341 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12342 to create a {1,2,3,...} vector.
12343 (vectorizable_reduction): Treat the number of units as polynomial.
12344 Choose vectype_in based on the largest scalar element size rather
12345 than the smallest number of units. Enforce the restrictions
12348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12349 Alan Hayward <alan.hayward@arm.com>
12350 David Sherwood <david.sherwood@arm.com>
12352 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12353 number of units as polynomial.
12355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12356 Alan Hayward <alan.hayward@arm.com>
12357 David Sherwood <david.sherwood@arm.com>
12359 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12360 * target.def (autovectorize_vector_sizes): Return the vector sizes
12361 by pointer, using vector_sizes rather than a bitmask.
12362 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12363 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12364 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12366 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12367 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12368 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12369 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12370 * omp-general.c (omp_max_vf): Likewise.
12371 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12372 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12373 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12374 * tree-vect-slp.c (vect_slp_bb): Likewise.
12375 * doc/tm.texi: Regenerate.
12376 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12378 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12379 the vector size as a poly_uint64 rather than an unsigned int.
12380 (current_vector_size): Change from an unsigned int to a poly_uint64.
12381 (get_vectype_for_scalar_type): Update accordingly.
12382 * tree.h (build_truth_vector_type): Take the size and number of
12383 units as a poly_uint64 rather than an unsigned int.
12384 (build_vector_type): Add a temporary overload that takes
12385 the number of units as a poly_uint64 rather than an unsigned int.
12386 * tree.c (make_vector_type): Likewise.
12387 (build_truth_vector_type): Take the number of units as a poly_uint64
12388 rather than an unsigned int.
12390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12391 Alan Hayward <alan.hayward@arm.com>
12392 David Sherwood <david.sherwood@arm.com>
12394 * target.def (get_mask_mode): Take the number of units and length
12395 as poly_uint64s rather than unsigned ints.
12396 * targhooks.h (default_get_mask_mode): Update accordingly.
12397 * targhooks.c (default_get_mask_mode): Likewise.
12398 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12399 * doc/tm.texi: Regenerate.
12401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12402 Alan Hayward <alan.hayward@arm.com>
12403 David Sherwood <david.sherwood@arm.com>
12405 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12406 * omp-general.c (omp_max_vf): Likewise.
12407 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12408 (expand_omp_simd): Handle polynomial safelen.
12409 * omp-low.c (omplow_simd_context): Add a default constructor.
12410 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12411 (lower_rec_simd_input_clauses): Update accordingly.
12412 (lower_rec_input_clauses): Likewise.
12414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12415 Alan Hayward <alan.hayward@arm.com>
12416 David Sherwood <david.sherwood@arm.com>
12418 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12419 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12420 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12421 (vect_analyze_slp_cost): Likewise.
12422 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12423 (vect_model_load_cost): Likewise.
12425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12426 Alan Hayward <alan.hayward@arm.com>
12427 David Sherwood <david.sherwood@arm.com>
12429 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12430 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12431 from an unsigned int * to a poly_uint64_pod *.
12432 (calculate_unrolling_factor): New function.
12433 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12436 Alan Hayward <alan.hayward@arm.com>
12437 David Sherwood <david.sherwood@arm.com>
12439 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12440 from an unsigned int to a poly_uint64.
12441 (_loop_vec_info::slp_unrolling_factor): Likewise.
12442 (_loop_vec_info::vectorization_factor): Change from an int
12444 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12445 (vect_get_num_vectors): New function.
12446 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12447 (vect_get_num_copies): Use vect_get_num_vectors.
12448 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12449 to an unsigned int *.
12450 (vect_analyze_data_refs): Change min_vf from an int * to a
12452 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12453 than an unsigned HOST_WIDE_INT.
12454 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12455 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12456 to an unsigned int *.
12457 (vect_analyze_data_ref_dependences): Likewise.
12458 (vect_compute_data_ref_alignment): Handle polynomial vf.
12459 (vect_enhance_data_refs_alignment): Likewise.
12460 (vect_prune_runtime_alias_test_list): Likewise.
12461 (vect_shift_permute_load_chain): Likewise.
12462 (vect_supportable_dr_alignment): Likewise.
12463 (dependence_distance_ge_vf): Take the vectorization factor as a
12464 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12465 (vect_analyze_data_refs): Change min_vf from an int * to a
12467 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12468 vfm1 as a poly_uint64 rather than an int. Make the same change
12469 for the returned bound_scalar.
12470 (vect_gen_vector_loop_niters): Handle polynomial vf.
12471 (vect_do_peeling): Likewise. Update call to
12472 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12473 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12475 * tree-vect-loop.c (vect_determine_vectorization_factor)
12476 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12477 (vect_get_known_peeling_cost): Likewise.
12478 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12479 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12480 (vect_transform_loop): Likewise. Use the lowest possible VF when
12481 updating the upper bounds of the loop.
12482 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12483 rather than an int.
12484 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12485 polynomial unroll factors.
12486 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12487 (vect_make_slp_decision): Likewise.
12488 (vect_supported_load_permutation_p): Likewise, and polynomial
12490 (vect_analyze_slp_cost): Handle polynomial vf.
12491 (vect_slp_analyze_node_operations): Likewise.
12492 (vect_slp_analyze_bb_1): Likewise.
12493 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12494 than an unsigned HOST_WIDE_INT.
12495 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12496 (vectorizable_load): Handle polynomial vf.
12497 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12499 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12502 Alan Hayward <alan.hayward@arm.com>
12503 David Sherwood <david.sherwood@arm.com>
12505 * match.pd: Handle bit operations involving three constants
12506 and try to fold one pair.
12508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12510 * tree-vect-loop-manip.c: Include gimple-fold.h.
12511 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12512 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12513 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12514 Add a path that uses a step of VF instead of 1, but disable it
12516 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12517 and niters_no_overflow parameters. Update calls to
12518 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12519 Create a new SSA name if the latter choses to use a ste other
12520 than zero, and return it via niters_vector_mult_vf_var.
12521 * tree-vect-loop.c (vect_transform_loop): Update calls to
12522 vect_do_peeling, vect_gen_vector_loop_niters and
12523 slpeel_make_loop_iterate_ntimes.
12524 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12525 (vect_gen_vector_loop_niters): Update declarations after above changes.
12527 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12529 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12530 128-bit round to integer instructions.
12531 (ceil<mode>2): Likewise.
12532 (btrunc<mode>2): Likewise.
12533 (round<mode>2): Likewise.
12535 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12537 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12538 unaligned VSX load/store on P8/P9.
12539 (expand_block_clear): Allow the use of unaligned VSX
12540 load/store on P8/P9.
12542 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12544 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12546 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12547 swap associated with both a load and a store.
12549 2018-01-02 Andrew Waterman <andrew@sifive.com>
12551 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12552 * config/riscv/riscv.md (clear_cache): Use it.
12554 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12556 * web.c: Remove out-of-date comment.
12558 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12560 * expr.c (fixup_args_size_notes): Check that any existing
12561 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12562 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12563 (emit_single_push_insn): ...here.
12565 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12567 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12568 (const_vector_encoded_nelts): New function.
12569 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12570 (const_vector_int_elt, const_vector_elt): Declare.
12571 * emit-rtl.c (const_vector_int_elt_1): New function.
12572 (const_vector_elt): Likewise.
12573 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12574 of CONST_VECTOR_ELT.
12576 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12578 * expr.c: Include rtx-vector-builder.h.
12579 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12580 directly on the tree encoding.
12581 (const_vector_from_tree): Likewise.
12582 * optabs.c: Include rtx-vector-builder.h.
12583 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12584 sequence of "u" values.
12585 * vec-perm-indices.c: Include rtx-vector-builder.h.
12586 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12587 directly on the vec_perm_indices encoding.
12589 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12591 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12592 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12593 * rtx-vector-builder.h: New file.
12594 * rtx-vector-builder.c: Likewise.
12595 * rtl.h (rtx_def::u2): Add a const_vector field.
12596 (CONST_VECTOR_NPATTERNS): New macro.
12597 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12598 (CONST_VECTOR_DUPLICATE_P): Likewise.
12599 (CONST_VECTOR_STEPPED_P): Likewise.
12600 (CONST_VECTOR_ENCODED_ELT): Likewise.
12601 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12602 (unwrap_const_vec_duplicate): Likewise.
12603 (const_vec_series_p): Check for a non-duplicated vector encoding.
12604 Say that the function only returns true for integer vectors.
12605 * emit-rtl.c: Include rtx-vector-builder.h.
12606 (gen_const_vec_duplicate_1): Delete.
12607 (gen_const_vector): Call gen_const_vec_duplicate instead of
12608 gen_const_vec_duplicate_1.
12609 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12610 (gen_const_vec_duplicate): Use rtx_vector_builder.
12611 (gen_const_vec_series): Likewise.
12612 (gen_rtx_CONST_VECTOR): Likewise.
12613 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12614 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12615 Build a new vector rather than modifying a CONST_VECTOR in-place.
12616 (handle_special_swappables): Update call accordingly.
12617 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12618 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12619 Build a new vector rather than modifying a CONST_VECTOR in-place.
12620 (handle_special_swappables): Update call accordingly.
12622 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12624 * simplify-rtx.c (simplify_const_binary_operation): Use
12625 CONST_VECTOR_ELT instead of XVECEXP.
12627 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12629 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12630 the selector elements to be different from the data elements
12631 if the selector is a VECTOR_CST.
12632 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12633 ssizetype for the selector.
12635 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12637 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12638 before testing each element individually.
12639 * tree-vect-generic.c (lower_vec_perm): Likewise.
12641 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12643 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12644 * selftest-run-tests.c (selftest::run_tests): Call it.
12645 * vector-builder.h (vector_builder::operator ==): New function.
12646 (vector_builder::operator !=): Likewise.
12647 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12648 (vec_perm_indices::all_from_input_p): New function.
12649 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12650 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12651 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12652 instead of reading the VECTOR_CST directly. Detect whether both
12653 vector inputs are the same before constructing the vec_perm_indices,
12654 and update the number of inputs argument accordingly. Use the
12655 utility functions added above. Only construct sel2 if we need to.
12657 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12659 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12660 the broadcast of the low byte.
12661 (expand_mult_highpart): Use an explicit encoding for the permutes.
12662 * optabs-query.c (can_mult_highpart_p): Likewise.
12663 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12664 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12665 (vectorizable_bswap): Likewise.
12666 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12667 explicit encoding for the power-of-2 permutes.
12668 (vect_permute_store_chain): Likewise.
12669 (vect_grouped_load_supported): Likewise.
12670 (vect_permute_load_chain): Likewise.
12672 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12674 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12675 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12676 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12677 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12678 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12679 (vect_gen_perm_mask_any): Likewise.
12681 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12683 * int-vector-builder.h: New file.
12684 * vec-perm-indices.h: Include int-vector-builder.h.
12685 (vec_perm_indices): Redefine as an int_vector_builder.
12686 (auto_vec_perm_indices): Delete.
12687 (vec_perm_builder): Redefine as a stand-alone class.
12688 (vec_perm_indices::vec_perm_indices): New function.
12689 (vec_perm_indices::clamp): Likewise.
12690 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12691 (vec_perm_indices::new_vector): New function.
12692 (vec_perm_indices::new_expanded_vector): Update for new
12693 vec_perm_indices class.
12694 (vec_perm_indices::rotate_inputs): New function.
12695 (vec_perm_indices::all_in_range_p): Operate directly on the
12696 encoded form, without computing elided elements.
12697 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12698 encoding. Update for new vec_perm_indices class.
12699 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12700 the given vec_perm_builder.
12701 (expand_vec_perm_var): Update vec_perm_builder constructor.
12702 (expand_mult_highpart): Use vec_perm_builder instead of
12703 auto_vec_perm_indices.
12704 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12705 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12706 or double series encoding as appropriate.
12707 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12708 vec_perm_indices instead of auto_vec_perm_indices.
12709 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12710 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12711 (vect_permute_store_chain): Likewise.
12712 (vect_grouped_load_supported): Likewise.
12713 (vect_permute_load_chain): Likewise.
12714 (vect_shift_permute_load_chain): Likewise.
12715 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12716 (vect_transform_slp_perm_load): Likewise.
12717 (vect_schedule_slp_instance): Likewise.
12718 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12719 (vectorizable_mask_load_store): Likewise.
12720 (vectorizable_bswap): Likewise.
12721 (vectorizable_store): Likewise.
12722 (vectorizable_load): Likewise.
12723 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12724 vec_perm_indices instead of auto_vec_perm_indices. Use
12725 tree_to_vec_perm_builder to read the vector from a tree.
12726 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12727 vec_perm_builder instead of a vec_perm_indices.
12728 (have_whole_vector_shift): Use vec_perm_builder and
12729 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12730 truncation to calc_vec_perm_mask_for_shift.
12731 (vect_create_epilog_for_reduction): Likewise.
12732 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12733 from auto_vec_perm_indices to vec_perm_indices.
12734 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12735 instead of changing individual elements.
12736 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12737 the vector in d.perm.
12738 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12739 from auto_vec_perm_indices to vec_perm_indices.
12740 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12741 instead of changing individual elements.
12742 (arm_vectorize_vec_perm_const): Use new_vector to install
12743 the vector in d.perm.
12744 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12745 Update vec_perm_builder constructor.
12746 (rs6000_expand_interleave): Likewise.
12747 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12748 (rs6000_expand_interleave): Likewise.
12750 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12752 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12753 to qimode could truncate the indices.
12754 * optabs.c (expand_vec_perm_var): Likewise.
12756 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12758 * Makefile.in (OBJS): Add vec-perm-indices.o.
12759 * vec-perm-indices.h: New file.
12760 * vec-perm-indices.c: Likewise.
12761 * target.h (vec_perm_indices): Replace with a forward class
12763 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12764 * optabs.h: Include vec-perm-indices.h.
12765 (expand_vec_perm): Delete.
12766 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12767 (expand_vec_perm_const): Declare.
12768 * target.def (vec_perm_const_ok): Replace with...
12769 (vec_perm_const): ...this new hook.
12770 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12771 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12772 * doc/tm.texi: Regenerate.
12773 * optabs.def (vec_perm_const): Delete.
12774 * doc/md.texi (vec_perm_const): Likewise.
12775 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12776 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12777 expand_vec_perm for constant permutation vectors. Assert that
12778 the mode of variable permutation vectors is the integer equivalent
12779 of the mode that is being permuted.
12780 * optabs-query.h (selector_fits_mode_p): Declare.
12781 * optabs-query.c: Include vec-perm-indices.h.
12782 (selector_fits_mode_p): New function.
12783 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12784 is defined, instead of checking whether the vec_perm_const_optab
12785 exists. Use targetm.vectorize.vec_perm_const instead of
12786 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12787 fit in the vector mode before using a variable permute.
12788 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12789 vec_perm_indices instead of an rtx.
12790 (expand_vec_perm): Replace with...
12791 (expand_vec_perm_const): ...this new function. Take the selector
12792 as a vec_perm_indices rather than an rtx. Also take the mode of
12793 the selector. Update call to shift_amt_for_vec_perm_mask.
12794 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12795 Use vec_perm_indices::new_expanded_vector to expand the original
12796 selector into bytes. Check whether the indices fit in the vector
12797 mode before using a variable permute.
12798 (expand_vec_perm_var): Make global.
12799 (expand_mult_highpart): Use expand_vec_perm_const.
12800 * fold-const.c: Includes vec-perm-indices.h.
12801 * tree-ssa-forwprop.c: Likewise.
12802 * tree-vect-data-refs.c: Likewise.
12803 * tree-vect-generic.c: Likewise.
12804 * tree-vect-loop.c: Likewise.
12805 * tree-vect-slp.c: Likewise.
12806 * tree-vect-stmts.c: Likewise.
12807 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12809 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12810 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12811 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12812 (aarch64_vectorize_vec_perm_const): ...this new function.
12813 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12814 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12815 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12816 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12817 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12818 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12819 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12821 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12822 check for NEON modes.
12823 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12824 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12825 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12826 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12828 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12829 the old VEC_PERM_CONST conditions.
12830 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12831 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12832 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12833 (ia64_vectorize_vec_perm_const_ok): Merge into...
12834 (ia64_vectorize_vec_perm_const): ...this new function.
12835 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12836 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12837 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12838 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12839 * config/mips/mips.c (mips_expand_vec_perm_const)
12840 (mips_vectorize_vec_perm_const_ok): Merge into...
12841 (mips_vectorize_vec_perm_const): ...this new function.
12842 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12843 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12844 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12845 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12846 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12847 (rs6000_expand_vec_perm_const): Delete.
12848 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12850 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12851 (altivec_expand_vec_perm_const_le): Take each operand individually.
12852 Operate on constant selectors rather than rtxes.
12853 (altivec_expand_vec_perm_const): Likewise. Update call to
12854 altivec_expand_vec_perm_const_le.
12855 (rs6000_expand_vec_perm_const): Delete.
12856 (rs6000_vectorize_vec_perm_const_ok): Delete.
12857 (rs6000_vectorize_vec_perm_const): New function.
12858 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12859 an element count and rtx array.
12860 (rs6000_expand_extract_even): Update call accordingly.
12861 (rs6000_expand_interleave): Likewise.
12862 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12863 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12864 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12865 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12866 (rs6000_expand_vec_perm_const): Delete.
12867 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12868 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12869 (altivec_expand_vec_perm_const_le): Take each operand individually.
12870 Operate on constant selectors rather than rtxes.
12871 (altivec_expand_vec_perm_const): Likewise. Update call to
12872 altivec_expand_vec_perm_const_le.
12873 (rs6000_expand_vec_perm_const): Delete.
12874 (rs6000_vectorize_vec_perm_const_ok): Delete.
12875 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12876 reference to the SPE evmerge intructions.
12877 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12878 an element count and rtx array.
12879 (rs6000_expand_extract_even): Update call accordingly.
12880 (rs6000_expand_interleave): Likewise.
12881 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12882 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12884 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12886 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12888 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12889 vector mode and that that mode matches the mode of the data
12891 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12892 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12893 directly using expand_vec_perm_1 when forcing selectors into
12895 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12897 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12899 * optabs-query.h (can_vec_perm_p): Delete.
12900 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12901 * optabs-query.c (can_vec_perm_p): Split into...
12902 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12903 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12904 particular selector is valid.
12905 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12906 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12907 (vect_grouped_load_supported): Likewise.
12908 (vect_shift_permute_load_chain): Likewise.
12909 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12910 (vect_transform_slp_perm_load): Likewise.
12911 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12912 (vectorizable_bswap): Likewise.
12913 (vect_gen_perm_mask_checked): Likewise.
12914 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12915 implementations of variable permutation vectors into account
12916 when deciding which selector to use.
12917 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12918 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12919 with a false third argument.
12920 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12921 to test whether the constant selector is valid and can_vec_perm_var_p
12922 to test whether a variable selector is valid.
12924 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12926 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12927 * optabs-query.c (can_vec_perm_p): Likewise.
12928 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12929 instead of vec_perm_indices.
12930 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12931 (vect_gen_perm_mask_checked): Likewise,
12932 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12933 (vect_gen_perm_mask_checked): Likewise,
12935 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12937 * optabs-query.h (qimode_for_vec_perm): Declare.
12938 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12939 (qimode_for_vec_perm): ...this new function.
12940 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12942 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12944 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12945 does not have a conditional at the top.
12947 2018-01-02 Richard Biener <rguenther@suse.de>
12949 * ipa-inline.c (big_speedup_p): Fix expression.
12951 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12954 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12957 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12961 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12962 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12963 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12964 cond_taken_branch_cost 3->4.
12966 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12968 PR tree-optimization/83581
12969 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12970 TODO_cleanup_cfg if any changes have been made.
12972 PR middle-end/83608
12973 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12974 convert_modes if target mode has the right side, but different mode
12977 PR middle-end/83609
12978 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12979 last argument when extracting from CONCAT. If either from_real or
12980 from_imag is NULL, use expansion through memory. If result is not
12981 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12982 the parts directly to inner mode, if even that fails, use expansion
12985 PR middle-end/83623
12986 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12987 check for bswap in mode rather than HImode and use that in expand_unop
12990 Copyright (C) 2018 Free Software Foundation, Inc.
12992 Copying and distribution of this file, with or without modification,
12993 are permitted in any medium without royalty provided the copyright
12994 notice and this notice are preserved.